Skip to contentSkip to conversation

CXl 2.0 emulation Support

PATCH Superseded Jonathan Cameron 6 participants 71 messages Mar 18, 2022 – Apr 5, 2022 Arm/Amlogic Meson SoC support

v8 of 11 Superseded Latest is v11 →

lore
Share & tools
lore
#0 lore raw

v8 cover CXl 2.0 emulation Support

Changes since v7:
* Fixed and added comments for _OSC definition (Michael Tsirkin)
* Dropped the RFC patch that added memops to a hostmem backend and replaced with a per CXL type 3 device address space and use of address_space_read() / address_space_write(). (Mark Cave-Ayland)
* Added switch documentation as a new patch.

Note that I haven't added tracepoints as Mark suggested. That will probably be the subject of a future patchset building on this support.

Posting now so that people can review it when then find time, but clearly this will need to wait for the next cycle.

Series organized to allow it to be taken in stages if the maintainers prefer that approach. Most sets end with the addition of appropriate tests (TBD for switch support)

Patches 0-15 - CXL PXB
Patches 16-22 - Type 3 Device, Root Port
Patches 23-39 - ACPI, board elements and interleave decoding to enable x86 hosts Patches 40-41 - arm64 support on virt.
Patch 42 - Initial documentation
Patches 43-46 - Switch support.

I'm resisting the urge to have this series continue to grow with additional features on the basis it is already huge and what we have here is useful + functional.

Updated background info:

Looking in particular for:
* Review of the PCI interactions
* x86 and ARM machine interactions (particularly the memory maps)
* Review of the interleaving approach - is the basic idea acceptable? It has changed since v7!
* Review of the command line interface.
* CXL related review welcome but much of that got reviewed in earlier versions and hasn't changed substantially.

Big TODOs:

* Volatile memory devices (easy but it's more code so left for now).
* Hotplug? May not need much but it's not tested yet!
* More tests and tighter verification that values written to hardware are actually valid - stuff that real hardware would check.
* Testing, testing and more testing. I have been running a basic set of ARM and x86 tests on this, but there is always room for more tests and greater automation.
* CFMWS flags as requested by Ben.

Why do we want QEMU emulation of CXL?

As Ben stated in V3, QEMU support has been critical to getting OS software written given lack of availability of hardware supporting the latest CXL features (coupled with very high demand for support being ready in a timely fashion). What has become clear since Ben's v3 is that situation is a continuous one. Whilst we can't talk about them yet, CXL 3.0 features and OS support have been prototyped on top of this support and a lot of the ongoing kernel work is being tested against these patches. The kernel CXL mocking code allows some forms of testing, but QEMU provides a more versatile and exensible platform.

Other features on the qemu-list that build on these include PCI-DOE
/CDAT support from the Avery Design team further showing how this
code is useful. Whilst not directly related this is also the test
platform for work on PCI IDE/CMA + related DMTF SPDM as CXL both
utilizes and extends those technologies and is likely to be an early
adopter.
Refs:
CMA Kernel: https://lore.kernel.org/all/[email protected]/
CMA Qemu: https://lore.kernel.org/qemu-devel/[email protected]/
DOE Qemu: https://lore.kernel.org/qemu-devel/[email protected]/

As can be seen there is non trivial interaction with other areas of Qemu, particularly PCI and keeping this set up to date is proving a burden we'd rather do without :)

Ben mentioned a few other good reasons in v3:
https://lore.kernel.org/qemu-devel/[email protected]/

What we have here is about what you need for it to be useful for testing currently kernel code. Note the kernel code is moving fast so since v4, some features have been introduced we don't yet support in QEMU (e.g. use of the PCIe serial number extended capability).

All comments welcome.

Additional info that was here in v5 is now in the documentation patch.

Thanks,

Jonathan

Ben Widawsky (24):
hw/pci/cxl: Add a CXL component type (interface) hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5) hw/cxl/device: Introduce a CXL device (8.2.8) hw/cxl/device: Implement the CAP array (8.2.8.1-2) hw/cxl/device: Implement basic mailbox (8.2.8.4) hw/cxl/device: Add memory device utilities hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1) hw/cxl/device: Timestamp implementation (8.2.9.3) hw/cxl/device: Add log commands (8.2.9.4) + CEL hw/pxb: Use a type for realizing expanders hw/pci/cxl: Create a CXL bus type
hw/pxb: Allow creation of a CXL PXB (host bridge) hw/cxl/rp: Add a root port
hw/cxl/device: Add a memory device (8.2.8.5) hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12) hw/cxl/device: Add some trivial commands hw/cxl/device: Plumb real Label Storage Area (LSA) sizing hw/cxl/device: Implement get/set Label Storage Area (LSA) hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142) acpi/cxl: Add _OSC implementation (9.14.2)
acpi/cxl: Create the CEDT (9.14.1)
acpi/cxl: Introduce CFMWS structures in CEDT hw/cxl/component Add a dumb HDM decoder handler qtest/cxl: Add more complex test cases with CFMWs

Jonathan Cameron (22):
MAINTAINERS: Add entry for Compute Express Link Emulation
cxl: Machine level control on whether CXL support is enabled qtest/cxl: Introduce initial test for pxb-cxl only. qtests/cxl: Add initial root port and CXL type3 tests hw/cxl/component: Add utils for interleave parameter encoding/decoding hw/cxl/host: Add support for CXL Fixed Memory Windows. hw/pci-host/gpex-acpi: Add support for dsdt construction for pxb-cxl pci/pcie_port: Add pci_find_port_by_pn() CXL/cxl_component: Add cxl_get_hb_cstate()
mem/cxl_type3: Add read and write functions for associated hostmem. cxl/cxl-host: Add memops for CFMWS region.
i386/pc: Enable CXL fixed memory windows
tests/acpi: q35: Allow addition of a CXL test. qtests/bios-tables-test: Add a test for CXL emulation. tests/acpi: Add tables for CXL emulation.
hw/arm/virt: Basic CXL enablement on pci_expander_bridge instances
pxb-cxl
qtest/cxl: Add aarch64 virt test for CXL
docs/cxl: Add initial Compute eXpress Link (CXL) documentation. pci-bridge/cxl_upstream: Add a CXL switch upstream port pci-bridge/cxl_downstream: Add a CXL switch downstream port cxl/cxl-host: Support interleave decoding with one level of switches. docs/cxl: Add switch documentation

 MAINTAINERS                         |   7 +
 docs/system/device-emulation.rst    |   1 +
 docs/system/devices/cxl.rst         | 386 ++++++++++++++++++++++
 hw/Kconfig                          |   1 +
 hw/acpi/Kconfig                     |   5 +
 hw/acpi/cxl-stub.c                  |  12 +
 hw/acpi/cxl.c                       | 257 +++++++++++++++
 hw/acpi/meson.build                 |   4 +-
 hw/arm/Kconfig                      |   1 +
 hw/arm/virt-acpi-build.c            |  33 ++
 hw/arm/virt.c                       |  40 ++-
 hw/core/machine.c                   |  28 ++
 hw/cxl/Kconfig                      |   3 +
 hw/cxl/cxl-component-utils.c        | 284 ++++++++++++++++
 hw/cxl/cxl-device-utils.c           | 265 +++++++++++++++
 hw/cxl/cxl-host-stubs.c             |  16 +
 hw/cxl/cxl-host.c                   | 262 +++++++++++++++
 hw/cxl/cxl-mailbox-utils.c          | 485 ++++++++++++++++++++++++++++
 hw/cxl/meson.build                  |  12 +
 hw/i386/acpi-build.c                |  57 +++-
 hw/i386/pc.c                        |  57 +++-
 hw/mem/Kconfig                      |   5 +
 hw/mem/cxl_type3.c                  | 359 ++++++++++++++++++++
 hw/mem/meson.build                  |   1 +
 hw/meson.build                      |   1 +
 hw/pci-bridge/Kconfig               |   5 +
 hw/pci-bridge/cxl_downstream.c      | 229 +++++++++++++
 hw/pci-bridge/cxl_root_port.c       | 231 +++++++++++++
 hw/pci-bridge/cxl_upstream.c        | 205 ++++++++++++
 hw/pci-bridge/meson.build           |   1 +
 hw/pci-bridge/pci_expander_bridge.c | 172 +++++++++-
 hw/pci-bridge/pcie_root_port.c      |   6 +-
 hw/pci-host/gpex-acpi.c             |  20 +-
 hw/pci/pci.c                        |  21 +-
 hw/pci/pcie_port.c                  |  25 ++
 include/hw/acpi/cxl.h               |  28 ++
 include/hw/arm/virt.h               |   1 +
 include/hw/boards.h                 |   2 +
 include/hw/cxl/cxl.h                |  54 ++++
 include/hw/cxl/cxl_component.h      | 207 ++++++++++++
 include/hw/cxl/cxl_device.h         | 271 ++++++++++++++++
 include/hw/cxl/cxl_pci.h            | 156 +++++++++
 include/hw/pci/pci.h                |  14 +
 include/hw/pci/pci_bridge.h         |  20 ++
 include/hw/pci/pci_bus.h            |   7 +
 include/hw/pci/pci_ids.h            |   1 +
 include/hw/pci/pcie_port.h          |   2 +
 qapi/machine.json                   |  18 ++
 qemu-options.hx                     |  38 +++
 scripts/device-crash-test           |   1 +
 softmmu/vl.c                        |  44 +++
 tests/data/acpi/q35/CEDT.cxl        | Bin 0 -> 184 bytes
 tests/data/acpi/q35/DSDT.cxl        | Bin 0 -> 9615 bytes
 tests/qtest/bios-tables-test.c      |  44 +++
 tests/qtest/cxl-test.c              | 181 +++++++++++
 tests/qtest/meson.build             |   5 +
 56 files changed, 4566 insertions(+), 25 deletions(-)
 create mode 100644 docs/system/devices/cxl.rst
 create mode 100644 hw/acpi/cxl-stub.c
 create mode 100644 hw/acpi/cxl.c
 create mode 100644 hw/cxl/Kconfig
 create mode 100644 hw/cxl/cxl-component-utils.c
 create mode 100644 hw/cxl/cxl-device-utils.c
 create mode 100644 hw/cxl/cxl-host-stubs.c
 create mode 100644 hw/cxl/cxl-host.c
 create mode 100644 hw/cxl/cxl-mailbox-utils.c
 create mode 100644 hw/cxl/meson.build
 create mode 100644 hw/mem/cxl_type3.c
 create mode 100644 hw/pci-bridge/cxl_downstream.c
 create mode 100644 hw/pci-bridge/cxl_root_port.c
 create mode 100644 hw/pci-bridge/cxl_upstream.c
 create mode 100644 include/hw/acpi/cxl.h
 create mode 100644 include/hw/cxl/cxl.h
 create mode 100644 include/hw/cxl/cxl_component.h
 create mode 100644 include/hw/cxl/cxl_device.h
 create mode 100644 include/hw/cxl/cxl_pci.h
 create mode 100644 tests/data/acpi/q35/CEDT.cxl
 create mode 100644 tests/data/acpi/q35/DSDT.cxl
 create mode 100644 tests/qtest/cxl-test.c
Signature · 1 line
2.32.0
#1 lore raw

PATCH v8 1/46 hw/pci/cxl: Add a CXL component type (interface)

From: Ben Widawsky <[email protected]>

A CXL component is a hardware entity that implements CXL component registers from the CXL 2.0 spec (8.2.3). Currently these represent 3 general types.
1. Host Bridge
2. Ports (root, upstream, downstream)
3. Devices (memory, other)

A CXL component can be conceptually thought of as a PCIe device with extra functionality when enumerated and enabled. For this reason, CXL does here, and will continue to add on to existing PCI code paths.

Host bridges will typically need to be handled specially and so they can implement this newly introduced interface or not. All other components should implement this interface. Implementing this interface allows the core PCI code to treat these devices as special where appropriate.

---
 hw/pci/pci.c         | 10 ++++++++++
 include/hw/pci/pci.h |  8 ++++++++
 2 files changed, 18 insertions(+)
hw/pci/pci.c +10 -0
index 5cb1232e27..7883778a99 100644
@@ -201,6 +201,11 @@ static const TypeInfo pci_bus_info = {
     .class_init = pci_bus_class_init,
 };
 
+static const TypeInfo cxl_interface_info = {
+    .name          = INTERFACE_CXL_DEVICE,
+    .parent        = TYPE_INTERFACE,
+};
+
 static const TypeInfo pcie_interface_info = {
     .name          = INTERFACE_PCIE_DEVICE,
     .parent        = TYPE_INTERFACE,
@@ -2182,6 +2187,10 @@ static void pci_qdev_realize(DeviceState *qdev, Error **errp)
         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
     }
 
+    if (object_class_dynamic_cast(klass, INTERFACE_CXL_DEVICE)) {
+        pci_dev->cap_present |= QEMU_PCIE_CAP_CXL;
+    }
+
     pci_dev = do_pci_register_device(pci_dev,
                                      object_get_typename(OBJECT(qdev)),
                                      pci_dev->devfn, errp);
@@ -2938,6 +2947,7 @@ static void pci_register_types(void)
     type_register_static(&pci_bus_info);
     type_register_static(&pcie_bus_info);
     type_register_static(&conventional_pci_interface_info);
+    type_register_static(&cxl_interface_info);
     type_register_static(&pcie_interface_info);
     type_register_static(&pci_device_type_info);
 }
include/hw/pci/pci.h +8 -0
index 3a32b8dd40..98f0d1b844 100644
@@ -194,6 +194,8 @@ enum {
     QEMU_PCIE_LNKSTA_DLLLA = (1 << QEMU_PCIE_LNKSTA_DLLLA_BITNR),
 #define QEMU_PCIE_EXTCAP_INIT_BITNR 9
     QEMU_PCIE_EXTCAP_INIT = (1 << QEMU_PCIE_EXTCAP_INIT_BITNR),
+#define QEMU_PCIE_CXL_BITNR 10
+    QEMU_PCIE_CAP_CXL = (1 << QEMU_PCIE_CXL_BITNR),
 };
 
 #define TYPE_PCI_DEVICE "pci-device"
@@ -201,6 +203,12 @@ typedef struct PCIDeviceClass PCIDeviceClass;
 DECLARE_OBJ_CHECKERS(PCIDevice, PCIDeviceClass,
                      PCI_DEVICE, TYPE_PCI_DEVICE)
 
+/*
+ * Implemented by devices that can be plugged on CXL buses. In the spec, this is
+ * actually a "CXL Component, but we name it device to match the PCI naming.
+ */
+#define INTERFACE_CXL_DEVICE "cxl-device"
+
 /* Implemented by devices that can be plugged on PCI Express buses */
 #define INTERFACE_PCIE_DEVICE "pci-express-device"
 
Signature · 1 line
2.32.0
#2 lore raw

Re: [PATCH v8 01/46] hw/pci/cxl: Add a CXL component type (interface)

Quoted 91 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL component is a hardware entity that implements CXL component registers from the CXL 2.0 spec (8.2.3). Currently these represent 3 general types.
1. Host Bridge
2. Ports (root, upstream, downstream)
3. Devices (memory, other)

A CXL component can be conceptually thought of as a PCIe device with extra functionality when enumerated and enabled. For this reason, CXL does here, and will continue to add on to existing PCI code paths.

Host bridges will typically need to be handled specially and so they can implement this newly introduced interface or not. All other components should implement this interface. Implementing this interface allows the core PCI code to treat these devices as special where appropriate.

---
 hw/pci/pci.c         | 10 ++++++++++
 include/hw/pci/pci.h |  8 ++++++++
 2 files changed, 18 insertions(+)
hw/pci/pci.c +10 -0
index 5cb1232e27..7883778a99 100644
@@ -201,6 +201,11 @@ static const TypeInfo pci_bus_info = {
     .class_init = pci_bus_class_init,
 };
 
+static const TypeInfo cxl_interface_info = {
+    .name          = INTERFACE_CXL_DEVICE,
+    .parent        = TYPE_INTERFACE,
+};
+
 static const TypeInfo pcie_interface_info = {
     .name          = INTERFACE_PCIE_DEVICE,
     .parent        = TYPE_INTERFACE,
@@ -2182,6 +2187,10 @@ static void pci_qdev_realize(DeviceState *qdev, Error **errp)
         pci_dev->cap_present |= QEMU_PCI_CAP_EXPRESS;
     }
 
+    if (object_class_dynamic_cast(klass, INTERFACE_CXL_DEVICE)) {
+        pci_dev->cap_present |= QEMU_PCIE_CAP_CXL;
+    }
+
     pci_dev = do_pci_register_device(pci_dev,
                                      object_get_typename(OBJECT(qdev)),
                                      pci_dev->devfn, errp);
@@ -2938,6 +2947,7 @@ static void pci_register_types(void)
     type_register_static(&pci_bus_info);
     type_register_static(&pcie_bus_info);
     type_register_static(&conventional_pci_interface_info);
+    type_register_static(&cxl_interface_info);
     type_register_static(&pcie_interface_info);
     type_register_static(&pci_device_type_info);
 }
include/hw/pci/pci.h +8 -0
index 3a32b8dd40..98f0d1b844 100644
@@ -194,6 +194,8 @@ enum {
     QEMU_PCIE_LNKSTA_DLLLA = (1 << QEMU_PCIE_LNKSTA_DLLLA_BITNR),
 #define QEMU_PCIE_EXTCAP_INIT_BITNR 9
     QEMU_PCIE_EXTCAP_INIT = (1 << QEMU_PCIE_EXTCAP_INIT_BITNR),
+#define QEMU_PCIE_CXL_BITNR 10
+    QEMU_PCIE_CAP_CXL = (1 << QEMU_PCIE_CXL_BITNR),
 };
 
 #define TYPE_PCI_DEVICE "pci-device"
@@ -201,6 +203,12 @@ typedef struct PCIDeviceClass PCIDeviceClass;
 DECLARE_OBJ_CHECKERS(PCIDevice, PCIDeviceClass,
                      PCI_DEVICE, TYPE_PCI_DEVICE)
 
+/*
+ * Implemented by devices that can be plugged on CXL buses. In the spec, this is
+ * actually a "CXL Component, but we name it device to match the PCI naming.
+ */
+#define INTERFACE_CXL_DEVICE "cxl-device"
+
 /* Implemented by devices that can be plugged on PCI Express buses */
 #define INTERFACE_PCIE_DEVICE "pci-express-device"
 
Signature · 1 line
2.32.0

Looks good.

Reviewed by: Adam Manzanares <[email protected]>

#3 lore raw

PATCH v8 2/46 hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL 2.0 component is any entity in the CXL topology. All components have a analogous function in PCIe. Except for the CXL host bridge, all have a PCIe config space that is accessible via the common PCIe mechanisms. CXL components are enumerated via DVSEC fields in the extended PCIe header space. CXL components will minimally implement some subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL 2.0 specification. Two headers and a utility library are introduced to support the minimum functionality needed to enumerate components.

The cxl_pci header manages bits associated with PCI, specifically the DVSEC and related fields. The cxl_component.h variant has data structures and APIs that are useful for drivers implementing any of the CXL 2.0 components. The library takes care of making use of the DVSEC bits and the CXL.[mem|cache] registers. Per spec, the registers are little endian.

None of the mechanisms required to enumerate a CXL capable hostbridge are introduced at this point.

Note that the CXL.mem and CXL.cache registers used are always 4B wide. It's possible in the future that this constraint will not hold.

---
 hw/Kconfig                     |   1 +
 hw/cxl/Kconfig                 |   3 +
 hw/cxl/cxl-component-utils.c   | 219 +++++++++++++++++++++++++++++++++
 hw/cxl/meson.build             |   4 +
 hw/meson.build                 |   1 +
 include/hw/cxl/cxl.h           |  16 +++
 include/hw/cxl/cxl_component.h | 197 +++++++++++++++++++++++++++++
 include/hw/cxl/cxl_pci.h       | 135 ++++++++++++++++++++
 8 files changed, 576 insertions(+)
hw/Kconfig +1 -0
index ad20cce0a9..50e0952889 100644
@@ -6,6 +6,7 @@ source audio/Kconfig
 source block/Kconfig
 source char/Kconfig
 source core/Kconfig
+source cxl/Kconfig
 source display/Kconfig
 source dma/Kconfig
 source gpio/Kconfig
hw/cxl/Kconfig new file +3 -0
new file mode 100644
index 0000000000..8e67519b16
@@ -0,0 +1,3 @@
+config CXL
+    bool
+    default y if PCI_EXPRESS
hw/cxl/cxl-component-utils.c new file +219 -0

219 lines not on this page · raw message

hw/cxl/meson.build new file +4 -0
new file mode 100644
index 0000000000..3231b5de1e
@@ -0,0 +1,4 @@
+softmmu_ss.add(when: 'CONFIG_CXL',
+               if_true: files(
+                   'cxl-component-utils.c',
+               ))
hw/meson.build +1 -0
index b3366c888e..9992c5101e 100644
@@ -6,6 +6,7 @@ subdir('block')
 subdir('char')
 subdir('core')
 subdir('cpu')
+subdir('cxl')
 subdir('display')
 subdir('dma')
 subdir('gpio')
include/hw/cxl/cxl.h new file +16 -0
new file mode 100644
index 0000000000..8c738c7a2b
@@ -0,0 +1,16 @@
+/*
+ * QEMU CXL Support
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_H
+#define CXL_H
+
+#include "cxl_pci.h"
+#include "cxl_component.h"
+
+#endif
include/hw/cxl/cxl_component.h new file +197 -0

197 lines not on this page · raw message

include/hw/cxl/cxl_pci.h new file +135 -0

135 lines not on this page · raw message

Signature · 1 line
2.32.0
#4 lore raw

Re: [PATCH v8 02/46] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)

Quoted 21 lines · On Fri, 18 Mar 2022 15:05:51 +0000 Jonathan Cameron…

From: Ben Widawsky <[email protected]>

A CXL 2.0 component is any entity in the CXL topology. All components have a analogous function in PCIe. Except for the CXL host bridge, all have a PCIe config space that is accessible via the common PCIe mechanisms. CXL components are enumerated via DVSEC fields in the extended PCIe header space. CXL components will minimally implement some subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL 2.0 specification. Two headers and a utility library are introduced to support the minimum functionality needed to enumerate components.

The cxl_pci header manages bits associated with PCI, specifically the DVSEC and related fields. The cxl_component.h variant has data structures and APIs that are useful for drivers implementing any of the CXL 2.0 components. The library takes care of making use of the DVSEC bits and the CXL.[mem|cache] registers. Per spec, the registers are little endian.

None of the mechanisms required to enumerate a CXL capable hostbridge are introduced at this point.

Quoted 6 lines

Note that the CXL.mem and CXL.cache registers used are always 4B wide. It's possible in the future that this constraint will not hold.

As a heads up for anyone follow current state of this series...

Whilst wiring up more of the PCI config space elements off these devices (for a future series, we don't need that to work yet) for kernel testing it's become clear we are getting lots of
offsetof(struct cxl_dvsec_port_extensions, whatever)

so I'm going to switch to typdefs and CamelCase naming though it gets a bit ugly for some of these as they are made up of a string of acronyms and we get things like

CXLDVSECDeviceGPF

However I think that's still fairly readable.

Thanks,

Jonathan

Quoted 645 lines
---
 hw/Kconfig                     |   1 +
 hw/cxl/Kconfig                 |   3 +
 hw/cxl/cxl-component-utils.c   | 219 +++++++++++++++++++++++++++++++++
 hw/cxl/meson.build             |   4 +
 hw/meson.build                 |   1 +
 include/hw/cxl/cxl.h           |  16 +++
 include/hw/cxl/cxl_component.h | 197 +++++++++++++++++++++++++++++
 include/hw/cxl/cxl_pci.h       | 135 ++++++++++++++++++++
 8 files changed, 576 insertions(+)
hw/Kconfig +1 -0
index ad20cce0a9..50e0952889 100644
@@ -6,6 +6,7 @@ source audio/Kconfig
 source block/Kconfig
 source char/Kconfig
 source core/Kconfig
+source cxl/Kconfig
 source display/Kconfig
 source dma/Kconfig
 source gpio/Kconfig
hw/cxl/Kconfig new file +3 -0
new file mode 100644
index 0000000000..8e67519b16
@@ -0,0 +1,3 @@
+config CXL
+    bool
+    default y if PCI_EXPRESS
hw/cxl/cxl-component-utils.c new file +219 -0
new file mode 100644
index 0000000000..410f8ef328
@@ -0,0 +1,219 @@
+/*
+ * CXL Utility library for components
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
+
+static uint64_t cxl_cache_mem_read_reg(void *opaque, hwaddr offset,
+                                       unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size == 8) {
+        qemu_log_mask(LOG_UNIMP,
+                      "CXL 8 byte cache mem registers not implemented\n");
+        return 0;
+    }
+
+    if (cregs->special_ops && cregs->special_ops->read) {
+        return cregs->special_ops->read(cxl_cstate, offset, size);
+    } else {
+        return cregs->cache_mem_registers[offset / 4];
+    }
+}
+
+static void cxl_cache_mem_write_reg(void *opaque, hwaddr offset, uint64_t value,
+                                    unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size == 8) {
+        qemu_log_mask(LOG_UNIMP,
+                      "CXL 8 byte cache mem registers not implemented\n");
+        return;
+    }
+    if (cregs->special_ops && cregs->special_ops->write) {
+        cregs->special_ops->write(cxl_cstate, offset, value, size);
+    } else {
+        cregs->cache_mem_registers[offset / 4] = value;
+    }
+}
+
+/*
+ * 8.2.3
+ *   The access restrictions specified in Section 8.2.2 also apply to CXL 2.0
+ *   Component Registers.
+ *
+ * 8.2.2
+ *   • A 32 bit register shall be accessed as a 4 Bytes quantity. Partial
+ *   reads are not permitted.
+ *   • A 64 bit register shall be accessed as a 8 Bytes quantity. Partial
+ *   reads are not permitted.
+ *
+ * As of the spec defined today, only 4 byte registers exist.
+ */
+static const MemoryRegionOps cache_mem_ops = {
+    .read = cxl_cache_mem_read_reg,
+    .write = cxl_cache_mem_write_reg,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+        .unaligned = false,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type)
+{
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    memory_region_init(&cregs->component_registers, obj, type,
+                       CXL2_COMPONENT_BLOCK_SIZE);
+
+    /* io registers controls link which we don't care about in QEMU */
+    memory_region_init_io(&cregs->io, obj, NULL, cregs, ".io",
+                          CXL2_COMPONENT_IO_REGION_SIZE);
+    memory_region_init_io(&cregs->cache_mem, obj, &cache_mem_ops, cregs,
+                          ".cache_mem", CXL2_COMPONENT_CM_REGION_SIZE);
+
+    memory_region_add_subregion(&cregs->component_registers, 0, &cregs->io);
+    memory_region_add_subregion(&cregs->component_registers,
+                                CXL2_COMPONENT_IO_REGION_SIZE,
+                                &cregs->cache_mem);
+}
+
+static void ras_init_common(uint32_t *reg_state)
+{
+    reg_state[R_CXL_RAS_UNC_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_UNC_ERR_MASK] = 0x1cfff;
+    reg_state[R_CXL_RAS_UNC_ERR_SEVERITY] = 0x1cfff;
+    reg_state[R_CXL_RAS_COR_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_COR_ERR_MASK] = 0x3f;
+
+    /* CXL switches and devices must set */
+    reg_state[R_CXL_RAS_ERR_CAP_CTRL] = 0;
+}
+
+static void hdm_init_common(uint32_t *reg_state)
+{
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_GLOBAL_CONTROL,
+                     HDM_DECODER_ENABLE, 0);
+}
+
+void cxl_component_register_init_common(uint32_t *reg_state, enum reg_type type)
+{
+    int caps = 0;
+    switch (type) {
+    case CXL2_DOWNSTREAM_PORT:
+    case CXL2_DEVICE:
+        /* CAP, RAS, Link */
+        caps = 2;
+        break;
+    case CXL2_UPSTREAM_PORT:
+    case CXL2_TYPE3_DEVICE:
+    case CXL2_LOGICAL_DEVICE:
+        /* + HDM */
+        caps = 3;
+        break;
+    case CXL2_ROOT_PORT:
+        /* + Extended Security, + Snoop */
+        caps = 5;
+        break;
+    default:
+        abort();
+    }
+
+    memset(reg_state, 0, CXL2_COMPONENT_CM_REGION_SIZE);
+
+    /* CXL Capability Header Register */
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ID, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ARRAY_SIZE, caps);
+
+
+#define init_cap_reg(reg, id, version)                                        \
+    QEMU_BUILD_BUG_ON(CXL_##reg##_REGISTERS_OFFSET == 0);                     \
+    do {                                                                      \
+        int which = R_CXL_##reg##_CAPABILITY_HEADER;                          \
+        reg_state[which] = FIELD_DP32(reg_state[which],                       \
+                                      CXL_##reg##_CAPABILITY_HEADER, ID, id); \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER,       \
+                       VERSION, version);                                     \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER, PTR,  \
+                       CXL_##reg##_REGISTERS_OFFSET);                         \
+    } while (0)
+
+    init_cap_reg(RAS, 2, 1);
+    ras_init_common(reg_state);
+
+    init_cap_reg(LINK, 4, 2);
+
+    if (caps < 3) {
+        return;
+    }
+
+    init_cap_reg(HDM, 5, 1);
+    hdm_init_common(reg_state);
+
+    if (caps < 5) {
+        return;
+    }
+
+    init_cap_reg(EXTSEC, 6, 1);
+    init_cap_reg(SNOOP, 8, 1);
+
+#undef init_cap_reg
+}
+
+/*
+ * Helper to creates a DVSEC header for a CXL entity. The caller is responsible
+ * for tracking the valid offset.
+ *
+ * This function will build the DVSEC header on behalf of the caller and then
+ * copy in the remaining data for the vendor specific bits.
+ */
+void cxl_component_create_dvsec(CXLComponentState *cxl, uint16_t length,
+                                uint16_t type, uint8_t rev, uint8_t *body)
+{
+    PCIDevice *pdev = cxl->pdev;
+    uint16_t offset = cxl->dvsec_offset;
+
+    assert(offset >= PCI_CFG_SPACE_SIZE &&
+           ((offset + length) < PCI_CFG_SPACE_EXP_SIZE));
+    assert((length & 0xf000) == 0);
+    assert((rev & ~0xf) == 0);
+
+    /* Create the DVSEC in the MCFG space */
+    pcie_add_capability(pdev, PCI_EXT_CAP_ID_DVSEC, 1, offset, length);
+    pci_set_long(pdev->config + offset + PCIE_DVSEC_HEADER1_OFFSET,
+                 (length << 20) | (rev << 16) | CXL_VENDOR_ID);
+    pci_set_word(pdev->config + offset + PCIE_DVSEC_ID_OFFSET, type);
+    memcpy(pdev->config + offset + sizeof(struct dvsec_header),
+           body + sizeof(struct dvsec_header),
+           length - sizeof(struct dvsec_header));
+
+    /* Update state for future DVSEC additions */
+    range_init_nofail(&cxl->dvsecs[type], cxl->dvsec_offset, length);
+    cxl->dvsec_offset += length;
+}
hw/cxl/meson.build new file +4 -0
new file mode 100644
index 0000000000..3231b5de1e
@@ -0,0 +1,4 @@
+softmmu_ss.add(when: 'CONFIG_CXL',
+               if_true: files(
+                   'cxl-component-utils.c',
+               ))
hw/meson.build +1 -0
index b3366c888e..9992c5101e 100644
@@ -6,6 +6,7 @@ subdir('block')
 subdir('char')
 subdir('core')
 subdir('cpu')
+subdir('cxl')
 subdir('display')
 subdir('dma')
 subdir('gpio')
include/hw/cxl/cxl.h new file +16 -0
new file mode 100644
index 0000000000..8c738c7a2b
@@ -0,0 +1,16 @@
+/*
+ * QEMU CXL Support
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_H
+#define CXL_H
+
+#include "cxl_pci.h"
+#include "cxl_component.h"
+
+#endif
include/hw/cxl/cxl_component.h new file +197 -0
new file mode 100644
index 0000000000..74e9bfe1ff
@@ -0,0 +1,197 @@
+/*
+ * QEMU CXL Component
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_COMPONENT_H
+#define CXL_COMPONENT_H
+
+/* CXL 2.0 - 8.2.4 */
+#define CXL2_COMPONENT_IO_REGION_SIZE 0x1000
+#define CXL2_COMPONENT_CM_REGION_SIZE 0x1000
+#define CXL2_COMPONENT_BLOCK_SIZE 0x10000
+
+#include "qemu/compiler.h"
+#include "qemu/range.h"
+#include "qemu/typedefs.h"
+#include "hw/register.h"
+
+enum reg_type {
+    CXL2_DEVICE,
+    CXL2_TYPE3_DEVICE,
+    CXL2_LOGICAL_DEVICE,
+    CXL2_ROOT_PORT,
+    CXL2_UPSTREAM_PORT,
+    CXL2_DOWNSTREAM_PORT
+};
+
+/*
+ * Capability registers are defined at the top of the CXL.cache/mem region and
+ * are packed. For our purposes we will always define the caps in the same
+ * order.
+ * CXL 2.0 - 8.2.5 Table 142 for details.
+ */
+
+/* CXL 2.0 - 8.2.5.1 */
+REG32(CXL_CAPABILITY_HEADER, 0)
+    FIELD(CXL_CAPABILITY_HEADER, ID, 0, 16)
+    FIELD(CXL_CAPABILITY_HEADER, VERSION, 16, 4)
+    FIELD(CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 20, 4)
+    FIELD(CXL_CAPABILITY_HEADER, ARRAY_SIZE, 24, 8)
+
+#define CXLx_CAPABILITY_HEADER(type, offset)                  \
+    REG32(CXL_##type##_CAPABILITY_HEADER, offset)             \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, ID, 0, 16)      \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, VERSION, 16, 4) \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, PTR, 20, 12)
+CXLx_CAPABILITY_HEADER(RAS, 0x4)
+CXLx_CAPABILITY_HEADER(LINK, 0x8)
+CXLx_CAPABILITY_HEADER(HDM, 0xc)
+CXLx_CAPABILITY_HEADER(EXTSEC, 0x10)
+CXLx_CAPABILITY_HEADER(SNOOP, 0x14)
+
+/*
+ * Capability structures contain the actual registers that the CXL component
+ * implements. Some of these are specific to certain types of components, but
+ * this implementation leaves enough space regardless.
+ */
+/* 8.2.5.9 - CXL RAS Capability Structure */
+
+/* Give ample space for caps before this */
+#define CXL_RAS_REGISTERS_OFFSET 0x80
+#define CXL_RAS_REGISTERS_SIZE   0x58
+REG32(CXL_RAS_UNC_ERR_STATUS, CXL_RAS_REGISTERS_OFFSET)
+REG32(CXL_RAS_UNC_ERR_MASK, CXL_RAS_REGISTERS_OFFSET + 0x4)
+REG32(CXL_RAS_UNC_ERR_SEVERITY, CXL_RAS_REGISTERS_OFFSET + 0x8)
+REG32(CXL_RAS_COR_ERR_STATUS, CXL_RAS_REGISTERS_OFFSET + 0xc)
+REG32(CXL_RAS_COR_ERR_MASK, CXL_RAS_REGISTERS_OFFSET + 0x10)
+REG32(CXL_RAS_ERR_CAP_CTRL, CXL_RAS_REGISTERS_OFFSET + 0x14)
+/* Offset 0x18 - 0x58 reserved for RAS logs */
+
+/* 8.2.5.10 - CXL Security Capability Structure */
+#define CXL_SEC_REGISTERS_OFFSET \
+    (CXL_RAS_REGISTERS_OFFSET + CXL_RAS_REGISTERS_SIZE)
+#define CXL_SEC_REGISTERS_SIZE   0 /* We don't implement 1.1 downstream ports */
+
+/* 8.2.5.11 - CXL Link Capability Structure */
+#define CXL_LINK_REGISTERS_OFFSET \
+    (CXL_SEC_REGISTERS_OFFSET + CXL_SEC_REGISTERS_SIZE)
+#define CXL_LINK_REGISTERS_SIZE   0x38
+
+/* 8.2.5.12 - CXL HDM Decoder Capability Structure */
+#define HDM_DECODE_MAX 10 /* 8.2.5.12.1 */
+#define CXL_HDM_REGISTERS_OFFSET \
+    (CXL_LINK_REGISTERS_OFFSET + CXL_LINK_REGISTERS_SIZE)
+#define CXL_HDM_REGISTERS_SIZE (0x20 + HDM_DECODE_MAX + 10)
+#define HDM_DECODER_INIT(n)                                                    \
+  REG32(CXL_HDM_DECODER##n##_BASE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x10)                          \
+            FIELD(CXL_HDM_DECODER##n##_BASE_LO, L, 28, 4)                      \
+  REG32(CXL_HDM_DECODER##n##_BASE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x14)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x18)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x1C)                          \
+  REG32(CXL_HDM_DECODER##n##_CTRL,                                             \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x20)                          \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IG, 0, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IW, 4, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, LOCK_ON_COMMIT, 8, 1)             \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMIT, 9, 1)                     \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMITTED, 10, 1)                 \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, ERR, 11, 1)                       \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, TYPE, 12, 1)                      \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_LO,                                   \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x24)                          \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_HI,                                   \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x28)
+
+REG32(CXL_HDM_DECODER_CAPABILITY, CXL_HDM_REGISTERS_OFFSET)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 4, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTERLEAVE_256B, 8, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTELEAVE_4K, 9, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, POISON_ON_ERR_CAP, 10, 1)
+REG32(CXL_HDM_DECODER_GLOBAL_CONTROL, CXL_HDM_REGISTERS_OFFSET + 4)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, POISON_ON_ERR_EN, 0, 1)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, HDM_DECODER_ENABLE, 1, 1)
+
+HDM_DECODER_INIT(0);
+
+/* 8.2.5.13 - CXL Extended Security Capability Structure (Root complex only) */
+#define EXTSEC_ENTRY_MAX        256
+#define CXL_EXTSEC_REGISTERS_OFFSET \
+    (CXL_HDM_REGISTERS_OFFSET + CXL_HDM_REGISTERS_SIZE)
+#define CXL_EXTSEC_REGISTERS_SIZE   (8 * EXTSEC_ENTRY_MAX + 4)
+
+/* 8.2.5.14 - CXL IDE Capability Structure */
+#define CXL_IDE_REGISTERS_OFFSET \
+    (CXL_EXTSEC_REGISTERS_OFFSET + CXL_EXTSEC_REGISTERS_SIZE)
+#define CXL_IDE_REGISTERS_SIZE   0x20
+
+/* 8.2.5.15 - CXL Snoop Filter Capability Structure */
+#define CXL_SNOOP_REGISTERS_OFFSET \
+    (CXL_IDE_REGISTERS_OFFSET + CXL_IDE_REGISTERS_SIZE)
+#define CXL_SNOOP_REGISTERS_SIZE   0x8
+
+QEMU_BUILD_BUG_MSG((CXL_SNOOP_REGISTERS_OFFSET + CXL_SNOOP_REGISTERS_SIZE) >= 0x1000,
+                   "No space for registers");
+
+typedef struct component_registers {
+    /*
+     * Main memory region to be registered with QEMU core.
+     */
+    MemoryRegion component_registers;
+
+    /*
+     * 8.2.4 Table 141:
+     *   0x0000 - 0x0fff CXL.io registers
+     *   0x1000 - 0x1fff CXL.cache and CXL.mem
+     *   0x2000 - 0xdfff Implementation specific
+     *   0xe000 - 0xe3ff CXL ARB/MUX registers
+     *   0xe400 - 0xffff RSVD
+     */
+    uint32_t io_registers[CXL2_COMPONENT_IO_REGION_SIZE >> 2];
+    MemoryRegion io;
+
+    uint32_t cache_mem_registers[CXL2_COMPONENT_CM_REGION_SIZE >> 2];
+    MemoryRegion cache_mem;
+
+    MemoryRegion impl_specific;
+    MemoryRegion arb_mux;
+    MemoryRegion rsvd;
+
+    /* special_ops is used for any component that needs any specific handling */
+    MemoryRegionOps *special_ops;
+} ComponentRegisters;
+
+/*
+ * A CXL component represents all entities in a CXL hierarchy. This includes,
+ * host bridges, root ports, upstream/downstream switch ports, and devices
+ */
+typedef struct cxl_component {
+    ComponentRegisters crb;
+    union {
+        struct {
+            Range dvsecs[CXL20_MAX_DVSEC];
+            uint16_t dvsec_offset;
+            struct PCIDevice *pdev;
+        };
+    };
+} CXLComponentState;
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type);
+void cxl_component_register_init_common(uint32_t *reg_state,
+                                        enum reg_type type);
+
+void cxl_component_create_dvsec(CXLComponentState *cxl_cstate, uint16_t length,
+                                uint16_t type, uint8_t rev, uint8_t *body);
+
+#endif
include/hw/cxl/cxl_pci.h new file +135 -0
new file mode 100644
index 0000000000..810a244fab
@@ -0,0 +1,135 @@
+/*
+ * QEMU CXL PCI interfaces
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_PCI_H
+#define CXL_PCI_H
+
+#include "qemu/compiler.h"
+#include "hw/pci/pci.h"
+#include "hw/pci/pcie.h"
+
+#define CXL_VENDOR_ID 0x1e98
+
+#define PCIE_DVSEC_HEADER1_OFFSET 0x4 /* Offset from start of extend cap */
+#define PCIE_DVSEC_ID_OFFSET 0x8
+
+#define PCIE_CXL_DEVICE_DVSEC_LENGTH 0x38
+#define PCIE_CXL1_DEVICE_DVSEC_REVID 0
+#define PCIE_CXL2_DEVICE_DVSEC_REVID 1
+
+#define EXTENSIONS_PORT_DVSEC_LENGTH 0x28
+#define EXTENSIONS_PORT_DVSEC_REVID 0
+
+#define GPF_PORT_DVSEC_LENGTH 0x10
+#define GPF_PORT_DVSEC_REVID  0
+
+#define PCIE_FLEXBUS_PORT_DVSEC_LENGTH_2_0 0x14
+#define PCIE_FLEXBUS_PORT_DVSEC_REVID_2_0  1
+
+#define REG_LOC_DVSEC_LENGTH 0x24
+#define REG_LOC_DVSEC_REVID  0
+
+enum {
+    PCIE_CXL_DEVICE_DVSEC      = 0,
+    NON_CXL_FUNCTION_MAP_DVSEC = 2,
+    EXTENSIONS_PORT_DVSEC      = 3,
+    GPF_PORT_DVSEC             = 4,
+    GPF_DEVICE_DVSEC           = 5,
+    PCIE_FLEXBUS_PORT_DVSEC    = 7,
+    REG_LOC_DVSEC              = 8,
+    MLD_DVSEC                  = 9,
+    CXL20_MAX_DVSEC
+};
+
+struct dvsec_header {
+    uint32_t cap_hdr;
+    uint32_t dv_hdr1;
+    uint16_t dv_hdr2;
+} QEMU_PACKED;
+QEMU_BUILD_BUG_ON(sizeof(struct dvsec_header) != 10);
+
+/*
+ * CXL 2.0 devices must implement certain DVSEC IDs, and can [optionally]
+ * implement others.
+ *
+ * CXL 2.0 Device: 0, [2], 5, 8
+ * CXL 2.0 RP: 3, 4, 7, 8
+ * CXL 2.0 Upstream Port: [2], 7, 8
+ * CXL 2.0 Downstream Port: 3, 4, 7, 8
+ */
+
+/* CXL 2.0 - 8.1.5 (ID 0003) */
+struct cxl_dvsec_port_extensions {
+    struct dvsec_header hdr;
+    uint16_t status;
+    uint16_t control;
+    uint8_t alt_bus_base;
+    uint8_t alt_bus_limit;
+    uint16_t alt_memory_base;
+    uint16_t alt_memory_limit;
+    uint16_t alt_prefetch_base;
+    uint16_t alt_prefetch_limit;
+    uint32_t alt_prefetch_base_high;
+    uint32_t alt_prefetch_base_low;
+    uint32_t rcrb_base;
+    uint32_t rcrb_base_high;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_port_extensions) != 0x28);
+
+#define PORT_CONTROL_OFFSET          0xc
+#define PORT_CONTROL_UNMASK_SBR      1
+#define PORT_CONTROL_ALT_MEMID_EN    4
+
+/* CXL 2.0 - 8.1.6 GPF DVSEC (ID 0004) */
+struct cxl_dvsec_port_gpf {
+    struct dvsec_header hdr;
+    uint16_t rsvd;
+    uint16_t phase1_ctrl;
+    uint16_t phase2_ctrl;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_port_gpf) != 0x10);
+
+/* CXL 2.0 - 8.1.8/8.2.1.3 Flexbus DVSEC (ID 0007) */
+struct cxl_dvsec_port_flexbus {
+    struct dvsec_header hdr;
+    uint16_t cap;
+    uint16_t ctrl;
+    uint16_t status;
+    uint32_t rcvd_mod_ts_data_phase1;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_port_flexbus) != 0x14);
+
+/* CXL 2.0 - 8.1.9 Register Locator DVSEC (ID 0008) */
+struct cxl_dvsec_register_locator {
+    struct dvsec_header hdr;
+    uint16_t rsvd;
+    uint32_t reg0_base_lo;
+    uint32_t reg0_base_hi;
+    uint32_t reg1_base_lo;
+    uint32_t reg1_base_hi;
+    uint32_t reg2_base_lo;
+    uint32_t reg2_base_hi;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_register_locator) != 0x24);
+
+/* BAR Equivalence Indicator */
+#define BEI_BAR_10H 0
+#define BEI_BAR_14H 1
+#define BEI_BAR_18H 2
+#define BEI_BAR_1cH 3
+#define BEI_BAR_20H 4
+#define BEI_BAR_24H 5
+
+/* Register Block Identifier */
+#define RBI_EMPTY          0
+#define RBI_COMPONENT_REG  (1 << 8)
+#define RBI_BAR_VIRT_ACL   (2 << 8)
+#define RBI_CXL_DEVICE_REG (3 << 8)
+
+#endif
#5 lore raw

Re: [PATCH v8 02/46] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)

In reply to Jonathan Cameron
Quoted 76 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL 2.0 component is any entity in the CXL topology. All components have a analogous function in PCIe. Except for the CXL host bridge, all have a PCIe config space that is accessible via the common PCIe mechanisms. CXL components are enumerated via DVSEC fields in the extended PCIe header space. CXL components will minimally implement some subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL 2.0 specification. Two headers and a utility library are introduced to support the minimum functionality needed to enumerate components.

The cxl_pci header manages bits associated with PCI, specifically the DVSEC and related fields. The cxl_component.h variant has data structures and APIs that are useful for drivers implementing any of the CXL 2.0 components. The library takes care of making use of the DVSEC bits and the CXL.[mem|cache] registers. Per spec, the registers are little endian.

None of the mechanisms required to enumerate a CXL capable hostbridge are introduced at this point.

Note that the CXL.mem and CXL.cache registers used are always 4B wide. It's possible in the future that this constraint will not hold.

---
 hw/Kconfig                     |   1 +
 hw/cxl/Kconfig                 |   3 +
 hw/cxl/cxl-component-utils.c   | 219 +++++++++++++++++++++++++++++++++
 hw/cxl/meson.build             |   4 +
 hw/meson.build                 |   1 +
 include/hw/cxl/cxl.h           |  16 +++
 include/hw/cxl/cxl_component.h | 197 +++++++++++++++++++++++++++++
 include/hw/cxl/cxl_pci.h       | 135 ++++++++++++++++++++
 8 files changed, 576 insertions(+)
hw/Kconfig +1 -0
index ad20cce0a9..50e0952889 100644
@@ -6,6 +6,7 @@ source audio/Kconfig
 source block/Kconfig
 source char/Kconfig
 source core/Kconfig
+source cxl/Kconfig
 source display/Kconfig
 source dma/Kconfig
 source gpio/Kconfig
hw/cxl/Kconfig new file +3 -0
new file mode 100644
index 0000000000..8e67519b16
@@ -0,0 +1,3 @@
+config CXL
+    bool
+    default y if PCI_EXPRESS
hw/cxl/cxl-component-utils.c new file +11 -0
new file mode 100644
index 0000000000..410f8ef328
@@ -0,0 +1,219 @@
+/*
+ * CXL Utility library for components
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"

hw/cxl/cxl-component-utils.c line 21

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+#include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
+
+static uint64_t cxl_cache_mem_read_reg(void *opaque, hwaddr offset,
+                                       unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size == 8) {

Is there a define that can be used instead of 8 for clarity?

hw/cxl/cxl-component-utils.c line 30

Quoted 9 lines
hw/cxl/cxl-component-utils.c new file +9 -0
+        qemu_log_mask(LOG_UNIMP,
+                      "CXL 8 byte cache mem registers not implemented\n");
+        return 0;
+    }
+
+    if (cregs->special_ops && cregs->special_ops->read) {
+        return cregs->special_ops->read(cxl_cstate, offset, size);
+    } else {
+        return cregs->cache_mem_registers[offset / 4];

I think this could benefit from a define as well.

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+    }
+}
+
+static void cxl_cache_mem_write_reg(void *opaque, hwaddr offset, uint64_t value,
+                                    unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size == 8) {

hw/cxl/cxl-component-utils.c line 50

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+        qemu_log_mask(LOG_UNIMP,
+                      "CXL 8 byte cache mem registers not implemented\n");
+        return;
+    }
+    if (cregs->special_ops && cregs->special_ops->write) {
+        cregs->special_ops->write(cxl_cstate, offset, value, size);
+    } else {
+        cregs->cache_mem_registers[offset / 4] = value;
+    }
+}

See comments for the read reg function they apply here as well.

Quoted 45 lines
hw/cxl/cxl-component-utils.c new file +45 -0
+
+/*
+ * 8.2.3
+ *   The access restrictions specified in Section 8.2.2 also apply to CXL 2.0
+ *   Component Registers.
+ *
+ * 8.2.2
+ *   • A 32 bit register shall be accessed as a 4 Bytes quantity. Partial
+ *   reads are not permitted.
+ *   • A 64 bit register shall be accessed as a 8 Bytes quantity. Partial
+ *   reads are not permitted.
+ *
+ * As of the spec defined today, only 4 byte registers exist.
+ */
+static const MemoryRegionOps cache_mem_ops = {
+    .read = cxl_cache_mem_read_reg,
+    .write = cxl_cache_mem_write_reg,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+        .unaligned = false,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type)
+{
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    memory_region_init(&cregs->component_registers, obj, type,
+                       CXL2_COMPONENT_BLOCK_SIZE);
+
+    /* io registers controls link which we don't care about in QEMU */
+    memory_region_init_io(&cregs->io, obj, NULL, cregs, ".io",
+                          CXL2_COMPONENT_IO_REGION_SIZE);
+    memory_region_init_io(&cregs->cache_mem, obj, &cache_mem_ops, cregs,
+                          ".cache_mem", CXL2_COMPONENT_CM_REGION_SIZE);
+
+    memory_region_add_subregion(&cregs->component_registers, 0, &cregs->io);

hw/cxl/cxl-component-utils.c line 105

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+    memory_region_add_subregion(&cregs->component_registers,
+                                CXL2_COMPONENT_IO_REGION_SIZE,
+                                &cregs->cache_mem);
+}
+
+static void ras_init_common(uint32_t *reg_state)
+{
+    reg_state[R_CXL_RAS_UNC_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_UNC_ERR_MASK] = 0x1cfff;
+    reg_state[R_CXL_RAS_UNC_ERR_SEVERITY] = 0x1cfff;

Should we add a comment that bits 12-13, 17-31 are reserved for the two previous register states, with all other bits set to 1 by default?

hw/cxl/cxl-component-utils.c line 107

Quoted 2 lines
hw/cxl/cxl-component-utils.c new file +2 -0
+    reg_state[R_CXL_RAS_COR_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_COR_ERR_MASK] = 0x3f;

I think this is supposed to be 0x7f

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+
+    /* CXL switches and devices must set */
+    reg_state[R_CXL_RAS_ERR_CAP_CTRL] = 0;
+}
+
+static void hdm_init_common(uint32_t *reg_state)
+{
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_GLOBAL_CONTROL,

hw/cxl/cxl-component-utils.c line 127

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+                     HDM_DECODER_ENABLE, 0);
+}
+
+void cxl_component_register_init_common(uint32_t *reg_state, enum reg_type type)
+{
+    int caps = 0;
+    switch (type) {
+    case CXL2_DOWNSTREAM_PORT:
+    case CXL2_DEVICE:
+        /* CAP, RAS, Link */

Minor nit, could move the CAP comment above the switch statement and provide a bit more info. Something like CAP HEADER ARRAY SIZE INITITALIZATION?

Quoted 30 lines
hw/cxl/cxl-component-utils.c new file +30 -0
+        caps = 2;
+        break;
+    case CXL2_UPSTREAM_PORT:
+    case CXL2_TYPE3_DEVICE:
+    case CXL2_LOGICAL_DEVICE:
+        /* + HDM */
+        caps = 3;
+        break;
+    case CXL2_ROOT_PORT:
+        /* + Extended Security, + Snoop */
+        caps = 5;
+        break;
+    default:
+        abort();
+    }
+
+    memset(reg_state, 0, CXL2_COMPONENT_CM_REGION_SIZE);
+
+    /* CXL Capability Header Register */
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ID, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ARRAY_SIZE, caps);
+
+
+#define init_cap_reg(reg, id, version)                                        \
+    QEMU_BUILD_BUG_ON(CXL_##reg##_REGISTERS_OFFSET == 0);                     \
+    do {                                                                      \
+        int which = R_CXL_##reg##_CAPABILITY_HEADER;                          \
+        reg_state[which] = FIELD_DP32(reg_state[which],                       \

hw/cxl/cxl-component-utils.c line 167

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+                                      CXL_##reg##_CAPABILITY_HEADER, ID, id); \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER,       \
+                       VERSION, version);                                     \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER, PTR,  \
+                       CXL_##reg##_REGISTERS_OFFSET);                         \
+    } while (0)
+
+    init_cap_reg(RAS, 2, 1);

Is the version of the RAS cap header supposed to be 1 here?

