Skip to contentSkip to conversation

ARM: Add support for a generic PCI Express host bridge

PATCH Superseded Alexander Graf 6 participants 24 messages Jan 21, 2015 – Jan 30, 2015

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v2 cover ARM: Add support for a generic PCI Express host bridge

Linux implements a nice binding to describe a "generic" PCI Express host bridge using only device tree.

This patch set adds enough emulation logic to expose the parts that are "generic" as a simple sysbus device and maps it into ARM's virt machine.

With this patch set, we can finally spawn PCI devices on ARM VMs. I was able to have a fully DRM enabled virtual machine with VGA, e1000 and XHCI (for keyboard and mouse) up and working.

It's only a small step for QEMU, but a big step for ARM VM's usability.

v1 -> v2:

- Add documentation links
- Remove mmio_window_size
- Add define for pci range types
- Use 4 PCI INTX IRQ lines

Alexander Graf (4):
pci: Split pcie_host_mmcfg_map()
pci: Add generic PCIe host bridge
arm: Add PCIe host bridge in virt machine
pci: Move PCI VGA to pci.mak

 default-configs/alpha-softmmu.mak    |   2 -
 default-configs/arm-softmmu.mak      |   2 +
 default-configs/i386-softmmu.mak     |   2 -
 default-configs/mips-softmmu.mak     |   2 -
 default-configs/mips64-softmmu.mak   |   2 -
 default-configs/mips64el-softmmu.mak |   2 -
 default-configs/mipsel-softmmu.mak   |   2 -
 default-configs/pci.mak              |   2 +
 default-configs/ppc-softmmu.mak      |   2 -
 default-configs/ppc64-softmmu.mak    |   2 -
 default-configs/ppcemb-softmmu.mak   |   2 -
 default-configs/sparc64-softmmu.mak  |   2 -
 default-configs/x86_64-softmmu.mak   |   2 -
 hw/arm/virt.c                        | 112 +++++++++++++++++++++++--
 hw/pci-host/Makefile.objs            |   1 +
 hw/pci-host/gpex.c                   | 153 +++++++++++++++++++++++++++++++++++
 hw/pci/pcie_host.c                   |   9 ++-
 include/hw/pci-host/gpex.h           |  54 +++++++++++++
 include/hw/pci/pcie_host.h           |   1 +
 include/sysemu/device_tree.h         |   9 +++
 20 files changed, 336 insertions(+), 29 deletions(-)
 create mode 100644 hw/pci-host/gpex.c
 create mode 100644 include/hw/pci-host/gpex.h
Signature · 1 line
1.7.12.4
#1 lore raw

PATCH v2 1/4 pci: Split pcie_host_mmcfg_map()

The mmcfg space is a memory region that allows access to PCI config space in the PCIe world. To maintain abstraction layers, I would like to expose the mmcfg space as a sysbus mmio region rather than have it mapped straight into the system's memory address space though.

So this patch splits the initialization of the mmcfg space from the actual mapping, allowing us to only have an mmfg memory region without the map.

---
 hw/pci/pcie_host.c         | 9 +++++++--
 include/hw/pci/pcie_host.h | 1 +
 2 files changed, 8 insertions(+), 2 deletions(-)
hw/pci/pcie_host.c +7 -2
index 3db038f..dfb4a2b 100644
@@ -98,8 +98,7 @@ void pcie_host_mmcfg_unmap(PCIExpressHost *e)
     }
 }
 
-void pcie_host_mmcfg_map(PCIExpressHost *e, hwaddr addr,
-                         uint32_t size)
+void pcie_host_mmcfg_init(PCIExpressHost *e, uint32_t size)
 {
     assert(!(size & (size - 1)));       /* power of 2 */
     assert(size >= PCIE_MMCFG_SIZE_MIN);
@@ -107,6 +106,12 @@ void pcie_host_mmcfg_map(PCIExpressHost *e, hwaddr addr,
     e->size = size;
     memory_region_init_io(&e->mmio, OBJECT(e), &pcie_mmcfg_ops, e,
                           "pcie-mmcfg", e->size);
+}
+
+void pcie_host_mmcfg_map(PCIExpressHost *e, hwaddr addr,
+                         uint32_t size)
+{
+    pcie_host_mmcfg_init(e, size);
     e->base_addr = addr;
     memory_region_add_subregion(get_system_memory(), e->base_addr, &e->mmio);
 }
include/hw/pci/pcie_host.h +1 -0
index ff44ef6..4d23c80 100644
@@ -50,6 +50,7 @@ struct PCIExpressHost {
 };
 
 void pcie_host_mmcfg_unmap(PCIExpressHost *e);
+void pcie_host_mmcfg_init(PCIExpressHost *e, uint32_t size);
 void pcie_host_mmcfg_map(PCIExpressHost *e, hwaddr addr, uint32_t size);
 void pcie_host_mmcfg_update(PCIExpressHost *e,
                             int enable,
Signature · 1 line
1.7.12.4
Reviewed-by on v2 1/4
#2 lore raw

Re: [Qemu-devel] [PATCH v2 1/4] pci: Split pcie_host_mmcfg_map()

Quoted 11 lines · Alexander Graf

The mmcfg space is a memory region that allows access to PCI config space in the PCIe world. To maintain abstraction layers, I would like to expose the mmcfg space as a sysbus mmio region rather than have it mapped straight into the system's memory address space though.

So this patch splits the initialization of the mmcfg space from the actual mapping, allowing us to only have an mmfg memory region without the map.

...as far as it goes, but:

Really the pcie_host_mmcfg_map/unmap/update() function is just totally misguided. This functionality should be pushed upwards into hw/pci-host/q35.c which can handle its own mapping of the MMIO region into the system address space at the appropriate location/size.

In particular, at the moment q35.c will leak a bunch of stuff every time the guest unmaps and remaps the mmcfg space, because we call memory_region_init_io() over and over again on the same MMIO object (which isn't valid).

Any time you see a device with its own base address in its device struct it's a red flag that the design's probably wrong...

The size of the MMCFG region should probably be a device property. Then the subclass realize could just rely on the baseclass realize to always create the mmio region, rather than having to explicitly call a function to get it to do the right thing.

thanks
-- PMM

Reviewed-by Peter Crosthwaite
#3 lore raw

Re: [Qemu-devel] [PATCH v2 1/4] pci: Split pcie_host_mmcfg_map()

Quoted 21 lines · Peter Maydell
Quoted 11 lines · Alexander Graf

The mmcfg space is a memory region that allows access to PCI config space in the PCIe world. To maintain abstraction layers, I would like to expose the mmcfg space as a sysbus mmio region rather than have it mapped straight into the system's memory address space though.

So this patch splits the initialization of the mmcfg space from the actual mapping, allowing us to only have an mmfg memory region without the map.

...as far as it goes, but:

Really the pcie_host_mmcfg_map/unmap/update() function is just totally misguided. This functionality should be pushed upwards into hw/pci-host/q35.c which can handle its own mapping of the MMIO region into the system address space at the appropriate location/size.

Quoted 4 lines

In particular, at the moment q35.c will leak a bunch of stuff every time the guest unmaps and remaps the mmcfg space, because we call memory_region_init_io() over and over again on the same MMIO object (which isn't valid).

I used to be fine before the QOM conversion I think?

Take a look at this one (and previous patch):

commit 469b046ead0671932ff3af8d6f95045b19b186ef

Author: Paolo Bonzini <[email protected]>
Date:   Wed Jun 11 12:50:43 2014 +0200

memory: remove memory_region_destroy

The function is empty after the previous patch, so remove it.

Quoted 3 lines

Any time you see a device with its own base address in its device struct it's a red flag that the design's probably wrong...

I suspect this is not the only device that leaks memory now. Paolo?

Quoted 7 lines

The size of the MMCFG region should probably be a device property. Then the subclass realize could just rely on the baseclass realize to always create the mmio region, rather than having to explicitly call a function to get it to do the right thing.

thanks
-- PMM

Signature · 1 line
MST
#4 lore raw

Re: [Qemu-devel] [PATCH v2 1/4] pci: Split pcie_host_mmcfg_map()

Quoted 2 lines · Michael S. Tsirkin

I suspect this is not the only device that leaks memory now. Paolo?

It should be, this is a bug.

Commit d8d95814609e89e5438a3318a647ec322fc4ff16 missed this place due to a rebase error. (I can tell it's a rebase error because I have the fixed patch in a private branch).

Paolo

Author: Paolo Bonzini <[email protected]>
Date:   Wed Jun 11 12:42:01 2014 +0200

memory: convert memory_region_destroy to object_unparent

Explicitly call object_unparent in the few places where we will re-create the memory region. If the memory region is simply being destroyed as part of device teardown, let QOM handle it.

#5 lore raw

PATCH v2 2/4 pci: Add generic PCIe host bridge

In reply to Alexander Graf

With simple exposure of MMFG, ioport window, mmio window and an IRQ line we can successfully create a workable PCIe host bridge that can be mapped anywhere and only needs to get described to the OS using whatever means it likes.

This patch implements such a "generic" host bridge. It only supports a single legacy IRQ line so far. MSIs need to be handled external to the host bridge.

This device is particularly useful for the "pci-host-ecam-generic" driver in Linux.

---

v1 -> v2:

- Add documentation links
- Remove mmio_window_size
- Expose 4 INTX IRQs

---
 hw/pci-host/Makefile.objs  |   1 +
 hw/pci-host/gpex.c         | 153 +++++++++++++++++++++++++++++++++++++++++++++
 include/hw/pci-host/gpex.h |  54 ++++++++++++++++
 3 files changed, 208 insertions(+)
 create mode 100644 hw/pci-host/gpex.c
 create mode 100644 include/hw/pci-host/gpex.h
hw/pci-host/Makefile.objs +1 -0
index bb65f9c..45f1f0e 100644
@@ -15,3 +15,4 @@ common-obj-$(CONFIG_PCI_APB) += apb.o
 common-obj-$(CONFIG_FULONG) += bonito.o
 common-obj-$(CONFIG_PCI_PIIX) += piix.o
 common-obj-$(CONFIG_PCI_Q35) += q35.o
+common-obj-$(CONFIG_PCI_GENERIC) += gpex.o
hw/pci-host/gpex.c new file +153 -0

153 lines not on this page · raw message

include/hw/pci-host/gpex.h new file +54 -0

54 lines not on this page · raw message

Signature · 1 line
1.7.12.4
#6 lore raw

Re: [Qemu-devel] [PATCH v2 2/4] pci: Add generic PCIe host bridge

Looks that you forgot the "single legacy IRQ line" comment from v1.

-Mike

Quoted 258 lines · Alexander Graf

With simple exposure of MMFG, ioport window, mmio window and an IRQ line we can successfully create a workable PCIe host bridge that can be mapped anywhere and only needs to get described to the OS using whatever means it likes.

This patch implements such a "generic" host bridge. It only supports a single legacy IRQ line so far. MSIs need to be handled external to the host bridge.

This device is particularly useful for the "pci-host-ecam-generic" driver in Linux.

