[PATCH v5 00/20] vfio/pci: Base Live Update support for VFIO

[PATCH v5 00/20] vfio/pci: Base Live Update support for VFIO

[Vipin Sharma]

Hello,

This series can be found on Github:

  https://github.com/shvipin/linux vfio/liveupdate/v5

This series provides the base support to enable Live Update feature for
VFIO PCI devices using the Live Update Orchestrator (LUO) and Kexec
Handover (KHO).

Overview
--------

This series makes the following changes to support VFIO Live Update:

 - Puts VFIO Live Update support behind the CONFIG_VFIO_PCI_LIVEUPDATE
   config option (which requires CONFIG_VFIO_DEVICE_CDEV).
 - Preserves VFIO cdev device FDs across a kexec Live Update session via
   LUO/KHO.
 - Enforces that incoming preserved devices can only be retrieved via
   LIVEUPDATE_SESSION_RETRIEVE_FD post-kexec (preventing legacy
   group/cdev opens from accessing uninitialized state).
 - Quiesces the device prior to kexec in freeze() by resetting hardware,
   disabling bus mastering, and marking the device frozen to reject
   post-freeze IO/MMIO.
 - Includes VFIO selftests (uapi & kexec tests with continuous DMA
   checks) and kernel documentation.

VFIO PCI Live Update Roadmap
----------------------------

We are planning to add VFIO PCI Live Update support in phases. Since full
support of this feature will come in phase, VFIO PCI Liveupdate is
controlled by CONFIG_VFIO_PCI_LIVEUPDATE config option. 

This series is phase 1 below:

 1. PCIe function (non-VF) preserved across Live Update without changing
    BDF
 2. PCIe function (non-VF) preserved across Live Update without resetting
 3. PCIe function (non-VF) preserved across Live Update with continuous
    DMA (noiommu)
 4. PCIe function (non-VF) preserved across Live Update with continuous
    DMA (iommufd)
 5. PCIe VF preserved across Live Update with continuous DMA (iommufd)

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

This series has following dependencies as shown in the diagram

    |---------------------------|
    |         VFIO v5           |  <--- Sent in this series.
    |---------------------------|
    |  Liveupdate Selftests[1]  |
    |---------------------------|
    |        PCI v7 [2]         |
    |---------------------------|

[1] https://lore.kernel.org/all/20260511201155.1488670-1-vipinsh@google.com/
    These selftests changes are needed by other selftests also (iommu,
    kvm), so, I have sent them separately.

[2] https://lore.kernel.org/linux-pci/20260710212616.1351130-1-dmatlack@google.com/
    This is the PCI v7, which adds the live update support in PCI core.

Github branch I have mentioned at the top has all of these dependencies.

Testing
-------

Tested using VFIO Live Update selftests in both QEMU and bare-metal
environment (Intel DSA PCIe device).

  $ ./setup.sh 0000:00:04.0
  $ ./vfio_pci_liveupdate_kexec_test --stage 1 0000:6a:01.0
  $ kexec --reuse-cmdline -l -s /boot/vmlinuz
  $ kexec -e
  $ ./setup.sh 0000:00:04.0
  $ ./vfio_pci_liveupdate_kexec_test --stage 2 0000:6a:01.0

Kernel command line options used:
  - liveupdate=on
  - kho=on
  - vfio_iommu_type1.allow_unsafe_interrupts=Y
  - vfio_pci.enable_sriov=Y
  - vfio_pci.disable_denylist=Y
  - vfio.enable_unsafe_noiommu_mode=Y


Changelog
---------

v5:
 - Addressed Sashiko feedback.
 - Removed condition to disable CONFIG_VFIO_PCI_DMABUF.
 - Expanded vfio_find_device() helper to safely discover a registered
   vfio_device by matching criteria. Internal Sashiko instance found
   data race.
 - Updated selftests to latest changes in VFIO selftests.
 - Rebased on top of PCI Live Update series v7.