Quoted 181 lines
hw/cxl/cxl-component-utils.c new file +52 -0
+    ras_init_common(reg_state);
+
+    init_cap_reg(LINK, 4, 2);
+
+    if (caps < 3) {
+        return;
+    }
+
+    init_cap_reg(HDM, 5, 1);
+    hdm_init_common(reg_state);
+
+    if (caps < 5) {
+        return;
+    }
+
+    init_cap_reg(EXTSEC, 6, 1);
+    init_cap_reg(SNOOP, 8, 1);
+
+#undef init_cap_reg
+}
+
+/*
+ * Helper to creates a DVSEC header for a CXL entity. The caller is responsible
+ * for tracking the valid offset.
+ *
+ * This function will build the DVSEC header on behalf of the caller and then
+ * copy in the remaining data for the vendor specific bits.
+ */
+void cxl_component_create_dvsec(CXLComponentState *cxl, uint16_t length,
+                                uint16_t type, uint8_t rev, uint8_t *body)
+{
+    PCIDevice *pdev = cxl->pdev;
+    uint16_t offset = cxl->dvsec_offset;
+
+    assert(offset >= PCI_CFG_SPACE_SIZE &&
+           ((offset + length) < PCI_CFG_SPACE_EXP_SIZE));
+    assert((length & 0xf000) == 0);
+    assert((rev & ~0xf) == 0);
+
+    /* Create the DVSEC in the MCFG space */
+    pcie_add_capability(pdev, PCI_EXT_CAP_ID_DVSEC, 1, offset, length);
+    pci_set_long(pdev->config + offset + PCIE_DVSEC_HEADER1_OFFSET,
+                 (length << 20) | (rev << 16) | CXL_VENDOR_ID);
+    pci_set_word(pdev->config + offset + PCIE_DVSEC_ID_OFFSET, type);
+    memcpy(pdev->config + offset + sizeof(struct dvsec_header),
+           body + sizeof(struct dvsec_header),
+           length - sizeof(struct dvsec_header));
+
+    /* Update state for future DVSEC additions */
+    range_init_nofail(&cxl->dvsecs[type], cxl->dvsec_offset, length);
+    cxl->dvsec_offset += length;
+}
hw/cxl/meson.build new file +4 -0
new file mode 100644
index 0000000000..3231b5de1e
@@ -0,0 +1,4 @@
+softmmu_ss.add(when: 'CONFIG_CXL',
+               if_true: files(
+                   'cxl-component-utils.c',
+               ))
hw/meson.build +1 -0
index b3366c888e..9992c5101e 100644
@@ -6,6 +6,7 @@ subdir('block')
 subdir('char')
 subdir('core')
 subdir('cpu')
+subdir('cxl')
 subdir('display')
 subdir('dma')
 subdir('gpio')
include/hw/cxl/cxl.h new file +16 -0
new file mode 100644
index 0000000000..8c738c7a2b
@@ -0,0 +1,16 @@
+/*
+ * QEMU CXL Support
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_H
+#define CXL_H
+
+#include "cxl_pci.h"
+#include "cxl_component.h"
+
+#endif
include/hw/cxl/cxl_component.h new file +79 -0
new file mode 100644
index 0000000000..74e9bfe1ff
@@ -0,0 +1,197 @@
+/*
+ * QEMU CXL Component
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_COMPONENT_H
+#define CXL_COMPONENT_H
+
+/* CXL 2.0 - 8.2.4 */
+#define CXL2_COMPONENT_IO_REGION_SIZE 0x1000
+#define CXL2_COMPONENT_CM_REGION_SIZE 0x1000
+#define CXL2_COMPONENT_BLOCK_SIZE 0x10000
+
+#include "qemu/compiler.h"
+#include "qemu/range.h"
+#include "qemu/typedefs.h"
+#include "hw/register.h"
+
+enum reg_type {
+    CXL2_DEVICE,
+    CXL2_TYPE3_DEVICE,
+    CXL2_LOGICAL_DEVICE,
+    CXL2_ROOT_PORT,
+    CXL2_UPSTREAM_PORT,
+    CXL2_DOWNSTREAM_PORT
+};
+
+/*
+ * Capability registers are defined at the top of the CXL.cache/mem region and
+ * are packed. For our purposes we will always define the caps in the same
+ * order.
+ * CXL 2.0 - 8.2.5 Table 142 for details.
+ */
+
+/* CXL 2.0 - 8.2.5.1 */
+REG32(CXL_CAPABILITY_HEADER, 0)
+    FIELD(CXL_CAPABILITY_HEADER, ID, 0, 16)
+    FIELD(CXL_CAPABILITY_HEADER, VERSION, 16, 4)
+    FIELD(CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 20, 4)
+    FIELD(CXL_CAPABILITY_HEADER, ARRAY_SIZE, 24, 8)
+
+#define CXLx_CAPABILITY_HEADER(type, offset)                  \
+    REG32(CXL_##type##_CAPABILITY_HEADER, offset)             \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, ID, 0, 16)      \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, VERSION, 16, 4) \
+        FIELD(CXL_##type##_CAPABILITY_HEADER, PTR, 20, 12)
+CXLx_CAPABILITY_HEADER(RAS, 0x4)
+CXLx_CAPABILITY_HEADER(LINK, 0x8)
+CXLx_CAPABILITY_HEADER(HDM, 0xc)
+CXLx_CAPABILITY_HEADER(EXTSEC, 0x10)
+CXLx_CAPABILITY_HEADER(SNOOP, 0x14)
+
+/*
+ * Capability structures contain the actual registers that the CXL component
+ * implements. Some of these are specific to certain types of components, but
+ * this implementation leaves enough space regardless.
+ */
+/* 8.2.5.9 - CXL RAS Capability Structure */
+
+/* Give ample space for caps before this */
+#define CXL_RAS_REGISTERS_OFFSET 0x80
+#define CXL_RAS_REGISTERS_SIZE   0x58
+REG32(CXL_RAS_UNC_ERR_STATUS, CXL_RAS_REGISTERS_OFFSET)
+REG32(CXL_RAS_UNC_ERR_MASK, CXL_RAS_REGISTERS_OFFSET + 0x4)
+REG32(CXL_RAS_UNC_ERR_SEVERITY, CXL_RAS_REGISTERS_OFFSET + 0x8)
+REG32(CXL_RAS_COR_ERR_STATUS, CXL_RAS_REGISTERS_OFFSET + 0xc)
+REG32(CXL_RAS_COR_ERR_MASK, CXL_RAS_REGISTERS_OFFSET + 0x10)
+REG32(CXL_RAS_ERR_CAP_CTRL, CXL_RAS_REGISTERS_OFFSET + 0x14)
+/* Offset 0x18 - 0x58 reserved for RAS logs */
+
+/* 8.2.5.10 - CXL Security Capability Structure */
+#define CXL_SEC_REGISTERS_OFFSET \
+    (CXL_RAS_REGISTERS_OFFSET + CXL_RAS_REGISTERS_SIZE)
+#define CXL_SEC_REGISTERS_SIZE   0 /* We don't implement 1.1 downstream ports */
+

include/hw/cxl/cxl_component.h line 89

Quoted 10 lines
include/hw/cxl/cxl_component.h new file +10 -0
+/* 8.2.5.11 - CXL Link Capability Structure */
+#define CXL_LINK_REGISTERS_OFFSET \
+    (CXL_SEC_REGISTERS_OFFSET + CXL_SEC_REGISTERS_SIZE)
+#define CXL_LINK_REGISTERS_SIZE   0x38
+
+/* 8.2.5.12 - CXL HDM Decoder Capability Structure */
+#define HDM_DECODE_MAX 10 /* 8.2.5.12.1 */
+#define CXL_HDM_REGISTERS_OFFSET \
+    (CXL_LINK_REGISTERS_OFFSET + CXL_LINK_REGISTERS_SIZE)
+#define CXL_HDM_REGISTERS_SIZE (0x20 + HDM_DECODE_MAX + 10)

Do we need to multiply HDM_DECODE_MAX here, I am under the impression that HDM_DECODE_MAX represents the number of decoders and we need register space for each of the possible decoders.

Quoted 183 lines
include/hw/cxl/cxl_component.h new file +108 -0
+#define HDM_DECODER_INIT(n)                                                    \
+  REG32(CXL_HDM_DECODER##n##_BASE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x10)                          \
+            FIELD(CXL_HDM_DECODER##n##_BASE_LO, L, 28, 4)                      \
+  REG32(CXL_HDM_DECODER##n##_BASE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x14)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x18)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x1C)                          \
+  REG32(CXL_HDM_DECODER##n##_CTRL,                                             \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x20)                          \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IG, 0, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IW, 4, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, LOCK_ON_COMMIT, 8, 1)             \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMIT, 9, 1)                     \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMITTED, 10, 1)                 \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, ERR, 11, 1)                       \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, TYPE, 12, 1)                      \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_LO,                                   \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x24)                          \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_HI,                                   \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x28)
+
+REG32(CXL_HDM_DECODER_CAPABILITY, CXL_HDM_REGISTERS_OFFSET)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 4, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTERLEAVE_256B, 8, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTELEAVE_4K, 9, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, POISON_ON_ERR_CAP, 10, 1)
+REG32(CXL_HDM_DECODER_GLOBAL_CONTROL, CXL_HDM_REGISTERS_OFFSET + 4)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, POISON_ON_ERR_EN, 0, 1)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, HDM_DECODER_ENABLE, 1, 1)
+
+HDM_DECODER_INIT(0);
+
+/* 8.2.5.13 - CXL Extended Security Capability Structure (Root complex only) */
+#define EXTSEC_ENTRY_MAX        256
+#define CXL_EXTSEC_REGISTERS_OFFSET \
+    (CXL_HDM_REGISTERS_OFFSET + CXL_HDM_REGISTERS_SIZE)
+#define CXL_EXTSEC_REGISTERS_SIZE   (8 * EXTSEC_ENTRY_MAX + 4)
+
+/* 8.2.5.14 - CXL IDE Capability Structure */
+#define CXL_IDE_REGISTERS_OFFSET \
+    (CXL_EXTSEC_REGISTERS_OFFSET + CXL_EXTSEC_REGISTERS_SIZE)
+#define CXL_IDE_REGISTERS_SIZE   0x20
+
+/* 8.2.5.15 - CXL Snoop Filter Capability Structure */
+#define CXL_SNOOP_REGISTERS_OFFSET \
+    (CXL_IDE_REGISTERS_OFFSET + CXL_IDE_REGISTERS_SIZE)
+#define CXL_SNOOP_REGISTERS_SIZE   0x8
+
+QEMU_BUILD_BUG_MSG((CXL_SNOOP_REGISTERS_OFFSET + CXL_SNOOP_REGISTERS_SIZE) >= 0x1000,
+                   "No space for registers");
+
+typedef struct component_registers {
+    /*
+     * Main memory region to be registered with QEMU core.
+     */
+    MemoryRegion component_registers;
+
+    /*
+     * 8.2.4 Table 141:
+     *   0x0000 - 0x0fff CXL.io registers
+     *   0x1000 - 0x1fff CXL.cache and CXL.mem
+     *   0x2000 - 0xdfff Implementation specific
+     *   0xe000 - 0xe3ff CXL ARB/MUX registers
+     *   0xe400 - 0xffff RSVD
+     */
+    uint32_t io_registers[CXL2_COMPONENT_IO_REGION_SIZE >> 2];
+    MemoryRegion io;
+
+    uint32_t cache_mem_registers[CXL2_COMPONENT_CM_REGION_SIZE >> 2];
+    MemoryRegion cache_mem;
+
+    MemoryRegion impl_specific;
+    MemoryRegion arb_mux;
+    MemoryRegion rsvd;
+
+    /* special_ops is used for any component that needs any specific handling */
+    MemoryRegionOps *special_ops;
+} ComponentRegisters;
+
+/*
+ * A CXL component represents all entities in a CXL hierarchy. This includes,
+ * host bridges, root ports, upstream/downstream switch ports, and devices
+ */
+typedef struct cxl_component {
+    ComponentRegisters crb;
+    union {
+        struct {
+            Range dvsecs[CXL20_MAX_DVSEC];
+            uint16_t dvsec_offset;
+            struct PCIDevice *pdev;
+        };
+    };
+} CXLComponentState;
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type);
+void cxl_component_register_init_common(uint32_t *reg_state,
+                                        enum reg_type type);
+
+void cxl_component_create_dvsec(CXLComponentState *cxl_cstate, uint16_t length,
+                                uint16_t type, uint8_t rev, uint8_t *body);
+
+#endif
include/hw/cxl/cxl_pci.h new file +69 -0
new file mode 100644
index 0000000000..810a244fab
@@ -0,0 +1,135 @@
+/*
+ * QEMU CXL PCI interfaces
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_PCI_H
+#define CXL_PCI_H
+
+#include "qemu/compiler.h"
+#include "hw/pci/pci.h"
+#include "hw/pci/pcie.h"
+
+#define CXL_VENDOR_ID 0x1e98
+
+#define PCIE_DVSEC_HEADER1_OFFSET 0x4 /* Offset from start of extend cap */
+#define PCIE_DVSEC_ID_OFFSET 0x8
+
+#define PCIE_CXL_DEVICE_DVSEC_LENGTH 0x38
+#define PCIE_CXL1_DEVICE_DVSEC_REVID 0
+#define PCIE_CXL2_DEVICE_DVSEC_REVID 1
+
+#define EXTENSIONS_PORT_DVSEC_LENGTH 0x28
+#define EXTENSIONS_PORT_DVSEC_REVID 0
+
+#define GPF_PORT_DVSEC_LENGTH 0x10
+#define GPF_PORT_DVSEC_REVID  0
+
+#define PCIE_FLEXBUS_PORT_DVSEC_LENGTH_2_0 0x14
+#define PCIE_FLEXBUS_PORT_DVSEC_REVID_2_0  1
+
+#define REG_LOC_DVSEC_LENGTH 0x24
+#define REG_LOC_DVSEC_REVID  0
+
+enum {
+    PCIE_CXL_DEVICE_DVSEC      = 0,
+    NON_CXL_FUNCTION_MAP_DVSEC = 2,
+    EXTENSIONS_PORT_DVSEC      = 3,
+    GPF_PORT_DVSEC             = 4,
+    GPF_DEVICE_DVSEC           = 5,
+    PCIE_FLEXBUS_PORT_DVSEC    = 7,
+    REG_LOC_DVSEC              = 8,
+    MLD_DVSEC                  = 9,
+    CXL20_MAX_DVSEC
+};
+
+struct dvsec_header {
+    uint32_t cap_hdr;
+    uint32_t dv_hdr1;
+    uint16_t dv_hdr2;
+} QEMU_PACKED;
+QEMU_BUILD_BUG_ON(sizeof(struct dvsec_header) != 10);
+
+/*
+ * CXL 2.0 devices must implement certain DVSEC IDs, and can [optionally]
+ * implement others.
+ *
+ * CXL 2.0 Device: 0, [2], 5, 8
+ * CXL 2.0 RP: 3, 4, 7, 8
+ * CXL 2.0 Upstream Port: [2], 7, 8
+ * CXL 2.0 Downstream Port: 3, 4, 7, 8
+ */
+
+/* CXL 2.0 - 8.1.5 (ID 0003) */
+struct cxl_dvsec_port_extensions {
+    struct dvsec_header hdr;

include/hw/cxl/cxl_pci.h line 79

Quoted 10 lines
include/hw/cxl/cxl_pci.h new file +10 -0
+    uint16_t status;
+    uint16_t control;
+    uint8_t alt_bus_base;
+    uint8_t alt_bus_limit;
+    uint16_t alt_memory_base;
+    uint16_t alt_memory_limit;
+    uint16_t alt_prefetch_base;
+    uint16_t alt_prefetch_limit;
+    uint32_t alt_prefetch_base_high;
+    uint32_t alt_prefetch_base_low;

Limit high?

Quoted 60 lines
include/hw/cxl/cxl_pci.h new file +56 -0
+    uint32_t rcrb_base;
+    uint32_t rcrb_base_high;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_port_extensions) != 0x28);
+
+#define PORT_CONTROL_OFFSET          0xc
+#define PORT_CONTROL_UNMASK_SBR      1
+#define PORT_CONTROL_ALT_MEMID_EN    4
+
+/* CXL 2.0 - 8.1.6 GPF DVSEC (ID 0004) */
+struct cxl_dvsec_port_gpf {
+    struct dvsec_header hdr;
+    uint16_t rsvd;
+    uint16_t phase1_ctrl;
+    uint16_t phase2_ctrl;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_port_gpf) != 0x10);
+
+/* CXL 2.0 - 8.1.8/8.2.1.3 Flexbus DVSEC (ID 0007) */
+struct cxl_dvsec_port_flexbus {
+    struct dvsec_header hdr;
+    uint16_t cap;
+    uint16_t ctrl;
+    uint16_t status;
+    uint32_t rcvd_mod_ts_data_phase1;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_port_flexbus) != 0x14);
+
+/* CXL 2.0 - 8.1.9 Register Locator DVSEC (ID 0008) */
+struct cxl_dvsec_register_locator {
+    struct dvsec_header hdr;
+    uint16_t rsvd;
+    uint32_t reg0_base_lo;
+    uint32_t reg0_base_hi;
+    uint32_t reg1_base_lo;
+    uint32_t reg1_base_hi;
+    uint32_t reg2_base_lo;
+    uint32_t reg2_base_hi;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_register_locator) != 0x24);
+
+/* BAR Equivalence Indicator */
+#define BEI_BAR_10H 0
+#define BEI_BAR_14H 1
+#define BEI_BAR_18H 2
+#define BEI_BAR_1cH 3
+#define BEI_BAR_20H 4
+#define BEI_BAR_24H 5
+
+/* Register Block Identifier */
+#define RBI_EMPTY          0
+#define RBI_COMPONENT_REG  (1 << 8)
+#define RBI_BAR_VIRT_ACL   (2 << 8)
+#define RBI_CXL_DEVICE_REG (3 << 8)
+
+#endif
Signature · 1 line
2.32.0

+cc (Klaus, Dave, Tong)

Other than the minor cleanups/nits.
Looks good.

Reviewed by: Adam Manzanares <[email protected]>

#6 lore raw

Re: [PATCH v8 02/46] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)

Quoted 39 lines · On Mon, 28 Mar 2022 14:28:41 +0000 Adam Manzanares <[email protected]>…
Quoted 38 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL 2.0 component is any entity in the CXL topology. All components have a analogous function in PCIe. Except for the CXL host bridge, all have a PCIe config space that is accessible via the common PCIe mechanisms. CXL components are enumerated via DVSEC fields in the extended PCIe header space. CXL components will minimally implement some subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL 2.0 specification. Two headers and a utility library are introduced to support the minimum functionality needed to enumerate components.

The cxl_pci header manages bits associated with PCI, specifically the DVSEC and related fields. The cxl_component.h variant has data structures and APIs that are useful for drivers implementing any of the CXL 2.0 components. The library takes care of making use of the DVSEC bits and the CXL.[mem|cache] registers. Per spec, the registers are little endian.

None of the mechanisms required to enumerate a CXL capable hostbridge are introduced at this point.

Note that the CXL.mem and CXL.cache registers used are always 4B wide. It's possible in the future that this constraint will not hold.

---
 hw/Kconfig                     |   1 +
 hw/cxl/Kconfig                 |   3 +
 hw/cxl/cxl-component-utils.c   | 219 +++++++++++++++++++++++++++++++++
 hw/cxl/meson.build             |   4 +
 hw/meson.build                 |   1 +
 include/hw/cxl/cxl.h           |  16 +++
 include/hw/cxl/cxl_component.h | 197 +++++++++++++++++++++++++++++
 include/hw/cxl/cxl_pci.h       | 135 ++++++++++++++++++++
 8 files changed, 576 insertions(+)

...

Quoted 21 lines
Quoted 21 lines
hw/cxl/cxl-component-utils.c new file +15 -0
new file mode 100644
index 0000000000..410f8ef328
@@ -0,0 +1,219 @@
+/*
+ * CXL Utility library for components
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
+
+static uint64_t cxl_cache_mem_read_reg(void *opaque, hwaddr offset,
Quoted 8 lines

hw/cxl/cxl-component-utils.c line 21

Quoted 6 lines
hw/cxl/cxl-component-utils.c new file +6 -0
+                                       unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size == 8) {  

Is there a define that can be used instead of 8 for clarity?

I can't think of one that would be clearer than the number as we just checking if it is 8.

Quoted 11 lines

hw/cxl/cxl-component-utils.c line 30

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +9 -0
+        qemu_log_mask(LOG_UNIMP,
+                      "CXL 8 byte cache mem registers not implemented\n");
+        return 0;
+    }
+
+    if (cregs->special_ops && cregs->special_ops->read) {
+        return cregs->special_ops->read(cxl_cstate, offset, size);
+    } else {
+        return cregs->cache_mem_registers[offset / 4];  
Quoted 1 line

I think this could benefit from a define as well.

For this one, we can use offset / sizeof(*cregs->cache_mem_registers) Bit of a long line but clear that we are indexing into the array.

Quoted 24 lines
Quoted 11 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+    }
+}
+
+static void cxl_cache_mem_write_reg(void *opaque, hwaddr offset, uint64_t value,
+                                    unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    if (size == 8) {

hw/cxl/cxl-component-utils.c line 50

Quoted 10 lines
hw/cxl/cxl-component-utils.c new file +10 -0
+        qemu_log_mask(LOG_UNIMP,
+                      "CXL 8 byte cache mem registers not implemented\n");
+        return;
+    }
+    if (cregs->special_ops && cregs->special_ops->write) {
+        cregs->special_ops->write(cxl_cstate, offset, value, size);
+    } else {
+        cregs->cache_mem_registers[offset / 4] = value;
+    }
+}  

See comments for the read reg function they apply here as well.

...

Quoted 22 lines
Quoted 22 lines
hw/cxl/cxl-component-utils.c new file +22 -0
+
+void cxl_component_register_block_init(Object *obj,
+                                       CXLComponentState *cxl_cstate,
+                                       const char *type)
+{
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+
+    memory_region_init(&cregs->component_registers, obj, type,
+                       CXL2_COMPONENT_BLOCK_SIZE);
+
+    /* io registers controls link which we don't care about in QEMU */
+    memory_region_init_io(&cregs->io, obj, NULL, cregs, ".io",
+                          CXL2_COMPONENT_IO_REGION_SIZE);
+    memory_region_init_io(&cregs->cache_mem, obj, &cache_mem_ops, cregs,
+                          ".cache_mem", CXL2_COMPONENT_CM_REGION_SIZE);
+
+    memory_region_add_subregion(&cregs->component_registers, 0, &cregs->io);
+    memory_region_add_subregion(&cregs->component_registers,
+                                CXL2_COMPONENT_IO_REGION_SIZE,
+                                &cregs->cache_mem);
+}
+
Quoted 8 lines

hw/cxl/cxl-component-utils.c line 105

Quoted 5 lines
hw/cxl/cxl-component-utils.c new file +5 -0
+static void ras_init_common(uint32_t *reg_state)
+{
+    reg_state[R_CXL_RAS_UNC_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_UNC_ERR_MASK] = 0x1cfff;
+    reg_state[R_CXL_RAS_UNC_ERR_SEVERITY] = 0x1cfff;  

Should we add a comment that bits 12-13, 17-31 are reserved for the two previous register states, with all other bits set to 1 by default?

Ok. This one is unusual enough that it's worth calling out he reserved bits. I'm not sure it's worth stating defaults of other bits though as we will need a lot of words to explain all the similar spaces.

Quoted 5 lines

hw/cxl/cxl-component-utils.c line 107

Quoted 3 lines
hw/cxl/cxl-component-utils.c new file +2 -0
+    reg_state[R_CXL_RAS_COR_ERR_STATUS] = 0;
+    reg_state[R_CXL_RAS_COR_ERR_MASK] = 0x3f;  

I think this is supposed to be 0x7f

Oops. Fixed now.

Quoted 16 lines
Quoted 16 lines
hw/cxl/cxl-component-utils.c new file +15 -0
+
+    /* CXL switches and devices must set */
+    reg_state[R_CXL_RAS_ERR_CAP_CTRL] = 0;
+}
+
+static void hdm_init_common(uint32_t *reg_state)
+{
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_GLOBAL_CONTROL,
+                     HDM_DECODER_ENABLE, 0);
+}
+
+void cxl_component_register_init_common(uint32_t *reg_state, enum reg_type type)
+{
Quoted 8 lines

hw/cxl/cxl-component-utils.c line 127

Quoted 5 lines
hw/cxl/cxl-component-utils.c new file +5 -0
+    int caps = 0;
+    switch (type) {
+    case CXL2_DOWNSTREAM_PORT:
+    case CXL2_DEVICE:
+        /* CAP, RAS, Link */  

Minor nit, could move the CAP comment above the switch statement and provide a bit more info. Something like CAP HEADER ARRAY SIZE INITITALIZATION?

This is weird enough (and kind of relies on a coincidence) that I agree more commentary is needed. I've introduced it with the following above the switch (type).

/*
* In CXL 2.0 the capabilities required for each CXL component are such that,
* with the ordering chosen here, a single number can be used to define
* which capabilities should be provided.
*/

and dropped CAP from this first comment.

Quoted 35 lines
Quoted 35 lines
hw/cxl/cxl-component-utils.c new file +34 -0
+        caps = 2;
+        break;
+    case CXL2_UPSTREAM_PORT:
+    case CXL2_TYPE3_DEVICE:
+    case CXL2_LOGICAL_DEVICE:
+        /* + HDM */
+        caps = 3;
+        break;
+    case CXL2_ROOT_PORT:
+        /* + Extended Security, + Snoop */
+        caps = 5;
+        break;
+    default:
+        abort();
+    }
+
+    memset(reg_state, 0, CXL2_COMPONENT_CM_REGION_SIZE);
+
+    /* CXL Capability Header Register */
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ID, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, CACHE_MEM_VERSION, 1);
+    ARRAY_FIELD_DP32(reg_state, CXL_CAPABILITY_HEADER, ARRAY_SIZE, caps);
+
+
+#define init_cap_reg(reg, id, version)                                        \
+    QEMU_BUILD_BUG_ON(CXL_##reg##_REGISTERS_OFFSET == 0);                     \
+    do {                                                                      \
+        int which = R_CXL_##reg##_CAPABILITY_HEADER;                          \
+        reg_state[which] = FIELD_DP32(reg_state[which],                       \
+                                      CXL_##reg##_CAPABILITY_HEADER, ID, id); \
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER,       \
+                       VERSION, version);                                     \
Quoted 8 lines

hw/cxl/cxl-component-utils.c line 167

Quoted 6 lines
hw/cxl/cxl-component-utils.c new file +6 -0
+        reg_state[which] =                                                    \
+            FIELD_DP32(reg_state[which], CXL_##reg##_CAPABILITY_HEADER, PTR,  \
+                       CXL_##reg##_REGISTERS_OFFSET);                         \
+    } while (0)
+
+    init_cap_reg(RAS, 2, 1);  

Is the version of the RAS cap header supposed to be 1 here?

Ah CXL 2.0 has this as version 2. Good spot.

Quoted 21 lines
Quoted 21 lines
hw/cxl/cxl-component-utils.c new file +20 -0
+    ras_init_common(reg_state);
+
+    init_cap_reg(LINK, 4, 2);
+
+    if (caps < 3) {
+        return;
+    }
+
+    init_cap_reg(HDM, 5, 1);
+    hdm_init_common(reg_state);
+
+    if (caps < 5) {
+        return;
+    }
+
+    init_cap_reg(EXTSEC, 6, 1);
+    init_cap_reg(SNOOP, 8, 1);
+
+#undef init_cap_reg
+}

...

Quoted 6 lines
Quoted 6 lines
include/hw/cxl/cxl_component.h new file +0 -0
new file mode 100644
index 0000000000..74e9bfe1ff
@@ -0,0 +1,197 @@

...

Quoted 9 lines

include/hw/cxl/cxl_component.h line 89

Quoted 5 lines
include/hw/cxl/cxl_component.h new file +5 -0
+/* 8.2.5.12 - CXL HDM Decoder Capability Structure */
+#define HDM_DECODE_MAX 10 /* 8.2.5.12.1 */
+#define CXL_HDM_REGISTERS_OFFSET \
+    (CXL_LINK_REGISTERS_OFFSET + CXL_LINK_REGISTERS_SIZE)
+#define CXL_HDM_REGISTERS_SIZE (0x20 + HDM_DECODE_MAX + 10)  

Do we need to multiply HDM_DECODE_MAX here, I am under the impression that HDM_DECODE_MAX represents the number of decoders and we need register space for each of the possible decoders.

Indeed curious. We don't currently hit this because the code only enables a single HDM decoder so far.

It should be
0x10 + 0x20 * HDM_DECODE_MAX assuming intent is to leave space for up to 10.

Previously the last few bytes would have spilled into the next cap. As those are either DPA skip which isn't implemented yet or the top of the target list (which I never hit as only did 4 way at most at a given decoder).

Quoted 37 lines
Quoted 37 lines
include/hw/cxl/cxl_component.h new file +36 -0
+#define HDM_DECODER_INIT(n)                                                    \
+  REG32(CXL_HDM_DECODER##n##_BASE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x10)                          \
+            FIELD(CXL_HDM_DECODER##n##_BASE_LO, L, 28, 4)                      \
+  REG32(CXL_HDM_DECODER##n##_BASE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x14)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_LO,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x18)                          \
+  REG32(CXL_HDM_DECODER##n##_SIZE_HI,                                          \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x1C)                          \
+  REG32(CXL_HDM_DECODER##n##_CTRL,                                             \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x20)                          \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IG, 0, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, IW, 4, 4)                         \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, LOCK_ON_COMMIT, 8, 1)             \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMIT, 9, 1)                     \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, COMMITTED, 10, 1)                 \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, ERR, 11, 1)                       \
+            FIELD(CXL_HDM_DECODER##n##_CTRL, TYPE, 12, 1)                      \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_LO,                                   \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x24)                          \
+  REG32(CXL_HDM_DECODER##n##_TARGET_LIST_HI,                                   \
+        CXL_HDM_REGISTERS_OFFSET + (0x20 * n) + 0x28)
+
+REG32(CXL_HDM_DECODER_CAPABILITY, CXL_HDM_REGISTERS_OFFSET)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, DECODER_COUNT, 0, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 4, 4)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTERLEAVE_256B, 8, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, INTELEAVE_4K, 9, 1)
+    FIELD(CXL_HDM_DECODER_CAPABILITY, POISON_ON_ERR_CAP, 10, 1)
+REG32(CXL_HDM_DECODER_GLOBAL_CONTROL, CXL_HDM_REGISTERS_OFFSET + 4)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, POISON_ON_ERR_EN, 0, 1)
+    FIELD(CXL_HDM_DECODER_GLOBAL_CONTROL, HDM_DECODER_ENABLE, 1, 1)
+
+HDM_DECODER_INIT(0);
+

...

Quoted 5 lines
Quoted 5 lines
include/hw/cxl/cxl_pci.h new file +0 -0
new file mode 100644
index 0000000000..810a244fab

...

Quoted 8 lines
Quoted 8 lines
include/hw/cxl/cxl_pci.h new file +8 -0
+
+/* CXL 2.0 - 8.1.5 (ID 0003) */
+struct cxl_dvsec_port_extensions {
+    struct dvsec_header hdr;
+    uint16_t status;
+    uint16_t control;
+    uint8_t alt_bus_base;
+    uint8_t alt_bus_limit;
Quoted 9 lines

include/hw/cxl/cxl_pci.h line 79

Quoted 6 lines
include/hw/cxl/cxl_pci.h new file +6 -0
+    uint16_t alt_memory_base;
+    uint16_t alt_memory_limit;
+    uint16_t alt_prefetch_base;
+    uint16_t alt_prefetch_limit;
+    uint32_t alt_prefetch_base_high;
+    uint32_t alt_prefetch_base_low;  

Limit high?

Good spot. Fixed. (this one turned up when I enabled the write masks for these registers, but I'd failed to push it down into this patch).

Quoted 7 lines

+cc (Klaus, Dave, Tong)

Other than the minor cleanups/nits.
Looks good.

Reviewed by: Adam Manzanares <[email protected]>

Thanks for taking such a close look!

Jonathan

#7 lore raw

Re: [PATCH v8 02/46] hw/cxl/component: Introduce CXL components (8.1.x, 8.2.5)

In reply to Jonathan Cameron
Quoted 21 lines · On Fri, 18 Mar 2022 15:05:51 +0000 Jonathan Cameron…

From: Ben Widawsky <[email protected]>

A CXL 2.0 component is any entity in the CXL topology. All components have a analogous function in PCIe. Except for the CXL host bridge, all have a PCIe config space that is accessible via the common PCIe mechanisms. CXL components are enumerated via DVSEC fields in the extended PCIe header space. CXL components will minimally implement some subset of CXL.mem and CXL.cache registers defined in 8.2.5 of the CXL 2.0 specification. Two headers and a utility library are introduced to support the minimum functionality needed to enumerate components.

The cxl_pci header manages bits associated with PCI, specifically the DVSEC and related fields. The cxl_component.h variant has data structures and APIs that are useful for drivers implementing any of the CXL 2.0 components. The library takes care of making use of the DVSEC bits and the CXL.[mem|cache] registers. Per spec, the registers are little endian.

None of the mechanisms required to enumerate a CXL capable hostbridge are introduced at this point.

Quoted 6 lines

Note that the CXL.mem and CXL.cache registers used are always 4B wide. It's possible in the future that this constraint will not hold.

There will be one other addition to this in v9 which is to add the device GPF dvsec structure definition.

I'll also add the relevant instantiation of this to the type 3 device and the Port GPF instantiation for the root port and switch DSP as these are mandatory (Even if they aren't that interesting to an OS)

Quoted 1 line
#8 lore raw

PATCH v8 3/46 MAINTAINERS: Add entry for Compute Express Link Emulation

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

The CXL emulation will be jointly maintained by Ben Widawsky and Jonathan Cameron. Broken out as a separate patch to improve visibility.

---
 MAINTAINERS | 7 +++++++
 1 file changed, 7 insertions(+)
MAINTAINERS +7 -0
index 9aed5f3e04..2eb481022f 100644
@@ -2544,6 +2544,13 @@ F: qapi/block*.json
 F: qapi/transaction.json
 T: git https://repo.or.cz/qemu/armbru.git block-next
 
+Compute Express Link
+M: Ben Widawsky <[email protected]>
+M: Jonathan Cameron <[email protected]>
+S: Supported
+F: hw/cxl/
+F: include/hw/cxl/
+
 Dirty Bitmaps
 M: Eric Blake <[email protected]>
 M: Vladimir Sementsov-Ogievskiy <[email protected]>
Signature · 1 line
2.32.0
#9 lore raw

PATCH v8 4/46 hw/cxl/device: Introduce a CXL device (8.2.8)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL device is a type of CXL component. Conceptually, a CXL device would be a leaf node in a CXL topology. From an emulation perspective, CXL devices are the most complex and so the actual implementation is reserved for discrete commits.

This new device type is specifically catered towards the eventual implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0 specification.

---
 include/hw/cxl/cxl.h        |   1 +
 include/hw/cxl/cxl_device.h | 165 ++++++++++++++++++++++++++++++++++++
 2 files changed, 166 insertions(+)
include/hw/cxl/cxl.h +1 -0
index 8c738c7a2b..b9d1ac3fad 100644
@@ -12,5 +12,6 @@
 
 #include "cxl_pci.h"
 #include "cxl_component.h"
+#include "cxl_device.h"
 
 #endif
include/hw/cxl/cxl_device.h new file +165 -0

165 lines not on this page · raw message

Signature · 1 line
2.32.0
#10 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 47 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL device is a type of CXL component. Conceptually, a CXL device would be a leaf node in a CXL topology. From an emulation perspective, CXL devices are the most complex and so the actual implementation is reserved for discrete commits.

This new device type is specifically catered towards the eventual implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0 specification.

---
 include/hw/cxl/cxl.h        |   1 +
 include/hw/cxl/cxl_device.h | 165 ++++++++++++++++++++++++++++++++++++
 2 files changed, 166 insertions(+)
include/hw/cxl/cxl.h +1 -0
index 8c738c7a2b..b9d1ac3fad 100644
@@ -12,5 +12,6 @@
 
 #include "cxl_pci.h"
 #include "cxl_component.h"
+#include "cxl_device.h"
 
 #endif
include/hw/cxl/cxl_device.h new file +11 -0
new file mode 100644
index 0000000000..b2416e45bf
@@ -0,0 +1,165 @@
+/*
+ * QEMU CXL Devices
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_DEVICE_H
+#define CXL_DEVICE_H

include/hw/cxl/cxl_device.h line 21

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+
+#include "hw/register.h"
+
+/*
+ * The following is how a CXL device's MMIO space is laid out. The only
+ * requirement from the spec is that the capabilities array and the capability
+ * headers start at offset 0 and are contiguously packed. The headers themselves
+ * provide offsets to the register fields. For this emulation, registers will
+ * start at offset 0x80 (m == 0x80). No secondary mailbox is implemented which
+ * means that n = m + sizeof(mailbox registers) + sizeof(device registers).

What is n here, the start offset of the mailbox registers, this question is based on the figure below?

Quoted 28 lines
include/hw/cxl/cxl_device.h new file +28 -0
+ *
+ * This is roughly described in 8.2.8 Figure 138 of the CXL 2.0 spec.
+ *
+ *                       +---------------------------------+
+ *                       |                                 |
+ *                       |    Memory Device Registers      |
+ *                       |                                 |
+ * n + PAYLOAD_SIZE_MAX  -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |         Mailbox Payload         |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    -----------------------------------
+ *                  |    |       Mailbox Registers         |
+ *                  |    |                                 |
+ *                  n    -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |        Device Registers         |
+ *                  |    |                                 |
+ *                  m    ---------------------------------->
+ *                  ^    |  Memory Device Capability Header|
+ *                  |    -----------------------------------
+ *                  |    |     Mailbox Capability Header   |

include/hw/cxl/cxl_device.h line 59

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+ *                  |    -------------- --------------------
+ *                  |    |     Device Capability Header    |
+ *                  |    -----------------------------------
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |      Device Cap Array[0..n]     |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                       |                                 |
+ *                  0    +---------------------------------+

Would it make sense to add CXL cap header register to the diagram? n also seems to be the size of the cap array, but it is also an offset so that could be clarified.

include/hw/cxl/cxl_device.h line 67

Quoted 8 lines
include/hw/cxl/cxl_device.h new file +8 -0
+ *
+ */
+
+#define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
+#define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
+#define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
+
+#define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */

Is this to plan for future capabilities? If we have CAPS MAX doesn't this allow us to remove the slack space.

include/hw/cxl/cxl_device.h line 68

Quoted 1 line
include/hw/cxl/cxl_device.h new file +1 -0
+#define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */

Should we add status to the name here, or would it get too long?

include/hw/cxl/cxl_device.h line 73

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+
+#define CXL_MAILBOX_REGISTERS_OFFSET \
+    (CXL_DEVICE_REGISTERS_OFFSET + CXL_DEVICE_REGISTERS_LENGTH)
+#define CXL_MAILBOX_REGISTERS_SIZE 0x20 /* 8.2.8.4, Figure 139 */
+#define CXL_MAILBOX_PAYLOAD_SHIFT 11

I see 20 in the spec.

Quoted 6 lines
include/hw/cxl/cxl_device.h new file +6 -0
+#define CXL_MAILBOX_MAX_PAYLOAD_SIZE (1 << CXL_MAILBOX_PAYLOAD_SHIFT)
+#define CXL_MAILBOX_REGISTERS_LENGTH \
+    (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
+
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;

include/hw/cxl/cxl_device.h line 89

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Quoted 36 lines
include/hw/cxl/cxl_device.h new file +36 -0
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);
+
+/* Set up default values for the register block */
+void cxl_device_register_init_common(CXLDeviceState *dev);
+
+/*
+ * CXL 2.0 - 8.2.8.1 including errata F4
+ * Documented as a 128 bit register, but 64 bit accesses and the second
+ * 64 bits are currently reserved.
+ */
+REG64(CXL_DEV_CAP_ARRAY, 0) /* Documented as 128 bit register but 64 byte accesses */
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_ID, 0, 16)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_VERSION, 16, 8)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_COUNT, 32, 16)
+
+/*
+ * Helper macro to initialize capability headers for CXL devices.
+ *
+ * In the 8.2.8.2, this is listed as a 128b register, but in 8.2.8, it says:
+ * > No registers defined in Section 8.2.8 are larger than 64-bits wide so that
+ * > is the maximum access size allowed for these registers. If this rule is not
+ * > followed, the behavior is undefined
+ *
+ * CXL 2.0 Errata F4 states futher that the layouts in the specification are
+ * shown as greater than 128 bits, but implementations are expected to
+ * use any size of access up to 64 bits.
+ *
+ * Here we've chosen to make it 4 dwords. The spec allows any pow2 multiple
+ * access to be used for a register up to 64 bits.
+ */
+#define CXL_DEVICE_CAPABILITY_HEADER_REGISTER(n, offset)  \
+    REG32(CXL_DEV_##n##_CAP_HDR0, offset)                 \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_ID, 0, 16)      \

include/hw/cxl/cxl_device.h line 135

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_VERSION, 16, 8) \
+    REG32(CXL_DEV_##n##_CAP_HDR1, offset + 4)             \
+        FIELD(CXL_DEV_##n##_CAP_HDR1, CAP_OFFSET, 0, 32)  \
+    REG32(CXL_DEV_##n##_CAP_HDR2, offset + 8)             \
+        FIELD(CXL_DEV_##n##_CAP_HDR2, CAP_LENGTH, 0, 32)
+
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
+                                               CXL_DEVICE_CAP_REG_SIZE)
+

Fig139 for the following registers.

8.2.8.4.3

include/hw/cxl/cxl_device.h line 141

Quoted 6 lines
include/hw/cxl/cxl_device.h new file +6 -0
+REG32(CXL_DEV_MAILBOX_CAP, 0)
+    FIELD(CXL_DEV_MAILBOX_CAP, PAYLOAD_SIZE, 0, 5)
+    FIELD(CXL_DEV_MAILBOX_CAP, INT_CAP, 5, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, BG_INT_CAP, 6, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, MSI_N, 7, 4)
+

8.2.8.4.4

include/hw/cxl/cxl_device.h line 146

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG32(CXL_DEV_MAILBOX_CTRL, 4)
+    FIELD(CXL_DEV_MAILBOX_CTRL, DOORBELL, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
+

8.2.8.4.5 + 8.2.9

include/hw/cxl/cxl_device.h line 151

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG64(CXL_DEV_MAILBOX_CMD, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
+

8.2.8.4.6

include/hw/cxl/cxl_device.h line 156

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG64(CXL_DEV_MAILBOX_STS, 0x10)
+    FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_STS, ERRNO, 32, 16)
+    FIELD(CXL_DEV_MAILBOX_STS, VENDOR_ERRNO, 48, 16)
+

8.2.8.4.7

include/hw/cxl/cxl_device.h line 158

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +2 -0
+REG64(CXL_DEV_BG_CMD_STS, 0x18)
+    FIELD(CXL_DEV_BG_CMD_STS, BG, 0, 16)

Should we call this OP since it is implied that we are BG given the register?

include/hw/cxl/cxl_device.h line 159

Quoted 1 line
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, DONE, 16, 7)

NUM_DONE? since this is a percentage.

include/hw/cxl/cxl_device.h line 160

Quoted 1 line
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, ERRNO, 32, 16)

Isn't this a RET_CODE since it is only valid if previous field is 100%

include/hw/cxl/cxl_device.h line 161

Quoted 1 line
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, VENDOR_ERRNO, 48, 16)

VENDOR_RET_CODE since the same rule for the previous field applies here.

Quoted 8 lines
include/hw/cxl/cxl_device.h new file +4 -0
+
+REG32(CXL_DEV_CMD_PAYLOAD, 0x20)
+
+#endif
Signature · 1 line
2.32.0

+cc Dave, Klaus, Tong
Other than the minor issues raised.

Looks good.

Reviewed by: Adam Manzanares <[email protected]>

#11 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 7 lines · Adam Manzanares
Quoted 7 lines
include/hw/cxl/cxl_device.h new file +7 -0
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
Quoted 8 lines

include/hw/cxl/cxl_device.h line 89

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Agreed, but ideally both volatile and persistent capacities would have been supported in this patchset. I'm also probably missing specific reasons as to why this isn't the case.

Looking at it briefly could it be just a matter of adding to cxl_type3_dev a new hostmem along with it's AddressSpace for the volatile? If so, I'm thinking something along these lines:

@@ -123,8 +123,8 @@ typedef struct cxl_device_state {
	 uint64_t host_set;
     } timestamp;
-    /* memory region for persistent memory, HDM */
-    uint64_t pmem_size;
+    /* memory region for persistent and volatile memory, HDM */
+    uint64_t pmem_size, mem_size;

} CXLDeviceState;

/* Initialize the register block for a device */

@@ -235,9 +235,9 @@ typedef struct cxl_type3_dev {
     PCIDevice parent_obj;

     /* Properties */
-    AddressSpace hostmem_as;
+    AddressSpace hostmem_as, hostmemv_as;
     uint64_t size;
-    HostMemoryBackend *hostmem;
+    HostMemoryBackend *hostmem, *hostmemv;
     HostMemoryBackend *lsa;
     uint64_t sn;

Then for cxl_setup_memory(), with ct3d->hostmem and/or ct3d->hostmemv non-nil, set the respective MemoryRegions:

+ if (ct3d->hostmem) {

+            memory_region_set_nonvolatile(mr, true);
+            memory_region_set_enabled(mr, true);
+            host_memory_backend_set_mapped(ct3d->hostmem, true);
+            address_space_init(&ct3d->hostmem_as, mr, name);
+            ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
+    }
+    if (ct3d->hostmemv) {
+            memory_region_set_nonvolatile(mrv, false);
+            memory_region_set_enabled(mrv, true);
+            host_memory_backend_set_mapped(ct3d->hostmemv, true);
+            address_space_init(&ct3d->hostmem_as, mrv, name);
+            ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;

+ }

For corresponding MB commands, it's mostly IDENTIFY_MEMORY_DEVICE that needs updating:

@@ -281,7 +281,7 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,

     CXLType3Dev *ct3d = container_of(cxl_dstate, CXLType3Dev, cxl_dstate);
     CXLType3Class *cvc = CXL_TYPE3_DEV_GET_CLASS(ct3d);

- uint64_t size = cxl_dstate->pmem_size;
+ uint64_t size = cxl_dstate->pmem_size + cxl_dstate->mem_size;

     if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
	 return CXL_MBOX_INTERNAL_ERROR;
@@ -290,11 +290,11 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
     id = (void *)cmd->payload;
     memset(id, 0, sizeof(*id));

- /* PMEM only */

     snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);

     id->total_capacity = size / (256 << 20);
-    id->persistent_capacity = size / (256 << 20);
+    id->persistent_capacity = cxl_dstate->pmem_size / (256 << 20);
+    id->volatile_capacity = cxl_dstate->mem_size / (256 << 20);
     id->lsa_size = cvc->get_lsa_size(ct3d);

     *len = sizeof(*id);
@@ -312,16 +312,16 @@ static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
	 uint64_t next_pmem;
     } QEMU_PACKED *part_info = (void *)cmd->payload;
     QEMU_BUILD_BUG_ON(sizeof(*part_info) != 0x20);
-    uint64_t size = cxl_dstate->pmem_size;
+    uint64_t psize = cxl_dstate->pmem_size;
+    uint64_t vsize = cxl_dstate->mem_size;

- if (!QEMU_IS_ALIGNED(size, 256 << 20)) {

+    if (!QEMU_IS_ALIGNED(psize + vsize, 256 << 20)) {
	 return CXL_MBOX_INTERNAL_ERROR;
     }
-    /* PMEM only */
-    part_info->active_vmem = 0;
-    part_info->next_vmem = 0;
-    part_info->active_pmem = size / (256 << 20);
+    part_info->active_vmem = vsize / (256 << 20);
+    part_info->next_vmem = part_info->active_vmem;
+    part_info->active_pmem = psize / (256 << 20);
     part_info->next_pmem = part_info->active_pmem;

Then for reads/writes, both cxl_type3_write() and _read() would, after computing the dpa_offset, first try the volatile region then upon error attempt the same in the persistent memory - this assuming the DPA space is consistent among both types of memory. Ie:

address_space_read(&ct3d->hostmemv_as, dpa_offset, attrs, data, size) or
address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size)

... but then again all this is probably just wishful thinking.

Thanks,
Davidlohr

#12 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 17 lines · On Tue, 29 Mar 2022 12:53:51 -0700 Davidlohr Bueso <[email protected]> wrote:
Quoted 7 lines · Adam Manzanares
Quoted 7 lines
include/hw/cxl/cxl_device.h new file +7 -0
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
Quoted 8 lines

include/hw/cxl/cxl_device.h line 89

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;  

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Quoted 3 lines

Agreed, but ideally both volatile and persistent capacities would have been supported in this patchset. I'm also probably missing specific reasons as to why this isn't the case.

Whilst it doesn't add a huge amount of complexity it does add some and the software paths in Linux we were developing this for are pmem focused. Hence volatile is on the todo list rather than in this first patch set. Not sensible to aim for feature complete in one go.

Quoted 98 lines

Looking at it briefly could it be just a matter of adding to cxl_type3_dev a new hostmem along with it's AddressSpace for the volatile? If so, I'm thinking something along these lines:

(quoted diff) +4 -4
@@ -123,8 +123,8 @@ typedef struct cxl_device_state {
 	 uint64_t host_set;
      } timestamp;
 
-    /* memory region for persistent memory, HDM */
-    uint64_t pmem_size;
+    /* memory region for persistent and volatile memory, HDM */
+    uint64_t pmem_size, mem_size;
  } CXLDeviceState;
 
  /* Initialize the register block for a device */
@@ -235,9 +235,9 @@ typedef struct cxl_type3_dev {
      PCIDevice parent_obj;
 
      /* Properties */
-    AddressSpace hostmem_as;
+    AddressSpace hostmem_as, hostmemv_as;
      uint64_t size;
-    HostMemoryBackend *hostmem;
+    HostMemoryBackend *hostmem, *hostmemv;
      HostMemoryBackend *lsa;
      uint64_t sn;
 

Then for cxl_setup_memory(), with ct3d->hostmem and/or ct3d->hostmemv non-nil, set the respective MemoryRegions:

(quoted diff) +14 -0
+    if (ct3d->hostmem) {
+            memory_region_set_nonvolatile(mr, true);
+            memory_region_set_enabled(mr, true);
+            host_memory_backend_set_mapped(ct3d->hostmem, true);
+            address_space_init(&ct3d->hostmem_as, mr, name);
+            ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
+    }
+    if (ct3d->hostmemv) {
+            memory_region_set_nonvolatile(mrv, false);
+            memory_region_set_enabled(mrv, true);
+            host_memory_backend_set_mapped(ct3d->hostmemv, true);
+            address_space_init(&ct3d->hostmem_as, mrv, name);
+            ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
+    }

For corresponding MB commands, it's mostly IDENTIFY_MEMORY_DEVICE that needs updating:

(quoted diff) +9 -9
@@ -281,7 +281,7 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
 
      CXLType3Dev *ct3d = container_of(cxl_dstate, CXLType3Dev, cxl_dstate);
      CXLType3Class *cvc = CXL_TYPE3_DEV_GET_CLASS(ct3d);
-    uint64_t size = cxl_dstate->pmem_size;
+    uint64_t size = cxl_dstate->pmem_size + cxl_dstate->mem_size;
 
      if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
 	 return CXL_MBOX_INTERNAL_ERROR;
@@ -290,11 +290,11 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
      id = (void *)cmd->payload;
      memset(id, 0, sizeof(*id));
 
-    /* PMEM only */
      snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
 
      id->total_capacity = size / (256 << 20);
-    id->persistent_capacity = size / (256 << 20);
+    id->persistent_capacity = cxl_dstate->pmem_size / (256 << 20);
+    id->volatile_capacity = cxl_dstate->mem_size / (256 << 20);
      id->lsa_size = cvc->get_lsa_size(ct3d);
 
      *len = sizeof(*id);
@@ -312,16 +312,16 @@ static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
 	 uint64_t next_pmem;
      } QEMU_PACKED *part_info = (void *)cmd->payload;
      QEMU_BUILD_BUG_ON(sizeof(*part_info) != 0x20);
-    uint64_t size = cxl_dstate->pmem_size;
+    uint64_t psize = cxl_dstate->pmem_size;
+    uint64_t vsize = cxl_dstate->mem_size;
 
-    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+    if (!QEMU_IS_ALIGNED(psize + vsize, 256 << 20)) {
 	 return CXL_MBOX_INTERNAL_ERROR;
      }
 
-    /* PMEM only */
-    part_info->active_vmem = 0;
-    part_info->next_vmem = 0;
-    part_info->active_pmem = size / (256 << 20);
+    part_info->active_vmem = vsize / (256 << 20);
+    part_info->next_vmem = part_info->active_vmem;
+    part_info->active_pmem = psize / (256 << 20);
      part_info->next_pmem = part_info->active_pmem;
 
 Then for reads/writes, both cxl_type3_write() and _read() would, after computing the dpa_offset,

first try the volatile region then upon error attempt the same in the persistent memory - this assuming the DPA space is consistent among both types of memory. Ie:

Quoted 5 lines

address_space_read(&ct3d->hostmemv_as, dpa_offset, attrs, data, size) or
address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size)

... but then again all this is probably just wishful thinking.

Without looking in detail, will indeed be something along those lines. Gets more fiddly if you include partitioning control that Alison was interested in adding.

Also, we probably need to support multiple HDM decoders. Also not a huge complexity to add, but left for now as main focus is to get the base patch set merged.

So I'm happy to queue stuff up on top of this series and carry it forward but I don't want to add features to what we try to merge initially. This set is already huge and hard to review even with what think is a minimum set of features to be useful.

Note I'm already carrying a few features on top if this on the gitlab branch gitlab.com/jic23/qemu (DOE + CDAT and serial numbers) and have a few other things out of tree for now (SPDM, emulating most of the PCI Config space controls).

Jonathan

Quoted 3 lines

Thanks,
Davidlohr

#13 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 20 lines · Jonathan Cameron
Quoted 17 lines · On Tue, 29 Mar 2022 12:53:51 -0700 Davidlohr Bueso <[email protected]> wrote:
Quoted 15 lines · Adam Manzanares

include/hw/cxl/cxl_device.h line 89

Quoted 12 lines
include/hw/cxl/cxl_device.h new file +12 -0
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;  

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Quoted 8 lines
Quoted 3 lines

Agreed, but ideally both volatile and persistent capacities would have been supported in this patchset. I'm also probably missing specific reasons as to why this isn't the case.

Whilst it doesn't add a huge amount of complexity it does add some and the software paths in Linux we were developing this for are pmem focused. Hence volatile is on the todo list rather than in this first patch set. Not sensible to aim for feature complete in one go.

Makes complete sense. We can help with the Linux development for the volatile side. I will add a couple of folks on cc. In addition, we would like to help the CXL ecosystem in general so I anticipate we will have more reviews and patches for CXL in general.