---

v1 -> v2:

- Add documentation links
- Remove mmio_window_size
- Expose 4 INTX IRQs

---
  hw/pci-host/Makefile.objs  |   1 +
  hw/pci-host/gpex.c         | 153 +++++++++++++++++++++++++++++++++++++++++++++
  include/hw/pci-host/gpex.h |  54 ++++++++++++++++
  3 files changed, 208 insertions(+)

create mode 100644 hw/pci-host/gpex.c
create mode 100644 include/hw/pci-host/gpex.h

hw/pci-host/Makefile.objs +1 -0
index bb65f9c..45f1f0e 100644
@@ -15,3 +15,4 @@ common-obj-$(CONFIG_PCI_APB) += apb.o
  common-obj-$(CONFIG_FULONG) += bonito.o
  common-obj-$(CONFIG_PCI_PIIX) += piix.o
  common-obj-$(CONFIG_PCI_Q35) += q35.o
+common-obj-$(CONFIG_PCI_GENERIC) += gpex.o
hw/pci-host/gpex.c new file +153 -0
new file mode 100644
index 0000000..11aaf5b
@@ -0,0 +1,153 @@
+/*
+ * QEMU Generic PCI Express Bridge Emulation
+ *
+ * Copyright (C) 2015 Alexander Graf <[email protected]>
+ *
+ * Code loosely based on q35.c.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ *
+ * Check out these documents for more information on the device:
+ *
+ * http://www.kernel.org/doc/Documentation/devicetree/bindings/pci/host-generic-pci.txt
+ * http://www.firmware.org/1275/practice/imap/imap0_9d.pdf
+ */
+#include "hw/hw.h"
+#include "hw/pci-host/gpex.h"
+
+/****************************************************************************
+ * GPEX host
+ */
+
+static void gpex_set_irq(void *opaque, int irq_num, int level)
+{
+    GPEXHost *s = opaque;
+
+    qemu_set_irq(s->irq[irq_num], level);
+}
+
+static void gpex_host_realize(DeviceState *dev, Error **errp)
+{
+    PCIHostState *pci = PCI_HOST_BRIDGE(dev);
+    GPEXHost *s = GPEX_HOST(dev);
+    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
+    PCIExpressHost *pex = PCIE_HOST_BRIDGE(dev);
+    int i;
+
+    pcie_host_mmcfg_init(pex, PCIE_MMCFG_SIZE_MIN);
+    memory_region_init(&s->io_mmio, OBJECT(s), "gpex_mmio", UINT64_MAX);
+    memory_region_init(&s->io_ioport, OBJECT(s), "gpex_ioport", 64 * 1024);
+
+    sysbus_init_mmio(sbd, &pex->mmio);
+    sysbus_init_mmio(sbd, &s->io_mmio);
+    sysbus_init_mmio(sbd, &s->io_ioport);
+    for (i = 0; i < 4; i++) {
+        sysbus_init_irq(sbd, &s->irq[i]);
+    }
+
+    pci->bus = pci_register_bus(dev, "pcie.0", gpex_set_irq,
+                                pci_swizzle_map_irq_fn, s, &s->io_mmio,
+                                &s->io_ioport, 0, 4, TYPE_PCIE_BUS);
+
+    qdev_set_parent_bus(DEVICE(&s->gpex_root), BUS(pci->bus));
+    qdev_init_nofail(DEVICE(&s->gpex_root));
+}
+
+static const char *gpex_host_root_bus_path(PCIHostState *host_bridge,
+                                          PCIBus *rootbus)
+{
+    return "0000:00";
+}
+
+static void gpex_host_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass);
+
+    hc->root_bus_path = gpex_host_root_bus_path;
+    dc->realize = gpex_host_realize;
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->fw_name = "pci";
+}
+
+static void gpex_host_initfn(Object *obj)
+{
+    GPEXHost *s = GPEX_HOST(obj);
+
+    object_initialize(&s->gpex_root, sizeof(s->gpex_root), TYPE_GPEX_ROOT_DEVICE);
+    object_property_add_child(OBJECT(s), "gpex_root", OBJECT(&s->gpex_root), NULL);
+    qdev_prop_set_uint32(DEVICE(&s->gpex_root), "addr", PCI_DEVFN(0, 0));
+    qdev_prop_set_bit(DEVICE(&s->gpex_root), "multifunction", false);
+}
+
+static const TypeInfo gpex_host_info = {
+    .name       = TYPE_GPEX_HOST,
+    .parent     = TYPE_PCIE_HOST_BRIDGE,
+    .instance_size = sizeof(GPEXHost),
+    .instance_init = gpex_host_initfn,
+    .class_init = gpex_host_class_init,
+};
+
+/****************************************************************************
+ * GPEX Root D0:F0
+ */
+
+static const VMStateDescription vmstate_gpex_root = {
+    .name = "gpex_root",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_PCI_DEVICE(parent_obj, GPEXRootState),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
+static void gpex_root_class_init(ObjectClass *klass, void *data)
+{
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+    DeviceClass *dc = DEVICE_CLASS(klass);
+
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->desc = "Host bridge";
+    dc->vmsd = &vmstate_gpex_root;
+    k->vendor_id = PCI_VENDOR_ID_REDHAT;
+    k->device_id = PCI_DEVICE_ID_REDHAT_BRIDGE;
+    k->revision = 0;
+    k->class_id = PCI_CLASS_BRIDGE_HOST;
+    /*
+     * PCI-facing part of the host bridge, not usable without the
+     * host-facing part, which can't be device_add'ed, yet.
+     */
+    dc->cannot_instantiate_with_device_add_yet = true;
+}
+
+static const TypeInfo gpex_root_info = {
+    .name = TYPE_GPEX_ROOT_DEVICE,
+    .parent = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(GPEXRootState),
+    .class_init = gpex_root_class_init,
+};
+
+static void gpex_register(void)
+{
+    type_register_static(&gpex_root_info);
+    type_register_static(&gpex_host_info);
+}
+
+type_init(gpex_register);
include/hw/pci-host/gpex.h new file +54 -0
new file mode 100644
index 0000000..b465667
@@ -0,0 +1,54 @@
+/*
+ * gpex.h
+ *
+ * Copyright (C) 2015 Alexander Graf <[email protected]>
+ *
+ * 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 Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>
+ */
+
+#ifndef HW_GPEX_H
+#define HW_GPEX_H
+
+#include "hw/hw.h"
+#include "hw/sysbus.h"
+#include "hw/pci/pci.h"
+#include "hw/pci/pcie_host.h"
+
+#define TYPE_GPEX_HOST "gpex-pcihost"
+#define GPEX_HOST(obj) \
+     OBJECT_CHECK(GPEXHost, (obj), TYPE_GPEX_HOST)
+
+#define TYPE_GPEX_ROOT_DEVICE "gpex-root"
+#define MCH_PCI_DEVICE(obj) \
+     OBJECT_CHECK(GPEXRootState, (obj), TYPE_GPEX_ROOT_DEVICE)
+
+typedef struct GPEXRootState {
+    /*< private >*/
+    PCIDevice parent_obj;
+    /*< public >*/
+} GPEXRootState;
+
+typedef struct GPEXHost {
+    /*< private >*/
+    PCIExpressHost parent_obj;
+    /*< public >*/
+
+    GPEXRootState gpex_root;
+
+    MemoryRegion io_ioport;
+    MemoryRegion io_mmio;
+    qemu_irq irq[4];
+} GPEXHost;
+
+#endif /* HW_GPEX_H */
#7 lore raw

Re: [Qemu-devel] [PATCH v2 2/4] pci: Add generic PCIe host bridge

In reply to Alexander Graf
Quoted 1 line · Alexander Graf

With simple exposure of MMFG, ioport window, mmio window and an IRQ line we

Four :-)

Quoted 6 lines

can successfully create a workable PCIe host bridge that can be mapped anywhere and only needs to get described to the OS using whatever means it likes.

This patch implements such a "generic" host bridge. It only supports a single legacy IRQ line so far. MSIs need to be handled external to the host bridge.

Quoted 6 lines

This device is particularly useful for the "pci-host-ecam-generic" driver in Linux.

Checkpatch complains about a few over-80-chars lines; the URL is an unavoidable one but could you fold the others, please?

hw/pci-host/gpex.c line 53

Quoted 9 lines
hw/pci-host/gpex.c new file +9 -0
+static void gpex_host_realize(DeviceState *dev, Error **errp)
+{
+    PCIHostState *pci = PCI_HOST_BRIDGE(dev);
+    GPEXHost *s = GPEX_HOST(dev);
+    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
+    PCIExpressHost *pex = PCIE_HOST_BRIDGE(dev);
+    int i;
+
+    pcie_host_mmcfg_init(pex, PCIE_MMCFG_SIZE_MIN);

So this gives us 1MB of ECAM (config) space. That means we can't specify a target bus, so we're restricted to the 31 devices directly attached to the root complex. Among other things, this will mean we can't do PCIe hotplug. I think we should make this at least a bit bigger, even if we don't go up to the full 256MB.

Probably the best thing for the gpex device itself is to make the config space be the spec maximum (PCIE_MMCFG_SIZE_MAX) and let the user map only a portion of that into their address space if they want.

Ideally we should just do that in the base class, and get the q35 subclass to do the right thing with aliases to handle the dynamic resizing it wants. Then we wouldn't need to track "size of cfg region" in the baseclass either.

hw/pci-host/gpex.c line 60

Quoted 7 lines
hw/pci-host/gpex.c new file +7 -0
+    memory_region_init(&s->io_mmio, OBJECT(s), "gpex_mmio", UINT64_MAX);
+    memory_region_init(&s->io_ioport, OBJECT(s), "gpex_ioport", 64 * 1024);
+
+    sysbus_init_mmio(sbd, &pex->mmio);
+    sysbus_init_mmio(sbd, &s->io_mmio);
+    sysbus_init_mmio(sbd, &s->io_ioport);
+    for (i = 0; i < 4; i++) {

Can we at least have a constant rather than hardcoded 4s ? (qdev property for number-of-irqs if you're really feeling enthusiastic...)

hw/pci-host/gpex.c line 109

Quoted 4 lines
hw/pci-host/gpex.c new file +4 -0
+
+/****************************************************************************
+ * GPEX Root D0:F0
+ */

What does "D0:F0" mean here?

Quoted 8 lines
hw/pci-host/gpex.c new file +8 -0
+
+static const VMStateDescription vmstate_gpex_root = {
+    .name = "gpex_root",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_PCI_DEVICE(parent_obj, GPEXRootState),
+        VMSTATE_END_OF_LIST()

hw/pci-host/gpex.c line 127

Quoted 10 lines
hw/pci-host/gpex.c new file +10 -0
+    }
+};
+
+static void gpex_root_class_init(ObjectClass *klass, void *data)
+{
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+    DeviceClass *dc = DEVICE_CLASS(klass);
+
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->desc = "Host bridge";

We can be a bit less terse: "QEMU generic PCIe host bridge".

hw/pci-host/gpex.c line 130

Quoted 3 lines
hw/pci-host/gpex.c new file +3 -0
+    dc->vmsd = &vmstate_gpex_root;
+    k->vendor_id = PCI_VENDOR_ID_REDHAT;
+    k->device_id = PCI_DEVICE_ID_REDHAT_BRIDGE;

Pretty sure we shouldn't be reusing the PCI bridge IDs for a host bridge. We should allocate ourselves another device ID in the range...

