Skip to contentSkip to conversation

vfio/pci: Base support to preserve a VFIO device file across Live Update

PATCH In review David Matlack 8 participants 62 messages Nov 26, 2025 – Jan 6

Subsystems (7)
KVM Kselftest PCI Live update +3 more subsystems
lore
Share & tools
lore

David Matlack

author maintainer
#0 lore raw

cover vfio/pci: Base support to preserve a VFIO device file across Live Update

This series adds the base support to preserve a VFIO device file across a Live Update. "Base support" means that this allows userspace to safetly preserve a VFIO device file with LIVEUPDATE_SESSION_PRESERVE_FD and retrieve a preserved VFIO device file with LIVEUPDATE_SESSION_RETRIEVE_FD, but the device itself is not preserved in a fully running state across Live Update.

This series unblocks 2 parallel but related streams of work:

- iommufd preservation across Live Update. This work spans iommufd, the IOMMU subsystem, and IOMMU drivers [1]

- Preservation of VFIO device state across Live Update (config space, BAR addresses, power state, SR-IOV state, etc.). This work spans both VFIO and the core PCI subsystem.

While we need all of the above to fully preserve a VFIO device across a Live Update without disrupting the workload on the device, this series aims to be functional and safe enough to merge as the first incremental step toward that goal.

Areas for Discussion
--------------------

BDF Stability across Live Update

The PCI support for tracking preserved devices across a Live Update to prevent auto-probing relies on PCI segment numbers and BDFs remaining stable. For now I have disallowed VFs, as the BDFs assigned to VFs can vary depending on how the kernel chooses to allocate bus numbers. For non-VFs I am wondering if there is any more needed to ensure BDF stability across Live Update.

While we would like to support many different systems and configurations in due time (including preserving VFs), I'd like to keep this first serses constrained to simple use-cases.

FLB Locking

I don't see a way to properly synchronize pci_flb_finish() with pci_liveupdate_incoming_is_preserved() since the incoming FLB mutex is dropped by liveupdate_flb_get_incoming() when it returns the pointer to the object, and taking pci_flb_incoming_lock in pci_flb_finish() could result in a deadlock due to reversing the lock ordering.

FLB Retrieving

The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

Testing
-------

The patches at the end of this series provide comprehensive selftests for the new code added by this series. The selftests have been validated in both a VM environment using a virtio-net PCIe device, and in a baremetal environment on an Intel EMR server with an Intel DSA device.

Here is an example of how to run the new selftests:

vfio_pci_liveupdate_uapi_test:

  $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
  $ tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test 0000:00:04.0
  $ tools/testing/selftests/vfio/scripts/cleanup.sh

vfio_pci_liveupdate_kexec_test:

  $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
  $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test --stage 1 0000:00:04.0
  $ kexec [...]  # NOTE: distro-dependent
  $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
  $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test --stage 2 0000:00:04.0
  $ tools/testing/selftests/vfio/scripts/cleanup.sh

Dependencies
------------

This series was constructed on top of several in-flight series and on top of mm-nonmm-unstable [2].

  +-- This series
  |
  +-- [PATCH v2 00/18] vfio: selftests: Support for multi-device tests
  |    https://lore.kernel.org/kvm/[email protected]/
  |
  +-- [PATCH v3 0/4] vfio: selftests: update DMA mapping tests to use queried IOVA ranges
  |   https://lore.kernel.org/kvm/[email protected]/
  |
  +-- [PATCH v8 0/2] Live Update: File-Lifecycle-Bound (FLB) State
  |   https://lore.kernel.org/linux-mm/[email protected]/
  |
  +-- [PATCH v8 00/18] Live Update Orchestrator
  |   https://lore.kernel.org/linux-mm/[email protected]/
  |

To simplify checking out the code, this series can be found on GitHub:

https://github.com/dmatlack/linux/tree/liveupdate/vfio/cdev/v1

Changelog
---------

v1:
- Rebase series on top of LUOv8 and VFIO selftests improvements
- Drop commits to preserve config space fields across Live Update. These changes require changes to the PCI layer. For exmaple, preserving rbars could lead to an inconsistent device state until device BARs addresses are preserved across Live Update.
- Drop commits to preserve Bus Master Enable on the device. There's no reason to preserve this until iommufd preservation is fully working. Furthermore, preserving Bus Master Enable could lead to memory corruption when the device if the device is bound to the default identity-map domain after Live Update.
- Drop commits to preserve saved PCI state. This work is not needed until we are ready to preserve the device's config space, and requires more thought to make the PCI state data layout ABI-friendly.
- Add support to skip auto-probing devices that are preserved by VFIO to avoid them getting bound to a different driver by the next kernel.
- Restrict device preservation further (no VFs, no intel-graphics).
- Various refactoring and small edits to improve readability and eliminate code duplication.

rfc: https://lore.kernel.org/kvm/[email protected]/

Cc Saeed Mahameed <[email protected]>
Cc Adithya Jayachandran <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Parav Pandit <[email protected]>
Cc Leon Romanovsky <[email protected]>
Cc William Tu <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Lukas Wunner <[email protected]>
Cc Pasha Tatashin <[email protected]>
Cc Mike Rapoport <[email protected]>
Cc Pratyush Yadav <[email protected]>
Cc Samiullah Khawaja <[email protected]>
Cc Chris Li <[email protected]>
Cc Josh Hilke <[email protected]>
Cc David Rientjes <[email protected]>

[1] https://lore.kernel.org/linux-iommu/[email protected]/
[2] https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git/log/?h=mm-nonmm-unstable

David Matlack (12):
liveupdate: luo_flb: Prevent retrieve() after finish()
PCI: Add API to track PCI devices preserved across Live Update
PCI: Require driver_override for incoming Live Update preserved
devices
vfio/pci: Notify PCI subsystem about devices preserved across Live
Update
vfio: Enforce preserved devices are retrieved via
LIVEUPDATE_SESSION_RETRIEVE_FD
vfio/pci: Store Live Update state in struct vfio_pci_core_device
vfio: selftests: Add Makefile support for TEST_GEN_PROGS_EXTENDED
vfio: selftests: Add vfio_pci_liveupdate_uapi_test
vfio: selftests: Expose iommu_modes to tests
vfio: selftests: Expose low-level helper routines for setting up
struct vfio_pci_device
vfio: selftests: Verify that opening VFIO device fails during Live
Update
vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

Vipin Sharma (9):
vfio/pci: Register a file handler with Live Update Orchestrator vfio/pci: Preserve vfio-pci device files across Live Update vfio/pci: Retrieve preserved device files after Live Update vfio/pci: Skip reset of preserved device after Live Update selftests/liveupdate: Move luo_test_utils.* into a reusable library selftests/liveupdate: Add helpers to preserve/retrieve FDs
vfio: selftests: Build liveupdate library in VFIO selftests
vfio: selftests: Initialize vfio_pci_device using a VFIO cdev FD
vfio: selftests: Add vfio_pci_liveupdate_kexec_test

 MAINTAINERS                                   |   1 +
 drivers/pci/Makefile                          |   1 +
 drivers/pci/liveupdate.c                      | 248 ++++++++++++++++
 drivers/pci/pci-driver.c                      |  12 +-
 drivers/vfio/device_cdev.c                    |  25 +-
 drivers/vfio/group.c                          |   9 +
 drivers/vfio/pci/Makefile                     |   1 +
 drivers/vfio/pci/vfio_pci.c                   |  11 +-
 drivers/vfio/pci/vfio_pci_core.c              |  23 +-
 drivers/vfio/pci/vfio_pci_liveupdate.c        | 278 ++++++++++++++++++
 drivers/vfio/pci/vfio_pci_priv.h              |  16 +
 drivers/vfio/vfio.h                           |  13 -
 drivers/vfio/vfio_main.c                      |  22 +-
 include/linux/kho/abi/pci.h                   |  53 ++++
 include/linux/kho/abi/vfio_pci.h              |  45 +++
 include/linux/liveupdate.h                    |   3 +
 include/linux/pci.h                           |  38 +++
 include/linux/vfio.h                          |  51 ++++
 include/linux/vfio_pci_core.h                 |   7 +
 kernel/liveupdate/luo_flb.c                   |   4 +
 tools/testing/selftests/liveupdate/.gitignore |   1 +
 tools/testing/selftests/liveupdate/Makefile   |  14 +-
 .../include/libliveupdate.h}                  |  11 +-
 .../selftests/liveupdate/lib/libliveupdate.mk |  20 ++
 .../{luo_test_utils.c => lib/liveupdate.c}    |  43 ++-
 .../selftests/liveupdate/luo_kexec_simple.c   |   2 +-
 .../selftests/liveupdate/luo_multi_session.c  |   2 +-
 tools/testing/selftests/vfio/Makefile         |  23 +-
 .../vfio/lib/include/libvfio/iommu.h          |   2 +
 .../lib/include/libvfio/vfio_pci_device.h     |   8 +
 tools/testing/selftests/vfio/lib/iommu.c      |   4 +-
 .../selftests/vfio/lib/vfio_pci_device.c      |  60 +++-
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 255 ++++++++++++++++
 .../vfio/vfio_pci_liveupdate_uapi_test.c      |  93 ++++++
 34 files changed, 1313 insertions(+), 86 deletions(-)
 create mode 100644 drivers/pci/liveupdate.c
 create mode 100644 drivers/vfio/pci/vfio_pci_liveupdate.c
 create mode 100644 include/linux/kho/abi/pci.h
 create mode 100644 include/linux/kho/abi/vfio_pci.h
 rename tools/testing/selftests/liveupdate/{luo_test_utils.h => lib/include/libliveupdate.h} (80%)
 create mode 100644 tools/testing/selftests/liveupdate/lib/libliveupdate.mk
 rename tools/testing/selftests/liveupdate/{luo_test_utils.c => lib/liveupdate.c} (89%)
 create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
 create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#1 lore raw

PATCH 1/21 liveupdate: luo_flb: Prevent retrieve() after finish()

Prevent an incoming FLB from being retrieved after its finish() callback has been run. An incoming FLB should only retrieved once, and once its finish() callback has run the data has been freed and cannot be (and should not be) retrieved again.

Fixes e8c57e582167 ("liveupdate: luo_flb: Introduce File-Lifecycle-Bound global state")
---
 include/linux/liveupdate.h  | 3 +++
 kernel/liveupdate/luo_flb.c | 4 ++++
 2 files changed, 7 insertions(+)
include/linux/liveupdate.h +3 -0
index ed81e7b31a9f..b913d63eab5f 100644
@@ -165,12 +165,15 @@ struct liveupdate_flb_ops {
  * @obj:   The live kernel object returned by .preserve() or .retrieve().
  * @lock:  A mutex that protects all fields within this structure, providing
  *         the synchronization service for the FLB's ops.
+ * @finished: True once the FLB's finish() callback has run, to prevent an FLB
+ *            from being retrieve()'d again after finish() (incoming only).
  */
 struct luo_flb_private_state {
 	long count;
 	u64 data;
 	void *obj;
 	struct mutex lock;
+	bool finished;
 };
 
 /*
kernel/liveupdate/luo_flb.c +4 -0
index e80ac5b575ec..072796fb75cb 100644
@@ -155,6 +155,9 @@ static int luo_flb_retrieve_one(struct liveupdate_flb *flb)
 
 	guard(mutex)(&private->incoming.lock);
 
+	if (private->incoming.finished)
+		return -ENODATA;
+
 	if (private->incoming.obj)
 		return 0;
 
@@ -213,6 +216,7 @@ static void luo_flb_file_finish_one(struct liveupdate_flb *flb)
 
 			private->incoming.data = 0;
 			private->incoming.obj = NULL;
+			private->incoming.finished = true;
 		}
 	}
 }
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#2 lore raw

PATCH 2/21 PCI: Add API to track PCI devices preserved across Live Update

In reply to David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device. VFs are not supported for now, since that requires preserving SR-IOV state on the device to ensure the same number of VFs appear after kexec and with the same BDFs.

Drivers that preserve devices across Live Update can now register their struct liveupdate_file_handler with the PCI subsystem so that the PCI subsystem can allocate and manage File-Lifecycle-Bound (FLB) global data to track the list of incoming and outgoing preserved devices.

pci_liveupdate_register_fh(driver_fh)
pci_liveupdate_unregister_fh(driver_fh)

Drivers can notify the PCI subsystem whenever a device is preserved and unpreserved with the following APIs:

pci_liveupdate_outgoing_preserve(pci_dev)
pci_liveupdate_outgoing_unpreserve(pci_dev)

After a Live Update, the PCI subsystem can fetch its FLB global data from the previous kernel from the Live Update Orchestrator (LUO) to determine which devices are preserved. This API is also made available for drivers to use to check if a device was preserved before userspace retrieves the file for it.

pci_liveupdate_incoming_is_preserved(pci_dev)

Once a driver has finished restoring an incoming preserved device, it can notify the PCI subsystem with the following call:

pci_liveupdate_incoming_finish(pci_dev)

This will be used in subsequent commits by the vfio-pci driver to preserve VFIO devices across Live Update.

---
 drivers/pci/Makefile        |   1 +
 drivers/pci/liveupdate.c    | 248 ++++++++++++++++++++++++++++++++++++
 include/linux/kho/abi/pci.h |  53 ++++++++
 include/linux/pci.h         |  38 ++++++
 4 files changed, 340 insertions(+)
 create mode 100644 drivers/pci/liveupdate.c
 create mode 100644 include/linux/kho/abi/pci.h
drivers/pci/Makefile +1 -0
index 67647f1880fb..0cb43e10e71d 100644
@@ -16,6 +16,7 @@ obj-$(CONFIG_PROC_FS)		+= proc.o
 obj-$(CONFIG_SYSFS)		+= pci-sysfs.o slot.o
 obj-$(CONFIG_ACPI)		+= pci-acpi.o
 obj-$(CONFIG_GENERIC_PCI_IOMAP) += iomap.o
+obj-$(CONFIG_LIVEUPDATE)	+= liveupdate.o
 endif
 
 obj-$(CONFIG_OF)		+= of.o
drivers/pci/liveupdate.c new file +248 -0

248 lines not on this page · raw message

include/linux/kho/abi/pci.h new file +53 -0

53 lines not on this page · raw message

include/linux/pci.h +38 -0
index d1fdf81fbe1e..6a3c2d7e5b82 100644
@@ -40,6 +40,7 @@
 #include <linux/resource_ext.h>
 #include <linux/msi_api.h>
 #include <uapi/linux/pci.h>
+#include <linux/liveupdate.h>
 
 #include <linux/pci_ids.h>
 
@@ -2795,4 +2796,41 @@ void pci_uevent_ers(struct pci_dev *pdev, enum  pci_ers_result err_type);
 	WARN_ONCE(condition, "%s %s: " fmt, \
 		  dev_driver_string(&(pdev)->dev), pci_name(pdev), ##arg)
 
+#ifdef CONFIG_LIVEUPDATE
+int pci_liveupdate_outgoing_preserve(struct pci_dev *dev);
+void pci_liveupdate_outgoing_unpreserve(struct pci_dev *dev);
+bool pci_liveupdate_incoming_is_preserved(struct pci_dev *dev);
+void pci_liveupdate_incoming_finish(struct pci_dev *dev);
+int pci_liveupdate_register_fh(struct liveupdate_file_handler *fh);
+int pci_liveupdate_unregister_fh(struct liveupdate_file_handler *fh);
+#else /* !CONFIG_LIVEUPDATE */
+static inline int pci_liveupdate_outgoing_preserve(struct pci_dev *dev)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline void pci_liveupdate_outgoing_unpreserve(struct pci_dev *dev)
+{
+}
+
+static inline bool pci_liveupdate_incoming_is_preserved(struct pci_dev *dev)
+{
+	return false;
+}
+
+static inline void pci_liveupdate_incoming_finish(struct pci_dev *dev)
+{
+}
+
+static inline int pci_liveupdate_register_fh(struct liveupdate_file_handler *fh)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline int pci_liveupdate_unregister_fh(struct liveupdate_file_handler *fh)
+{
+	return -EOPNOTSUPP;
+}
+#endif /* !CONFIG_LIVEUPDATE */
+
 #endif /* LINUX_PCI_H */
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
#3 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 8 lines · David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device.

While bus numbers will *usually* stay the same across next and previous kernel, there are exceptions. E.g. if "pci=assign-busses" is specified on the command line, the kernel will re-assign bus numbers on every boot.

The most portable way to identify PCI devices across kernels is to store their path from the root down the hierarchy. Because the bus number might change but the device/function number on each bus stays the same.

This is what EFI does with device paths:
https://uefi.org/specs/UEFI/2.10/10_Protocols_Device_Path_Protocol.html

Example:
Acpi(PNP0A03,0)/Pci(1E|0)/Pci(0|0)

Source:
https://raw.githubusercontent.com/tianocore-docs/edk2-UefiDriverWritersGuide/main/3_foundation/39_uefi_device_paths/README.9.md

We've got a device path *parser* in drivers/firmware/efi/dev-path-parser.c, but we don't have a *generator* for device paths in the kernel yet.

Thanks,

Lukas

maintainer
#4 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 23 lines · Lukas Wunner
Quoted 8 lines · David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device.

While bus numbers will *usually* stay the same across next and previous kernel, there are exceptions. E.g. if "pci=assign-busses" is specified on the command line, the kernel will re-assign bus numbers on every boot.

The most portable way to identify PCI devices across kernels is to store their path from the root down the hierarchy. Because the bus number might change but the device/function number on each bus stays the same.

This is what EFI does with device paths:
https://uefi.org/specs/UEFI/2.10/10_Protocols_Device_Path_Protocol.html

Quoted 8 lines

Example:
Acpi(PNP0A03,0)/Pci(1E|0)/Pci(0|0)

Source:
https://raw.githubusercontent.com/tianocore-docs/edk2-UefiDriverWritersGuide/main/3_foundation/39_uefi_device_paths/README.9.md

We've got a device path *parser* in drivers/firmware/efi/dev-path-parser.c, but we don't have a *generator* for device paths in the kernel yet.

Hi Lukas,

Thanks for the input.

You are right that bus numbers can change in standard boot scenarios. However, for Live Update, we skip firmware, and we would likely list pci=assign-busses as an unsupported parameter. So, BDF should be sufficient.

That said, if there is a better method using a stable hierarchical path, and more importantly, if that method can be extended to other bus types, we are open to considering it. The main hurdle is that we would need a way to generate this stable path in the kernel and also parse it during early boot.

Thanks,
Pasha

#5 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

In reply to Lukas Wunner
Quoted 6 lines · Lukas Wunner
Quoted 5 lines · David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Quoted 7 lines
Quoted 3 lines

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device.

While bus numbers will *usually* stay the same across next and previous kernel, there are exceptions. E.g. if "pci=assign-busses" is specified on the command line, the kernel will re-assign bus numbers on every boot.

Stuff like this has to be disabled for this live update stuff, if the bus numbers are changed it will break the active use of the iommu across the kexec.

So while what you say is all technically true, I'm not sure this is necessary.

Jason

maintainer
#6 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 16 lines · Jason Gunthorpe
Quoted 6 lines · Lukas Wunner
Quoted 5 lines · David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Quoted 7 lines
Quoted 3 lines

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device.

While bus numbers will *usually* stay the same across next and previous kernel, there are exceptions. E.g. if "pci=assign-busses" is specified on the command line, the kernel will re-assign bus numbers on every boot.

Quoted 6 lines

Stuff like this has to be disabled for this live update stuff, if the bus numbers are changed it will break the active use of the iommu across the kexec.

So while what you say is all technically true, I'm not sure this is necessary.