Quoted 118 lines
Quoted 99 lines

Looking at it briefly could it be just a matter of adding to cxl_type3_dev a new hostmem along with it's AddressSpace for the volatile? If so, I'm thinking something along these lines:

(quoted diff) +4 -4
@@ -123,8 +123,8 @@ typedef struct cxl_device_state {
 	 uint64_t host_set;
      } timestamp;
 
-    /* memory region for persistent memory, HDM */
-    uint64_t pmem_size;
+    /* memory region for persistent and volatile memory, HDM */
+    uint64_t pmem_size, mem_size;
  } CXLDeviceState;
 
  /* Initialize the register block for a device */
@@ -235,9 +235,9 @@ typedef struct cxl_type3_dev {
      PCIDevice parent_obj;
 
      /* Properties */
-    AddressSpace hostmem_as;
+    AddressSpace hostmem_as, hostmemv_as;
      uint64_t size;
-    HostMemoryBackend *hostmem;
+    HostMemoryBackend *hostmem, *hostmemv;
      HostMemoryBackend *lsa;
      uint64_t sn;
 

Then for cxl_setup_memory(), with ct3d->hostmem and/or ct3d->hostmemv non-nil, set the respective MemoryRegions:

(quoted diff) +14 -0
+    if (ct3d->hostmem) {
+            memory_region_set_nonvolatile(mr, true);
+            memory_region_set_enabled(mr, true);
+            host_memory_backend_set_mapped(ct3d->hostmem, true);
+            address_space_init(&ct3d->hostmem_as, mr, name);
+            ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
+    }
+    if (ct3d->hostmemv) {
+            memory_region_set_nonvolatile(mrv, false);
+            memory_region_set_enabled(mrv, true);
+            host_memory_backend_set_mapped(ct3d->hostmemv, true);
+            address_space_init(&ct3d->hostmem_as, mrv, name);
+            ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
+    }

For corresponding MB commands, it's mostly IDENTIFY_MEMORY_DEVICE that needs updating:

(quoted diff) +9 -9
@@ -281,7 +281,7 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
 
      CXLType3Dev *ct3d = container_of(cxl_dstate, CXLType3Dev, cxl_dstate);
      CXLType3Class *cvc = CXL_TYPE3_DEV_GET_CLASS(ct3d);
-    uint64_t size = cxl_dstate->pmem_size;
+    uint64_t size = cxl_dstate->pmem_size + cxl_dstate->mem_size;
 
      if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
 	 return CXL_MBOX_INTERNAL_ERROR;
@@ -290,11 +290,11 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
      id = (void *)cmd->payload;
      memset(id, 0, sizeof(*id));
 
-    /* PMEM only */
      snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
 
      id->total_capacity = size / (256 << 20);
-    id->persistent_capacity = size / (256 << 20);
+    id->persistent_capacity = cxl_dstate->pmem_size / (256 << 20);
+    id->volatile_capacity = cxl_dstate->mem_size / (256 << 20);
      id->lsa_size = cvc->get_lsa_size(ct3d);
 
      *len = sizeof(*id);
@@ -312,16 +312,16 @@ static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
 	 uint64_t next_pmem;
      } QEMU_PACKED *part_info = (void *)cmd->payload;
      QEMU_BUILD_BUG_ON(sizeof(*part_info) != 0x20);
-    uint64_t size = cxl_dstate->pmem_size;
+    uint64_t psize = cxl_dstate->pmem_size;
+    uint64_t vsize = cxl_dstate->mem_size;
 
-    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+    if (!QEMU_IS_ALIGNED(psize + vsize, 256 << 20)) {
 	 return CXL_MBOX_INTERNAL_ERROR;
      }
 
-    /* PMEM only */
-    part_info->active_vmem = 0;
-    part_info->next_vmem = 0;
-    part_info->active_pmem = size / (256 << 20);
+    part_info->active_vmem = vsize / (256 << 20);
+    part_info->next_vmem = part_info->active_vmem;
+    part_info->active_pmem = psize / (256 << 20);
      part_info->next_pmem = part_info->active_pmem;
 
 Then for reads/writes, both cxl_type3_write() and _read() would, after computing the dpa_offset,

first try the volatile region then upon error attempt the same in the persistent memory - this assuming the DPA space is consistent among both types of memory. Ie:

Quoted 5 lines

address_space_read(&ct3d->hostmemv_as, dpa_offset, attrs, data, size) or
address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size)

... but then again all this is probably just wishful thinking.

Without looking in detail, will indeed be something along those lines. Gets more fiddly if you include partitioning control that Alison was interested in adding.

Also, we probably need to support multiple HDM decoders. Also not a huge complexity to add, but left for now as main focus is to get the base patch set merged.

So I'm happy to queue stuff up on top of this series and carry it forward but I don't want to add features to what we try to merge initially. This set is already huge and hard to review even with what think is a minimum set of features to be useful.

Quoted 5 lines

Note I'm already carrying a few features on top if this on the gitlab branch gitlab.com/jic23/qemu (DOE + CDAT and serial numbers) and have a few other things out of tree for now (SPDM, emulating most of the PCI Config space controls).

Thanks for the updates. Do you have any suggestions on how to coordinate efforts? Ideally we can have a list of features that need to be developed and some names of people that will lead the work.

Quoted 6 lines

Jonathan

Quoted 3 lines

Thanks,
Davidlohr

#14 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

In reply to Adam Manzanares
Quoted 15 lines · On Tue, 29 Mar 2022 18:13:59 +0000 Adam Manzanares <[email protected]>…
Quoted 14 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL device is a type of CXL component. Conceptually, a CXL device would be a leaf node in a CXL topology. From an emulation perspective, CXL devices are the most complex and so the actual implementation is reserved for discrete commits.

This new device type is specifically catered towards the eventual implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0 specification.

...

Quoted 22 lines
Quoted 22 lines
include/hw/cxl/cxl_device.h new file +16 -0
new file mode 100644
index 0000000000..b2416e45bf
@@ -0,0 +1,165 @@
+/*
+ * QEMU CXL Devices
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_DEVICE_H
+#define CXL_DEVICE_H
+
+#include "hw/register.h"
+
+/*
+ * The following is how a CXL device's MMIO space is laid out. The only
Quoted 8 lines

include/hw/cxl/cxl_device.h line 21

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+ * requirement from the spec is that the capabilities array and the capability
+ * headers start at offset 0 and are contiguously packed. The headers themselves
+ * provide offsets to the register fields. For this emulation, registers will
+ * start at offset 0x80 (m == 0x80). No secondary mailbox is implemented which
+ * means that n = m + sizeof(mailbox registers) + sizeof(device registers).  

What is n here, the start offset of the mailbox registers, this question is based on the figure below?

I'll expand on this to say

means that the offset of the start of the mailbox payload (n) is given by n = m + sizeof....

Which means the diagram below is wrong as should align with top of mailbox registers.

Quoted 3 lines

include/hw/cxl/cxl_device.h line 23

Quoted 3 lines
include/hw/cxl/cxl_device.h new file +2 -0
+ *
+ * This is roughly described in 8.2.8 Figure 138 of the CXL 2.0 spec

I'm going drop this comment as that figure appears unrelated to me.

Quoted 37 lines
Quoted 26 lines
include/hw/cxl/cxl_device.h new file +26 -0
+ *
+ *                       +---------------------------------+
+ *                       |                                 |
+ *                       |    Memory Device Registers      |
+ *                       |                                 |
+ * n + PAYLOAD_SIZE_MAX  -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |         Mailbox Payload         |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    -----------------------------------
+ *                  |    |       Mailbox Registers         |
+ *                  |    |                                 |
+ *                  n    -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |        Device Registers         |
+ *                  |    |                                 |
+ *                  m    ---------------------------------->
+ *                  ^    |  Memory Device Capability Header|
+ *                  |    -----------------------------------
+ *                  |    |     Mailbox Capability Header   |

include/hw/cxl/cxl_device.h line 59

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+ *                  |    -------------- --------------------
+ *                  |    |     Device Capability Header    |
+ *                  |    -----------------------------------
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |      Device Cap Array[0..n]     |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                       |                                 |
+ *                  0    +---------------------------------+  
Quoted 1 line

Would it make sense to add CXL cap header register to the diagram?

Too many similar names in the CXL spec. I'm not sure which one you mean, could you let me have a reference? If you mean the one that is at the start of the CXL.cache and CXL.mem registers that whole region isn't covered by this diagram and might be in a different BAR. Here we are only dealing with the Memory Device Registers. I'll add statement to the initial comment block to make that clear as it definitely isn't currently!

Quoted 3 lines

n also
seems to be the size of the cap array, but it is also an offset so that could be clarified.

Ah. Letter reuse. good point. Looking more closely it isn't an array anyway in the diagram (the array would have to include the Device Capability Header and Mailbox Capability headers. Renamed as simply Device Cap Array Register

Quoted 17 lines

include/hw/cxl/cxl_device.h line 67

Quoted 9 lines
include/hw/cxl/cxl_device.h new file +8 -0
+ *
+ */
+
+#define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
+#define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
+#define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
+
+#define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */  

Is this to plan for future capabilities? If we have CAPS MAX doesn't this allow us to remove the slack space.

include/hw/cxl/cxl_device.h line 68

Quoted 1 line
include/hw/cxl/cxl_device.h new file +1 -0
+#define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */  

Should we add status to the name here, or would it get too long?

Quoted 7 lines

include/hw/cxl/cxl_device.h line 73

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+
+#define CXL_MAILBOX_REGISTERS_OFFSET \
+    (CXL_DEVICE_REGISTERS_OFFSET + CXL_DEVICE_REGISTERS_LENGTH)
+#define CXL_MAILBOX_REGISTERS_SIZE 0x20 /* 8.2.8.4, Figure 139 */
+#define CXL_MAILBOX_PAYLOAD_SHIFT 11  

I see 20 in the spec.

It's an implementation choice between 8 and 20. For now, this code goes with 11 for no particularly strong reason.

Quoted 12 lines
Quoted 12 lines
include/hw/cxl/cxl_device.h new file +11 -0
+#define CXL_MAILBOX_MAX_PAYLOAD_SIZE (1 << CXL_MAILBOX_PAYLOAD_SHIFT)
+#define CXL_MAILBOX_REGISTERS_LENGTH \
+    (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
+
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
Quoted 8 lines

include/hw/cxl/cxl_device.h line 89

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;  

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Discussed in the other branch of this thread. Short answer is we don't support non persistent yet but it's on the todo list. What exactly that looks like is to be determined. One aspect of that is there isn't currently a software stack to test volatile memory.

Quoted 41 lines
Quoted 41 lines
include/hw/cxl/cxl_device.h new file +40 -0
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);
+
+/* Set up default values for the register block */
+void cxl_device_register_init_common(CXLDeviceState *dev);
+
+/*
+ * CXL 2.0 - 8.2.8.1 including errata F4
+ * Documented as a 128 bit register, but 64 bit accesses and the second
+ * 64 bits are currently reserved.
+ */
+REG64(CXL_DEV_CAP_ARRAY, 0) /* Documented as 128 bit register but 64 byte accesses */
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_ID, 0, 16)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_VERSION, 16, 8)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_COUNT, 32, 16)
+
+/*
+ * Helper macro to initialize capability headers for CXL devices.
+ *
+ * In the 8.2.8.2, this is listed as a 128b register, but in 8.2.8, it says:
+ * > No registers defined in Section 8.2.8 are larger than 64-bits wide so that
+ * > is the maximum access size allowed for these registers. If this rule is not
+ * > followed, the behavior is undefined
+ *
+ * CXL 2.0 Errata F4 states futher that the layouts in the specification are
+ * shown as greater than 128 bits, but implementations are expected to
+ * use any size of access up to 64 bits.
+ *
+ * Here we've chosen to make it 4 dwords. The spec allows any pow2 multiple
+ * access to be used for a register up to 64 bits.
+ */
+#define CXL_DEVICE_CAPABILITY_HEADER_REGISTER(n, offset)  \
+    REG32(CXL_DEV_##n##_CAP_HDR0, offset)                 \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_ID, 0, 16)      \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_VERSION, 16, 8) \
+    REG32(CXL_DEV_##n##_CAP_HDR1, offset + 4)             \
+        FIELD(CXL_DEV_##n##_CAP_HDR1, CAP_OFFSET, 0, 32)  \
+    REG32(CXL_DEV_##n##_CAP_HDR2, offset + 8)             \
Quoted 8 lines

include/hw/cxl/cxl_device.h line 135

Quoted 6 lines
include/hw/cxl/cxl_device.h new file +6 -0
+        FIELD(CXL_DEV_##n##_CAP_HDR2, CAP_LENGTH, 0, 32)
+
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
+                                               CXL_DEVICE_CAP_REG_SIZE)
+  

Fig139 for the following registers.

Added ref

Quoted 2 lines

8.2.8.4.3

Good idea. Added all these references.

Quoted 28 lines

include/hw/cxl/cxl_device.h line 141

Quoted 6 lines
include/hw/cxl/cxl_device.h new file +6 -0
+REG32(CXL_DEV_MAILBOX_CAP, 0)
+    FIELD(CXL_DEV_MAILBOX_CAP, PAYLOAD_SIZE, 0, 5)
+    FIELD(CXL_DEV_MAILBOX_CAP, INT_CAP, 5, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, BG_INT_CAP, 6, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, MSI_N, 7, 4)
+  

8.2.8.4.4

include/hw/cxl/cxl_device.h line 146

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG32(CXL_DEV_MAILBOX_CTRL, 4)
+    FIELD(CXL_DEV_MAILBOX_CTRL, DOORBELL, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
+  

8.2.8.4.5 + 8.2.9

include/hw/cxl/cxl_device.h line 151

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG64(CXL_DEV_MAILBOX_CMD, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
+  

8.2.8.4.6

include/hw/cxl/cxl_device.h line 156

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG64(CXL_DEV_MAILBOX_STS, 0x10)
+    FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_STS, ERRNO, 32, 16)
+    FIELD(CXL_DEV_MAILBOX_STS, VENDOR_ERRNO, 48, 16)
+  
Quoted 5 lines

8.2.8.4.7

include/hw/cxl/cxl_device.h line 158

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +2 -0
+REG64(CXL_DEV_BG_CMD_STS, 0x18)
+    FIELD(CXL_DEV_BG_CMD_STS, BG, 0, 16)  

Should we call this OP since it is implied that we are BG given the register?

Sure. It's a better name than BG.

Quoted 4 lines

include/hw/cxl/cxl_device.h line 159

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, DONE, 16, 7)  

NUM_DONE? since this is a percentage.

Let's be verbose as NUM_DONE still seems confusing to me. PERCENTAGE_COMP

I hadn't really noticed these names as I don't think any of them are used yet.

Quoted 4 lines

include/hw/cxl/cxl_device.h line 160

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, ERRNO, 32, 16)  

Isn't this a RET_CODE since it is only valid if previous field is 100%

Changed

Quoted 4 lines

include/hw/cxl/cxl_device.h line 161

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, VENDOR_ERRNO, 48, 16)  

VENDOR_RET_CODE since the same rule for the previous field applies here.

Changed

Quoted 13 lines
Quoted 9 lines
include/hw/cxl/cxl_device.h new file +4 -0
+
+REG32(CXL_DEV_CMD_PAYLOAD, 0x20)
+
+#endif
Signature · 1 line
2.32.0

+cc Dave, Klaus, Tong
Other than the minor issues raised.

Quoted 3 lines

Looks good.

Reviewed by: Adam Manzanares <[email protected]>

Btw I haven't accepted all changes, but have been picking up your RB. Shout if that's not fine with you.

Thanks.

Jonathan

#15 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 104 lines · Jonathan Cameron
Quoted 15 lines · On Tue, 29 Mar 2022 18:13:59 +0000 Adam Manzanares <[email protected]>…
Quoted 14 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL device is a type of CXL component. Conceptually, a CXL device would be a leaf node in a CXL topology. From an emulation perspective, CXL devices are the most complex and so the actual implementation is reserved for discrete commits.

This new device type is specifically catered towards the eventual implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0 specification.

...

Quoted 22 lines
Quoted 22 lines
include/hw/cxl/cxl_device.h new file +16 -0
new file mode 100644
index 0000000000..b2416e45bf
@@ -0,0 +1,165 @@
+/*
+ * QEMU CXL Devices
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_DEVICE_H
+#define CXL_DEVICE_H
+
+#include "hw/register.h"
+
+/*
+ * The following is how a CXL device's MMIO space is laid out. The only
Quoted 8 lines

include/hw/cxl/cxl_device.h line 21

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+ * requirement from the spec is that the capabilities array and the capability
+ * headers start at offset 0 and are contiguously packed. The headers themselves
+ * provide offsets to the register fields. For this emulation, registers will
+ * start at offset 0x80 (m == 0x80). No secondary mailbox is implemented which
+ * means that n = m + sizeof(mailbox registers) + sizeof(device registers).  

What is n here, the start offset of the mailbox registers, this question is based on the figure below?

I'll expand on this to say

means that the offset of the start of the mailbox payload (n) is given by n = m + sizeof....

Which means the diagram below is wrong as should align with top of mailbox registers.

Quoted 3 lines

include/hw/cxl/cxl_device.h line 23

Quoted 3 lines
include/hw/cxl/cxl_device.h new file +2 -0
+ *
+ * This is roughly described in 8.2.8 Figure 138 of the CXL 2.0 spec

I'm going drop this comment as that figure appears unrelated to me.

Quoted 37 lines
Quoted 26 lines
include/hw/cxl/cxl_device.h new file +26 -0
+ *
+ *                       +---------------------------------+
+ *                       |                                 |
+ *                       |    Memory Device Registers      |
+ *                       |                                 |
+ * n + PAYLOAD_SIZE_MAX  -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |         Mailbox Payload         |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    -----------------------------------
+ *                  |    |       Mailbox Registers         |
+ *                  |    |                                 |
+ *                  n    -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |        Device Registers         |
+ *                  |    |                                 |
+ *                  m    ---------------------------------->
+ *                  ^    |  Memory Device Capability Header|
+ *                  |    -----------------------------------
+ *                  |    |     Mailbox Capability Header   |

include/hw/cxl/cxl_device.h line 59

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+ *                  |    -------------- --------------------
+ *                  |    |     Device Capability Header    |
+ *                  |    -----------------------------------
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |      Device Cap Array[0..n]     |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                       |                                 |
+ *                  0    +---------------------------------+  
Quoted 1 line

Would it make sense to add CXL cap header register to the diagram?

Quoted 7 lines

Too many similar names in the CXL spec. I'm not sure which one you mean, could you let me have a reference? If you mean the one that is at the start of the CXL.cache and CXL.mem registers that whole region isn't covered by this diagram and might be in a different BAR. Here we are only dealing with the Memory Device Registers. I'll add statement to the initial comment block to make that clear as it definitely isn't currently!

I was thinking 0 in your figure is the device capabilities array register, which tells us how many capabilites that are in the array. This would be 8.2.8.1. After that comes 8.2.8.2 with n capability header registers which point to the device registers.

Quoted 34 lines
Quoted 4 lines

n also
seems to be the size of the cap array, but it is also an offset so that could be clarified.

Ah. Letter reuse. good point. Looking more closely it isn't an array anyway in the diagram (the array would have to include the Device Capability Header and Mailbox Capability headers. Renamed as simply Device Cap Array Register

Quoted 17 lines

include/hw/cxl/cxl_device.h line 67

Quoted 9 lines
include/hw/cxl/cxl_device.h new file +8 -0
+ *
+ */
+
+#define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
+#define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
+#define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
+
+#define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */  

Is this to plan for future capabilities? If we have CAPS MAX doesn't this allow us to remove the slack space.

include/hw/cxl/cxl_device.h line 68

Quoted 1 line
include/hw/cxl/cxl_device.h new file +1 -0
+#define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */  

Should we add status to the name here, or would it get too long?

Quoted 7 lines

include/hw/cxl/cxl_device.h line 73

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+
+#define CXL_MAILBOX_REGISTERS_OFFSET \
+    (CXL_DEVICE_REGISTERS_OFFSET + CXL_DEVICE_REGISTERS_LENGTH)
+#define CXL_MAILBOX_REGISTERS_SIZE 0x20 /* 8.2.8.4, Figure 139 */
+#define CXL_MAILBOX_PAYLOAD_SHIFT 11  

I see 20 in the spec.

Quoted 2 lines

It's an implementation choice between 8 and 20. For now, this code goes with 11 for no particularly strong reason.

Got it.

Quoted 22 lines
Quoted 13 lines
Quoted 13 lines
include/hw/cxl/cxl_device.h new file +11 -0
+#define CXL_MAILBOX_MAX_PAYLOAD_SIZE (1 << CXL_MAILBOX_PAYLOAD_SHIFT)
+#define CXL_MAILBOX_REGISTERS_LENGTH \
+    (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
+
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
Quoted 8 lines

include/hw/cxl/cxl_device.h line 89

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;  

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Quoted 4 lines

Discussed in the other branch of this thread. Short answer is we don't support non persistent yet but it's on the todo list. What exactly that looks like is to be determined. One aspect of that is there isn't currently a software stack to test volatile memory.

If you can elaborate more here on what is needed to test the volatile memory stack we may be able to help out.

Quoted 88 lines
Quoted 42 lines
Quoted 42 lines
include/hw/cxl/cxl_device.h new file +40 -0
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);
+
+/* Set up default values for the register block */
+void cxl_device_register_init_common(CXLDeviceState *dev);
+
+/*
+ * CXL 2.0 - 8.2.8.1 including errata F4
+ * Documented as a 128 bit register, but 64 bit accesses and the second
+ * 64 bits are currently reserved.
+ */
+REG64(CXL_DEV_CAP_ARRAY, 0) /* Documented as 128 bit register but 64 byte accesses */
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_ID, 0, 16)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_VERSION, 16, 8)
+    FIELD(CXL_DEV_CAP_ARRAY, CAP_COUNT, 32, 16)
+
+/*
+ * Helper macro to initialize capability headers for CXL devices.
+ *
+ * In the 8.2.8.2, this is listed as a 128b register, but in 8.2.8, it says:
+ * > No registers defined in Section 8.2.8 are larger than 64-bits wide so that
+ * > is the maximum access size allowed for these registers. If this rule is not
+ * > followed, the behavior is undefined
+ *
+ * CXL 2.0 Errata F4 states futher that the layouts in the specification are
+ * shown as greater than 128 bits, but implementations are expected to
+ * use any size of access up to 64 bits.
+ *
+ * Here we've chosen to make it 4 dwords. The spec allows any pow2 multiple
+ * access to be used for a register up to 64 bits.
+ */
+#define CXL_DEVICE_CAPABILITY_HEADER_REGISTER(n, offset)  \
+    REG32(CXL_DEV_##n##_CAP_HDR0, offset)                 \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_ID, 0, 16)      \
+        FIELD(CXL_DEV_##n##_CAP_HDR0, CAP_VERSION, 16, 8) \
+    REG32(CXL_DEV_##n##_CAP_HDR1, offset + 4)             \
+        FIELD(CXL_DEV_##n##_CAP_HDR1, CAP_OFFSET, 0, 32)  \
+    REG32(CXL_DEV_##n##_CAP_HDR2, offset + 8)             \
Quoted 8 lines

include/hw/cxl/cxl_device.h line 135

Quoted 6 lines
include/hw/cxl/cxl_device.h new file +6 -0
+        FIELD(CXL_DEV_##n##_CAP_HDR2, CAP_LENGTH, 0, 32)
+
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
+                                               CXL_DEVICE_CAP_REG_SIZE)
+  

Fig139 for the following registers.

Added ref

Quoted 2 lines

8.2.8.4.3

Good idea. Added all these references.

Quoted 32 lines

include/hw/cxl/cxl_device.h line 141

Quoted 6 lines
include/hw/cxl/cxl_device.h new file +6 -0
+REG32(CXL_DEV_MAILBOX_CAP, 0)
+    FIELD(CXL_DEV_MAILBOX_CAP, PAYLOAD_SIZE, 0, 5)
+    FIELD(CXL_DEV_MAILBOX_CAP, INT_CAP, 5, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, BG_INT_CAP, 6, 1)
+    FIELD(CXL_DEV_MAILBOX_CAP, MSI_N, 7, 4)
+  

8.2.8.4.4

include/hw/cxl/cxl_device.h line 146

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG32(CXL_DEV_MAILBOX_CTRL, 4)
+    FIELD(CXL_DEV_MAILBOX_CTRL, DOORBELL, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, INT_EN, 1, 1)
+    FIELD(CXL_DEV_MAILBOX_CTRL, BG_INT_EN, 2, 1)
+  

8.2.8.4.5 + 8.2.9

include/hw/cxl/cxl_device.h line 151

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG64(CXL_DEV_MAILBOX_CMD, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND, 0, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, COMMAND_SET, 8, 8)
+    FIELD(CXL_DEV_MAILBOX_CMD, LENGTH, 16, 20)
+  

8.2.8.4.6

include/hw/cxl/cxl_device.h line 156

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+REG64(CXL_DEV_MAILBOX_STS, 0x10)
+    FIELD(CXL_DEV_MAILBOX_STS, BG_OP, 0, 1)
+    FIELD(CXL_DEV_MAILBOX_STS, ERRNO, 32, 16)
+    FIELD(CXL_DEV_MAILBOX_STS, VENDOR_ERRNO, 48, 16)
+  

8.2.8.4.7

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +2 -0
+REG64(CXL_DEV_BG_CMD_STS, 0x18)
+    FIELD(CXL_DEV_BG_CMD_STS, BG, 0, 16)  
Quoted 8 lines
Quoted 1 line

Should we call this OP since it is implied that we are BG given the register?

Sure. It's a better name than BG.

Quoted 4 lines

include/hw/cxl/cxl_device.h line 159

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, DONE, 16, 7)  

NUM_DONE? since this is a percentage.

Let's be verbose as NUM_DONE still seems confusing to me. PERCENTAGE_COMP

Even better.

Quoted 29 lines

I hadn't really noticed these names as I don't think any of them are used yet.

Quoted 4 lines

include/hw/cxl/cxl_device.h line 160

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, ERRNO, 32, 16)  

Isn't this a RET_CODE since it is only valid if previous field is 100%

Changed

Quoted 4 lines

include/hw/cxl/cxl_device.h line 161

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+    FIELD(CXL_DEV_BG_CMD_STS, VENDOR_ERRNO, 48, 16)  

VENDOR_RET_CODE since the same rule for the previous field applies here.

Changed

Quoted 13 lines
Quoted 9 lines
include/hw/cxl/cxl_device.h new file +4 -0
+
+REG32(CXL_DEV_CMD_PAYLOAD, 0x20)
+
+#endif
Signature · 1 line
2.32.0

+cc Dave, Klaus, Tong
Other than the minor issues raised.

Quoted 6 lines
Quoted 3 lines

Looks good.

Reviewed by: Adam Manzanares <[email protected]>

Btw I haven't accepted all changes, but have been picking up your RB. Shout if that's not fine with you.

Definitely fine with me and was my intention. Let us know how we can help move the work forward. I am kick starting reviewing and will try to bring others in.

Quoted 5 lines

Thanks.

Jonathan

#16 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 114 lines · On Thu, 31 Mar 2022 22:13:20 +0000 Adam Manzanares <[email protected]>…
Quoted 104 lines · Jonathan Cameron
Quoted 15 lines · On Tue, 29 Mar 2022 18:13:59 +0000 Adam Manzanares <[email protected]>…
Quoted 14 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL device is a type of CXL component. Conceptually, a CXL device would be a leaf node in a CXL topology. From an emulation perspective, CXL devices are the most complex and so the actual implementation is reserved for discrete commits.

This new device type is specifically catered towards the eventual implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0 specification.

...

Quoted 22 lines
Quoted 22 lines
include/hw/cxl/cxl_device.h new file +16 -0
new file mode 100644
index 0000000000..b2416e45bf
@@ -0,0 +1,165 @@
+/*
+ * QEMU CXL Devices
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_DEVICE_H
+#define CXL_DEVICE_H
+
+#include "hw/register.h"
+
+/*
+ * The following is how a CXL device's MMIO space is laid out. The only
Quoted 8 lines

include/hw/cxl/cxl_device.h line 21

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+ * requirement from the spec is that the capabilities array and the capability
+ * headers start at offset 0 and are contiguously packed. The headers themselves
+ * provide offsets to the register fields. For this emulation, registers will
+ * start at offset 0x80 (m == 0x80). No secondary mailbox is implemented which
+ * means that n = m + sizeof(mailbox registers) + sizeof(device registers).    

What is n here, the start offset of the mailbox registers, this question is based on the figure below?

I'll expand on this to say

means that the offset of the start of the mailbox payload (n) is given by n = m + sizeof....

Which means the diagram below is wrong as should align with top of mailbox registers.

Quoted 3 lines

include/hw/cxl/cxl_device.h line 23

Quoted 3 lines
include/hw/cxl/cxl_device.h new file +2 -0
+ *
+ * This is roughly described in 8.2.8 Figure 138 of the CXL 2.0 spec  

I'm going drop this comment as that figure appears unrelated to me.

Quoted 37 lines
Quoted 26 lines
include/hw/cxl/cxl_device.h new file +26 -0
+ *
+ *                       +---------------------------------+
+ *                       |                                 |
+ *                       |    Memory Device Registers      |
+ *                       |                                 |
+ * n + PAYLOAD_SIZE_MAX  -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |         Mailbox Payload         |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    -----------------------------------
+ *                  |    |       Mailbox Registers         |
+ *                  |    |                                 |
+ *                  n    -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |        Device Registers         |
+ *                  |    |                                 |
+ *                  m    ---------------------------------->
+ *                  ^    |  Memory Device Capability Header|
+ *                  |    -----------------------------------
+ *                  |    |     Mailbox Capability Header   |

include/hw/cxl/cxl_device.h line 59

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+ *                  |    -------------- --------------------
+ *                  |    |     Device Capability Header    |
+ *                  |    -----------------------------------
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |      Device Cap Array[0..n]     |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                       |                                 |
+ *                  0    +---------------------------------+    
Quoted 1 line

Would it make sense to add CXL cap header register to the diagram?

Quoted 7 lines

Too many similar names in the CXL spec. I'm not sure which one you mean, could you let me have a reference? If you mean the one that is at the start of the CXL.cache and CXL.mem registers that whole region isn't covered by this diagram and might be in a different BAR. Here we are only dealing with the Memory Device Registers. I'll add statement to the initial comment block to make that clear as it definitely isn't currently!

Quoted 4 lines

I was thinking 0 in your figure is the device capabilities array register, which tells us how many capabilites that are in the array. This would be 8.2.8.1. After that comes 8.2.8.2 with n capability header registers which point to the device registers.

Got it. See below.

Quoted 9 lines
Quoted 9 lines
Quoted 5 lines

n also
seems to be the size of the cap array, but it is also an offset so that could be clarified.

Ah. Letter reuse. good point. Looking more closely it isn't an array anyway in the diagram (the array would have to include the Device Capability Header and Mailbox Capability headers. Renamed as simply Device Cap Array Register

As mentioned here, the array is misleading anyway because we have the actual entries listed directly above it rather than 'inside' the array. Hence the change described above.

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines

include/hw/cxl/cxl_device.h line 67

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +8 -0
+ *
+ */
+
+#define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
+#define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
+#define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
+
+#define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */    

Is this to plan for future capabilities? If we have CAPS MAX doesn't this allow us to remove the slack space.

I missed replying to this before.

So far CAPS MAX covers everything in the spec. (room for secondary mailbox
+ the 3 we have implemented).
We don't support migration etc yet (and I'm not sure we ever will) anyway so I'm not hugely bothered about backwards compatibility. Hence we can just move things if needed later.

Quoted 40 lines
Quoted 13 lines
Quoted 5 lines

include/hw/cxl/cxl_device.h line 68

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+#define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */    

Should we add status to the name here, or would it get too long?

Quoted 7 lines

include/hw/cxl/cxl_device.h line 73

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+
+#define CXL_MAILBOX_REGISTERS_OFFSET \
+    (CXL_DEVICE_REGISTERS_OFFSET + CXL_DEVICE_REGISTERS_LENGTH)
+#define CXL_MAILBOX_REGISTERS_SIZE 0x20 /* 8.2.8.4, Figure 139 */
+#define CXL_MAILBOX_PAYLOAD_SHIFT 11    

I see 20 in the spec.

Quoted 2 lines

It's an implementation choice between 8 and 20. For now, this code goes with 11 for no particularly strong reason.

Got it.

Quoted 22 lines
Quoted 21 lines
Quoted 8 lines
include/hw/cxl/cxl_device.h new file +6 -0
+#define CXL_MAILBOX_MAX_PAYLOAD_SIZE (1 << CXL_MAILBOX_PAYLOAD_SHIFT)
+#define CXL_MAILBOX_REGISTERS_LENGTH \
+    (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
+
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;

include/hw/cxl/cxl_device.h line 89

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;    

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Quoted 7 lines
Quoted 4 lines

Discussed in the other branch of this thread. Short answer is we don't support non persistent yet but it's on the todo list. What exactly that looks like is to be determined. One aspect of that is there isn't currently a software stack to test volatile memory.

If you can elaborate more here on what is needed to test the volatile memory stack we may be able to help out.

There are a bunch of different ways this could be done - ultimate we probably want to do all of them.

https://cdrdv2.intel.com/v1/dl/getContent/643805?wapkw=CXL%20memory%20device%20sw%20guide
has some suggestions (though no one is obliged to follow them!) See 2.4

First assumption is that for volatile devices, a common approach will be to do all the setup in firmware before the OS boots and just present normal SRAT, HMAT and memory tables as if it were any other memory. If we want to go that way for testing purposes then we'd need an open source firmware to implement setup similar to that done in Linux - probably EDK2.

Of course, volatile memory might be hot added, in which case the OS may be involved. In that case I think the main missing part would be actually doing the final memory hotplug event to expose it to the OS + the necessary dynamic updating of the OS numa description etc. There is work on going to get the information needed but I think we are still some way off actually tying everything together.

Dan / Ben and team may be able to share more status information.

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
include/hw/cxl/cxl_device.h new file +4 -0
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);

...

Quoted 5 lines
Quoted 5 lines
Quoted 5 lines

+cc Dave, Klaus, Tong
Other than the minor issues raised.

Looks good.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Reviewed by: Adam Manzanares <[email protected]>

Btw I haven't accepted all changes, but have been picking up your RB. Shout if that's not fine with you.

Definitely fine with me and was my intention. Let us know how we can help move the work forward. I am kick starting reviewing and will try to bring others in.

Great. For various reasons I'll not bother mention here (see my employer ;) I need to keep any discussions on mailing list or in a 'published' form. So discussion on mailing list + at conferences works best for me but we can organize some suitably hosted public calls if needed to align plans. There is a plan for uconf at Plumbers this year which will hopefully let us do any longer term planning. Shorter term my aims around QEMU side of things are:

1) Get the initial support upstream as I'm getting bored of rebasing :) I think we are in a fairly good state for doing that once qemu 7.0 is out.
2) Improved tests so it doesn't break when no one is paying attention.
3) Expand out the feature set to keep up with what is going on Linux kernel wise (personally no other OS of interest, but it would be great if anyone wanted to help deal with other operating systems that care).
* RAS
* CDAT for switches etc, host table updates for generic port definition
- What ever else I've missed recently. When the region code finalizes I suspect we'll want to add a load more tests to stress various corners of that.
* Alison may help with partitioning support.
4) Expand features where we have currently taken a short cut such as enabling multiple HDM decoders.
5) Use it as a path for testing spec features before publication (obviously can't talk about that on list but I've open in appropriate venue about that).

Happy to have help on any of the above, but 'features' that are reasonably separate such as RAS support might be a good place for contributions that won't be greatly affected by any other refactoring going on.

I've pushed all but SPDM support and stuff for which the spec isn't public yet up on
https://gitlab.com/jic23/qemu/-/commits/cxl-v9-draft-1
(as you can see CI found a segfault today so I'll push the fix out for that
shortly - that also highlighted a build breakage mid series that I've fixed up.).

Jonathan

Quoted 6 lines
Quoted 6 lines

Thanks.

Jonathan

#17 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

Quoted 219 lines · Jonathan Cameron
Quoted 114 lines · On Thu, 31 Mar 2022 22:13:20 +0000 Adam Manzanares <[email protected]>…
Quoted 104 lines · Jonathan Cameron
Quoted 15 lines · On Tue, 29 Mar 2022 18:13:59 +0000 Adam Manzanares <[email protected]>…
Quoted 14 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL device is a type of CXL component. Conceptually, a CXL device would be a leaf node in a CXL topology. From an emulation perspective, CXL devices are the most complex and so the actual implementation is reserved for discrete commits.

This new device type is specifically catered towards the eventual implementation of a Type3 CXL.mem device, 8.2.8.5 in the CXL 2.0 specification.

...

Quoted 22 lines
Quoted 22 lines
include/hw/cxl/cxl_device.h new file +16 -0
new file mode 100644
index 0000000000..b2416e45bf
@@ -0,0 +1,165 @@
+/*
+ * QEMU CXL Devices
+ *
+ * Copyright (c) 2020 Intel
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#ifndef CXL_DEVICE_H
+#define CXL_DEVICE_H
+
+#include "hw/register.h"
+
+/*
+ * The following is how a CXL device's MMIO space is laid out. The only
Quoted 8 lines

include/hw/cxl/cxl_device.h line 21

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+ * requirement from the spec is that the capabilities array and the capability
+ * headers start at offset 0 and are contiguously packed. The headers themselves
+ * provide offsets to the register fields. For this emulation, registers will
+ * start at offset 0x80 (m == 0x80). No secondary mailbox is implemented which
+ * means that n = m + sizeof(mailbox registers) + sizeof(device registers).    

What is n here, the start offset of the mailbox registers, this question is based on the figure below?

I'll expand on this to say

means that the offset of the start of the mailbox payload (n) is given by n = m + sizeof....

Which means the diagram below is wrong as should align with top of mailbox registers.

Quoted 3 lines

include/hw/cxl/cxl_device.h line 23

Quoted 3 lines
include/hw/cxl/cxl_device.h new file +2 -0
+ *
+ * This is roughly described in 8.2.8 Figure 138 of the CXL 2.0 spec  

I'm going drop this comment as that figure appears unrelated to me.

Quoted 37 lines
Quoted 26 lines
include/hw/cxl/cxl_device.h new file +26 -0
+ *
+ *                       +---------------------------------+
+ *                       |                                 |
+ *                       |    Memory Device Registers      |
+ *                       |                                 |
+ * n + PAYLOAD_SIZE_MAX  -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |         Mailbox Payload         |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    -----------------------------------
+ *                  |    |       Mailbox Registers         |
+ *                  |    |                                 |
+ *                  n    -----------------------------------
+ *                  ^    |                                 |
+ *                  |    |        Device Registers         |
+ *                  |    |                                 |
+ *                  m    ---------------------------------->
+ *                  ^    |  Memory Device Capability Header|
+ *                  |    -----------------------------------
+ *                  |    |     Mailbox Capability Header   |

include/hw/cxl/cxl_device.h line 59

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+ *                  |    -------------- --------------------
+ *                  |    |     Device Capability Header    |
+ *                  |    -----------------------------------
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                  |    |      Device Cap Array[0..n]     |
+ *                  |    |                                 |
+ *                  |    |                                 |
+ *                       |                                 |
+ *                  0    +---------------------------------+    
Quoted 1 line

Would it make sense to add CXL cap header register to the diagram?

Quoted 7 lines

Too many similar names in the CXL spec. I'm not sure which one you mean, could you let me have a reference? If you mean the one that is at the start of the CXL.cache and CXL.mem registers that whole region isn't covered by this diagram and might be in a different BAR. Here we are only dealing with the Memory Device Registers. I'll add statement to the initial comment block to make that clear as it definitely isn't currently!

Quoted 4 lines

I was thinking 0 in your figure is the device capabilities array register, which tells us how many capabilites that are in the array. This would be 8.2.8.1. After that comes 8.2.8.2 with n capability header registers which point to the device registers.

Got it. See below.

Quoted 9 lines
Quoted 9 lines
Quoted 5 lines

n also
seems to be the size of the cap array, but it is also an offset so that could be clarified.

Ah. Letter reuse. good point. Looking more closely it isn't an array anyway in the diagram (the array would have to include the Device Capability Header and Mailbox Capability headers. Renamed as simply Device Cap Array Register

As mentioned here, the array is misleading anyway because we have the actual entries listed directly above it rather than 'inside' the array. Hence the change described above.

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines

include/hw/cxl/cxl_device.h line 67

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +8 -0
+ *
+ */
+
+#define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
+#define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
+#define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
+
+#define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */    

Is this to plan for future capabilities? If we have CAPS MAX doesn't this allow us to remove the slack space.

I missed replying to this before.

So far CAPS MAX covers everything in the spec. (room for secondary mailbox
+ the 3 we have implemented).
We don't support migration etc yet (and I'm not sure we ever will) anyway so I'm not hugely bothered about backwards compatibility. Hence we can just move things if needed later.

Quoted 40 lines
Quoted 13 lines
Quoted 5 lines

include/hw/cxl/cxl_device.h line 68

Quoted 2 lines
include/hw/cxl/cxl_device.h new file +1 -0
+#define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */    

Should we add status to the name here, or would it get too long?

Quoted 7 lines

include/hw/cxl/cxl_device.h line 73

Quoted 5 lines
include/hw/cxl/cxl_device.h new file +5 -0
+
+#define CXL_MAILBOX_REGISTERS_OFFSET \
+    (CXL_DEVICE_REGISTERS_OFFSET + CXL_DEVICE_REGISTERS_LENGTH)
+#define CXL_MAILBOX_REGISTERS_SIZE 0x20 /* 8.2.8.4, Figure 139 */
+#define CXL_MAILBOX_PAYLOAD_SHIFT 11    

I see 20 in the spec.

Quoted 2 lines

It's an implementation choice between 8 and 20. For now, this code goes with 11 for no particularly strong reason.

Got it.

Quoted 22 lines
Quoted 21 lines
Quoted 8 lines
include/hw/cxl/cxl_device.h new file +6 -0
+#define CXL_MAILBOX_MAX_PAYLOAD_SIZE (1 << CXL_MAILBOX_PAYLOAD_SHIFT)
+#define CXL_MAILBOX_REGISTERS_LENGTH \
+    (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
+
+typedef struct cxl_device_state {
+    MemoryRegion device_registers;

include/hw/cxl/cxl_device.h line 89

Quoted 10 lines
include/hw/cxl/cxl_device.h new file +10 -0
+
+    /* mmio for device capabilities array - 8.2.8.2 */
+    MemoryRegion device;
+    MemoryRegion caps;
+
+    /* mmio for the mailbox registers 8.2.8.4 */
+    MemoryRegion mailbox;
+
+    /* memory region for persistent memory, HDM */
+    uint64_t pmem_size;    

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Quoted 7 lines
Quoted 4 lines

Discussed in the other branch of this thread. Short answer is we don't support non persistent yet but it's on the todo list. What exactly that looks like is to be determined. One aspect of that is there isn't currently a software stack to test volatile memory.

If you can elaborate more here on what is needed to test the volatile memory stack we may be able to help out.

There are a bunch of different ways this could be done - ultimate we probably want to do all of them.

https://urldefense.com/v3/__https://cdrdv2.intel.com/v1/dl/getContent/643805?wapkw=CXL*20memory*20device*20sw*20guide__;JSUlJQ!!EwVzqGoTKBqv-0DWAJBm!HzD_Dh_I9m9MydppOSSyhuzvwTawlg7LE77bEYiZ1i3AMgxV_YOI56VeZgkg-KuX7XMA$
has some suggestions (though no one is obliged to follow them!) See 2.4

First assumption is that for volatile devices, a common approach will be to do all the setup in firmware before the OS boots and just present normal SRAT, HMAT and memory tables as if it were any other memory. If we want to go that way for testing purposes then we'd need an open source firmware to implement setup similar to that done in Linux - probably EDK2.

Quoted 7 lines

Of course, volatile memory might be hot added, in which case the OS may be involved. In that case I think the main missing part would be actually doing the final memory hotplug event to expose it to the OS + the necessary dynamic updating of the OS numa description etc. There is work on going to get the information needed but I think we are still some way off actually tying everything together.

Dan / Ben and team may be able to share more status information.

Great, thanks for all of the information. We will start planning out our next steps. I'll add Luis on cc since he has chatted with me about setting up a test framework for the CXL kernel code that will rely on QEMU.

Quoted 21 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
include/hw/cxl/cxl_device.h new file +4 -0
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);

...

Quoted 10 lines
Quoted 9 lines
Quoted 6 lines

+cc Dave, Klaus, Tong
Other than the minor issues raised.

Looks good.

Reviewed by: Adam Manzanares <[email protected]>

Btw I haven't accepted all changes, but have been picking up your RB. Shout if that's not fine with you.

Quoted 8 lines
Quoted 2 lines

Definitely fine with me and was my intention. Let us know how we can help move the work forward. I am kick starting reviewing and will try to bring others in.

Great. For various reasons I'll not bother mention here (see my employer ;) I need to keep any discussions on mailing list or in a 'published' form. So discussion on mailing list + at conferences works best for me but we can organize some suitably hosted public calls if needed to align plans. There is a plan for uconf at Plumbers this year which will hopefully let

We would also prefer to keep discussions in the public domain. We have plans to attend Plumbers this year, so we look forward to discussing in person.

Quoted 7 lines

us do any longer term planning. Shorter term my aims around QEMU side of things are:

1) Get the initial support upstream as I'm getting bored of rebasing :) I think we are in a fairly good state for doing that once qemu 7.0 is out.
2) Improved tests so it doesn't break when no one is paying attention.

Luis may have some thoughts here.

Quoted 14 lines

3) Expand out the feature set to keep up with what is going on Linux kernel wise (personally no other OS of interest, but it would be great if anyone wanted to help deal with other operating systems that care).
* RAS
* CDAT for switches etc, host table updates for generic port definition
- What ever else I've missed recently. When the region code finalizes I suspect we'll want to add a load more tests to stress various corners of that.
* Alison may help with partitioning support.
4) Expand features where we have currently taken a short cut such as enabling multiple HDM decoders.
5) Use it as a path for testing spec features before publication (obviously can't talk about that on list but I've open in appropriate venue about that).

Quoted 9 lines

Happy to have help on any of the above, but 'features' that are reasonably separate such as RAS support might be a good place for contributions that won't be greatly affected by any other refactoring going on.

I've pushed all but SPDM support and stuff for which the spec isn't public yet up on
https://urldefense.com/v3/__https://gitlab.com/jic23/qemu/-/commits/cxl-v9-draft-1__;!!EwVzqGoTKBqv-0DWAJBm!HzD_Dh_I9m9MydppOSSyhuzvwTawlg7LE77bEYiZ1i3AMgxV_YOI56VeZgkg-EMwmPTV$
(as you can see CI found a segfault today so I'll push the fix out for that
shortly - that also highlighted a build breakage mid series that I've fixed up.).

Once again thanks for all of the pointers.

Quoted 10 lines

Jonathan

Quoted 6 lines
Quoted 6 lines

Thanks.

Jonathan

#18 lore raw

Re: [PATCH v8 04/46] hw/cxl/device: Introduce a CXL device (8.2.8)

...

Quoted 60 lines
Quoted 24 lines
Quoted 11 lines
Quoted 8 lines
Quoted 3 lines

Can we switch this to mem_size and drop the persistent comment? It is my understanding that HDM is independent of persistence.

Discussed in the other branch of this thread. Short answer is we don't support non persistent yet but it's on the todo list. What exactly that looks like is to be determined. One aspect of that is there isn't currently a software stack to test volatile memory.

If you can elaborate more here on what is needed to test the volatile memory stack we may be able to help out.

There are a bunch of different ways this could be done - ultimate we probably want to do all of them.

https://urldefense.com/v3/__https://cdrdv2.intel.com/v1/dl/getContent/643805?wapkw=CXL*20memory*20device*20sw*20guide__;JSUlJQ!!EwVzqGoTKBqv-0DWAJBm!HzD_Dh_I9m9MydppOSSyhuzvwTawlg7LE77bEYiZ1i3AMgxV_YOI56VeZgkg-KuX7XMA$
has some suggestions (though no one is obliged to follow them!) See 2.4

First assumption is that for volatile devices, a common approach will be to do all the setup in firmware before the OS boots and just present normal SRAT, HMAT and memory tables as if it were any other memory. If we want to go that way for testing purposes then we'd need an open source firmware to implement setup similar to that done in Linux - probably EDK2.

Quoted 7 lines

Of course, volatile memory might be hot added, in which case the OS may be involved. In that case I think the main missing part would be actually doing the final memory hotplug event to expose it to the OS + the necessary dynamic updating of the OS numa description etc. There is work on going to get the information needed but I think we are still some way off actually tying everything together.

Dan / Ben and team may be able to share more status information.

Great, thanks for all of the information. We will start planning out our next steps. I'll add Luis on cc since he has chatted with me about setting up a test framework for the CXL kernel code that will rely on QEMU.

Quoted 24 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
include/hw/cxl/cxl_device.h new file +4 -0
+} CXLDeviceState;
+
+/* Initialize the register block for a device */
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *dev);  

...

Quoted 12 lines
Quoted 9 lines
Quoted 6 lines

+cc Dave, Klaus, Tong
Other than the minor issues raised.

Looks good.

Reviewed by: Adam Manzanares <[email protected]>

Btw I haven't accepted all changes, but have been picking up your RB. Shout if that's not fine with you.

Definitely fine with me and was my intention. Let us know how we can help move the work forward. I am kick starting reviewing and will try to bring others in.

Quoted 8 lines
Quoted 5 lines

Great. For various reasons I'll not bother mention here (see my employer ;) I need to keep any discussions on mailing list or in a 'published' form. So discussion on mailing list + at conferences works best for me but we can organize some suitably hosted public calls if needed to align plans. There is a plan for uconf at Plumbers this year which will hopefully let

We would also prefer to keep discussions in the public domain. We have plans to attend Plumbers this year, so we look forward to discussing in person.

Excellent. If it's useful to have a public discussion before plumbers then the nice folk at Linaro have been kind enough to host similar discussion in the past (and deal with posting recordings etc afterwards for those who missed the live call) and I expect they'd help us out again (Hi Alex ;)

Quoted 10 lines
Quoted 8 lines

us do any longer term planning. Shorter term my aims around QEMU side of things are:

1) Get the initial support upstream as I'm getting bored of rebasing :) I think we are in a fairly good state for doing that once qemu 7.0 is out.
2) Improved tests so it doesn't break when no one is paying attention.

Luis may have some thoughts here.

Excellent. A testing expert is always useful. It would be nice to think about getting something beyond a basic 'does it boot' test into the qemu CI but I've not really looked into how one might do that.

Quoted 19 lines
Quoted 19 lines

3) Expand out the feature set to keep up with what is going on Linux kernel wise (personally no other OS of interest, but it would be great if anyone wanted to help deal with other operating systems that care).
* RAS
* CDAT for switches etc, host table updates for generic port definition
- What ever else I've missed recently. When the region code finalizes I suspect we'll want to add a load more tests to stress various corners of that.
* Alison may help with partitioning support.
4) Expand features where we have currently taken a short cut such as enabling multiple HDM decoders.
5) Use it as a path for testing spec features before publication (obviously can't talk about that on list but I've open in appropriate venue about that).

Happy to have help on any of the above, but 'features' that are reasonably separate such as RAS support might be a good place for contributions that won't be greatly affected by any other refactoring going on.

Quoted 7 lines
Quoted 5 lines

I've pushed all but SPDM support and stuff for which the spec isn't public yet up on
https://urldefense.com/v3/__https://gitlab.com/jic23/qemu/-/commits/cxl-v9-draft-1__;!!EwVzqGoTKBqv-0DWAJBm!HzD_Dh_I9m9MydppOSSyhuzvwTawlg7LE77bEYiZ1i3AMgxV_YOI56VeZgkg-EMwmPTV$
(as you can see CI found a segfault today so I'll push the fix out for that
shortly - that also highlighted a build breakage mid series that I've fixed up.).

Once again thanks for all of the pointers.

You are welcome. It's nice to see this work gain traction :)

Anyhow, v9 is on it's way (slowly) through our firewall (got log anti spam send rate limits) so fingers crossed we are nearly ready with this first bit of support to build more fun stuff on top of.

Jonathan

Quoted 2 lines
Quoted 2 lines

Jonathan

#19 lore raw

PATCH v8 5/46 hw/cxl/device: Implement the CAP array (8.2.8.1-2)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

This implements all device MMIO up to the first capability. That includes the CXL Device Capabilities Array Register, as well as all of the CXL Device Capability Header Registers. The latter are filled in as they are implemented in the following patches.

Endianness and alignment are managed by softmmu memory core.

---
 hw/cxl/cxl-device-utils.c   | 109 ++++++++++++++++++++++++++++++++++++
 hw/cxl/meson.build          |   1 +
 include/hw/cxl/cxl_device.h |  31 +++++++++-
 3 files changed, 140 insertions(+), 1 deletion(-)
hw/cxl/cxl-device-utils.c new file +109 -0
new file mode 100644
index 0000000000..0895b9d78b
@@ -0,0 +1,109 @@
+/*
+ * CXL Utility library for devices
+ *
+ * Copyright(C) 2020 Intel Corporation.
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See the
+ * COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu/log.h"
+#include "hw/cxl/cxl.h"
+
+/*
+ * Device registers have no restrictions per the spec, and so fall back to the
+ * default memory mapped register rules in 8.2:
+ *   Software shall use CXL.io Memory Read and Write to access memory mapped
+ *   register defined in this section. Unless otherwise specified, software
+ *   shall restrict the accesses width based on the following:
+ *   • A 32 bit register shall be accessed as a 1 Byte, 2 Bytes or 4 Bytes
+ *     quantity.
+ *   • A 64 bit register shall be accessed as a 1 Byte, 2 Bytes, 4 Bytes or 8
+ *     Bytes
+ *   • The address shall be a multiple of the access width, e.g. when
+ *     accessing a register as a 4 Byte quantity, the address shall be
+ *     multiple of 4.
+ *   • The accesses shall map to contiguous bytes.If these rules are not
+ *     followed, the behavior is undefined
+ */
+
+static uint64_t caps_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    CXLDeviceState *cxl_dstate = opaque;
+
+    if (size == 4) {
+        return cxl_dstate->caps_reg_state32[offset / 4];
+    } else {
+        return cxl_dstate->caps_reg_state64[offset / 8];
+    }
+}
+
+static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    return 0;
+}
+
+static const MemoryRegionOps dev_ops = {
+    .read = dev_reg_read,
+    .write = NULL, /* status register is read only */
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+        .unaligned = false,
+    },
+    .impl = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+    },
+};
+
+static const MemoryRegionOps caps_ops = {
+    .read = caps_reg_read,
+    .write = NULL, /* caps registers are read only */
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+        .unaligned = false,
+    },
+    .impl = {
+        .min_access_size = 4,
+        .max_access_size = 8,
+    },
+};
+
+void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
+{
+    /* This will be a BAR, so needs to be rounded up to pow2 for PCI spec */
+    memory_region_init(&cxl_dstate->device_registers, obj, "device-registers",
+                       pow2ceil(CXL_MMIO_SIZE));
+
+    memory_region_init_io(&cxl_dstate->caps, obj, &caps_ops, cxl_dstate,
+                          "cap-array", CXL_CAPS_SIZE);
+    memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
+                          "device-status", CXL_DEVICE_REGISTERS_LENGTH);
+
+    memory_region_add_subregion(&cxl_dstate->device_registers, 0,
+                                &cxl_dstate->caps);
+    memory_region_add_subregion(&cxl_dstate->device_registers,
+                                CXL_DEVICE_REGISTERS_OFFSET,
+                                &cxl_dstate->device);
+}
+
+static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
+
+void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
+{
+    uint64_t *cap_hdrs = cxl_dstate->caps_reg_state64;
+    const int cap_count = 1;
+
+    /* CXL Device Capabilities Array Register */
+    ARRAY_FIELD_DP64(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
+    ARRAY_FIELD_DP64(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_VERSION, 1);
+    ARRAY_FIELD_DP64(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_COUNT, cap_count);
+
+    cxl_device_cap_init(cxl_dstate, DEVICE, 1);
+    device_reg_init_common(cxl_dstate);
+}
hw/cxl/meson.build +1 -0
index 3231b5de1e..dd7c6f8e5a 100644
@@ -1,4 +1,5 @@
 softmmu_ss.add(when: 'CONFIG_CXL',
                if_true: files(
                    'cxl-component-utils.c',
+                   'cxl-device-utils.c',
                ))
include/hw/cxl/cxl_device.h +30 -1
index b2416e45bf..1ac0dcd97e 100644
@@ -63,6 +63,8 @@
 #define CXL_DEVICE_CAP_HDR1_OFFSET 0x10 /* Figure 138 */
 #define CXL_DEVICE_CAP_REG_SIZE 0x10 /* 8.2.8.2 */
 #define CXL_DEVICE_CAPS_MAX 4 /* 8.2.8.2.1 + 8.2.8.5 */
+#define CXL_CAPS_SIZE \
+    (CXL_DEVICE_CAP_REG_SIZE * (CXL_DEVICE_CAPS_MAX + 1)) /* +1 for header */
 
 #define CXL_DEVICE_REGISTERS_OFFSET 0x80 /* Read comment above */
 #define CXL_DEVICE_REGISTERS_LENGTH 0x8 /* 8.2.8.3.1 */
@@ -75,12 +77,22 @@
 #define CXL_MAILBOX_REGISTERS_LENGTH \
     (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
 
+#define CXL_MMIO_SIZE                                           \
+    (CXL_DEVICE_CAP_REG_SIZE + CXL_DEVICE_REGISTERS_LENGTH +    \
+     CXL_MAILBOX_REGISTERS_LENGTH)
+
 typedef struct cxl_device_state {
     MemoryRegion device_registers;
 
     /* mmio for device capabilities array - 8.2.8.2 */
     MemoryRegion device;
-    MemoryRegion caps;
+    struct {
+        MemoryRegion caps;
+        union {
+            uint32_t caps_reg_state32[CXL_CAPS_SIZE / 4];
+            uint64_t caps_reg_state64[CXL_CAPS_SIZE / 8];
+        };
+    };
 
     /* mmio for the mailbox registers 8.2.8.4 */
     MemoryRegion mailbox;
@@ -133,6 +145,23 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
                                                CXL_DEVICE_CAP_REG_SIZE)
 
+#define cxl_device_cap_init(dstate, reg, cap_id)                           \
+    do {                                                                   \
+        uint32_t *cap_hdrs = dstate->caps_reg_state32;                     \
+        int which = R_CXL_DEV_##reg##_CAP_HDR0;                            \
+        cap_hdrs[which] =                                                  \
+            FIELD_DP32(cap_hdrs[which], CXL_DEV_##reg##_CAP_HDR0,          \
+                       CAP_ID, cap_id);                                    \
+        cap_hdrs[which] = FIELD_DP32(                                      \
+            cap_hdrs[which], CXL_DEV_##reg##_CAP_HDR0, CAP_VERSION, 1);    \
+        cap_hdrs[which + 1] =                                              \
+            FIELD_DP32(cap_hdrs[which + 1], CXL_DEV_##reg##_CAP_HDR1,      \
+                       CAP_OFFSET, CXL_##reg##_REGISTERS_OFFSET);          \
+        cap_hdrs[which + 2] =                                              \
+            FIELD_DP32(cap_hdrs[which + 2], CXL_DEV_##reg##_CAP_HDR2,      \
+                       CAP_LENGTH, CXL_##reg##_REGISTERS_LENGTH);          \
+    } while (0)
+
 REG32(CXL_DEV_MAILBOX_CAP, 0)
     FIELD(CXL_DEV_MAILBOX_CAP, PAYLOAD_SIZE, 0, 5)
     FIELD(CXL_DEV_MAILBOX_CAP, INT_CAP, 5, 1)
Signature · 1 line
2.32.0
#20 lore raw

PATCH v8 6/46 hw/cxl/device: Implement basic mailbox (8.2.8.4)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

This is the beginning of implementing mailbox support for CXL 2.0 devices. The implementation recognizes when the doorbell is rung, handles the command/payload, clears the doorbell while returning error codes and data.