Quoted 13 lines
hw/pci-host/gpex.c new file +13 -0
+    k->revision = 0;
+    k->class_id = PCI_CLASS_BRIDGE_HOST;
+    /*
+     * PCI-facing part of the host bridge, not usable without the
+     * host-facing part, which can't be device_add'ed, yet.
+     */
+    dc->cannot_instantiate_with_device_add_yet = true;
+}
+
+static const TypeInfo gpex_root_info = {
+    .name = TYPE_GPEX_ROOT_DEVICE,
+    .parent = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(GPEXRootState),

hw/pci-host/gpex.c line 153

Quoted 10 lines
hw/pci-host/gpex.c new file +10 -0
+    .class_init = gpex_root_class_init,
+};
+
+static void gpex_register(void)
+{
+    type_register_static(&gpex_root_info);
+    type_register_static(&gpex_host_info);
+}
+
+type_init(gpex_register);

We seem to prefer no trailing ';' on type_init by a ratio of more than 10:1.

include/hw/pci-host/gpex.h line 2

Quoted 8 lines
include/hw/pci-host/gpex.h new file +2 -0
new file mode 100644
index 0000000..b465667
@@ -0,0 +1,54 @@
+/*
+ * gpex.h

Would be nice to say "QEMU Generic PCI Express Bridge Emulation" here (like the .c file's header does).

thanks
-- PMM

#8 lore raw

Re: [Qemu-devel] [PATCH v2 2/4] pci: Add generic PCIe host bridge

Quoted 14 lines · Peter Maydell
Quoted 1 line · Alexander Graf

With simple exposure of MMFG, ioport window, mmio window and an IRQ line we

Four :-)

Quoted 9 lines

can successfully create a workable PCIe host bridge that can be mapped anywhere and only needs to get described to the OS using whatever means it likes.

This patch implements such a "generic" host bridge. It only supports a single legacy IRQ line so far. MSIs need to be handled external to the host bridge.

This device is particularly useful for the "pci-host-ecam-generic" driver in Linux.

Quoted 7 lines
Quoted 3 lines

Checkpatch complains about a few over-80-chars lines; the URL is an unavoidable one but could you fold the others, please?

Sure

Quoted 18 lines

hw/pci-host/gpex.c line 53

Quoted 10 lines
hw/pci-host/gpex.c new file +9 -0
+static void gpex_host_realize(DeviceState *dev, Error **errp)
+{
+    PCIHostState *pci = PCI_HOST_BRIDGE(dev);
+    GPEXHost *s = GPEX_HOST(dev);
+    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
+    PCIExpressHost *pex = PCIE_HOST_BRIDGE(dev);
+    int i;
+
+    pcie_host_mmcfg_init(pex, PCIE_MMCFG_SIZE_MIN);

So this gives us 1MB of ECAM (config) space. That means we can't specify a target bus, so we're restricted to the 31 devices directly attached to the root complex. Among other things, this will mean we can't do PCIe hotplug. I think we should make this at least a bit bigger, even if we don't go up to the full 256MB.

Quoted 4 lines

Probably the best thing for the gpex device itself is to make the config space be the spec maximum (PCIE_MMCFG_SIZE_MAX) and let the user map only a portion of that into their address space if they want.

Sounds like a good plan. Will do.

Quoted 14 lines

Ideally we should just do that in the base class, and get the q35 subclass to do the right thing with aliases to handle the dynamic resizing it wants. Then we wouldn't need to track "size of cfg region" in the baseclass either.

hw/pci-host/gpex.c line 60

Quoted 7 lines
hw/pci-host/gpex.c new file +7 -0
+    memory_region_init(&s->io_mmio, OBJECT(s), "gpex_mmio", UINT64_MAX);
+    memory_region_init(&s->io_ioport, OBJECT(s), "gpex_ioport", 64 * 1024);
+
+    sysbus_init_mmio(sbd, &pex->mmio);
+    sysbus_init_mmio(sbd, &s->io_mmio);
+    sysbus_init_mmio(sbd, &s->io_ioport);
+    for (i = 0; i < 4; i++) {
Quoted 3 lines

Can we at least have a constant rather than hardcoded 4s ? (qdev property for number-of-irqs if you're really feeling enthusiastic...)

Sure.

Quoted 7 lines

hw/pci-host/gpex.c line 109

Quoted 5 lines
hw/pci-host/gpex.c new file +4 -0
+
+/****************************************************************************
+ * GPEX Root D0:F0
+ */

What does "D0:F0" mean here?

Device 0 Function 0. It's pretty common PCI speech ;).

Quoted 22 lines
Quoted 9 lines
hw/pci-host/gpex.c new file +8 -0
+
+static const VMStateDescription vmstate_gpex_root = {
+    .name = "gpex_root",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_PCI_DEVICE(parent_obj, GPEXRootState),
+        VMSTATE_END_OF_LIST()

hw/pci-host/gpex.c line 127

Quoted 10 lines
hw/pci-host/gpex.c new file +10 -0
+    }
+};
+
+static void gpex_root_class_init(ObjectClass *klass, void *data)
+{
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+    DeviceClass *dc = DEVICE_CLASS(klass);
+
+    set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+    dc->desc = "Host bridge";

We can be a bit less terse: "QEMU generic PCIe host bridge".

Quoted 7 lines

hw/pci-host/gpex.c line 130

Quoted 3 lines
hw/pci-host/gpex.c new file +3 -0
+    dc->vmsd = &vmstate_gpex_root;
+    k->vendor_id = PCI_VENDOR_ID_REDHAT;
+    k->device_id = PCI_DEVICE_ID_REDHAT_BRIDGE;

Pretty sure we shouldn't be reusing the PCI bridge IDs for a host bridge. We should allocate ourselves another device ID in the range...

Ah, right, I forgot about that. Unfortunately I can't reserve PCI IDs in the Red Hat PCI range.

Michael, what's the process here?

Quoted 19 lines
Quoted 19 lines
hw/pci-host/gpex.c new file +18 -0
+    k->revision = 0;
+    k->class_id = PCI_CLASS_BRIDGE_HOST;
+    /*
+     * PCI-facing part of the host bridge, not usable without the
+     * host-facing part, which can't be device_add'ed, yet.
+     */
+    dc->cannot_instantiate_with_device_add_yet = true;
+}
+
+static const TypeInfo gpex_root_info = {
+    .name = TYPE_GPEX_ROOT_DEVICE,
+    .parent = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(GPEXRootState),
+    .class_init = gpex_root_class_init,
+};
+
+static void gpex_register(void)
+{
Quoted 8 lines

hw/pci-host/gpex.c line 153

Quoted 5 lines
hw/pci-host/gpex.c new file +5 -0
+    type_register_static(&gpex_root_info);
+    type_register_static(&gpex_host_info);
+}
+
+type_init(gpex_register);

We seem to prefer no trailing ';' on type_init by a ratio of more than 10:1.

Ok

Quoted 12 lines

include/hw/pci-host/gpex.h line 2

Quoted 9 lines
include/hw/pci-host/gpex.h new file +2 -0
new file mode 100644
index 0000000..b465667
@@ -0,0 +1,54 @@
+/*
+ * gpex.h

Would be nice to say "QEMU Generic PCI Express Bridge Emulation" here (like the .c file's header does).

Ok

Alex

#9 lore raw

Re: [Qemu-devel] [PATCH v2 2/4] pci: Add generic PCIe host bridge

Quoted 6 lines · Alexander Graf
Quoted 5 lines · Peter Maydell
Quoted 3 lines · Alexander Graf
hw/pci-host/gpex.c new file +3 -0
+    dc->vmsd = &vmstate_gpex_root;
+    k->vendor_id = PCI_VENDOR_ID_REDHAT;
+    k->device_id = PCI_DEVICE_ID_REDHAT_BRIDGE;
Quoted 8 lines
Quoted 3 lines

Pretty sure we shouldn't be reusing the PCI bridge IDs for a host bridge. We should allocate ourselves another device ID in the range...

Ah, right, I forgot about that. Unfortunately I can't reserve PCI IDs in the Red Hat PCI range.

Michael, what's the process here?

Last time this came up I think Paolo said the official registration was getting a patch into QEMU master that added a line to the list in pci.h :-) [ie QEMU is the master source for these IDs]

-- PMM

#10 lore raw

Re: [Qemu-devel] [PATCH v2 2/4] pci: Add generic PCIe host bridge

Quoted 8 lines · Peter Maydell
Quoted 4 lines

Ah, right, I forgot about that. Unfortunately I can't reserve PCI IDs in the Red Hat PCI range.

Michael, what's the process here?

Last time this came up I think Paolo said the official registration was getting a patch into QEMU master that added a line to the list in pci.h :-) [ie QEMU is the master source for these IDs]

Yeah, rocker was the first case of someone not a QEMU developer needing an id. So let's make QEMU the master source.

Paolo

#11 lore raw

PATCH v2 3/4 arm: Add PCIe host bridge in virt machine

In reply to Alexander Graf

Now that we have a working "generic" PCIe host bridge driver, we can plug it into ARMs virt machine to always have PCIe available to normal ARM VMs.

I've successfully managed to expose a Bochs VGA device, XHCI and an e1000 into an AArch64 VM with this and they all lived happily ever after.

---

Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM systems. If you want to use it with AArch64 guests, please apply the following patch or wait until upstream cleaned up the code properly:

http://csgraf.de/agraf/pci/pci-3.19.patch

v1 -> v2:

- Add define for pci range types
- Remove mmio_window_size
- Use 4 PCI INTX IRQ lines

---
 default-configs/arm-softmmu.mak |   2 +
 hw/arm/virt.c                   | 112 ++++++++++++++++++++++++++++++++++++++--
 include/sysemu/device_tree.h    |   9 ++++
 3 files changed, 118 insertions(+), 5 deletions(-)
default-configs/arm-softmmu.mak +2 -0
index f3513fa..7671ee2 100644
@@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
 CONFIG_VERSATILE_PCI=y
 CONFIG_VERSATILE_I2C=y
 
+CONFIG_PCI_GENERIC=y
+
 CONFIG_SDHCI=y
 CONFIG_INTEGRATOR_DEBUG=y
 
hw/arm/virt.c +107 -5

179 lines not on this page · raw message

include/sysemu/device_tree.h +9 -0
index 899f05c..979169b 100644
@@ -110,4 +110,13 @@ int qemu_fdt_setprop_sized_cells_from_array(void *fdt,
                                                 qdt_tmp);                 \
     })
 
+#define FDT_PCI_RANGE_RELOCATABLE          0x80000000
+#define FDT_PCI_RANGE_PREFETCHABLE         0x40000000
+#define FDT_PCI_RANGE_ALIASED              0x20000000
+#define FDT_PCI_RANGE_TYPE                 0x03000000
+#define FDT_PCI_RANGE_MMIO_64BIT           0x03000000
+#define FDT_PCI_RANGE_MMIO                 0x02000000
+#define FDT_PCI_RANGE_IOPORT               0x01000000
+#define FDT_PCI_RANGE_CONFIG               0x00000000
+
 #endif /* __DEVICE_TREE_H__ */
Signature · 1 line
1.7.12.4
#12 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Hi Alexander,

thank you for the respin. I retested with the new mappings on OSv for AArch64, and they show up ok in the guest.

Just a couple minor comments below, otherwise I think this is good.