I agree. However, Lukas's comment made me wonder about the future: if we eventually need to preserve non-PCI devices (like a TPM), should we be designing a common identification mechanism for all buses now? Or should we settle on BDF for PCI and invent stable identifiers for other bus types as they become necessary?

Pasha

Quoted 2 lines

Jason

#7 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 21 lines · Pasha Tatashin
Quoted 20 lines · Jason Gunthorpe
Quoted 6 lines · Lukas Wunner
Quoted 5 lines · David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Quoted 7 lines
Quoted 3 lines

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device.

While bus numbers will *usually* stay the same across next and previous kernel, there are exceptions. E.g. if "pci=assign-busses" is specified on the command line, the kernel will re-assign bus numbers on every boot.

Stuff like this has to be disabled for this live update stuff, if the bus numbers are changed it will break the active use of the iommu across the kexec.

Quoted 8 lines
Quoted 2 lines

So while what you say is all technically true, I'm not sure this is necessary.

I agree. However, Lukas's comment made me wonder about the future: if we eventually need to preserve non-PCI devices (like a TPM), should we be designing a common identification mechanism for all buses now? Or should we settle on BDF for PCI and invent stable identifiers for other bus types as they become necessary?

Well, at least PCI subsystem should use BDF..

You are probably right that the matching of preserved data to a struct device should be more general though.

Jason

author maintainer
#8 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 31 lines · Jason Gunthorpe
Quoted 21 lines · Pasha Tatashin
Quoted 20 lines · Jason Gunthorpe
Quoted 6 lines · Lukas Wunner
Quoted 5 lines · David Matlack

Add an API to enable the PCI subsystem to track all devices that are preserved across a Live Update, including both incoming devices (passed from the previous kernel) and outgoing devices (passed to the next kernel).

Quoted 7 lines
Quoted 3 lines

Use PCI segment number and BDF to keep track of devices across Live Update. This means the kernel must keep both identifiers constant across a Live Update for any preserved device.

While bus numbers will *usually* stay the same across next and previous kernel, there are exceptions. E.g. if "pci=assign-busses" is specified on the command line, the kernel will re-assign bus numbers on every boot.

Stuff like this has to be disabled for this live update stuff, if the bus numbers are changed it will break the active use of the iommu across the kexec.

Quoted 8 lines
Quoted 2 lines

So while what you say is all technically true, I'm not sure this is necessary.

I agree. However, Lukas's comment made me wonder about the future: if we eventually need to preserve non-PCI devices (like a TPM), should we be designing a common identification mechanism for all buses now? Or should we settle on BDF for PCI and invent stable identifiers for other bus types as they become necessary?

Quoted 4 lines

Well, at least PCI subsystem should use BDF..

You are probably right that the matching of preserved data to a struct device should be more general though.

Lukas' suggestion would also make it more reliable to detect bus numbers changing during a Live Update. We can play whack-a-mole with things like assign-busses, but there will be a risk that we miss something or something changes in the future.

Perhaps it would make sense to rely on BDF in the PCI subsystem in the short term and enforce bus number stability manually (e.g. see patch at the bottom), and then explore stable device paths as a future improvement to make PCI device preservation more reliable and also to enable other bus types?

To handle pci=assign-busses, perhaps something like this? Are there any other places where the kernel could change busses?

drivers/pci/probe.c +20 -3
index 0ce98e18b5a8..2e1e1aa385a8 100644
@@ -1331,6 +1331,20 @@ static bool pci_ea_fixed_busnrs(struct pci_dev *dev, u8 *sec, u8 *sub)
 	return true;
 }
 
+static bool pci_assign_all_busses(void)
+{
+	/*
+	 * During a Live Update, do not assign new bus numbers. Use bus numbers
+	 * assigned by the firmware and the previous kernel. Bus numbers must
+	 * remain constant so that devices preserved across the Live Update can
+	 * use the IOMMU uninterrupted.
+	 */
+	if (liveupdate_count())
+		return false;
+
+	return pcibios_assign_all_busses();
+}
+
 /*
  * pci_scan_bridge_extend() - Scan buses behind a bridge
  * @bus: Parent bus the bridge is on
@@ -1404,7 +1418,7 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
 	pci_write_config_word(dev, PCI_BRIDGE_CONTROL,
 			      bctl & ~PCI_BRIDGE_CTL_MASTER_ABORT);
 
-	if ((secondary || subordinate) && !pcibios_assign_all_busses() &&
+	if ((secondary || subordinate) && !pci_assign_all_busses() &&
 	    !is_cardbus && !broken) {
 		unsigned int cmax, buses;
 
@@ -1441,13 +1455,16 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
 		if (subordinate > max)
 			max = subordinate;
 	} else {
+		pci_WARN_ONCE(dev, liveupdate_count(),
+			      "Assigning new bus numbers during a Live Update! [%u %u %u %u]\n",
+			      secondary, subordinate, is_cardbus, broken);
 
 		/*
 		 * We need to assign a number to this bus which we always
 		 * do in the second pass.
 		 */
 		if (!pass) {
-			if (pcibios_assign_all_busses() || broken || is_cardbus)
+			if (pci_assign_all_busses() || broken || is_cardbus)
 
 				/*
 				 * Temporarily disable forwarding of the
@@ -1522,7 +1539,7 @@ static int pci_scan_bridge_extend(struct pci_bus *bus, struct pci_dev *dev,
 							max+i+1))
 					break;
 				while (parent->parent) {
-					if ((!pcibios_assign_all_busses()) &&
+					if ((!pci_assign_all_busses()) &&
 					    (parent->busn_res.end > max) &&
 					    (parent->busn_res.end <= max+i)) {
 						j = 1;
include/linux/liveupdate.h +6 -0
index b913d63eab5f..87a4982d0eb1 100644
@@ -219,6 +219,7 @@ struct liveupdate_flb {
 
 /* Return true if live update orchestrator is enabled */
 bool liveupdate_enabled(void);
+int liveupdate_count(void);
 
 /* Called during kexec to tell LUO that entered into reboot */
 int liveupdate_reboot(void);
@@ -241,6 +242,11 @@ static inline bool liveupdate_enabled(void)
 	return false;
 }
 
+static inline int liveupdate_count(void)
+{
+	return 0;
+}
+
 static inline int liveupdate_reboot(void)
 {
 	return 0;
kernel/liveupdate/luo_core.c +7 -0
index 69298d82f404..2f273397bd41 100644
@@ -256,6 +256,13 @@ bool liveupdate_enabled(void)
 {
 	return luo_global.enabled;
 }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
+
+int liveupdate_count(void)
+{
+	return luo_global.liveupdate_num;
+}
+EXPORT_SYMBOL_GPL(liveupdate_count);
 
 /**
  * DOC: LUO ioctl Interface
#9 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 2 lines · David Matlack

To handle pci=assign-busses, perhaps something like this? Are there any other places where the kernel could change busses?

In theory the algorithm to assign bus numbers could change from one kernel version to the next. Ilpo (+cc) is currently reworking the resource allocation algorithm. That work primarily covers MMIO window sizing, but bus numbers are resources as well and could be affected by changes. Resource allocation code is already quite convoluted and sprinkling liveupdate special cases all over it may not be received with enthusiasm. ;)

Of course in practice, changes to the algorithm do not happen often and the kernel will preserve bus numbers as set by BIOS. Only if it detects incorrect bus assignments or if forced via the command line will the kernel re-assign bus numbers.

But you do gain a bit of reliability if you don't assume bus numbers to stay the same and instead use the "path from root" approach to identify devices.

Thanks,

Lukas

#10 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 3 lines · Lukas Wunner

But you do gain a bit of reliability if you don't assume bus numbers to stay the same and instead use the "path from root" approach to identify devices.

Again, that's not reliability it is subtle bugs. The device is active during KHO, you CAN NOT do any resource reassignment, not bus numbers, not mmio. It must be fully disabled.

Jason

#11 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Hi Lukas,

Sorry I am late to the party.

Quoted 10 lines · Jason Gunthorpe
Quoted 3 lines · Lukas Wunner

But you do gain a bit of reliability if you don't assume bus numbers to stay the same and instead use the "path from root" approach to identify devices.

Again, that's not reliability it is subtle bugs. The device is active during KHO, you CAN NOT do any resource reassignment, not bus numbers, not mmio. It must be fully disabled.

I agree with Jason. The bus number is used in the low level hardware to do the DMA transfer. The bus number can not change for a device during livedupate with pending DMA transfer. The BDF MUST remain the same as the liveupdate with DMA transfer requirement. Given the BDF remains the same. Using the path from root doesn't buy you more protections. It just makes the patch more complicated but achieves the same thing. That is why I chose the BDF approach for the PCI liveupdate subsystem in the first place. To keep it simple.

Jason, please correct me if I am wrong. My understanding is that not only the device that is actively doing the DMA requires the bus number to stay the same, I think all the parent bridge, all the way to the root PCI host bridge, bus number must remain the same. After all, the DMA will need to route through the parent bridges.

Another point is that, on the same machine it can have multiple PCI host bridges. Each PCI host bridge bus number is acquired from the ACPI table walk. I am not aware of any way to get the slot number of the PCI host bridge. Lukas, do you know how to get the PCI host bridge slot number to form a path?

Chris

#12 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 5 lines · Chris Li

Jason, please correct me if I am wrong. My understanding is that not only the device that is actively doing the DMA requires the bus number to stay the same, I think all the parent bridge, all the way to the root PCI host bridge, bus number must remain the same. After all, the DMA will need to route through the parent bridges.

The completions need to route back through the parent bridges, so yes you cannot do anything to disturb RID based routing in the active fabric either, with also means few changes to the subordinate bus range of any bridge are possible.

Jason

#13 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 9 lines · Jason Gunthorpe
Quoted 5 lines · Chris Li

Jason, please correct me if I am wrong. My understanding is that not only the device that is actively doing the DMA requires the bus number to stay the same, I think all the parent bridge, all the way to the root PCI host bridge, bus number must remain the same. After all, the DMA will need to route through the parent bridges.

Quoted 4 lines

The completions need to route back through the parent bridges, so yes you cannot do anything to disturb RID based routing in the active fabric either, with also means few changes to the subordinate bus range of any bridge are possible.

Thank you Jason for the confirmation.

Lukas, that means if we are using the path, we will need to save the bus number along each path node. Different liveupdate devices might share the parent bridges, we might want to de-duplicate that. Then you end up with something very similar to the BDF design, where the path part is just redundant if you have BDF.

That is what I mean previously, using the BDF has the same protections as path design, just simpler.

Chris

#14 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

In reply to Chris Li
Quoted 9 lines · Chris Li
Quoted 3 lines · Jason Gunthorpe

The device is active
during KHO, you CAN NOT do any resource reassignment, not bus numbers, not mmio. It must be fully disabled.

I agree with Jason. The bus number is used in the low level hardware to do the DMA transfer. The bus number can not change for a device during livedupate with pending DMA transfer. The BDF MUST remain the same as the liveupdate with DMA transfer requirement.

Thank you both for the explanation.

Quoted 5 lines

Another point is that, on the same machine it can have multiple PCI host bridges. Each PCI host bridge bus number is acquired from the ACPI table walk. I am not aware of any way to get the slot number of the PCI host bridge. Lukas, do you know how to get the PCI host bridge slot number to form a path?

Host bridges are identified by the segment number. On ACPI-based systems, it's retrieved by acpi_pci_root_add() through invocation of the _SEG method.

Thanks,

Lukas

maintainer
#15 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

In reply to Jason Gunthorpe
Quoted 10 lines
Quoted 8 lines
Quoted 2 lines

So while what you say is all technically true, I'm not sure this is necessary.

I agree. However, Lukas's comment made me wonder about the future: if we eventually need to preserve non-PCI devices (like a TPM), should we be designing a common identification mechanism for all buses now? Or should we settle on BDF for PCI and invent stable identifiers for other bus types as they become necessary?

Well, at least PCI subsystem should use BDF..

BDF must be stable for PCI live update to work correctly.

Quoted 2 lines

You are probably right that the matching of preserved data to a struct device should be more general though.

Right, we need a mechanism to ensure early in boot that any preserved device does not auto-bind to a driver later in boot.

Using the UEFI Device Path format seems like a good way not to re-invent something that already exists. For example, while a preserved PCI device looks like this:
Acpi(PNP0A03,0)/Pci(1E|0)/Pci(0|0) (Luka's example)

We can seamlessly support other device types later using their native paths without changing the identification schema:

TPM: Acpi(PNP0C31,0)
IPMI/BMC: Acpi(PNP0A03,0)/Pci(1F|0)/BMC(1,0xCA2)
NVMe (PCI-attached): Acpi(PNP0A03,0)/Pci(1C|0)/Pci(0|0)/NVMe(1,00-00-...) etc...

Pasha

maintainer
#16 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

In reply to David Matlack

drivers/pci/liveupdate.c line 68

Quoted 8 lines
drivers/pci/liveupdate.c new file +8 -0
+static void pci_flb_unpreserve(struct liveupdate_flb_op_args *args)
+{
+       struct pci_ser *ser = args->obj;
+       struct folio *folio = virt_to_folio(ser);
+
+       WARN_ON_ONCE(ser->nr_devices);
+       kho_unpreserve_folio(folio);
+       folio_put(folio);

Here, and in other places in this series, I would use:
https://lore.kernel.org/all/[email protected]

kho_alloc_preserve(size_t size)
kho_unpreserve_free(void *mem)
kho_restore_free(void *mem)

Pasha

author maintainer
#17 lore raw

Re: [PATCH 02/21] PCI: Add API to track PCI devices preserved across Live Update

Quoted 13 lines · Pasha Tatashin

drivers/pci/liveupdate.c line 68

Quoted 9 lines
drivers/pci/liveupdate.c new file +8 -0
+static void pci_flb_unpreserve(struct liveupdate_flb_op_args *args)
+{
+       struct pci_ser *ser = args->obj;
+       struct folio *folio = virt_to_folio(ser);
+
+       WARN_ON_ONCE(ser->nr_devices);
+       kho_unpreserve_folio(folio);
+       folio_put(folio);

Here, and in other places in this series, I would use:
https://lore.kernel.org/all/[email protected]

Quoted 3 lines

kho_alloc_preserve(size_t size)
kho_unpreserve_free(void *mem)
kho_restore_free(void *mem)

Will do, thanks for the suggestion.

author maintainer
#18 lore raw

PATCH 3/21 PCI: Require driver_override for incoming Live Update preserved devices

In reply to David Matlack

Require driver_override to be set to bind an incoming Live Update preserved device to a driver. Auto-probing could lead to the device being bound to a different driver than what was bound to the device prior to Live Update.

Delegate binding preserved devices to the right driver to userspace by requiring driver_override to be set on the device.

This restriction is relaxed once a driver calls
pci_liveupdate_incoming_finish().

---
 drivers/pci/pci-driver.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)
drivers/pci/pci-driver.c +10 -2
index 302d61783f6c..294ce92331a8 100644
@@ -420,18 +420,26 @@ static int __pci_device_probe(struct pci_driver *drv, struct pci_dev *pci_dev)
 }
 
 #ifdef CONFIG_PCI_IOV
-static inline bool pci_device_can_probe(struct pci_dev *pdev)
+static inline bool pci_iov_device_can_probe(struct pci_dev *pdev)
 {
 	return (!pdev->is_virtfn || pdev->physfn->sriov->drivers_autoprobe ||
 		pdev->driver_override);
 }
 #else
-static inline bool pci_device_can_probe(struct pci_dev *pdev)
+static inline bool pci_iov_device_can_probe(struct pci_dev *pdev)
 {
 	return true;
 }
 #endif
 
+static inline bool pci_device_can_probe(struct pci_dev *pdev)
+{
+	if (pci_liveupdate_incoming_is_preserved(pdev))
+		return pdev->driver_override;
+
+	return pci_iov_device_can_probe(pdev);
+}
+
 static int pci_device_probe(struct device *dev)
 {
 	int error;
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#19 lore raw

Re: [PATCH 03/21] PCI: Require driver_override for incoming Live Update preserved devices

Quoted 5 lines · David Matlack

Require driver_override to be set to bind an incoming Live Update preserved device to a driver. Auto-probing could lead to the device being bound to a different driver than what was bound to the device prior to Live Update.

Quoted 7 lines

Delegate binding preserved devices to the right driver to userspace by requiring driver_override to be set on the device.

This restriction is relaxed once a driver calls
pci_liveupdate_incoming_finish().

After some offline discussion with Jason and Pasha, I will drop this patch from the next version to avoid hard-coding a policy in the kernel.

Instead, for the time being, it will be the user's responsibility to only preserve devices that do not have any built-in drivers that will bind to the device during boot following a Live Update.

#20 lore raw

Re: [PATCH 03/21] PCI: Require driver_override for incoming Live Update preserved devices

Quoted 13 lines · David Matlack
Quoted 11 lines · David Matlack

Require driver_override to be set to bind an incoming Live Update preserved device to a driver. Auto-probing could lead to the device being bound to a different driver than what was bound to the device prior to Live Update.

Delegate binding preserved devices to the right driver to userspace by requiring driver_override to be set on the device.

This restriction is relaxed once a driver calls
pci_liveupdate_incoming_finish().

Quoted 8 lines
Quoted 1 line

After some offline discussion with Jason and Pasha, I will drop this patch from the next version to avoid hard-coding a policy in the kernel.

Instead, for the time being, it will be the user's responsibility to only preserve devices that do not have any built-in drivers that will bind to the device during boot following a Live Update.

Ack, that certainly can simplify the PCI kernel side of the code, the kernel doesn't need to deal with PCI driver binding.

Chris

author maintainer
#21 lore raw

PATCH 4/21 vfio/pci: Register a file handler with Live Update Orchestrator

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Register a live update file handler for vfio-pci device files. Add stub implementations of all required callbacks so that registration does not fail (i.e. to avoid breaking git-bisect).

This file handler will be extended in subsequent commits to enable a device bound to vfio-pci to run without interruption while the host is going through a kexec Live Update.