Generally the mailbox mechanism is designed to permit communication between the host OS and the firmware running on the device. For our purposes, we emulate both the firmware, implemented primarily in cxl-mailbox-utils.c, and the hardware.

No commands are implemented yet.

---
 hw/cxl/cxl-device-utils.c   | 122 ++++++++++++++++++++++++-
 hw/cxl/cxl-mailbox-utils.c  | 171 ++++++++++++++++++++++++++++++++++++
 hw/cxl/meson.build          |   1 +
 include/hw/cxl/cxl.h        |   3 +
 include/hw/cxl/cxl_device.h |  19 +++-
 5 files changed, 314 insertions(+), 2 deletions(-)
hw/cxl/cxl-device-utils.c +121 -1
index 0895b9d78b..4b995beba7 100644
@@ -44,6 +44,108 @@ static uint64_t dev_reg_read(void *opaque, hwaddr offset, unsigned size)
     return 0;
 }
 
+static uint64_t mailbox_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    CXLDeviceState *cxl_dstate = opaque;
+
+    switch (size) {
+    case 1:
+        return cxl_dstate->mbox_reg_state[offset];
+    case 2:
+        return cxl_dstate->mbox_reg_state16[offset / 2];
+    case 4:
+        return cxl_dstate->mbox_reg_state32[offset / 4];
+    case 8:
+        return cxl_dstate->mbox_reg_state64[offset / 8];
+    default:
+        g_assert_not_reached();
+    }
+}
+
+static void mailbox_mem_writel(uint32_t *reg_state, hwaddr offset,
+                               uint64_t value)
+{
+    switch (offset) {
+    case A_CXL_DEV_MAILBOX_CTRL:
+        /* fallthrough */
+    case A_CXL_DEV_MAILBOX_CAP:
+        /* RO register */
+        break;
+    default:
+        qemu_log_mask(LOG_UNIMP,
+                      "%s Unexpected 32-bit access to 0x%" PRIx64 " (WI)\n",
+                      __func__, offset);
+        return;
+    }
+
+    reg_state[offset / 4] = value;
+}
+
+static void mailbox_mem_writeq(uint64_t *reg_state, hwaddr offset,
+                               uint64_t value)
+{
+    switch (offset) {
+    case A_CXL_DEV_MAILBOX_CMD:
+        break;
+    case A_CXL_DEV_BG_CMD_STS:
+        /* BG not supported */
+        /* fallthrough */
+    case A_CXL_DEV_MAILBOX_STS:
+        /* Read only register, will get updated by the state machine */
+        return;
+    default:
+        qemu_log_mask(LOG_UNIMP,
+                      "%s Unexpected 64-bit access to 0x%" PRIx64 " (WI)\n",
+                      __func__, offset);
+        return;
+    }
+
+
+    reg_state[offset / 8] = value;
+}
+
+static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
+                              unsigned size)
+{
+    CXLDeviceState *cxl_dstate = opaque;
+
+    if (offset >= A_CXL_DEV_CMD_PAYLOAD) {
+        memcpy(cxl_dstate->mbox_reg_state + offset, &value, size);
+        return;
+    }
+
+    switch (size) {
+    case 4:
+        mailbox_mem_writel(cxl_dstate->mbox_reg_state32, offset, value);
+        break;
+    case 8:
+        mailbox_mem_writeq(cxl_dstate->mbox_reg_state64, offset, value);
+        break;
+    default:
+        g_assert_not_reached();
+    }
+
+    if (ARRAY_FIELD_EX32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CTRL,
+                         DOORBELL)) {
+        cxl_process_mailbox(cxl_dstate);
+    }
+}
+
+static const MemoryRegionOps mailbox_ops = {
+    .read = mailbox_reg_read,
+    .write = mailbox_reg_write,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+        .unaligned = false,
+    },
+    .impl = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+    },
+};
+
 static const MemoryRegionOps dev_ops = {
     .read = dev_reg_read,
     .write = NULL, /* status register is read only */
@@ -84,20 +186,33 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
                           "cap-array", CXL_CAPS_SIZE);
     memory_region_init_io(&cxl_dstate->device, obj, &dev_ops, cxl_dstate,
                           "device-status", CXL_DEVICE_REGISTERS_LENGTH);
+    memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
+                          "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
 
     memory_region_add_subregion(&cxl_dstate->device_registers, 0,
                                 &cxl_dstate->caps);
     memory_region_add_subregion(&cxl_dstate->device_registers,
                                 CXL_DEVICE_REGISTERS_OFFSET,
                                 &cxl_dstate->device);
+    memory_region_add_subregion(&cxl_dstate->device_registers,
+                                CXL_MAILBOX_REGISTERS_OFFSET,
+                                &cxl_dstate->mailbox);
 }
 
 static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
 
+static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
+{
+    /* 2048 payload size, with no interrupt or background support */
+    ARRAY_FIELD_DP32(cxl_dstate->mbox_reg_state32, CXL_DEV_MAILBOX_CAP,
+                     PAYLOAD_SIZE, CXL_MAILBOX_PAYLOAD_SHIFT);
+    cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
+}
+
 void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 {
     uint64_t *cap_hdrs = cxl_dstate->caps_reg_state64;
-    const int cap_count = 1;
+    const int cap_count = 2;
 
     /* CXL Device Capabilities Array Register */
     ARRAY_FIELD_DP64(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
@@ -106,4 +221,9 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 
     cxl_device_cap_init(cxl_dstate, DEVICE, 1);
     device_reg_init_common(cxl_dstate);
+
+    cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
+    mailbox_reg_init_common(cxl_dstate);
+
+    assert(cxl_initialize_mailbox(cxl_dstate) == 0);
 }
hw/cxl/cxl-mailbox-utils.c new file +171 -0

171 lines not on this page · raw message

hw/cxl/meson.build +1 -0
index dd7c6f8e5a..e68eea2358 100644
@@ -2,4 +2,5 @@ softmmu_ss.add(when: 'CONFIG_CXL',
                if_true: files(
                    'cxl-component-utils.c',
                    'cxl-device-utils.c',
+                   'cxl-mailbox-utils.c',
                ))
include/hw/cxl/cxl.h +3 -0
index b9d1ac3fad..554ad93b6b 100644
@@ -14,4 +14,7 @@
 #include "cxl_component.h"
 #include "cxl_device.h"
 
+#define CXL_COMPONENT_REG_BAR_IDX 0
+#define CXL_DEVICE_REG_BAR_IDX 2
+
 #endif
include/hw/cxl/cxl_device.h +18 -1
index 1ac0dcd97e..49dcca7e44 100644
@@ -95,7 +95,21 @@ typedef struct cxl_device_state {
     };
 
     /* mmio for the mailbox registers 8.2.8.4 */
-    MemoryRegion mailbox;
+    struct {
+        MemoryRegion mailbox;
+        uint16_t payload_size;
+        union {
+            uint8_t mbox_reg_state[CXL_MAILBOX_REGISTERS_LENGTH];
+            uint16_t mbox_reg_state16[CXL_MAILBOX_REGISTERS_LENGTH / 2];
+            uint32_t mbox_reg_state32[CXL_MAILBOX_REGISTERS_LENGTH / 4];
+            uint64_t mbox_reg_state64[CXL_MAILBOX_REGISTERS_LENGTH / 8];
+        };
+        struct cel_log {
+            uint16_t opcode;
+            uint16_t effect;
+        } cel_log[1 << 16];
+        size_t cel_size;
+    };
 
     /* memory region for persistent memory, HDM */
     uint64_t pmem_size;
@@ -145,6 +159,9 @@ CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
                                                CXL_DEVICE_CAP_REG_SIZE)
 
+int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
+void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
+
 #define cxl_device_cap_init(dstate, reg, cap_id)                           \
     do {                                                                   \
         uint32_t *cap_hdrs = dstate->caps_reg_state32;                     \
Signature · 1 line
2.32.0
#21 lore raw

PATCH v8 7/46 hw/cxl/device: Add memory device utilities

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

Memory devices implement extra capabilities on top of CXL devices. This adds support for that.

A large part of memory devices is the mailbox/command interface. All of the mailbox handling is done in the mailbox-utils library. Longer term, new CXL devices that are being emulated may want to handle commands differently, and therefore would need a mechanism to opt in/out of the specific generic handlers. As such, this is considered sufficient for now, but may need more depth in the future.

---
 hw/cxl/cxl-device-utils.c   | 38 ++++++++++++++++++++++++++++++++++++-
 include/hw/cxl/cxl_device.h | 22 ++++++++++++++++++---
 2 files changed, 56 insertions(+), 4 deletions(-)
hw/cxl/cxl-device-utils.c +37 -1
index 4b995beba7..1f5587ffcd 100644
@@ -131,6 +131,31 @@ static void mailbox_reg_write(void *opaque, hwaddr offset, uint64_t value,
     }
 }
 
+static uint64_t mdev_reg_read(void *opaque, hwaddr offset, unsigned size)
+{
+    uint64_t retval = 0;
+
+    retval = FIELD_DP64(retval, CXL_MEM_DEV_STS, MEDIA_STATUS, 1);
+    retval = FIELD_DP64(retval, CXL_MEM_DEV_STS, MBOX_READY, 1);
+
+    return retval;
+}
+
+static const MemoryRegionOps mdev_ops = {
+    .read = mdev_reg_read,
+    .write = NULL, /* memory device register is read only */
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .valid = {
+        .min_access_size = 1,
+        .max_access_size = 8,
+        .unaligned = false,
+    },
+    .impl = {
+        .min_access_size = 8,
+        .max_access_size = 8,
+    },
+};
+
 static const MemoryRegionOps mailbox_ops = {
     .read = mailbox_reg_read,
     .write = mailbox_reg_write,
@@ -188,6 +213,9 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
                           "device-status", CXL_DEVICE_REGISTERS_LENGTH);
     memory_region_init_io(&cxl_dstate->mailbox, obj, &mailbox_ops, cxl_dstate,
                           "mailbox", CXL_MAILBOX_REGISTERS_LENGTH);
+    memory_region_init_io(&cxl_dstate->memory_device, obj, &mdev_ops,
+                          cxl_dstate, "memory device caps",
+                          CXL_MEMORY_DEVICE_REGISTERS_LENGTH);
 
     memory_region_add_subregion(&cxl_dstate->device_registers, 0,
                                 &cxl_dstate->caps);
@@ -197,6 +225,9 @@ void cxl_device_register_block_init(Object *obj, CXLDeviceState *cxl_dstate)
     memory_region_add_subregion(&cxl_dstate->device_registers,
                                 CXL_MAILBOX_REGISTERS_OFFSET,
                                 &cxl_dstate->mailbox);
+    memory_region_add_subregion(&cxl_dstate->device_registers,
+                                CXL_MEMORY_DEVICE_REGISTERS_OFFSET,
+                                &cxl_dstate->memory_device);
 }
 
 static void device_reg_init_common(CXLDeviceState *cxl_dstate) { }
@@ -209,10 +240,12 @@ static void mailbox_reg_init_common(CXLDeviceState *cxl_dstate)
     cxl_dstate->payload_size = CXL_MAILBOX_MAX_PAYLOAD_SIZE;
 }
 
+static void memdev_reg_init_common(CXLDeviceState *cxl_dstate) { }
+
 void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
 {
     uint64_t *cap_hdrs = cxl_dstate->caps_reg_state64;
-    const int cap_count = 2;
+    const int cap_count = 3;
 
     /* CXL Device Capabilities Array Register */
     ARRAY_FIELD_DP64(cap_hdrs, CXL_DEV_CAP_ARRAY, CAP_ID, 0);
@@ -225,5 +258,8 @@ void cxl_device_register_init_common(CXLDeviceState *cxl_dstate)
     cxl_device_cap_init(cxl_dstate, MAILBOX, 2);
     mailbox_reg_init_common(cxl_dstate);
 
+    cxl_device_cap_init(cxl_dstate, MEMORY_DEVICE, 0x4000);
+    memdev_reg_init_common(cxl_dstate);
+
     assert(cxl_initialize_mailbox(cxl_dstate) == 0);
 }
include/hw/cxl/cxl_device.h +19 -3
index 49dcca7e44..7fd8d0f616 100644
@@ -77,15 +77,21 @@
 #define CXL_MAILBOX_REGISTERS_LENGTH \
     (CXL_MAILBOX_REGISTERS_SIZE + CXL_MAILBOX_MAX_PAYLOAD_SIZE)
 
-#define CXL_MMIO_SIZE                                           \
-    (CXL_DEVICE_CAP_REG_SIZE + CXL_DEVICE_REGISTERS_LENGTH +    \
-     CXL_MAILBOX_REGISTERS_LENGTH)
+
+#define CXL_MEMORY_DEVICE_REGISTERS_OFFSET \
+    (CXL_MAILBOX_REGISTERS_OFFSET + CXL_MAILBOX_REGISTERS_LENGTH)
+#define CXL_MEMORY_DEVICE_REGISTERS_LENGTH 0x8
+
+#define CXL_MMIO_SIZE                                                   \
+    (CXL_DEVICE_CAP_REG_SIZE + CXL_DEVICE_REGISTERS_LENGTH +            \
+     CXL_MAILBOX_REGISTERS_LENGTH + CXL_MEMORY_DEVICE_REGISTERS_LENGTH)
 
 typedef struct cxl_device_state {
     MemoryRegion device_registers;
 
     /* mmio for device capabilities array - 8.2.8.2 */
     MemoryRegion device;
+    MemoryRegion memory_device;
     struct {
         MemoryRegion caps;
         union {
@@ -158,6 +164,9 @@ REG64(CXL_DEV_CAP_ARRAY, 0) /* Documented as 128 bit register but 64 byte access
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(DEVICE, CXL_DEVICE_CAP_HDR1_OFFSET)
 CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MAILBOX, CXL_DEVICE_CAP_HDR1_OFFSET + \
                                                CXL_DEVICE_CAP_REG_SIZE)
+CXL_DEVICE_CAPABILITY_HEADER_REGISTER(MEMORY_DEVICE,
+                                      CXL_DEVICE_CAP_HDR1_OFFSET +
+                                          CXL_DEVICE_CAP_REG_SIZE * 2)
 
 int cxl_initialize_mailbox(CXLDeviceState *cxl_dstate);
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate);
@@ -208,4 +217,11 @@ REG64(CXL_DEV_BG_CMD_STS, 0x18)
 
 REG32(CXL_DEV_CMD_PAYLOAD, 0x20)
 
+REG64(CXL_MEM_DEV_STS, 0)
+    FIELD(CXL_MEM_DEV_STS, FATAL, 0, 1)
+    FIELD(CXL_MEM_DEV_STS, FW_HALT, 1, 1)
+    FIELD(CXL_MEM_DEV_STS, MEDIA_STATUS, 2, 2)
+    FIELD(CXL_MEM_DEV_STS, MBOX_READY, 4, 1)
+    FIELD(CXL_MEM_DEV_STS, RESET_NEEDED, 5, 3)
+
 #endif
Signature · 1 line
2.32.0
#22 lore raw

PATCH v8 8/46 hw/cxl/device: Add cheap EVENTS implementation (8.2.9.1)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

Using the previously implemented stubbed helpers, it is now possible to easily add the missing, required commands to the implementation.

---
 hw/cxl/cxl-mailbox-utils.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)
hw/cxl/cxl-mailbox-utils.c +26 -1
index 7e03dc224a..51af9282e2 100644
@@ -38,6 +38,14 @@
  *  a register interface that already deals with it.
  */
 
+enum {
+    EVENTS      = 0x01,
+        #define GET_RECORDS   0x0
+        #define CLEAR_RECORDS   0x1
+        #define GET_INTERRUPT_POLICY   0x2
+        #define SET_INTERRUPT_POLICY   0x3
+};
+
 /* 8.2.8.4.5.1 Command Return Codes */
 typedef enum {
     CXL_MBOX_SUCCESS = 0x0,
@@ -93,9 +101,26 @@ struct cxl_cmd {
         return CXL_MBOX_SUCCESS;                                          \
     }
 
+DEFINE_MAILBOX_HANDLER_ZEROED(events_get_records, 0x20);
+DEFINE_MAILBOX_HANDLER_NOP(events_clear_records);
+DEFINE_MAILBOX_HANDLER_ZEROED(events_get_interrupt_policy, 4);
+DEFINE_MAILBOX_HANDLER_NOP(events_set_interrupt_policy);
+
 static QemuUUID cel_uuid;
 
-static struct cxl_cmd cxl_cmd_set[256][256] = {};
+#define IMMEDIATE_CONFIG_CHANGE (1 << 1)
+#define IMMEDIATE_LOG_CHANGE (1 << 4)
+
+static struct cxl_cmd cxl_cmd_set[256][256] = {
+    [EVENTS][GET_RECORDS] = { "EVENTS_GET_RECORDS",
+        cmd_events_get_records, 1, 0 },
+    [EVENTS][CLEAR_RECORDS] = { "EVENTS_CLEAR_RECORDS",
+        cmd_events_clear_records, ~0, IMMEDIATE_LOG_CHANGE },
+    [EVENTS][GET_INTERRUPT_POLICY] = { "EVENTS_GET_INTERRUPT_POLICY",
+        cmd_events_get_interrupt_policy, 0, 0 },
+    [EVENTS][SET_INTERRUPT_POLICY] = { "EVENTS_SET_INTERRUPT_POLICY",
+        cmd_events_set_interrupt_policy, 4, IMMEDIATE_CONFIG_CHANGE },
+};
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
 {
Signature · 1 line
2.32.0
#23 lore raw

PATCH v8 9/46 hw/cxl/device: Timestamp implementation (8.2.9.3)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

Errata F4 to CXL 2.0 clarified the meaning of the timer as the sum of the value set with the timestamp set command and the number of nano seconds since it was last set.

---
 hw/cxl/cxl-mailbox-utils.c  | 42 +++++++++++++++++++++++++++++++++++++
 include/hw/cxl/cxl_device.h |  6 ++++++
 2 files changed, 48 insertions(+)
hw/cxl/cxl-mailbox-utils.c +42 -0
index 51af9282e2..05e6bbdd6f 100644
@@ -44,6 +44,9 @@ enum {
         #define CLEAR_RECORDS   0x1
         #define GET_INTERRUPT_POLICY   0x2
         #define SET_INTERRUPT_POLICY   0x3
+    TIMESTAMP   = 0x03,
+        #define GET           0x0
+        #define SET           0x1
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -106,9 +109,46 @@ DEFINE_MAILBOX_HANDLER_NOP(events_clear_records);
 DEFINE_MAILBOX_HANDLER_ZEROED(events_get_interrupt_policy, 4);
 DEFINE_MAILBOX_HANDLER_NOP(events_set_interrupt_policy);
 
+/* 8.2.9.3.1 */
+static ret_code cmd_timestamp_get(struct cxl_cmd *cmd,
+                                  CXLDeviceState *cxl_dstate,
+                                  uint16_t *len)
+{
+    uint64_t time, delta;
+    uint64_t final_time = 0;
+
+    if (cxl_dstate->timestamp.set) {
+        /* First find the delta from the last time the host set the time. */
+        time = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+        delta = time - cxl_dstate->timestamp.last_set;
+        final_time = cxl_dstate->timestamp.host_set + delta;
+    }
+
+    /* Then adjust the actual time */
+    stq_le_p(cmd->payload, final_time);
+    *len = 8;
+
+    return CXL_MBOX_SUCCESS;
+}
+
+/* 8.2.9.3.2 */
+static ret_code cmd_timestamp_set(struct cxl_cmd *cmd,
+                                  CXLDeviceState *cxl_dstate,
+                                  uint16_t *len)
+{
+    cxl_dstate->timestamp.set = true;
+    cxl_dstate->timestamp.last_set = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+
+    cxl_dstate->timestamp.host_set = le64_to_cpu(*(uint64_t *)cmd->payload);
+
+    *len = 0;
+    return CXL_MBOX_SUCCESS;
+}
+
 static QemuUUID cel_uuid;
 
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
+#define IMMEDIATE_POLICY_CHANGE (1 << 3)
 #define IMMEDIATE_LOG_CHANGE (1 << 4)
 
 static struct cxl_cmd cxl_cmd_set[256][256] = {
@@ -120,6 +160,8 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
         cmd_events_get_interrupt_policy, 0, 0 },
     [EVENTS][SET_INTERRUPT_POLICY] = { "EVENTS_SET_INTERRUPT_POLICY",
         cmd_events_set_interrupt_policy, 4, IMMEDIATE_CONFIG_CHANGE },
+    [TIMESTAMP][GET] = { "TIMESTAMP_GET", cmd_timestamp_get, 0, 0 },
+    [TIMESTAMP][SET] = { "TIMESTAMP_SET", cmd_timestamp_set, 8, IMMEDIATE_POLICY_CHANGE },
 };
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
include/hw/cxl/cxl_device.h +6 -0
index 7fd8d0f616..8102d2a813 100644
@@ -117,6 +117,12 @@ typedef struct cxl_device_state {
         size_t cel_size;
     };
 
+    struct {
+        bool set;
+        uint64_t last_set;
+        uint64_t host_set;
+    } timestamp;
+
     /* memory region for persistent memory, HDM */
     uint64_t pmem_size;
 } CXLDeviceState;
Signature · 1 line
2.32.0
#24 lore raw

PATCH v8 10/46 hw/cxl/device: Add log commands (8.2.9.4) + CEL

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

CXL specification provides for the ability to obtain logs from the device. Logs are either spec defined, like the "Command Effects Log" (CEL), or vendor specific. UUIDs are defined for all log types.

The CEL is a mechanism to provide information to the host about which commands are supported. It is useful both to determine which spec'd optional commands are supported, as well as provide a list of vendor specified commands that might be used. The CEL is already created as part of mailbox initialization, but here it is now exported to hosts that use these log commands.

---
 hw/cxl/cxl-mailbox-utils.c | 69 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 69 insertions(+)
hw/cxl/cxl-mailbox-utils.c +69 -0
index 05e6bbdd6f..4e9cb2bccd 100644
@@ -47,6 +47,9 @@ enum {
     TIMESTAMP   = 0x03,
         #define GET           0x0
         #define SET           0x1
+    LOGS        = 0x04,
+        #define GET_SUPPORTED 0x0
+        #define GET_LOG       0x1
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -147,6 +150,70 @@ static ret_code cmd_timestamp_set(struct cxl_cmd *cmd,
 
 static QemuUUID cel_uuid;
 
+/* 8.2.9.4.1 */
+static ret_code cmd_logs_get_supported(struct cxl_cmd *cmd,
+                                       CXLDeviceState *cxl_dstate,
+                                       uint16_t *len)
+{
+    struct {
+        uint16_t entries;
+        uint8_t rsvd[6];
+        struct {
+            QemuUUID uuid;
+            uint32_t size;
+        } log_entries[1];
+    } QEMU_PACKED *supported_logs = (void *)cmd->payload;
+    QEMU_BUILD_BUG_ON(sizeof(*supported_logs) != 0x1c);
+
+    supported_logs->entries = 1;
+    supported_logs->log_entries[0].uuid = cel_uuid;
+    supported_logs->log_entries[0].size = 4 * cxl_dstate->cel_size;
+
+    *len = sizeof(*supported_logs);
+    return CXL_MBOX_SUCCESS;
+}
+
+/* 8.2.9.4.2 */
+static ret_code cmd_logs_get_log(struct cxl_cmd *cmd,
+                                 CXLDeviceState *cxl_dstate,
+                                 uint16_t *len)
+{
+    struct {
+        QemuUUID uuid;
+        uint32_t offset;
+        uint32_t length;
+    } QEMU_PACKED QEMU_ALIGNED(16) *get_log = (void *)cmd->payload;
+
+    /*
+     * 8.2.9.4.2
+     *   The device shall return Invalid Parameter if the Offset or Length
+     *   fields attempt to access beyond the size of the log as reported by Get
+     *   Supported Logs.
+     *
+     * XXX: Spec is wrong, "Invalid Parameter" isn't a thing.
+     * XXX: Spec doesn't address incorrect UUID incorrectness.
+     *
+     * The CEL buffer is large enough to fit all commands in the emulation, so
+     * the only possible failure would be if the mailbox itself isn't big
+     * enough.
+     */
+    if (get_log->offset + get_log->length > cxl_dstate->payload_size) {
+        return CXL_MBOX_INVALID_INPUT;
+    }
+
+    if (!qemu_uuid_is_equal(&get_log->uuid, &cel_uuid)) {
+        return CXL_MBOX_UNSUPPORTED;
+    }
+
+    /* Store off everything to local variables so we can wipe out the payload */
+    *len = get_log->length;
+
+    memmove(cmd->payload, cxl_dstate->cel_log + get_log->offset,
+           get_log->length);
+
+    return CXL_MBOX_SUCCESS;
+}
+
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
 #define IMMEDIATE_POLICY_CHANGE (1 << 3)
 #define IMMEDIATE_LOG_CHANGE (1 << 4)
@@ -162,6 +229,8 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
         cmd_events_set_interrupt_policy, 4, IMMEDIATE_CONFIG_CHANGE },
     [TIMESTAMP][GET] = { "TIMESTAMP_GET", cmd_timestamp_get, 0, 0 },
     [TIMESTAMP][SET] = { "TIMESTAMP_SET", cmd_timestamp_set, 8, IMMEDIATE_POLICY_CHANGE },
+    [LOGS][GET_SUPPORTED] = { "LOGS_GET_SUPPORTED", cmd_logs_get_supported, 0, 0 },
+    [LOGS][GET_LOG] = { "LOGS_GET_LOG", cmd_logs_get_log, 0x18, 0 },
 };
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
Signature · 1 line
2.32.0
#25 lore raw

PATCH v8 11/46 hw/pxb: Use a type for realizing expanders

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

This opens up the possibility for more types of expanders (other than PCI and PCIe). We'll need this to create a CXL expander.

---
 hw/pci-bridge/pci_expander_bridge.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
hw/pci-bridge/pci_expander_bridge.c +7 -4
index de932286b5..d4514227a8 100644
@@ -24,6 +24,8 @@
 #include "hw/boards.h"
 #include "qom/object.h"
 
+enum BusType { PCI, PCIE };
+
 #define TYPE_PXB_BUS "pxb-bus"
 typedef struct PXBBus PXBBus;
 DECLARE_INSTANCE_CHECKER(PXBBus, PXB_BUS,
@@ -221,7 +223,8 @@ static gint pxb_compare(gconstpointer a, gconstpointer b)
            0;
 }
 
-static void pxb_dev_realize_common(PCIDevice *dev, bool pcie, Error **errp)
+static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
+                                   Error **errp)
 {
     PXBDev *pxb = convert_to_pxb(dev);
     DeviceState *ds, *bds = NULL;
@@ -246,7 +249,7 @@ static void pxb_dev_realize_common(PCIDevice *dev, bool pcie, Error **errp)
     }
 
     ds = qdev_new(TYPE_PXB_HOST);
-    if (pcie) {
+    if (type == PCIE) {
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
     } else {
         bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
@@ -295,7 +298,7 @@ static void pxb_dev_realize(PCIDevice *dev, Error **errp)
         return;
     }
 
-    pxb_dev_realize_common(dev, false, errp);
+    pxb_dev_realize_common(dev, PCI, errp);
 }
 
 static void pxb_dev_exitfn(PCIDevice *pci_dev)
@@ -348,7 +351,7 @@ static void pxb_pcie_dev_realize(PCIDevice *dev, Error **errp)
         return;
     }
 
-    pxb_dev_realize_common(dev, true, errp);
+    pxb_dev_realize_common(dev, PCIE, errp);
 }
 
 static void pxb_pcie_dev_class_init(ObjectClass *klass, void *data)
Signature · 1 line
2.32.0
#26 lore raw

PATCH v8 12/46 hw/pci/cxl: Create a CXL bus type

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

The easiest way to differentiate a CXL bus, and a PCIE bus is using a flag. A CXL bus, in hardware, is backward compatible with PCIE, and therefore the code tries pretty hard to keep them in sync as much as possible.

The other way to implement this would be to try to cast the bus to the correct type. This is less code and useful for debugging via simply looking at the flags.

---
 hw/pci-bridge/pci_expander_bridge.c | 9 ++++++++-
 include/hw/pci/pci_bus.h            | 7 +++++++
 2 files changed, 15 insertions(+), 1 deletion(-)
hw/pci-bridge/pci_expander_bridge.c +8 -1
index d4514227a8..a6caa1e7b5 100644
@@ -24,7 +24,7 @@
 #include "hw/boards.h"
 #include "qom/object.h"
 
-enum BusType { PCI, PCIE };
+enum BusType { PCI, PCIE, CXL };
 
 #define TYPE_PXB_BUS "pxb-bus"
 typedef struct PXBBus PXBBus;
@@ -35,6 +35,10 @@ DECLARE_INSTANCE_CHECKER(PXBBus, PXB_BUS,
 DECLARE_INSTANCE_CHECKER(PXBBus, PXB_PCIE_BUS,
                          TYPE_PXB_PCIE_BUS)
 
+#define TYPE_PXB_CXL_BUS "pxb-cxl-bus"
+DECLARE_INSTANCE_CHECKER(PXBBus, PXB_CXL_BUS,
+                         TYPE_PXB_CXL_BUS)
+
 struct PXBBus {
     /*< private >*/
     PCIBus parent_obj;
@@ -251,6 +255,9 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
     ds = qdev_new(TYPE_PXB_HOST);
     if (type == PCIE) {
         bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
+    } else if (type == CXL) {
+        bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_CXL_BUS);
+        bus->flags |= PCI_BUS_CXL;
     } else {
         bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
         bds = qdev_new("pci-bridge");
include/hw/pci/pci_bus.h +7 -0
index 347440d42c..eb94e7e85c 100644
@@ -24,6 +24,8 @@ enum PCIBusFlags {
     PCI_BUS_IS_ROOT                                         = 0x0001,
     /* PCIe extended configuration space is accessible on this bus */
     PCI_BUS_EXTENDED_CONFIG_SPACE                           = 0x0002,
+    /* This is a CXL Type BUS */
+    PCI_BUS_CXL                                             = 0x0004,
 };
 
 struct PCIBus {
@@ -53,6 +55,11 @@ struct PCIBus {
     Notifier machine_done;
 };
 
+static inline bool pci_bus_is_cxl(PCIBus *bus)
+{
+    return !!(bus->flags & PCI_BUS_CXL);
+}
+
 static inline bool pci_bus_is_root(PCIBus *bus)
 {
     return !!(bus->flags & PCI_BUS_IS_ROOT);
Signature · 1 line
2.32.0
#27 lore raw

PATCH v8 13/46 cxl: Machine level control on whether CXL support is enabled

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

There are going to be some potential overheads to CXL enablement, for example the host bridge region reserved in memory maps. Add a machine level control so that CXL is disabled by default.

---
 hw/core/machine.c    | 28 ++++++++++++++++++++++++++++
 hw/i386/pc.c         |  1 +
 include/hw/boards.h  |  2 ++
 include/hw/cxl/cxl.h |  4 ++++
 4 files changed, 35 insertions(+)
hw/core/machine.c +28 -0
index d856485cb4..6ff5dba64e 100644
@@ -31,6 +31,7 @@
 #include "sysemu/qtest.h"
 #include "hw/pci/pci.h"
 #include "hw/mem/nvdimm.h"
+#include "hw/cxl/cxl.h"
 #include "migration/global_state.h"
 #include "migration/vmstate.h"
 #include "exec/confidential-guest-support.h"
@@ -545,6 +546,20 @@ static void machine_set_nvdimm_persistence(Object *obj, const char *value,
     nvdimms_state->persistence_string = g_strdup(value);
 }
 
+static bool machine_get_cxl(Object *obj, Error **errp)
+{
+    MachineState *ms = MACHINE(obj);
+
+    return ms->cxl_devices_state->is_enabled;
+}
+
+static void machine_set_cxl(Object *obj, bool value, Error **errp)
+{
+    MachineState *ms = MACHINE(obj);
+
+    ms->cxl_devices_state->is_enabled = value;
+}
+
 void machine_class_allow_dynamic_sysbus_dev(MachineClass *mc, const char *type)
 {
     QAPI_LIST_PREPEND(mc->allowed_dynamic_sysbus_devices, g_strdup(type));
@@ -777,6 +792,8 @@ static void machine_class_init(ObjectClass *oc, void *data)
     mc->default_ram_size = 128 * MiB;
     mc->rom_file_has_mr = true;
 
+    /* Few machines support CXL, so default to off */
+    mc->cxl_supported = false;
     /* numa node memory size aligned on 8MB by default.
      * On Linux, each node's border has to be 8MB aligned
      */
@@ -922,6 +939,16 @@ static void machine_initfn(Object *obj)
                                         "Valid values are cpu, mem-ctrl");
     }
 
+    if (mc->cxl_supported) {
+        Object *obj = OBJECT(ms);
+
+        ms->cxl_devices_state = g_new0(CXLState, 1);
+        object_property_add_bool(obj, "cxl", machine_get_cxl, machine_set_cxl);
+        object_property_set_description(obj, "cxl",
+                                        "Set on/off to enable/disable "
+                                        "CXL instantiation");
+    }
+
     if (mc->cpu_index_to_instance_props && mc->get_default_cpu_node_id) {
         ms->numa_state = g_new0(NumaState, 1);
         object_property_add_bool(obj, "hmat",
@@ -956,6 +983,7 @@ static void machine_finalize(Object *obj)
     g_free(ms->device_memory);
     g_free(ms->nvdimms_state);
     g_free(ms->numa_state);
+    g_free(ms->cxl_devices_state);
 }
 
 bool machine_usb(MachineState *machine)
hw/i386/pc.c +1 -0
index fd55fc725c..e2849fc741 100644
@@ -1758,6 +1758,7 @@ static void pc_machine_class_init(ObjectClass *oc, void *data)
     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
     mc->nvdimm_supported = true;
     mc->smp_props.dies_supported = true;
+    mc->cxl_supported = true;
     mc->default_ram_id = "pc.ram";
 
     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
include/hw/boards.h +2 -0
index c92ac8815c..680718dafc 100644
@@ -269,6 +269,7 @@ struct MachineClass {
     bool ignore_boot_device_suffixes;
     bool smbus_no_migration_support;
     bool nvdimm_supported;
+    bool cxl_supported;
     bool numa_mem_supported;
     bool auto_enable_numa;
     SMPCompatProps smp_props;
@@ -360,6 +361,7 @@ struct MachineState {
     CPUArchIdList *possible_cpus;
     CpuTopology smp;
     struct NVDIMMState *nvdimms_state;
+    struct CXLState *cxl_devices_state;
     struct NumaState *numa_state;
 };
 
include/hw/cxl/cxl.h +4 -0
index 554ad93b6b..31af92fd5e 100644
@@ -17,4 +17,8 @@
 #define CXL_COMPONENT_REG_BAR_IDX 0
 #define CXL_DEVICE_REG_BAR_IDX 2
 
+typedef struct CXLState {
+    bool is_enabled;
+} CXLState;
+
 #endif
Signature · 1 line
2.32.0
#28 lore raw

PATCH v8 14/46 hw/pxb: Allow creation of a CXL PXB (host bridge)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

This works like adding a typical pxb device, except the name is 'pxb-cxl' instead of 'pxb-pcie'. An example command line would be as follows:
-device pxb-cxl,id=cxl.0,bus="pcie.0",bus_nr=1

A CXL PXB is backward compatible with PCIe. What this means in practice is that an operating system that is unaware of CXL should still be able to enumerate this topology as if it were PCIe.

One can create multiple CXL PXB host bridges, but a host bridge can only be connected to the main root bus. Host bridges cannot appear elsewhere in the topology.

Note that as of this patch, the ACPI tables needed for the host bridge (specifically, an ACPI object in _SB named ACPI0016 and the CEDT) aren't created. So while this patch internally creates it, it cannot be properly used by an operating system or other system software.

Also necessary is to add an exception to scripts/device-crash-test similar to that for exiting pxb as both must created on a PCIexpress host bus.

---
 hw/pci-bridge/pci_expander_bridge.c | 86 ++++++++++++++++++++++++++++-
 hw/pci/pci.c                        |  7 +++
 include/hw/pci/pci.h                |  6 ++
 scripts/device-crash-test           |  1 +
 4 files changed, 98 insertions(+), 2 deletions(-)
hw/pci-bridge/pci_expander_bridge.c +84 -2
index a6caa1e7b5..f762eb4a6e 100644
@@ -17,6 +17,7 @@
 #include "hw/pci/pci_host.h"
 #include "hw/qdev-properties.h"
 #include "hw/pci/pci_bridge.h"
+#include "hw/cxl/cxl.h"
 #include "qemu/range.h"
 #include "qemu/error-report.h"
 #include "qemu/module.h"
@@ -56,6 +57,16 @@ DECLARE_INSTANCE_CHECKER(PXBDev, PXB_DEV,
 DECLARE_INSTANCE_CHECKER(PXBDev, PXB_PCIE_DEV,
                          TYPE_PXB_PCIE_DEVICE)
 
+#define TYPE_PXB_CXL_DEVICE "pxb-cxl"
+DECLARE_INSTANCE_CHECKER(PXBDev, PXB_CXL_DEV,
+                         TYPE_PXB_CXL_DEVICE)
+
+typedef struct CXLHost {
+    PCIHostState parent_obj;
+
+    CXLComponentState cxl_cstate;
+} CXLHost;
+
 struct PXBDev {
     /*< private >*/
     PCIDevice parent_obj;
@@ -68,6 +79,11 @@ struct PXBDev {
 
 static PXBDev *convert_to_pxb(PCIDevice *dev)
 {
+    /* A CXL PXB's parent bus is PCIe, so the normal check won't work */
+    if (object_dynamic_cast(OBJECT(dev), TYPE_PXB_CXL_DEVICE)) {
+        return PXB_CXL_DEV(dev);
+    }
+
     return pci_bus_is_express(pci_get_bus(dev))
         ? PXB_PCIE_DEV(dev) : PXB_DEV(dev);
 }
@@ -112,11 +128,20 @@ static const TypeInfo pxb_pcie_bus_info = {
     .class_init    = pxb_bus_class_init,
 };
 
+static const TypeInfo pxb_cxl_bus_info = {
+    .name          = TYPE_PXB_CXL_BUS,
+    .parent        = TYPE_CXL_BUS,
+    .instance_size = sizeof(PXBBus),
+    .class_init    = pxb_bus_class_init,
+};
+
 static const char *pxb_host_root_bus_path(PCIHostState *host_bridge,
                                           PCIBus *rootbus)
 {
-    PXBBus *bus = pci_bus_is_express(rootbus) ?
-                  PXB_PCIE_BUS(rootbus) : PXB_BUS(rootbus);
+    PXBBus *bus = pci_bus_is_cxl(rootbus) ?
+                      PXB_CXL_BUS(rootbus) :
+                      pci_bus_is_express(rootbus) ? PXB_PCIE_BUS(rootbus) :
+                                                    PXB_BUS(rootbus);
 
     snprintf(bus->bus_path, 8, "0000:%02x", pxb_bus_num(rootbus));
     return bus->bus_path;
@@ -218,6 +243,10 @@ static int pxb_map_irq_fn(PCIDevice *pci_dev, int pin)
     return pin - PCI_SLOT(pxb->devfn);
 }
 
+static void pxb_dev_reset(DeviceState *dev)
+{
+}
+
 static gint pxb_compare(gconstpointer a, gconstpointer b)
 {
     const PXBDev *pxb_a = a, *pxb_b = b;
@@ -389,13 +418,66 @@ static const TypeInfo pxb_pcie_dev_info = {
     },
 };
 
+static void pxb_cxl_dev_realize(PCIDevice *dev, Error **errp)
+{
+    MachineState *ms = MACHINE(qdev_get_machine());
+
+    /* A CXL PXB's parent bus is still PCIe */
+    if (!pci_bus_is_express(pci_get_bus(dev))) {
+        error_setg(errp, "pxb-cxl devices cannot reside on a PCI bus");
+        return;
+    }
+    if (!ms->cxl_devices_state->is_enabled) {
+        error_setg(errp, "Machine does not have cxl=on");
+        return;
+    }
+
+    pxb_dev_realize_common(dev, CXL, errp);
+    pxb_dev_reset(DEVICE(dev));
+}
+
+static void pxb_cxl_dev_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc   = DEVICE_CLASS(klass);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+
+    k->realize             = pxb_cxl_dev_realize;
+    k->exit                = pxb_dev_exitfn;
+    /*
+     * XXX: These types of bridges don't actually show up in the hierarchy so
+     * vendor, device, class, etc. ids are intentionally left out.
+     */
+
+    dc->desc = "CXL Host Bridge";
+    device_class_set_props(dc, pxb_dev_properties);
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+
+    /* Host bridges aren't hotpluggable. FIXME: spec reference */
+    dc->hotpluggable = false;
+    dc->reset = pxb_dev_reset;
+}
+
+static const TypeInfo pxb_cxl_dev_info = {
+    .name          = TYPE_PXB_CXL_DEVICE,
+    .parent        = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(PXBDev),
+    .class_init    = pxb_cxl_dev_class_init,
+    .interfaces =
+        (InterfaceInfo[]){
+            { INTERFACE_CONVENTIONAL_PCI_DEVICE },
+            {},
+        },
+};
+
 static void pxb_register_types(void)
 {
     type_register_static(&pxb_bus_info);
     type_register_static(&pxb_pcie_bus_info);
+    type_register_static(&pxb_cxl_bus_info);
     type_register_static(&pxb_host_info);
     type_register_static(&pxb_dev_info);
     type_register_static(&pxb_pcie_dev_info);
+    type_register_static(&pxb_cxl_dev_info);
 }
 
 type_init(pxb_register_types)
hw/pci/pci.c +7 -0
index 7883778a99..3b738d6f1f 100644
@@ -229,6 +229,12 @@ static const TypeInfo pcie_bus_info = {
     .class_init = pcie_bus_class_init,
 };
 
+static const TypeInfo cxl_bus_info = {
+    .name       = TYPE_CXL_BUS,
+    .parent     = TYPE_PCIE_BUS,
+    .class_init = pcie_bus_class_init,
+};
+
 static PCIBus *pci_find_bus_nr(PCIBus *bus, int bus_num);
 static void pci_update_mappings(PCIDevice *d);
 static void pci_irq_handler(void *opaque, int irq_num, int level);
@@ -2946,6 +2952,7 @@ static void pci_register_types(void)
 {
     type_register_static(&pci_bus_info);
     type_register_static(&pcie_bus_info);
+    type_register_static(&cxl_bus_info);
     type_register_static(&conventional_pci_interface_info);
     type_register_static(&cxl_interface_info);
     type_register_static(&pcie_interface_info);
include/hw/pci/pci.h +6 -0
index 98f0d1b844..44dacfa224 100644
@@ -408,6 +408,7 @@ typedef PCIINTxRoute (*pci_route_irq_fn)(void *opaque, int pin);
 #define TYPE_PCI_BUS "PCI"
 OBJECT_DECLARE_TYPE(PCIBus, PCIBusClass, PCI_BUS)
 #define TYPE_PCIE_BUS "PCIE"
+#define TYPE_CXL_BUS "CXL"
 
 typedef void (*pci_bus_dev_fn)(PCIBus *b, PCIDevice *d, void *opaque);
 typedef void (*pci_bus_fn)(PCIBus *b, void *opaque);
@@ -770,6 +771,11 @@ static inline void pci_irq_pulse(PCIDevice *pci_dev)
     pci_irq_deassert(pci_dev);
 }
 
+static inline int pci_is_cxl(const PCIDevice *d)
+{
+    return d->cap_present & QEMU_PCIE_CAP_CXL;
+}
+
 static inline int pci_is_express(const PCIDevice *d)
 {
     return d->cap_present & QEMU_PCI_CAP_EXPRESS;
scripts/device-crash-test +1 -0
index 7fbd99158b..52bd3d8f71 100755
@@ -93,6 +93,7 @@ ERROR_RULE_LIST = [
     {'device':'pci-bridge', 'expected':True},              # Bridge chassis not specified. Each bridge is required to be assigned a unique chassis id > 0.
     {'device':'pci-bridge-seat', 'expected':True},         # Bridge chassis not specified. Each bridge is required to be assigned a unique chassis id > 0.
     {'device':'pxb', 'expected':True},                     # Bridge chassis not specified. Each bridge is required to be assigned a unique chassis id > 0.
+    {'device':'pxb-cxl', 'expected':True},                 # pxb-cxl devices cannot reside on a PCI bus.
     {'device':'scsi-block', 'expected':True},              # drive property not set
     {'device':'scsi-generic', 'expected':True},            # drive property not set
     {'device':'scsi-hd', 'expected':True},                 # drive property not set
Signature · 1 line
2.32.0
#29 lore raw

PATCH v8 15/46 qtest/cxl: Introduce initial test for pxb-cxl only.

In reply to Jonathan Cameron

Initial test with just pxb-cxl. Other tests will be added alongside functionality.

---
 tests/qtest/cxl-test.c  | 23 +++++++++++++++++++++++
 tests/qtest/meson.build |  4 ++++
 2 files changed, 27 insertions(+)
tests/qtest/cxl-test.c new file +23 -0
new file mode 100644
index 0000000000..1006c8ae4e
@@ -0,0 +1,23 @@
+/*
+ * QTest testcase for CXL
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+#include "qemu/osdep.h"
+#include "libqtest-single.h"
+
+
+static void cxl_basic_pxb(void)
+{
+    qtest_start("-machine q35,cxl=on -device pxb-cxl,bus=pcie.0");
+    qtest_end();
+}
+
+int main(int argc, char **argv)
+{
+    g_test_init(&argc, &argv, NULL);
+    qtest_add_func("/pci/cxl/basic_pxb", cxl_basic_pxb);
+    return g_test_run();
+}
tests/qtest/meson.build +4 -0
index d25f82bb5a..6e1ad4dc9a 100644
@@ -41,6 +41,9 @@ qtests_pci = \
   (config_all_devices.has_key('CONFIG_VGA') ? ['display-vga-test'] : []) +                  \
   (config_all_devices.has_key('CONFIG_IVSHMEM_DEVICE') ? ['ivshmem-test'] : [])
 
+qtests_cxl = \
+  (config_all_devices.has_key('CONFIG_CXL') ? ['cxl-test'] : [])
+
 qtests_i386 = \
   (slirp.found() ? ['pxe-test', 'test-netfilter'] : []) +             \
   (config_host.has_key('CONFIG_POSIX') ? ['test-filter-mirror'] : []) +                     \
@@ -75,6 +78,7 @@ qtests_i386 = \
    slirp.found() ? ['virtio-net-failover'] : []) +                                          \
   (unpack_edk2_blobs ? ['bios-tables-test'] : []) +                                         \
   qtests_pci +                                                                              \
+  qtests_cxl +                                                                              \
   ['fdc-test',
    'ide-test',
    'hd-geo-test',
Signature · 1 line
2.32.0
#30 lore raw

PATCH v8 16/46 hw/cxl/rp: Add a root port

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

This adds just enough of a root port implementation to be able to enumerate root ports (creating the required DVSEC entries). What's not here yet is the MMIO nor the ability to write some of the DVSEC entries.

This can be added with the qemu commandline by adding a rootport to a specific CXL host bridge. For example:
-device cxl-rp,id=rp0,bus="cxl.0",addr=0.0,chassis=4

Like the host bridge patch, the ACPI tables aren't generated at this point and so system software cannot use it.

---
 hw/pci-bridge/Kconfig          |   5 +
 hw/pci-bridge/cxl_root_port.c  | 231 +++++++++++++++++++++++++++++++++
 hw/pci-bridge/meson.build      |   1 +
 hw/pci-bridge/pcie_root_port.c |   6 +-
 hw/pci/pci.c                   |   4 +-
 5 files changed, 245 insertions(+), 2 deletions(-)
hw/pci-bridge/Kconfig +5 -0
index f8df4315ba..02614f49aa 100644
@@ -27,3 +27,8 @@ config DEC_PCI
 
 config SIMBA
     bool
+
+config CXL
+    bool
+    default y if PCI_EXPRESS && PXB
+    depends on PCI_EXPRESS && MSI_NONBROKEN && PXB
hw/pci-bridge/cxl_root_port.c new file +231 -0

231 lines not on this page · raw message

hw/pci-bridge/meson.build +1 -0
index daab8acf2a..b6d26a03d5 100644
@@ -5,6 +5,7 @@ pci_ss.add(when: 'CONFIG_IOH3420', if_true: files('ioh3420.c'))
 pci_ss.add(when: 'CONFIG_PCIE_PORT', if_true: files('pcie_root_port.c', 'gen_pcie_root_port.c', 'pcie_pci_bridge.c'))
 pci_ss.add(when: 'CONFIG_PXB', if_true: files('pci_expander_bridge.c'))
 pci_ss.add(when: 'CONFIG_XIO3130', if_true: files('xio3130_upstream.c', 'xio3130_downstream.c'))
+pci_ss.add(when: 'CONFIG_CXL', if_true: files('cxl_root_port.c'))
 
 # NewWorld PowerMac
 pci_ss.add(when: 'CONFIG_DEC_PCI', if_true: files('dec.c'))
hw/pci-bridge/pcie_root_port.c +5 -1
index f1cfe9d14a..460e48269d 100644
@@ -67,7 +67,11 @@ static void rp_realize(PCIDevice *d, Error **errp)
     int rc;
 
     pci_config_set_interrupt_pin(d->config, 1);
-    pci_bridge_initfn(d, TYPE_PCIE_BUS);
+    if (d->cap_present & QEMU_PCIE_CAP_CXL) {
+        pci_bridge_initfn(d, TYPE_CXL_BUS);
+    } else {
+        pci_bridge_initfn(d, TYPE_PCIE_BUS);
+    }
     pcie_port_init_reg(d);
 
     rc = pci_bridge_ssvid_init(d, rpc->ssvid_offset, dc->vendor_id,
hw/pci/pci.c +3 -1
index 3b738d6f1f..54b9ad9c28 100644
@@ -2762,7 +2762,9 @@ static void pci_device_class_base_init(ObjectClass *klass, void *data)
             object_class_dynamic_cast(klass, INTERFACE_CONVENTIONAL_PCI_DEVICE);
         ObjectClass *pcie =
             object_class_dynamic_cast(klass, INTERFACE_PCIE_DEVICE);
-        assert(conventional || pcie);
+        ObjectClass *cxl =
+            object_class_dynamic_cast(klass, INTERFACE_CXL_DEVICE);
+        assert(conventional || pcie || cxl);
     }
 }
 
Signature · 1 line
2.32.0
#31 lore raw

PATCH v8 17/46 hw/cxl/device: Add a memory device (8.2.8.5)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

A CXL memory device (AKA Type 3) is a CXL component that contains some combination of volatile and persistent memory. It also implements the previously defined mailbox interface as well as the memory device firmware interface.

Although the memory device is configured like a normal PCIe device, the memory traffic is on an entirely separate bus conceptually (using the same physical wires as PCIe, but different protocol).

Once the CXL topology is fully configure and address decoders committed, the guest physical address for the memory device is part of a larger window which is owned by the platform. The creation of these windows is later in this series.

The following example will create a 256M device in a 512M window:
-object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M" -device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0"

Note: Dropped PCDIMM info interfaces for now. They can be added if appropriate at a later date.