Quoted 110 lines · Alexander Graf

Now that we have a working "generic" PCIe host bridge driver, we can plug it into ARMs virt machine to always have PCIe available to normal ARM VMs.

I've successfully managed to expose a Bochs VGA device, XHCI and an e1000 into an AArch64 VM with this and they all lived happily ever after.

---

Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM systems. If you want to use it with AArch64 guests, please apply the following patch or wait until upstream cleaned up the code properly:

http://csgraf.de/agraf/pci/pci-3.19.patch

v1 -> v2:

- Add define for pci range types
- Remove mmio_window_size
- Use 4 PCI INTX IRQ lines

---
 default-configs/arm-softmmu.mak |   2 +
 hw/arm/virt.c                   | 112 ++++++++++++++++++++++++++++++++++++++--
 include/sysemu/device_tree.h    |   9 ++++
 3 files changed, 118 insertions(+), 5 deletions(-)
default-configs/arm-softmmu.mak +2 -0
index f3513fa..7671ee2 100644
@@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
 CONFIG_VERSATILE_PCI=y
 CONFIG_VERSATILE_I2C=y
 
+CONFIG_PCI_GENERIC=y
+
 CONFIG_SDHCI=y
 CONFIG_INTEGRATOR_DEBUG=y
 
hw/arm/virt.c +9 -4
index 2353440..a727b4e 100644
@@ -42,6 +42,7 @@
 #include "exec/address-spaces.h"
 #include "qemu/bitops.h"
 #include "qemu/error-report.h"
+#include "hw/pci-host/gpex.h"
 
 #define NUM_VIRTIO_TRANSPORTS 32
 
@@ -69,6 +70,7 @@ enum {
     VIRT_MMIO,
     VIRT_RTC,
     VIRT_FW_CFG,
+    VIRT_PCIE,
 };
 
 typedef struct MemMapEntry {
@@ -129,13 +131,14 @@ static const MemMapEntry a15memmap[] = {
     [VIRT_FW_CFG] =     { 0x09020000, 0x0000000a },
     [VIRT_MMIO] =       { 0x0a000000, 0x00000200 },
     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
-    /* 0x10000000 .. 0x40000000 reserved for PCI */
+    [VIRT_PCIE] =       { 0x10000000, 0x30000000 },
     [VIRT_MEM] =        { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
 };
 
 static const int a15irqmap[] = {
     [VIRT_UART] = 1,
     [VIRT_RTC] = 2,
+    [VIRT_PCIE] = 3, /* ... to 6 */
     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 };
 
@@ -312,7 +315,7 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
     }
 }
 
-static void fdt_add_gic_node(const VirtBoardInfo *vbi)
+static uint32_t fdt_add_gic_node(const VirtBoardInfo *vbi)
 {
     uint32_t gic_phandle;
 
@@ -331,9 +334,11 @@ static void fdt_add_gic_node(const VirtBoardInfo *vbi)
                                      2, vbi->memmap[VIRT_GIC_CPU].base,
                                      2, vbi->memmap[VIRT_GIC_CPU].size);
     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", gic_phandle);
+
+    return gic_phandle;
 }
 
-static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
+static uint32_t create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
 {
     /* We create a standalone GIC v2 */
     DeviceState *gicdev;
@@ -380,7 +385,7 @@ static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
         pic[i] = qdev_get_gpio_in(gicdev, i);
     }
 
-    fdt_add_gic_node(vbi);
+    return fdt_add_gic_node(vbi);
 }
 
 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
@@ -556,6 +561,100 @@ static void create_fw_cfg(const VirtBoardInfo *vbi)
     g_free(nodename);

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 571

Quoted 10 lines
hw/arm/virt.c +8 -0
 }
 
+static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
+                                int first_irq, const char *nodename)
+{
+    int devfn, pin;
+    uint32_t full_irq_map[4 * 4 * 8] = { 0 };
+    uint32_t *irq_map = full_irq_map;
+
+    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {

devfn += 0x8 (spaces)

Quoted 12 lines
hw/arm/virt.c +12 -0
+        for (pin = 0; pin < 4; pin++) {
+            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
+            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
+            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
+            int i;
+
+            uint32_t map[] = {
+                devfn << 8, 0, 0,                           /* devfn */
+                pin + 1,                                    /* PCI pin */
+                gic_phandle, irq_type, irq_nr, irq_level }; /* GIC irq */
+
+            /* Convert map to big endian */

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 593

Quoted 10 lines
hw/arm/virt.c +10 -0
+            for (i = 0; i < 8; i++) {
+                irq_map[i] = cpu_to_be32(map[i]);
+            }
+            irq_map += 8;
+        }
+    }
+
+    qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",
+                     full_irq_map, sizeof(full_irq_map));
+}

I raise again (? or maybe for the first time) the suggestion to add a comment like:

/* see the openfirmware specs at
* http://www.firmware.org/1275/practice/imap/imap0_9d.pdf
*/

and optionally also a reference to the linux host-generic-pci bindings:

/* see the linux Documentation about device tree bindings at:
* https://www.kernel.org/doc/Documentation/devicetree/bindings/pci/host-generic-pci.txt
*/

Peter mentioned that the documentation there is not perfect, but at least it's a starting point for the reader to understand what's going on.

Quoted 100 lines
hw/arm/virt.c +68 -1
+
+static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic,
+                        uint32_t gic_phandle)
+{
+    hwaddr base = vbi->memmap[VIRT_PCIE].base;
+    hwaddr size = vbi->memmap[VIRT_PCIE].size;
+    hwaddr size_ioport = 64 * 1024;
+    hwaddr size_ecam = PCIE_MMCFG_SIZE_MIN;
+    hwaddr size_mmio = size - size_ecam - size_ioport;
+    hwaddr base_mmio = base;
+    hwaddr base_ioport = base_mmio + size_mmio;
+    hwaddr base_ecam = base_ioport + size_ioport;
+    int irq = vbi->irqmap[VIRT_PCIE];
+    MemoryRegion *mmio_alias;
+    MemoryRegion *mmio_reg;
+    DeviceState *dev;
+    char *nodename;
+    int i;
+
+    dev = qdev_create(NULL, TYPE_GPEX_HOST);
+    qdev_init_nofail(dev);
+
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base_ecam);
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_ioport);
+
+    /* Map the MMIO window at the same spot in bus and cpu layouts */
+    mmio_alias = g_new0(MemoryRegion, 1);
+    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
+    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
+                             mmio_reg, base_mmio, size_mmio);
+    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
+
+    for (i = 0; i < 4; i++) {
+        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
+    }
+
+    nodename = g_strdup_printf("/pcie@%" PRIx64, base);
+    qemu_fdt_add_subnode(vbi->fdt, nodename);
+    qemu_fdt_setprop_string(vbi->fdt, nodename,
+                            "compatible", "pci-host-ecam-generic");
+    qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "pci");
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#address-cells", 3);
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#size-cells", 2);
+    qemu_fdt_setprop_cells(vbi->fdt, nodename, "bus-range", 0, 1);
+
+    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
+                                 2, base_ecam, 2, size_ecam);
+    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
+                                 1, FDT_PCI_RANGE_IOPORT, 2, 0,
+                                 2, base_ioport, 2, size_ioport,
+
+                                 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
+                                 2, base_mmio, 2, size_mmio);
+
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#interrupt-cells", 1);
+    create_pcie_irq_map(vbi, gic_phandle, irq, nodename);
+
+    qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupt-map-mask",
+                           0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
+                           0x7           /* PCI irq */);
+
+    g_free(nodename);
+}
+
 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
 {
     const VirtBoardInfo *board = (const VirtBoardInfo *)binfo;
@@ -573,6 +672,7 @@ static void machvirt_init(MachineState *machine)
     MemoryRegion *ram = g_new(MemoryRegion, 1);
     const char *cpu_model = machine->cpu_model;
     VirtBoardInfo *vbi;
+    uint32_t gic_phandle;
 
     if (!cpu_model) {
         cpu_model = "cortex-a15";
@@ -634,12 +734,14 @@ static void machvirt_init(MachineState *machine)
 
     create_flash(vbi);
 
-    create_gic(vbi, pic);
+    gic_phandle = create_gic(vbi, pic);
 
     create_uart(vbi, pic);
 
     create_rtc(vbi, pic);
 
+    create_pcie(vbi, pic, gic_phandle);
+
     /* Create mmio transports, so the user can create virtio backends
      * (which will be automatically plugged in to the transports). If
      * no backend is created the transport will just sit harmlessly idle.
include/sysemu/device_tree.h +0 -0
index 899f05c..979169b 100644
@@ -110,4 +110,13 @@ int qemu_fdt_setprop_sized_cells_from_array(void *fdt,
                                                 qdt_tmp);                 \
     })
 

include/sysemu/device_tree.h int qemu_fdt_setprop_sized_cells_from_array(void *fdt, · line 122

Quoted 10 lines
include/sysemu/device_tree.h +9 -0
+#define FDT_PCI_RANGE_RELOCATABLE          0x80000000
+#define FDT_PCI_RANGE_PREFETCHABLE         0x40000000
+#define FDT_PCI_RANGE_ALIASED              0x20000000
+#define FDT_PCI_RANGE_TYPE                 0x03000000
+#define FDT_PCI_RANGE_MMIO_64BIT           0x03000000
+#define FDT_PCI_RANGE_MMIO                 0x02000000
+#define FDT_PCI_RANGE_IOPORT               0x01000000
+#define FDT_PCI_RANGE_CONFIG               0x00000000
+
 #endif /* __DEVICE_TREE_H__ */

I appreciate those defines.

Thank you,

Claudio

#13 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 120 lines · Claudio Fontana

Hi Alexander,

thank you for the respin. I retested with the new mappings on OSv for AArch64, and they show up ok in the guest.

Just a couple minor comments below, otherwise I think this is good.

Quoted 112 lines · Alexander Graf

Now that we have a working "generic" PCIe host bridge driver, we can plug it into ARMs virt machine to always have PCIe available to normal ARM VMs.

I've successfully managed to expose a Bochs VGA device, XHCI and an e1000 into an AArch64 VM with this and they all lived happily ever after.