---
 MAINTAINERS                            |  1 +
 drivers/vfio/pci/Makefile              |  1 +
 drivers/vfio/pci/vfio_pci.c            |  9 +++-
 drivers/vfio/pci/vfio_pci_liveupdate.c | 69 ++++++++++++++++++++++++++
 drivers/vfio/pci/vfio_pci_priv.h       | 14 ++++++
 include/linux/kho/abi/vfio_pci.h       | 28 +++++++++++
 6 files changed, 121 insertions(+), 1 deletion(-)
 create mode 100644 drivers/vfio/pci/vfio_pci_liveupdate.c
 create mode 100644 include/linux/kho/abi/vfio_pci.h
MAINTAINERS +1 -0
index 2722f98d0ed7..ff50977277c4 100644
@@ -26933,6 +26933,7 @@ F:	Documentation/ABI/testing/debugfs-vfio
 F:	Documentation/ABI/testing/sysfs-devices-vfio-dev
 F:	Documentation/driver-api/vfio.rst
 F:	drivers/vfio/
+F:	include/linux/kho/abi/vfio_pci.h
 F:	include/linux/vfio.h
 F:	include/linux/vfio_pci_core.h
 F:	include/uapi/linux/vfio.h
drivers/vfio/pci/Makefile +1 -0
index cf00c0a7e55c..929df22c079b 100644
@@ -2,6 +2,7 @@
 
 vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
+vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
 vfio-pci-y := vfio_pci.o
drivers/vfio/pci/vfio_pci.c +8 -1
index ac10f14417f2..c2fe34a830d8 100644
@@ -252,6 +252,10 @@ static int __init vfio_pci_init(void)
 	int ret;
 	bool is_disable_vga = true;
 
+	ret = vfio_pci_liveupdate_init();
+	if (ret)
+		return ret;
+
 #ifdef CONFIG_VFIO_PCI_VGA
 	is_disable_vga = disable_vga;
 #endif
@@ -260,8 +264,10 @@ static int __init vfio_pci_init(void)
 
 	/* Register and scan for devices */
 	ret = pci_register_driver(&vfio_pci_driver);
-	if (ret)
+	if (ret) {
+		vfio_pci_liveupdate_cleanup();
 		return ret;
+	}
 
 	vfio_pci_fill_ids();
 
@@ -275,6 +281,7 @@ module_init(vfio_pci_init);
 static void __exit vfio_pci_cleanup(void)
 {
 	pci_unregister_driver(&vfio_pci_driver);
+	vfio_pci_liveupdate_cleanup();
 }
 module_exit(vfio_pci_cleanup);
 
drivers/vfio/pci/vfio_pci_liveupdate.c new file +69 -0

69 lines not on this page · raw message

drivers/vfio/pci/vfio_pci_priv.h +14 -0
index a9972eacb293..b9f7c4e2b4df 100644
@@ -107,4 +107,18 @@ static inline bool vfio_pci_is_vga(struct pci_dev *pdev)
 	return (pdev->class >> 8) == PCI_CLASS_DISPLAY_VGA;
 }
 
+#ifdef CONFIG_LIVEUPDATE
+int __init vfio_pci_liveupdate_init(void);
+void vfio_pci_liveupdate_cleanup(void);
+#else
+static inline int vfio_pci_liveupdate_init(void)
+{
+	return 0;
+}
+
+static inline void vfio_pci_liveupdate_cleanup(void)
+{
+}
+#endif /* CONFIG_LIVEUPDATE */
+
 #endif
include/linux/kho/abi/vfio_pci.h new file +28 -0
new file mode 100644
index 000000000000..37a845eed972
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+/*
+ * Copyright (c) 2025, Google LLC.
+ * Vipin Sharma <[email protected]>
+ * David Matlack <[email protected]>
+ */
+
+#ifndef _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H
+#define _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H
+
+/**
+ * DOC: VFIO PCI Live Update ABI
+ *
+ * This header defines the ABI for preserving the state of a VFIO PCI device
+ * files across a kexec reboot using LUO.
+ *
+ * Device metadata is serialized into memory which is then handed to the next
+ * kernel via KHO.
+ *
+ * This interface is a contract. Any modification to any of the serialization
+ * structs defined here constitutes a breaking change. Such changes require
+ * incrementing the version number in the VFIO_PCI_LUO_FH_COMPATIBLE string.
+ */
+
+#define VFIO_PCI_LUO_FH_COMPATIBLE "vfio-pci-v1"
+
+#endif /* _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H */
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#22 lore raw

PATCH 5/21 vfio/pci: Preserve vfio-pci device files across Live Update

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Implement the live update file handler callbacks to preserve a vfio-pci device across a Live Update. Subsequent commits will enable userspace to then retrieve this file after the Live Update.

Live Update support is scoped only to cdev files (i.e. not VFIO_GROUP_GET_DEVICE_FD files).

State about each device is serialized into a new ABI struct vfio_pci_core_device_ser. The contents of this struct are preserved across the Live Update to the next kernel using a combination of Kexec-Handover (KHO) to preserve the page(s) holding the struct and the Live Update Orchestrator (LUO) to preserve the physical address of the struct.

For now the only contents of struct vfio_pci_core_device_ser the device's PCI segment number and BDF, so that the device can be uniquely identified after the Live Update.

Require that userspace disables interrupts on the device prior to freeze() so that the device does not send any interrupts until new interrupt handlers have been set up by the next kernel.

Reset the device and restore its state in the freeze() callback. This ensures the device can be received by the next kernel in a consistent state. Eventually this will be dropped and the device can be preserved across in a running state, but that requires further work in VFIO and the core PCI layer.

Note that LUO holds a reference to this file when it is preserved. So VFIO is guaranteed that vfio_df_device_last_close() will not be called on this device no matter what userspace does.

---
 drivers/vfio/pci/vfio_pci.c            |   2 +-
 drivers/vfio/pci/vfio_pci_liveupdate.c | 100 ++++++++++++++++++++++++-
 drivers/vfio/pci/vfio_pci_priv.h       |   2 +
 drivers/vfio/vfio.h                    |  13 ----
 drivers/vfio/vfio_main.c               |   9 ---
 include/linux/kho/abi/vfio_pci.h       |  15 ++++
 include/linux/vfio.h                   |  28 +++++++
 7 files changed, 144 insertions(+), 25 deletions(-)
drivers/vfio/pci/vfio_pci.c +1 -1
index c2fe34a830d8..281c69c086d3 100644
@@ -125,7 +125,7 @@ static int vfio_pci_open_device(struct vfio_device *core_vdev)
 	return 0;
 }
 
-static const struct vfio_device_ops vfio_pci_ops = {
+const struct vfio_device_ops vfio_pci_ops = {
 	.name		= "vfio-pci",
 	.init		= vfio_pci_core_init_dev,
 	.release	= vfio_pci_core_release_dev,
drivers/vfio/pci/vfio_pci_liveupdate.c +98 -2

129 lines not on this page · raw message

drivers/vfio/pci/vfio_pci_priv.h +2 -0
index b9f7c4e2b4df..7f189e5e6c0a 100644
@@ -11,6 +11,8 @@
 /* Cap maximum number of ioeventfds per device (arbitrary) */
 #define VFIO_PCI_IOEVENTFD_MAX		1000
 
+extern const struct vfio_device_ops vfio_pci_ops;
+
 struct vfio_pci_ioeventfd {
 	struct list_head	next;
 	struct vfio_pci_core_device	*vdev;
drivers/vfio/vfio.h +0 -13
index 50128da18bca..6b89edbbf174 100644
@@ -16,17 +16,6 @@ struct iommufd_ctx;
 struct iommu_group;
 struct vfio_container;
 
-struct vfio_device_file {
-	struct vfio_device *device;
-	struct vfio_group *group;
-
-	u8 access_granted;
-	u32 devid; /* only valid when iommufd is valid */
-	spinlock_t kvm_ref_lock; /* protect kvm field */
-	struct kvm *kvm;
-	struct iommufd_ctx *iommufd; /* protected by struct vfio_device_set::lock */
-};
-
 void vfio_device_put_registration(struct vfio_device *device);
 bool vfio_device_try_get_registration(struct vfio_device *device);
 int vfio_df_open(struct vfio_device_file *df);
@@ -34,8 +23,6 @@ void vfio_df_close(struct vfio_device_file *df);
 struct vfio_device_file *
 vfio_allocate_device_file(struct vfio_device *device);
 
-extern const struct file_operations vfio_device_fops;
-
 #ifdef CONFIG_VFIO_NOIOMMU
 extern bool vfio_noiommu __read_mostly;
 #else
drivers/vfio/vfio_main.c +0 -9
index 38c8e9350a60..9182dc46d73f 100644
@@ -1386,15 +1386,6 @@ const struct file_operations vfio_device_fops = {
 #endif
 };
 
-static struct vfio_device *vfio_device_from_file(struct file *file)
-{
-	struct vfio_device_file *df = file->private_data;
-
-	if (file->f_op != &vfio_device_fops)
-		return NULL;
-	return df->device;
-}
-
 /**
  * vfio_file_is_valid - True if the file is valid vfio file
  * @file: VFIO group file or VFIO device file
include/linux/kho/abi/vfio_pci.h +15 -0
index 37a845eed972..9bf58a2f3820 100644
@@ -9,6 +9,9 @@
 #ifndef _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H
 #define _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H
 
+#include <linux/compiler.h>
+#include <linux/types.h>
+
 /**
  * DOC: VFIO PCI Live Update ABI
  *
@@ -25,4 +28,16 @@
 
 #define VFIO_PCI_LUO_FH_COMPATIBLE "vfio-pci-v1"
 
+/**
+ * struct vfio_pci_core_device_ser - Serialized state of a single VFIO PCI
+ * device.
+ *
+ * @bdf: The device's PCI bus, device, and function number.
+ * @domain: The device's PCI domain number (segment).
+ */
+struct vfio_pci_core_device_ser {
+	u16 bdf;
+	u16 domain;
+} __packed;
+
 #endif /* _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H */
include/linux/vfio.h +28 -0
index eb563f538dee..f09da3bdf786 100644
@@ -80,6 +80,34 @@ struct vfio_device {
 #endif
 };
 
+struct vfio_device_file {
+	struct vfio_device *device;
+	struct vfio_group *group;
+
+	u8 access_granted;
+	u32 devid; /* only valid when iommufd is valid */
+	spinlock_t kvm_ref_lock; /* protect kvm field */
+	struct kvm *kvm;
+	struct iommufd_ctx *iommufd; /* protected by struct vfio_device_set::lock */
+};
+
+extern const struct file_operations vfio_device_fops;
+
+static inline struct vfio_device_file *to_vfio_device_file(struct file *file)
+{
+	if (file->f_op != &vfio_device_fops)
+		return NULL;
+
+	return file->private_data;
+}
+
+static inline struct vfio_device *vfio_device_from_file(struct file *file)
+{
+	struct vfio_device_file *df = to_vfio_device_file(file);
+
+	return df ? df->device : NULL;
+}
+
 /**
  * struct vfio_device_ops - VFIO bus driver device callbacks
  *
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#23 lore raw

PATCH 6/21 vfio/pci: Retrieve preserved device files after Live Update

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Enable userspace to retrieve preserved VFIO device files from VFIO after a Live Update by implementing the retrieve() and finish() file handler callbacks.

Use an anonymous inode when creating the file, since the retrieved device file is not opened through any particular cdev inode, and the cdev inode does not matter in practice.

For now the retrieved file is functionally equivalent a opening the corresponding VFIO cdev file. Subsequent commits will leverage the preserved state associated with the retrieved file to preserve bits of the device across Live Update.

---
 drivers/vfio/device_cdev.c             | 21 +++++---
 drivers/vfio/pci/vfio_pci_liveupdate.c | 75 +++++++++++++++++++++++++-
 drivers/vfio/vfio_main.c               | 13 +++++
 include/linux/vfio.h                   | 12 +++++
 4 files changed, 113 insertions(+), 8 deletions(-)
drivers/vfio/device_cdev.c +14 -7
index 480cac3a0c27..0a6e972f322b 100644
@@ -16,14 +16,8 @@ void vfio_init_device_cdev(struct vfio_device *device)
 	device->cdev.owner = THIS_MODULE;
 }
 
-/*
- * device access via the fd opened by this function is blocked until
- * .open_device() is called successfully during BIND_IOMMUFD.
- */
-int vfio_device_fops_cdev_open(struct inode *inode, struct file *filep)
+int __vfio_device_fops_cdev_open(struct vfio_device *device, struct file *filep)
 {
-	struct vfio_device *device = container_of(inode->i_cdev,
-						  struct vfio_device, cdev);
 	struct vfio_device_file *df;
 	int ret;
 
@@ -52,6 +46,19 @@ int vfio_device_fops_cdev_open(struct inode *inode, struct file *filep)
 	vfio_device_put_registration(device);
 	return ret;
 }
+EXPORT_SYMBOL_GPL(__vfio_device_fops_cdev_open);
+
+/*
+ * device access via the fd opened by this function is blocked until
+ * .open_device() is called successfully during BIND_IOMMUFD.
+ */
+int vfio_device_fops_cdev_open(struct inode *inode, struct file *filep)
+{
+	struct vfio_device *device = container_of(inode->i_cdev,
+						  struct vfio_device, cdev);
+
+	return __vfio_device_fops_cdev_open(device, filep);
+}
 
 static void vfio_df_get_kvm_safe(struct vfio_device_file *df)
 {
drivers/vfio/pci/vfio_pci_liveupdate.c +74 -1

99 lines not on this page · raw message

drivers/vfio/vfio_main.c +13 -0
index 9182dc46d73f..c5b5eb509474 100644
@@ -13,6 +13,7 @@
 #include <linux/cdev.h>
 #include <linux/compat.h>
 #include <linux/device.h>
+#include <linux/device/class.h>
 #include <linux/fs.h>
 #include <linux/idr.h>
 #include <linux/iommu.h>
@@ -1706,6 +1707,18 @@ int vfio_dma_rw(struct vfio_device *device, dma_addr_t iova, void *data,
 }
 EXPORT_SYMBOL(vfio_dma_rw);
 
+struct vfio_device *vfio_find_device(const void *data, device_match_t match)
+{
+	struct device *device;
+
+	device = class_find_device(vfio.device_class, NULL, data, match);
+	if (!device)
+		return NULL;
+
+	return container_of(device, struct vfio_device, device);
+}
+EXPORT_SYMBOL_GPL(vfio_find_device);
+
 /*
  * Module/class support
  */
include/linux/vfio.h +12 -0
index f09da3bdf786..4e400a7219ea 100644
@@ -413,4 +413,16 @@ int vfio_virqfd_enable(void *opaque, int (*handler)(void *, void *),
 void vfio_virqfd_disable(struct virqfd **pvirqfd);
 void vfio_virqfd_flush_thread(struct virqfd **pvirqfd);
 
+#if IS_ENABLED(CONFIG_VFIO_DEVICE_CDEV)
+int __vfio_device_fops_cdev_open(struct vfio_device *device, struct file *filep);
+#else
+static inline int __vfio_device_fops_cdev_open(struct vfio_device *device,
+					       struct file *filep)
+{
+	return -EOPNOTSUPP;
+}
+#endif /* IS_ENABLED(CONFIG_VFIO_DEVICE_CDEV) */
+
+struct vfio_device *vfio_find_device(const void *data, device_match_t match);
+
 #endif /* VFIO_H */
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
#24 lore raw

Re: [PATCH 06/21] vfio/pci: Retrieve preserved device files after Live Update

drivers/vfio/pci/vfio_pci_liveupdate.c static int vfio_pci_liveupdate_freeze(struct liveupdate_file_op_args *args) · line 152

Quoted 13 lines · David Matlack

From: Vipin Sharma <[email protected]>
static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args) {

drivers/vfio/pci/vfio_pci_liveupdate.c +9 -1
-	return -EOPNOTSUPP;
+	struct vfio_pci_core_device_ser *ser;
+	struct vfio_device *device;
+	struct folio *folio;
+	struct file *file;
+	int ret;
+
+	folio = kho_restore_folio(args->serialized_data);
+	if (!folio)
+		return -ENOENT;

Should this be consistent with the behavior of pci_flb_retrieve() which panics on failure? The short circuit failure paths which follow leak the folio, which seems like a hygiene issue, but the practical significance is moot if vfio_pci_liveupdate_retrieve() failure is catastrophic anyways?

Quoted 41 lines
drivers/vfio/pci/vfio_pci_liveupdate.c +41 -0
+
+	ser = folio_address(folio);
+
+	device = vfio_find_device(ser, match_device);
+	if (!device)
+		return -ENODEV;
+
+	/*
+	 * During a Live Update userspace retrieves preserved VFIO cdev files by
+	 * issuing an ioctl on /dev/liveupdate rather than by opening VFIO
+	 * character devices.
+	 *
+	 * To handle that scenario, this routine simulates opening the VFIO
+	 * character device for userspace with an anonymous inode. The returned
+	 * file has the same properties as a cdev file (e.g. operations are
+	 * blocked until BIND_IOMMUFD is called), aside from the inode
+	 * association.
+	 */
+	file = anon_inode_getfile_fmode("[vfio-device-liveupdate]",
+					&vfio_device_fops, NULL,
+					O_RDWR, FMODE_PREAD | FMODE_PWRITE);
+
+	if (IS_ERR(file)) {
+		ret = PTR_ERR(file);
+		goto out;
+	}
+
+	ret = __vfio_device_fops_cdev_open(device, file);
+	if (ret) {
+		fput(file);
+		goto out;
+	}
+
+	args->file = file;
+
+out:
+	/* Drop the reference from vfio_find_device() */
+	put_device(&device->device);
+
+	return ret;
+}
maintainer
#25 lore raw

Re: [PATCH 06/21] vfio/pci: Retrieve preserved device files after Live Update

Quoted 16 lines · Alex Mastro

drivers/vfio/pci/vfio_pci_liveupdate.c static int vfio_pci_liveupdate_freeze(struct liveupdate_file_op_args *args) · line 152

Quoted 13 lines · David Matlack

From: Vipin Sharma <[email protected]>
static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args) {

drivers/vfio/pci/vfio_pci_liveupdate.c +9 -1
-     return -EOPNOTSUPP;
+     struct vfio_pci_core_device_ser *ser;
+     struct vfio_device *device;
+     struct folio *folio;
+     struct file *file;
+     int ret;
+
+     folio = kho_restore_folio(args->serialized_data);
+     if (!folio)
+             return -ENOENT;
Quoted 4 lines

Should this be consistent with the behavior of pci_flb_retrieve() which panics on failure? The short circuit failure paths which follow leak the folio, which seems like a hygiene issue, but the practical significance is moot if vfio_pci_liveupdate_retrieve() failure is catastrophic anyways?

pci_flb_retrieve() is used during boot. If it fails, we risk DMA corrupting any memory region, so a panic makes sense. In contrast, this retrieval happens once we are already in userspace, allowing the user to decide how to handle the failure to recover the preserved cdev.

Pasha

Quoted 42 lines
Quoted 42 lines
drivers/vfio/pci/vfio_pci_liveupdate.c +41 -0
+
+     ser = folio_address(folio);
+
+     device = vfio_find_device(ser, match_device);
+     if (!device)
+             return -ENODEV;
+
+     /*
+      * During a Live Update userspace retrieves preserved VFIO cdev files by
+      * issuing an ioctl on /dev/liveupdate rather than by opening VFIO
+      * character devices.
+      *
+      * To handle that scenario, this routine simulates opening the VFIO
+      * character device for userspace with an anonymous inode. The returned
+      * file has the same properties as a cdev file (e.g. operations are
+      * blocked until BIND_IOMMUFD is called), aside from the inode
+      * association.
+      */
+     file = anon_inode_getfile_fmode("[vfio-device-liveupdate]",
+                                     &vfio_device_fops, NULL,
+                                     O_RDWR, FMODE_PREAD | FMODE_PWRITE);
+
+     if (IS_ERR(file)) {
+             ret = PTR_ERR(file);
+             goto out;
+     }
+
+     ret = __vfio_device_fops_cdev_open(device, file);
+     if (ret) {
+             fput(file);
+             goto out;
+     }
+
+     args->file = file;
+
+out:
+     /* Drop the reference from vfio_find_device() */
+     put_device(&device->device);
+
+     return ret;
+}
author maintainer
#26 lore raw

Re: [PATCH 06/21] vfio/pci: Retrieve preserved device files after Live Update

Quoted 18 lines · Pasha Tatashin
Quoted 16 lines · Alex Mastro
Quoted 13 lines · David Matlack

From: Vipin Sharma <[email protected]>
static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args) {

drivers/vfio/pci/vfio_pci_liveupdate.c +9 -1
-     return -EOPNOTSUPP;
+     struct vfio_pci_core_device_ser *ser;
+     struct vfio_device *device;
+     struct folio *folio;
+     struct file *file;
+     int ret;
+
+     folio = kho_restore_folio(args->serialized_data);
+     if (!folio)
+             return -ENOENT;
Quoted 2 lines
Quoted 2 lines

Should this be consistent with the behavior of pci_flb_retrieve() which panics on failure? The short circuit failure paths which follow leak the folio,

Thanks for catching the leaked folio. I'll fix that in the next version.