---
 hw/cxl/cxl-mailbox-utils.c  |  46 +++++++++++
 hw/mem/Kconfig              |   5 ++
 hw/mem/cxl_type3.c          | 153 ++++++++++++++++++++++++++++++++++++
 hw/mem/meson.build          |   1 +
 include/hw/cxl/cxl_device.h |  17 ++++
 include/hw/cxl/cxl_pci.h    |  21 +++++
 include/hw/pci/pci_ids.h    |   1 +
 7 files changed, 244 insertions(+)
hw/cxl/cxl-mailbox-utils.c +46 -0
index 4e9cb2bccd..7ed9606c6b 100644
@@ -50,6 +50,8 @@ enum {
     LOGS        = 0x04,
         #define GET_SUPPORTED 0x0
         #define GET_LOG       0x1
+    IDENTIFY    = 0x40,
+        #define MEMORY_DEVICE 0x0
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -214,6 +216,48 @@ static ret_code cmd_logs_get_log(struct cxl_cmd *cmd,
     return CXL_MBOX_SUCCESS;
 }
 
+/* 8.2.9.5.1.1 */
+static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
+                                           CXLDeviceState *cxl_dstate,
+                                           uint16_t *len)
+{
+    struct {
+        char fw_revision[0x10];
+        uint64_t total_capacity;
+        uint64_t volatile_capacity;
+        uint64_t persistent_capacity;
+        uint64_t partition_align;
+        uint16_t info_event_log_size;
+        uint16_t warning_event_log_size;
+        uint16_t failure_event_log_size;
+        uint16_t fatal_event_log_size;
+        uint32_t lsa_size;
+        uint8_t poison_list_max_mer[3];
+        uint16_t inject_poison_limit;
+        uint8_t poison_caps;
+        uint8_t qos_telemetry_caps;
+    } QEMU_PACKED *id;
+    QEMU_BUILD_BUG_ON(sizeof(*id) != 0x43);
+
+    uint64_t size = cxl_dstate->pmem_size;
+
+    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    id = (void *)cmd->payload;
+    memset(id, 0, sizeof(*id));
+
+    /* PMEM only */
+    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
+
+    id->total_capacity = size / (256 << 20);
+    id->persistent_capacity = size / (256 << 20);
+
+    *len = sizeof(*id);
+    return CXL_MBOX_SUCCESS;
+}
+
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
 #define IMMEDIATE_POLICY_CHANGE (1 << 3)
 #define IMMEDIATE_LOG_CHANGE (1 << 4)
@@ -231,6 +275,8 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
     [TIMESTAMP][SET] = { "TIMESTAMP_SET", cmd_timestamp_set, 8, IMMEDIATE_POLICY_CHANGE },
     [LOGS][GET_SUPPORTED] = { "LOGS_GET_SUPPORTED", cmd_logs_get_supported, 0, 0 },
     [LOGS][GET_LOG] = { "LOGS_GET_LOG", cmd_logs_get_log, 0x18, 0 },
+    [IDENTIFY][MEMORY_DEVICE] = { "IDENTIFY_MEMORY_DEVICE",
+        cmd_identify_memory_device, 0, 0 },
 };
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
hw/mem/Kconfig +5 -0
index 03dbb3c7df..73c5ae8ad9 100644
@@ -11,3 +11,8 @@ config NVDIMM
 
 config SPARSE_MEM
     bool
+
+config CXL_MEM_DEVICE
+    bool
+    default y if CXL
+    select MEM_DEVICE
hw/mem/cxl_type3.c new file +153 -0

153 lines not on this page · raw message

hw/mem/meson.build +1 -0
index 82f86d117e..609b2b36fc 100644
@@ -3,6 +3,7 @@ mem_ss.add(files('memory-device.c'))
 mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
 mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
 mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
+mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
 
 softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
 
include/hw/cxl/cxl_device.h +17 -0
index 8102d2a813..72da811c52 100644
@@ -230,4 +230,21 @@ REG64(CXL_MEM_DEV_STS, 0)
     FIELD(CXL_MEM_DEV_STS, MBOX_READY, 4, 1)
     FIELD(CXL_MEM_DEV_STS, RESET_NEEDED, 5, 3)
 
+typedef struct cxl_type3_dev {
+    /* Private */
+    PCIDevice parent_obj;
+
+    /* Properties */
+    uint64_t size;
+    HostMemoryBackend *hostmem;
+
+    /* State */
+    CXLComponentState cxl_cstate;
+    CXLDeviceState cxl_dstate;
+} CXLType3Dev;
+
+#define TYPE_CXL_TYPE3_DEV "cxl-type3"
+
+#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
+
 #endif
include/hw/cxl/cxl_pci.h +21 -0
index 810a244fab..cf53fe5425 100644
@@ -64,6 +64,27 @@ QEMU_BUILD_BUG_ON(sizeof(struct dvsec_header) != 10);
  * CXL 2.0 Downstream Port: 3, 4, 7, 8
  */
 
+/* CXL 2.0 - 8.1.3 (ID 0001) */
+struct cxl_dvsec_device {
+    struct dvsec_header hdr;
+    uint16_t cap;
+    uint16_t ctrl;
+    uint16_t status;
+    uint16_t ctrl2;
+    uint16_t status2;
+    uint16_t lock;
+    uint16_t cap2;
+    uint32_t range1_size_hi;
+    uint32_t range1_size_lo;
+    uint32_t range1_base_hi;
+    uint32_t range1_base_lo;
+    uint32_t range2_size_hi;
+    uint32_t range2_size_lo;
+    uint32_t range2_base_hi;
+    uint32_t range2_base_lo;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_device) != 0x38);
+
 /* CXL 2.0 - 8.1.5 (ID 0003) */
 struct cxl_dvsec_port_extensions {
     struct dvsec_header hdr;
include/hw/pci/pci_ids.h +1 -0
index 11abe22d46..898083b86f 100644
@@ -53,6 +53,7 @@
 #define PCI_BASE_CLASS_MEMORY            0x05
 #define PCI_CLASS_MEMORY_RAM             0x0500
 #define PCI_CLASS_MEMORY_FLASH           0x0501
+#define PCI_CLASS_MEMORY_CXL             0x0502
 #define PCI_CLASS_MEMORY_OTHER           0x0580
 
 #define PCI_BASE_CLASS_BRIDGE            0x06
Signature · 1 line
2.32.0
#32 lore raw

Re: [PATCH v8 17/46] hw/cxl/device: Add a memory device (8.2.8.5)

Quoted 145 lines · Jonathan Cameron via

From: Ben Widawsky <[email protected]>

A CXL memory device (AKA Type 3) is a CXL component that contains some combination of volatile and persistent memory. It also implements the previously defined mailbox interface as well as the memory device firmware interface.

Although the memory device is configured like a normal PCIe device, the memory traffic is on an entirely separate bus conceptually (using the same physical wires as PCIe, but different protocol).

Once the CXL topology is fully configure and address decoders committed, the guest physical address for the memory device is part of a larger window which is owned by the platform. The creation of these windows is later in this series.

The following example will create a 256M device in a 512M window:

(quoted diff) +0 -2
-object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
-device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0"

Note: Dropped PCDIMM info interfaces for now. They can be added if appropriate at a later date.

---
  hw/cxl/cxl-mailbox-utils.c  |  46 +++++++++++
  hw/mem/Kconfig              |   5 ++
  hw/mem/cxl_type3.c          | 153 ++++++++++++++++++++++++++++++++++++
  hw/mem/meson.build          |   1 +
  include/hw/cxl/cxl_device.h |  17 ++++
  include/hw/cxl/cxl_pci.h    |  21 +++++
  include/hw/pci/pci_ids.h    |   1 +
  7 files changed, 244 insertions(+)
hw/cxl/cxl-mailbox-utils.c +46 -0
index 4e9cb2bccd..7ed9606c6b 100644
@@ -50,6 +50,8 @@ enum {
      LOGS        = 0x04,
          #define GET_SUPPORTED 0x0
          #define GET_LOG       0x1
+    IDENTIFY    = 0x40,
+        #define MEMORY_DEVICE 0x0
  };
 
  /* 8.2.8.4.5.1 Command Return Codes */
@@ -214,6 +216,48 @@ static ret_code cmd_logs_get_log(struct cxl_cmd *cmd,
      return CXL_MBOX_SUCCESS;
  }
 
+/* 8.2.9.5.1.1 */
+static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
+                                           CXLDeviceState *cxl_dstate,
+                                           uint16_t *len)
+{
+    struct {
+        char fw_revision[0x10];
+        uint64_t total_capacity;
+        uint64_t volatile_capacity;
+        uint64_t persistent_capacity;
+        uint64_t partition_align;
+        uint16_t info_event_log_size;
+        uint16_t warning_event_log_size;
+        uint16_t failure_event_log_size;
+        uint16_t fatal_event_log_size;
+        uint32_t lsa_size;
+        uint8_t poison_list_max_mer[3];
+        uint16_t inject_poison_limit;
+        uint8_t poison_caps;
+        uint8_t qos_telemetry_caps;
+    } QEMU_PACKED *id;
+    QEMU_BUILD_BUG_ON(sizeof(*id) != 0x43);
+
+    uint64_t size = cxl_dstate->pmem_size;
+
+    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    id = (void *)cmd->payload;
+    memset(id, 0, sizeof(*id));
+
+    /* PMEM only */
+    snprintf(id->fw_revision, 0x10, "BWFW VERSION %02d", 0);
+
+    id->total_capacity = size / (256 << 20);
+    id->persistent_capacity = size / (256 << 20);
+
+    *len = sizeof(*id);
+    return CXL_MBOX_SUCCESS;
+}
+
  #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
  #define IMMEDIATE_POLICY_CHANGE (1 << 3)
  #define IMMEDIATE_LOG_CHANGE (1 << 4)
@@ -231,6 +275,8 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
      [TIMESTAMP][SET] = { "TIMESTAMP_SET", cmd_timestamp_set, 8, IMMEDIATE_POLICY_CHANGE },
      [LOGS][GET_SUPPORTED] = { "LOGS_GET_SUPPORTED", cmd_logs_get_supported, 0, 0 },
      [LOGS][GET_LOG] = { "LOGS_GET_LOG", cmd_logs_get_log, 0x18, 0 },
+    [IDENTIFY][MEMORY_DEVICE] = { "IDENTIFY_MEMORY_DEVICE",
+        cmd_identify_memory_device, 0, 0 },
  };
 
  void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
hw/mem/Kconfig +5 -0
index 03dbb3c7df..73c5ae8ad9 100644
@@ -11,3 +11,8 @@ config NVDIMM
 
  config SPARSE_MEM
      bool
+
+config CXL_MEM_DEVICE
+    bool
+    default y if CXL
+    select MEM_DEVICE
hw/mem/cxl_type3.c new file +20 -0
new file mode 100644
index 0000000000..a8d7cfcc81
@@ -0,0 +1,153 @@
+#include "qemu/osdep.h"
+#include "qemu/units.h"
+#include "qemu/error-report.h"
+#include "hw/mem/memory-device.h"
+#include "hw/mem/pc-dimm.h"
+#include "hw/pci/pci.h"
+#include "hw/qdev-properties.h"
+#include "qapi/error.h"
+#include "qemu/log.h"
+#include "qemu/module.h"
+#include "qemu/range.h"
+#include "qemu/rcu.h"
+#include "sysemu/hostmem.h"
+#include "hw/cxl/cxl.h"
+
+static void build_dvsecs(CXLType3Dev *ct3d)
+{
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    uint8_t *dvsec;
+

hw/mem/cxl_type3.c line 30

Quoted 10 lines
hw/mem/cxl_type3.c new file +10 -0
+    dvsec = (uint8_t *)&(struct cxl_dvsec_device){
+        .cap = 0x1e,
+        .ctrl = 0x6,
+        .status2 = 0x2,
+        .range1_size_hi = 0,
+#ifdef SET_PMEM_PADDR
+        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
+#else
+        .range1_size_lo = 0x3,
+#endif

Generally you don't want #ifdefs like this to control functionality: it should either be removed if it is for unimplemented features, or controlled via a qdev property in ct3_props below.

Quoted 150 lines
hw/mem/cxl_type3.c new file +123 -0
+        .range1_base_hi = 0,
+        .range1_base_lo = 0,
+    };
+    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
+                               PCIE_CXL_DEVICE_DVSEC,
+                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_register_locator){
+        .rsvd         = 0,
+        .reg0_base_lo = RBI_COMPONENT_REG | CXL_COMPONENT_REG_BAR_IDX,
+        .reg0_base_hi = 0,
+        .reg1_base_lo = RBI_CXL_DEVICE_REG | CXL_DEVICE_REG_BAR_IDX,
+        .reg1_base_hi = 0,
+    };
+    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
+                               REG_LOC_DVSEC_REVID, dvsec);
+}
+
+static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
+{
+    MemoryRegion *mr;
+
+    if (!ct3d->hostmem) {
+        error_setg(errp, "memdev property must be set");
+        return;
+    }
+
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        error_setg(errp, "memdev property must be set");
+        return;
+    }
+    memory_region_set_nonvolatile(mr, true);
+    memory_region_set_enabled(mr, true);
+    host_memory_backend_set_mapped(ct3d->hostmem, true);
+    ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
+}
+
+
+static void ct3_realize(PCIDevice *pci_dev, Error **errp)
+{
+    CXLType3Dev *ct3d = CT3(pci_dev);
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    ComponentRegisters *regs = &cxl_cstate->crb;
+    MemoryRegion *mr = &regs->component_registers;
+    uint8_t *pci_conf = pci_dev->config;
+
+    cxl_setup_memory(ct3d, errp);
+
+    pci_config_set_prog_interface(pci_conf, 0x10);
+    pci_config_set_class(pci_conf, PCI_CLASS_MEMORY_CXL);
+
+    pcie_endpoint_cap_init(pci_dev, 0x80);
+    cxl_cstate->dvsec_offset = 0x100;
+
+    ct3d->cxl_cstate.pdev = pci_dev;
+    build_dvsecs(ct3d);
+
+    cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
+                                      TYPE_CXL_TYPE3_DEV);
+
+    pci_register_bar(
+        pci_dev, CXL_COMPONENT_REG_BAR_IDX,
+        PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64, mr);
+
+    cxl_device_register_block_init(OBJECT(pci_dev), &ct3d->cxl_dstate);
+    pci_register_bar(pci_dev, CXL_DEVICE_REG_BAR_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY |
+                         PCI_BASE_ADDRESS_MEM_TYPE_64,
+                     &ct3d->cxl_dstate.device_registers);
+}
+
+static void ct3d_reset(DeviceState *dev)
+{
+    CXLType3Dev *ct3d = CT3(dev);
+    uint32_t *reg_state = ct3d->cxl_cstate.crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_TYPE3_DEVICE);
+    cxl_device_register_init_common(&ct3d->cxl_dstate);
+}
+
+static Property ct3_props[] = {
+    DEFINE_PROP_SIZE("size", CXLType3Dev, size, -1),
+    DEFINE_PROP_LINK("memdev", CXLType3Dev, hostmem, TYPE_MEMORY_BACKEND,
+                     HostMemoryBackend *),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void ct3_class_init(ObjectClass *oc, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(oc);
+    PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
+
+    pc->realize = ct3_realize;
+    pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
+    pc->vendor_id = PCI_VENDOR_ID_INTEL;
+    pc->device_id = 0xd93; /* LVF for now */
+    pc->revision = 1;
+
+    set_bit(DEVICE_CATEGORY_STORAGE, dc->categories);
+    dc->desc = "CXL PMEM Device (Type 3)";
+    dc->reset = ct3d_reset;
+    device_class_set_props(dc, ct3_props);
+}
+
+static const TypeInfo ct3d_info = {
+    .name = TYPE_CXL_TYPE3_DEV,
+    .parent = TYPE_PCI_DEVICE,
+    .class_init = ct3_class_init,
+    .instance_size = sizeof(CXLType3Dev),
+    .interfaces = (InterfaceInfo[]) {
+        { INTERFACE_CXL_DEVICE },
+        { INTERFACE_PCIE_DEVICE },
+        {}
+    },
+};
+
+static void ct3d_registers(void)
+{
+    type_register_static(&ct3d_info);
+}
+
+type_init(ct3d_registers);
hw/mem/meson.build +1 -0
index 82f86d117e..609b2b36fc 100644
@@ -3,6 +3,7 @@ mem_ss.add(files('memory-device.c'))
  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
+mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
 
  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
 
include/hw/cxl/cxl_device.h +7 -0
index 8102d2a813..72da811c52 100644
@@ -230,4 +230,21 @@ REG64(CXL_MEM_DEV_STS, 0)
      FIELD(CXL_MEM_DEV_STS, MBOX_READY, 4, 1)
      FIELD(CXL_MEM_DEV_STS, RESET_NEEDED, 5, 3)
 
+typedef struct cxl_type3_dev {
+    /* Private */
+    PCIDevice parent_obj;
+
+    /* Properties */
+    uint64_t size;
+    HostMemoryBackend *hostmem;

include/hw/cxl/cxl_device.h REG64(CXL_MEM_DEV_STS, 0) · line 249

Quoted 10 lines
include/hw/cxl/cxl_device.h +10 -0
+
+    /* State */
+    CXLComponentState cxl_cstate;
+    CXLDeviceState cxl_dstate;
+} CXLType3Dev;
+
+#define TYPE_CXL_TYPE3_DEV "cxl-type3"
+
+#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
+

Ah okay this is an old style initialiser, and new code shouldn't be using them anymore (I believe there should be no remaining instances in git master). Drop the typedef from struct cxl_type3_dev and replace with:

#define TYPE_CXL_TYPE3_DEV "cxl-type3" OBJECT_DECLARE_SIMPLE_TYPE(CXLType3Dev, CXL_TYPE3_DEV))

Note this will alter the generated QOM CAST from CT3() to CXL_TYPE3_DEV(): I would argue that the _DEV suffix is QOM legacy naming and recommend removing it from both of the above to give you TYPE_CXL_TYPE3 and CXL_TYPE3 respectively.

Quoted 38 lines
include/hw/cxl/cxl_device.h +0 -0
  #endif
include/hw/cxl/cxl_pci.h +21 -0
index 810a244fab..cf53fe5425 100644
@@ -64,6 +64,27 @@ QEMU_BUILD_BUG_ON(sizeof(struct dvsec_header) != 10);
   * CXL 2.0 Downstream Port: 3, 4, 7, 8
   */
 
+/* CXL 2.0 - 8.1.3 (ID 0001) */
+struct cxl_dvsec_device {
+    struct dvsec_header hdr;
+    uint16_t cap;
+    uint16_t ctrl;
+    uint16_t status;
+    uint16_t ctrl2;
+    uint16_t status2;
+    uint16_t lock;
+    uint16_t cap2;
+    uint32_t range1_size_hi;
+    uint32_t range1_size_lo;
+    uint32_t range1_base_hi;
+    uint32_t range1_base_lo;
+    uint32_t range2_size_hi;
+    uint32_t range2_size_lo;
+    uint32_t range2_base_hi;
+    uint32_t range2_base_lo;
+};
+QEMU_BUILD_BUG_ON(sizeof(struct cxl_dvsec_device) != 0x38);
+
  /* CXL 2.0 - 8.1.5 (ID 0003) */
  struct cxl_dvsec_port_extensions {
      struct dvsec_header hdr;
include/hw/pci/pci_ids.h +0 -0
index 11abe22d46..898083b86f 100644

include/hw/pci/pci_ids.h line 59

Quoted 7 lines
include/hw/pci/pci_ids.h +1 -0
@@ -53,6 +53,7 @@
  #define PCI_BASE_CLASS_MEMORY            0x05
  #define PCI_CLASS_MEMORY_RAM             0x0500
  #define PCI_CLASS_MEMORY_FLASH           0x0501
+#define PCI_CLASS_MEMORY_CXL             0x0502
  #define PCI_CLASS_MEMORY_OTHER           0x0580
 
  #define PCI_BASE_CLASS_BRIDGE            0x06

ATB,

Mark.

#33 lore raw

Re: [PATCH v8 17/46] hw/cxl/device: Add a memory device (8.2.8.5)

Quoted 28 lines · On Sat, 19 Mar 2022 08:32:29 +0000 Mark Cave-Ayland…
Quoted 26 lines · Jonathan Cameron via

From: Ben Widawsky <[email protected]>

A CXL memory device (AKA Type 3) is a CXL component that contains some combination of volatile and persistent memory. It also implements the previously defined mailbox interface as well as the memory device firmware interface.

Although the memory device is configured like a normal PCIe device, the memory traffic is on an entirely separate bus conceptually (using the same physical wires as PCIe, but different protocol).

Once the CXL topology is fully configure and address decoders committed, the guest physical address for the memory device is part of a larger window which is owned by the platform. The creation of these windows is later in this series.

The following example will create a 256M device in a 512M window:

(quoted diff) +0 -2
-object "memory-backend-file,id=cxl-mem1,share,mem-path=cxl-type3,size=512M"
-device "cxl-type3,bus=rp0,memdev=cxl-mem1,id=cxl-pmem0"

Note: Dropped PCDIMM info interfaces for now. They can be added if appropriate at a later date.

---

...

Quoted 32 lines
Quoted 32 lines
hw/mem/cxl_type3.c new file +26 -0
new file mode 100644
index 0000000000..a8d7cfcc81
@@ -0,0 +1,153 @@
+#include "qemu/osdep.h"
+#include "qemu/units.h"
+#include "qemu/error-report.h"
+#include "hw/mem/memory-device.h"
+#include "hw/mem/pc-dimm.h"
+#include "hw/pci/pci.h"
+#include "hw/qdev-properties.h"
+#include "qapi/error.h"
+#include "qemu/log.h"
+#include "qemu/module.h"
+#include "qemu/range.h"
+#include "qemu/rcu.h"
+#include "sysemu/hostmem.h"
+#include "hw/cxl/cxl.h"
+
+static void build_dvsecs(CXLType3Dev *ct3d)
+{
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    uint8_t *dvsec;
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_device){
+        .cap = 0x1e,
+        .ctrl = 0x6,
+        .status2 = 0x2,
+        .range1_size_hi = 0,
+#ifdef SET_PMEM_PADDR
Quoted 8 lines

hw/mem/cxl_type3.c line 30

Quoted 4 lines
hw/mem/cxl_type3.c new file +4 -0
+        .range1_size_lo = (2 << 5) | (2 << 2) | 0x3 | ct3d->size,
+#else
+        .range1_size_lo = 0x3,
+#endif  

Generally you don't want #ifdefs like this to control functionality: it should either be removed if it is for unimplemented features, or controlled via a qdev property in ct3_props below.

Oops. That should have been long gone. It's a leftover from a much earlier attempt to handle the memory address spaces.
The size property of a type3 devices is also not used for anything worthwhile any more as there seems little reason to not just use the hostmem region size instead.
So I'll drop that whilst tidying this up which also involves a bunch of changes to tests and docs.

Quoted 19 lines
Quoted 19 lines
hw/mem/cxl_type3.c new file +18 -0
+        .range1_base_hi = 0,
+        .range1_base_lo = 0,
+    };
+    cxl_component_create_dvsec(cxl_cstate, PCIE_CXL_DEVICE_DVSEC_LENGTH,
+                               PCIE_CXL_DEVICE_DVSEC,
+                               PCIE_CXL2_DEVICE_DVSEC_REVID, dvsec);
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_register_locator){
+        .rsvd         = 0,
+        .reg0_base_lo = RBI_COMPONENT_REG | CXL_COMPONENT_REG_BAR_IDX,
+        .reg0_base_hi = 0,
+        .reg1_base_lo = RBI_CXL_DEVICE_REG | CXL_DEVICE_REG_BAR_IDX,
+        .reg1_base_hi = 0,
+    };
+    cxl_component_create_dvsec(cxl_cstate, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
+                               REG_LOC_DVSEC_REVID, dvsec);
+}
+

...

Quoted 42 lines
Quoted 27 lines
hw/mem/meson.build +1 -0
index 82f86d117e..609b2b36fc 100644
@@ -3,6 +3,7 @@ mem_ss.add(files('memory-device.c'))
  mem_ss.add(when: 'CONFIG_DIMM', if_true: files('pc-dimm.c'))
  mem_ss.add(when: 'CONFIG_NPCM7XX', if_true: files('npcm7xx_mc.c'))
  mem_ss.add(when: 'CONFIG_NVDIMM', if_true: files('nvdimm.c'))
+mem_ss.add(when: 'CONFIG_CXL_MEM_DEVICE', if_true: files('cxl_type3.c'))
 
  softmmu_ss.add_all(when: 'CONFIG_MEM_DEVICE', if_true: mem_ss)
 
include/hw/cxl/cxl_device.h +7 -0
index 8102d2a813..72da811c52 100644
@@ -230,4 +230,21 @@ REG64(CXL_MEM_DEV_STS, 0)
      FIELD(CXL_MEM_DEV_STS, MBOX_READY, 4, 1)
      FIELD(CXL_MEM_DEV_STS, RESET_NEEDED, 5, 3)
 
+typedef struct cxl_type3_dev {
+    /* Private */
+    PCIDevice parent_obj;
+
+    /* Properties */
+    uint64_t size;
+    HostMemoryBackend *hostmem;

include/hw/cxl/cxl_device.h REG64(CXL_MEM_DEV_STS, 0) · line 249

Quoted 10 lines
include/hw/cxl/cxl_device.h +10 -0
+
+    /* State */
+    CXLComponentState cxl_cstate;
+    CXLDeviceState cxl_dstate;
+} CXLType3Dev;
+
+#define TYPE_CXL_TYPE3_DEV "cxl-type3"
+
+#define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
+  

Ah okay this is an old style initialiser, and new code shouldn't be using them anymore (I believe there should be no remaining instances in git master). Drop the typedef from struct cxl_type3_dev and replace with:

Quoted 6 lines

#define TYPE_CXL_TYPE3_DEV "cxl-type3" OBJECT_DECLARE_SIMPLE_TYPE(CXLType3Dev, CXL_TYPE3_DEV))

Note this will alter the generated QOM CAST from CT3() to CXL_TYPE3_DEV(): I would argue that the _DEV suffix is QOM legacy naming and recommend removing it from both of the above to give you TYPE_CXL_TYPE3 and CXL_TYPE3 respectively.

Code has been around a while and being out of tree this sort of stuff is only likely to get picked up by reviewers (thanks!)
Now updated to this newer approach - this code mostly went away in patch 20 where a class is introduced, but I've switched to your new suggested naming and gotten rid of CT3() which was left behind.

Thanks,

Jonathan

#34 lore raw

PATCH v8 18/46 hw/cxl/device: Implement MMIO HDM decoding (8.2.5.12)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

A device's volatile and persistent memory are known Host Defined Memory (HDM) regions. The mechanism by which the device is programmed to claim the addresses associated with those regions is through dedicated logic known as the HDM decoder. In order to allow the OS to properly program the HDMs, the HDM decoders must be modeled.

There are two ways the HDM decoders can be implemented, the legacy mechanism is through the PCIe DVSEC programming from CXL 1.1 (8.1.3.8), and MMIO is found in 8.2.5.12 of the spec. For now, 8.1.3.8 is not implemented.

Much of CXL device logic is implemented in cxl-utils. The HDM decoder however is implemented directly by the device implementation. Whilst the implementation currently does no validity checks on the encoder set up, future work will add sanity checking specific to the type of cxl component.

---
 hw/mem/cxl_type3.c | 55 ++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 55 insertions(+)
hw/mem/cxl_type3.c +55 -0
index a8d7cfcc81..16b113d5ed 100644
@@ -46,6 +46,57 @@ static void build_dvsecs(CXLType3Dev *ct3d)
                                REG_LOC_DVSEC_REVID, dvsec);
 }
 
+static void hdm_decoder_commit(CXLType3Dev *ct3d, int which)
+{
+    ComponentRegisters *cregs = &ct3d->cxl_cstate.crb;
+    uint32_t *cache_mem = cregs->cache_mem_registers;
+
+    assert(which == 0);
+
+    /* TODO: Sanity checks that the decoder is possible */
+    ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, COMMIT, 0);
+    ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, ERR, 0);
+
+    ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, COMMITTED, 1);
+}
+
+static void ct3d_reg_write(void *opaque, hwaddr offset, uint64_t value,
+                           unsigned size)
+{
+    CXLComponentState *cxl_cstate = opaque;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    CXLType3Dev *ct3d = container_of(cxl_cstate, CXLType3Dev, cxl_cstate);
+    uint32_t *cache_mem = cregs->cache_mem_registers;
+    bool should_commit = false;
+    int which_hdm = -1;
+
+    assert(size == 4);
+    g_assert(offset <= CXL2_COMPONENT_CM_REGION_SIZE);
+
+    switch (offset) {
+    case A_CXL_HDM_DECODER0_CTRL:
+        should_commit = FIELD_EX32(value, CXL_HDM_DECODER0_CTRL, COMMIT);
+        which_hdm = 0;
+        break;
+    default:
+        break;
+    }
+
+    stl_le_p((uint8_t *)cache_mem + offset, value);
+    if (should_commit) {
+        hdm_decoder_commit(ct3d, which_hdm);
+    }
+}
+
+static void ct3_finalize(Object *obj)
+{
+    CXLType3Dev *ct3d = CT3(obj);
+    CXLComponentState *cxl_cstate = &ct3d->cxl_cstate;
+    ComponentRegisters *regs = &cxl_cstate->crb;
+
+    g_free(regs->special_ops);
+}
+
 static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
 {
     MemoryRegion *mr;
@@ -86,6 +137,9 @@ static void ct3_realize(PCIDevice *pci_dev, Error **errp)
     ct3d->cxl_cstate.pdev = pci_dev;
     build_dvsecs(ct3d);
 
+    regs->special_ops = g_new0(MemoryRegionOps, 1);
+    regs->special_ops->write = ct3d_reg_write;
+
     cxl_component_register_block_init(OBJECT(pci_dev), cxl_cstate,
                                       TYPE_CXL_TYPE3_DEV);
 
@@ -138,6 +192,7 @@ static const TypeInfo ct3d_info = {
     .parent = TYPE_PCI_DEVICE,
     .class_init = ct3_class_init,
     .instance_size = sizeof(CXLType3Dev),
+    .instance_finalize = ct3_finalize,
     .interfaces = (InterfaceInfo[]) {
         { INTERFACE_CXL_DEVICE },
         { INTERFACE_PCIE_DEVICE },
Signature · 1 line
2.32.0
#35 lore raw

PATCH v8 19/46 hw/cxl/device: Add some trivial commands

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

GET_FW_INFO and GET_PARTITION_INFO, for this emulation, is equivalent to info already returned in the IDENTIFY command. To have a more robust implementation, add those.

---
 hw/cxl/cxl-mailbox-utils.c | 69 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 69 insertions(+)
hw/cxl/cxl-mailbox-utils.c +69 -0
index 7ed9606c6b..fcd41d9a9d 100644
@@ -10,6 +10,7 @@
 #include "qemu/osdep.h"
 #include "hw/cxl/cxl.h"
 #include "hw/pci/pci.h"
+#include "qemu/cutils.h"
 #include "qemu/log.h"
 #include "qemu/uuid.h"
 
@@ -44,6 +45,8 @@ enum {
         #define CLEAR_RECORDS   0x1
         #define GET_INTERRUPT_POLICY   0x2
         #define SET_INTERRUPT_POLICY   0x3
+    FIRMWARE_UPDATE = 0x02,
+        #define GET_INFO      0x0
     TIMESTAMP   = 0x03,
         #define GET           0x0
         #define SET           0x1
@@ -52,6 +55,8 @@ enum {
         #define GET_LOG       0x1
     IDENTIFY    = 0x40,
         #define MEMORY_DEVICE 0x0
+    CCLS        = 0x41,
+        #define GET_PARTITION_INFO     0x0
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -114,6 +119,39 @@ DEFINE_MAILBOX_HANDLER_NOP(events_clear_records);
 DEFINE_MAILBOX_HANDLER_ZEROED(events_get_interrupt_policy, 4);
 DEFINE_MAILBOX_HANDLER_NOP(events_set_interrupt_policy);
 
+/* 8.2.9.2.1 */
+static ret_code cmd_firmware_update_get_info(struct cxl_cmd *cmd,
+                                             CXLDeviceState *cxl_dstate,
+                                             uint16_t *len)
+{
+    struct {
+        uint8_t slots_supported;
+        uint8_t slot_info;
+        uint8_t caps;
+        uint8_t rsvd[0xd];
+        char fw_rev1[0x10];
+        char fw_rev2[0x10];
+        char fw_rev3[0x10];
+        char fw_rev4[0x10];
+    } QEMU_PACKED *fw_info;
+    QEMU_BUILD_BUG_ON(sizeof(*fw_info) != 0x50);
+
+    if (cxl_dstate->pmem_size < (256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    fw_info = (void *)cmd->payload;
+    memset(fw_info, 0, sizeof(*fw_info));
+
+    fw_info->slots_supported = 2;
+    fw_info->slot_info = BIT(0) | BIT(3);
+    fw_info->caps = 0;
+    pstrcpy(fw_info->fw_rev1, sizeof(fw_info->fw_rev1), "BWFW VERSION 0");
+
+    *len = sizeof(*fw_info);
+    return CXL_MBOX_SUCCESS;
+}
+
 /* 8.2.9.3.1 */
 static ret_code cmd_timestamp_get(struct cxl_cmd *cmd,
                                   CXLDeviceState *cxl_dstate,
@@ -258,6 +296,33 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
     return CXL_MBOX_SUCCESS;
 }
 
+static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
+                                           CXLDeviceState *cxl_dstate,
+                                           uint16_t *len)
+{
+    struct {
+        uint64_t active_vmem;
+        uint64_t active_pmem;
+        uint64_t next_vmem;
+        uint64_t next_pmem;
+    } QEMU_PACKED *part_info = (void *)cmd->payload;
+    QEMU_BUILD_BUG_ON(sizeof(*part_info) != 0x20);
+    uint64_t size = cxl_dstate->pmem_size;
+
+    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    /* PMEM only */
+    part_info->active_vmem = 0;
+    part_info->next_vmem = 0;
+    part_info->active_pmem = size / (256 << 20);
+    part_info->next_pmem = part_info->active_pmem;
+
+    *len = sizeof(*part_info);
+    return CXL_MBOX_SUCCESS;
+}
+
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
 #define IMMEDIATE_POLICY_CHANGE (1 << 3)
 #define IMMEDIATE_LOG_CHANGE (1 << 4)
@@ -271,12 +336,16 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
         cmd_events_get_interrupt_policy, 0, 0 },
     [EVENTS][SET_INTERRUPT_POLICY] = { "EVENTS_SET_INTERRUPT_POLICY",
         cmd_events_set_interrupt_policy, 4, IMMEDIATE_CONFIG_CHANGE },
+    [FIRMWARE_UPDATE][GET_INFO] = { "FIRMWARE_UPDATE_GET_INFO",
+        cmd_firmware_update_get_info, 0, 0 },
     [TIMESTAMP][GET] = { "TIMESTAMP_GET", cmd_timestamp_get, 0, 0 },
     [TIMESTAMP][SET] = { "TIMESTAMP_SET", cmd_timestamp_set, 8, IMMEDIATE_POLICY_CHANGE },
     [LOGS][GET_SUPPORTED] = { "LOGS_GET_SUPPORTED", cmd_logs_get_supported, 0, 0 },
     [LOGS][GET_LOG] = { "LOGS_GET_LOG", cmd_logs_get_log, 0x18, 0 },
     [IDENTIFY][MEMORY_DEVICE] = { "IDENTIFY_MEMORY_DEVICE",
         cmd_identify_memory_device, 0, 0 },
+    [CCLS][GET_PARTITION_INFO] = { "CCLS_GET_PARTITION_INFO",
+        cmd_ccls_get_partition_info, 0, 0 },
 };
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
Signature · 1 line
2.32.0
#36 lore raw

Re: [PATCH v8 19/46] hw/cxl/device: Add some trivial commands

Quoted 12 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

GET_FW_INFO and GET_PARTITION_INFO, for this emulation, is equivalent to info already returned in the IDENTIFY command. To have a more robust implementation, add those.

---
 hw/cxl/cxl-mailbox-utils.c | 69 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 69 insertions(+)

snip

Quoted 13 lines
hw/cxl/cxl-mailbox-utils.c +12 -0
+static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
+                                           CXLDeviceState *cxl_dstate,
+                                           uint16_t *len)
+{
+    struct {
+        uint64_t active_vmem;
+        uint64_t active_pmem;
+        uint64_t next_vmem;
+        uint64_t next_pmem;
+    } QEMU_PACKED *part_info = (void *)cmd->payload;
+    QEMU_BUILD_BUG_ON(sizeof(*part_info) != 0x20);
+    uint64_t size = cxl_dstate->pmem_size;

hw/cxl/cxl-mailbox-utils.c static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd, · line 320

Quoted 10 lines
hw/cxl/cxl-mailbox-utils.c +10 -0
+
+    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    /* PMEM only */
+    part_info->active_vmem = 0;
+    part_info->next_vmem = 0;
+    part_info->active_pmem = size / (256 << 20);
+    part_info->next_pmem = part_info->active_pmem;

Setting next like this is logical, but it's not per the CXL spec:

8.2.9.5.2.1
"Next Persistent Capacity: If non-zero, this value shall become the Active Persistent Capacity on the next cold reset. If both this field and the Next Volatile Capacity field are zero, there is no pending change to the partitioning."

next_(vmem|pmem) should start as zero and only change as the result of a successful set_partition_info command.

From your cover letter:
* Volatile memory devices (easy but it's more code so left for now). Wondering if this is something I could do, and follow that with set_partition support. Does that sound reasonable?

Alison

Quoted 5 lines
hw/cxl/cxl-mailbox-utils.c +5 -0
+
+    *len = sizeof(*part_info);
+    return CXL_MBOX_SUCCESS;
+}
+

snip

#37 lore raw

Re: [PATCH v8 19/46] hw/cxl/device: Add some trivial commands

Quoted 42 lines · On Fri, 18 Mar 2022 09:56:22 -0700 Alison Schofield…
Quoted 12 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

GET_FW_INFO and GET_PARTITION_INFO, for this emulation, is equivalent to info already returned in the IDENTIFY command. To have a more robust implementation, add those.

---
 hw/cxl/cxl-mailbox-utils.c | 69 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 69 insertions(+)

snip

Quoted 13 lines
hw/cxl/cxl-mailbox-utils.c +12 -0
+static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
+                                           CXLDeviceState *cxl_dstate,
+                                           uint16_t *len)
+{
+    struct {
+        uint64_t active_vmem;
+        uint64_t active_pmem;
+        uint64_t next_vmem;
+        uint64_t next_pmem;
+    } QEMU_PACKED *part_info = (void *)cmd->payload;
+    QEMU_BUILD_BUG_ON(sizeof(*part_info) != 0x20);
+    uint64_t size = cxl_dstate->pmem_size;

hw/cxl/cxl-mailbox-utils.c static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd, · line 320

Quoted 10 lines
hw/cxl/cxl-mailbox-utils.c +10 -0
+
+    if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
+        return CXL_MBOX_INTERNAL_ERROR;
+    }
+
+    /* PMEM only */
+    part_info->active_vmem = 0;
+    part_info->next_vmem = 0;
+    part_info->active_pmem = size / (256 << 20);
+    part_info->next_pmem = part_info->active_pmem;  

Setting next like this is logical, but it's not per the CXL spec:

Quoted 8 lines

8.2.9.5.2.1
"Next Persistent Capacity: If non-zero, this value shall become the Active Persistent Capacity on the next cold reset. If both this field and the Next Volatile Capacity field are zero, there is no pending change to the partitioning."

next_(vmem|pmem) should start as zero and only change as the result of a successful set_partition_info command.

Ah. Good point and now fixed up.

Quoted 5 lines

From your cover letter:
* Volatile memory devices (easy but it's more code so left for now). Wondering if this is something I could do, and follow that with set_partition support. Does that sound reasonable?

Sure, that would be great.

It raises an interesting question for what the volatile / non volatile backend will look like. So what is the command line?

So far I'd been assuming we'd do it as separate memdev devices for volatile from those for non volatile because chances are someone testing would back the volatile with RAM the non volatile with a file. If we enable the partitioning control that approach won't make much sense. Various options come to mind.
1) Just back with one memdev and present that as persistent or not but actually always have it persistent under the hood.
2) Allow backing with 2 memdevs and don't support the set_partition command. Lazy approach I was planing on doing eventualy.
3) Support 3 memdevs. The middle one of which is the part we allow to be repartitioned if it is provided

Also, to actually enable this we'd probably want more HDM decoders than currently supported and flags for CFMWS which aren't there yet (Ben asked for them but I've left it for a future patch set).

Lots still to do once the initial patch set is moving forwards. I don't mind queuing stuff up it gitlab though and then we can send suitable series to the list once the earlier part has been applied. In the meantime we would have a moderately stable based to build on top of.

Thanks,

Jonathan

Quoted 12 lines

Alison

Quoted 5 lines
hw/cxl/cxl-mailbox-utils.c +5 -0
+
+    *len = sizeof(*part_info);
+    return CXL_MBOX_SUCCESS;
+}
+  

snip

#38 lore raw

PATCH v8 20/46 hw/cxl/device: Plumb real Label Storage Area (LSA) sizing

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

This should introduce no change. Subsequent work will make use of this new class member.

---
 hw/cxl/cxl-mailbox-utils.c  |  3 +++
 hw/mem/cxl_type3.c          |  9 +++++++++
 include/hw/cxl/cxl_device.h | 10 ++++++++++
 3 files changed, 22 insertions(+)
hw/cxl/cxl-mailbox-utils.c +3 -0
index fcd41d9a9d..771b1cfe90 100644
@@ -277,6 +277,8 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
     } QEMU_PACKED *id;
     QEMU_BUILD_BUG_ON(sizeof(*id) != 0x43);
 
+    CXLType3Dev *ct3d = container_of(cxl_dstate, CXLType3Dev, cxl_dstate);
+    CXLType3Class *cvc = CXL_TYPE3_DEV_GET_CLASS(ct3d);
     uint64_t size = cxl_dstate->pmem_size;
 
     if (!QEMU_IS_ALIGNED(size, 256 << 20)) {
@@ -291,6 +293,7 @@ static ret_code cmd_identify_memory_device(struct cxl_cmd *cmd,
 
     id->total_capacity = size / (256 << 20);
     id->persistent_capacity = size / (256 << 20);
+    id->lsa_size = cvc->get_lsa_size(ct3d);
 
     *len = sizeof(*id);
     return CXL_MBOX_SUCCESS;
hw/mem/cxl_type3.c +9 -0
index 16b113d5ed..7cd3041eb3 100644
@@ -170,10 +170,16 @@ static Property ct3_props[] = {
     DEFINE_PROP_END_OF_LIST(),
 };
 
+static uint64_t get_lsa_size(CXLType3Dev *ct3d)
+{
+    return 0;
+}
+
 static void ct3_class_init(ObjectClass *oc, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(oc);
     PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc);
+    CXLType3Class *cvc = CXL_TYPE3_DEV_CLASS(oc);
 
     pc->realize = ct3_realize;
     pc->class_id = PCI_CLASS_STORAGE_EXPRESS;
@@ -185,11 +191,14 @@ static void ct3_class_init(ObjectClass *oc, void *data)
     dc->desc = "CXL PMEM Device (Type 3)";
     dc->reset = ct3d_reset;
     device_class_set_props(dc, ct3_props);
+
+    cvc->get_lsa_size = get_lsa_size;
 }
 
 static const TypeInfo ct3d_info = {
     .name = TYPE_CXL_TYPE3_DEV,
     .parent = TYPE_PCI_DEVICE,
+    .class_size = sizeof(struct CXLType3Class),
     .class_init = ct3_class_init,
     .instance_size = sizeof(CXLType3Dev),
     .instance_finalize = ct3_finalize,
include/hw/cxl/cxl_device.h +10 -0
index 72da811c52..cf4c110f7e 100644
@@ -237,6 +237,7 @@ typedef struct cxl_type3_dev {
     /* Properties */
     uint64_t size;
     HostMemoryBackend *hostmem;
+    HostMemoryBackend *lsa;
 
     /* State */
     CXLComponentState cxl_cstate;
@@ -246,5 +247,14 @@ typedef struct cxl_type3_dev {
 #define TYPE_CXL_TYPE3_DEV "cxl-type3"
 
 #define CT3(obj) OBJECT_CHECK(CXLType3Dev, (obj), TYPE_CXL_TYPE3_DEV)
+OBJECT_DECLARE_TYPE(CXLType3Device, CXLType3Class, CXL_TYPE3_DEV)
+
+struct CXLType3Class {
+    /* Private */
+    PCIDeviceClass parent_class;
+
+    /* public */
+    uint64_t (*get_lsa_size)(CXLType3Dev *ct3d);
+};
 
 #endif
Signature · 1 line
2.32.0
#39 lore raw

PATCH v8 21/46 hw/cxl/device: Implement get/set Label Storage Area (LSA)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

Implement get and set handlers for the Label Storage Area used to hold data describing persistent memory configuration so that it can be ensured it is seen in the same configuration after reboot.

---
 hw/cxl/cxl-mailbox-utils.c  | 60 +++++++++++++++++++++++++++++++++++++
 hw/mem/cxl_type3.c          | 56 +++++++++++++++++++++++++++++++++-
 include/hw/cxl/cxl_device.h |  5 ++++
 3 files changed, 120 insertions(+), 1 deletion(-)
hw/cxl/cxl-mailbox-utils.c +60 -0
index 771b1cfe90..acb73c7a68 100644
@@ -57,6 +57,8 @@ enum {
         #define MEMORY_DEVICE 0x0
     CCLS        = 0x41,
         #define GET_PARTITION_INFO     0x0
+        #define GET_LSA       0x2
+        #define SET_LSA       0x3
 };
 
 /* 8.2.8.4.5.1 Command Return Codes */
@@ -326,7 +328,62 @@ static ret_code cmd_ccls_get_partition_info(struct cxl_cmd *cmd,
     return CXL_MBOX_SUCCESS;
 }
 
+static ret_code cmd_ccls_get_lsa(struct cxl_cmd *cmd,
+                                 CXLDeviceState *cxl_dstate,
+                                 uint16_t *len)
+{
+    struct {
+        uint32_t offset;
+        uint32_t length;
+    } QEMU_PACKED *get_lsa;
+    CXLType3Dev *ct3d = container_of(cxl_dstate, CXLType3Dev, cxl_dstate);
+    CXLType3Class *cvc = CXL_TYPE3_DEV_GET_CLASS(ct3d);
+    uint32_t offset, length;
+
+    get_lsa = (void *)cmd->payload;
+    offset = get_lsa->offset;
+    length = get_lsa->length;
+
+    if (offset + length > cvc->get_lsa_size(ct3d)) {
+        *len = 0;
+        return CXL_MBOX_INVALID_INPUT;
+    }
+
+    *len = cvc->get_lsa(ct3d, get_lsa, length, offset);
+    return CXL_MBOX_SUCCESS;
+}
+
+static ret_code cmd_ccls_set_lsa(struct cxl_cmd *cmd,
+                                 CXLDeviceState *cxl_dstate,
+                                 uint16_t *len)
+{
+    struct set_lsa_pl {
+        uint32_t offset;
+        uint32_t rsvd;
+        uint8_t data[];
+    } QEMU_PACKED;
+    struct set_lsa_pl *set_lsa_payload = (void *)cmd->payload;
+    CXLType3Dev *ct3d = container_of(cxl_dstate, CXLType3Dev, cxl_dstate);
+    CXLType3Class *cvc = CXL_TYPE3_DEV_GET_CLASS(ct3d);
+    const size_t hdr_len = offsetof(struct set_lsa_pl, data);
+    uint16_t plen = *len;
+
+    *len = 0;
+    if (!plen) {
+        return CXL_MBOX_SUCCESS;
+    }
+
+    if (set_lsa_payload->offset + plen > cvc->get_lsa_size(ct3d) + hdr_len) {
+        return CXL_MBOX_INVALID_INPUT;
+    }
+    plen -= hdr_len;
+
+    cvc->set_lsa(ct3d, set_lsa_payload->data, plen, set_lsa_payload->offset);
+    return CXL_MBOX_SUCCESS;
+}
+
 #define IMMEDIATE_CONFIG_CHANGE (1 << 1)
+#define IMMEDIATE_DATA_CHANGE (1 << 2)
 #define IMMEDIATE_POLICY_CHANGE (1 << 3)
 #define IMMEDIATE_LOG_CHANGE (1 << 4)
 
@@ -349,6 +406,9 @@ static struct cxl_cmd cxl_cmd_set[256][256] = {
         cmd_identify_memory_device, 0, 0 },
     [CCLS][GET_PARTITION_INFO] = { "CCLS_GET_PARTITION_INFO",
         cmd_ccls_get_partition_info, 0, 0 },
+    [CCLS][GET_LSA] = { "CCLS_GET_LSA", cmd_ccls_get_lsa, 0, 0 },
+    [CCLS][SET_LSA] = { "CCLS_SET_LSA", cmd_ccls_set_lsa,
+        ~0, IMMEDIATE_CONFIG_CHANGE | IMMEDIATE_DATA_CHANGE },
 };
 
 void cxl_process_mailbox(CXLDeviceState *cxl_dstate)
hw/mem/cxl_type3.c +55 -1

85 lines not on this page · raw message

include/hw/cxl/cxl_device.h +5 -0
index cf4c110f7e..288cc11772 100644
@@ -255,6 +255,11 @@ struct CXLType3Class {
 
     /* public */
     uint64_t (*get_lsa_size)(CXLType3Dev *ct3d);
+
+    uint64_t (*get_lsa)(CXLType3Dev *ct3d, void *buf, uint64_t size,
+                        uint64_t offset);
+    void (*set_lsa)(CXLType3Dev *ct3d, const void *buf, uint64_t size,
+                    uint64_t offset);
 };
 
 #endif
Signature · 1 line
2.32.0
#40 lore raw

PATCH v8 22/46 qtests/cxl: Add initial root port and CXL type3 tests

In reply to Jonathan Cameron

At this stage we can boot configurations with host bridges, root ports and type 3 memory devices, so add appropriate tests.

---
 tests/qtest/cxl-test.c | 126 +++++++++++++++++++++++++++++++++++++++++
 1 file changed, 126 insertions(+)
tests/qtest/cxl-test.c +126 -0
index 1006c8ae4e..148bc94340 100644
@@ -8,6 +8,54 @@
 #include "qemu/osdep.h"
 #include "libqtest-single.h"
 
+#define QEMU_PXB_CMD "-machine q35,cxl=on " \
+                     "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52 "
+
+#define QEMU_2PXB_CMD "-machine q35,cxl=on " \
+                      "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52 "  \
+                      "-device pxb-cxl,id=cxl.1,bus=pcie.0,bus_nr=53 "
+
+#define QEMU_RP "-device cxl-rp,id=rp0,bus=cxl.0,chassis=0,slot=0 "
+
+/* Dual ports on first pxb */
+#define QEMU_2RP "-device cxl-rp,id=rp0,bus=cxl.0,chassis=0,slot=0 " \
+                 "-device cxl-rp,id=rp1,bus=cxl.0,chassis=0,slot=1 "
+
+/* Dual ports on each of the pxb instances */
+#define QEMU_4RP "-device cxl-rp,id=rp0,bus=cxl.0,chassis=0,slot=0 " \
+                 "-device cxl-rp,id=rp1,bus=cxl.0,chassis=0,slot=1 " \
+                 "-device cxl-rp,id=rp2,bus=cxl.1,chassis=0,slot=2 " \
+                 "-device cxl-rp,id=rp3,bus=cxl.1,chassis=0,slot=3 "
+
+#define QEMU_T3D "-object memory-backend-file,id=cxl-mem0,mem-path=%s,size=256M " \
+                 "-object memory-backend-file,id=lsa0,mem-path=%s,size=256M "    \
+                 "-device cxl-type3,bus=rp0,memdev=cxl-mem0,lsa=lsa0,id=cxl-pmem0,size=256M "
+
+#define QEMU_2T3D "-object memory-backend-file,id=cxl-mem0,mem-path=%s,size=256M "    \
+                  "-object memory-backend-file,id=lsa0,mem-path=%s,size=256M "    \
+                  "-device cxl-type3,bus=rp0,memdev=cxl-mem0,lsa=lsa0,id=cxl-pmem0,size=256M " \
+                  "-object memory-backend-file,id=cxl-mem1,mem-path=%s,size=256M "    \
+                  "-object memory-backend-file,id=lsa1,mem-path=%s,size=256M "    \
+                  "-device cxl-type3,bus=rp1,memdev=cxl-mem1,lsa=lsa1,id=cxl-pmem1,size=256M "
+
+#define QEMU_4T3D "-object memory-backend-file,id=cxl-mem0,mem-path=%s,size=256M " \
+                  "-object memory-backend-file,id=lsa0,mem-path=%s,size=256M "    \
+                  "-device cxl-type3,bus=rp0,memdev=cxl-mem0,lsa=lsa0,id=cxl-pmem0,size=256M " \
+                  "-object memory-backend-file,id=cxl-mem1,mem-path=%s,size=256M "    \
+                  "-object memory-backend-file,id=lsa1,mem-path=%s,size=256M "    \
+                  "-device cxl-type3,bus=rp1,memdev=cxl-mem1,lsa=lsa1,id=cxl-pmem1,size=256M " \
+                  "-object memory-backend-file,id=cxl-mem2,mem-path=%s,size=256M "    \
+                  "-object memory-backend-file,id=lsa2,mem-path=%s,size=256M "    \
+                  "-device cxl-type3,bus=rp2,memdev=cxl-mem2,lsa=lsa2,id=cxl-pmem2,size=256M " \
+                  "-object memory-backend-file,id=cxl-mem3,mem-path=%s,size=256M "    \
+                  "-object memory-backend-file,id=lsa3,mem-path=%s,size=256M "    \
+                  "-device cxl-type3,bus=rp3,memdev=cxl-mem3,lsa=lsa3,id=cxl-pmem3,size=256M "
+
+static void cxl_basic_hb(void)
+{
+    qtest_start("-machine q35,cxl=on");
+    qtest_end();
+}
 
 static void cxl_basic_pxb(void)
 {
@@ -15,9 +63,87 @@ static void cxl_basic_pxb(void)
     qtest_end();
 }
 
+static void cxl_pxb_with_window(void)
+{
+    qtest_start(QEMU_PXB_CMD);
+    qtest_end();
+}
+
+static void cxl_2pxb_with_window(void)
+{
+    qtest_start(QEMU_2PXB_CMD);
+    qtest_end();
+}
+
+static void cxl_root_port(void)
+{
+    qtest_start(QEMU_PXB_CMD QEMU_RP);
+    qtest_end();
+}
+
+static void cxl_2root_port(void)
+{
+    qtest_start(QEMU_PXB_CMD QEMU_2RP);
+    qtest_end();
+}
+
+static void cxl_t3d(void)
+{
+    g_autoptr(GString) cmdline = g_string_new(NULL);
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    g_string_printf(cmdline, QEMU_PXB_CMD QEMU_RP QEMU_T3D, tmpfs, tmpfs);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+}
+
+static void cxl_1pxb_2rp_2t3d(void)
+{
+    g_autoptr(GString) cmdline = g_string_new(NULL);
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    g_string_printf(cmdline, QEMU_PXB_CMD QEMU_2RP QEMU_2T3D,
+                    tmpfs, tmpfs, tmpfs, tmpfs);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+}
+
+static void cxl_2pxb_4rp_4t3d(void)
+{
+    g_autoptr(GString) cmdline = g_string_new(NULL);
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    g_string_printf(cmdline, QEMU_2PXB_CMD QEMU_4RP QEMU_4T3D,
+                    tmpfs, tmpfs, tmpfs, tmpfs, tmpfs, tmpfs,
+                    tmpfs, tmpfs);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+}
+
 int main(int argc, char **argv)
 {
     g_test_init(&argc, &argv, NULL);
+
+    qtest_add_func("/pci/cxl/basic_hostbridge", cxl_basic_hb);
     qtest_add_func("/pci/cxl/basic_pxb", cxl_basic_pxb);
+    qtest_add_func("/pci/cxl/pxb_with_window", cxl_pxb_with_window);
+    qtest_add_func("/pci/cxl/pxb_x2_with_window", cxl_2pxb_with_window);
+    qtest_add_func("/pci/cxl/rp", cxl_root_port);
+    qtest_add_func("/pci/cxl/rp_x2", cxl_2root_port);
+    qtest_add_func("/pci/cxl/type3_device", cxl_t3d);
+    qtest_add_func("/pci/cxl/rp_x2_type3_x2", cxl_1pxb_2rp_2t3d);
+    qtest_add_func("/pci/cxl/pxb_x2_root_port_x4_type3_x4", cxl_2pxb_4rp_4t3d);
     return g_test_run();
 }
Signature · 1 line
2.32.0
#41 lore raw

PATCH v8 23/46 hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

CXL host bridges themselves may have MMIO. Since host bridges don't have a BAR they are treated as special for MMIO. This patch includes i386/pc support.
Also hook up the device reset now that we have have the MMIO space in which the results are visible.