---

Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM systems. If you want to use it with AArch64 guests, please apply the following patch or wait until upstream cleaned up the code properly:

http://csgraf.de/agraf/pci/pci-3.19.patch

v1 -> v2:

- Add define for pci range types
- Remove mmio_window_size
- Use 4 PCI INTX IRQ lines

---
 default-configs/arm-softmmu.mak |   2 +
 hw/arm/virt.c                   | 112 ++++++++++++++++++++++++++++++++++++++--
 include/sysemu/device_tree.h    |   9 ++++
 3 files changed, 118 insertions(+), 5 deletions(-)
default-configs/arm-softmmu.mak +2 -0
index f3513fa..7671ee2 100644
@@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
 CONFIG_VERSATILE_PCI=y
 CONFIG_VERSATILE_I2C=y
 
+CONFIG_PCI_GENERIC=y
+
 CONFIG_SDHCI=y
 CONFIG_INTEGRATOR_DEBUG=y
 
hw/arm/virt.c +9 -4
index 2353440..a727b4e 100644
@@ -42,6 +42,7 @@
 #include "exec/address-spaces.h"
 #include "qemu/bitops.h"
 #include "qemu/error-report.h"
+#include "hw/pci-host/gpex.h"
 
 #define NUM_VIRTIO_TRANSPORTS 32
 
@@ -69,6 +70,7 @@ enum {
     VIRT_MMIO,
     VIRT_RTC,
     VIRT_FW_CFG,
+    VIRT_PCIE,
 };
 
 typedef struct MemMapEntry {
@@ -129,13 +131,14 @@ static const MemMapEntry a15memmap[] = {
     [VIRT_FW_CFG] =     { 0x09020000, 0x0000000a },
     [VIRT_MMIO] =       { 0x0a000000, 0x00000200 },
     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
-    /* 0x10000000 .. 0x40000000 reserved for PCI */
+    [VIRT_PCIE] =       { 0x10000000, 0x30000000 },
     [VIRT_MEM] =        { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
 };
 
 static const int a15irqmap[] = {
     [VIRT_UART] = 1,
     [VIRT_RTC] = 2,
+    [VIRT_PCIE] = 3, /* ... to 6 */
     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 };
 
@@ -312,7 +315,7 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
     }
 }
 
-static void fdt_add_gic_node(const VirtBoardInfo *vbi)
+static uint32_t fdt_add_gic_node(const VirtBoardInfo *vbi)
 {
     uint32_t gic_phandle;
 
@@ -331,9 +334,11 @@ static void fdt_add_gic_node(const VirtBoardInfo *vbi)
                                      2, vbi->memmap[VIRT_GIC_CPU].base,
                                      2, vbi->memmap[VIRT_GIC_CPU].size);
     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", gic_phandle);
+
+    return gic_phandle;
 }
 
-static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
+static uint32_t create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
 {
     /* We create a standalone GIC v2 */
     DeviceState *gicdev;
@@ -380,7 +385,7 @@ static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
         pic[i] = qdev_get_gpio_in(gicdev, i);
     }
 
-    fdt_add_gic_node(vbi);
+    return fdt_add_gic_node(vbi);
 }
 
 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
@@ -556,6 +561,100 @@ static void create_fw_cfg(const VirtBoardInfo *vbi)
     g_free(nodename);
 }
 
Quoted 10 lines

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 571

Quoted 8 lines
hw/arm/virt.c +8 -0
+static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
+                                int first_irq, const char *nodename)
+{
+    int devfn, pin;
+    uint32_t full_irq_map[4 * 4 * 8] = { 0 };
+    uint32_t *irq_map = full_irq_map;
+
+    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {

devfn += 0x8 (spaces)

Yeah, I had it like that and it looked uglier. I guess it's a matter of personal preference?

Quoted 30 lines
Quoted 13 lines
hw/arm/virt.c +12 -0
+        for (pin = 0; pin < 4; pin++) {
+            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
+            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
+            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
+            int i;
+
+            uint32_t map[] = {
+                devfn << 8, 0, 0,                           /* devfn */
+                pin + 1,                                    /* PCI pin */
+                gic_phandle, irq_type, irq_nr, irq_level }; /* GIC irq */
+
+            /* Convert map to big endian */

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 593

Quoted 10 lines
hw/arm/virt.c +10 -0
+            for (i = 0; i < 8; i++) {
+                irq_map[i] = cpu_to_be32(map[i]);
+            }
+            irq_map += 8;
+        }
+    }
+
+    qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",
+                     full_irq_map, sizeof(full_irq_map));
+}

I raise again (? or maybe for the first time) the suggestion to add a comment like:

/* see the openfirmware specs at
* http://www.firmware.org/1275/practice/imap/imap0_9d.pdf
*/

Quoted 5 lines

and optionally also a reference to the linux host-generic-pci bindings:

/* see the linux Documentation about device tree bindings at:
* https://www.kernel.org/doc/Documentation/devicetree/bindings/pci/host-generic-pci.txt
*/

I added those links to the big description comment at the top of this file.

Alex

#14 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 134 lines · Alexander Graf
Quoted 120 lines · Claudio Fontana

Hi Alexander,

thank you for the respin. I retested with the new mappings on OSv for AArch64, and they show up ok in the guest.

Just a couple minor comments below, otherwise I think this is good.

Quoted 112 lines · Alexander Graf

Now that we have a working "generic" PCIe host bridge driver, we can plug it into ARMs virt machine to always have PCIe available to normal ARM VMs.

I've successfully managed to expose a Bochs VGA device, XHCI and an e1000 into an AArch64 VM with this and they all lived happily ever after.

---

Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM systems. If you want to use it with AArch64 guests, please apply the following patch or wait until upstream cleaned up the code properly:

http://csgraf.de/agraf/pci/pci-3.19.patch

v1 -> v2:

- Add define for pci range types
- Remove mmio_window_size
- Use 4 PCI INTX IRQ lines

---
 default-configs/arm-softmmu.mak |   2 +
 hw/arm/virt.c                   | 112 ++++++++++++++++++++++++++++++++++++++--
 include/sysemu/device_tree.h    |   9 ++++
 3 files changed, 118 insertions(+), 5 deletions(-)
default-configs/arm-softmmu.mak +2 -0
index f3513fa..7671ee2 100644
@@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
 CONFIG_VERSATILE_PCI=y
 CONFIG_VERSATILE_I2C=y
 
+CONFIG_PCI_GENERIC=y
+
 CONFIG_SDHCI=y
 CONFIG_INTEGRATOR_DEBUG=y
 
hw/arm/virt.c +9 -4
index 2353440..a727b4e 100644
@@ -42,6 +42,7 @@
 #include "exec/address-spaces.h"
 #include "qemu/bitops.h"
 #include "qemu/error-report.h"
+#include "hw/pci-host/gpex.h"
 
 #define NUM_VIRTIO_TRANSPORTS 32
 
@@ -69,6 +70,7 @@ enum {
     VIRT_MMIO,
     VIRT_RTC,
     VIRT_FW_CFG,
+    VIRT_PCIE,
 };
 
 typedef struct MemMapEntry {
@@ -129,13 +131,14 @@ static const MemMapEntry a15memmap[] = {
     [VIRT_FW_CFG] =     { 0x09020000, 0x0000000a },
     [VIRT_MMIO] =       { 0x0a000000, 0x00000200 },
     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
-    /* 0x10000000 .. 0x40000000 reserved for PCI */
+    [VIRT_PCIE] =       { 0x10000000, 0x30000000 },
     [VIRT_MEM] =        { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
 };
 
 static const int a15irqmap[] = {
     [VIRT_UART] = 1,
     [VIRT_RTC] = 2,
+    [VIRT_PCIE] = 3, /* ... to 6 */
     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 };
 
@@ -312,7 +315,7 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
     }
 }
 
-static void fdt_add_gic_node(const VirtBoardInfo *vbi)
+static uint32_t fdt_add_gic_node(const VirtBoardInfo *vbi)
 {
     uint32_t gic_phandle;
 
@@ -331,9 +334,11 @@ static void fdt_add_gic_node(const VirtBoardInfo *vbi)
                                      2, vbi->memmap[VIRT_GIC_CPU].base,
                                      2, vbi->memmap[VIRT_GIC_CPU].size);
     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", gic_phandle);
+
+    return gic_phandle;
 }
 
-static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
+static uint32_t create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
 {
     /* We create a standalone GIC v2 */
     DeviceState *gicdev;
@@ -380,7 +385,7 @@ static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
         pic[i] = qdev_get_gpio_in(gicdev, i);
     }
 
-    fdt_add_gic_node(vbi);
+    return fdt_add_gic_node(vbi);
 }
 
 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
@@ -556,6 +561,100 @@ static void create_fw_cfg(const VirtBoardInfo *vbi)
     g_free(nodename);
 }
 
Quoted 10 lines

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 571

Quoted 8 lines
hw/arm/virt.c +8 -0
+static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
+                                int first_irq, const char *nodename)
+{
+    int devfn, pin;
+    uint32_t full_irq_map[4 * 4 * 8] = { 0 };
+    uint32_t *irq_map = full_irq_map;
+
+    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {

devfn += 0x8 (spaces)

Quoted 2 lines

Yeah, I had it like that and it looked uglier. I guess it's a matter of personal preference?

You don't have coding standards for this ?

Quoted 31 lines
Quoted 31 lines
Quoted 14 lines
hw/arm/virt.c +12 -0
+        for (pin = 0; pin < 4; pin++) {
+            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
+            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
+            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
+            int i;
+
+            uint32_t map[] = {
+                devfn << 8, 0, 0,                           /* devfn */
+                pin + 1,                                    /* PCI pin */
+                gic_phandle, irq_type, irq_nr, irq_level }; /* GIC irq */
+
+            /* Convert map to big endian */

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 593

Quoted 10 lines
hw/arm/virt.c +10 -0
+            for (i = 0; i < 8; i++) {
+                irq_map[i] = cpu_to_be32(map[i]);
+            }
+            irq_map += 8;
+        }
+    }
+
+    qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",
+                     full_irq_map, sizeof(full_irq_map));
+}

I raise again (? or maybe for the first time) the suggestion to add a comment like:

/* see the openfirmware specs at
* http://www.firmware.org/1275/practice/imap/imap0_9d.pdf
*/

Quoted 8 lines
Quoted 5 lines

and optionally also a reference to the linux host-generic-pci bindings:

/* see the linux Documentation about device tree bindings at:
* https://www.kernel.org/doc/Documentation/devicetree/bindings/pci/host-generic-pci.txt
*/

I added those links to the big description comment at the top of this file.

It is not on top of this file unfortunately.
You put the description comment in an unrelated .h file: the pointers to the docs belong here I think.

Ciao,

Claudio

#15 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 11 lines · Claudio Fontana
Quoted 8 lines · Alexander Graf
Quoted 4 lines · Claudio Fontana

Alex wrote;

Quoted 1 line

+ for (devfn = 0; devfn <= 0x18; devfn+=0x8) {

devfn += 0x8 (spaces)

Yeah, I had it like that and it looked uglier. I guess it's a matter of personal preference?

You don't have coding standards for this ?

Not formal ones, but (a) I think += should have spaces aronud it and (b) I'm a bit surprised if checkpatch doesn't catch this (it certainly attempts to...)

-- PMM

#16 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 11 lines · Peter Maydell
Quoted 10 lines · Claudio Fontana
Quoted 8 lines · Alexander Graf
Quoted 4 lines · Claudio Fontana

Alex wrote;

Quoted 1 line

+ for (devfn = 0; devfn <= 0x18; devfn+=0x8) {

devfn += 0x8 (spaces)

Yeah, I had it like that and it looked uglier. I guess it's a matter of personal preference?

Quoted 7 lines
Quoted 1 line

You don't have coding standards for this ?

Not formal ones, but (a) I think += should have spaces aronud it and (b) I'm a bit surprised if checkpatch doesn't catch this (it certainly attempts to...)

-- PMM

It does:

ERROR: spaces required around that '+=' (ctx:VxV)
#167: FILE: hw/arm/virt.c:571:

+    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {
                                         ^

Claudio

#17 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

In reply to Alexander Graf
Quoted 2 lines · Alexander Graf

Now that we have a working "generic" PCIe host bridge driver, we can plug it into ARMs virt machine to always have PCIe available to normal ARM VMs.

"ARM's".