Quoted 8 lines
Quoted 2 lines

which seems like a hygiene issue, but the practical significance is moot if vfio_pci_liveupdate_retrieve() failure is catastrophic anyways?

pci_flb_retrieve() is used during boot. If it fails, we risk DMA corrupting any memory region, so a panic makes sense. In contrast, this retrieval happens once we are already in userspace, allowing the user to decide how to handle the failure to recover the preserved cdev.

This is what I was thinking as well. vfio_pci_liveupdate_retrieve() runs in the context of the ioctl LIVEUPDATE_SESSION_RETRIEVE_FD, so we can just return an error up to userspace if anything goes wrong and let userspace initiate the reboot to recover the device if/when it's ready.

OTOH, pci_flb_retrieve() gets called by the kernel during early boot to determine what devices the previous kernel preserved. If the kernel can't determine which devices were preserved by the previous kernel and once the kernel starts preserving I/O page tables, that could lead to corruption, so panicking is warranted.

#27 lore raw

Re: [PATCH 06/21] vfio/pci: Retrieve preserved device files after Live Update

Quoted 39 lines · David Matlack
Quoted 18 lines · Pasha Tatashin
Quoted 16 lines · Alex Mastro
Quoted 13 lines · David Matlack

From: Vipin Sharma <[email protected]>
static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args) {

drivers/vfio/pci/vfio_pci_liveupdate.c +9 -1
-     return -EOPNOTSUPP;
+     struct vfio_pci_core_device_ser *ser;
+     struct vfio_device *device;
+     struct folio *folio;
+     struct file *file;
+     int ret;
+
+     folio = kho_restore_folio(args->serialized_data);
+     if (!folio)
+             return -ENOENT;
Quoted 2 lines
Quoted 2 lines

Should this be consistent with the behavior of pci_flb_retrieve() which panics on failure? The short circuit failure paths which follow leak the folio,

Thanks for catching the leaked folio. I'll fix that in the next version.

Quoted 8 lines
Quoted 2 lines

which seems like a hygiene issue, but the practical significance is moot if vfio_pci_liveupdate_retrieve() failure is catastrophic anyways?

pci_flb_retrieve() is used during boot. If it fails, we risk DMA corrupting any memory region, so a panic makes sense. In contrast, this retrieval happens once we are already in userspace, allowing the user to decide how to handle the failure to recover the preserved cdev.

This is what I was thinking as well. vfio_pci_liveupdate_retrieve() runs in the context of the ioctl LIVEUPDATE_SESSION_RETRIEVE_FD, so we can just return an error up to userspace if anything goes wrong and let userspace initiate the reboot to recover the device if/when it's ready.

Quoted 5 lines

OTOH, pci_flb_retrieve() gets called by the kernel during early boot to determine what devices the previous kernel preserved. If the kernel can't determine which devices were preserved by the previous kernel and once the kernel starts preserving I/O page tables, that could lead to corruption, so panicking is warranted.

Make sense, thanks for elaborating David and Pasha.

author maintainer
#28 lore raw

PATCH 7/21 vfio/pci: Notify PCI subsystem about devices preserved across Live Update

In reply to David Matlack

Notify the PCI subsystem about devices vfio-pci is preserving across Live Update by registering the vfio-pci liveupdate file handler with the PCI subsystem's FLB handler.

Notably this will ensure that devices preserved through vfio-pci do not go through auto-probing in the next kernel (i.e. preventing them from being bound to a different driver).

This also enables VFIO to detect that a device was preserved before userspace first retrieves the file from it, which will be used in subsequent commits.

---
 drivers/vfio/pci/vfio_pci_liveupdate.c | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)
drivers/vfio/pci/vfio_pci_liveupdate.c +22 -1
index b7451007fca4..7669c65bde17 100644
@@ -64,6 +64,7 @@ static int vfio_pci_liveupdate_preserve(struct liveupdate_file_op_args *args)
 	if (err)
 		goto error;
 
+	pci_liveupdate_outgoing_preserve(pdev);
 	args->serialized_data = virt_to_phys(ser);
 	return 0;
 
@@ -75,8 +76,10 @@ static int vfio_pci_liveupdate_preserve(struct liveupdate_file_op_args *args)
 static void vfio_pci_liveupdate_unpreserve(struct liveupdate_file_op_args *args)
 {
 	struct vfio_pci_core_device_ser *ser = phys_to_virt(args->serialized_data);
+	struct vfio_device *device = vfio_device_from_file(args->file);
 	struct folio *folio = virt_to_folio(ser);
 
+	pci_liveupdate_outgoing_unpreserve(to_pci_dev(device->dev));
 	kho_unpreserve_folio(folio);
 	folio_put(folio);
 }
@@ -199,8 +202,11 @@ static bool vfio_pci_liveupdate_can_finish(struct liveupdate_file_op_args *args)
 
 static void vfio_pci_liveupdate_finish(struct liveupdate_file_op_args *args)
 {
+	struct vfio_device *device = vfio_device_from_file(args->file);
 	struct folio *folio;
 
+	pci_liveupdate_incoming_finish(to_pci_dev(device->dev));
+
 	folio = virt_to_folio(phys_to_virt(args->serialized_data));
 	folio_put(folio);
 }
@@ -223,10 +229,24 @@ static struct liveupdate_file_handler vfio_pci_liveupdate_fh = {
 
 int __init vfio_pci_liveupdate_init(void)
 {
+	int ret;
+
 	if (!liveupdate_enabled())
 		return 0;
 
-	return liveupdate_register_file_handler(&vfio_pci_liveupdate_fh);
+	ret = liveupdate_register_file_handler(&vfio_pci_liveupdate_fh);
+	if (ret)
+		return ret;
+
+	ret = pci_liveupdate_register_fh(&vfio_pci_liveupdate_fh);
+	if (ret)
+		goto error;
+
+	return 0;
+
+error:
+	liveupdate_unregister_file_handler(&vfio_pci_liveupdate_fh);
+	return ret;
 }
 
 void vfio_pci_liveupdate_cleanup(void)
@@ -234,5 +254,6 @@ void vfio_pci_liveupdate_cleanup(void)
 	if (!liveupdate_enabled())
 		return;
 
+	WARN_ON_ONCE(pci_liveupdate_unregister_fh(&vfio_pci_liveupdate_fh));
 	liveupdate_unregister_file_handler(&vfio_pci_liveupdate_fh);
 }
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#29 lore raw

PATCH 8/21 vfio: Enforce preserved devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD

In reply to David Matlack

Enforce that files for incoming (preserved by previous kernel) VFIO devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD rather than by opening the corresponding VFIO character device or via VFIO_GROUP_GET_DEVICE_FD.

Both of these methods would result in VFIO initializing the device without access to the preserved state of the device passed by the previous kernel.

---
 drivers/vfio/device_cdev.c |  4 ++++
 drivers/vfio/group.c       |  9 +++++++++
 include/linux/vfio.h       | 11 +++++++++++
 3 files changed, 24 insertions(+)
drivers/vfio/device_cdev.c +4 -0
index 0a6e972f322b..f1e54dcc04e7 100644
@@ -57,6 +57,10 @@ int vfio_device_fops_cdev_open(struct inode *inode, struct file *filep)
 	struct vfio_device *device = container_of(inode->i_cdev,
 						  struct vfio_device, cdev);
 
+	/* Device file must be retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD */
+	if (vfio_liveupdate_incoming_is_preserved(device))
+		return -EBUSY;
+
 	return __vfio_device_fops_cdev_open(device, filep);
 }
 
drivers/vfio/group.c +9 -0
index c376a6279de0..93dbbf37ec4c 100644
@@ -313,6 +313,15 @@ static int vfio_group_ioctl_get_device_fd(struct vfio_group *group,
 	if (IS_ERR(device))
 		return PTR_ERR(device);
 
+	/*
+	 * This device was preserved across a Live Update. Accessing it via
+	 * VFIO_GROUP_GET_DEVICE_FD is not allowed.
+	 */
+	if (vfio_liveupdate_incoming_is_preserved(device)) {
+		ret = -EBUSY;
+		goto err_put_device;
+	}
+
 	fdno = get_unused_fd_flags(O_CLOEXEC);
 	if (fdno < 0) {
 		ret = fdno;
include/linux/vfio.h +11 -0
index 4e400a7219ea..040c2b2ee074 100644
@@ -16,6 +16,7 @@
 #include <linux/cdev.h>
 #include <uapi/linux/vfio.h>
 #include <linux/iova_bitmap.h>
+#include <linux/pci.h>
 
 struct kvm;
 struct iommufd_ctx;
@@ -425,4 +426,14 @@ static inline int __vfio_device_fops_cdev_open(struct vfio_device *device,
 
 struct vfio_device *vfio_find_device(const void *data, device_match_t match);
 
+static inline bool vfio_liveupdate_incoming_is_preserved(struct vfio_device *device)
+{
+	struct device *d = device->dev;
+
+	if (dev_is_pci(d))
+		return pci_liveupdate_incoming_is_preserved(to_pci_dev(d));
+
+	return false;
+}
+
 #endif /* VFIO_H */
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#30 lore raw

PATCH 9/21 vfio/pci: Store Live Update state in struct vfio_pci_core_device

In reply to David Matlack

Stash a pointer to a device's Live Updated state in struct vfio_pci_core_device. This will enable subsequent commits to use the preserved state when enabling the device.

To enable VFIO to safely access this pointer during device enablement, require that the device is fully enabled before returning true from can_finish().

---
 drivers/vfio/pci/vfio_pci_core.c       |  1 +
 drivers/vfio/pci/vfio_pci_liveupdate.c | 20 +++++++++++++++++++-
 include/linux/vfio_pci_core.h          |  6 ++++++
 3 files changed, 26 insertions(+), 1 deletion(-)
drivers/vfio/pci/vfio_pci_core.c +1 -0
index 7dcf5439dedc..b09fe0993e04 100644
@@ -536,6 +536,7 @@ int vfio_pci_core_enable(struct vfio_pci_core_device *vdev)
 	if (!vfio_vga_disabled() && vfio_pci_is_vga(pdev))
 		vdev->has_vga = true;
 
+	vdev->liveupdate_state = NULL;
 
 	return 0;
 
drivers/vfio/pci/vfio_pci_liveupdate.c +19 -1
index 7669c65bde17..0fb29bd3ae3b 100644
@@ -145,6 +145,7 @@ static int match_device(struct device *dev, const void *arg)
 static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args)
 {
 	struct vfio_pci_core_device_ser *ser;
+	struct vfio_pci_core_device *vdev;
 	struct vfio_device *device;
 	struct folio *folio;
 	struct file *file;
@@ -186,6 +187,9 @@ static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args)
 		goto out;
 	}
 
+	vdev = container_of(device, struct vfio_pci_core_device, vdev);
+	vdev->liveupdate_state = ser;
+
 	args->file = file;
 
 out:
@@ -197,7 +201,21 @@ static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args)
 
 static bool vfio_pci_liveupdate_can_finish(struct liveupdate_file_op_args *args)
 {
-	return args->retrieved;
+	struct vfio_pci_core_device *vdev;
+	struct vfio_device *device;
+
+	if (!args->retrieved)
+		return false;
+
+	device = vfio_device_from_file(args->file);
+	vdev = container_of(device, struct vfio_pci_core_device, vdev);
+
+	/*
+	 * Ensure VFIO is done using vdev->liveupdate_state, which means its
+	 * safe for vfio_pci_liveupdate_finish() to free it.
+	 */
+	guard(mutex)(&device->dev_set->lock);
+	return !vdev->liveupdate_state;
 }
 
 static void vfio_pci_liveupdate_finish(struct liveupdate_file_op_args *args)
include/linux/vfio_pci_core.h +6 -0
index f541044e42a2..56ff6452562d 100644
@@ -94,6 +94,12 @@ struct vfio_pci_core_device {
 	struct vfio_pci_core_device	*sriov_pf_core_dev;
 	struct notifier_block	nb;
 	struct rw_semaphore	memory_lock;
+
+	/*
+	 * State passed by the previous kernel during a Live Update. Only
+	 * safe to access when first opening the device.
+	 */
+	struct vfio_pci_core_device_ser *liveupdate_state;
 };
 
 /* Will be exported for vfio pci drivers usage */
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#31 lore raw

PATCH 10/21 vfio/pci: Skip reset of preserved device after Live Update

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Do not reset the device when a Live Update preserved vfio-pci device is retrieved and first enabled. vfio_pci_liveupdate_freeze() guarantees the device is reset prior to Live Update, so there's no reason to reset it again after Live Update.

Since VFIO normally uses the initial reset to detect if the device supports function resets, pass that from the previous kernel via struct vfio_pci_core_dev_ser.

---
 drivers/vfio/pci/vfio_pci_core.c       | 22 +++++++++++++++++-----
 drivers/vfio/pci/vfio_pci_liveupdate.c |  1 +
 include/linux/kho/abi/vfio_pci.h       |  2 ++
 include/linux/vfio_pci_core.h          |  1 +
 4 files changed, 21 insertions(+), 5 deletions(-)
drivers/vfio/pci/vfio_pci_core.c +17 -5
index b09fe0993e04..c3b30f08a788 100644
@@ -482,12 +482,24 @@ int vfio_pci_core_enable(struct vfio_pci_core_device *vdev)
 	if (ret)
 		goto out_power;
 
-	/* If reset fails because of the device lock, fail this path entirely */
-	ret = pci_try_reset_function(pdev);
-	if (ret == -EAGAIN)
-		goto out_disable_device;
+	if (vdev->liveupdate_state) {
+		/*
+		 * This device was handed off by vfio-pci from a previous kernel
+		 * via Live Update, so it does not need to be reset.
+		 */
+		vdev->reset_works = vdev->liveupdate_state->reset_works;
+	} else {
+		/*
+		 * If reset fails because of the device lock, fail this path
+		 * entirely.
+		 */
+		ret = pci_try_reset_function(pdev);
+		if (ret == -EAGAIN)
+			goto out_disable_device;
+
+		vdev->reset_works = !ret;
+	}
 
-	vdev->reset_works = !ret;
 	pci_save_state(pdev);
 	vdev->pci_saved_state = pci_store_saved_state(pdev);
 	if (!vdev->pci_saved_state)
drivers/vfio/pci/vfio_pci_liveupdate.c +1 -0
index 0fb29bd3ae3b..bcaf9de8a823 100644
@@ -59,6 +59,7 @@ static int vfio_pci_liveupdate_preserve(struct liveupdate_file_op_args *args)
 
 	ser->bdf = pci_dev_id(pdev);
 	ser->domain = pci_domain_nr(pdev->bus);
+	ser->reset_works = vdev->reset_works;
 
 	err = kho_preserve_folio(folio);
 	if (err)
include/linux/kho/abi/vfio_pci.h +2 -0
index 9bf58a2f3820..6c3d3c6dfc09 100644
@@ -34,10 +34,12 @@
  *
  * @bdf: The device's PCI bus, device, and function number.
  * @domain: The device's PCI domain number (segment).
+ * @reset_works: Non-zero if the device supports function resets.
  */
 struct vfio_pci_core_device_ser {
 	u16 bdf;
 	u16 domain;
+	u8 reset_works;
 } __packed;
 
 #endif /* _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H */
include/linux/vfio_pci_core.h +1 -0
index 56ff6452562d..3421721a1615 100644
@@ -15,6 +15,7 @@
 #include <linux/types.h>
 #include <linux/uuid.h>
 #include <linux/notifier.h>
+#include <linux/kho/abi/vfio_pci.h>
 
 #ifndef VFIO_PCI_CORE_H
 #define VFIO_PCI_CORE_H
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#32 lore raw

PATCH 11/21 selftests/liveupdate: Move luo_test_utils.* into a reusable library

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Move luo_test_utils.[ch] into a lib/ directory and pull the rules to build them out into a separate make script. This will enable these utilities to be also built by and used within other selftests (such as VFIO) in subsequent commits.

No functional change intended.

---
 tools/testing/selftests/liveupdate/.gitignore |  1 +
 tools/testing/selftests/liveupdate/Makefile   | 14 ++++---------
 .../include/libliveupdate.h}                  |  8 ++++----
 .../selftests/liveupdate/lib/libliveupdate.mk | 20 +++++++++++++++++++
 .../{luo_test_utils.c => lib/liveupdate.c}    |  2 +-
 .../selftests/liveupdate/luo_kexec_simple.c   |  2 +-
 .../selftests/liveupdate/luo_multi_session.c  |  2 +-
 7 files changed, 32 insertions(+), 17 deletions(-)
 rename tools/testing/selftests/liveupdate/{luo_test_utils.h => lib/include/libliveupdate.h} (87%)
 create mode 100644 tools/testing/selftests/liveupdate/lib/libliveupdate.mk
 rename tools/testing/selftests/liveupdate/{luo_test_utils.c => lib/liveupdate.c} (99%)
tools/testing/selftests/liveupdate/.gitignore +1 -0
index 661827083ab6..18a0c7036cf3 100644
@@ -3,6 +3,7 @@
 !/**/
 !*.c
 !*.h
+!*.mk
 !*.sh
 !.gitignore
 !config
tools/testing/selftests/liveupdate/Makefile +4 -10
index 080754787ede..a060cc21f27f 100644
@@ -1,7 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
 
-LIB_C += luo_test_utils.c
-
 TEST_GEN_PROGS += liveupdate
 
 TEST_GEN_PROGS_EXTENDED += luo_kexec_simple
@@ -10,25 +8,21 @@ TEST_GEN_PROGS_EXTENDED += luo_multi_session
 TEST_FILES += do_kexec.sh
 
 include ../lib.mk
+include lib/libliveupdate.mk
 
 CFLAGS += $(KHDR_INCLUDES)
 CFLAGS += -Wall -O2 -Wno-unused-function
 CFLAGS += -MD
 
-LIB_O := $(patsubst %.c, $(OUTPUT)/%.o, $(LIB_C))
 TEST_O := $(patsubst %, %.o, $(TEST_GEN_PROGS))
 TEST_O += $(patsubst %, %.o, $(TEST_GEN_PROGS_EXTENDED))
 
-TEST_DEP_FILES := $(patsubst %.o, %.d, $(LIB_O))
+TEST_DEP_FILES := $(patsubst %.o, %.d, $(LIBLIVEUPDATE_O))
 TEST_DEP_FILES += $(patsubst %.o, %.d, $(TEST_O))
 -include $(TEST_DEP_FILES)
 
-$(LIB_O): $(OUTPUT)/%.o: %.c
-	$(CC) $(CFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c $< -o $@
-
-$(TEST_GEN_PROGS) $(TEST_GEN_PROGS_EXTENDED): $(OUTPUT)/%: %.o $(LIB_O)
-	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH) $< $(LIB_O) $(LDLIBS) -o $@
+$(TEST_GEN_PROGS) $(TEST_GEN_PROGS_EXTENDED): $(OUTPUT)/%: %.o $(LIBLIVEUPDATE_O)
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH) $< $(LIBLIVEUPDATE_O) $(LDLIBS) -o $@
 
-EXTRA_CLEAN += $(LIB_O)
 EXTRA_CLEAN += $(TEST_O)
 EXTRA_CLEAN += $(TEST_DEP_FILES)
tools/testing/selftests/liveupdate/lib/include/libliveupdate.h renamed from tools/testing/selftests/liveupdate/luo_test_utils.h +4 -4
similarity index 87%
rename from tools/testing/selftests/liveupdate/luo_test_utils.h
rename to tools/testing/selftests/liveupdate/lib/include/libliveupdate.h
index 90099bf49577..4390a2737930 100644
@@ -7,13 +7,13 @@
  * Utility functions for LUO kselftests.
  */
 
-#ifndef LUO_TEST_UTILS_H
-#define LUO_TEST_UTILS_H
+#ifndef SELFTESTS_LIVEUPDATE_LIB_LIVEUPDATE_H
+#define SELFTESTS_LIVEUPDATE_LIB_LIVEUPDATE_H
 