Note that we duplicate the PCI express case for the aml_build but the implementations will diverge when the CXL specific _OSC is introduced.

---
 hw/i386/acpi-build.c                | 25 +++++++++++-
 hw/i386/pc.c                        | 27 ++++++++++++-
 hw/pci-bridge/pci_expander_bridge.c | 62 ++++++++++++++++++++++++++++-
 include/hw/cxl/cxl.h                |  4 ++
 4 files changed, 114 insertions(+), 4 deletions(-)
hw/i386/acpi-build.c +23 -2
index 4ad4d7286c..0b7d06e107 100644
@@ -28,6 +28,7 @@
 #include "qemu/bitmap.h"
 #include "qemu/error-report.h"
 #include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
 #include "hw/core/cpu.h"
 #include "target/i386/cpu.h"
 #include "hw/misc/pvpanic.h"
@@ -1572,10 +1573,21 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
             }
 
             scope = aml_scope("\\_SB");
-            dev = aml_device("PC%.02X", bus_num);
+
+            if (pci_bus_is_cxl(bus)) {
+                dev = aml_device("CL%.02X", bus_num);
+            } else {
+                dev = aml_device("PC%.02X", bus_num);
+            }
             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
-            if (pci_bus_is_express(bus)) {
+            if (pci_bus_is_cxl(bus)) {
+                aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
+                aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
+
+                /* Expander bridges do not have ACPI PCI Hot-plug enabled */
+                aml_append(dev, build_q35_osc_method(true));
+            } else if (pci_bus_is_express(bus)) {
                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
 
@@ -1595,6 +1607,15 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
             aml_append(dev, aml_name_decl("_CRS", crs));
             aml_append(scope, dev);
             aml_append(dsdt, scope);
+
+            /* Handle the ranges for the PXB expanders */
+            if (pci_bus_is_cxl(bus)) {
+                MemoryRegion *mr = &machine->cxl_devices_state->host_mr;
+                uint64_t base = mr->addr;
+
+                crs_range_insert(crs_range_set.mem_ranges, base,
+                                 base + memory_region_size(mr) - 1);
+            }
         }
     }
 
hw/i386/pc.c +26 -1
index e2849fc741..da74f08f9e 100644
@@ -75,6 +75,7 @@
 #include "acpi-build.h"
 #include "hw/mem/pc-dimm.h"
 #include "hw/mem/nvdimm.h"
+#include "hw/cxl/cxl.h"
 #include "qapi/error.h"
 #include "qapi/qapi-visit-common.h"
 #include "qapi/qapi-visit-machine.h"
@@ -813,6 +814,7 @@ void pc_memory_init(PCMachineState *pcms,
     MachineClass *mc = MACHINE_GET_CLASS(machine);
     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
     X86MachineState *x86ms = X86_MACHINE(pcms);
+    hwaddr cxl_base;
 
     assert(machine->ram_size == x86ms->below_4g_mem_size +
                                 x86ms->above_4g_mem_size);
@@ -902,6 +904,26 @@ void pc_memory_init(PCMachineState *pcms,
                                     &machine->device_memory->mr);
     }
 
+    if (machine->cxl_devices_state->is_enabled) {
+        MemoryRegion *mr = &machine->cxl_devices_state->host_mr;
+        hwaddr cxl_size = MiB;
+
+        if (pcmc->has_reserved_memory && machine->device_memory->base) {
+            cxl_base = machine->device_memory->base;
+            if (!pcmc->broken_reserved_end) {
+                cxl_base += memory_region_size(&machine->device_memory->mr);
+            }
+        } else if (pcms->sgx_epc.size != 0) {
+            cxl_base = sgx_epc_above_4g_end(&pcms->sgx_epc);
+        } else {
+            cxl_base = 0x100000000ULL + x86ms->above_4g_mem_size;
+        }
+
+        e820_add_entry(cxl_base, cxl_size, E820_RESERVED);
+        memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
+        memory_region_add_subregion(system_memory, cxl_base, mr);
+    }
+
     /* Initialize PC system firmware */
     pc_system_firmware_init(pcms, rom_memory);
 
@@ -962,7 +984,10 @@ uint64_t pc_pci_hole64_start(void)
     X86MachineState *x86ms = X86_MACHINE(pcms);
     uint64_t hole64_start = 0;
 
-    if (pcmc->has_reserved_memory && ms->device_memory->base) {
+    if (ms->cxl_devices_state->host_mr.addr) {
+        hole64_start = ms->cxl_devices_state->host_mr.addr +
+            memory_region_size(&ms->cxl_devices_state->host_mr);
+    } else if (pcmc->has_reserved_memory && ms->device_memory->base) {
         hole64_start = ms->device_memory->base;
         if (!pcmc->broken_reserved_end) {
             hole64_start += memory_region_size(&ms->device_memory->mr);
hw/pci-bridge/pci_expander_bridge.c +61 -1

103 lines not on this page · raw message

include/hw/cxl/cxl.h +4 -0
index 31af92fd5e..75e5bf71e1 100644
@@ -17,8 +17,12 @@
 #define CXL_COMPONENT_REG_BAR_IDX 0
 #define CXL_DEVICE_REG_BAR_IDX 2
 
+#define CXL_WINDOW_MAX 10
+
 typedef struct CXLState {
     bool is_enabled;
+    MemoryRegion host_mr;
+    unsigned int next_mr_idx;
 } CXLState;
 
 #endif
Signature · 1 line
2.32.0
#42 lore raw

Re: [PATCH v8 23/46] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)

Quoted 149 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

CXL host bridges themselves may have MMIO. Since host bridges don't have a BAR they are treated as special for MMIO. This patch includes i386/pc support.
Also hook up the device reset now that we have have the MMIO space in which the results are visible.

Note that we duplicate the PCI express case for the aml_build but the implementations will diverge when the CXL specific _OSC is introduced.

---
  hw/i386/acpi-build.c                | 25 +++++++++++-
  hw/i386/pc.c                        | 27 ++++++++++++-
  hw/pci-bridge/pci_expander_bridge.c | 62 ++++++++++++++++++++++++++++-
  include/hw/cxl/cxl.h                |  4 ++
  4 files changed, 114 insertions(+), 4 deletions(-)
hw/i386/acpi-build.c +23 -2
index 4ad4d7286c..0b7d06e107 100644
@@ -28,6 +28,7 @@
  #include "qemu/bitmap.h"
  #include "qemu/error-report.h"
  #include "hw/pci/pci.h"
+#include "hw/cxl/cxl.h"
  #include "hw/core/cpu.h"
  #include "target/i386/cpu.h"
  #include "hw/misc/pvpanic.h"
@@ -1572,10 +1573,21 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
              }
 
              scope = aml_scope("\\_SB");
-            dev = aml_device("PC%.02X", bus_num);
+
+            if (pci_bus_is_cxl(bus)) {
+                dev = aml_device("CL%.02X", bus_num);
+            } else {
+                dev = aml_device("PC%.02X", bus_num);
+            }
              aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
              aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
-            if (pci_bus_is_express(bus)) {
+            if (pci_bus_is_cxl(bus)) {
+                aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
+                aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
+
+                /* Expander bridges do not have ACPI PCI Hot-plug enabled */
+                aml_append(dev, build_q35_osc_method(true));
+            } else if (pci_bus_is_express(bus)) {
                  aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
                  aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
 
@@ -1595,6 +1607,15 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
              aml_append(dev, aml_name_decl("_CRS", crs));
              aml_append(scope, dev);
              aml_append(dsdt, scope);
+
+            /* Handle the ranges for the PXB expanders */
+            if (pci_bus_is_cxl(bus)) {
+                MemoryRegion *mr = &machine->cxl_devices_state->host_mr;
+                uint64_t base = mr->addr;
+
+                crs_range_insert(crs_range_set.mem_ranges, base,
+                                 base + memory_region_size(mr) - 1);
+            }
          }
      }
 
hw/i386/pc.c +26 -1
index e2849fc741..da74f08f9e 100644
@@ -75,6 +75,7 @@
  #include "acpi-build.h"
  #include "hw/mem/pc-dimm.h"
  #include "hw/mem/nvdimm.h"
+#include "hw/cxl/cxl.h"
  #include "qapi/error.h"
  #include "qapi/qapi-visit-common.h"
  #include "qapi/qapi-visit-machine.h"
@@ -813,6 +814,7 @@ void pc_memory_init(PCMachineState *pcms,
      MachineClass *mc = MACHINE_GET_CLASS(machine);
      PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
      X86MachineState *x86ms = X86_MACHINE(pcms);
+    hwaddr cxl_base;
 
      assert(machine->ram_size == x86ms->below_4g_mem_size +
                                  x86ms->above_4g_mem_size);
@@ -902,6 +904,26 @@ void pc_memory_init(PCMachineState *pcms,
                                      &machine->device_memory->mr);
      }
 
+    if (machine->cxl_devices_state->is_enabled) {
+        MemoryRegion *mr = &machine->cxl_devices_state->host_mr;
+        hwaddr cxl_size = MiB;
+
+        if (pcmc->has_reserved_memory && machine->device_memory->base) {
+            cxl_base = machine->device_memory->base;
+            if (!pcmc->broken_reserved_end) {
+                cxl_base += memory_region_size(&machine->device_memory->mr);
+            }
+        } else if (pcms->sgx_epc.size != 0) {
+            cxl_base = sgx_epc_above_4g_end(&pcms->sgx_epc);
+        } else {
+            cxl_base = 0x100000000ULL + x86ms->above_4g_mem_size;
+        }
+
+        e820_add_entry(cxl_base, cxl_size, E820_RESERVED);
+        memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
+        memory_region_add_subregion(system_memory, cxl_base, mr);
+    }
+
      /* Initialize PC system firmware */
      pc_system_firmware_init(pcms, rom_memory);
 
@@ -962,7 +984,10 @@ uint64_t pc_pci_hole64_start(void)
      X86MachineState *x86ms = X86_MACHINE(pcms);
      uint64_t hole64_start = 0;
 
-    if (pcmc->has_reserved_memory && ms->device_memory->base) {
+    if (ms->cxl_devices_state->host_mr.addr) {
+        hole64_start = ms->cxl_devices_state->host_mr.addr +
+            memory_region_size(&ms->cxl_devices_state->host_mr);
+    } else if (pcmc->has_reserved_memory && ms->device_memory->base) {
          hole64_start = ms->device_memory->base;
          if (!pcmc->broken_reserved_end) {
              hole64_start += memory_region_size(&ms->device_memory->mr);
hw/pci-bridge/pci_expander_bridge.c +3 -0
index f762eb4a6e..b3b5f93650 100644
@@ -75,6 +75,9 @@ struct PXBDev {
      uint8_t bus_nr;
      uint16_t numa_node;
      bool bypass_iommu;
+    struct cxl_dev {
+        CXLHost *cxl_host_bridge;
+    } cxl;
  };
 
  static PXBDev *convert_to_pxb(PCIDevice *dev)

hw/pci-bridge/pci_expander_bridge.c static GList *pxb_dev_list; · line 99

Quoted 5 lines
hw/pci-bridge/pci_expander_bridge.c +2 -0
@@ -92,6 +95,9 @@ static GList *pxb_dev_list;
 
  #define TYPE_PXB_HOST "pxb-host"
 
+#define TYPE_PXB_CXL_HOST "pxb-cxl-host"
+#define PXB_CXL_HOST(obj) OBJECT_CHECK(CXLHost, (obj), TYPE_PXB_CXL_HOST)

Again this is a legacy QOM type declaration and should be replaced with OBJECT_DECLARE_SIMPLE_TYPE(). It should also be located in the associated header file if it exists (include/hw/cxl/cxl.h in this case).

Quoted 97 lines
hw/pci-bridge/pci_expander_bridge.c +55 -1
  static int pxb_bus_num(PCIBus *bus)
  {
      PXBDev *pxb = convert_to_pxb(bus->parent_dev);
@@ -197,6 +203,52 @@ static const TypeInfo pxb_host_info = {
      .class_init    = pxb_host_class_init,
  };
 
+static void pxb_cxl_realize(DeviceState *dev, Error **errp)
+{
+    MachineState *ms = MACHINE(qdev_get_machine());
+    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
+    CXLHost *cxl = PXB_CXL_HOST(dev);
+    CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
+    struct MemoryRegion *mr = &cxl_cstate->crb.component_registers;
+    hwaddr offset;
+
+    cxl_component_register_block_init(OBJECT(dev), cxl_cstate,
+                                      TYPE_PXB_CXL_HOST);
+    sysbus_init_mmio(sbd, mr);
+
+    offset = memory_region_size(mr) * ms->cxl_devices_state->next_mr_idx;
+    if (offset > memory_region_size(&ms->cxl_devices_state->host_mr)) {
+        error_setg(errp, "Insufficient space for pxb cxl host register space");
+        return;
+    }
+
+    memory_region_add_subregion(&ms->cxl_devices_state->host_mr, offset, mr);
+    ms->cxl_devices_state->next_mr_idx++;
+}
+
+static void pxb_cxl_host_class_init(ObjectClass *class, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(class);
+    PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(class);
+
+    hc->root_bus_path = pxb_host_root_bus_path;
+    dc->fw_name = "cxl";
+    dc->realize = pxb_cxl_realize;
+    /* Reason: Internal part of the pxb/pxb-pcie device, not usable by itself */
+    dc->user_creatable = false;
+}
+
+/*
+ * This is a device to handle the MMIO for a CXL host bridge. It does nothing
+ * else.
+ */
+static const TypeInfo cxl_host_info = {
+    .name          = TYPE_PXB_CXL_HOST,
+    .parent        = TYPE_PCI_HOST_BRIDGE,
+    .instance_size = sizeof(CXLHost),
+    .class_init    = pxb_cxl_host_class_init,
+};
+
  /*
   * Registers the PXB bus as a child of pci host root bus.
   */
@@ -245,6 +297,12 @@ static int pxb_map_irq_fn(PCIDevice *pci_dev, int pin)
 
  static void pxb_dev_reset(DeviceState *dev)
  {
+    CXLHost *cxl = PXB_CXL_DEV(dev)->cxl.cxl_host_bridge;
+    CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
+    uint32_t *reg_state = cxl_cstate->crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_ROOT_PORT);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 8);
  }
 
  static gint pxb_compare(gconstpointer a, gconstpointer b)
@@ -281,12 +339,13 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
          dev_name = dev->qdev.id;
      }
 
-    ds = qdev_new(TYPE_PXB_HOST);
+    ds = qdev_new(type == CXL ? TYPE_PXB_CXL_HOST : TYPE_PXB_HOST);
      if (type == PCIE) {
          bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
      } else if (type == CXL) {
          bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_CXL_BUS);
          bus->flags |= PCI_BUS_CXL;
+        PXB_CXL_DEV(dev)->cxl.cxl_host_bridge = PXB_CXL_HOST(ds);
      } else {
          bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
          bds = qdev_new("pci-bridge");
@@ -475,6 +534,7 @@ static void pxb_register_types(void)
      type_register_static(&pxb_pcie_bus_info);
      type_register_static(&pxb_cxl_bus_info);
      type_register_static(&pxb_host_info);
+    type_register_static(&cxl_host_info);
      type_register_static(&pxb_dev_info);
      type_register_static(&pxb_pcie_dev_info);
      type_register_static(&pxb_cxl_dev_info);
include/hw/cxl/cxl.h +0 -0
index 31af92fd5e..75e5bf71e1 100644

include/hw/cxl/cxl.h line 26

Quoted 10 lines
include/hw/cxl/cxl.h +4 -0
@@ -17,8 +17,12 @@
  #define CXL_COMPONENT_REG_BAR_IDX 0
  #define CXL_DEVICE_REG_BAR_IDX 2
 
+#define CXL_WINDOW_MAX 10
+
  typedef struct CXLState {
      bool is_enabled;
+    MemoryRegion host_mr;
+    unsigned int next_mr_idx;
  } CXLState;

... so simply drop the typedef and place OBJECT_DECLARE_SIMPLE_TYPE here instead.

include/hw/cxl/cxl.h line 27

Quoted 1 line
include/hw/cxl/cxl.h +0 -0
  #endif

ATB,

Mark.

#43 lore raw

Re: [PATCH v8 23/46] hw/cxl/component: Implement host bridge MMIO (8.2.5, table 142)

Quoted 18 lines · On Sat, 19 Mar 2022 08:35:54 +0000 Mark Cave-Ayland…
Quoted 16 lines · Jonathan Cameron

From: Ben Widawsky <[email protected]>

CXL host bridges themselves may have MMIO. Since host bridges don't have a BAR they are treated as special for MMIO. This patch includes i386/pc support.
Also hook up the device reset now that we have have the MMIO space in which the results are visible.

Note that we duplicate the PCI express case for the aml_build but the implementations will diverge when the CXL specific _OSC is introduced.

...

Quoted 16 lines
Quoted 16 lines
hw/pci-bridge/pci_expander_bridge.c +3 -0
index f762eb4a6e..b3b5f93650 100644
@@ -75,6 +75,9 @@ struct PXBDev {
      uint8_t bus_nr;
      uint16_t numa_node;
      bool bypass_iommu;
+    struct cxl_dev {
+        CXLHost *cxl_host_bridge;
+    } cxl;
  };
 
  static PXBDev *convert_to_pxb(PCIDevice *dev)
@@ -92,6 +95,9 @@ static GList *pxb_dev_list;
 
Quoted 8 lines

hw/pci-bridge/pci_expander_bridge.c static GList *pxb_dev_list; · line 99

Quoted 4 lines
hw/pci-bridge/pci_expander_bridge.c +2 -0
  #define TYPE_PXB_HOST "pxb-host"
 
+#define TYPE_PXB_CXL_HOST "pxb-cxl-host"
+#define PXB_CXL_HOST(obj) OBJECT_CHECK(CXLHost, (obj), TYPE_PXB_CXL_HOST)  

Again this is a legacy QOM type declaration and should be replaced with OBJECT_DECLARE_SIMPLE_TYPE(). It should also be located in the associated header file if it exists (include/hw/cxl/cxl.h in this case).

Sure, though as noted below that means dragging the structure there as well. I'm not quite sure why we want to put this in a header given it's only used locally in this file, but fair enough that is best practice.

Quoted 101 lines
Quoted 101 lines
hw/pci-bridge/pci_expander_bridge.c +55 -1
  static int pxb_bus_num(PCIBus *bus)
  {
      PXBDev *pxb = convert_to_pxb(bus->parent_dev);
@@ -197,6 +203,52 @@ static const TypeInfo pxb_host_info = {
      .class_init    = pxb_host_class_init,
  };
 
+static void pxb_cxl_realize(DeviceState *dev, Error **errp)
+{
+    MachineState *ms = MACHINE(qdev_get_machine());
+    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
+    CXLHost *cxl = PXB_CXL_HOST(dev);
+    CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
+    struct MemoryRegion *mr = &cxl_cstate->crb.component_registers;
+    hwaddr offset;
+
+    cxl_component_register_block_init(OBJECT(dev), cxl_cstate,
+                                      TYPE_PXB_CXL_HOST);
+    sysbus_init_mmio(sbd, mr);
+
+    offset = memory_region_size(mr) * ms->cxl_devices_state->next_mr_idx;
+    if (offset > memory_region_size(&ms->cxl_devices_state->host_mr)) {
+        error_setg(errp, "Insufficient space for pxb cxl host register space");
+        return;
+    }
+
+    memory_region_add_subregion(&ms->cxl_devices_state->host_mr, offset, mr);
+    ms->cxl_devices_state->next_mr_idx++;
+}
+
+static void pxb_cxl_host_class_init(ObjectClass *class, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(class);
+    PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(class);
+
+    hc->root_bus_path = pxb_host_root_bus_path;
+    dc->fw_name = "cxl";
+    dc->realize = pxb_cxl_realize;
+    /* Reason: Internal part of the pxb/pxb-pcie device, not usable by itself */
+    dc->user_creatable = false;
+}
+
+/*
+ * This is a device to handle the MMIO for a CXL host bridge. It does nothing
+ * else.
+ */
+static const TypeInfo cxl_host_info = {
+    .name          = TYPE_PXB_CXL_HOST,
+    .parent        = TYPE_PCI_HOST_BRIDGE,
+    .instance_size = sizeof(CXLHost),
+    .class_init    = pxb_cxl_host_class_init,
+};
+
  /*
   * Registers the PXB bus as a child of pci host root bus.
   */
@@ -245,6 +297,12 @@ static int pxb_map_irq_fn(PCIDevice *pci_dev, int pin)
 
  static void pxb_dev_reset(DeviceState *dev)
  {
+    CXLHost *cxl = PXB_CXL_DEV(dev)->cxl.cxl_host_bridge;
+    CXLComponentState *cxl_cstate = &cxl->cxl_cstate;
+    uint32_t *reg_state = cxl_cstate->crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_ROOT_PORT);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 8);
  }
 
  static gint pxb_compare(gconstpointer a, gconstpointer b)
@@ -281,12 +339,13 @@ static void pxb_dev_realize_common(PCIDevice *dev, enum BusType type,
          dev_name = dev->qdev.id;
      }
 
-    ds = qdev_new(TYPE_PXB_HOST);
+    ds = qdev_new(type == CXL ? TYPE_PXB_CXL_HOST : TYPE_PXB_HOST);
      if (type == PCIE) {
          bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_PCIE_BUS);
      } else if (type == CXL) {
          bus = pci_root_bus_new(ds, dev_name, NULL, NULL, 0, TYPE_PXB_CXL_BUS);
          bus->flags |= PCI_BUS_CXL;
+        PXB_CXL_DEV(dev)->cxl.cxl_host_bridge = PXB_CXL_HOST(ds);
      } else {
          bus = pci_root_bus_new(ds, "pxb-internal", NULL, NULL, 0, TYPE_PXB_BUS);
          bds = qdev_new("pci-bridge");
@@ -475,6 +534,7 @@ static void pxb_register_types(void)
      type_register_static(&pxb_pcie_bus_info);
      type_register_static(&pxb_cxl_bus_info);
      type_register_static(&pxb_host_info);
+    type_register_static(&cxl_host_info);
      type_register_static(&pxb_dev_info);
      type_register_static(&pxb_pcie_dev_info);
      type_register_static(&pxb_cxl_dev_info);
include/hw/cxl/cxl.h +0 -0
index 31af92fd5e..75e5bf71e1 100644
@@ -17,8 +17,12 @@
  #define CXL_COMPONENT_REG_BAR_IDX 0
  #define CXL_DEVICE_REG_BAR_IDX 2
 
Quoted 9 lines

include/hw/cxl/cxl.h line 26

Quoted 7 lines
include/hw/cxl/cxl.h +4 -0
+#define CXL_WINDOW_MAX 10
+
  typedef struct CXLState {
      bool is_enabled;
+    MemoryRegion host_mr;
+    unsigned int next_mr_idx;
  } CXLState;  

... so simply drop the typedef and place OBJECT_DECLARE_SIMPLE_TYPE here instead.

This isn't the right structure.

I've dragged the
struct CXLHost definition to here.

which need an additional include but that's fine.

Quoted 7 lines

include/hw/cxl/cxl.h line 27

Quoted 2 lines
include/hw/cxl/cxl.h +0 -0
  #endif  

ATB,

Mark.

#44 lore raw

PATCH v8 24/46 acpi/cxl: Add _OSC implementation (9.14.2)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

CXL 2.0 specification adds 2 new dwords to the existing _OSC definition from PCIe. The new dwords are accessed with a new uuid. This implementation supports what is in the specification.

iasl -d decodes the result of this patch as:

Name (SUPP, Zero)
Name (CTRL, Zero)
Name (SUPC, Zero)
Name (CTRC, Zero)
Method (_OSC, 4, NotSerialized) // _OSC: Operating System Capabilities

{
    CreateDWordField (Arg3, Zero, CDW1)
    If (((Arg0 == ToUUID ("33db4d5b-1ff7-401c-9657-7441c03dd766") /* PCI Host Bridge Device */) || (Arg0 == ToUUID ("68f2d50b-c469-4d8a-bd3d-941a103fd3fc") /* Unknown UUID */)))
    {
        CreateDWordField (Arg3, 0x04, CDW2)
        CreateDWordField (Arg3, 0x08, CDW3)
        Local0 = CDW3 /* \_SB_.PC0C._OSC.CDW3 */
        Local0 &= 0x1F
        If ((Arg1 != One))
        {
            CDW1 |= 0x08
        }

        If ((CDW3 != Local0))
        {
            CDW1 |= 0x10
        }

        SUPP = CDW2 /* \_SB_.PC0C._OSC.CDW2 */
        CTRL = CDW3 /* \_SB_.PC0C._OSC.CDW3 */
        CDW3 = Local0
        If ((Arg0 == ToUUID ("68f2d50b-c469-4d8a-bd3d-941a103fd3fc") /* Unknown UUID */))
        {
            CreateDWordField (Arg3, 0x0C, CDW4)
            CreateDWordField (Arg3, 0x10, CDW5)
            SUPC = CDW4 /* \_SB_.PC0C._OSC.CDW4 */
            CTRC = CDW5 /* \_SB_.PC0C._OSC.CDW5 */
            CDW5 |= One
        }

        Return (Arg3)
    }
    Else
    {
        CDW1 |= 0x04
        Return (Arg3)
    }
---
 hw/acpi/Kconfig       |   5 ++
 hw/acpi/cxl-stub.c    |  12 ++++
 hw/acpi/cxl.c         | 130 ++++++++++++++++++++++++++++++++++++++++++
 hw/acpi/meson.build   |   4 +-
 hw/i386/acpi-build.c  |  15 +++--
 include/hw/acpi/cxl.h |  23 ++++++++
 6 files changed, 183 insertions(+), 6 deletions(-)
hw/acpi/Kconfig +5 -0
index 19caebde6c..3703aca212 100644
@@ -5,6 +5,7 @@ config ACPI_X86
     bool
     select ACPI
     select ACPI_NVDIMM
+    select ACPI_CXL
     select ACPI_CPU_HOTPLUG
     select ACPI_MEMORY_HOTPLUG
     select ACPI_HMAT
@@ -66,3 +67,7 @@ config ACPI_ERST
     bool
     default y
     depends on ACPI && PCI
+
+config ACPI_CXL
+    bool
+    depends on ACPI
hw/acpi/cxl-stub.c new file +12 -0
new file mode 100644
index 0000000000..15bc21076b
@@ -0,0 +1,12 @@
+
+/*
+ * Stubs for ACPI platforms that don't support CXl
+ */
+#include "qemu/osdep.h"
+#include "hw/acpi/aml-build.h"
+#include "hw/acpi/cxl.h"
+
+void build_cxl_osc_method(Aml *dev)
+{
+    g_assert_not_reached();
+}
hw/acpi/cxl.c new file +130 -0

130 lines not on this page · raw message

hw/acpi/meson.build +3 -1
index 8bea2e6933..cea2f5f93a 100644
@@ -13,6 +13,7 @@ acpi_ss.add(when: 'CONFIG_ACPI_MEMORY_HOTPLUG', if_false: files('acpi-mem-hotplu
 acpi_ss.add(when: 'CONFIG_ACPI_NVDIMM', if_true: files('nvdimm.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_NVDIMM', if_false: files('acpi-nvdimm-stub.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_PCI', if_true: files('pci.c'))
+acpi_ss.add(when: 'CONFIG_ACPI_CXL', if_true: files('cxl.c'), if_false: files('cxl-stub.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_VMGENID', if_true: files('vmgenid.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_HW_REDUCED', if_true: files('generic_event_device.c'))
 acpi_ss.add(when: 'CONFIG_ACPI_HMAT', if_true: files('hmat.c'))
@@ -33,4 +34,5 @@ softmmu_ss.add_all(when: 'CONFIG_ACPI', if_true: acpi_ss)
 softmmu_ss.add(when: 'CONFIG_ALL', if_true: files('acpi-stub.c', 'aml-build-stub.c',
                                                   'acpi-x86-stub.c', 'ipmi-stub.c', 'ghes-stub.c',
                                                   'acpi-mem-hotplug-stub.c', 'acpi-cpu-hotplug-stub.c',
-                                                  'acpi-pci-hotplug-stub.c', 'acpi-nvdimm-stub.c'))
+                                                  'acpi-pci-hotplug-stub.c', 'acpi-nvdimm-stub.c',
+                                                  'cxl-stub.c'))
hw/i386/acpi-build.c +10 -5
index 0b7d06e107..042532bd6d 100644
@@ -67,6 +67,7 @@
 #include "hw/acpi/aml-build.h"
 #include "hw/acpi/utils.h"
 #include "hw/acpi/pci.h"
+#include "hw/acpi/cxl.h"
 
 #include "qom/qom-qobject.h"
 #include "hw/i386/amd_iommu.h"
@@ -1582,11 +1583,15 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
             if (pci_bus_is_cxl(bus)) {
-                aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
-                aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
-
-                /* Expander bridges do not have ACPI PCI Hot-plug enabled */
-                aml_append(dev, build_q35_osc_method(true));
+                struct Aml *pkg = aml_package(2);
+
+                aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0016")));
+                aml_append(pkg, aml_eisaid("PNP0A08"));
+                aml_append(pkg, aml_eisaid("PNP0A03"));
+                aml_append(dev, aml_name_decl("_CID", pkg));
+                aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
+                aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
+                build_cxl_osc_method(dev);
             } else if (pci_bus_is_express(bus)) {
                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
include/hw/acpi/cxl.h new file +23 -0
new file mode 100644
index 0000000000..7b8f3b8a2e
@@ -0,0 +1,23 @@
+/*
+ * Copyright (C) 2020 Intel Corporation
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef HW_ACPI_CXL_H
+#define HW_ACPI_CXL_H
+
+void build_cxl_osc_method(Aml *dev);
+
+#endif
Signature · 1 line
2.32.0
#45 lore raw

PATCH v8 25/46 acpi/cxl: Create the CEDT (9.14.1)

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

The CXL Early Discovery Table is defined in the CXL 2.0 specification as a way for the OS to get CXL specific information from the system firmware.

CXL 2.0 specification adds an _HID, ACPI0016, for CXL capable host bridges, with a _CID of PNP0A08 (PCIe host bridge). CXL aware software is able to use this initiate the proper _OSC method, and get the _UID which is referenced by the CEDT. Therefore the existence of an ACPI0016 device allows a CXL aware driver perform the necessary actions. For a CXL capable OS, this works. For a CXL unaware OS, this works.

CEDT awaremess requires more. The motivation for ACPI0017 is to provide the possibility of having a Linux CXL module that can work on a legacy Linux kernel. Linux core PCI/ACPI which won't be built as a module, will see the _CID of PNP0A08 and bind a driver to it. If we later loaded a driver for ACPI0016, Linux won't be able to bind it to the hardware because it has already bound the PNP0A08 driver. The ACPI0017 device is an opportunity to have an object to bind a driver will be used by a Linux driver to walk the CXL topology and do everything that we would have preferred to do with ACPI0016.

There is another motivation for an ACPI0017 device which isn't implemented here. An operating system needs an attach point for a non-volatile region provider that understands cross-hostbridge interleaving. Since QEMU emulation doesn't support interleaving yet, this is more important on the OS side, for now.

As of CXL 2.0 spec, only 1 sub structure is defined, the CXL Host Bridge Structure (CHBS) which is primarily useful for telling the OS exactly where the MMIO for the host bridge is.

---
 hw/acpi/cxl.c                       | 68 +++++++++++++++++++++++++++++
 hw/i386/acpi-build.c                | 27 ++++++++++++
 hw/pci-bridge/pci_expander_bridge.c | 17 --------
 include/hw/acpi/cxl.h               |  5 +++
 include/hw/pci/pci_bridge.h         | 20 +++++++++
 5 files changed, 120 insertions(+), 17 deletions(-)
hw/acpi/cxl.c +68 -0
index ca1f04f359..aa4af86a4c 100644
@@ -18,7 +18,11 @@
  */
 
 #include "qemu/osdep.h"
+#include "hw/sysbus.h"
+#include "hw/pci/pci_bridge.h"
+#include "hw/pci/pci_host.h"
 #include "hw/cxl/cxl.h"
+#include "hw/mem/memory-device.h"
 #include "hw/acpi/acpi.h"
 #include "hw/acpi/aml-build.h"
 #include "hw/acpi/bios-linker-loader.h"
@@ -26,6 +30,70 @@
 #include "qapi/error.h"
 #include "qemu/uuid.h"
 
+static void cedt_build_chbs(GArray *table_data, PXBDev *cxl)
+{
+    SysBusDevice *sbd = SYS_BUS_DEVICE(cxl->cxl.cxl_host_bridge);
+    struct MemoryRegion *mr = sbd->mmio[0].memory;
+
+    /* Type */
+    build_append_int_noprefix(table_data, 0, 1);
+
+    /* Reserved */
+    build_append_int_noprefix(table_data, 0, 1);
+
+    /* Record Length */
+    build_append_int_noprefix(table_data, 32, 2);
+
+    /* UID - currently equal to bus number */
+    build_append_int_noprefix(table_data, cxl->bus_nr, 4);
+
+    /* Version */
+    build_append_int_noprefix(table_data, 1, 4);
+
+    /* Reserved */
+    build_append_int_noprefix(table_data, 0, 4);
+
+    /* Base - subregion within a container that is in PA space */
+    build_append_int_noprefix(table_data, mr->container->addr + mr->addr, 8);
+
+    /* Length */
+    build_append_int_noprefix(table_data, memory_region_size(mr), 8);
+}
+
+static int cxl_foreach_pxb_hb(Object *obj, void *opaque)
+{
+    Aml *cedt = opaque;
+
+    if (object_dynamic_cast(obj, TYPE_PXB_CXL_DEVICE)) {
+        cedt_build_chbs(cedt->buf, PXB_CXL_DEV(obj));
+    }
+
+    return 0;
+}
+
+void cxl_build_cedt(MachineState *ms, GArray *table_offsets, GArray *table_data,
+                    BIOSLinker *linker, const char *oem_id,
+                    const char *oem_table_id)
+{
+    Aml *cedt;
+    AcpiTable table = { .sig = "CEDT", .rev = 1, .oem_id = oem_id,
+                        .oem_table_id = oem_table_id };
+
+    acpi_add_table(table_offsets, table_data);
+    acpi_table_begin(&table, table_data);
+    cedt = init_aml_allocator();
+
+    /* reserve space for CEDT header */
+
+    object_child_foreach_recursive(object_get_root(), cxl_foreach_pxb_hb, cedt);
+
+    /* copy AML table into ACPI tables blob and patch header there */
+    g_array_append_vals(table_data, cedt->buf->data, cedt->buf->len);
+    free_aml_allocator();
+
+    acpi_table_end(linker, &table);
+}
+
 static Aml *__build_cxl_osc_method(void)
 {
     Aml *method, *if_uuid, *else_uuid, *if_arg1_not_1, *if_cxl, *if_caps_masked;
hw/i386/acpi-build.c +27 -0
index 042532bd6d..1b49190663 100644
@@ -77,6 +77,7 @@
 #include "hw/acpi/ipmi.h"
 #include "hw/acpi/hmat.h"
 #include "hw/acpi/viot.h"
+#include "hw/acpi/cxl.h"
 
 #include CONFIG_DEVICES
 
@@ -1411,6 +1412,22 @@ static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
     aml_append(table, scope);
 }
 
+static void build_acpi0017(Aml *table)
+{
+    Aml *dev, *scope, *method;
+
+    scope =  aml_scope("_SB");
+    dev = aml_device("CXLM");
+    aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0017")));
+
+    method = aml_method("_STA", 0, AML_NOTSERIALIZED);
+    aml_append(method, aml_return(aml_int(0x01)));
+    aml_append(dev, method);
+
+    aml_append(scope, dev);
+    aml_append(table, scope);
+}
+
 static void
 build_dsdt(GArray *table_data, BIOSLinker *linker,
            AcpiPmInfo *pm, AcpiMiscInfo *misc,
@@ -1430,6 +1447,7 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
 #ifdef CONFIG_TPM
     TPMIf *tpm = tpm_find();
 #endif
+    bool cxl_present = false;
     int i;
     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
     AcpiTable table = { .sig = "DSDT", .rev = 1, .oem_id = x86ms->oem_id,
@@ -1618,12 +1636,17 @@ build_dsdt(GArray *table_data, BIOSLinker *linker,
                 MemoryRegion *mr = &machine->cxl_devices_state->host_mr;
                 uint64_t base = mr->addr;
 
+                cxl_present = true;
                 crs_range_insert(crs_range_set.mem_ranges, base,
                                  base + memory_region_size(mr) - 1);
             }
         }
     }
 
+    if (cxl_present) {
+        build_acpi0017(dsdt);
+    }
+
     /*
      * At this point crs_range_set has all the ranges used by pci
      * busses *other* than PCI0.  These ranges will be excluded from
@@ -2688,6 +2711,10 @@ void acpi_build(AcpiBuildTables *tables, MachineState *machine)
                           machine->nvdimms_state, machine->ram_slots,
                           x86ms->oem_id, x86ms->oem_table_id);
     }
+    if (machine->cxl_devices_state->is_enabled) {
+        cxl_build_cedt(machine, table_offsets, tables_blob, tables->linker,
+                       x86ms->oem_id, x86ms->oem_table_id);
+    }
 
     acpi_add_table(table_offsets, tables_blob);
     build_waet(tables_blob, tables->linker, x86ms->oem_id, x86ms->oem_table_id);
hw/pci-bridge/pci_expander_bridge.c +0 -17
index b3b5f93650..e11a967916 100644
@@ -57,29 +57,12 @@ DECLARE_INSTANCE_CHECKER(PXBDev, PXB_DEV,
 DECLARE_INSTANCE_CHECKER(PXBDev, PXB_PCIE_DEV,
                          TYPE_PXB_PCIE_DEVICE)
 
-#define TYPE_PXB_CXL_DEVICE "pxb-cxl"
-DECLARE_INSTANCE_CHECKER(PXBDev, PXB_CXL_DEV,
-                         TYPE_PXB_CXL_DEVICE)
-
 typedef struct CXLHost {
     PCIHostState parent_obj;
 
     CXLComponentState cxl_cstate;
 } CXLHost;
 
-struct PXBDev {
-    /*< private >*/
-    PCIDevice parent_obj;
-    /*< public >*/
-
-    uint8_t bus_nr;
-    uint16_t numa_node;
-    bool bypass_iommu;
-    struct cxl_dev {
-        CXLHost *cxl_host_bridge;
-    } cxl;
-};
-
 static PXBDev *convert_to_pxb(PCIDevice *dev)
 {
     /* A CXL PXB's parent bus is PCIe, so the normal check won't work */
include/hw/acpi/cxl.h +5 -0
index 7b8f3b8a2e..0c496538c0 100644
@@ -18,6 +18,11 @@
 #ifndef HW_ACPI_CXL_H
 #define HW_ACPI_CXL_H
 
+#include "hw/acpi/bios-linker-loader.h"
+
+void cxl_build_cedt(MachineState *ms, GArray *table_offsets, GArray *table_data,
+                    BIOSLinker *linker, const char *oem_id,
+                    const char *oem_table_id);
 void build_cxl_osc_method(Aml *dev);
 
 #endif
include/hw/pci/pci_bridge.h +20 -0
index 30691a6e57..ba4bafac7c 100644
@@ -28,6 +28,7 @@
 
 #include "hw/pci/pci.h"
 #include "hw/pci/pci_bus.h"
+#include "hw/cxl/cxl.h"
 #include "qom/object.h"
 
 typedef struct PCIBridgeWindows PCIBridgeWindows;
@@ -80,6 +81,25 @@ struct PCIBridge {
 #define PCI_BRIDGE_DEV_PROP_CHASSIS_NR "chassis_nr"
 #define PCI_BRIDGE_DEV_PROP_MSI        "msi"
 #define PCI_BRIDGE_DEV_PROP_SHPC       "shpc"
+typedef struct CXLHost CXLHost;
+
+struct PXBDev {
+    /*< private >*/
+    PCIDevice parent_obj;
+    /*< public >*/
+
+    uint8_t bus_nr;
+    uint16_t numa_node;
+    bool bypass_iommu;
+    struct cxl_dev {
+        CXLHost *cxl_host_bridge; /* Pointer to a CXLHost */
+    } cxl;
+};
+
+typedef struct PXBDev PXBDev;
+#define TYPE_PXB_CXL_DEVICE "pxb-cxl"
+DECLARE_INSTANCE_CHECKER(PXBDev, PXB_CXL_DEV,
+                         TYPE_PXB_CXL_DEVICE)
 
 int pci_bridge_ssvid_init(PCIDevice *dev, uint8_t offset,
                           uint16_t svid, uint16_t ssid,
Signature · 1 line
2.32.0
#46 lore raw

PATCH v8 26/46 hw/cxl/component: Add utils for interleave parameter encoding/decoding

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Both registers and the CFMWS entries in CDAT use simple encodings for the number of interleave ways and the interleave granularity. Introduce simple conversion functions to/from the unencoded number / size. So far the iw decode has not been needed so is it not implemented.

---
 hw/cxl/cxl-component-utils.c   | 34 ++++++++++++++++++++++++++++++++++
 include/hw/cxl/cxl_component.h |  8 ++++++++
 2 files changed, 42 insertions(+)
hw/cxl/cxl-component-utils.c +34 -0
index 410f8ef328..443a11c837 100644
@@ -9,6 +9,7 @@
 
 #include "qemu/osdep.h"
 #include "qemu/log.h"
+#include "qapi/error.h"
 #include "hw/pci/pci.h"
 #include "hw/cxl/cxl.h"
 
@@ -217,3 +218,36 @@ void cxl_component_create_dvsec(CXLComponentState *cxl, uint16_t length,
     range_init_nofail(&cxl->dvsecs[type], cxl->dvsec_offset, length);
     cxl->dvsec_offset += length;
 }
+
+uint8_t cxl_interleave_ways_enc(int iw, Error **errp)
+{
+    switch (iw) {
+    case 1: return 0x0;
+    case 2: return 0x1;
+    case 4: return 0x2;
+    case 8: return 0x3;
+    case 16: return 0x4;
+    case 3: return 0x8;
+    case 6: return 0x9;
+    case 12: return 0xa;
+    default:
+        error_setg(errp, "Interleave ways: %d not supported", iw);
+        return 0;
+    }
+}
+
+uint8_t cxl_interleave_granularity_enc(uint64_t gran, Error **errp)
+{
+    switch (gran) {
+    case 256: return 0;
+    case 512: return 1;
+    case 1024: return 2;
+    case 2048: return 3;
+    case 4096: return 4;
+    case 8192: return 5;
+    case 16384: return 6;
+    default:
+        error_setg(errp, "Interleave granularity: %" PRIu64 " invalid", gran);
+        return 0;
+    }
+}
include/hw/cxl/cxl_component.h +8 -0
index 74e9bfe1ff..8dae21cfc6 100644
@@ -194,4 +194,12 @@ void cxl_component_register_init_common(uint32_t *reg_state,
 void cxl_component_create_dvsec(CXLComponentState *cxl_cstate, uint16_t length,
                                 uint16_t type, uint8_t rev, uint8_t *body);
 
+uint8_t cxl_interleave_ways_enc(int iw, Error **errp);
+uint8_t cxl_interleave_granularity_enc(uint64_t gran, Error **errp);
+
+static inline hwaddr cxl_decode_ig(int ig)
+{
+    return 1 << (ig + 8);
+}
+
 #endif
Signature · 1 line
2.32.0
#47 lore raw

PATCH v8 27/46 hw/cxl/host: Add support for CXL Fixed Memory Windows.

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

The concept of these is introduced in [1] in terms of the description the CEDT ACPI table. The principal is more general. Unlike once traffic hits the CXL root bridges, the host system memory address routing is implementation defined and effectively static once observable by standard / generic system software. Each CXL Fixed Memory Windows (CFMW) is a region of PA space which has fixed system dependent routing configured so that accesses can be routed to the CXL devices below a set of target root bridges. The accesses may be interleaved across multiple root bridges.

For QEMU we could have fully specified these regions in terms of a base PA + size, but as the absolute address does not matter it is simpler to let individual platforms place the memory regions.

ExampleS:
-cxl-fixed-memory-window targets.0=cxl.0,size=128G
-cxl-fixed-memory-window targets.0=cxl.1,size=128G
-cxl-fixed-memory-window targets.0=cxl0,targets.1=cxl.1,size=256G,interleave-granularity=2k

Specifies
* 2x 128G regions not interleaved across root bridges, one for each of the root bridges with ids cxl.0 and cxl.1
* 256G region interleaved across root bridges with ids cxl.0 and cxl.1 with a 2k interleave granularity.

When system software enumerates the devices below a given root bridge it can then decide which CFMW to use. If non interleave is desired (or possible) it can use the appropriate CFMW for the root bridge in question. If there are suitable devices to interleave across the two root bridges then it may use the 3rd CFMS.

A number of other designs were considered but the following constraints made it hard to adapt existing QEMU approaches to this particular problem.
1) The size must be known before a specific architecture / board brings up it's PA memory map. We need to set up an appropriate region.
2) Using links to the host bridges provides a clean command line interface but these links cannot be established until command line devices have been added.

Hence the two step process used here of first establishing the size, interleave-ways and granularity + caching the ids of the host bridges and then, once available finding the actual host bridges so they can be used later to support interleave decoding.

[1] CXL 2.0 ECN: CEDT CFMWS & QTG DSM (computeexpresslink.org / specifications)

---
 hw/cxl/cxl-host-stubs.c | 14 ++++++
 hw/cxl/cxl-host.c       | 94 +++++++++++++++++++++++++++++++++++++++++
 hw/cxl/meson.build      |  6 +++
 include/hw/cxl/cxl.h    | 20 +++++++++
 qapi/machine.json       | 18 ++++++++
 qemu-options.hx         | 38 +++++++++++++++++
 softmmu/vl.c            | 44 +++++++++++++++++++
 7 files changed, 234 insertions(+)
hw/cxl/cxl-host-stubs.c new file +14 -0
new file mode 100644
index 0000000000..d24282ec1c
@@ -0,0 +1,14 @@
+/*
+ * CXL host parameter parsing routine stubs
+ *
+ * Copyright (c) 2022 Huawei
+ */
+#include "qemu/osdep.h"
+#include "qapi/error.h"
+#include "hw/cxl/cxl.h"
+
+void cxl_fixed_memory_window_options_set(MachineState *ms,
+                                         CXLFixedMemoryWindowOptions *object,
+                                         Error **errp) {};
+
+void cxl_fixed_memory_window_link_targets(Error **errp) {};
hw/cxl/cxl-host.c new file +94 -0

94 lines not on this page · raw message

hw/cxl/meson.build +6 -0
index e68eea2358..f117b99949 100644
@@ -3,4 +3,10 @@ softmmu_ss.add(when: 'CONFIG_CXL',
                    'cxl-component-utils.c',
                    'cxl-device-utils.c',
                    'cxl-mailbox-utils.c',
+                   'cxl-host.c',
+               ),
+               if_false: files(
+                   'cxl-host-stubs.c',
                ))
+
+softmmu_ss.add(when: 'CONFIG_ALL', if_true: files('cxl-host-stubs.c'))
include/hw/cxl/cxl.h +20 -0
index 75e5bf71e1..5abc307ef4 100644
@@ -10,6 +10,8 @@
 #ifndef CXL_H
 #define CXL_H
 
+#include "qapi/qapi-types-machine.h"
+#include "hw/pci/pci_bridge.h"
 #include "cxl_pci.h"
 #include "cxl_component.h"
 #include "cxl_device.h"
@@ -19,10 +21,28 @@
 
 #define CXL_WINDOW_MAX 10
 
+typedef struct CXLFixedWindow {
+    uint64_t size;
+    char **targets;
+    struct PXBDev *target_hbs[8];
+    uint8_t num_targets;
+    uint8_t enc_int_ways;
+    uint8_t enc_int_gran;
+    /* Todo: XOR based interleaving */
+    MemoryRegion mr;
+    hwaddr base;
+} CXLFixedWindow;
+
 typedef struct CXLState {
     bool is_enabled;
     MemoryRegion host_mr;
     unsigned int next_mr_idx;
+    GList *fixed_windows;
 } CXLState;
 
+void cxl_fixed_memory_window_options_set(MachineState *ms,
+                                         CXLFixedMemoryWindowOptions *object,
+                                         Error **errp);
+void cxl_fixed_memory_window_link_targets(Error **errp);
+
 #endif
qapi/machine.json +18 -0
index 42fc68403d..e4e64096ca 100644
@@ -504,6 +504,24 @@
    'dst': 'uint16',
    'val': 'uint8' }}
 
+##
+# @CXLFixedMemoryWindowOptions:
+#
+# Create a CXL Fixed Memory Window (for OptsVisitor)
+#
+# @size: Size in bytes of the Fixed Memory Window
+# @interleave-granularity: Number of contiguous bytes for which
+#                          accesses will go to a given interleave target.
+# @targets: Target root bridge IDs
+#
+# Since 6.3
+##
+{ 'struct': 'CXLFixedMemoryWindowOptions',
+  'data': {
+      'size': 'size',
+      '*interleave-granularity': 'size',
+      'targets': ['str'] }}
+
 ##
 # @X86CPURegister32:
 #
qemu-options.hx +38 -0
index 58f2f76775..764f57606d 100644
@@ -467,6 +467,44 @@ SRST
         -numa hmat-cache,node-id=1,size=10K,level=1,associativity=direct,policy=write-back,line=8
 ERST
 
+DEF("cxl-fixed-memory-window", HAS_ARG, QEMU_OPTION_cxl_fixed_memory_window,
+    "-cxl-fixed-memory-window targets.0=firsttarget,targets.1=secondtarget,size=size[,interleave-granularity=granularity]\n",
+    QEMU_ARCH_ALL)
+SRST
+``-cxl-fixed-memory-window targets.0=firsttarget,targets.1=secondtarget,size=size[,interleave-granularity=granularity]``
+    Define a CXL Fixed Memory Window (CFMW).
+
+    Described in the CXL 2.0 ECN: CEDT CFMWS & QTG _DSM.
+
+    They are regions of Host Physical Addresses (HPA) on a system which
+    may be interleaved across one or more CXL host bridges.  The system
+    software will assign particular devices into these windows and
+    configure the downstream Host-managed Device Memory (HDM) decoders
+    in root ports, switch ports and devices appropriately to meet the
+    interleave requirements before enabling the memory devices.
+
+    ``targets.X=firsttarget`` provides the mapping to CXL host bridges
+    which may be identified by the id provied in the -device entry.
+    Multiple entries are needed to specify all the targets when
+    the fixed memory window represents interleaved memory. X is the
+    target index from 0.
+
+    ``size=size`` sets the size of the CFMW. This must be a multiple of
+    256MiB. The region will be aligned to 256MiB but the location is
+    platform and configuration dependent.
+
+    ``interleave-granularity=granularity`` sets the granularity of
+    interleave. Default 256KiB. Only 256KiB, 512KiB, 1024KiB, 2048KiB
+    4096KiB, 8192KiB and 16384KiB granularities supported.
+
+    Example:
+
+    ::
+
+        -cxl-fixed-memory-window -targets.0=cxl.0,-targets.1=cxl.1,size=128G,interleave-granularity=512k
+
+ERST
+
 DEF("add-fd", HAS_ARG, QEMU_OPTION_add_fd,
     "-add-fd fd=fd,set=set[,opaque=opaque]\n"
     "                Add 'fd' to fd 'set'\n", QEMU_ARCH_ALL)
softmmu/vl.c +44 -0

98 lines not on this page · raw message

Signature · 1 line
2.32.0
#48 lore raw

Re: [PATCH v8 27/46] hw/cxl/host: Add support for CXL Fixed Memory Windows.

Quoted 50 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

The concept of these is introduced in [1] in terms of the description the CEDT ACPI table. The principal is more general. Unlike once traffic hits the CXL root bridges, the host system memory address routing is implementation defined and effectively static once observable by standard / generic system software. Each CXL Fixed Memory Windows (CFMW) is a region of PA space which has fixed system dependent routing configured so that accesses can be routed to the CXL devices below a set of target root bridges. The accesses may be interleaved across multiple root bridges.

For QEMU we could have fully specified these regions in terms of a base PA + size, but as the absolute address does not matter it is simpler to let individual platforms place the memory regions.

ExampleS:

(quoted diff) +0 -3
-cxl-fixed-memory-window targets.0=cxl.0,size=128G
-cxl-fixed-memory-window targets.0=cxl.1,size=128G
-cxl-fixed-memory-window targets.0=cxl0,targets.1=cxl.1,size=256G,interleave-granularity=2k

Specifies
* 2x 128G regions not interleaved across root bridges, one for each of the root bridges with ids cxl.0 and cxl.1
* 256G region interleaved across root bridges with ids cxl.0 and cxl.1 with a 2k interleave granularity.

When system software enumerates the devices below a given root bridge it can then decide which CFMW to use. If non interleave is desired (or possible) it can use the appropriate CFMW for the root bridge in question. If there are suitable devices to interleave across the two root bridges then it may use the 3rd CFMS.

A number of other designs were considered but the following constraints made it hard to adapt existing QEMU approaches to this particular problem.
1) The size must be known before a specific architecture / board brings up it's PA memory map. We need to set up an appropriate region.
2) Using links to the host bridges provides a clean command line interface but these links cannot be established until command line devices have been added.

Hence the two step process used here of first establishing the size, interleave-ways and granularity + caching the ids of the host bridges and then, once available finding the actual host bridges so they can be used later to support interleave decoding.

[1] CXL 2.0 ECN: CEDT CFMWS & QTG DSM (computeexpresslink.org / specifications)

[...]

qapi/machine.json line 510

Quoted 12 lines
qapi/machine.json +4 -0
index 42fc68403d..e4e64096ca 100644
@@ -504,6 +504,24 @@
    'dst': 'uint16',
    'val': 'uint8' }}
 
+##
+# @CXLFixedMemoryWindowOptions:
+#
+# Create a CXL Fixed Memory Window (for OptsVisitor)

Please drop "(for OptsVisitor)". It's no longer true.

Aside: I'd ask to drop it even if it was true, because it's about an implementation detail. Such details don't belong in doc comments, which become QMP *user* documentation. I know we have the same parenthesis elsewhere. It should be dropped there as well, but that's not this series' problem.

qapi/machine.json line 512

Quoted 2 lines
qapi/machine.json +2 -0
+#
+# @size: Size in bytes of the Fixed Memory Window

I'm not a native speaker, but "Size of the fixed memory window in bytes" sounds better to me.