v4: https://lore.kernel.org/kvm/20260511234802.2280368-1-vipinsh@google.com/

v3: https://lore.kernel.org/kvm/20260323235817.1960573-1-dmatlack@google.com/

v2: https://lore.kernel.org/kvm/20260129212510.967611-1-dmatlack@google.com/

v1: https://lore.kernel.org/kvm/20251126193608.2678510-1-dmatlack@google.com/

rfc: https://lore.kernel.org/kvm/20251018000713.677779-1-vipinsh@google.com/


David Matlack (7):
  vfio: Enforce preserved devices are retrieved via
    LIVEUPDATE_SESSION_RETRIEVE_FD
  docs: liveupdate: Add documentation for VFIO PCI
  vfio: selftests: Add vfio_pci_liveupdate_uapi_test
  vfio: selftests: Add Makefile support for TEST_GEN_PROGS_EXTENDED
  vfio: selftests: Expose iommu_modes to tests
  vfio: selftests: Verify that opening VFIO device fails during Live
    Update
  vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

Vipin Sharma (13):
  vfio/pci: Factor out the reset logic in VFIO PCI device close path
  vfio: Export various helpers from VFIO
  vfio/pci: Export vfio_pci_dma_buf_move for vfio-pci module
  liveupdate: Export symbols needed by modules
  vfio/pci: Register a file handler with Live Update Orchestrator
  vfio/pci: Preserve vfio-pci device files across Live Update
  vfio: Introduce vfio_find_device() helper
  vfio: Refactor vfio_device_fops_cdev_open()
  vfio: Add API to open cdev device for Live Update restore
  vfio/pci: Retrieve preserved device files after Live Update
  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

 Documentation/core-api/liveupdate.rst         |   1 +
 .../driver-api/vfio_pci_liveupdate.rst        |  23 ++
 MAINTAINERS                                   |   2 +
 drivers/vfio/device_cdev.c                    |  67 +++-
 drivers/vfio/group.c                          |   9 +
 drivers/vfio/pci/Kconfig                      |  12 +
 drivers/vfio/pci/Makefile                     |   1 +
 drivers/vfio/pci/vfio_pci.c                   |  13 +-
 drivers/vfio/pci/vfio_pci_config.c            |   3 +
 drivers/vfio/pci/vfio_pci_core.c              | 119 +++---
 drivers/vfio/pci/vfio_pci_dmabuf.c            |   1 +
 drivers/vfio/pci/vfio_pci_liveupdate.c        | 340 ++++++++++++++++++
 drivers/vfio/pci/vfio_pci_priv.h              |  28 ++
 drivers/vfio/vfio.h                           |  18 +
 drivers/vfio/vfio_main.c                      |  67 +++-
 include/linux/kho/abi/vfio_pci.h              |  43 +++
 include/linux/vfio.h                          |  12 +
 include/linux/vfio_pci_core.h                 |   3 +
 kernel/liveupdate/luo_file.c                  |   2 +
 tools/testing/selftests/vfio/Makefile         |  23 +-
 .../vfio/lib/include/libvfio/iommu.h          |   2 +
 .../lib/include/libvfio/vfio_pci_device.h     |   3 +
 tools/testing/selftests/vfio/lib/iommu.c      |   4 +-
 .../selftests/vfio/lib/vfio_pci_device.c      |  25 +-
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 256 +++++++++++++
 .../vfio/vfio_pci_liveupdate_uapi_test.c      |  97 +++++
 26 files changed, 1097 insertions(+), 77 deletions(-)
 create mode 100644 Documentation/driver-api/vfio_pci_liveupdate.rst
 create mode 100644 drivers/vfio/pci/vfio_pci_liveupdate.c
 create mode 100644 include/linux/kho/abi/vfio_pci.h
 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


base-commit: cbf37fff757e87376e80e11a8378393a8f25145f
prerequisite-patch-id: 85705fb54d3065efe1d87ab4b69e828a9f3404e7
prerequisite-patch-id: a33ae79b5fc0259bd013b430c18cc3d14a678599
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 01/20] vfio/pci: Factor out the reset logic in VFIO PCI device close path

[Vipin Sharma]

vfio_pci_core_disable() is a very long function doing multiple things
like exiting runtime power management, setting device to D0 power,
clearing bus master, clearing irqs, freeing eventfds, releasing device
regions, unmapping PCI bars, and performing device/bus resets.

The upcoming VFIO Liveupdate feature will also need to utilize same
reset logic.

Extract device reset code to a new helper function,
vfio_pci_core_try_reset(), and update vfio_pci_core_disable() to use
this new function.