 #include <errno.h>
 #include <string.h>
 #include <linux/liveupdate.h>
-#include "../kselftest.h"
+#include "../../../kselftest.h"
 
 #define LUO_DEVICE "/dev/liveupdate"
 
@@ -41,4 +41,4 @@ typedef void (*luo_test_stage2_fn)(int luo_fd, int state_session_fd);
 int luo_test(int argc, char *argv[], const char *state_session_name,
 	     luo_test_stage1_fn stage1, luo_test_stage2_fn stage2);
 
-#endif /* LUO_TEST_UTILS_H */
+#endif /* SELFTESTS_LIVEUPDATE_LIB_LIVEUPDATE_H */
tools/testing/selftests/liveupdate/lib/libliveupdate.mk new file +20 -0
new file mode 100644
index 000000000000..fffd95b085b6
@@ -0,0 +1,20 @@
+include $(top_srcdir)/scripts/subarch.include
+ARCH ?= $(SUBARCH)
+
+LIBLIVEUPDATE_SRCDIR := $(selfdir)/liveupdate/lib
+
+LIBLIVEUPDATE_C := liveupdate.c
+
+LIBLIVEUPDATE_OUTPUT := $(OUTPUT)/libliveupdate
+
+LIBLIVEUPDATE_O := $(patsubst %.c, $(LIBLIVEUPDATE_OUTPUT)/%.o, $(LIBLIVEUPDATE_C))
+
+LIBLIVEUPDATE_O_DIRS := $(shell dirname $(LIBLIVEUPDATE_O) | uniq)
+$(shell mkdir -p $(LIBLIVEUPDATE_O_DIRS))
+
+CFLAGS += -I$(LIBLIVEUPDATE_SRCDIR)/include
+
+$(LIBLIVEUPDATE_O): $(LIBLIVEUPDATE_OUTPUT)/%.o : $(LIBLIVEUPDATE_SRCDIR)/%.c
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c $< -o $@
+
+EXTRA_CLEAN += $(LIBLIVEUPDATE_OUTPUT)
tools/testing/selftests/liveupdate/lib/liveupdate.c renamed from tools/testing/selftests/liveupdate/luo_test_utils.c +1 -1
similarity index 99%
rename from tools/testing/selftests/liveupdate/luo_test_utils.c
rename to tools/testing/selftests/liveupdate/lib/liveupdate.c
index 3c8721c505df..60121873f685 100644
@@ -21,7 +21,7 @@
 #include <errno.h>
 #include <stdarg.h>
 
-#include "luo_test_utils.h"
+#include <libliveupdate.h>
 
 int luo_open_device(void)
 {
tools/testing/selftests/liveupdate/luo_kexec_simple.c +1 -1
index d7ac1f3dc4cb..786ac93b9ae3 100644
@@ -8,7 +8,7 @@
  * across a single kexec reboot.
  */
 
-#include "luo_test_utils.h"
+#include <libliveupdate.h>
 
 #define TEST_SESSION_NAME "test-session"
 #define TEST_MEMFD_TOKEN 0x1A
tools/testing/selftests/liveupdate/luo_multi_session.c +1 -1
index 0ee2d795beef..aac24a5f5ce3 100644
@@ -9,7 +9,7 @@
  * files.
  */
 
-#include "luo_test_utils.h"
+#include <libliveupdate.h>
 
 #define SESSION_EMPTY_1 "multi-test-empty-1"
 #define SESSION_EMPTY_2 "multi-test-empty-2"
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#33 lore raw

PATCH 12/21 selftests/liveupdate: Add helpers to preserve/retrieve FDs

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Add helper functions to preserve and retrieve file descriptors from an LUO session. These will be used be used in subsequent commits to preserve FDs other than memfd.

No functional change intended.

---
 .../liveupdate/lib/include/libliveupdate.h    |  3 ++
 .../selftests/liveupdate/lib/liveupdate.c     | 41 +++++++++++++++----
 2 files changed, 35 insertions(+), 9 deletions(-)
tools/testing/selftests/liveupdate/lib/include/libliveupdate.h +3 -0
index 4390a2737930..4c93d043d2b3 100644
@@ -26,6 +26,9 @@ int luo_create_session(int luo_fd, const char *name);
 int luo_retrieve_session(int luo_fd, const char *name);
 int luo_session_finish(int session_fd);
 
+int luo_session_preserve_fd(int session_fd, int fd, int token);
+int luo_session_retrieve_fd(int session_fd, int token);
+
 int create_and_preserve_memfd(int session_fd, int token, const char *data);
 int restore_and_verify_memfd(int session_fd, int token, const char *expected_data);
 
tools/testing/selftests/liveupdate/lib/liveupdate.c +32 -9

65 lines not on this page · raw message

Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#34 lore raw

PATCH 13/21 vfio: selftests: Build liveupdate library in VFIO selftests

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Import and build liveupdate selftest library in VFIO selftests.

It allows to use liveupdate ioctls in VFIO selftests

---
 tools/testing/selftests/vfio/Makefile | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)
tools/testing/selftests/vfio/Makefile +8 -3
index 8bb0b1e2d3a3..8e6e2cc2d8fd 100644
@@ -11,6 +11,7 @@ TEST_PROGS_EXTENDED := scripts/run.sh
 TEST_PROGS_EXTENDED := scripts/setup.sh
 include ../lib.mk
 include lib/libvfio.mk
+include ../liveupdate/lib/libliveupdate.mk
 
 CFLAGS += -I$(top_srcdir)/tools/include
 CFLAGS += -MD
@@ -18,11 +19,15 @@ CFLAGS += $(EXTRA_CFLAGS)
 
 LDFLAGS += -pthread
 
-$(TEST_GEN_PROGS): %: %.o $(LIBVFIO_O)
-	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $< $(LIBVFIO_O) $(LDLIBS) -o $@
+LIBS_O := $(LIBVFIO_O)
+LIBS_O += $(LIBLIVEUPDATE_O)
+
+$(TEST_GEN_PROGS): %: %.o $(LIBS_O)
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH) $< $(LIBS_O) $(LDLIBS) -o $@
 
 TEST_GEN_PROGS_O = $(patsubst %, %.o, $(TEST_GEN_PROGS))
-TEST_DEP_FILES = $(patsubst %.o, %.d, $(TEST_GEN_PROGS_O) $(LIBVFIO_O))
+TEST_DEP_FILES := $(patsubst %.o, %.d, $(TEST_GEN_PROGS_O))
+TEST_DEP_FILES += $(patsubst %.o, %.d, $(LIBS_O))
 -include $(TEST_DEP_FILES)
 
 EXTRA_CLEAN += $(TEST_GEN_PROGS_O) $(TEST_DEP_FILES)
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#35 lore raw

PATCH 14/21 vfio: selftests: Add Makefile support for TEST_GEN_PROGS_EXTENDED

In reply to David Matlack

Add Makefile support for TEST_GEN_PROGS_EXTENDED targets. These tests are not run by default.

TEST_GEN_PROGS_EXTENDED will be used for Live Update selftests in subsequent commits. These selftests must be run manually because they require the user/runner to perform additional actions, such as kexec, during the test.

---
 tools/testing/selftests/vfio/Makefile | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)
tools/testing/selftests/vfio/Makefile +7 -4
index 8e6e2cc2d8fd..424649049d94 100644
@@ -22,12 +22,15 @@ LDFLAGS += -pthread
 LIBS_O := $(LIBVFIO_O)
 LIBS_O += $(LIBLIVEUPDATE_O)
 
-$(TEST_GEN_PROGS): %: %.o $(LIBS_O)
+$(TEST_GEN_PROGS) $(TEST_GEN_PROGS_EXTENDED): %: %.o $(LIBS_O)
 	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $(TARGET_ARCH) $< $(LIBS_O) $(LDLIBS) -o $@
 
-TEST_GEN_PROGS_O = $(patsubst %, %.o, $(TEST_GEN_PROGS))
-TEST_DEP_FILES := $(patsubst %.o, %.d, $(TEST_GEN_PROGS_O))
+TESTS_O := $(patsubst %, %.o, $(TEST_GEN_PROGS))
+TESTS_O += $(patsubst %, %.o, $(TEST_GEN_PROGS_EXTENDED))
+
+TEST_DEP_FILES := $(patsubst %.o, %.d, $(TESTS_O))
 TEST_DEP_FILES += $(patsubst %.o, %.d, $(LIBS_O))
 -include $(TEST_DEP_FILES)
 
-EXTRA_CLEAN += $(TEST_GEN_PROGS_O) $(TEST_DEP_FILES)
+EXTRA_CLEAN += $(TESTS_O)
+EXTRA_CLEAN += $(TEST_DEP_FILES)
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#36 lore raw

PATCH 15/21 vfio: selftests: Add vfio_pci_liveupdate_uapi_test

In reply to David Matlack

Add a selftest to exercise preserving a various VFIO files through /dev/liveupdate. Ensure that VFIO cdev device files can be preserved and everything else (group-based device files, group files, and container files) all fail.

---
 tools/testing/selftests/vfio/Makefile         |  1 +
 .../vfio/vfio_pci_liveupdate_uapi_test.c      | 93 +++++++++++++++++++
 2 files changed, 94 insertions(+)
 create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c
tools/testing/selftests/vfio/Makefile +1 -0
index 424649049d94..9b2a4f10c558 100644
@@ -4,6 +4,7 @@ TEST_GEN_PROGS += vfio_iommufd_setup_test
 TEST_GEN_PROGS += vfio_pci_device_test
 TEST_GEN_PROGS += vfio_pci_device_init_perf_test
 TEST_GEN_PROGS += vfio_pci_driver_test
+TEST_GEN_PROGS += vfio_pci_liveupdate_uapi_test
 
 TEST_PROGS_EXTENDED := scripts/cleanup.sh
 TEST_PROGS_EXTENDED := scripts/lib.sh
tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c new file +93 -0

93 lines not on this page · raw message

Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#37 lore raw

PATCH 16/21 vfio: selftests: Initialize vfio_pci_device using a VFIO cdev FD

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Use the given VFIO cdev FD to initialize vfio_pci_device in VFIO selftests. Add the assertion to make sure that passed cdev FD is not used with legacy VFIO APIs. If VFIO cdev FD is provided then do not open the device instead use the FD for any interaction with the device.

This API will allow to write selftests where VFIO device FD is preserved using liveupdate and retrieved later using liveupdate ioctl after kexec.

---
 .../lib/include/libvfio/vfio_pci_device.h     |  3 ++
 .../selftests/vfio/lib/vfio_pci_device.c      | 33 ++++++++++++++-----
 2 files changed, 27 insertions(+), 9 deletions(-)
tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h +3 -0
index 2858885a89bb..896dfde88118 100644
@@ -38,6 +38,9 @@ struct vfio_pci_device {
 #define dev_info(_dev, _fmt, ...) printf("%s: " _fmt, (_dev)->bdf, ##__VA_ARGS__)
 #define dev_err(_dev, _fmt, ...) fprintf(stderr, "%s: " _fmt, (_dev)->bdf, ##__VA_ARGS__)
 
+struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
+					       struct iommu *iommu,
+					       int device_fd);
 struct vfio_pci_device *vfio_pci_device_init(const char *bdf, struct iommu *iommu);
 void vfio_pci_device_cleanup(struct vfio_pci_device *device);
 
tools/testing/selftests/vfio/lib/vfio_pci_device.c +24 -9
index 468ee1c61b0c..e9423dc3864a 100644
@@ -319,19 +319,27 @@ static void vfio_device_attach_iommufd_pt(int device_fd, u32 pt_id)
 	ioctl_assert(device_fd, VFIO_DEVICE_ATTACH_IOMMUFD_PT, &args);
 }
 
-static void vfio_pci_iommufd_setup(struct vfio_pci_device *device, const char *bdf)
+static void vfio_pci_iommufd_setup(struct vfio_pci_device *device,
+				   const char *bdf, int device_fd)
 {
-	const char *cdev_path = vfio_pci_get_cdev_path(bdf);
+	const char *cdev_path;
 
-	device->fd = open(cdev_path, O_RDWR);
-	VFIO_ASSERT_GE(device->fd, 0);
-	free((void *)cdev_path);
+	if (device_fd >= 0) {
+		device->fd = device_fd;
+	} else {
+		cdev_path = vfio_pci_get_cdev_path(bdf);
+		device->fd = open(cdev_path, O_RDWR);
+		VFIO_ASSERT_GE(device->fd, 0);
+		free((void *)cdev_path);
+	}
 
 	vfio_device_bind_iommufd(device->fd, device->iommu->iommufd);
 	vfio_device_attach_iommufd_pt(device->fd, device->iommu->ioas_id);
 }
 
-struct vfio_pci_device *vfio_pci_device_init(const char *bdf, struct iommu *iommu)
+struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
+					       struct iommu *iommu,
+					       int device_fd)
 {
 	struct vfio_pci_device *device;
 
@@ -341,10 +349,12 @@ struct vfio_pci_device *vfio_pci_device_init(const char *bdf, struct iommu *iomm
 	device->bdf = bdf;
 	device->iommu = iommu;
 
-	if (device->iommu->mode->container_path)
+	if (device->iommu->mode->container_path) {
+		VFIO_ASSERT_EQ(device_fd, -1);
 		vfio_pci_container_setup(device, bdf);
-	else
-		vfio_pci_iommufd_setup(device, bdf);
+	} else {
+		vfio_pci_iommufd_setup(device, bdf, device_fd);
+	}
 
 	vfio_pci_device_setup(device);
 	vfio_pci_driver_probe(device);
@@ -352,6 +362,11 @@ struct vfio_pci_device *vfio_pci_device_init(const char *bdf, struct iommu *iomm
 	return device;
 }
 
+struct vfio_pci_device *vfio_pci_device_init(const char *bdf, struct iommu *iommu)
+{
+	return __vfio_pci_device_init(bdf, iommu, /*device_fd=*/-1);
+}
+
 void vfio_pci_device_cleanup(struct vfio_pci_device *device)
 {
 	int i;
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#38 lore raw

PATCH 17/21 vfio: selftests: Add vfio_pci_liveupdate_kexec_test

In reply to David Matlack

From: Vipin Sharma <[email protected]>

Add a selftest to exercise preserving a vfio-pci device across a Live Update. For now the test is extremely simple and just verifies that the device file can be preserved and retrieved. In the future this test will be extended to verify more parts about device preservation as they are implemented.

This test is added to TEST_GEN_PROGS_EXTENDED since it must be run manually along with a kexec.

To run this test manually:

 $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
 $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test --stage 1 0000:00:04.0
 $ kexec ...   # NOTE: Exact method will be distro-dependent
 $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
 $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test --stage 2 0000:00:04.0

The second call to setup.sh is necessary because preserved devices are not bound to a driver after Live Update. Such devices must be manually bound by userspace after Live Update via driver_override.

This test is considered passing if all commands exit with 0.

---
 tools/testing/selftests/vfio/Makefile         |  4 +
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 82 +++++++++++++++++++
 2 files changed, 86 insertions(+)
 create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
tools/testing/selftests/vfio/Makefile +4 -0
index 9b2a4f10c558..177aa5cd4376 100644
@@ -6,6 +6,10 @@ TEST_GEN_PROGS += vfio_pci_device_init_perf_test
 TEST_GEN_PROGS += vfio_pci_driver_test
 TEST_GEN_PROGS += vfio_pci_liveupdate_uapi_test
 
+# This test must be run manually since it requires the user/automation to
+# perform a kexec during the test.
+TEST_GEN_PROGS_EXTENDED += vfio_pci_liveupdate_kexec_test
+
 TEST_PROGS_EXTENDED := scripts/cleanup.sh
 TEST_PROGS_EXTENDED := scripts/lib.sh
 TEST_PROGS_EXTENDED := scripts/run.sh
tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c new file +82 -0

82 lines not on this page · raw message

Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#39 lore raw

PATCH 18/21 vfio: selftests: Expose iommu_modes to tests

In reply to David Matlack

Expose the list of iommu_modes to enable tests that want to iterate through all possible iommu modes.

---
 tools/testing/selftests/vfio/lib/include/libvfio/iommu.h | 2 ++
 tools/testing/selftests/vfio/lib/iommu.c                 | 4 +++-
 2 files changed, 5 insertions(+), 1 deletion(-)
tools/testing/selftests/vfio/lib/include/libvfio/iommu.h +2 -0
index 5c9b9dc6d993..a03ff2281f11 100644
@@ -15,6 +15,8 @@ struct iommu_mode {
 	unsigned long iommu_type;
 };
 
+extern const struct iommu_mode iommu_modes[];
+extern const int nr_iommu_modes;
 extern const char *default_iommu_mode;
 
 struct dma_region {
tools/testing/selftests/vfio/lib/iommu.c +3 -1
index 8079d43523f3..bf54f95ec833 100644
@@ -24,7 +24,7 @@
 const char *default_iommu_mode = "iommufd";
 
 /* Reminder: Keep in sync with FIXTURE_VARIANT_ADD_ALL_IOMMU_MODES(). */
-static const struct iommu_mode iommu_modes[] = {
+const struct iommu_mode iommu_modes[] = {
 	{
 		.name = "vfio_type1_iommu",
 		.container_path = "/dev/vfio/vfio",
@@ -50,6 +50,8 @@ static const struct iommu_mode iommu_modes[] = {
 	},
 };
 
+const int nr_iommu_modes = ARRAY_SIZE(iommu_modes);
+
 static const struct iommu_mode *lookup_iommu_mode(const char *iommu_mode)
 {
 	int i;
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#40 lore raw

PATCH 19/21 vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device

In reply to David Matlack

Expose a few low-level helper routings for setting up vfio_pci_device structs. These routines will be used in a subsequent commit to assert that VFIO_GROUP_GET_DEVICE_FD fails under certain conditions.