Are arbitrary sizes accepted?

qapi/machine.json line 515

Quoted 3 lines
qapi/machine.json +3 -0
+# @interleave-granularity: Number of contiguous bytes for which
+#                          accesses will go to a given interleave target.
+# @targets: Target root bridge IDs

What kind of IDs are these? Hmm, the CLI help text below suggest qdev IDs (the ones in -device id=...). Correct?

qapi/machine.json line 517

Quoted 2 lines
qapi/machine.json +2 -0
+#
+# Since 6.3

7.1 most likely.

Quoted 108 lines
qapi/machine.json +7 -0
+##
+{ 'struct': 'CXLFixedMemoryWindowOptions',
+  'data': {
+      'size': 'size',
+      '*interleave-granularity': 'size',
+      'targets': ['str'] }}
+
 ##
 # @X86CPURegister32:
 #
qemu-options.hx +38 -0
index 58f2f76775..764f57606d 100644
@@ -467,6 +467,44 @@ SRST
         -numa hmat-cache,node-id=1,size=10K,level=1,associativity=direct,policy=write-back,line=8
 ERST
 
+DEF("cxl-fixed-memory-window", HAS_ARG, QEMU_OPTION_cxl_fixed_memory_window,
+    "-cxl-fixed-memory-window targets.0=firsttarget,targets.1=secondtarget,size=size[,interleave-granularity=granularity]\n",
+    QEMU_ARCH_ALL)
+SRST
+``-cxl-fixed-memory-window targets.0=firsttarget,targets.1=secondtarget,size=size[,interleave-granularity=granularity]``
+    Define a CXL Fixed Memory Window (CFMW).
+
+    Described in the CXL 2.0 ECN: CEDT CFMWS & QTG _DSM.
+
+    They are regions of Host Physical Addresses (HPA) on a system which
+    may be interleaved across one or more CXL host bridges.  The system
+    software will assign particular devices into these windows and
+    configure the downstream Host-managed Device Memory (HDM) decoders
+    in root ports, switch ports and devices appropriately to meet the
+    interleave requirements before enabling the memory devices.
+
+    ``targets.X=firsttarget`` provides the mapping to CXL host bridges
+    which may be identified by the id provied in the -device entry.
+    Multiple entries are needed to specify all the targets when
+    the fixed memory window represents interleaved memory. X is the
+    target index from 0.
+
+    ``size=size`` sets the size of the CFMW. This must be a multiple of
+    256MiB. The region will be aligned to 256MiB but the location is
+    platform and configuration dependent.
+
+    ``interleave-granularity=granularity`` sets the granularity of
+    interleave. Default 256KiB. Only 256KiB, 512KiB, 1024KiB, 2048KiB
+    4096KiB, 8192KiB and 16384KiB granularities supported.
+
+    Example:
+
+    ::
+
+        -cxl-fixed-memory-window -targets.0=cxl.0,-targets.1=cxl.1,size=128G,interleave-granularity=512k
+
+ERST
+
 DEF("add-fd", HAS_ARG, QEMU_OPTION_add_fd,
     "-add-fd fd=fd,set=set[,opaque=opaque]\n"
     "                Add 'fd' to fd 'set'\n", QEMU_ARCH_ALL)
softmmu/vl.c +13 -0
index 0b81f61535..dab1eb3380 100644
@@ -92,6 +92,7 @@
 #include "qemu/config-file.h"
 #include "qemu/qemu-options.h"
 #include "qemu/main-loop.h"
+#include "hw/cxl/cxl.h"
 #ifdef CONFIG_VIRTFS
 #include "fsdev/qemu-fsdev.h"
 #endif
@@ -117,6 +118,7 @@
 #include "qapi/qapi-events-run-state.h"
 #include "qapi/qapi-visit-block-core.h"
 #include "qapi/qapi-visit-compat.h"
+#include "qapi/qapi-visit-machine.h"
 #include "qapi/qapi-visit-ui.h"
 #include "qapi/qapi-commands-block-core.h"
 #include "qapi/qapi-commands-migration.h"
@@ -140,6 +142,11 @@ typedef struct BlockdevOptionsQueueEntry {
 
 typedef QSIMPLEQ_HEAD(, BlockdevOptionsQueueEntry) BlockdevOptionsQueue;
 
+typedef struct CXLFMWOptionQueueEntry {
+    CXLFixedMemoryWindowOptions *opts;
+    QSIMPLEQ_ENTRY(CXLFMWOptionQueueEntry) entry;
+} CXLFMWOptionQueueEntry;
+
 typedef struct ObjectOption {
     ObjectOptions *opts;
     QTAILQ_ENTRY(ObjectOption) next;
@@ -166,6 +173,8 @@ static int snapshot;
 static bool preconfig_requested;
 static QemuPluginList plugin_list = QTAILQ_HEAD_INITIALIZER(plugin_list);
 static BlockdevOptionsQueue bdo_queue = QSIMPLEQ_HEAD_INITIALIZER(bdo_queue);
+static QSIMPLEQ_HEAD(, CXLFMWOptionQueueEntry) CXLFMW_opts =
+    QSIMPLEQ_HEAD_INITIALIZER(CXLFMW_opts);
 static bool nographic = false;
 static int mem_prealloc; /* force preallocation of physical target memory */
 static ram_addr_t ram_size;
@@ -1149,6 +1158,23 @@ static void parse_display(const char *p)
     }
 }
 
+static void parse_cxl_fixed_memory_window(const char *optarg)
+{
+    CXLFMWOptionQueueEntry *cfmws_entry;
+    Visitor *v;

softmmu/vl.c static void parse_display(const char *p) · line 1174

Quoted 10 lines
softmmu/vl.c +10 -0
+
+    v = qobject_input_visitor_new_str(optarg, "cxl-fixed-memory-window",
+                                      &error_fatal);
+    cfmws_entry = g_new(CXLFMWOptionQueueEntry, 1);
+    visit_type_CXLFixedMemoryWindowOptions(v, NULL, &cfmws_entry->opts,
+                                           &error_fatal);
+    if (!cfmws_entry->opts) {
+        exit(1);
+    }
+    visit_free(v);

If you loc_save() here like we do for QEMU_OPTION_blockdev, and ...

Quoted 7 lines
softmmu/vl.c +3 -0
+    QSIMPLEQ_INSERT_TAIL(&CXLFMW_opts, cfmws_entry, entry);
+}
+
 static inline bool nonempty_str(const char *str)
 {
     return str && *str;

softmmu/vl.c static void qemu_create_late_backends(void) · line 2054

Quoted 9 lines
softmmu/vl.c +6 -0
@@ -2020,6 +2046,19 @@ static void qemu_create_late_backends(void)
     qemu_semihosting_console_init();
 }
 
+static void cxl_set_opts(void)
+{
+    while (!QSIMPLEQ_EMPTY(&CXLFMW_opts)) {
+        CXLFMWOptionQueueEntry *cfmws_entry = QSIMPLEQ_FIRST(&CXLFMW_opts);
+
+        QSIMPLEQ_REMOVE_HEAD(&CXLFMW_opts, entry);

... loc_pop() here, like we do in configure_blockdev(), then the error messages from cxl_fixed_memory_window_options_set() will point to the offending option, which is nice. Give it a try, please.

softmmu/vl.c static void qemu_create_late_backends(void) · line 2055

Quoted 1 line
softmmu/vl.c +1 -0
+        cxl_fixed_memory_window_options_set(current_machine, cfmws_entry->opts,

Line is a bit long. Name the function cxl_fixed_memory_window_config()?

Quoted 35 lines
softmmu/vl.c +11 -0
+                                            &error_fatal);
+        qapi_free_CXLFixedMemoryWindowOptions(cfmws_entry->opts);
+        g_free(cfmws_entry);
+    }
+}
+
 static bool have_custom_ram_size(void)
 {
     QemuOpts *opts = qemu_find_opts_singleton("memory");
@@ -2745,6 +2784,7 @@ void qmp_x_exit_preconfig(Error **errp)
 
     qemu_init_board();
     qemu_create_cli_devices();
+    cxl_fixed_memory_window_link_targets(errp);
     qemu_machine_creation_done();
 
     if (loadvm) {
@@ -2925,6 +2965,9 @@ void qemu_init(int argc, char **argv, char **envp)
                     exit(1);
                 }
                 break;
+            case QEMU_OPTION_cxl_fixed_memory_window:
+                parse_cxl_fixed_memory_window(optarg);
+                break;
             case QEMU_OPTION_display:
                 parse_display(optarg);
                 break;
@@ -3762,6 +3805,7 @@ void qemu_init(int argc, char **argv, char **envp)
 
     qemu_resolve_machine_memdev();
     parse_numa_opts(current_machine);
+    cxl_set_opts();
 
     if (vmstate_dump_file) {
         /* dump and exit */
#49 lore raw

Re: [PATCH v8 27/46] hw/cxl/host: Add support for CXL Fixed Memory Windows.

Quoted 44 lines · On Mon, 28 Mar 2022 14:50:23 +0200 Markus Armbruster <[email protected]> wrote:
Quoted 42 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

The concept of these is introduced in [1] in terms of the description the CEDT ACPI table. The principal is more general. Unlike once traffic hits the CXL root bridges, the host system memory address routing is implementation defined and effectively static once observable by standard / generic system software. Each CXL Fixed Memory Windows (CFMW) is a region of PA space which has fixed system dependent routing configured so that accesses can be routed to the CXL devices below a set of target root bridges. The accesses may be interleaved across multiple root bridges.

For QEMU we could have fully specified these regions in terms of a base PA + size, but as the absolute address does not matter it is simpler to let individual platforms place the memory regions.

ExampleS:

(quoted diff) +0 -3
-cxl-fixed-memory-window targets.0=cxl.0,size=128G
-cxl-fixed-memory-window targets.0=cxl.1,size=128G
-cxl-fixed-memory-window targets.0=cxl0,targets.1=cxl.1,size=256G,interleave-granularity=2k

Specifies
* 2x 128G regions not interleaved across root bridges, one for each of the root bridges with ids cxl.0 and cxl.1
* 256G region interleaved across root bridges with ids cxl.0 and cxl.1 with a 2k interleave granularity.

When system software enumerates the devices below a given root bridge it can then decide which CFMW to use. If non interleave is desired (or possible) it can use the appropriate CFMW for the root bridge in question. If there are suitable devices to interleave across the two root bridges then it may use the 3rd CFMS.

A number of other designs were considered but the following constraints made it hard to adapt existing QEMU approaches to this particular problem.
1) The size must be known before a specific architecture / board brings up it's PA memory map. We need to set up an appropriate region.
2) Using links to the host bridges provides a clean command line interface but these links cannot be established until command line devices have been added.

Quoted 8 lines
Quoted 8 lines

Hence the two step process used here of first establishing the size, interleave-ways and granularity + caching the ids of the host bridges and then, once available finding the actual host bridges so they can be used later to support interleave decoding.

[1] CXL 2.0 ECN: CEDT CFMWS & QTG DSM (computeexpresslink.org / specifications)

Agree with everything you've raised. Thanks!

Jonathan

Quoted 11 lines

[...]

Quoted 8 lines
qapi/machine.json +0 -0
index 42fc68403d..e4e64096ca 100644
@@ -504,6 +504,24 @@
    'dst': 'uint16',
    'val': 'uint8' }}
 
Quoted 7 lines

qapi/machine.json line 510

Quoted 4 lines
qapi/machine.json +4 -0
+##
+# @CXLFixedMemoryWindowOptions:
+#
+# Create a CXL Fixed Memory Window (for OptsVisitor)  

Please drop "(for OptsVisitor)". It's no longer true.

Done

Quoted 11 lines

Aside: I'd ask to drop it even if it was true, because it's about an implementation detail. Such details don't belong in doc comments, which become QMP *user* documentation. I know we have the same parenthesis elsewhere. It should be dropped there as well, but that's not this series' problem.

qapi/machine.json line 512

Quoted 2 lines
qapi/machine.json +2 -0
+#
+# @size: Size in bytes of the Fixed Memory Window  

I'm not a native speaker, but "Size of the fixed memory window in bytes" sounds better to me.

Done.

Quoted 2 lines

Are arbitrary sizes accepted?

Good point - no. Multiple of 256MiB. Added
"Must be a multiple of 256MiB."

For interleave-granularity I've listed the possible values out as "Accepted values [256, 512, 1k, 2k, 4k, 8k, 16k]"

Quoted 7 lines

qapi/machine.json line 515

Quoted 4 lines
qapi/machine.json +3 -0
+# @interleave-granularity: Number of contiguous bytes for which
+#                          accesses will go to a given interleave target.
+# @targets: Target root bridge IDs  

What kind of IDs are these? Hmm, the CLI help text below suggest qdev IDs (the ones in -device id=...). Correct?

Yes. I've added "from -device ...,id=<ID> for each root bridge." That leaves room for other root bridge implementations than pxb-cxl in future without having to update this text.

Quoted 5 lines

qapi/machine.json line 517

Quoted 3 lines
qapi/machine.json +2 -0
+#
+# Since 6.3  

7.1 most likely.

Updated.

Quoted 132 lines
Quoted 109 lines
qapi/machine.json +7 -0
+##
+{ 'struct': 'CXLFixedMemoryWindowOptions',
+  'data': {
+      'size': 'size',
+      '*interleave-granularity': 'size',
+      'targets': ['str'] }}
+
 ##
 # @X86CPURegister32:
 #
qemu-options.hx +38 -0
index 58f2f76775..764f57606d 100644
@@ -467,6 +467,44 @@ SRST
         -numa hmat-cache,node-id=1,size=10K,level=1,associativity=direct,policy=write-back,line=8
 ERST
 
+DEF("cxl-fixed-memory-window", HAS_ARG, QEMU_OPTION_cxl_fixed_memory_window,
+    "-cxl-fixed-memory-window targets.0=firsttarget,targets.1=secondtarget,size=size[,interleave-granularity=granularity]\n",
+    QEMU_ARCH_ALL)
+SRST
+``-cxl-fixed-memory-window targets.0=firsttarget,targets.1=secondtarget,size=size[,interleave-granularity=granularity]``
+    Define a CXL Fixed Memory Window (CFMW).
+
+    Described in the CXL 2.0 ECN: CEDT CFMWS & QTG _DSM.
+
+    They are regions of Host Physical Addresses (HPA) on a system which
+    may be interleaved across one or more CXL host bridges.  The system
+    software will assign particular devices into these windows and
+    configure the downstream Host-managed Device Memory (HDM) decoders
+    in root ports, switch ports and devices appropriately to meet the
+    interleave requirements before enabling the memory devices.
+
+    ``targets.X=firsttarget`` provides the mapping to CXL host bridges
+    which may be identified by the id provied in the -device entry.
+    Multiple entries are needed to specify all the targets when
+    the fixed memory window represents interleaved memory. X is the
+    target index from 0.
+
+    ``size=size`` sets the size of the CFMW. This must be a multiple of
+    256MiB. The region will be aligned to 256MiB but the location is
+    platform and configuration dependent.
+
+    ``interleave-granularity=granularity`` sets the granularity of
+    interleave. Default 256KiB. Only 256KiB, 512KiB, 1024KiB, 2048KiB
+    4096KiB, 8192KiB and 16384KiB granularities supported.
+
+    Example:
+
+    ::
+
+        -cxl-fixed-memory-window -targets.0=cxl.0,-targets.1=cxl.1,size=128G,interleave-granularity=512k
+
+ERST
+
 DEF("add-fd", HAS_ARG, QEMU_OPTION_add_fd,
     "-add-fd fd=fd,set=set[,opaque=opaque]\n"
     "                Add 'fd' to fd 'set'\n", QEMU_ARCH_ALL)
softmmu/vl.c +13 -0
index 0b81f61535..dab1eb3380 100644
@@ -92,6 +92,7 @@
 #include "qemu/config-file.h"
 #include "qemu/qemu-options.h"
 #include "qemu/main-loop.h"
+#include "hw/cxl/cxl.h"
 #ifdef CONFIG_VIRTFS
 #include "fsdev/qemu-fsdev.h"
 #endif
@@ -117,6 +118,7 @@
 #include "qapi/qapi-events-run-state.h"
 #include "qapi/qapi-visit-block-core.h"
 #include "qapi/qapi-visit-compat.h"
+#include "qapi/qapi-visit-machine.h"
 #include "qapi/qapi-visit-ui.h"
 #include "qapi/qapi-commands-block-core.h"
 #include "qapi/qapi-commands-migration.h"
@@ -140,6 +142,11 @@ typedef struct BlockdevOptionsQueueEntry {
 
 typedef QSIMPLEQ_HEAD(, BlockdevOptionsQueueEntry) BlockdevOptionsQueue;
 
+typedef struct CXLFMWOptionQueueEntry {
+    CXLFixedMemoryWindowOptions *opts;
+    QSIMPLEQ_ENTRY(CXLFMWOptionQueueEntry) entry;
+} CXLFMWOptionQueueEntry;
+
 typedef struct ObjectOption {
     ObjectOptions *opts;
     QTAILQ_ENTRY(ObjectOption) next;
@@ -166,6 +173,8 @@ static int snapshot;
 static bool preconfig_requested;
 static QemuPluginList plugin_list = QTAILQ_HEAD_INITIALIZER(plugin_list);
 static BlockdevOptionsQueue bdo_queue = QSIMPLEQ_HEAD_INITIALIZER(bdo_queue);
+static QSIMPLEQ_HEAD(, CXLFMWOptionQueueEntry) CXLFMW_opts =
+    QSIMPLEQ_HEAD_INITIALIZER(CXLFMW_opts);
 static bool nographic = false;
 static int mem_prealloc; /* force preallocation of physical target memory */
 static ram_addr_t ram_size;
@@ -1149,6 +1158,23 @@ static void parse_display(const char *p)
     }
 }
 
+static void parse_cxl_fixed_memory_window(const char *optarg)
+{
+    CXLFMWOptionQueueEntry *cfmws_entry;
+    Visitor *v;

softmmu/vl.c static void parse_display(const char *p) · line 1174

Quoted 10 lines
softmmu/vl.c +10 -0
+
+    v = qobject_input_visitor_new_str(optarg, "cxl-fixed-memory-window",
+                                      &error_fatal);
+    cfmws_entry = g_new(CXLFMWOptionQueueEntry, 1);
+    visit_type_CXLFixedMemoryWindowOptions(v, NULL, &cfmws_entry->opts,
+                                           &error_fatal);
+    if (!cfmws_entry->opts) {
+        exit(1);
+    }
+    visit_free(v);  

If you loc_save() here like we do for QEMU_OPTION_blockdev, and ...

Quoted 10 lines
softmmu/vl.c +3 -0
+    QSIMPLEQ_INSERT_TAIL(&CXLFMW_opts, cfmws_entry, entry);
+}
+
 static inline bool nonempty_str(const char *str)
 {
     return str && *str;
@@ -2020,6 +2046,19 @@ static void qemu_create_late_backends(void)
     qemu_semihosting_console_init();
 }
 
Quoted 10 lines

softmmu/vl.c static void qemu_create_late_backends(void) · line 2054

Quoted 6 lines
softmmu/vl.c +6 -0
+static void cxl_set_opts(void)
+{
+    while (!QSIMPLEQ_EMPTY(&CXLFMW_opts)) {
+        CXLFMWOptionQueueEntry *cfmws_entry = QSIMPLEQ_FIRST(&CXLFMW_opts);
+
+        QSIMPLEQ_REMOVE_HEAD(&CXLFMW_opts, entry);  

... loc_pop() here, like we do in configure_blockdev(), then the error messages from cxl_fixed_memory_window_options_set() will point to the offending option, which is nice. Give it a try, please.

Nice :)

Quoted 4 lines

softmmu/vl.c static void qemu_create_late_backends(void) · line 2055

Quoted 2 lines
softmmu/vl.c +1 -0
+        cxl_fixed_memory_window_options_set(current_machine, cfmws_entry->opts,  

Line is a bit long. Name the function cxl_fixed_memory_window_config()?

Done

Quoted 37 lines
Quoted 36 lines
softmmu/vl.c +11 -0
+                                            &error_fatal);
+        qapi_free_CXLFixedMemoryWindowOptions(cfmws_entry->opts);
+        g_free(cfmws_entry);
+    }
+}
+
 static bool have_custom_ram_size(void)
 {
     QemuOpts *opts = qemu_find_opts_singleton("memory");
@@ -2745,6 +2784,7 @@ void qmp_x_exit_preconfig(Error **errp)
 
     qemu_init_board();
     qemu_create_cli_devices();
+    cxl_fixed_memory_window_link_targets(errp);
     qemu_machine_creation_done();
 
     if (loadvm) {
@@ -2925,6 +2965,9 @@ void qemu_init(int argc, char **argv, char **envp)
                     exit(1);
                 }
                 break;
+            case QEMU_OPTION_cxl_fixed_memory_window:
+                parse_cxl_fixed_memory_window(optarg);
+                break;
             case QEMU_OPTION_display:
                 parse_display(optarg);
                 break;
@@ -3762,6 +3805,7 @@ void qemu_init(int argc, char **argv, char **envp)
 
     qemu_resolve_machine_memdev();
     parse_numa_opts(current_machine);
+    cxl_set_opts();
 
     if (vmstate_dump_file) {
         /* dump and exit */  
#50 lore raw

PATCH v8 28/46 acpi/cxl: Introduce CFMWS structures in CEDT

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

The CEDT CXL Fixed Window Memory Window Structures (CFMWs) define regions of the host phyiscal address map which (via an impdef means) are configured such that they have a particular interleave setup across one or more CXL Host Bridges.

---
 hw/acpi/cxl.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 59 insertions(+)
hw/acpi/cxl.c +59 -0
index aa4af86a4c..31d5235136 100644
@@ -60,6 +60,64 @@ static void cedt_build_chbs(GArray *table_data, PXBDev *cxl)
     build_append_int_noprefix(table_data, memory_region_size(mr), 8);
 }
 
+/*
+ * CFMWS entries in CXL 2.0 ECN: CEDT CFMWS & QTG _DSM.
+ * Interleave ways encoding in CXL 2.0 ECN: 3, 6, 12 and 16-way memory
+ * interleaving.
+ */
+static void cedt_build_cfmws(GArray *table_data, MachineState *ms)
+{
+    CXLState *cxls = ms->cxl_devices_state;
+    GList *it;
+
+    for (it = cxls->fixed_windows; it; it = it->next) {
+        CXLFixedWindow *fw = it->data;
+        int i;
+
+        /* Type */
+        build_append_int_noprefix(table_data, 1, 1);
+
+        /* Reserved */
+        build_append_int_noprefix(table_data, 0, 1);
+
+        /* Record Length */
+        build_append_int_noprefix(table_data, 36 + 4 * fw->num_targets, 2);
+
+        /* Reserved */
+        build_append_int_noprefix(table_data, 0, 4);
+
+        /* Base HPA */
+        build_append_int_noprefix(table_data, fw->mr.addr, 8);
+
+        /* Window Size */
+        build_append_int_noprefix(table_data, fw->size, 8);
+
+        /* Host Bridge Interleave Ways */
+        build_append_int_noprefix(table_data, fw->enc_int_ways, 1);
+
+        /* Host Bridge Interleave Arithmetic */
+        build_append_int_noprefix(table_data, 0, 1);
+
+        /* Reserved */
+        build_append_int_noprefix(table_data, 0, 2);
+
+        /* Host Bridge Interleave Granularity */
+        build_append_int_noprefix(table_data, fw->enc_int_gran, 4);
+
+        /* Window Restrictions */
+        build_append_int_noprefix(table_data, 0x0f, 2); /* No restrictions */
+
+        /* QTG ID */
+        build_append_int_noprefix(table_data, 0, 2);
+
+        /* Host Bridge List (list of UIDs - currently bus_nr) */
+        for (i = 0; i < fw->num_targets; i++) {
+            g_assert(fw->target_hbs[i]);
+            build_append_int_noprefix(table_data, fw->target_hbs[i]->bus_nr, 4);
+        }
+    }
+}
+
 static int cxl_foreach_pxb_hb(Object *obj, void *opaque)
 {
     Aml *cedt = opaque;
@@ -86,6 +144,7 @@ void cxl_build_cedt(MachineState *ms, GArray *table_offsets, GArray *table_data,
     /* reserve space for CEDT header */
 
     object_child_foreach_recursive(object_get_root(), cxl_foreach_pxb_hb, cedt);
+    cedt_build_cfmws(cedt->buf, ms);
 
     /* copy AML table into ACPI tables blob and patch header there */
     g_array_append_vals(table_data, cedt->buf->data, cedt->buf->len);
Signature · 1 line
2.32.0
#51 lore raw

PATCH v8 29/46 hw/pci-host/gpex-acpi: Add support for dsdt construction for pxb-cxl

In reply to Jonathan Cameron

This adds code to instantiate the slightly extended ACPI root port description in DSDT as per the CXL 2.0 specification.

Basically a cut and paste job from the i386/pc code.

---
 hw/arm/Kconfig          |  1 +
 hw/pci-host/gpex-acpi.c | 20 +++++++++++++++++---
 2 files changed, 18 insertions(+), 3 deletions(-)
hw/arm/Kconfig +1 -0
index 97f3b38019..219262a8da 100644
@@ -29,6 +29,7 @@ config ARM_VIRT
     select ACPI_APEI
     select ACPI_VIOT
     select VIRTIO_MEM_SUPPORTED
+    select ACPI_CXL
 
 config CHEETAH
     bool
hw/pci-host/gpex-acpi.c +17 -3
index e7e162a00a..7c7316bc96 100644
@@ -5,6 +5,7 @@
 #include "hw/pci/pci_bus.h"
 #include "hw/pci/pci_bridge.h"
 #include "hw/pci/pcie_host.h"
+#include "hw/acpi/cxl.h"
 
 static void acpi_dsdt_add_pci_route_table(Aml *dev, uint32_t irq)
 {
@@ -139,6 +140,7 @@ void acpi_dsdt_add_gpex(Aml *scope, struct GPEXConfig *cfg)
         QLIST_FOREACH(bus, &bus->child, sibling) {
             uint8_t bus_num = pci_bus_num(bus);
             uint8_t numa_node = pci_bus_numa_node(bus);
+            bool is_cxl = pci_bus_is_cxl(bus);
 
             if (!pci_bus_is_root(bus)) {
                 continue;
@@ -154,8 +156,16 @@ void acpi_dsdt_add_gpex(Aml *scope, struct GPEXConfig *cfg)
             }
 
             dev = aml_device("PC%.02X", bus_num);
-            aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A08")));
-            aml_append(dev, aml_name_decl("_CID", aml_string("PNP0A03")));
+            if (is_cxl) {
+                struct Aml *pkg = aml_package(2);
+                aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0016")));
+                aml_append(pkg, aml_eisaid("PNP0A08"));
+                aml_append(pkg, aml_eisaid("PNP0A03"));
+                aml_append(dev, aml_name_decl("_CID", pkg));
+            } else {
+                aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A08")));
+                aml_append(dev, aml_name_decl("_CID", aml_string("PNP0A03")));
+            }
             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
             aml_append(dev, aml_name_decl("_STR", aml_unicode("pxb Device")));
@@ -175,7 +185,11 @@ void acpi_dsdt_add_gpex(Aml *scope, struct GPEXConfig *cfg)
                             cfg->pio.base, 0, 0, 0);
             aml_append(dev, aml_name_decl("_CRS", crs));
 
-            acpi_dsdt_add_pci_osc(dev);
+            if (is_cxl) {
+                build_cxl_osc_method(dev);
+            } else {
+                acpi_dsdt_add_pci_osc(dev);
+            }
 
             aml_append(scope, dev);
         }
Signature · 1 line
2.32.0
#52 lore raw

PATCH v8 30/46 pci/pcie_port: Add pci_find_port_by_pn()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Simple function to search a PCIBus to find a port by it's port number.

CXL interleave decoding uses the port number as a target so it is necessary to locate the port when doing interleave decoding.

---
 hw/pci/pcie_port.c         | 25 +++++++++++++++++++++++++
 include/hw/pci/pcie_port.h |  2 ++
 2 files changed, 27 insertions(+)
hw/pci/pcie_port.c +25 -0
index e95c1e5519..687e4e763a 100644
@@ -136,6 +136,31 @@ static void pcie_port_class_init(ObjectClass *oc, void *data)
     device_class_set_props(dc, pcie_port_props);
 }
 
+PCIDevice *pcie_find_port_by_pn(PCIBus *bus, uint8_t pn)
+{
+    int devfn;
+
+    for (devfn = 0; devfn < ARRAY_SIZE(bus->devices); devfn++) {
+        PCIDevice *d = bus->devices[devfn];
+        PCIEPort *port;
+
+        if (!d || !pci_is_express(d) || !d->exp.exp_cap) {
+            continue;
+        }
+
+        if (!object_dynamic_cast(OBJECT(d), TYPE_PCIE_PORT)) {
+            continue;
+        }
+
+        port = PCIE_PORT(d);
+        if (port->port == pn) {
+            return d;
+        }
+    }
+
+    return NULL;
+}
+
 static const TypeInfo pcie_port_type_info = {
     .name = TYPE_PCIE_PORT,
     .parent = TYPE_PCI_BRIDGE,
include/hw/pci/pcie_port.h +2 -0
index e25b289ce8..7b8193061a 100644
@@ -39,6 +39,8 @@ struct PCIEPort {
 
 void pcie_port_init_reg(PCIDevice *d);
 
+PCIDevice *pcie_find_port_by_pn(PCIBus *bus, uint8_t pn);
+
 #define TYPE_PCIE_SLOT "pcie-slot"
 OBJECT_DECLARE_SIMPLE_TYPE(PCIESlot, PCIE_SLOT)
 
Signature · 1 line
2.32.0
#53 lore raw

PATCH v8 31/46 CXL/cxl_component: Add cxl_get_hb_cstate()

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Accessor to get hold of the cxl state for a CXL host bridge without exposing the internals of the implementation.

---
 hw/pci-bridge/pci_expander_bridge.c | 7 +++++++
 include/hw/cxl/cxl_component.h      | 2 ++
 2 files changed, 9 insertions(+)
hw/pci-bridge/pci_expander_bridge.c +7 -0
index e11a967916..de534c44ab 100644
@@ -81,6 +81,13 @@ static GList *pxb_dev_list;
 #define TYPE_PXB_CXL_HOST "pxb-cxl-host"
 #define PXB_CXL_HOST(obj) OBJECT_CHECK(CXLHost, (obj), TYPE_PXB_CXL_HOST)
 
+CXLComponentState *cxl_get_hb_cstate(PCIHostState *hb)
+{
+    CXLHost *host = PXB_CXL_HOST(hb);
+
+    return &host->cxl_cstate;
+}
+
 static int pxb_bus_num(PCIBus *bus)
 {
     PXBDev *pxb = convert_to_pxb(bus->parent_dev);
include/hw/cxl/cxl_component.h +2 -0
index 8dae21cfc6..65779ce7ed 100644
@@ -202,4 +202,6 @@ static inline hwaddr cxl_decode_ig(int ig)
     return 1 << (ig + 8);
 }
 
+CXLComponentState *cxl_get_hb_cstate(PCIHostState *hb);
+
 #endif
Signature · 1 line
2.32.0
#54 lore raw

PATCH v8 32/46 mem/cxl_type3: Add read and write functions for associated hostmem.

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Once a read or write reaches a CXL type 3 device, the HDM decoders on the device are used to establish the Device Physical Address which should be accessed. These functions peform the required maths and then use a device specific address space to access the hostmem->mr to fullfil the actual operation. Note that failed writes are silent, but failed reads return poison. Note this is based loosely on:

https://lore.kernel.org/qemu-devel/[email protected]/
[RFC PATCH 0/9] hw/misc: Add support for interleaved memory accesses

Only lightly tested so far. More complex test cases yet to be written.

---
 hw/mem/cxl_type3.c          | 88 +++++++++++++++++++++++++++++++++++++
 include/hw/cxl/cxl_device.h |  6 +++
 2 files changed, 94 insertions(+)
hw/mem/cxl_type3.c +88 -0
index 244eb5dc91..225155dac5 100644
@@ -100,7 +100,9 @@ static void ct3_finalize(Object *obj)
 
 static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
 {
+    DeviceState *ds = DEVICE(ct3d);
     MemoryRegion *mr;
+    g_autofree char *name = NULL;
 
     if (!ct3d->hostmem) {
         error_setg(errp, "memdev property must be set");
@@ -115,6 +117,13 @@ static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
     memory_region_set_nonvolatile(mr, true);
     memory_region_set_enabled(mr, true);
     host_memory_backend_set_mapped(ct3d->hostmem, true);
+
+    if (ds->id) {
+        name = g_strdup_printf("cxl-type3-dpa-space-%s", ds->id);
+    } else {
+        name = g_strdup("cxl-type3-dpa-space");
+    }
+    address_space_init(&ct3d->hostmem_as, mr, name);
     ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
 
     if (!ct3d->lsa) {
@@ -160,6 +169,85 @@ static void ct3_realize(PCIDevice *pci_dev, Error **errp)
                      &ct3d->cxl_dstate.device_registers);
 }
 
+/* TODO: Support multiple HDM decoders and DPA skip */
+static bool cxl_type3_dpa(CXLType3Dev *ct3d, hwaddr host_addr, uint64_t *dpa)
+{
+    uint32_t *cache_mem = ct3d->cxl_cstate.crb.cache_mem_registers;
+    uint64_t decoder_base, decoder_size, hpa_offset;
+    uint32_t hdm0_ctrl;
+    int ig, iw;
+
+    decoder_base = (((uint64_t)cache_mem[R_CXL_HDM_DECODER0_BASE_HI] << 32) |
+                    cache_mem[R_CXL_HDM_DECODER0_BASE_LO]);
+    if ((uint64_t)host_addr < decoder_base) {
+        return false;
+    }
+
+    hpa_offset = (uint64_t)host_addr - decoder_base;
+
+    decoder_size = ((uint64_t)cache_mem[R_CXL_HDM_DECODER0_SIZE_HI] << 32) |
+        cache_mem[R_CXL_HDM_DECODER0_SIZE_LO];
+    if (hpa_offset >= decoder_size) {
+        return false;
+    }
+
+    hdm0_ctrl = cache_mem[R_CXL_HDM_DECODER0_CTRL];
+    iw = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IW);
+    ig = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IG);
+
+    *dpa = (MAKE_64BIT_MASK(0, 8 + ig) & hpa_offset) |
+        ((MAKE_64BIT_MASK(8 + ig + iw, 64 - 8 - ig - iw) & hpa_offset) >> iw);
+
+    return true;
+}
+
+MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
+                           unsigned size, MemTxAttrs attrs)
+{
+    CXLType3Dev *ct3d = CT3(d);
+    uint64_t dpa_offset;
+    MemoryRegion *mr;
+
+    /* TODO support volatile region */
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        return MEMTX_ERROR;
+    }
+
+    if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
+        return MEMTX_ERROR;
+    }
+
+    if (dpa_offset > int128_get64(mr->size)) {
+        return MEMTX_ERROR;
+    }
+
+    return address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size);
+}
+
+MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
+                            unsigned size, MemTxAttrs attrs)
+{
+    CXLType3Dev *ct3d = CT3(d);
+    uint64_t dpa_offset;
+    MemoryRegion *mr;
+
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        return MEMTX_OK;
+    }
+
+    if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
+        return MEMTX_OK;
+    }
+
+    if (dpa_offset > int128_get64(mr->size)) {
+        return MEMTX_OK;
+    }
+    return address_space_write(&ct3d->hostmem_as, dpa_offset, attrs,
+                               &data, size);
+}
+
 static void ct3d_reset(DeviceState *dev)
 {
     CXLType3Dev *ct3d = CT3(dev);
include/hw/cxl/cxl_device.h +6 -0
index 288cc11772..eb998791d7 100644
@@ -235,6 +235,7 @@ typedef struct cxl_type3_dev {
     PCIDevice parent_obj;
 
     /* Properties */
+    AddressSpace hostmem_as;
     uint64_t size;
     HostMemoryBackend *hostmem;
     HostMemoryBackend *lsa;
@@ -262,4 +263,9 @@ struct CXLType3Class {
                     uint64_t offset);
 };
 
+MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
+                           unsigned size, MemTxAttrs attrs);
+MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
+                            unsigned size, MemTxAttrs attrs);
+
 #endif
Signature · 1 line
2.32.0
#55 lore raw

Re: [PATCH v8 32/46] mem/cxl_type3: Add read and write functions for associated hostmem.

Quoted 36 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Once a read or write reaches a CXL type 3 device, the HDM decoders on the device are used to establish the Device Physical Address which should be accessed. These functions peform the required maths and then use a device specific address space to access the hostmem->mr to fullfil the actual operation. Note that failed writes are silent, but failed reads return poison. Note this is based loosely on:

https://lore.kernel.org/qemu-devel/[email protected]/
[RFC PATCH 0/9] hw/misc: Add support for interleaved memory accesses

Only lightly tested so far. More complex test cases yet to be written.

---
  hw/mem/cxl_type3.c          | 88 +++++++++++++++++++++++++++++++++++++
  include/hw/cxl/cxl_device.h |  6 +++
  2 files changed, 94 insertions(+)
hw/mem/cxl_type3.c +2 -0
index 244eb5dc91..225155dac5 100644
@@ -100,7 +100,9 @@ static void ct3_finalize(Object *obj)
 
  static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
  {
+    DeviceState *ds = DEVICE(ct3d);
      MemoryRegion *mr;
+    g_autofree char *name = NULL;
 
      if (!ct3d->hostmem) {
          error_setg(errp, "memdev property must be set");

hw/mem/cxl_type3.c static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp) · line 119

Quoted 3 lines
hw/mem/cxl_type3.c +0 -0
@@ -115,6 +117,13 @@ static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
      memory_region_set_nonvolatile(mr, true);
      memory_region_set_enabled(mr, true);
      host_memory_backend_set_mapped(ct3d->hostmem, true);

There is an existing example for generating names for PCI devices in SPAPR which you can borrow which looks something like this (not compile tested!):

static char *cxl_type3_get_id(CXLType3Dev *ct3d)

{
    uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(ct3d))));
    PCIDevice *pd = PCI_DEVICE(ct3d);
    DeviceState *ds = DEVICE(ct3d);

    if (ds->id) {
        return g_strdup_printf("%s:%02x:%02x.%x", ds->id, busnr,
                               PCI_SLOT(pd->devfn), PCI_FUNC(pd->devfn));
    } else {
        return g_strdup_printf("%02x:%02x.%x", busnr,
                               PCI_SLOT(pd->devfn), PCI_FUNC(pd->devfn));
    }
}

hw/mem/cxl_type3.c static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp) · line 125

Quoted 6 lines
hw/mem/cxl_type3.c +6 -0
+
+    if (ds->id) {
+        name = g_strdup_printf("cxl-type3-dpa-space-%s", ds->id);
+    } else {
+        name = g_strdup("cxl-type3-dpa-space");
+    }

This then becomes:

    char *id, *name;
    ...

    id = cxl_type3_get_id(ct3d);
    name = g_strdup_printf("cxl-type3-dpa-space:%s", id);
    address_space_init(&ct3d->hostmem_as, mr, name);
    g_free(id);
    g_free(name);

hw/mem/cxl_type3.c static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp) · line 126

Quoted 1 line
hw/mem/cxl_type3.c +1 -0
+    address_space_init(&ct3d->hostmem_as, mr, name);

There is an address_space_init() here but no associated address_space_destroy() - you'll need to add a ct3_finalize() function to remove the address space, otherwise there will be a memory leak when the device is removed because of the dangling reference.

Quoted 102 lines
hw/mem/cxl_type3.c +79 -0
      ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
 
      if (!ct3d->lsa) {
@@ -160,6 +169,85 @@ static void ct3_realize(PCIDevice *pci_dev, Error **errp)
                       &ct3d->cxl_dstate.device_registers);
  }
 
+/* TODO: Support multiple HDM decoders and DPA skip */
+static bool cxl_type3_dpa(CXLType3Dev *ct3d, hwaddr host_addr, uint64_t *dpa)
+{
+    uint32_t *cache_mem = ct3d->cxl_cstate.crb.cache_mem_registers;
+    uint64_t decoder_base, decoder_size, hpa_offset;
+    uint32_t hdm0_ctrl;
+    int ig, iw;
+
+    decoder_base = (((uint64_t)cache_mem[R_CXL_HDM_DECODER0_BASE_HI] << 32) |
+                    cache_mem[R_CXL_HDM_DECODER0_BASE_LO]);
+    if ((uint64_t)host_addr < decoder_base) {
+        return false;
+    }
+
+    hpa_offset = (uint64_t)host_addr - decoder_base;
+
+    decoder_size = ((uint64_t)cache_mem[R_CXL_HDM_DECODER0_SIZE_HI] << 32) |
+        cache_mem[R_CXL_HDM_DECODER0_SIZE_LO];
+    if (hpa_offset >= decoder_size) {
+        return false;
+    }
+
+    hdm0_ctrl = cache_mem[R_CXL_HDM_DECODER0_CTRL];
+    iw = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IW);
+    ig = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IG);
+
+    *dpa = (MAKE_64BIT_MASK(0, 8 + ig) & hpa_offset) |
+        ((MAKE_64BIT_MASK(8 + ig + iw, 64 - 8 - ig - iw) & hpa_offset) >> iw);
+
+    return true;
+}
+
+MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
+                           unsigned size, MemTxAttrs attrs)
+{
+    CXLType3Dev *ct3d = CT3(d);
+    uint64_t dpa_offset;
+    MemoryRegion *mr;
+
+    /* TODO support volatile region */
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        return MEMTX_ERROR;
+    }
+
+    if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
+        return MEMTX_ERROR;
+    }
+
+    if (dpa_offset > int128_get64(mr->size)) {
+        return MEMTX_ERROR;
+    }
+
+    return address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size);
+}
+
+MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
+                            unsigned size, MemTxAttrs attrs)
+{
+    CXLType3Dev *ct3d = CT3(d);
+    uint64_t dpa_offset;
+    MemoryRegion *mr;
+
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        return MEMTX_OK;
+    }
+
+    if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
+        return MEMTX_OK;
+    }
+
+    if (dpa_offset > int128_get64(mr->size)) {
+        return MEMTX_OK;
+    }
+    return address_space_write(&ct3d->hostmem_as, dpa_offset, attrs,
+                               &data, size);
+}
+
  static void ct3d_reset(DeviceState *dev)
  {
      CXLType3Dev *ct3d = CT3(dev);
include/hw/cxl/cxl_device.h +1 -0
index 288cc11772..eb998791d7 100644
@@ -235,6 +235,7 @@ typedef struct cxl_type3_dev {
      PCIDevice parent_obj;
 
      /* Properties */
+    AddressSpace hostmem_as;
      uint64_t size;
      HostMemoryBackend *hostmem;
      HostMemoryBackend *lsa;

include/hw/cxl/cxl_device.h struct CXLType3Class { · line 271

Quoted 9 lines
include/hw/cxl/cxl_device.h +5 -0
@@ -262,4 +263,9 @@ struct CXLType3Class {
                      uint64_t offset);
  };
 
+MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
+                           unsigned size, MemTxAttrs attrs);
+MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
+                            unsigned size, MemTxAttrs attrs);
+
  #endif

ATB,

Mark.

#56 lore raw

Re: [PATCH v8 32/46] mem/cxl_type3: Add read and write functions for associated hostmem.

Quoted 46 lines · On Sat, 19 Mar 2022 08:53:40 +0000 Mark Cave-Ayland…
Quoted 36 lines · Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Once a read or write reaches a CXL type 3 device, the HDM decoders on the device are used to establish the Device Physical Address which should be accessed. These functions peform the required maths and then use a device specific address space to access the hostmem->mr to fullfil the actual operation. Note that failed writes are silent, but failed reads return poison. Note this is based loosely on:

https://lore.kernel.org/qemu-devel/[email protected]/
[RFC PATCH 0/9] hw/misc: Add support for interleaved memory accesses

Only lightly tested so far. More complex test cases yet to be written.

---
  hw/mem/cxl_type3.c          | 88 +++++++++++++++++++++++++++++++++++++
  include/hw/cxl/cxl_device.h |  6 +++
  2 files changed, 94 insertions(+)
hw/mem/cxl_type3.c +2 -0
index 244eb5dc91..225155dac5 100644
@@ -100,7 +100,9 @@ static void ct3_finalize(Object *obj)
 
  static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
  {
+    DeviceState *ds = DEVICE(ct3d);
      MemoryRegion *mr;
+    g_autofree char *name = NULL;
 
      if (!ct3d->hostmem) {
          error_setg(errp, "memdev property must be set");

hw/mem/cxl_type3.c static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp) · line 119

Quoted 3 lines
hw/mem/cxl_type3.c +0 -0
@@ -115,6 +117,13 @@ static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp)
      memory_region_set_nonvolatile(mr, true);
      memory_region_set_enabled(mr, true);
      host_memory_backend_set_mapped(ct3d->hostmem, true);  

There is an existing example for generating names for PCI devices in SPAPR which you can borrow which looks something like this (not compile tested!):

static char *cxl_type3_get_id(CXLType3Dev *ct3d)

Quoted 13 lines
{
     uint32_t busnr = pci_bus_num(PCI_BUS(qdev_get_parent_bus(DEVICE(ct3d))));
     PCIDevice *pd = PCI_DEVICE(ct3d);
     DeviceState *ds = DEVICE(ct3d);

     if (ds->id) {
         return g_strdup_printf("%s:%02x:%02x.%x", ds->id, busnr,
                                PCI_SLOT(pd->devfn), PCI_FUNC(pd->devfn));
     } else {
         return g_strdup_printf("%02x:%02x.%x", busnr,
                                PCI_SLOT(pd->devfn), PCI_FUNC(pd->devfn));
     }
}

The snag is at this point the PCI bus hasn't been enumerated so all of these numbers are 0 for endpoints. They can change at any point (in theory) so using them to help identify a device is unstable.
I switched to using a : as the separator as that is clear than yet another -

Quoted 19 lines

hw/mem/cxl_type3.c static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp) · line 125

Quoted 7 lines
hw/mem/cxl_type3.c +6 -0
+
+    if (ds->id) {
+        name = g_strdup_printf("cxl-type3-dpa-space-%s", ds->id);
+    } else {
+        name = g_strdup("cxl-type3-dpa-space");
+    }  

This then becomes:

     char *id, *name;
     ...

     id = cxl_type3_get_id(ct3d);
     name = g_strdup_printf("cxl-type3-dpa-space:%s", id);
     address_space_init(&ct3d->hostmem_as, mr, name);
     g_free(id);
     g_free(name);
Quoted 5 lines

hw/mem/cxl_type3.c static void cxl_setup_memory(CXLType3Dev *ct3d, Error **errp) · line 126

Quoted 1 line
hw/mem/cxl_type3.c +1 -0
+    address_space_init(&ct3d->hostmem_as, mr, name);  

There is an address_space_init() here but no associated address_space_destroy() - you'll need to add a ct3_finalize() function to remove the address space, otherwise there will be a memory leak when the device is removed because of the dangling reference.

I was lazy on this for two reasons:
a) I could actually figure out to do a finalize for a PCI device. I think after digging more it's via the pc->exit callback (which is very oddly named - I guess for historic reasons - why is the unwind of realize called exit? Ah well. I had previously had it in instance_finalize which resulted in a qtest crash as it would destroy an address space we hadn't created.
b) Only a tiny percentage of all the address spaces in qemu are ever destroyed (or at least I couldn't find where they were destroyed). I guess using bad practice elsewhere isn't the best way to write code :)

Having chased it through it seems to me that what I had in instance_finalize in 23 should have been in pc->exit as it is unwinding stuff done in pc->realize() not the instance_init.

Thanks,

Jonathan

Quoted 117 lines
Quoted 103 lines
hw/mem/cxl_type3.c +79 -0
      ct3d->cxl_dstate.pmem_size = ct3d->hostmem->size;
 
      if (!ct3d->lsa) {
@@ -160,6 +169,85 @@ static void ct3_realize(PCIDevice *pci_dev, Error **errp)
                       &ct3d->cxl_dstate.device_registers);
  }
 
+/* TODO: Support multiple HDM decoders and DPA skip */
+static bool cxl_type3_dpa(CXLType3Dev *ct3d, hwaddr host_addr, uint64_t *dpa)
+{
+    uint32_t *cache_mem = ct3d->cxl_cstate.crb.cache_mem_registers;
+    uint64_t decoder_base, decoder_size, hpa_offset;
+    uint32_t hdm0_ctrl;
+    int ig, iw;
+
+    decoder_base = (((uint64_t)cache_mem[R_CXL_HDM_DECODER0_BASE_HI] << 32) |
+                    cache_mem[R_CXL_HDM_DECODER0_BASE_LO]);
+    if ((uint64_t)host_addr < decoder_base) {
+        return false;
+    }
+
+    hpa_offset = (uint64_t)host_addr - decoder_base;
+
+    decoder_size = ((uint64_t)cache_mem[R_CXL_HDM_DECODER0_SIZE_HI] << 32) |
+        cache_mem[R_CXL_HDM_DECODER0_SIZE_LO];
+    if (hpa_offset >= decoder_size) {
+        return false;
+    }
+
+    hdm0_ctrl = cache_mem[R_CXL_HDM_DECODER0_CTRL];
+    iw = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IW);
+    ig = FIELD_EX32(hdm0_ctrl, CXL_HDM_DECODER0_CTRL, IG);
+
+    *dpa = (MAKE_64BIT_MASK(0, 8 + ig) & hpa_offset) |
+        ((MAKE_64BIT_MASK(8 + ig + iw, 64 - 8 - ig - iw) & hpa_offset) >> iw);
+
+    return true;
+}
+
+MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
+                           unsigned size, MemTxAttrs attrs)
+{
+    CXLType3Dev *ct3d = CT3(d);
+    uint64_t dpa_offset;
+    MemoryRegion *mr;
+
+    /* TODO support volatile region */
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        return MEMTX_ERROR;
+    }
+
+    if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
+        return MEMTX_ERROR;
+    }
+
+    if (dpa_offset > int128_get64(mr->size)) {
+        return MEMTX_ERROR;
+    }
+
+    return address_space_read(&ct3d->hostmem_as, dpa_offset, attrs, data, size);
+}
+
+MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
+                            unsigned size, MemTxAttrs attrs)
+{
+    CXLType3Dev *ct3d = CT3(d);
+    uint64_t dpa_offset;
+    MemoryRegion *mr;
+
+    mr = host_memory_backend_get_memory(ct3d->hostmem);
+    if (!mr) {
+        return MEMTX_OK;
+    }
+
+    if (!cxl_type3_dpa(ct3d, host_addr, &dpa_offset)) {
+        return MEMTX_OK;
+    }
+
+    if (dpa_offset > int128_get64(mr->size)) {
+        return MEMTX_OK;
+    }
+    return address_space_write(&ct3d->hostmem_as, dpa_offset, attrs,
+                               &data, size);
+}
+
  static void ct3d_reset(DeviceState *dev)
  {
      CXLType3Dev *ct3d = CT3(dev);
include/hw/cxl/cxl_device.h +1 -0
index 288cc11772..eb998791d7 100644
@@ -235,6 +235,7 @@ typedef struct cxl_type3_dev {
      PCIDevice parent_obj;
 
      /* Properties */
+    AddressSpace hostmem_as;
      uint64_t size;
      HostMemoryBackend *hostmem;
      HostMemoryBackend *lsa;

include/hw/cxl/cxl_device.h struct CXLType3Class { · line 271

Quoted 9 lines
include/hw/cxl/cxl_device.h +5 -0
@@ -262,4 +263,9 @@ struct CXLType3Class {
                      uint64_t offset);
  };
 
+MemTxResult cxl_type3_read(PCIDevice *d, hwaddr host_addr, uint64_t *data,
+                           unsigned size, MemTxAttrs attrs);
+MemTxResult cxl_type3_write(PCIDevice *d, hwaddr host_addr, uint64_t data,
+                            unsigned size, MemTxAttrs attrs);
+
  #endif  

ATB,

Mark.

#57 lore raw

PATCH v8 33/46 cxl/cxl-host: Add memops for CFMWS region.

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

These memops perform interleave decoding, walking down the CXL topology from CFMWS described host interleave decoder via CXL host bridge HDM decoders, through the CXL root ports and finally call CXL type 3 specific read and write functions.

Note that, whilst functional the current implementation does not support:
* switches
* multiple HDM decoders at a given level.
* unaligned accesses across the interleave boundaries

---
 hw/cxl/cxl-host-stubs.c |   2 +
 hw/cxl/cxl-host.c       | 128 ++++++++++++++++++++++++++++++++++++++++
 include/hw/cxl/cxl.h    |   2 +
 3 files changed, 132 insertions(+)
hw/cxl/cxl-host-stubs.c +2 -0
index d24282ec1c..fcb7aded1f 100644
@@ -12,3 +12,5 @@ void cxl_fixed_memory_window_options_set(MachineState *ms,
                                          Error **errp) {};
 
 void cxl_fixed_memory_window_link_targets(Error **errp) {};
+
+const MemoryRegionOps cfmws_ops;
hw/cxl/cxl-host.c +128 -0

137 lines not on this page · raw message

include/hw/cxl/cxl.h +2 -0
index 5abc307ef4..14194acead 100644
@@ -45,4 +45,6 @@ void cxl_fixed_memory_window_options_set(MachineState *ms,
                                          Error **errp);
 void cxl_fixed_memory_window_link_targets(Error **errp);
 
+extern const MemoryRegionOps cfmws_ops;
+
 #endif
Signature · 1 line
2.32.0
#58 lore raw

PATCH v8 34/46 hw/cxl/component Add a dumb HDM decoder handler

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

Add a trivial handler for now to cover the root bridge where we could do some error checking in future.

---
 hw/cxl/cxl-component-utils.c | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)
hw/cxl/cxl-component-utils.c +31 -0
index 443a11c837..110ec9864e 100644
@@ -32,6 +32,31 @@ static uint64_t cxl_cache_mem_read_reg(void *opaque, hwaddr offset,
     }
 }
 
+static void dumb_hdm_handler(CXLComponentState *cxl_cstate, hwaddr offset,
+                             uint32_t value)
+{
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    uint32_t *cache_mem = cregs->cache_mem_registers;
+    bool should_commit = false;
+
+    switch (offset) {
+    case A_CXL_HDM_DECODER0_CTRL:
+        should_commit = FIELD_EX32(value, CXL_HDM_DECODER0_CTRL, COMMIT);
+        break;
+    default:
+        break;
+    }
+
+    memory_region_transaction_begin();
+    stl_le_p((uint8_t *)cache_mem + offset, value);
+    if (should_commit) {
+        ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, COMMIT, 0);
+        ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, ERR, 0);
+        ARRAY_FIELD_DP32(cache_mem, CXL_HDM_DECODER0_CTRL, COMMITTED, 1);
+    }
+    memory_region_transaction_commit();
+}
+
 static void cxl_cache_mem_write_reg(void *opaque, hwaddr offset, uint64_t value,
                                     unsigned size)
 {
@@ -45,6 +70,12 @@ static void cxl_cache_mem_write_reg(void *opaque, hwaddr offset, uint64_t value,
     }
     if (cregs->special_ops && cregs->special_ops->write) {
         cregs->special_ops->write(cxl_cstate, offset, value, size);
+        return;
+    }
+
+    if (offset >= A_CXL_HDM_DECODER_CAPABILITY &&
+        offset <= A_CXL_HDM_DECODER0_TARGET_LIST_HI) {
+        dumb_hdm_handler(cxl_cstate, offset, value);
     } else {
         cregs->cache_mem_registers[offset / 4] = value;
     }
Signature · 1 line
2.32.0
#59 lore raw

PATCH v8 35/46 i386/pc: Enable CXL fixed memory windows

In reply to Jonathan Cameron

From: Jonathan Cameron <[email protected]>

Add the CFMWs memory regions to the memorymap and adjust the PCI window to avoid hitting the same memory.