Quoted 60 lines

I've successfully managed to expose a Bochs VGA device, XHCI and an e1000 into an AArch64 VM with this and they all lived happily ever after.

---

Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM systems. If you want to use it with AArch64 guests, please apply the following patch or wait until upstream cleaned up the code properly:

http://csgraf.de/agraf/pci/pci-3.19.patch

v1 -> v2:

- Add define for pci range types
- Remove mmio_window_size
- Use 4 PCI INTX IRQ lines

---
 default-configs/arm-softmmu.mak |   2 +
 hw/arm/virt.c                   | 112 ++++++++++++++++++++++++++++++++++++++--
 include/sysemu/device_tree.h    |   9 ++++
 3 files changed, 118 insertions(+), 5 deletions(-)
default-configs/arm-softmmu.mak +2 -0
index f3513fa..7671ee2 100644
@@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
 CONFIG_VERSATILE_PCI=y
 CONFIG_VERSATILE_I2C=y
 
+CONFIG_PCI_GENERIC=y
+
 CONFIG_SDHCI=y
 CONFIG_INTEGRATOR_DEBUG=y
 
hw/arm/virt.c +2 -0
index 2353440..a727b4e 100644
@@ -42,6 +42,7 @@
 #include "exec/address-spaces.h"
 #include "qemu/bitops.h"
 #include "qemu/error-report.h"
+#include "hw/pci-host/gpex.h"
 
 #define NUM_VIRTIO_TRANSPORTS 32
 
@@ -69,6 +70,7 @@ enum {
     VIRT_MMIO,
     VIRT_RTC,
     VIRT_FW_CFG,
+    VIRT_PCIE,
 };
 
 typedef struct MemMapEntry {

hw/arm/virt.c static const MemMapEntry a15memmap[] = { · line 134

Quoted 5 lines
hw/arm/virt.c +1 -1
@@ -129,13 +131,14 @@ static const MemMapEntry a15memmap[] = {
     [VIRT_FW_CFG] =     { 0x09020000, 0x0000000a },
     [VIRT_MMIO] =       { 0x0a000000, 0x00000200 },
     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
-    /* 0x10000000 .. 0x40000000 reserved for PCI */
+    [VIRT_PCIE] =       { 0x10000000, 0x30000000 },

Might be nice to have a comment about how this is split up between MMIO, IO and config space.

Quoted 73 lines
hw/arm/virt.c +34 -3
     [VIRT_MEM] =        { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
 };
 
 static const int a15irqmap[] = {
     [VIRT_UART] = 1,
     [VIRT_RTC] = 2,
+    [VIRT_PCIE] = 3, /* ... to 6 */
     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 };
 
@@ -312,7 +315,7 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
     }
 }
 
-static void fdt_add_gic_node(const VirtBoardInfo *vbi)
+static uint32_t fdt_add_gic_node(const VirtBoardInfo *vbi)
 {
     uint32_t gic_phandle;
 
@@ -331,9 +334,11 @@ static void fdt_add_gic_node(const VirtBoardInfo *vbi)
                                      2, vbi->memmap[VIRT_GIC_CPU].base,
                                      2, vbi->memmap[VIRT_GIC_CPU].size);
     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", gic_phandle);
+
+    return gic_phandle;
 }
 
-static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
+static uint32_t create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
 {
     /* We create a standalone GIC v2 */
     DeviceState *gicdev;
@@ -380,7 +385,7 @@ static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
         pic[i] = qdev_get_gpio_in(gicdev, i);
     }
 
-    fdt_add_gic_node(vbi);
+    return fdt_add_gic_node(vbi);
 }
 
 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
@@ -556,6 +561,100 @@ static void create_fw_cfg(const VirtBoardInfo *vbi)
     g_free(nodename);
 }
 
+static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
+                                int first_irq, const char *nodename)
+{
+    int devfn, pin;
+    uint32_t full_irq_map[4 * 4 * 8] = { 0 };
+    uint32_t *irq_map = full_irq_map;
+
+    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {
+        for (pin = 0; pin < 4; pin++) {
+            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
+            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
+            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
+            int i;
+
+            uint32_t map[] = {
+                devfn << 8, 0, 0,                           /* devfn */
+                pin + 1,                                    /* PCI pin */
+                gic_phandle, irq_type, irq_nr, irq_level }; /* GIC irq */
+
+            /* Convert map to big endian */
+            for (i = 0; i < 8; i++) {
+                irq_map[i] = cpu_to_be32(map[i]);
+            }
+            irq_map += 8;
+        }
+    }
+
+    qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 601

Quoted 10 lines
hw/arm/virt.c +10 -0
+                     full_irq_map, sizeof(full_irq_map));
+}
+
+static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic,
+                        uint32_t gic_phandle)
+{
+    hwaddr base = vbi->memmap[VIRT_PCIE].base;
+    hwaddr size = vbi->memmap[VIRT_PCIE].size;
+    hwaddr size_ioport = 64 * 1024;
+    hwaddr size_ecam = PCIE_MMCFG_SIZE_MIN;

See remarks in previous patch about wanting more than the minimum ECAM space.

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 605

Quoted 4 lines
hw/arm/virt.c +4 -0
+    hwaddr size_mmio = size - size_ecam - size_ioport;
+    hwaddr base_mmio = base;
+    hwaddr base_ioport = base_mmio + size_mmio;
+    hwaddr base_ecam = base_ioport + size_ioport;

PCIe spec says the ECAM window has to be aligned to its size (ie if it's 1MB in size it has to be 1MB aligned, if it's 256MB aligned it has to be 256MB aligned). I think this is dumb, but it's what the spec says...

Quoted 9 lines
hw/arm/virt.c +9 -0
+    int irq = vbi->irqmap[VIRT_PCIE];
+    MemoryRegion *mmio_alias;
+    MemoryRegion *mmio_reg;
+    DeviceState *dev;
+    char *nodename;
+    int i;
+
+    dev = qdev_create(NULL, TYPE_GPEX_HOST);
+    qdev_init_nofail(dev);

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 624

Quoted 10 lines
hw/arm/virt.c +10 -0
+
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base_ecam);
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_ioport);
+
+    /* Map the MMIO window at the same spot in bus and cpu layouts */
+    mmio_alias = g_new0(MemoryRegion, 1);
+    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
+    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
+                             mmio_reg, base_mmio, size_mmio);
+    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);

The comment claims to be mapping the MMIO window twice (in the system memory space and in the PCI mmio address space) but the code only seems to be mapping something into system memory space?

Quoted 19 lines
hw/arm/virt.c +19 -0
+
+    for (i = 0; i < 4; i++) {
+        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
+    }
+
+    nodename = g_strdup_printf("/pcie@%" PRIx64, base);
+    qemu_fdt_add_subnode(vbi->fdt, nodename);
+    qemu_fdt_setprop_string(vbi->fdt, nodename,
+                            "compatible", "pci-host-ecam-generic");
+    qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "pci");
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#address-cells", 3);
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#size-cells", 2);
+    qemu_fdt_setprop_cells(vbi->fdt, nodename, "bus-range", 0, 1);
+
+    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
+                                 2, base_ecam, 2, size_ecam);
+    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
+                                 1, FDT_PCI_RANGE_IOPORT, 2, 0,
+                                 2, base_ioport, 2, size_ioport,

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 653

Quoted 10 lines
hw/arm/virt.c +10 -0
+
+                                 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
+                                 2, base_mmio, 2, size_mmio);
+
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#interrupt-cells", 1);
+    create_pcie_irq_map(vbi, gic_phandle, irq, nodename);
+
+    qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupt-map-mask",
+                           0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
+                           0x7           /* PCI irq */);

I think at least the interrupt-map-mask and maybe also the #interrupt-cells setprop calls should go inside the create_pcie_irq_map() function. In particular the interrupt-map and interrupt-map-mask are meaningless unless interpreted together so they should go in the same function.

Quoted 36 lines
hw/arm/virt.c +8 -1
+
+    g_free(nodename);
+}
+
 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
 {
     const VirtBoardInfo *board = (const VirtBoardInfo *)binfo;
@@ -573,6 +672,7 @@ static void machvirt_init(MachineState *machine)
     MemoryRegion *ram = g_new(MemoryRegion, 1);
     const char *cpu_model = machine->cpu_model;
     VirtBoardInfo *vbi;
+    uint32_t gic_phandle;
 
     if (!cpu_model) {
         cpu_model = "cortex-a15";
@@ -634,12 +734,14 @@ static void machvirt_init(MachineState *machine)
 
     create_flash(vbi);
 
-    create_gic(vbi, pic);
+    gic_phandle = create_gic(vbi, pic);
 
     create_uart(vbi, pic);
 
     create_rtc(vbi, pic);
 
+    create_pcie(vbi, pic, gic_phandle);
+
     /* Create mmio transports, so the user can create virtio backends
      * (which will be automatically plugged in to the transports). If
      * no backend is created the transport will just sit harmlessly idle.
include/sysemu/device_tree.h +0 -0
index 899f05c..979169b 100644

include/sysemu/device_tree.h int qemu_fdt_setprop_sized_cells_from_array(void *fdt, · line 117

Quoted 8 lines
include/sysemu/device_tree.h +5 -0
@@ -110,4 +110,13 @@ int qemu_fdt_setprop_sized_cells_from_array(void *fdt,
                                                 qdt_tmp);                 \
     })
 
+#define FDT_PCI_RANGE_RELOCATABLE          0x80000000
+#define FDT_PCI_RANGE_PREFETCHABLE         0x40000000
+#define FDT_PCI_RANGE_ALIASED              0x20000000
+#define FDT_PCI_RANGE_TYPE                 0x03000000
+#define FDT_PCI_RANGE_MMIO_64BIT           0x03000000

I had to go dig up the spec to figure out that it wasn't an error that these two had the same values. Calling the first _TYPE_MASK might help?

include/sysemu/device_tree.h int qemu_fdt_setprop_sized_cells_from_array(void *fdt, · line 120

Quoted 3 lines
include/sysemu/device_tree.h +3 -0
+#define FDT_PCI_RANGE_MMIO                 0x02000000
+#define FDT_PCI_RANGE_IOPORT               0x01000000
+#define FDT_PCI_RANGE_CONFIG               0x00000000

thanks
-- PMM

#18 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 64 lines · Peter Maydell
Quoted 2 lines · Alexander Graf

Now that we have a working "generic" PCIe host bridge driver, we can plug it into ARMs virt machine to always have PCIe available to normal ARM VMs.

"ARM's".

Quoted 58 lines

I've successfully managed to expose a Bochs VGA device, XHCI and an e1000 into an AArch64 VM with this and they all lived happily ever after.