---
 .../lib/include/libvfio/vfio_pci_device.h     |  5 +++
 .../selftests/vfio/lib/vfio_pci_device.c      | 33 +++++++++++++------
 2 files changed, 28 insertions(+), 10 deletions(-)
tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h +5 -0
index 896dfde88118..2389c7698335 100644
@@ -125,4 +125,9 @@ static inline bool vfio_pci_device_match(struct vfio_pci_device *device,
 
 const char *vfio_pci_get_cdev_path(const char *bdf);
 
+/* Low-level routines for setting up a struct vfio_pci_device */
+struct vfio_pci_device *vfio_pci_device_alloc(const char *bdf, struct iommu *iommu);
+void vfio_pci_group_setup(struct vfio_pci_device *device);
+void vfio_pci_iommu_setup(struct vfio_pci_device *device);
+
 #endif /* SELFTESTS_VFIO_LIB_INCLUDE_LIBVFIO_VFIO_PCI_DEVICE_H */
tools/testing/selftests/vfio/lib/vfio_pci_device.c +23 -10
index e9423dc3864a..c1a3886dee30 100644
@@ -199,7 +199,7 @@ static unsigned int vfio_pci_get_group_from_dev(const char *bdf)
 	return group;
 }
 
-static void vfio_pci_group_setup(struct vfio_pci_device *device, const char *bdf)
+void vfio_pci_group_setup(struct vfio_pci_device *device)
 {
 	struct vfio_group_status group_status = {
 		.argsz = sizeof(group_status),
@@ -207,7 +207,7 @@ static void vfio_pci_group_setup(struct vfio_pci_device *device, const char *bdf
 	char group_path[32];
 	int group;
 
-	group = vfio_pci_get_group_from_dev(bdf);
+	group = vfio_pci_get_group_from_dev(device->bdf);
 	snprintf(group_path, sizeof(group_path), "/dev/vfio/%d", group);
 
 	device->group_fd = open(group_path, O_RDWR);
@@ -219,14 +219,12 @@ static void vfio_pci_group_setup(struct vfio_pci_device *device, const char *bdf
 	ioctl_assert(device->group_fd, VFIO_GROUP_SET_CONTAINER, &device->iommu->container_fd);
 }
 
-static void vfio_pci_container_setup(struct vfio_pci_device *device, const char *bdf)
+void vfio_pci_iommu_setup(struct vfio_pci_device *device)
 {
 	struct iommu *iommu = device->iommu;
 	unsigned long iommu_type = iommu->mode->iommu_type;
 	int ret;
 
-	vfio_pci_group_setup(device, bdf);
-
 	ret = ioctl(iommu->container_fd, VFIO_CHECK_EXTENSION, iommu_type);
 	VFIO_ASSERT_GT(ret, 0, "VFIO IOMMU type %lu not supported\n", iommu_type);
 
@@ -236,8 +234,14 @@ static void vfio_pci_container_setup(struct vfio_pci_device *device, const char
 	 * because the IOMMU type is already set.
 	 */
 	(void)ioctl(iommu->container_fd, VFIO_SET_IOMMU, (void *)iommu_type);
+}
 
-	device->fd = ioctl(device->group_fd, VFIO_GROUP_GET_DEVICE_FD, bdf);
+static void vfio_pci_container_setup(struct vfio_pci_device *device)
+{
+	vfio_pci_group_setup(device);
+	vfio_pci_iommu_setup(device);
+
+	device->fd = ioctl(device->group_fd, VFIO_GROUP_GET_DEVICE_FD, device->bdf);
 	VFIO_ASSERT_GE(device->fd, 0);
 }
 
@@ -337,9 +341,7 @@ static void vfio_pci_iommufd_setup(struct vfio_pci_device *device,
 	vfio_device_attach_iommufd_pt(device->fd, device->iommu->ioas_id);
 }
 
-struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
-					       struct iommu *iommu,
-					       int device_fd)
+struct vfio_pci_device *vfio_pci_device_alloc(const char *bdf, struct iommu *iommu)
 {
 	struct vfio_pci_device *device;
 
@@ -349,9 +351,20 @@ struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
 	device->bdf = bdf;
 	device->iommu = iommu;
 
+	return device;
+}
+
+struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
+					       struct iommu *iommu,
+					       int device_fd)
+{
+	struct vfio_pci_device *device;
+
+	device = vfio_pci_device_alloc(bdf, iommu);
+
 	if (device->iommu->mode->container_path) {
 		VFIO_ASSERT_EQ(device_fd, -1);
-		vfio_pci_container_setup(device, bdf);
+		vfio_pci_container_setup(device);
 	} else {
 		vfio_pci_iommufd_setup(device, bdf, device_fd);
 	}
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
#41 lore raw

Re: [PATCH 19/21] vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device

Quoted 90 lines · David Matlack

Expose a few low-level helper routings for setting up vfio_pci_device structs. These routines will be used in a subsequent commit to assert that VFIO_GROUP_GET_DEVICE_FD fails under certain conditions.

---
  .../lib/include/libvfio/vfio_pci_device.h     |  5 +++
  .../selftests/vfio/lib/vfio_pci_device.c      | 33 +++++++++++++------
  2 files changed, 28 insertions(+), 10 deletions(-)
tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h +5 -0
index 896dfde88118..2389c7698335 100644
@@ -125,4 +125,9 @@ static inline bool vfio_pci_device_match(struct vfio_pci_device *device,
 
  const char *vfio_pci_get_cdev_path(const char *bdf);
 
+/* Low-level routines for setting up a struct vfio_pci_device */
+struct vfio_pci_device *vfio_pci_device_alloc(const char *bdf, struct iommu *iommu);
+void vfio_pci_group_setup(struct vfio_pci_device *device);
+void vfio_pci_iommu_setup(struct vfio_pci_device *device);
+
  #endif /* SELFTESTS_VFIO_LIB_INCLUDE_LIBVFIO_VFIO_PCI_DEVICE_H */
tools/testing/selftests/vfio/lib/vfio_pci_device.c +11 -9
index e9423dc3864a..c1a3886dee30 100644
@@ -199,7 +199,7 @@ static unsigned int vfio_pci_get_group_from_dev(const char *bdf)
  	return group;
  }
 
-static void vfio_pci_group_setup(struct vfio_pci_device *device, const char *bdf)
+void vfio_pci_group_setup(struct vfio_pci_device *device)
  {
  	struct vfio_group_status group_status = {
  		.argsz = sizeof(group_status),
@@ -207,7 +207,7 @@ static void vfio_pci_group_setup(struct vfio_pci_device *device, const char *bdf
  	char group_path[32];
  	int group;
 
-	group = vfio_pci_get_group_from_dev(bdf);
+	group = vfio_pci_get_group_from_dev(device->bdf);
  	snprintf(group_path, sizeof(group_path), "/dev/vfio/%d", group);
 
  	device->group_fd = open(group_path, O_RDWR);
@@ -219,14 +219,12 @@ static void vfio_pci_group_setup(struct vfio_pci_device *device, const char *bdf
  	ioctl_assert(device->group_fd, VFIO_GROUP_SET_CONTAINER, &device->iommu->container_fd);
  }
 
-static void vfio_pci_container_setup(struct vfio_pci_device *device, const char *bdf)
+void vfio_pci_iommu_setup(struct vfio_pci_device *device)
  {
  	struct iommu *iommu = device->iommu;
  	unsigned long iommu_type = iommu->mode->iommu_type;
  	int ret;
 
-	vfio_pci_group_setup(device, bdf);
-
  	ret = ioctl(iommu->container_fd, VFIO_CHECK_EXTENSION, iommu_type);
  	VFIO_ASSERT_GT(ret, 0, "VFIO IOMMU type %lu not supported\n", iommu_type);
 
@@ -236,8 +234,14 @@ static void vfio_pci_container_setup(struct vfio_pci_device *device, const char
  	 * because the IOMMU type is already set.
  	 */
  	(void)ioctl(iommu->container_fd, VFIO_SET_IOMMU, (void *)iommu_type);
+}
 
-	device->fd = ioctl(device->group_fd, VFIO_GROUP_GET_DEVICE_FD, bdf);
+static void vfio_pci_container_setup(struct vfio_pci_device *device)
+{
+	vfio_pci_group_setup(device);
+	vfio_pci_iommu_setup(device);
+
+	device->fd = ioctl(device->group_fd, VFIO_GROUP_GET_DEVICE_FD, device->bdf);
  	VFIO_ASSERT_GE(device->fd, 0);
  }
 
@@ -337,9 +341,7 @@ static void vfio_pci_iommufd_setup(struct vfio_pci_device *device,
  	vfio_device_attach_iommufd_pt(device->fd, device->iommu->ioas_id);
  }
 
-struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
-					       struct iommu *iommu,
-					       int device_fd)
+struct vfio_pci_device *vfio_pci_device_alloc(const char *bdf, struct iommu *iommu)
  {
  	struct vfio_pci_device *device;
 

tools/testing/selftests/vfio/lib/vfio_pci_device.c struct vfio_pci_device *__vfio_pci_device_init(const char *bdf, · line 356

Quoted 6 lines
tools/testing/selftests/vfio/lib/vfio_pci_device.c +3 -0
@@ -349,9 +351,20 @@ struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
  	device->bdf = bdf;
  	device->iommu = iommu;
 
+	return device;
+}
+

In the latest kernel, this part changes too much.

Yanjun.Zhu

Quoted 15 lines
tools/testing/selftests/vfio/lib/vfio_pci_device.c +9 -1
+struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
+					       struct iommu *iommu,
+					       int device_fd)
+{
+	struct vfio_pci_device *device;
+
+	device = vfio_pci_device_alloc(bdf, iommu);
+
  	if (device->iommu->mode->container_path) {
  		VFIO_ASSERT_EQ(device_fd, -1);
-		vfio_pci_container_setup(device, bdf);
+		vfio_pci_container_setup(device);
  	} else {
  		vfio_pci_iommufd_setup(device, bdf, device_fd);
  	}
Signature · 2 lines
Best Regards,
Yanjun.Zhu
author maintainer
#42 lore raw

Re: [PATCH 19/21] vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device

Quoted 10 lines · Zhu Yanjun

tools/testing/selftests/vfio/lib/vfio_pci_device.c struct vfio_pci_device *__vfio_pci_device_init(const char *bdf, · line 356

Quoted 7 lines · David Matlack
tools/testing/selftests/vfio/lib/vfio_pci_device.c +3 -0
@@ -349,9 +351,20 @@ struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
      device->bdf = bdf;
      device->iommu = iommu;
 
+     return device;
+}
+

In the latest kernel, this part changes too much.

Can you clarify what you mean by "changes too much"? What is the issue?

#43 lore raw

Re: [PATCH 19/21] vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device

Quoted 11 lines · David Matlack
Quoted 9 lines · Zhu Yanjun

tools/testing/selftests/vfio/lib/vfio_pci_device.c struct vfio_pci_device *__vfio_pci_device_init(const char *bdf, · line 356

Quoted 7 lines · David Matlack
tools/testing/selftests/vfio/lib/vfio_pci_device.c +3 -0
@@ -349,9 +351,20 @@ struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
       device->bdf = bdf;
       device->iommu = iommu;
 
+     return device;
+}
+

In the latest kernel, this part changes too much.

Can you clarify what you mean by "changes too much"? What is the issue?

I tried to apply this commit to the linux and linux-next repositories and run tests.

However, I’m unable to apply [PATCH 19/21] vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device, because the related source code has changed significantly in both linux and linux-next.

If you plan to resend this patch series based on the latest linux or linux-next, please feel free to ignore this comment.

I look forward to testing the updated patch series once it is available.

Best Regards,

Yanjun.Zhu

author maintainer
#44 lore raw

Re: [PATCH 19/21] vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device

Quoted 15 lines · Yanjun.Zhu
Quoted 11 lines · David Matlack
Quoted 9 lines · Zhu Yanjun

tools/testing/selftests/vfio/lib/vfio_pci_device.c struct vfio_pci_device *__vfio_pci_device_init(const char *bdf, · line 356

Quoted 7 lines · David Matlack
tools/testing/selftests/vfio/lib/vfio_pci_device.c +3 -0
@@ -349,9 +351,20 @@ struct vfio_pci_device *__vfio_pci_device_init(const char *bdf,
       device->bdf = bdf;
       device->iommu = iommu;
 
+     return device;
+}
+

In the latest kernel, this part changes too much.

Can you clarify what you mean by "changes too much"? What is the issue?

Quoted 7 lines

I tried to apply this commit to the linux and linux-next repositories and run tests.

However, I’m unable to apply [PATCH 19/21] vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device, because the related source code has changed significantly in both linux and linux-next.

Ahhh. This series depends on several in-flight series, so I'm not surprised it doesn't apply cleanly. There is this blurb in the cover letter:
---
This series was constructed on top of several in-flight series and on top of mm-nonmm-unstable [2].

  +-- This series
  |
  +-- [PATCH v2 00/18] vfio: selftests: Support for multi-device tests
  |    https://lore.kernel.org/kvm/[email protected]/
  |
  +-- [PATCH v3 0/4] vfio: selftests: update DMA mapping tests to use
queried IOVA ranges
  |   https://lore.kernel.org/kvm/[email protected]/
  |
  +-- [PATCH v8 0/2] Live Update: File-Lifecycle-Bound (FLB) State
  |   https://lore.kernel.org/linux-mm/[email protected]/
  |
  +-- [PATCH v8 00/18] Live Update Orchestrator
  |   https://lore.kernel.org/linux-mm/[email protected]/
  |

To simplify checking out the code, this series can be found on GitHub:

https://github.com/dmatlack/linux/tree/liveupdate/vfio/cdev/v1
---

Cloning the GitHub repo is probably your simplest option if you want to check out the code and run some tests.

Quoted 5 lines

If you plan to resend this patch series based on the latest linux or linux-next, please feel free to ignore this comment.

I look forward to testing the updated patch series once it is available.

I will send out an updated patch set hopefully within the next 2 weeks.

Thanks for taking a look!

author maintainer
#45 lore raw

PATCH 20/21 vfio: selftests: Verify that opening VFIO device fails during Live Update

In reply to David Matlack

Verify that opening a VFIO device through its cdev file and via VFIO_GROUP_GET_DEVICE_FD both fail with -EBUSY if the device was preserved across a Live Update. When a device file is preserve across a Live Update, the file must be retrieved from /dev/liveupdate, not from VFIO directly.

---
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 38 +++++++++++++++++++
 1 file changed, 38 insertions(+)
tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c +38 -0
index 95644c3bd2d3..925c5fc30d56 100644
@@ -36,6 +36,42 @@ static void before_kexec(int luo_fd)
 	daemonize_and_wait();
 }
 
+static void check_open_vfio_device_fails(void)
+{
+	const char *cdev_path = vfio_pci_get_cdev_path(device_bdf);
+	struct vfio_pci_device *device;
+	struct iommu *iommu;
+	int ret, i;
+
+	printf("Checking open(%s) fails\n", cdev_path);
+	ret = open(cdev_path, O_RDWR);
+	VFIO_ASSERT_EQ(ret, -1);
+	VFIO_ASSERT_EQ(errno, EBUSY);
+	free((void *)cdev_path);
+
+	for (i = 0; i < nr_iommu_modes; i++) {
+		if (!iommu_modes[i].container_path)
+			continue;
+
+		iommu = iommu_init(iommu_modes[i].name);
+
+		device = vfio_pci_device_alloc(device_bdf, iommu);
+		vfio_pci_group_setup(device);
+		vfio_pci_iommu_setup(device);
+
+		printf("Checking ioctl(group_fd, VFIO_GROUP_GET_DEVICE_FD, \"%s\") fails (%s)\n",
+		       device_bdf, iommu_modes[i].name);
+
+		ret = ioctl(device->group_fd, VFIO_GROUP_GET_DEVICE_FD, device->bdf);
+		VFIO_ASSERT_EQ(ret, -1);
+		VFIO_ASSERT_EQ(errno, EBUSY);
+
+		close(device->group_fd);
+		free(device);
+		iommu_cleanup(iommu);
+	}
+}
+
 static void after_kexec(int luo_fd, int state_session_fd)
 {
 	struct vfio_pci_device *device;
@@ -44,6 +80,8 @@ static void after_kexec(int luo_fd, int state_session_fd)
 	int device_fd;
 	int stage;
 
+	check_open_vfio_device_fails();
+
 	restore_and_read_stage(state_session_fd, STATE_TOKEN, &stage);
 	VFIO_ASSERT_EQ(stage, 2);
 
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
author maintainer
#46 lore raw

PATCH 21/21 vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

In reply to David Matlack

Add a long-running DMA memcpy operation to vfio_pci_liveupdate_kexec_test so that the device attempts to perform DMAs continuously during the Live Update.

At this point iommufd preservation is not supported and bus mastering is not kept enabled on the device during across the kexec, so most of these DMAs will be dropped. However this test ensures that the current device preservation support does not lead to system instability or crashes if the device is active. And once iommufd and bus mastering are preserved, this test can be relaxed to check that the DMA operations completed successfully.

---
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 135 ++++++++++++++++++
 1 file changed, 135 insertions(+)
tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c +135 -0
index 925c5fc30d56..3af50110a65d 100644
@@ -1,8 +1,16 @@
 // SPDX-License-Identifier: GPL-2.0-only
 
+#include <linux/sizes.h>
+#include <sys/mman.h>
+
 #include <libliveupdate.h>
 #include <libvfio.h>
 
+#define MEMCPY_SIZE SZ_1G
+#define DRIVER_SIZE SZ_1M
+#define MEMFD_SIZE (MEMCPY_SIZE + DRIVER_SIZE)
+
+static struct dma_region memcpy_region;
 static const char *device_bdf;
 
 #define STATE_SESSION "567e1b8d-daf1-4a4a-ac13-35d6d275f646"
@@ -11,8 +19,100 @@ static const char *device_bdf;
 enum {
 	STATE_TOKEN,
 	DEVICE_TOKEN,
+	MEMFD_TOKEN,
 };
 
+static void dma_memcpy_one(struct vfio_pci_device *device)
+{
+	void *src = memcpy_region.vaddr, *dst;
+	u64 size;
+
+	size = min_t(u64, memcpy_region.size / 2, device->driver.max_memcpy_size);
+	dst = src + size;
+
+	memset(src, 1, size);
+	memset(dst, 0, size);
+
+	printf("Kicking off 1 DMA memcpy operations of size 0x%lx...\n", size);
+	vfio_pci_driver_memcpy(device,
+			       to_iova(device, src),
+			       to_iova(device, dst),
+			       size);
+
+	VFIO_ASSERT_EQ(memcmp(src, dst, size), 0);
+}
+
+static void dma_memcpy_start(struct vfio_pci_device *device)
+{
+	void *src = memcpy_region.vaddr, *dst;
+	u64 count, size;
+
+	size = min_t(u64, memcpy_region.size / 2, device->driver.max_memcpy_size);
+	dst = src + size;
+
+	/*
+	 * Rough Math: If we assume the device will perform memcpy at a rate of
+	 * 30GB/s then 7200GB of transfers will run for about 4 minutes.
+	 */
+	count = (u64)7200 * SZ_1G / size;
+	count = min_t(u64, count, device->driver.max_memcpy_count);
+
+	memset(src, 1, size / 2);
+	memset(dst, 0, size / 2);
+
+	printf("Kicking off %lu DMA memcpy operations of size 0x%lx...\n", count, size);
+	vfio_pci_driver_memcpy_start(device,
+				     to_iova(device, src),
+				     to_iova(device, dst),
+				     size, count);
+}
+
+static void dma_memfd_map(struct vfio_pci_device *device, int fd)
+{
+	void *vaddr;
+
+	vaddr = mmap(NULL, MEMFD_SIZE, PROT_WRITE, MAP_SHARED, fd, 0);
+	VFIO_ASSERT_NE(vaddr, MAP_FAILED);
+
+	memcpy_region.iova = SZ_4G;
+	memcpy_region.size = MEMCPY_SIZE;
+	memcpy_region.vaddr = vaddr;
+	iommu_map(device->iommu, &memcpy_region);
+
+	device->driver.region.iova = memcpy_region.iova + memcpy_region.size;
+	device->driver.region.size = DRIVER_SIZE;
+	device->driver.region.vaddr = vaddr + memcpy_region.size;
+	iommu_map(device->iommu, &device->driver.region);
+}
+
+static void dma_memfd_setup(struct vfio_pci_device *device, int session_fd)
+{
+	int fd, ret;
+
+	fd = memfd_create("dma-buffer", 0);
+	VFIO_ASSERT_GE(fd, 0);
+
+	ret = fallocate(fd, 0, 0, MEMFD_SIZE);
+	VFIO_ASSERT_EQ(ret, 0);
+
+	printf("Preserving memfd of size 0x%x in session\n", MEMFD_SIZE);
+	ret = luo_session_preserve_fd(session_fd, fd, MEMFD_TOKEN);
+	VFIO_ASSERT_EQ(ret, 0);
+
+	dma_memfd_map(device, fd);
+}
+
+static void dma_memfd_restore(struct vfio_pci_device *device, int session_fd)
+{
+	int fd;
+
+	printf("Retrieving memfd from LUO\n");
+	fd = luo_session_retrieve_fd(session_fd, MEMFD_TOKEN);
+	VFIO_ASSERT_GE(fd, 0);
+
+	dma_memfd_map(device, fd);
+}
+
 static void before_kexec(int luo_fd)
 {
 	struct vfio_pci_device *device;
@@ -32,6 +132,27 @@ static void before_kexec(int luo_fd)
 	ret = luo_session_preserve_fd(session_fd, device->fd, DEVICE_TOKEN);
 	VFIO_ASSERT_EQ(ret, 0);
 
+	dma_memfd_setup(device, session_fd);
+
+	/*
+	 * If the device has a selftests driver, kick off a long-running DMA
+	 * operation to exercise the device trying to DMA during a Live Update.
+	 * Since iommufd preservation is not supported yet, these DMAs should be
+	 * dropped. So this is just looking to verify that the system does not
+	 * fall over and crash as a result of a busy device being preserved.
+	 */
+	if (device->driver.ops) {
+		vfio_pci_driver_init(device);
+		dma_memcpy_start(device);
+
+		/*
+		 * Disable interrupts on the device or freeze() will fail.
+		 * Unfortunately there isn't a way to easily have a test for
+		 * that here since the check happens during shutdown.
+		 */
+		vfio_pci_msix_disable(device);
+	}
+
 	close(luo_fd);
 	daemonize_and_wait();
 }
@@ -106,9 +227,23 @@ static void after_kexec(int luo_fd, int state_session_fd)
 	 */
 	device = __vfio_pci_device_init(device_bdf, iommu, device_fd);
 
+	dma_memfd_restore(device, session_fd);
+
 	printf("Finishing the session\n");
 	VFIO_ASSERT_EQ(luo_session_finish(session_fd), 0);
 
+	/*
+	 * Once iommufd preservation is supported and the device is kept fully
+	 * running across the Live Update, this should wait for the long-
+	 * running DMA memcpy operation kicked off in before_kexec() to
+	 * complete. But for now we expect the device to be reset so just
+	 * trigger a single memcpy to make sure it's still functional.
+	 */
+	if (device->driver.ops) {
+		vfio_pci_driver_init(device);
+		dma_memcpy_one(device);
+	}
+
 	vfio_pci_device_cleanup(device);
 	iommu_cleanup(iommu);
 }
Signature · 1 line
2.52.0.487.g5c8c507ade-goog
#47 lore raw
In reply to David Matlack
Quoted 55 lines · David Matlack

This series adds the base support to preserve a VFIO device file across a Live Update. "Base support" means that this allows userspace to safetly preserve a VFIO device file with LIVEUPDATE_SESSION_PRESERVE_FD and retrieve a preserved VFIO device file with LIVEUPDATE_SESSION_RETRIEVE_FD, but the device itself is not preserved in a fully running state across Live Update.