---
 hw/i386/pc.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
hw/i386/pc.c +30 -1
index da74f08f9e..48a86ac8a4 100644
@@ -814,7 +814,7 @@ void pc_memory_init(PCMachineState *pcms,
     MachineClass *mc = MACHINE_GET_CLASS(machine);
     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
     X86MachineState *x86ms = X86_MACHINE(pcms);
-    hwaddr cxl_base;
+    hwaddr cxl_base, cxl_resv_end = 0;
 
     assert(machine->ram_size == x86ms->below_4g_mem_size +
                                 x86ms->above_4g_mem_size);
@@ -922,6 +922,24 @@ void pc_memory_init(PCMachineState *pcms,
         e820_add_entry(cxl_base, cxl_size, E820_RESERVED);
         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
         memory_region_add_subregion(system_memory, cxl_base, mr);
+        cxl_resv_end = cxl_base + cxl_size;
+        if (machine->cxl_devices_state->fixed_windows) {
+            hwaddr cxl_fmw_base;
+            GList *it;
+
+            cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
+            for (it = machine->cxl_devices_state->fixed_windows; it; it = it->next) {
+                CXLFixedWindow *fw = it->data;
+
+                fw->base = cxl_fmw_base;
+                memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
+                                      "cxl-fixed-memory-region", fw->size);
+                memory_region_add_subregion(system_memory, fw->base, &fw->mr);
+                e820_add_entry(fw->base, fw->size, E820_RESERVED);
+                cxl_fmw_base += fw->size;
+                cxl_resv_end = cxl_fmw_base;
+            }
+        }
     }
 
     /* Initialize PC system firmware */
@@ -951,6 +969,10 @@ void pc_memory_init(PCMachineState *pcms,
         if (!pcmc->broken_reserved_end) {
             res_mem_end += memory_region_size(&machine->device_memory->mr);
         }
+
+        if (machine->cxl_devices_state->is_enabled) {
+            res_mem_end = cxl_resv_end;
+        }
         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
     }
@@ -987,6 +1009,13 @@ uint64_t pc_pci_hole64_start(void)
     if (ms->cxl_devices_state->host_mr.addr) {
         hole64_start = ms->cxl_devices_state->host_mr.addr +
             memory_region_size(&ms->cxl_devices_state->host_mr);
+        if (ms->cxl_devices_state->fixed_windows) {
+            GList *it;
+            for (it = ms->cxl_devices_state->fixed_windows; it; it = it->next) {
+                CXLFixedWindow *fw = it->data;
+                hole64_start = fw->mr.addr + memory_region_size(&fw->mr);
+            }
+        }
     } else if (pcmc->has_reserved_memory && ms->device_memory->base) {
         hole64_start = ms->device_memory->base;
         if (!pcmc->broken_reserved_end) {
Signature · 1 line
2.32.0
#60 lore raw

PATCH v8 36/46 tests/acpi: q35: Allow addition of a CXL test.

In reply to Jonathan Cameron

Add exceptions for the DSDT and the new CEDT tables specific to a new CXL test in the following patch.

---
 tests/data/acpi/q35/CEDT.cxl                | 0
 tests/data/acpi/q35/DSDT.cxl                | 0
 tests/qtest/bios-tables-test-allowed-diff.h | 2 ++
 3 files changed, 2 insertions(+)
tests/data/acpi/q35/CEDT.cxl new file +0 -0
new file mode 100644
index 0000000000..e69de29bb2
tests/data/acpi/q35/DSDT.cxl new file +0 -0
new file mode 100644
index 0000000000..e69de29bb2
tests/qtest/bios-tables-test-allowed-diff.h +2 -0
index dfb8523c8b..7c7f9fbc44 100644
@@ -1 +1,3 @@
 /* List of comma-separated changed AML files to ignore */
+"tests/data/acpi/q35/DSDT.cxl",
+"tests/data/acpi/q35/CEDT.cxl",
Signature · 1 line
2.32.0
#61 lore raw

PATCH v8 37/46 qtests/bios-tables-test: Add a test for CXL emulation.

In reply to Jonathan Cameron

The DSDT includes several CXL specific elements and the CEDT table is only present if we enable CXL.

The test exercises all current functionality with several CFMWS, CHBS structures in CEDT and ACPI0016/ACPI00017 and _OSC entries in DSDT.

---
 tests/qtest/bios-tables-test.c | 44 ++++++++++++++++++++++++++++++++++
 1 file changed, 44 insertions(+)
tests/qtest/bios-tables-test.c +44 -0
index c4a2d1e166..d43503a42d 100644
@@ -1537,6 +1537,49 @@ static void test_acpi_q35_viot(void)
     free_test_data(&data);
 }
 
+static void test_acpi_q35_cxl(void)
+{
+    gchar *tmp_path = g_dir_make_tmp("qemu-test-cxl.XXXXXX", NULL);
+    gchar *params;
+
+    test_data data = {
+        .machine = MACHINE_Q35,
+        .variant = ".cxl",
+    };
+    /*
+     * A complex CXL setup.
+     */
+    params = g_strdup_printf(" -machine cxl=on"
+                             " -object memory-backend-file,id=cxl-mem1,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=cxl-mem2,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=cxl-mem3,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=cxl-mem4,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=lsa1,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=lsa2,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=lsa3,mem-path=%s,size=256M"
+                             " -object memory-backend-file,id=lsa4,mem-path=%s,size=256M"
+                             " -device pxb-cxl,bus_nr=12,bus=pcie.0,id=cxl.1"
+                             " -device pxb-cxl,bus_nr=222,bus=pcie.0,id=cxl.2"
+                             " -device cxl-rp,port=0,bus=cxl.1,id=rp1,chassis=0,slot=2"
+                             " -device cxl-type3,bus=rp1,memdev=cxl-mem1,lsa=lsa1,size=256M"
+                             " -device cxl-rp,port=1,bus=cxl.1,id=rp2,chassis=0,slot=3"
+                             " -device cxl-type3,bus=rp2,memdev=cxl-mem2,lsa=lsa2,size=256M"
+                             " -device cxl-rp,port=0,bus=cxl.2,id=rp3,chassis=0,slot=5"
+                             " -device cxl-type3,bus=rp3,memdev=cxl-mem3,lsa=lsa3,size=256M"
+                             " -device cxl-rp,port=1,bus=cxl.2,id=rp4,chassis=0,slot=6"
+                             " -device cxl-type3,bus=rp4,memdev=cxl-mem4,lsa=lsa4,size=256M"
+                             " -cxl-fixed-memory-window targets.0=cxl.1,size=4G,interleave-granularity=8k"
+                             " -cxl-fixed-memory-window targets.0=cxl.1,targets.1=cxl.2,size=4G,interleave-granularity=8k",
+                             tmp_path, tmp_path, tmp_path, tmp_path,
+                             tmp_path, tmp_path, tmp_path, tmp_path);
+    test_acpi_one(params, &data);
+
+    g_free(params);
+    g_assert(g_rmdir(tmp_path) == 0);
+    g_free(tmp_path);
+    free_test_data(&data);
+}
+
 static void test_acpi_virt_viot(void)
 {
     test_data data = {
@@ -1742,6 +1785,7 @@ int main(int argc, char *argv[])
             qtest_add_func("acpi/q35/kvm/dmar", test_acpi_q35_kvm_dmar);
         }
         qtest_add_func("acpi/q35/viot", test_acpi_q35_viot);
+        qtest_add_func("acpi/q35/cxl", test_acpi_q35_cxl);
         qtest_add_func("acpi/q35/slic", test_acpi_q35_slic);
     } else if (strcmp(arch, "aarch64") == 0) {
         if (has_tcg) {
Signature · 1 line
2.32.0
#62 lore raw

PATCH v8 38/46 tests/acpi: Add tables for CXL emulation.

In reply to Jonathan Cameron

Tables that differ from normal Q35 tables when running the CXL test.

---
 tests/data/acpi/q35/CEDT.cxl                | Bin 0 -> 184 bytes
 tests/data/acpi/q35/DSDT.cxl                | Bin 0 -> 9615 bytes
 tests/qtest/bios-tables-test-allowed-diff.h |   2 --
 3 files changed, 2 deletions(-)
tests/data/acpi/q35/CEDT.cxl binary +0 -0
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..b8fa06b00e65712e91e0a5ea0d9277e0146d1c00 100644
GIT binary patch
tests/data/acpi/q35/DSDT.cxl binary +0 -0
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..c1206defed0154e9024702bba88453b2790a306d 100644
GIT binary patch
tests/qtest/bios-tables-test-allowed-diff.h +0 -2
index 7c7f9fbc44..dfb8523c8b 100644
@@ -1,3 +1 @@
 /* List of comma-separated changed AML files to ignore */
-"tests/data/acpi/q35/DSDT.cxl",
-"tests/data/acpi/q35/CEDT.cxl",
Signature · 1 line
2.32.0
#63 lore raw

PATCH v8 39/46 qtest/cxl: Add more complex test cases with CFMWs

In reply to Jonathan Cameron

From: Ben Widawsky <[email protected]>

Add CXL Fixed Memory Windows to the CXL tests.

---
 tests/qtest/cxl-test.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
tests/qtest/cxl-test.c +5 -3
index 148bc94340..0bb93b0191 100644
@@ -9,11 +9,13 @@
 #include "libqtest-single.h"
 
 #define QEMU_PXB_CMD "-machine q35,cxl=on " \
-                     "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52 "
+                     "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52 "  \
+                     "-cxl-fixed-memory-window targets.0=cxl.0,size=4G "
 
-#define QEMU_2PXB_CMD "-machine q35,cxl=on " \
+#define QEMU_2PXB_CMD "-machine q35,cxl=on "                            \
                       "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52 "  \
-                      "-device pxb-cxl,id=cxl.1,bus=pcie.0,bus_nr=53 "
+                      "-device pxb-cxl,id=cxl.1,bus=pcie.0,bus_nr=53 " \
+                      "-cxl-fixed-memory-window targets.0=cxl.0,targets.1=cxl.1,size=4G "
 
 #define QEMU_RP "-device cxl-rp,id=rp0,bus=cxl.0,chassis=0,slot=0 "
 
Signature · 1 line
2.32.0
#64 lore raw

PATCH v8 40/46 hw/arm/virt: Basic CXL enablement on pci_expander_bridge instances pxb-cxl

In reply to Jonathan Cameron

Code based on i386/pc enablement.
The memory layout places space for 16 host bridge register regions after the GIC_REDIST2 in the extended memmap.
The CFMWs are placed above the extended memmap.

Only create the CEDT table if cxl=on set for the machine.

---
 hw/arm/virt-acpi-build.c | 33 +++++++++++++++++++++++++++++++++
 hw/arm/virt.c            | 40 +++++++++++++++++++++++++++++++++++++++-
 include/hw/arm/virt.h    |  1 +
 3 files changed, 73 insertions(+), 1 deletion(-)
hw/arm/virt-acpi-build.c +33 -0
index 449fab0080..86a2f40437 100644
@@ -39,9 +39,11 @@
 #include "hw/acpi/aml-build.h"
 #include "hw/acpi/utils.h"
 #include "hw/acpi/pci.h"
+#include "hw/acpi/cxl.h"
 #include "hw/acpi/memory_hotplug.h"
 #include "hw/acpi/generic_event_device.h"
 #include "hw/acpi/tpm.h"
+#include "hw/cxl/cxl.h"
 #include "hw/pci/pcie_host.h"
 #include "hw/pci/pci.h"
 #include "hw/pci/pci_bus.h"
@@ -157,10 +159,29 @@ static void acpi_dsdt_add_virtio(Aml *scope,
     }
 }
 
+/* Uses local definition of AcpiBuildState so can't easily be common code */
+static void build_acpi0017(Aml *table)
+{
+    Aml *dev, *scope, *method;
+
+    scope =  aml_scope("_SB");
+    dev = aml_device("CXLM");
+    aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0017")));
+
+    method = aml_method("_STA", 0, AML_NOTSERIALIZED);
+    aml_append(method, aml_return(aml_int(0x01)));
+    aml_append(dev, method);
+
+    aml_append(scope, dev);
+    aml_append(table, scope);
+}
+
 static void acpi_dsdt_add_pci(Aml *scope, const MemMapEntry *memmap,
                               uint32_t irq, VirtMachineState *vms)
 {
     int ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
+    bool cxl_present = false;
+    PCIBus *bus = vms->bus;
     struct GPEXConfig cfg = {
         .mmio32 = memmap[VIRT_PCIE_MMIO],
         .pio    = memmap[VIRT_PCIE_PIO],
@@ -174,6 +195,14 @@ static void acpi_dsdt_add_pci(Aml *scope, const MemMapEntry *memmap,
     }
 
     acpi_dsdt_add_gpex(scope, &cfg);
+    QLIST_FOREACH(bus, &vms->bus->child, sibling) {
+        if (pci_bus_is_cxl(bus)) {
+            cxl_present = true;
+        }
+    }
+    if (cxl_present) {
+        build_acpi0017(scope);
+    }
 }
 
 static void acpi_dsdt_add_gpio(Aml *scope, const MemMapEntry *gpio_memmap,
@@ -991,6 +1020,10 @@ void virt_acpi_build(VirtMachineState *vms, AcpiBuildTables *tables)
                        vms->oem_table_id);
         }
     }
+    if (ms->cxl_devices_state->is_enabled) {
+        cxl_build_cedt(ms, table_offsets, tables_blob, tables->linker,
+                       vms->oem_id, vms->oem_table_id);
+    }
 
     if (ms->nvdimms_state->is_enabled) {
         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
hw/arm/virt.c +39 -1
index 5f7b0f2607..02ae7ca750 100644
@@ -78,6 +78,7 @@
 #include "hw/virtio/virtio-mem-pci.h"
 #include "hw/virtio/virtio-iommu.h"
 #include "hw/char/pl011.h"
+#include "hw/cxl/cxl.h"
 #include "qemu/guest-random.h"
 
 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
@@ -178,6 +179,7 @@ static const MemMapEntry base_memmap[] = {
 static MemMapEntry extended_memmap[] = {
     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
+    [VIRT_CXL_HOST] =           { 0x0, 64 * KiB * 16 }, /* 16 UID */
     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
     /* Second PCIe window */
     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
@@ -1508,6 +1510,17 @@ static void create_pcie(VirtMachineState *vms)
     }
 }
 
+static void create_cxl_host_reg_region(VirtMachineState *vms)
+{
+    MemoryRegion *sysmem = get_system_memory();
+    MachineState *ms = MACHINE(vms);
+    MemoryRegion *mr = &ms->cxl_devices_state->host_mr;
+
+    memory_region_init(mr, OBJECT(ms), "cxl_host_reg",
+                       vms->memmap[VIRT_CXL_HOST].size);
+    memory_region_add_subregion(sysmem, vms->memmap[VIRT_CXL_HOST].base, mr);
+}
+
 static void create_platform_bus(VirtMachineState *vms)
 {
     DeviceState *dev;
@@ -1670,7 +1683,7 @@ static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
 {
     MachineState *ms = MACHINE(vms);
-    hwaddr base, device_memory_base, device_memory_size, memtop;
+    hwaddr base, device_memory_base, device_memory_size, memtop, cxl_fmw_base;
     int i;
 
     vms->memmap = extended_memmap;
@@ -1762,6 +1775,20 @@ static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
                            "device-memory", device_memory_size);
     }
+
+    if (ms->cxl_devices_state->fixed_windows) {
+        GList *it;
+
+        cxl_fmw_base = ROUND_UP(base, 256 * MiB);
+        for (it = ms->cxl_devices_state->fixed_windows; it; it = it->next) {
+            CXLFixedWindow *fw = it->data;
+
+            fw->base = cxl_fmw_base;
+            memory_region_init_io(&fw->mr, OBJECT(vms), &cfmws_ops, fw,
+                                  "cxl-fixed-memory-region", fw->size);
+            cxl_fmw_base += fw->size;
+        }
+    }
 }
 
 /*
@@ -2168,6 +2195,15 @@ static void machvirt_init(MachineState *machine)
         memory_region_add_subregion(sysmem, machine->device_memory->base,
                                     &machine->device_memory->mr);
     }
+    if (machine->cxl_devices_state->fixed_windows) {
+        GList *it;
+        for (it = machine->cxl_devices_state->fixed_windows; it;
+             it = it->next) {
+            CXLFixedWindow *fw = it->data;
+
+            memory_region_add_subregion(sysmem, fw->base, &fw->mr);
+        }
+    }
 
     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
 
@@ -2194,6 +2230,7 @@ static void machvirt_init(MachineState *machine)
     create_rtc(vms);
 
     create_pcie(vms);
+    create_cxl_host_reg_region(vms);
 
     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
         vms->acpi_dev = create_acpi_ged(vms);
@@ -2849,6 +2886,7 @@ static void virt_machine_class_init(ObjectClass *oc, void *data)
     hc->unplug = virt_machine_device_unplug_cb;
     mc->nvdimm_supported = true;
     mc->smp_props.clusters_supported = true;
+    mc->cxl_supported = true;
     mc->auto_enable_numa_with_memhp = true;
     mc->auto_enable_numa_with_memdev = true;
     mc->default_ram_id = "mach-virt.ram";
include/hw/arm/virt.h +1 -0
index 7e76ee2619..a10b950452 100644
@@ -92,6 +92,7 @@ enum {
 /* indices of IO regions located after the RAM */
 enum {
     VIRT_HIGH_GIC_REDIST2 =  VIRT_LOWMEMMAP_LAST,
+    VIRT_CXL_HOST,
     VIRT_HIGH_PCIE_ECAM,
     VIRT_HIGH_PCIE_MMIO,
 };
Signature · 1 line
2.32.0
#65 lore raw

PATCH v8 41/46 qtest/cxl: Add aarch64 virt test for CXL

In reply to Jonathan Cameron

Add a single complex case for aarch64 virt machine.

---
 tests/qtest/cxl-test.c  | 48 +++++++++++++++++++++++++++++++++--------
 tests/qtest/meson.build |  1 +
 2 files changed, 40 insertions(+), 9 deletions(-)
tests/qtest/cxl-test.c +39 -9
index 0bb93b0191..c96469a435 100644
@@ -17,6 +17,11 @@
                       "-device pxb-cxl,id=cxl.1,bus=pcie.0,bus_nr=53 " \
                       "-cxl-fixed-memory-window targets.0=cxl.0,targets.1=cxl.1,size=4G "
 
+#define QEMU_VIRT_2PXB_CMD "-machine virt,cxl=on "                      \
+                      "-device pxb-cxl,id=cxl.0,bus=pcie.0,bus_nr=52 "  \
+                      "-device pxb-cxl,id=cxl.1,bus=pcie.0,bus_nr=53 "  \
+                      "-cxl-fixed-memory-window targets.0=cxl.0,targets.1=cxl.1,size=4G "
+
 #define QEMU_RP "-device cxl-rp,id=rp0,bus=cxl.0,chassis=0,slot=0 "
 
 /* Dual ports on first pxb */
@@ -134,18 +139,43 @@ static void cxl_2pxb_4rp_4t3d(void)
     qtest_end();
 }
 
+static void cxl_virt_2pxb_4rp_4t3d(void)
+{
+    g_autoptr(GString) cmdline = g_string_new(NULL);
+    char template[] = "/tmp/cxl-test-XXXXXX";
+    const char *tmpfs;
+
+    tmpfs = mkdtemp(template);
+
+    g_string_printf(cmdline, QEMU_VIRT_2PXB_CMD QEMU_4RP QEMU_4T3D,
+                    tmpfs, tmpfs, tmpfs, tmpfs, tmpfs, tmpfs,
+                    tmpfs, tmpfs);
+
+    qtest_start(cmdline->str);
+    qtest_end();
+}
+
 int main(int argc, char **argv)
 {
+    const char *arch = qtest_get_arch();
+
     g_test_init(&argc, &argv, NULL);
 
-    qtest_add_func("/pci/cxl/basic_hostbridge", cxl_basic_hb);
-    qtest_add_func("/pci/cxl/basic_pxb", cxl_basic_pxb);
-    qtest_add_func("/pci/cxl/pxb_with_window", cxl_pxb_with_window);
-    qtest_add_func("/pci/cxl/pxb_x2_with_window", cxl_2pxb_with_window);
-    qtest_add_func("/pci/cxl/rp", cxl_root_port);
-    qtest_add_func("/pci/cxl/rp_x2", cxl_2root_port);
-    qtest_add_func("/pci/cxl/type3_device", cxl_t3d);
-    qtest_add_func("/pci/cxl/rp_x2_type3_x2", cxl_1pxb_2rp_2t3d);
-    qtest_add_func("/pci/cxl/pxb_x2_root_port_x4_type3_x4", cxl_2pxb_4rp_4t3d);
+    if (strcmp(arch, "i386") == 0 || strcmp(arch, "x86_64") == 0) {
+        qtest_add_func("/pci/cxl/basic_hostbridge", cxl_basic_hb);
+        qtest_add_func("/pci/cxl/basic_pxb", cxl_basic_pxb);
+        qtest_add_func("/pci/cxl/pxb_with_window", cxl_pxb_with_window);
+        qtest_add_func("/pci/cxl/pxb_x2_with_window", cxl_2pxb_with_window);
+        qtest_add_func("/pci/cxl/rp", cxl_root_port);
+        qtest_add_func("/pci/cxl/rp_x2", cxl_2root_port);
+        qtest_add_func("/pci/cxl/type3_device", cxl_t3d);
+        qtest_add_func("/pci/cxl/rp_x2_type3_x2", cxl_1pxb_2rp_2t3d);
+        qtest_add_func("/pci/cxl/pxb_x2_root_port_x4_type3_x4",
+                       cxl_2pxb_4rp_4t3d);
+    } else if (strcmp(arch, "aarch64") == 0) {
+        qtest_add_func("/pci/cxl/virt/pxb_x2_root_port_x4_type3_x4",
+                       cxl_virt_2pxb_4rp_4t3d);
+    }
+
     return g_test_run();
 }
tests/qtest/meson.build +1 -0
index 6e1ad4dc9a..6aefde4584 100644
@@ -224,6 +224,7 @@ qtests_aarch64 = \
   (config_all_devices.has_key('CONFIG_TPM_TIS_SYSBUS') ? ['tpm-tis-device-test'] : []) +        \
   (config_all_devices.has_key('CONFIG_TPM_TIS_SYSBUS') ? ['tpm-tis-device-swtpm-test'] : []) +  \
   (config_all_devices.has_key('CONFIG_XLNX_ZYNQMP_ARM') ? ['xlnx-can-test', 'fuzz-xlnx-dp-test'] : []) + \
+  qtests_cxl +                                                                                  \
   ['arm-cpu-features',
    'numa-test',
    'boot-serial-test',
Signature · 1 line
2.32.0
#66 lore raw

PATCH v8 42/46 docs/cxl: Add initial Compute eXpress Link (CXL) documentation.

In reply to Jonathan Cameron

Provide an introduction to the main components of a CXL system, with detailed explanation of memory interleaving, example command lines and kernel configuration.

This was a challenging document to write due to the need to extract only that subset of CXL information which is relevant to either users of QEMU emulation of CXL or to those interested in the implementation. Much of CXL is concerned with specific elements of the protocol, management of memory pooling etc which is simply not relevant to what is currently planned for CXL emulation in QEMU. All comments welcome

---
 docs/system/device-emulation.rst |   1 +
 docs/system/devices/cxl.rst      | 302 +++++++++++++++++++++++++++++++
 2 files changed, 303 insertions(+)
docs/system/device-emulation.rst +1 -0
index 0b3a3d73ad..2da2bd5d64 100644
@@ -83,6 +83,7 @@ Emulated Devices
    :maxdepth: 1
 
    devices/can.rst
+   devices/cxl.rst
    devices/ivshmem.rst
    devices/net.rst
    devices/nvme.rst
docs/system/devices/cxl.rst new file +302 -0

302 lines not on this page · raw message

Signature · 1 line
2.32.0
#67 lore raw

PATCH v8 43/46 pci-bridge/cxl_upstream: Add a CXL switch upstream port

In reply to Jonathan Cameron

An initial simple upstream port emulation to allow the creation of CXL switches. The Device ID has been allocated for this use.

---
 hw/pci-bridge/cxl_upstream.c | 205 +++++++++++++++++++++++++++++++++++
 hw/pci-bridge/meson.build    |   2 +-
 include/hw/cxl/cxl.h         |   4 +
 3 files changed, 210 insertions(+), 1 deletion(-)
hw/pci-bridge/cxl_upstream.c new file +205 -0
new file mode 100644
index 0000000000..f95e01b373
@@ -0,0 +1,205 @@
+/*
+ * Emulated CXL Switch Upstream Port
+ *
+ * Copyright (c) 2022 Huawei Technologies.
+ *
+ * Based on xio31130_upstream.c
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include "hw/pci/msi.h"
+#include "hw/pci/pcie.h"
+#include "hw/pci/pcie_port.h"
+
+#define CXL_UPSTREAM_PORT_MSI_NR_VECTOR 1
+
+#define CXL_UPSTREAM_PORT_MSI_OFFSET 0x70
+#define CXL_UPSTREAM_PORT_PCIE_CAP_OFFSET 0x90
+#define CXL_UPSTREAM_PORT_AER_OFFSET 0x100
+#define CXL_UPSTREAM_PORT_DVSEC_OFFSET \
+    (CXL_UPSTREAM_PORT_AER_OFFSET + PCI_ERR_SIZEOF)
+
+typedef struct CXLUpstreamPort {
+    /*< private >*/
+    PCIEPort parent_obj;
+
+    /*< public >*/
+    CXLComponentState cxl_cstate;
+} CXLUpstreamPort;
+
+CXLComponentState *cxl_usp_to_cstate(CXLUpstreamPort *usp)
+{
+    return &usp->cxl_cstate;
+}
+
+static void cxl_usp_dvsec_write_config(PCIDevice *dev, uint32_t addr,
+                                       uint32_t val, int len)
+{
+    CXLUpstreamPort *usp = CXL_USP(dev);
+
+    if (range_contains(&usp->cxl_cstate.dvsecs[EXTENSIONS_PORT_DVSEC], addr)) {
+        uint8_t *reg = &dev->config[addr];
+        addr -= usp->cxl_cstate.dvsecs[EXTENSIONS_PORT_DVSEC].lob;
+        if (addr == PORT_CONTROL_OFFSET) {
+            if (pci_get_word(reg) & PORT_CONTROL_UNMASK_SBR) {
+                /* unmask SBR */
+            }
+            if (pci_get_word(reg) & PORT_CONTROL_ALT_MEMID_EN) {
+                /* Alt Memory & ID Space Enable */
+            }
+        }
+    }
+}
+
+static void cxl_usp_write_config(PCIDevice *d, uint32_t address,
+                                 uint32_t val, int len)
+{
+    pci_bridge_write_config(d, address, val, len);
+    pcie_cap_flr_write_config(d, address, val, len);
+    pcie_aer_write_config(d, address, val, len);
+
+    cxl_usp_dvsec_write_config(d, address, val, len);
+}
+
+static void latch_registers(CXLUpstreamPort *usp)
+{
+    uint32_t *reg_state = usp->cxl_cstate.crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_UPSTREAM_PORT);
+    ARRAY_FIELD_DP32(reg_state, CXL_HDM_DECODER_CAPABILITY, TARGET_COUNT, 8);
+}
+
+static void cxl_usp_reset(DeviceState *qdev)
+{
+    PCIDevice *d = PCI_DEVICE(qdev);
+    CXLUpstreamPort *usp = CXL_USP(qdev);
+
+    pci_bridge_reset(qdev);
+    pcie_cap_deverr_reset(d);
+    latch_registers(usp);
+}
+
+static void build_dvsecs(CXLComponentState *cxl)
+{
+    uint8_t *dvsec;
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_port_extensions){ 0 };
+    cxl_component_create_dvsec(cxl, EXTENSIONS_PORT_DVSEC_LENGTH,
+                               EXTENSIONS_PORT_DVSEC,
+                               EXTENSIONS_PORT_DVSEC_REVID, dvsec);
+    dvsec = (uint8_t *)&(struct cxl_dvsec_port_flexbus){
+        .cap                     = 0x26, /* IO, Mem, non-MLD */
+        .ctrl                    = 0,
+        .status                  = 0x26, /* same */
+        .rcvd_mod_ts_data_phase1 = 0xef, /* WTF? */
+    };
+    cxl_component_create_dvsec(cxl, PCIE_FLEXBUS_PORT_DVSEC_LENGTH_2_0,
+                               PCIE_FLEXBUS_PORT_DVSEC,
+                               PCIE_FLEXBUS_PORT_DVSEC_REVID_2_0, dvsec);
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_register_locator){
+        .rsvd         = 0,
+        .reg0_base_lo = RBI_COMPONENT_REG | CXL_COMPONENT_REG_BAR_IDX,
+        .reg0_base_hi = 0,
+    };
+    cxl_component_create_dvsec(cxl, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
+                               REG_LOC_DVSEC_REVID, dvsec);
+}
+
+static void cxl_usp_realize(PCIDevice *d, Error **errp)
+{
+    PCIEPort *p = PCIE_PORT(d);
+    CXLUpstreamPort *usp = CXL_USP(d);
+    CXLComponentState *cxl_cstate = &usp->cxl_cstate;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    MemoryRegion *component_bar = &cregs->component_registers;
+    int rc;
+
+    pci_bridge_initfn(d, TYPE_PCIE_BUS);
+    pcie_port_init_reg(d);
+
+    rc = msi_init(d, CXL_UPSTREAM_PORT_MSI_OFFSET,
+                  CXL_UPSTREAM_PORT_MSI_NR_VECTOR, true, true, errp);
+    if (rc) {
+        assert(rc == -ENOTSUP);
+        goto err_bridge;
+    }
+
+    rc = pcie_cap_init(d, CXL_UPSTREAM_PORT_PCIE_CAP_OFFSET,
+                       PCI_EXP_TYPE_UPSTREAM, p->port, errp);
+    if (rc < 0) {
+        goto err_msi;
+    }
+
+    pcie_cap_flr_init(d);
+    pcie_cap_deverr_init(d);
+    rc = pcie_aer_init(d, PCI_ERR_VER, CXL_UPSTREAM_PORT_AER_OFFSET,
+                       PCI_ERR_SIZEOF, errp);
+    if (rc) {
+        goto err_cap;
+    }
+
+    cxl_cstate->dvsec_offset = CXL_UPSTREAM_PORT_DVSEC_OFFSET;
+    cxl_cstate->pdev = d;
+    build_dvsecs(cxl_cstate);
+    cxl_component_register_block_init(OBJECT(d), cxl_cstate, TYPE_CXL_USP);
+    pci_register_bar(d, CXL_COMPONENT_REG_BAR_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY |
+                     PCI_BASE_ADDRESS_MEM_TYPE_64,
+                     component_bar);
+
+    return;
+
+err_cap:
+    pcie_cap_exit(d);
+err_msi:
+    msi_uninit(d);
+err_bridge:
+    pci_bridge_exitfn(d);
+}
+
+static void cxl_usp_exitfn(PCIDevice *d)
+{
+    pcie_aer_exit(d);
+    pcie_cap_exit(d);
+    msi_uninit(d);
+    pci_bridge_exitfn(d);
+}
+
+static void cxl_upstream_class_init(ObjectClass *oc, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(oc);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(oc);
+
+    k->is_bridge = true;
+    k->config_write = cxl_usp_write_config;
+    k->realize = cxl_usp_realize;
+    k->exit = cxl_usp_exitfn;
+    k->vendor_id = 0x19e5; /* Huawei */
+    k->device_id = 0xa128; /* Emulated CXL Switch Upstream Port */
+    k->revision = 0;
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->desc = "CXL Switch Upstream Port";
+    dc->reset = cxl_usp_reset;
+}
+
+static const TypeInfo cxl_usp_info = {
+    .name = TYPE_CXL_USP,
+    .parent = TYPE_PCIE_PORT,
+    .instance_size = sizeof(CXLUpstreamPort),
+    .class_init = cxl_upstream_class_init,
+    .interfaces = (InterfaceInfo[]) {
+        { INTERFACE_PCIE_DEVICE },
+        { INTERFACE_CXL_DEVICE },
+        { }
+    },
+};
+
+static void cxl_usp_register_type(void)
+{
+    type_register_static(&cxl_usp_info);
+}
+
+type_init(cxl_usp_register_type);
hw/pci-bridge/meson.build +1 -1
index b6d26a03d5..4a8c26b5a1 100644
@@ -5,7 +5,7 @@ pci_ss.add(when: 'CONFIG_IOH3420', if_true: files('ioh3420.c'))
 pci_ss.add(when: 'CONFIG_PCIE_PORT', if_true: files('pcie_root_port.c', 'gen_pcie_root_port.c', 'pcie_pci_bridge.c'))
 pci_ss.add(when: 'CONFIG_PXB', if_true: files('pci_expander_bridge.c'))
 pci_ss.add(when: 'CONFIG_XIO3130', if_true: files('xio3130_upstream.c', 'xio3130_downstream.c'))
-pci_ss.add(when: 'CONFIG_CXL', if_true: files('cxl_root_port.c'))
+pci_ss.add(when: 'CONFIG_CXL', if_true: files('cxl_root_port.c', 'cxl_upstream.c'))
 
 # NewWorld PowerMac
 pci_ss.add(when: 'CONFIG_DEC_PCI', if_true: files('dec.c'))
include/hw/cxl/cxl.h +4 -0
index 14194acead..a9a2e6c405 100644
@@ -46,5 +46,9 @@ void cxl_fixed_memory_window_options_set(MachineState *ms,
 void cxl_fixed_memory_window_link_targets(Error **errp);
 
 extern const MemoryRegionOps cfmws_ops;
+#define TYPE_CXL_USP "cxl-upstream"
 
+typedef struct CXLUpstreamPort CXLUpstreamPort;
+DECLARE_INSTANCE_CHECKER(CXLUpstreamPort, CXL_USP, TYPE_CXL_USP)
+CXLComponentState *cxl_usp_to_cstate(CXLUpstreamPort *usp);
 #endif
Signature · 1 line
2.32.0
#68 lore raw

PATCH v8 44/46 pci-bridge/cxl_downstream: Add a CXL switch downstream port

In reply to Jonathan Cameron

Emulation of a simple CXL Switch downstream port.
The Device ID has been allocated for this use.

---
 hw/pci-bridge/cxl_downstream.c | 229 +++++++++++++++++++++++++++++++++
 hw/pci-bridge/meson.build      |   2 +-
 2 files changed, 230 insertions(+), 1 deletion(-)
hw/pci-bridge/cxl_downstream.c new file +229 -0
new file mode 100644
index 0000000000..8e66f632a5
@@ -0,0 +1,229 @@
+/*
+ * Emulated CXL Switch Downstream Port
+ *
+ * Copyright (c) 2022 Huawei Technologies.
+ *
+ * Based on xio31130_downstream.c
+ *
+ * SPDX-License-Identifier: GPL-2.0-or-later
+ */
+
+#include "qemu/osdep.h"
+#include "hw/pci/msi.h"
+#include "hw/pci/pcie.h"
+#include "hw/pci/pcie_port.h"
+#include "qapi/error.h"
+
+typedef struct CXLDownStreamPort {
+    /*< private >*/
+    PCIESlot parent_obj;
+
+    /*< public >*/
+    CXLComponentState cxl_cstate;
+} CXLDownstreamPort;
+
+#define TYPE_CXL_DSP "cxl-downstream"
+DECLARE_INSTANCE_CHECKER(CXLDownstreamPort, CXL_DSP, TYPE_CXL_DSP)
+
+#define CXL_DOWNSTREAM_PORT_MSI_OFFSET 0x70
+#define CXL_DOWNSTREAM_PORT_MSI_NR_VECTOR 1
+#define CXL_DOWNSTREAM_PORT_EXP_OFFSET 0x90
+#define CXL_DOWNSTREAM_PORT_AER_OFFSET 0x100
+#define CXL_DOWNSTREAM_PORT_DVSEC_OFFSET        \
+    (CXL_DOWNSTREAM_PORT_AER_OFFSET + PCI_ERR_SIZEOF)
+
+static void latch_registers(CXLDownstreamPort *dsp)
+{
+    uint32_t *reg_state = dsp->cxl_cstate.crb.cache_mem_registers;
+
+    cxl_component_register_init_common(reg_state, CXL2_DOWNSTREAM_PORT);
+}
+
+/* TODO: Look at sharing this code acorss all CXL port types */
+static void cxl_dsp_dvsec_write_config(PCIDevice *dev, uint32_t addr,
+                                      uint32_t val, int len)
+{
+    CXLDownstreamPort *dsp = CXL_DSP(dev);
+    CXLComponentState *cxl_cstate = &dsp->cxl_cstate;
+
+    if (range_contains(&cxl_cstate->dvsecs[EXTENSIONS_PORT_DVSEC], addr)) {
+        uint8_t *reg = &dev->config[addr];
+        addr -= cxl_cstate->dvsecs[EXTENSIONS_PORT_DVSEC].lob;
+        if (addr == PORT_CONTROL_OFFSET) {
+            if (pci_get_word(reg) & PORT_CONTROL_UNMASK_SBR) {
+                /* unmask SBR */
+            }
+            if (pci_get_word(reg) & PORT_CONTROL_ALT_MEMID_EN) {
+                /* Alt Memory & ID Space Enable */
+            }
+        }
+    }
+}
+
+static void cxl_dsp_config_write(PCIDevice *d, uint32_t address,
+                                 uint32_t val, int len)
+{
+    uint16_t slt_ctl, slt_sta;
+
+    pcie_cap_slot_get(d, &slt_ctl, &slt_sta);
+    pci_bridge_write_config(d, address, val, len);
+    pcie_cap_flr_write_config(d, address, val, len);
+    pcie_cap_slot_write_config(d, slt_ctl, slt_sta, address, val, len);
+    pcie_aer_write_config(d, address, val, len);
+
+    cxl_dsp_dvsec_write_config(d, address, val, len);
+}
+
+static void cxl_dsp_reset(DeviceState *qdev)
+{
+    PCIDevice *d = PCI_DEVICE(qdev);
+    CXLDownstreamPort *dsp = CXL_DSP(qdev);
+
+    pcie_cap_deverr_reset(d);
+    pcie_cap_slot_reset(d);
+    pcie_cap_arifwd_reset(d);
+    pci_bridge_reset(qdev);
+
+    latch_registers(dsp);
+}
+
+static void build_dvsecs(CXLComponentState *cxl)
+{
+    uint8_t *dvsec;
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_port_extensions){ 0 };
+    cxl_component_create_dvsec(cxl, EXTENSIONS_PORT_DVSEC_LENGTH,
+                               EXTENSIONS_PORT_DVSEC,
+                               EXTENSIONS_PORT_DVSEC_REVID, dvsec);
+    dvsec = (uint8_t *)&(struct cxl_dvsec_port_flexbus){
+        .cap                     = 0x26, /* IO, Mem, non-MLD */
+        .ctrl                    = 0,
+        .status                  = 0x26, /* same */
+        .rcvd_mod_ts_data_phase1 = 0xef, /* WTF? */
+    };
+    cxl_component_create_dvsec(cxl, PCIE_FLEXBUS_PORT_DVSEC_LENGTH_2_0,
+                               PCIE_FLEXBUS_PORT_DVSEC,
+                               PCIE_FLEXBUS_PORT_DVSEC_REVID_2_0, dvsec);
+
+    dvsec = (uint8_t *)&(struct cxl_dvsec_register_locator){
+        .rsvd         = 0,
+        .reg0_base_lo = RBI_COMPONENT_REG | CXL_COMPONENT_REG_BAR_IDX,
+        .reg0_base_hi = 0,
+    };
+    cxl_component_create_dvsec(cxl, REG_LOC_DVSEC_LENGTH, REG_LOC_DVSEC,
+                               REG_LOC_DVSEC_REVID, dvsec);
+}
+
+static void cxl_dsp_realize(PCIDevice *d, Error **errp)
+{
+    PCIEPort *p = PCIE_PORT(d);
+    PCIESlot *s = PCIE_SLOT(d);
+    CXLDownstreamPort *dsp = CXL_DSP(d);
+    CXLComponentState *cxl_cstate = &dsp->cxl_cstate;
+    ComponentRegisters *cregs = &cxl_cstate->crb;
+    MemoryRegion *component_bar = &cregs->component_registers;
+    int rc;
+
+    pci_bridge_initfn(d, TYPE_PCIE_BUS);
+    pcie_port_init_reg(d);
+
+    rc = msi_init(d, CXL_DOWNSTREAM_PORT_MSI_OFFSET,
+                  CXL_DOWNSTREAM_PORT_MSI_NR_VECTOR,
+                  true, true, errp);
+    if (rc) {
+        assert(rc == -ENOTSUP);
+        goto err_bridge;
+    }
+
+    rc = pcie_cap_init(d, CXL_DOWNSTREAM_PORT_EXP_OFFSET,
+                       PCI_EXP_TYPE_DOWNSTREAM, p->port,
+                       errp);
+    if (rc < 0) {
+        goto err_msi;
+    }
+
+    pcie_cap_flr_init(d);
+    pcie_cap_deverr_init(d);
+    pcie_cap_slot_init(d, s);
+    pcie_cap_arifwd_init(d);
+
+    pcie_chassis_create(s->chassis);
+    rc = pcie_chassis_add_slot(s);
+    if (rc < 0) {
+        error_setg(errp, "Can't add chassis slot, error %d", rc);
+        goto err_pcie_cap;
+    }
+
+    rc = pcie_aer_init(d, PCI_ERR_VER, CXL_DOWNSTREAM_PORT_AER_OFFSET,
+                       PCI_ERR_SIZEOF, errp);
+    if (rc < 0) {
+        goto err_chassis;
+    }
+
+    cxl_cstate->dvsec_offset = CXL_DOWNSTREAM_PORT_DVSEC_OFFSET;
+    cxl_cstate->pdev = d;
+    build_dvsecs(cxl_cstate);
+    cxl_component_register_block_init(OBJECT(d), cxl_cstate, TYPE_CXL_DSP);
+    pci_register_bar(d, CXL_COMPONENT_REG_BAR_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY |
+                         PCI_BASE_ADDRESS_MEM_TYPE_64,
+                     component_bar);
+
+    return;
+
+ err_chassis:
+    pcie_chassis_del_slot(s);
+ err_pcie_cap:
+    pcie_cap_exit(d);
+ err_msi:
+    msi_uninit(d);
+ err_bridge:
+    pci_bridge_exitfn(d);
+}
+
+static void cxl_dsp_exitfn(PCIDevice *d)
+{
+    PCIESlot *s = PCIE_SLOT(d);
+
+    pcie_aer_exit(d);
+    pcie_chassis_del_slot(s);
+    pcie_cap_exit(d);
+    msi_uninit(d);
+    pci_bridge_exitfn(d);
+}
+
+static void cxl_dsp_class_init(ObjectClass *oc, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(oc);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(oc);
+
+    k->is_bridge = true;
+    k->config_write = cxl_dsp_config_write;
+    k->realize = cxl_dsp_realize;
+    k->exit = cxl_dsp_exitfn;
+    k->vendor_id = 0x19e5; /* Huawei */
+    k->device_id = 0xa129; /* Emulated CXL Switch Downstream Port */
+    k->revision = 0;
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->desc = "CXL Switch Downstream Port";
+    dc->reset = cxl_dsp_reset;
+}
+
+static const TypeInfo cxl_dsp_info = {
+    .name = TYPE_CXL_DSP,
+    .instance_size = sizeof(CXLDownstreamPort),
+    .parent = TYPE_PCIE_SLOT,
+    .class_init = cxl_dsp_class_init,
+    .interfaces = (InterfaceInfo[]) {
+        { INTERFACE_PCIE_DEVICE },
+        { INTERFACE_CXL_DEVICE },
+        { }
+    },
+};
+
+static void cxl_dsp_register_type(void)
+{
+    type_register_static(&cxl_dsp_info);
+}
+
+type_init(cxl_dsp_register_type);
hw/pci-bridge/meson.build +1 -1
index 4a8c26b5a1..53cc53314a 100644
@@ -5,7 +5,7 @@ pci_ss.add(when: 'CONFIG_IOH3420', if_true: files('ioh3420.c'))
 pci_ss.add(when: 'CONFIG_PCIE_PORT', if_true: files('pcie_root_port.c', 'gen_pcie_root_port.c', 'pcie_pci_bridge.c'))
 pci_ss.add(when: 'CONFIG_PXB', if_true: files('pci_expander_bridge.c'))
 pci_ss.add(when: 'CONFIG_XIO3130', if_true: files('xio3130_upstream.c', 'xio3130_downstream.c'))
-pci_ss.add(when: 'CONFIG_CXL', if_true: files('cxl_root_port.c', 'cxl_upstream.c'))
+pci_ss.add(when: 'CONFIG_CXL', if_true: files('cxl_root_port.c', 'cxl_upstream.c', 'cxl_downstream.c'))
 
 # NewWorld PowerMac
 pci_ss.add(when: 'CONFIG_DEC_PCI', if_true: files('dec.c'))
Signature · 1 line
2.32.0
#69 lore raw

PATCH v8 45/46 cxl/cxl-host: Support interleave decoding with one level of switches.

In reply to Jonathan Cameron

Extend the walk of the CXL bus during interleave decoding to take into account one layer of switches.

Whilst theoretically CXL 2.0 allows multiple switch levels, in the vast majority of usecases only one level is expected and currently that is all the proposed Linux support provides.

---
 hw/cxl/cxl-host.c | 44 ++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 42 insertions(+), 2 deletions(-)
hw/cxl/cxl-host.c +42 -2
index a1eafa89bb..ac20d9e2f5 100644
@@ -130,8 +130,9 @@ static bool cxl_hdm_find_target(uint32_t *cache_mem, hwaddr addr,
 
 static PCIDevice *cxl_cfmws_find_device(CXLFixedWindow *fw, hwaddr addr)
 {
-    CXLComponentState *hb_cstate;
+    CXLComponentState *hb_cstate, *usp_cstate;
     PCIHostState *hb;
+    CXLUpstreamPort *usp;
     int rb_index;
     uint32_t *cache_mem;
     uint8_t target;
@@ -166,7 +167,46 @@ static PCIDevice *cxl_cfmws_find_device(CXLFixedWindow *fw, hwaddr addr)
 
     d = pci_bridge_get_sec_bus(PCI_BRIDGE(rp))->devices[0];
 
-    if (!d || !object_dynamic_cast(OBJECT(d), TYPE_CXL_TYPE3_DEV)) {
+    if (!d) {
+        return NULL;
+    }
+
+    if (object_dynamic_cast(OBJECT(d), TYPE_CXL_TYPE3_DEV)) {
+        return d;
+    }
+
+    /*
+     * Could also be a switch.  Note only one level of switching currently
+     * supported.
+     */
+    if (!object_dynamic_cast(OBJECT(d), TYPE_CXL_USP)) {
+        return NULL;
+    }
+    usp = CXL_USP(d);
+
+    usp_cstate = cxl_usp_to_cstate(usp);
+    if (!usp_cstate) {
+        return NULL;
+    }
+
+    cache_mem = usp_cstate->crb.cache_mem_registers;
+
+    target_found = cxl_hdm_find_target(cache_mem, addr, &target);
+    if (!target_found) {
+        return NULL;
+    }
+
+    d = pcie_find_port_by_pn(&PCI_BRIDGE(d)->sec_bus, target);
+    if (!d) {
+        return NULL;
+    }
+
+    d = pci_bridge_get_sec_bus(PCI_BRIDGE(d))->devices[0];
+    if (!d) {
+        return NULL;
+    }
+
+    if (!object_dynamic_cast(OBJECT(d), TYPE_CXL_TYPE3_DEV)) {
         return NULL;
     }
 
Signature · 1 line
2.32.0
#70 lore raw

PATCH v8 46/46 docs/cxl: Add switch documentation

In reply to Jonathan Cameron

Switches were already introduced, but now we support them update the documentation to provide an example in diagram and qemu command line parameter forms.

---
 docs/system/devices/cxl.rst | 88 ++++++++++++++++++++++++++++++++++++-
 1 file changed, 86 insertions(+), 2 deletions(-)
docs/system/devices/cxl.rst +86 -2
index 6871c26efd..ebc5fb3862 100644
@@ -118,8 +118,6 @@ and associated component register access via PCI bars.
 
 CXL Switch
 ~~~~~~~~~~
-Not yet implemented in QEMU.
-
 Here we consider a simple CXL switch with only a single
 virtual hierarchy. Whilst more complex devices exist, their
 visibility to a particular host is generally the same as for
@@ -137,6 +135,10 @@ BARs.  The Upstream Port has the configuration interfaces for
 the HDM decoders which route incoming memory accesses to the
 appropriate downstream port.
 
+A CXL switch is created in a similar fashion to PCI switches
+by creating an upstream port (cxl-upstream) and a number of
+downstream ports on the internal switch bus (cxl-downstream).
+
 CXL Memory Devices - Type 3
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
 CXL type 3 devices use a PCI class code and are intended to be supported
@@ -240,6 +242,62 @@ Notes:
     they will take the Host Physical Addresses of accesses and map
     them to their own local Device Physical Address Space (DPA).
 
+Example topology involving a switch::
+
+  |<------------------SYSTEM PHYSICAL ADDRESS MAP (1)----------------->|
+  |    __________   __________________________________   __________    |
+  |   |          | |                                  | |          |   |
+  |   | CFMW 0   | |  CXL Fixed Memory Window 1       | | CFMW 1   |   |
+  |   | HB0 only | |  Configured to interleave memory | | HB1 only |   |
+  |   |          | |  memory accesses across HB0/HB1  | |          |   |
+  |   |____x_____| |__________________________________| |__________|   |
+           |             |                     |             |
+           |             |                     |             |
+           |             |                     |
+  Interleave Decoder     |                     |             |
+   Matches this HB       |                     |             |
+           \_____________|                     |_____________/
+               __________|__________      _____|_______________
+              |                     |    |                     |
+              | CXL HB 0            |    | CXL HB 1            |
+              | HB IntLv Decoders   |    | HB IntLv Decoders   |
+              | PCI/CXL Root Bus 0c |    | PCI/CXL Root Bus 0d |
+              |                     |    |                     |
+              |___x_________________|    |_____________________|
+                  |              |          |               |
+                  |
+       A HB 0 HDM Decoder
+       matches this Port
+       ___________|___
+      |  Root Port 0  |
+      |  Appears in   |
+      |  PCI topology |
+      |  As 0c:00.0   |
+      |___________x___|
+                  |
+                  |
+                  \_____________________
+                                        |
+                                        |
+            ---------------------------------------------------
+           |    Switch 0  USP as PCI 0d:00.0                   |
+           |    USP has HDM decoder which direct traffic to    |
+           |    appropiate downstream port                     |
+           |    Switch BUS appears as 0e                       |
+           |x__________________________________________________|
+            |                  |               |              |
+            |                  |               |              |
+       _____|_________   ______|______   ______|_____   ______|_______
+   (4)|     x         | |             | |            | |              |
+      | CXL Type3 0   | | CXL Type3 1 | | CXL type3 2| | CLX Type 3 3 |
+      |               | |             | |            | |              |
+      | PMEM0(Vol LSA)| | PMEM1 (...) | | PMEM2 (...)| | PMEM3 (...)  |
+      | Decoder to go | |             | |            | |              |
+      | from host PA  | | PCI 10:00.0 | | PCI 11:00.0| | PCI 12:00.0  |
+      | to device PA  | |             | |            | |              |
+      | PCI as 0f:00.0| |             | |            | |              |
+      |_______________| |_____________| |____________| |______________|
+
 Example command lines
 ---------------------
 A very simple setup with just one directly attached CXL Type 3 device::
@@ -279,6 +337,32 @@ the CXL Type3 device directly attached (no switches).::
   -device cxl-type3,bus=root_port16,memdev=cxl-mem4,lsa=cxl-lsa4,id=cxl-pmem3,size=256M \
   -cxl-fixed-memory-window targets.0=cxl.1,targets.1=cxl.2,size=4G,interleave-granularity=8k
 
+An example of 4 devices below a switch suitable for 1, 2 or 4 way interleave::
+
+  qemu-system-aarch64 -M virt,gic-version=3,cxl=on -m 4g,maxmem=8G,slots=8 -cpu max \
+  ...
+  -object memory-backend-file,id=cxl-mem0,share=on,mem-path=/tmp/cxltest.raw,size=256M \
+  -object memory-backend-file,id=cxl-mem1,share=on,mem-path=/tmp/cxltest1.raw,size=256M \
+  -object memory-backend-file,id=cxl-mem2,share=on,mem-path=/tmp/cxltest2.raw,size=256M \
+  -object memory-backend-file,id=cxl-mem3,share=on,mem-path=/tmp/cxltest3.raw,size=256M \
+  -object memory-backend-file,id=cxl-lsa0,share=on,mem-path=/tmp/lsa0.raw,size=256M \
+  -object memory-backend-file,id=cxl-lsa1,share=on,mem-path=/tmp/lsa1.raw,size=256M \
+  -object memory-backend-file,id=cxl-lsa2,share=on,mem-path=/tmp/lsa2.raw,size=256M \
+  -object memory-backend-file,id=cxl-lsa3,share=on,mem-path=/tmp/lsa3.raw,size=256M \
+  -device pxb-cxl,bus_nr=12,bus=pcie.0,id=cxl.1 \
+  -device cxl-rp,port=0,bus=cxl.1,id=root_port0,chassis=0,slot=0 \
+  -device cxl-rp,port=1,bus=cxl.1,id=root_port1,chassis=0,slot=1 \
+  -device cxl-upstream,bus=root_port0,id=us0 \
+  -device cxl-downstream,port=0,bus=us0,id=swport0,chassis=0,slot=4 \
+  -device cxl-type3,bus=swport0,memdev=cxl-mem0,lsa=cxl-lsa0,id=cxl-pmem0,size=256M \
+  -device cxl-downstream,port=1,bus=us0,id=swport1,chassis=0,slot=5 \
+  -device cxl-type3,bus=swport1,memdev=cxl-mem1,lsa=cxl-lsa1,id=cxl-pmem1,size=256M \
+  -device cxl-downstream,port=2,bus=us0,id=swport2,chassis=0,slot=6 \
+  -device cxl-type3,bus=swport2,memdev=cxl-mem2,lsa=cxl-lsa2,id=cxl-pmem2,size=256M \
+  -device cxl-downstream,port=3,bus=us0,id=swport3,chassis=0,slot=7 \
+  -device cxl-type3,bus=swport3,memdev=cxl-mem3,lsa=cxl-lsa3,id=cxl-pmem3,size=256M \
+  -cxl-fixed-memory-window targets.0=cxl.1,size=4G,interleave-granularity=4k
+  
 Kernel Configuration Options
 ----------------------------
 
Signature · 1 line
2.32.0