---

Linux 3.19 only supports the generic PCIe host bridge driver for 32bit ARM systems. If you want to use it with AArch64 guests, please apply the following patch or wait until upstream cleaned up the code properly:

http://csgraf.de/agraf/pci/pci-3.19.patch

v1 -> v2:

- Add define for pci range types
- Remove mmio_window_size
- Use 4 PCI INTX IRQ lines

---
 default-configs/arm-softmmu.mak |   2 +
 hw/arm/virt.c                   | 112 ++++++++++++++++++++++++++++++++++++++--
 include/sysemu/device_tree.h    |   9 ++++
 3 files changed, 118 insertions(+), 5 deletions(-)
default-configs/arm-softmmu.mak +2 -0
index f3513fa..7671ee2 100644
@@ -82,6 +82,8 @@ CONFIG_ZYNQ=y
 CONFIG_VERSATILE_PCI=y
 CONFIG_VERSATILE_I2C=y
 
+CONFIG_PCI_GENERIC=y
+
 CONFIG_SDHCI=y
 CONFIG_INTEGRATOR_DEBUG=y
 
hw/arm/virt.c +2 -0
index 2353440..a727b4e 100644
@@ -42,6 +42,7 @@
 #include "exec/address-spaces.h"
 #include "qemu/bitops.h"
 #include "qemu/error-report.h"
+#include "hw/pci-host/gpex.h"
 
 #define NUM_VIRTIO_TRANSPORTS 32
 
@@ -69,6 +70,7 @@ enum {
     VIRT_MMIO,
     VIRT_RTC,
     VIRT_FW_CFG,
+    VIRT_PCIE,
 };
 
Quoted 10 lines

hw/arm/virt.c static const MemMapEntry a15memmap[] = { · line 134

Quoted 7 lines
hw/arm/virt.c +1 -1
 typedef struct MemMapEntry {
@@ -129,13 +131,14 @@ static const MemMapEntry a15memmap[] = {
     [VIRT_FW_CFG] =     { 0x09020000, 0x0000000a },
     [VIRT_MMIO] =       { 0x0a000000, 0x00000200 },
     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
-    /* 0x10000000 .. 0x40000000 reserved for PCI */
+    [VIRT_PCIE] =       { 0x10000000, 0x30000000 },

Might be nice to have a comment about how this is split up between MMIO, IO and config space.

I'd prefer to keep the details of that in the allocation code so that we don't potentially have 2 places that diverge eventually. But I'll document the allocation code a bit nicer and add a small overview split here.

Quoted 93 lines
Quoted 74 lines
hw/arm/virt.c +34 -3
     [VIRT_MEM] =        { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
 };
 
 static const int a15irqmap[] = {
     [VIRT_UART] = 1,
     [VIRT_RTC] = 2,
+    [VIRT_PCIE] = 3, /* ... to 6 */
     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 };
 
@@ -312,7 +315,7 @@ static void fdt_add_cpu_nodes(const VirtBoardInfo *vbi)
     }
 }
 
-static void fdt_add_gic_node(const VirtBoardInfo *vbi)
+static uint32_t fdt_add_gic_node(const VirtBoardInfo *vbi)
 {
     uint32_t gic_phandle;
 
@@ -331,9 +334,11 @@ static void fdt_add_gic_node(const VirtBoardInfo *vbi)
                                      2, vbi->memmap[VIRT_GIC_CPU].base,
                                      2, vbi->memmap[VIRT_GIC_CPU].size);
     qemu_fdt_setprop_cell(vbi->fdt, "/intc", "phandle", gic_phandle);
+
+    return gic_phandle;
 }
 
-static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
+static uint32_t create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
 {
     /* We create a standalone GIC v2 */
     DeviceState *gicdev;
@@ -380,7 +385,7 @@ static void create_gic(const VirtBoardInfo *vbi, qemu_irq *pic)
         pic[i] = qdev_get_gpio_in(gicdev, i);
     }
 
-    fdt_add_gic_node(vbi);
+    return fdt_add_gic_node(vbi);
 }
 
 static void create_uart(const VirtBoardInfo *vbi, qemu_irq *pic)
@@ -556,6 +561,100 @@ static void create_fw_cfg(const VirtBoardInfo *vbi)
     g_free(nodename);
 }
 
+static void create_pcie_irq_map(const VirtBoardInfo *vbi, uint32_t gic_phandle,
+                                int first_irq, const char *nodename)
+{
+    int devfn, pin;
+    uint32_t full_irq_map[4 * 4 * 8] = { 0 };
+    uint32_t *irq_map = full_irq_map;
+
+    for (devfn = 0; devfn <= 0x18; devfn+=0x8) {
+        for (pin = 0; pin < 4; pin++) {
+            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
+            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
+            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
+            int i;
+
+            uint32_t map[] = {
+                devfn << 8, 0, 0,                           /* devfn */
+                pin + 1,                                    /* PCI pin */
+                gic_phandle, irq_type, irq_nr, irq_level }; /* GIC irq */
+
+            /* Convert map to big endian */
+            for (i = 0; i < 8; i++) {
+                irq_map[i] = cpu_to_be32(map[i]);
+            }
+            irq_map += 8;
+        }
+    }
+
+    qemu_fdt_setprop(vbi->fdt, nodename, "interrupt-map",

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 601

Quoted 10 lines
hw/arm/virt.c +10 -0
+                     full_irq_map, sizeof(full_irq_map));
+}
+
+static void create_pcie(const VirtBoardInfo *vbi, qemu_irq *pic,
+                        uint32_t gic_phandle)
+{
+    hwaddr base = vbi->memmap[VIRT_PCIE].base;
+    hwaddr size = vbi->memmap[VIRT_PCIE].size;
+    hwaddr size_ioport = 64 * 1024;
+    hwaddr size_ecam = PCIE_MMCFG_SIZE_MIN;

See remarks in previous patch about wanting more than the minimum ECAM space.

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 605

Quoted 4 lines
hw/arm/virt.c +4 -0
+    hwaddr size_mmio = size - size_ecam - size_ioport;
+    hwaddr base_mmio = base;
+    hwaddr base_ioport = base_mmio + size_mmio;
+    hwaddr base_ecam = base_ioport + size_ioport;
Quoted 4 lines

PCIe spec says the ECAM window has to be aligned to its size (ie if it's 1MB in size it has to be 1MB aligned, if it's 256MB aligned it has to be 256MB aligned). I think this is dumb, but it's what the spec says...

No problem, fortunately we have a good number of nice macros for this ;)

Quoted 16 lines
Quoted 16 lines
hw/arm/virt.c +15 -0
+    int irq = vbi->irqmap[VIRT_PCIE];
+    MemoryRegion *mmio_alias;
+    MemoryRegion *mmio_reg;
+    DeviceState *dev;
+    char *nodename;
+    int i;
+
+    dev = qdev_create(NULL, TYPE_GPEX_HOST);
+    qdev_init_nofail(dev);
+
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base_ecam);
+    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_ioport);
+
+    /* Map the MMIO window at the same spot in bus and cpu layouts */
+    mmio_alias = g_new0(MemoryRegion, 1);
Quoted 8 lines

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 624

Quoted 4 lines
hw/arm/virt.c +4 -0
+    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
+    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
+                             mmio_reg, base_mmio, size_mmio);
+    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);

The comment claims to be mapping the MMIO window twice (in the system memory space and in the PCI mmio address space) but the code only seems to be mapping something into system memory space?

The comment claims to map it at the same spot. It means the offset in system memory is the same offset as the one in the mmio window that gets exported by the PHB.

The PHB exports a UINT64_MAX MMIO region as id 1. This is basically a full window into the PCI address space. What we do here is to map a 1:1 window between CPU address space and PCI address space.

Quoted 28 lines
Quoted 28 lines
hw/arm/virt.c +27 -0
+
+    for (i = 0; i < 4; i++) {
+        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
+    }
+
+    nodename = g_strdup_printf("/pcie@%" PRIx64, base);
+    qemu_fdt_add_subnode(vbi->fdt, nodename);
+    qemu_fdt_setprop_string(vbi->fdt, nodename,
+                            "compatible", "pci-host-ecam-generic");
+    qemu_fdt_setprop_string(vbi->fdt, nodename, "device_type", "pci");
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#address-cells", 3);
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#size-cells", 2);
+    qemu_fdt_setprop_cells(vbi->fdt, nodename, "bus-range", 0, 1);
+
+    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "reg",
+                                 2, base_ecam, 2, size_ecam);
+    qemu_fdt_setprop_sized_cells(vbi->fdt, nodename, "ranges",
+                                 1, FDT_PCI_RANGE_IOPORT, 2, 0,
+                                 2, base_ioport, 2, size_ioport,
+
+                                 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
+                                 2, base_mmio, 2, size_mmio);
+
+    qemu_fdt_setprop_cell(vbi->fdt, nodename, "#interrupt-cells", 1);
+    create_pcie_irq_map(vbi, gic_phandle, irq, nodename);
+
+    qemu_fdt_setprop_cells(vbi->fdt, nodename, "interrupt-map-mask",
Quoted 8 lines

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 653

Quoted 2 lines
hw/arm/virt.c +2 -0
+                           0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
+                           0x7           /* PCI irq */);

I think at least the interrupt-map-mask and maybe also the #interrupt-cells setprop calls should go inside the create_pcie_irq_map() function. In particular the interrupt-map and interrupt-map-mask are meaningless unless interpreted together so they should go in the same function.

Sure, works for me.

Quoted 40 lines
Quoted 40 lines
hw/arm/virt.c +8 -1
+
+    g_free(nodename);
+}
+
 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
 {
     const VirtBoardInfo *board = (const VirtBoardInfo *)binfo;
@@ -573,6 +672,7 @@ static void machvirt_init(MachineState *machine)
     MemoryRegion *ram = g_new(MemoryRegion, 1);
     const char *cpu_model = machine->cpu_model;
     VirtBoardInfo *vbi;
+    uint32_t gic_phandle;
 
     if (!cpu_model) {
         cpu_model = "cortex-a15";
@@ -634,12 +734,14 @@ static void machvirt_init(MachineState *machine)
 
     create_flash(vbi);
 
-    create_gic(vbi, pic);
+    gic_phandle = create_gic(vbi, pic);
 
     create_uart(vbi, pic);
 
     create_rtc(vbi, pic);
 
+    create_pcie(vbi, pic, gic_phandle);
+
     /* Create mmio transports, so the user can create virtio backends
      * (which will be automatically plugged in to the transports). If
      * no backend is created the transport will just sit harmlessly idle.
include/sysemu/device_tree.h +0 -0
index 899f05c..979169b 100644
@@ -110,4 +110,13 @@ int qemu_fdt_setprop_sized_cells_from_array(void *fdt,
                                                 qdt_tmp);                 \
     })
 
Quoted 8 lines

include/sysemu/device_tree.h int qemu_fdt_setprop_sized_cells_from_array(void *fdt, · line 117

Quoted 5 lines
include/sysemu/device_tree.h +5 -0
+#define FDT_PCI_RANGE_RELOCATABLE          0x80000000
+#define FDT_PCI_RANGE_PREFETCHABLE         0x40000000
+#define FDT_PCI_RANGE_ALIASED              0x20000000
+#define FDT_PCI_RANGE_TYPE                 0x03000000
+#define FDT_PCI_RANGE_MMIO_64BIT           0x03000000

I had to go dig up the spec to figure out that it wasn't an error that these two had the same values. Calling the first _TYPE_MASK might help?

Probably :)

Alex

#19 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 15 lines · Alexander Graf
Quoted 11 lines · Peter Maydell

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 624

Quoted 6 lines · Alexander Graf
hw/arm/virt.c +6 -0
+    /* Map the MMIO window at the same spot in bus and cpu layouts */
+    mmio_alias = g_new0(MemoryRegion, 1);
+    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
+    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
+                             mmio_reg, base_mmio, size_mmio);
+    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);

The comment claims to be mapping the MMIO window twice (in the system memory space and in the PCI mmio address space) but the code only seems to be mapping something into system memory space?