This series unblocks 2 parallel but related streams of work:

- iommufd preservation across Live Update. This work spans iommufd, the IOMMU subsystem, and IOMMU drivers [1]

- Preservation of VFIO device state across Live Update (config space, BAR addresses, power state, SR-IOV state, etc.). This work spans both VFIO and the core PCI subsystem.

While we need all of the above to fully preserve a VFIO device across a Live Update without disrupting the workload on the device, this series aims to be functional and safe enough to merge as the first incremental step toward that goal.

Areas for Discussion
--------------------

BDF Stability across Live Update

The PCI support for tracking preserved devices across a Live Update to prevent auto-probing relies on PCI segment numbers and BDFs remaining stable. For now I have disallowed VFs, as the BDFs assigned to VFs can vary depending on how the kernel chooses to allocate bus numbers. For non-VFs I am wondering if there is any more needed to ensure BDF stability across Live Update.

While we would like to support many different systems and configurations in due time (including preserving VFs), I'd like to keep this first serses constrained to simple use-cases.

FLB Locking

I don't see a way to properly synchronize pci_flb_finish() with pci_liveupdate_incoming_is_preserved() since the incoming FLB mutex is dropped by liveupdate_flb_get_incoming() when it returns the pointer to the object, and taking pci_flb_incoming_lock in pci_flb_finish() could result in a deadlock due to reversing the lock ordering.

FLB Retrieving

The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

Testing
-------

Quoted 6 lines

The patches at the end of this series provide comprehensive selftests for the new code added by this series. The selftests have been validated in both a VM environment using a virtio-net PCIe device, and in a baremetal environment on an Intel EMR server with an Intel DSA device.

Here is an example of how to run the new selftests:

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio-pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio-pci.ko] undefined! make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1
make[3]: *** [Makefile:1960: modpost] Error 2

After I git clone the source code from the link https://github.com/dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

Thanks,

Yanjun.Zhu

Quoted 161 lines

vfio_pci_liveupdate_uapi_test:

   $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
   $ tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test 0000:00:04.0
   $ tools/testing/selftests/vfio/scripts/cleanup.sh

vfio_pci_liveupdate_kexec_test:

   $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
   $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test --stage 1 0000:00:04.0
   $ kexec [...]  # NOTE: distro-dependent
   $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
   $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test --stage 2 0000:00:04.0
   $ tools/testing/selftests/vfio/scripts/cleanup.sh

Dependencies
------------

This series was constructed on top of several in-flight series and on top of mm-nonmm-unstable [2].

   +-- This series
   |
   +-- [PATCH v2 00/18] vfio: selftests: Support for multi-device tests
   |    https://lore.kernel.org/kvm/[email protected]/
   |
   +-- [PATCH v3 0/4] vfio: selftests: update DMA mapping tests to use queried IOVA ranges
   |   https://lore.kernel.org/kvm/[email protected]/
   |
   +-- [PATCH v8 0/2] Live Update: File-Lifecycle-Bound (FLB) State
   |   https://lore.kernel.org/linux-mm/[email protected]/
   |
   +-- [PATCH v8 00/18] Live Update Orchestrator
   |   https://lore.kernel.org/linux-mm/[email protected]/
   |

To simplify checking out the code, this series can be found on GitHub:

https://github.com/dmatlack/linux/tree/liveupdate/vfio/cdev/v1

Changelog
---------

v1:
- Rebase series on top of LUOv8 and VFIO selftests improvements
- Drop commits to preserve config space fields across Live Update. These changes require changes to the PCI layer. For exmaple, preserving rbars could lead to an inconsistent device state until device BARs addresses are preserved across Live Update.
- Drop commits to preserve Bus Master Enable on the device. There's no reason to preserve this until iommufd preservation is fully working. Furthermore, preserving Bus Master Enable could lead to memory corruption when the device if the device is bound to the default identity-map domain after Live Update.
- Drop commits to preserve saved PCI state. This work is not needed until we are ready to preserve the device's config space, and requires more thought to make the PCI state data layout ABI-friendly.
- Add support to skip auto-probing devices that are preserved by VFIO to avoid them getting bound to a different driver by the next kernel.
- Restrict device preservation further (no VFs, no intel-graphics).
- Various refactoring and small edits to improve readability and eliminate code duplication.

rfc: https://lore.kernel.org/kvm/[email protected]/

Cc Saeed Mahameed <[email protected]>
Cc Adithya Jayachandran <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Parav Pandit <[email protected]>
Cc Leon Romanovsky <[email protected]>
Cc William Tu <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Lukas Wunner <[email protected]>
Cc Pasha Tatashin <[email protected]>
Cc Mike Rapoport <[email protected]>
Cc Pratyush Yadav <[email protected]>
Cc Samiullah Khawaja <[email protected]>
Cc Chris Li <[email protected]>
Cc Josh Hilke <[email protected]>
Cc David Rientjes <[email protected]>

[1] https://lore.kernel.org/linux-iommu/[email protected]/
[2] https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git/log/?h=mm-nonmm-unstable

David Matlack (12):
liveupdate: luo_flb: Prevent retrieve() after finish()
PCI: Add API to track PCI devices preserved across Live Update
PCI: Require driver_override for incoming Live Update preserved
devices
vfio/pci: Notify PCI subsystem about devices preserved across Live
Update
vfio: Enforce preserved devices are retrieved via
LIVEUPDATE_SESSION_RETRIEVE_FD
vfio/pci: Store Live Update state in struct vfio_pci_core_device
vfio: selftests: Add Makefile support for TEST_GEN_PROGS_EXTENDED
vfio: selftests: Add vfio_pci_liveupdate_uapi_test
vfio: selftests: Expose iommu_modes to tests
vfio: selftests: Expose low-level helper routines for setting up
struct vfio_pci_device
vfio: selftests: Verify that opening VFIO device fails during Live
Update
vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

Vipin Sharma (9):
vfio/pci: Register a file handler with Live Update Orchestrator vfio/pci: Preserve vfio-pci device files across Live Update vfio/pci: Retrieve preserved device files after Live Update vfio/pci: Skip reset of preserved device after Live Update selftests/liveupdate: Move luo_test_utils.* into a reusable library selftests/liveupdate: Add helpers to preserve/retrieve FDs
vfio: selftests: Build liveupdate library in VFIO selftests
vfio: selftests: Initialize vfio_pci_device using a VFIO cdev FD
vfio: selftests: Add vfio_pci_liveupdate_kexec_test

  MAINTAINERS                                   |   1 +
  drivers/pci/Makefile                          |   1 +
  drivers/pci/liveupdate.c                      | 248 ++++++++++++++++
  drivers/pci/pci-driver.c                      |  12 +-
  drivers/vfio/device_cdev.c                    |  25 +-
  drivers/vfio/group.c                          |   9 +
  drivers/vfio/pci/Makefile                     |   1 +
  drivers/vfio/pci/vfio_pci.c                   |  11 +-
  drivers/vfio/pci/vfio_pci_core.c              |  23 +-
  drivers/vfio/pci/vfio_pci_liveupdate.c        | 278 ++++++++++++++++++
  drivers/vfio/pci/vfio_pci_priv.h              |  16 +
  drivers/vfio/vfio.h                           |  13 -
  drivers/vfio/vfio_main.c                      |  22 +-
  include/linux/kho/abi/pci.h                   |  53 ++++
  include/linux/kho/abi/vfio_pci.h              |  45 +++
  include/linux/liveupdate.h                    |   3 +
  include/linux/pci.h                           |  38 +++
  include/linux/vfio.h                          |  51 ++++
  include/linux/vfio_pci_core.h                 |   7 +
  kernel/liveupdate/luo_flb.c                   |   4 +
  tools/testing/selftests/liveupdate/.gitignore |   1 +
  tools/testing/selftests/liveupdate/Makefile   |  14 +-
  .../include/libliveupdate.h}                  |  11 +-
  .../selftests/liveupdate/lib/libliveupdate.mk |  20 ++
  .../{luo_test_utils.c => lib/liveupdate.c}    |  43 ++-
  .../selftests/liveupdate/luo_kexec_simple.c   |   2 +-
  .../selftests/liveupdate/luo_multi_session.c  |   2 +-
  tools/testing/selftests/vfio/Makefile         |  23 +-
  .../vfio/lib/include/libvfio/iommu.h          |   2 +
  .../lib/include/libvfio/vfio_pci_device.h     |   8 +
  tools/testing/selftests/vfio/lib/iommu.c      |   4 +-
  .../selftests/vfio/lib/vfio_pci_device.c      |  60 +++-
  .../vfio/vfio_pci_liveupdate_kexec_test.c     | 255 ++++++++++++++++
  .../vfio/vfio_pci_liveupdate_uapi_test.c      |  93 ++++++
  34 files changed, 1313 insertions(+), 86 deletions(-)

create mode 100644 drivers/pci/liveupdate.c
create mode 100644 drivers/vfio/pci/vfio_pci_liveupdate.c
create mode 100644 include/linux/kho/abi/pci.h
create mode 100644 include/linux/kho/abi/vfio_pci.h
rename tools/testing/selftests/liveupdate/{luo_test_utils.h => lib/include/libliveupdate.h} (80%) create mode 100644 tools/testing/selftests/liveupdate/lib/libliveupdate.mk
rename tools/testing/selftests/liveupdate/{luo_test_utils.c => lib/liveupdate.c} (89%) create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c

Signature · 2 lines
Best Regards,
Yanjun.Zhu
#48 lore raw
Quoted 78 lines · Zhu Yanjun
Quoted 55 lines · David Matlack

This series adds the base support to preserve a VFIO device file across a Live Update. "Base support" means that this allows userspace to safetly preserve a VFIO device file with LIVEUPDATE_SESSION_PRESERVE_FD and retrieve a preserved VFIO device file with LIVEUPDATE_SESSION_RETRIEVE_FD, but the device itself is not preserved in a fully running state across Live Update.

This series unblocks 2 parallel but related streams of work:

  - iommufd preservation across Live Update. This work spans iommufd, the IOMMU subsystem, and IOMMU drivers [1]

  - Preservation of VFIO device state across Live Update (config space, BAR addresses, power state, SR-IOV state, etc.). This work spans both VFIO and the core PCI subsystem.

While we need all of the above to fully preserve a VFIO device across a Live Update without disrupting the workload on the device, this series aims to be functional and safe enough to merge as the first incremental step toward that goal.

Areas for Discussion
--------------------

BDF Stability across Live Update

   The PCI support for tracking preserved devices across a Live Update to prevent auto-probing relies on PCI segment numbers and BDFs remaining stable. For now I have disallowed VFs, as the BDFs assigned to VFs can vary depending on how the kernel chooses to allocate bus numbers. For non-VFs I am wondering if there is any more needed to ensure BDF stability across Live Update.

   While we would like to support many different systems and configurations in due time (including preserving VFs), I'd like to keep this first serses constrained to simple use-cases.

FLB Locking

   I don't see a way to properly synchronize pci_flb_finish() with pci_liveupdate_incoming_is_preserved() since the incoming FLB mutex is dropped by liveupdate_flb_get_incoming() when it returns the pointer to the object, and taking pci_flb_incoming_lock in pci_flb_finish() could result in a deadlock due to reversing the lock ordering.

FLB Retrieving

   The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

Testing
-------

Quoted 6 lines

The patches at the end of this series provide comprehensive selftests for the new code added by this series. The selftests have been validated in both a VM environment using a virtio-net PCIe device, and in a baremetal environment on an Intel EMR server with an Intel DSA device.

Here is an example of how to run the new selftests:

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined! make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1
make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 7 lines

After I git clone the source code from the link https://github.com/ dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Yanjun.Zhu

Quoted 166 lines

Thanks,

Yanjun.Zhu

Quoted 161 lines

vfio_pci_liveupdate_uapi_test:

   $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
   $ tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test 0000:00:04.0
   $ tools/testing/selftests/vfio/scripts/cleanup.sh

vfio_pci_liveupdate_kexec_test:

   $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
   $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test -- stage 1 0000:00:04.0
   $ kexec [...]  # NOTE: distro-dependent
   $ tools/testing/selftests/vfio/scripts/setup.sh 0000:00:04.0
   $ tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test -- stage 2 0000:00:04.0
   $ tools/testing/selftests/vfio/scripts/cleanup.sh

Dependencies
------------

This series was constructed on top of several in-flight series and on top of mm-nonmm-unstable [2].

   +-- This series
   |
   +-- [PATCH v2 00/18] vfio: selftests: Support for multi-device tests
   |    https://lore.kernel.org/kvm/20251112192232.442761-1- [email protected]/
   |
   +-- [PATCH v3 0/4] vfio: selftests: update DMA mapping tests to use queried IOVA ranges
   |   https://lore.kernel.org/kvm/20251111-iova-ranges- [email protected]/
   |
   +-- [PATCH v8 0/2] Live Update: File-Lifecycle-Bound (FLB) State
   |   https://lore.kernel.org/linux-mm/20251125225006.3722394-1- [email protected]/
   |
   +-- [PATCH v8 00/18] Live Update Orchestrator
   |   https://lore.kernel.org/linux-mm/20251125165850.3389713-1- [email protected]/
   |

To simplify checking out the code, this series can be found on GitHub:

   https://github.com/dmatlack/linux/tree/liveupdate/vfio/cdev/v1

Changelog
---------

v1:
  - Rebase series on top of LUOv8 and VFIO selftests improvements
  - Drop commits to preserve config space fields across Live Update. These changes require changes to the PCI layer. For exmaple, preserving rbars could lead to an inconsistent device state until device BARs addresses are preserved across Live Update.
  - Drop commits to preserve Bus Master Enable on the device. There's no reason to preserve this until iommufd preservation is fully working. Furthermore, preserving Bus Master Enable could lead to memory corruption when the device if the device is bound to the default identity-map domain after Live Update.
  - Drop commits to preserve saved PCI state. This work is not needed until we are ready to preserve the device's config space, and requires more thought to make the PCI state data layout ABI-friendly.
  - Add support to skip auto-probing devices that are preserved by VFIO to avoid them getting bound to a different driver by the next kernel.
  - Restrict device preservation further (no VFs, no intel-graphics).
  - Various refactoring and small edits to improve readability and eliminate code duplication.

rfc: https://lore.kernel.org/kvm/20251018000713.677779-1- [email protected]/

Cc Saeed Mahameed <[email protected]>
Cc Adithya Jayachandran <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Parav Pandit <[email protected]>
Cc Leon Romanovsky <[email protected]>
Cc William Tu <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Lukas Wunner <[email protected]>
Cc Pasha Tatashin <[email protected]>
Cc Mike Rapoport <[email protected]>
Cc Pratyush Yadav <[email protected]>
Cc Samiullah Khawaja <[email protected]>
Cc Chris Li <[email protected]>
Cc Josh Hilke <[email protected]>
Cc David Rientjes <[email protected]>

[1] https://lore.kernel.org/linux-iommu/20250928190624.3735830-1- [email protected]/
[2] https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git/log/? h=mm-nonmm-unstable

David Matlack (12):
   liveupdate: luo_flb: Prevent retrieve() after finish()
   PCI: Add API to track PCI devices preserved across Live Update
   PCI: Require driver_override for incoming Live Update preserved devices
   vfio/pci: Notify PCI subsystem about devices preserved across Live Update
   vfio: Enforce preserved devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD
   vfio/pci: Store Live Update state in struct vfio_pci_core_device
   vfio: selftests: Add Makefile support for TEST_GEN_PROGS_EXTENDED
   vfio: selftests: Add vfio_pci_liveupdate_uapi_test
   vfio: selftests: Expose iommu_modes to tests
   vfio: selftests: Expose low-level helper routines for setting up struct vfio_pci_device
   vfio: selftests: Verify that opening VFIO device fails during Live Update
   vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

Vipin Sharma (9):
   vfio/pci: Register a file handler with Live Update Orchestrator vfio/pci: Preserve vfio-pci device files across Live Update vfio/pci: Retrieve preserved device files after Live Update vfio/pci: Skip reset of preserved device after Live Update selftests/liveupdate: Move luo_test_utils.* into a reusable library selftests/liveupdate: Add helpers to preserve/retrieve FDs
   vfio: selftests: Build liveupdate library in VFIO selftests
   vfio: selftests: Initialize vfio_pci_device using a VFIO cdev FD
   vfio: selftests: Add vfio_pci_liveupdate_kexec_test