Quoted 7 lines

The comment claims to map it at the same spot. It means the offset in system memory is the same offset as the one in the mmio window that gets exported by the PHB.

The PHB exports a UINT64_MAX MMIO region as id 1. This is basically a full window into the PCI address space. What we do here is to map a 1:1 window between CPU address space and PCI address space.

I kind of see, but isn't this just a window from CPU address space into PCI address space, not vice-versa?

DMA by PCI devices bus-mastering into system memory must be being set up elsewhere, I think.

-- PMM

#20 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 20 lines · Peter Maydell
Quoted 19 lines · Alexander Graf
Quoted 11 lines · Peter Maydell

hw/arm/virt.c static void create_fw_cfg(const VirtBoardInfo *vbi) · line 624

Quoted 6 lines · Alexander Graf
hw/arm/virt.c +6 -0
+    /* Map the MMIO window at the same spot in bus and cpu layouts */
+    mmio_alias = g_new0(MemoryRegion, 1);
+    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
+    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
+                             mmio_reg, base_mmio, size_mmio);
+    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);

The comment claims to be mapping the MMIO window twice (in the system memory space and in the PCI mmio address space) but the code only seems to be mapping something into system memory space?

The comment claims to map it at the same spot. It means the offset in system memory is the same offset as the one in the mmio window that gets exported by the PHB.

Quoted 6 lines
Quoted 3 lines

The PHB exports a UINT64_MAX MMIO region as id 1. This is basically a full window into the PCI address space. What we do here is to map a 1:1 window between CPU address space and PCI address space.

I kind of see, but isn't this just a window from CPU address space into PCI address space, not vice-versa?

Yup, exactly. But PCI devices need to map themselves somewhere into the PCI address space. So if I configure a BAR to live at 0x10000000, it should also show up at 0x10000000 when accessed from the CPU. That's what the mapping above is about.

Quoted 2 lines

DMA by PCI devices bus-mastering into system memory must be being set up elsewhere, I think.

Yes, that's a different mechanism that's not implemented yet for GPEX :). On ARM this would happen via SMMU emulation.

Alex

#21 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 8 lines · Alexander Graf
Quoted 2 lines · Peter Maydell

I kind of see, but isn't this just a window from CPU address space into PCI address space, not vice-versa?

Yup, exactly. But PCI devices need to map themselves somewhere into the PCI address space. So if I configure a BAR to live at 0x10000000, it should also show up at 0x10000000 when accessed from the CPU. That's what the mapping above is about.

No, it doesn't have to. It's a choice to make the mapping be such that the system address for a BAR matches the address in PCI memory space, not a requirement. I agree it's a sensible choice, though.

But as I say, this code is setting up one mapping (the system address -> PCI space mapping), not two.

Quoted 5 lines
Quoted 2 lines

DMA by PCI devices bus-mastering into system memory must be being set up elsewhere, I think.

Yes, that's a different mechanism that's not implemented yet for GPEX :).

We can't not implement DMA, it would break lots of the usual PCI devices people want to use. In fact I thought the PCI core code implemented a default of "DMA by PCI devices goes to the system address space" if you didn't specifically set up something else by calling pci_setup_iommu(). This is definitely how it works for plain PCI host bridges, are PCIe bridges different?

Quoted 1 line

On ARM this would happen via SMMU emulation.

There's no requirement for a PCI host controller to be sat behind an SMMU -- that's a system design choice. We don't need to implement the SMMU yet (or perhaps ever?); we definitely need to support PCI DMA.

-- PMM

#22 lore raw

Re: [Qemu-devel] [PATCH v2 3/4] arm: Add PCIe host bridge in virt machine

Quoted 10 lines · Peter Maydell
Quoted 8 lines · Alexander Graf
Quoted 2 lines · Peter Maydell

I kind of see, but isn't this just a window from CPU address space into PCI address space, not vice-versa?

Yup, exactly. But PCI devices need to map themselves somewhere into the PCI address space. So if I configure a BAR to live at 0x10000000, it should also show up at 0x10000000 when accessed from the CPU. That's what the mapping above is about.

Quoted 7 lines

No, it doesn't have to. It's a choice to make the mapping be such that the system address for a BAR matches the address in PCI memory space, not a requirement. I agree it's a sensible choice, though.

But as I say, this code is setting up one mapping (the system address -> PCI space mapping), not two.

Yes, the other one is done implicitly by the OS based on what device tree tells it to do. If we map it at 0, our good old if (BAR == 0) break; friend hits us again though - and any other arbitrary offset is as good as a 1:1 map.

Quoted 7 lines
Quoted 6 lines
Quoted 3 lines

DMA by PCI devices bus-mastering into system memory must be being set up elsewhere, I think.

Yes, that's a different mechanism that's not implemented yet for GPEX :).

Quoted 7 lines

We can't not implement DMA, it would break lots of the usual PCI devices people want to use. In fact I thought the PCI core code implemented a default of "DMA by PCI devices goes to the system address space" if you didn't specifically set up something else by calling pci_setup_iommu(). This is definitely how it works for plain PCI host bridges, are PCIe bridges different?

Nono, this is exactly the way it works. The thing that's not implemented is the SMMU to make that dynamic.

Quoted 5 lines
Quoted 1 line

On ARM this would happen via SMMU emulation.

There's no requirement for a PCI host controller to be sat behind an SMMU -- that's a system design choice. We don't need to implement the SMMU yet (or perhaps ever?);

The main benefit of implementing a guest SMMU is that you don't have to pin all guest memory at all times. Apart from that and the usual security aid it only makes things slower ;).

Quoted 1 line

we definitely need to support PCI DMA.

We do.

Alex

#23 lore raw

PATCH v2 4/4 pci: Move PCI VGA to pci.mak

In reply to Alexander Graf

Every platform that supports PCI can also spawn the Bochs VGA PCI adapter. Move it to pci.mak to enable it for everyone.

---
 default-configs/alpha-softmmu.mak    | 2 --
 default-configs/i386-softmmu.mak     | 2 --
 default-configs/mips-softmmu.mak     | 2 --
 default-configs/mips64-softmmu.mak   | 2 --
 default-configs/mips64el-softmmu.mak | 2 --
 default-configs/mipsel-softmmu.mak   | 2 --
 default-configs/pci.mak              | 2 ++
 default-configs/ppc-softmmu.mak      | 2 --
 default-configs/ppc64-softmmu.mak    | 2 --
 default-configs/ppcemb-softmmu.mak   | 2 --
 default-configs/sparc64-softmmu.mak  | 2 --
 default-configs/x86_64-softmmu.mak   | 2 --
 12 files changed, 2 insertions(+), 22 deletions(-)
default-configs/alpha-softmmu.mak +0 -2
index bc07600..7f6161e 100644
@@ -5,8 +5,6 @@ include usb.mak
 CONFIG_SERIAL=y
 CONFIG_I8254=y
 CONFIG_PCKBD=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_VGA_CIRRUS=y
 CONFIG_IDE_CORE=y
 CONFIG_IDE_QDEV=y
default-configs/i386-softmmu.mak +0 -2
index 8e08841..bd99af9 100644
@@ -3,9 +3,7 @@
 include pci.mak
 include sound.mak
 include usb.mak
-CONFIG_VGA=y
 CONFIG_QXL=$(CONFIG_SPICE)
-CONFIG_VGA_PCI=y
 CONFIG_VGA_ISA=y
 CONFIG_VGA_CIRRUS=y
 CONFIG_VMWARE_VGA=y
default-configs/mips-softmmu.mak +0 -2
index 2a80b04..cce2c81 100644
@@ -4,8 +4,6 @@ include pci.mak
 include sound.mak
 include usb.mak
 CONFIG_ESP=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_VGA_ISA=y
 CONFIG_VGA_ISA_MM=y
 CONFIG_VGA_CIRRUS=y
default-configs/mips64-softmmu.mak +0 -2
index f1f933b..7a88a08 100644
@@ -4,8 +4,6 @@ include pci.mak
 include sound.mak
 include usb.mak
 CONFIG_ESP=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_VGA_ISA=y
 CONFIG_VGA_ISA_MM=y
 CONFIG_VGA_CIRRUS=y
default-configs/mips64el-softmmu.mak +0 -2
index 317b151..095de43 100644
@@ -4,8 +4,6 @@ include pci.mak
 include sound.mak
 include usb.mak
 CONFIG_ESP=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_VGA_ISA=y
 CONFIG_VGA_ISA_MM=y
 CONFIG_VGA_CIRRUS=y
default-configs/mipsel-softmmu.mak +0 -2
index 7708185..0e25108 100644
@@ -4,8 +4,6 @@ include pci.mak
 include sound.mak
 include usb.mak
 CONFIG_ESP=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_VGA_ISA=y
 CONFIG_VGA_ISA_MM=y
 CONFIG_VGA_CIRRUS=y
default-configs/pci.mak +2 -0
index a186c39..89986f1 100644
@@ -32,3 +32,5 @@ CONFIG_PCI_TESTDEV=y
 CONFIG_NVME_PCI=y
 CONFIG_SD=y
 CONFIG_SDHCI=y
+CONFIG_VGA=y
+CONFIG_VGA_PCI=y
default-configs/ppc-softmmu.mak +0 -2
index d725b23..aebfab9 100644
@@ -6,8 +6,6 @@ include usb.mak
 CONFIG_ISA_MMIO=y
 CONFIG_ESCC=y
 CONFIG_M48T59=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_SERIAL=y
 CONFIG_PARALLEL=y
 CONFIG_I8254=y
default-configs/ppc64-softmmu.mak +0 -2
index bd30d69..f195a87 100644
@@ -6,8 +6,6 @@ include usb.mak
 CONFIG_ISA_MMIO=y
 CONFIG_ESCC=y
 CONFIG_M48T59=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_SERIAL=y
 CONFIG_PARALLEL=y
 CONFIG_I8254=y
default-configs/ppcemb-softmmu.mak +0 -2
index e032761..a1b3d5f 100644
@@ -4,8 +4,6 @@ include pci.mak
 include sound.mak
 include usb.mak
 CONFIG_M48T59=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_SERIAL=y
 CONFIG_I8257=y
 CONFIG_OPENPIC=y
default-configs/sparc64-softmmu.mak +0 -2
index 299c97b..123bb99 100644
@@ -5,8 +5,6 @@ include usb.mak
 CONFIG_ISA_MMIO=y
 CONFIG_M48T59=y
 CONFIG_PTIMER=y
-CONFIG_VGA=y
-CONFIG_VGA_PCI=y
 CONFIG_SERIAL=y
 CONFIG_PARALLEL=y
 CONFIG_PCKBD=y
default-configs/x86_64-softmmu.mak +0 -2
index 66557ac..e7c2734 100644
@@ -3,9 +3,7 @@
 include pci.mak
 include sound.mak
 include usb.mak
-CONFIG_VGA=y
 CONFIG_QXL=$(CONFIG_SPICE)
-CONFIG_VGA_PCI=y
 CONFIG_VGA_ISA=y
 CONFIG_VGA_CIRRUS=y
 CONFIG_VMWARE_VGA=y
Signature · 1 line
1.7.12.4