  MAINTAINERS                                   |   1 + drivers/pci/Makefile                          |   1 + drivers/pci/liveupdate.c                      | 248 ++++++++++++++++ drivers/pci/pci-driver.c                      |  12 +- drivers/vfio/device_cdev.c                    |  25 +- drivers/vfio/group.c                          |   9 + drivers/vfio/pci/Makefile                     |   1 + drivers/vfio/pci/vfio_pci.c                   |  11 +- drivers/vfio/pci/vfio_pci_core.c              |  23 +- drivers/vfio/pci/vfio_pci_liveupdate.c        | 278 ++++++++++++++++++ drivers/vfio/pci/vfio_pci_priv.h              |  16 + drivers/vfio/vfio.h                           |  13 - drivers/vfio/vfio_main.c                      |  22 +- include/linux/kho/abi/pci.h                   |  53 ++++ include/linux/kho/abi/vfio_pci.h              |  45 +++ include/linux/liveupdate.h                    |   3 + include/linux/pci.h                           |  38 +++ include/linux/vfio.h                          |  51 ++++ include/linux/vfio_pci_core.h                 |   7 + kernel/liveupdate/luo_flb.c                   |   4 + tools/testing/selftests/liveupdate/.gitignore |   1 + tools/testing/selftests/liveupdate/Makefile   |  14 +- .../include/libliveupdate.h}                  |  11 +- .../selftests/liveupdate/lib/libliveupdate.mk |  20 ++ .../{luo_test_utils.c => lib/liveupdate.c}    |  43 ++- .../selftests/liveupdate/luo_kexec_simple.c   |   2 +- .../selftests/liveupdate/luo_multi_session.c  |   2 +- tools/testing/selftests/vfio/Makefile         |  23 +- .../vfio/lib/include/libvfio/iommu.h          |   2 + .../lib/include/libvfio/vfio_pci_device.h     |   8 + tools/testing/selftests/vfio/lib/iommu.c      |   4 +- .../selftests/vfio/lib/vfio_pci_device.c      |  60 +++- .../vfio/vfio_pci_liveupdate_kexec_test.c     | 255 ++++++++++++++++ .../vfio/vfio_pci_liveupdate_uapi_test.c      |  93 ++++++

  34 files changed, 1313 insertions(+), 86 deletions(-)

  create mode 100644 drivers/pci/liveupdate.c
  create mode 100644 drivers/vfio/pci/vfio_pci_liveupdate.c
  create mode 100644 include/linux/kho/abi/pci.h
  create mode 100644 include/linux/kho/abi/vfio_pci.h
  rename tools/testing/selftests/liveupdate/{luo_test_utils.h => lib/ include/libliveupdate.h} (80%) create mode 100644 tools/testing/selftests/liveupdate/lib/ libliveupdate.mk
  rename tools/testing/selftests/liveupdate/{luo_test_utils.c => lib/ liveupdate.c} (89%) create mode 100644 tools/testing/selftests/vfio/ vfio_pci_liveupdate_kexec_test.c create mode 100644 tools/testing/selftests/vfio/ vfio_pci_liveupdate_uapi_test.c

Signature · 2 lines
Best Regards,
Yanjun.Zhu
author maintainer
#49 lore raw
Quoted 32 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

Quoted 7 lines

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

#50 lore raw
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Yanjun.Zhu

Signature · 2 lines
Best Regards,
Yanjun.Zhu
author maintainer
#51 lore raw
Quoted 39 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 9 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

drivers/vfio/pci/Makefile +0 -0
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
 vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
 vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
 vfio-pci-y := vfio_pci.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
 obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
        .show_fdinfo    = vfio_device_show_fdinfo,
 #endif
 };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
 /**
  * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
 {
        return luo_global.enabled;
 }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
 /**
  * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +1 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
        luo_session_resume();
        return err;
 }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
author maintainer
#52 lore raw
Quoted 50 lines · David Matlack
Quoted 39 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 9 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Quoted 7 lines

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

Oops, sorry, something went wrong when I posted that diff. Here's the correct diff:

drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
 vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
 vfio-pci-y := vfio_pci.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
 obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
        .show_fdinfo    = vfio_device_show_fdinfo,
 #endif
 };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
 /**
  * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
 {
        return luo_global.enabled;
 }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
 /**
  * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +2 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
        luo_session_resume();
        return err;
 }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
 /**
  * liveupdate_unregister_file_handler - Unregister a liveupdate file handler
@@ -913,3 +914,4 @@ int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
        liveupdate_test_register(fh);
        return err;
 }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);
maintainer
#53 lore raw
Quoted 117 lines · David Matlack
Quoted 50 lines · David Matlack
Quoted 39 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 9 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Quoted 7 lines

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

Oops, sorry, something went wrong when I posted that diff. Here's the correct diff:

drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
 vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
 vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
 vfio-pci-y := vfio_pci.o
 vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
 obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
        .show_fdinfo    = vfio_device_show_fdinfo,
 #endif
 };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
 /**
  * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
 {
        return luo_global.enabled;
 }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
 /**
  * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +1 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
        luo_session_resume();
        return err;
 }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
 /**
  * liveupdate_unregister_file_handler - Unregister a liveupdate file handler

kernel/liveupdate/luo_file.c int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh) · line 917

Quoted 4 lines
kernel/liveupdate/luo_file.c +1 -0
@@ -913,3 +914,4 @@ int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
        liveupdate_test_register(fh);
        return err;
 }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);

Please add the EXPORT_SYMBOL_GPL() to the public LUO API as needed, I was asked to stay conservative and not export until it is actually needed by modules.

#54 lore raw
Quoted 110 lines · Pasha Tatashin
Quoted 109 lines · David Matlack
Quoted 47 lines · David Matlack
Quoted 38 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Quoted 7 lines

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

Oops, sorry, something went wrong when I posted that diff. Here's the correct diff:

drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
  vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
  vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
  obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
  vfio-pci-y := vfio_pci.o
  vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
  obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
  obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
         .show_fdinfo    = vfio_device_show_fdinfo,
  #endif
  };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
  /**
   * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
  {
         return luo_global.enabled;
  }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
  /**
   * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +1 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
         luo_session_resume();
         return err;
  }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
Quoted 7 lines

kernel/liveupdate/luo_file.c int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh) · line 917

Quoted 7 lines
kernel/liveupdate/luo_file.c +1 -0
  /**
   * liveupdate_unregister_file_handler - Unregister a liveupdate file handler
@@ -913,3 +914,4 @@ int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
         liveupdate_test_register(fh);
         return err;
  }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);

Sure. Exactly. The above is the same with my solution. But after that EXPORT_SYMBOL_GPL, a cyclic dependency issue will occur.

Yanjun.Zhu

Quoted 3 lines

Please add the EXPORT_SYMBOL_GPL() to the public LUO API as needed, I was asked to stay conservative and not export until it is actually needed by modules.

author maintainer
#55 lore raw
Quoted 113 lines · Yanjun.Zhu
Quoted 110 lines · Pasha Tatashin
Quoted 109 lines · David Matlack
Quoted 47 lines · David Matlack
Quoted 38 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Quoted 7 lines

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

Oops, sorry, something went wrong when I posted that diff. Here's the correct diff:

drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
  vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
  vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
  obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
  vfio-pci-y := vfio_pci.o
  vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
  obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
  obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
         .show_fdinfo    = vfio_device_show_fdinfo,
  #endif
  };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
  /**
   * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
  {
         return luo_global.enabled;
  }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
  /**
   * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +1 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
         luo_session_resume();
         return err;
  }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
Quoted 11 lines
Quoted 7 lines

kernel/liveupdate/luo_file.c int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh) · line 917

Quoted 7 lines
kernel/liveupdate/luo_file.c +1 -0
  /**
   * liveupdate_unregister_file_handler - Unregister a liveupdate file handler
@@ -913,3 +914,4 @@ int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
         liveupdate_test_register(fh);
         return err;
  }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);

Sure. Exactly. The above is the same with my solution. But after that EXPORT_SYMBOL_GPL, a cyclic dependency issue will occur.

Did see the change to drivers/vfio/pci/Makefile? That fixes the circular dependency between vfio-pci and vfio-pci-core that I created in this series.

#56 lore raw
Quoted 122 lines · David Matlack
Quoted 111 lines · Yanjun.Zhu
Quoted 109 lines · Pasha Tatashin
Quoted 108 lines · David Matlack
Quoted 46 lines · David Matlack
Quoted 38 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Quoted 7 lines

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

Oops, sorry, something went wrong when I posted that diff. Here's the correct diff:

drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
   vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
   vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
   obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
   vfio-pci-y := vfio_pci.o
   vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
   obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
   obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
          .show_fdinfo    = vfio_device_show_fdinfo,
   #endif
   };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
   /**
    * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
   {
          return luo_global.enabled;
   }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
   /**
    * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +1 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
          luo_session_resume();
          return err;
   }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
Quoted 10 lines
Quoted 7 lines

kernel/liveupdate/luo_file.c int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh) · line 917

Quoted 7 lines
kernel/liveupdate/luo_file.c +1 -0
   /**
    * liveupdate_unregister_file_handler - Unregister a liveupdate file handler
@@ -913,3 +914,4 @@ int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
          liveupdate_test_register(fh);
          return err;
   }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);

Sure. Exactly. The above is the same with my solution. But after that EXPORT_SYMBOL_GPL, a cyclic dependency issue will occur.

Quoted 1 line

Did see the change to drivers/vfio/pci/Makefile? That fixes the

Ah, I overlooked this change. I'll conduct a test.

Thanks,

Yanjun.Zhu

Quoted 2 lines

circular dependency between vfio-pci and vfio-pci-core that I created in this series.

#57 lore raw
In reply to David Matlack
Quoted 112 lines · David Matlack
Quoted 47 lines · David Matlack
Quoted 38 lines · Zhu Yanjun
Quoted 37 lines · David Matlack
Quoted 36 lines · Zhu Yanjun
Quoted 22 lines · Zhu Yanjun

Hi, David

ERROR: modpost: "liveupdate_register_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!

ERROR: modpost: "vfio_pci_ops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_enabled" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "liveupdate_unregister_file_handler" [drivers/vfio/pci/ vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_device_fops" [drivers/vfio/pci/vfio-pci-core.ko] undefined!
ERROR: modpost: "vfio_pci_is_intel_display" [drivers/vfio/pci/vfio-pci- core.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_init" [drivers/vfio/pci/vfio- pci.ko] undefined!
ERROR: modpost: "vfio_pci_liveupdate_cleanup" [drivers/vfio/pci/vfio- pci.ko] undefined!
make[4]: *** [scripts/Makefile.modpost:147: Module.symvers] Error 1 make[3]: *** [Makefile:1960: modpost] Error 2

Quoted 8 lines

After I git clone the source code from the link https://github.com/
dmatlack/linux/tree/liveupdate/vfio/cdev/v1,

I found the above errors when I built the source code.

Perhaps the above errors can be solved by EXPORT_SYMBOL.

But I am not sure if a better solution can solve the above problems or not.

I reviewed this patch series in detail. If I’m understanding it correctly, there appears to be a cyclic dependency issue. Specifically, some functions in kernel module A depend on kernel module B, while at the same time certain functions in module B depend on module A.

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines

I’m not entirely sure whether this constitutes a real problem or if it’s intentional design.

Thanks for your report. Can you share the .config file you used to generate these errors?

IIRC, I used FC 42 default config. Perhaps you can make tests with it. If this problem can not be reproduced, I will share my config with you.

Quoted 7 lines

What does "FC 42 default config" mean?

Either way I was able to reproduce the errors you posted above by changing CONFIG_VFIO_PCI{_CORE} from "y" to "m".

To unblock building and testing this series you can change these configs from "m" to "y", or the following patch (which fixed things for me):

Oops, sorry, something went wrong when I posted that diff. Here's the correct diff:

drivers/vfio/pci/Makefile +1 -1
index 929df22c079b..c2cca16e99a8 100644
@@ -2,11 +2,11 @@
 
  vfio-pci-core-y := vfio_pci_core.o vfio_pci_intrs.o vfio_pci_rdwr.o vfio_pci_config.o
  vfio-pci-core-$(CONFIG_VFIO_PCI_ZDEV_KVM) += vfio_pci_zdev.o
-vfio-pci-core-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
  obj-$(CONFIG_VFIO_PCI_CORE) += vfio-pci-core.o
 
  vfio-pci-y := vfio_pci.o
  vfio-pci-$(CONFIG_VFIO_PCI_IGD) += vfio_pci_igd.o
+vfio-pci-$(CONFIG_LIVEUPDATE) += vfio_pci_liveupdate.o
  obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
  obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
drivers/vfio/vfio_main.c +1 -0
index c5b5eb509474..b9805861763a 100644
@@ -1386,6 +1386,7 @@ const struct file_operations vfio_device_fops = {
         .show_fdinfo    = vfio_device_show_fdinfo,
  #endif
  };
+EXPORT_SYMBOL_GPL(vfio_device_fops);
 
  /**
   * vfio_file_is_valid - True if the file is valid vfio file
kernel/liveupdate/luo_core.c +1 -0
index 69298d82f404..c7a0c9c3b6a8 100644
@@ -256,6 +256,7 @@ bool liveupdate_enabled(void)
  {
         return luo_global.enabled;
  }
+EXPORT_SYMBOL_GPL(liveupdate_enabled);
 
  /**
   * DOC: LUO ioctl Interface
kernel/liveupdate/luo_file.c +1 -0
index fca3806dae28..9baa88966f04 100644
@@ -868,6 +868,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
         luo_session_resume();
         return err;
  }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
  /**
   * liveupdate_unregister_file_handler - Unregister a liveupdate file handler

kernel/liveupdate/luo_file.c int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh) · line 917

Quoted 4 lines
kernel/liveupdate/luo_file.c +1 -0
@@ -913,3 +914,4 @@ int liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
         liveupdate_test_register(fh);
         return err;
  }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);

Hi, David

Thanks. The above changes resolve the EXPORT_SYMBOL_GPL issue as well as the cyclic dependency problem.

I recommend submitting this as a new commit, as it effectively addresses the issues present in the patch series.

Yanjun.Zhu

Signature · 2 lines
Best Regards,
Yanjun.Zhu
maintainer
#58 lore raw
In reply to David Matlack
Quoted 38 lines · David Matlack

This series adds the base support to preserve a VFIO device file across a Live Update. "Base support" means that this allows userspace to safetly preserve a VFIO device file with LIVEUPDATE_SESSION_PRESERVE_FD and retrieve a preserved VFIO device file with LIVEUPDATE_SESSION_RETRIEVE_FD, but the device itself is not preserved in a fully running state across Live Update.

This series unblocks 2 parallel but related streams of work:

- iommufd preservation across Live Update. This work spans iommufd, the IOMMU subsystem, and IOMMU drivers [1]

- Preservation of VFIO device state across Live Update (config space, BAR addresses, power state, SR-IOV state, etc.). This work spans both VFIO and the core PCI subsystem.

While we need all of the above to fully preserve a VFIO device across a Live Update without disrupting the workload on the device, this series aims to be functional and safe enough to merge as the first incremental step toward that goal.

Areas for Discussion
--------------------

BDF Stability across Live Update

The PCI support for tracking preserved devices across a Live Update to prevent auto-probing relies on PCI segment numbers and BDFs remaining stable. For now I have disallowed VFs, as the BDFs assigned to VFs can vary depending on how the kernel chooses to allocate bus numbers. For non-VFs I am wondering if there is any more needed to ensure BDF stability across Live Update.

While we would like to support many different systems and configurations in due time (including preserving VFs), I'd like to keep this first serses constrained to simple use-cases.

Quoted 7 lines

FLB Locking

I don't see a way to properly synchronize pci_flb_finish() with pci_liveupdate_incoming_is_preserved() since the incoming FLB mutex is dropped by liveupdate_flb_get_incoming() when it returns the pointer to the object, and taking pci_flb_incoming_lock in pci_flb_finish() could result in a deadlock due to reversing the lock ordering.

I will re-introduce _lock/_unlock API to solve this issue.

Quoted 7 lines

FLB Retrieving

The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

Thanks, I will include this fix in the next version of FLB.

maintainer
#59 lore raw
Quoted 1 line · Pasha Tatashin

On Wed, Nov 26, 2025 at 2:36 PM David Matlack <[email protected]> wrote:

[...]

Quoted 7 lines
Quoted 7 lines

FLB Locking

I don't see a way to properly synchronize pci_flb_finish() with pci_liveupdate_incoming_is_preserved() since the incoming FLB mutex is dropped by liveupdate_flb_get_incoming() when it returns the pointer to the object, and taking pci_flb_incoming_lock in pci_flb_finish() could result in a deadlock due to reversing the lock ordering.

My mental model for FLB is that it is a dependency for files, so it should always be created (aka prepare) before _any_ of the files, and always destroyed (aka finish) after _all_ of the files.

By the time the FLB is being finished, all the files for that FLB should also be finished, so there should no longer be a user of the FLB.

Once all of the files are finished, it should be LUO's responsibility to make sure liveupdate_flb_get_incoming() returns an error _before_ it starts doing the FLB finish. And in FLB finish you should not be needing to take any locks.

Why do you want to use the FLB when it is being finished?

Quoted 10 lines

I will re-introduce _lock/_unlock API to solve this issue.

Quoted 7 lines

FLB Retrieving

The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

IMO once the FLB is finished, LUO should make sure it cannot be retrieved, mainly so subsystem code is simpler and less bug-prone.

Quoted 2 lines

Thanks, I will include this fix in the next version of FLB.

Signature · 2 lines
Regards,
Pratyush Yadav
author maintainer
#60 lore raw
Quoted 20 lines · Pratyush Yadav
Quoted 1 line · Pasha Tatashin

On Wed, Nov 26, 2025 at 2:36 PM David Matlack <[email protected]> wrote:

[...]

Quoted 7 lines
Quoted 7 lines

FLB Locking

I don't see a way to properly synchronize pci_flb_finish() with pci_liveupdate_incoming_is_preserved() since the incoming FLB mutex is dropped by liveupdate_flb_get_incoming() when it returns the pointer to the object, and taking pci_flb_incoming_lock in pci_flb_finish() could result in a deadlock due to reversing the lock ordering.

My mental model for FLB is that it is a dependency for files, so it should always be created (aka prepare) before _any_ of the files, and always destroyed (aka finish) after _all_ of the files.

By the time the FLB is being finished, all the files for that FLB should also be finished, so there should no longer be a user of the FLB.

Quoted 6 lines

Once all of the files are finished, it should be LUO's responsibility to make sure liveupdate_flb_get_incoming() returns an error _before_ it starts doing the FLB finish. And in FLB finish you should not be needing to take any locks.

Why do you want to use the FLB when it is being finished?

The next patch looks at the PCI FLB anytime a device is probed, which could could race with the last device file getting finished causing the FLB to be freed.

However, it looks like I am going to drop that patch. But the PCI FLB is still used in PATCH 08 [1] whenever userspace opens a VFIO cdev or issues the VFIO_GROUP_GET_DEVICE_FD ioctl to check of the underlying PCI device was preserved. Offline Jason suggested decoupling those checks from the FLB, so I'll look into doing that in the next version.

[1]https://lore.kernel.org/kvm/[email protected]/

Quoted 7 lines
Quoted 7 lines

I will re-introduce _lock/_unlock API to solve this issue.

Quoted 3 lines

FLB Retrieving

Quoted 7 lines
Quoted 4 lines
Quoted 4 lines

The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

IMO once the FLB is finished, LUO should make sure it cannot be retrieved, mainly so subsystem code is simpler and less bug-prone.

+1, and I think Pasha is going to do that in the next version of FLB.

maintainer
#61 lore raw
Quoted 11 lines
Quoted 9 lines
Quoted 6 lines
Quoted 6 lines

FLB Retrieving

The first patch of this series includes a fix to prevent an FLB from being retrieved again it is finished. I am wondering if this is the right approach or if subsystems are expected to stop calling liveupdate_flb_get_incoming() after an FLB is finished.

IMO once the FLB is finished, LUO should make sure it cannot be retrieved, mainly so subsystem code is simpler and less bug-prone.

+1, and I think Pasha is going to do that in the next version of FLB.

Yes, I will add this change in the next version of FLB; however, I will send the next version of FLB only once list_private.h [1] is added to linux-next, so I can replace luo_list_for_each_private() with list_private_for_each_entry().

Pasha

[1] https://lore.kernel.org/all/[email protected]