No functional change intended.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/pci/vfio_pci_core.c | 100 ++++++++++++++++---------------
 1 file changed, 53 insertions(+), 47 deletions(-)

diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index a28f1e99362c..3c8b990ee92f 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -623,9 +623,60 @@ int vfio_pci_core_enable(struct vfio_pci_core_device *vdev)
 }
 EXPORT_SYMBOL_GPL(vfio_pci_core_enable);
 
-void vfio_pci_core_disable(struct vfio_pci_core_device *vdev)
+static void vfio_pci_core_try_reset(struct vfio_pci_core_device *vdev)
 {
+	struct pci_dev *pdev = vdev->pdev;
 	struct pci_dev *bridge;
+
+	vdev->needs_reset = true;
+
+	/*
+	 * If we have saved state, restore it.  If we can reset the device,
+	 * even better.  Resetting with current state seems better than
+	 * nothing, but saving and restoring current state without reset
+	 * is just busy work.
+	 */
+	if (pci_load_and_free_saved_state(pdev, &vdev->pci_saved_state)) {
+		pci_info(pdev, "%s: Couldn't reload saved state\n", __func__);
+
+		if (!vdev->reset_works)
+			return;
+
+		pci_save_state(pdev);
+	}
+
+	/*
+	 * Disable INTx and MSI, presumably to avoid spurious interrupts
+	 * during reset.  Stolen from pci_reset_function()
+	 */
+	pci_write_config_word(pdev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
+
+	/*
+	 * Try to get the locks ourselves to prevent a deadlock. The
+	 * success of this is dependent on being able to lock the device,
+	 * which is not always possible.
+	 * We can not use the "try" reset interface here, which will
+	 * overwrite the previously restored configuration information.
+	 */
+	if (vdev->reset_works) {
+		bridge = pci_upstream_bridge(pdev);
+		if (bridge && !pci_dev_trylock(bridge))
+			goto out_restore_state;
+		if (pci_dev_trylock(pdev)) {
+			if (!__pci_reset_function_locked(pdev))
+				vdev->needs_reset = false;
+			pci_dev_unlock(pdev);
+		}
+		if (bridge)
+			pci_dev_unlock(bridge);
+	}
+
+out_restore_state:
+	pci_restore_state(pdev);
+}
+
+void vfio_pci_core_disable(struct vfio_pci_core_device *vdev)
+{
 	struct pci_dev *pdev = vdev->pdev;
 	struct vfio_pci_dummy_resource *dummy_res, *tmp;
 	struct vfio_pci_ioeventfd *ioeventfd, *ioeventfd_tmp;
@@ -700,54 +751,9 @@ void vfio_pci_core_disable(struct vfio_pci_core_device *vdev)
 		kfree(dummy_res);
 	}
 
-	vdev->needs_reset = true;
-
 	vfio_pci_zdev_close_device(vdev);
 
-	/*
-	 * If we have saved state, restore it.  If we can reset the device,
-	 * even better.  Resetting with current state seems better than
-	 * nothing, but saving and restoring current state without reset
-	 * is just busy work.
-	 */
-	if (pci_load_and_free_saved_state(pdev, &vdev->pci_saved_state)) {
-		pci_info(pdev, "%s: Couldn't reload saved state\n", __func__);
-
-		if (!vdev->reset_works)
-			goto out;
-
-		pci_save_state(pdev);
-	}
-
-	/*
-	 * Disable INTx and MSI, presumably to avoid spurious interrupts
-	 * during reset.  Stolen from pci_reset_function()
-	 */
-	pci_write_config_word(pdev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
-
-	/*
-	 * Try to get the locks ourselves to prevent a deadlock. The
-	 * success of this is dependent on being able to lock the device,
-	 * which is not always possible.
-	 * We can not use the "try" reset interface here, which will
-	 * overwrite the previously restored configuration information.
-	 */
-	if (vdev->reset_works) {
-		bridge = pci_upstream_bridge(pdev);
-		if (bridge && !pci_dev_trylock(bridge))
-			goto out_restore_state;
-		if (pci_dev_trylock(pdev)) {
-			if (!__pci_reset_function_locked(pdev))
-				vdev->needs_reset = false;
-			pci_dev_unlock(pdev);
-		}
-		if (bridge)
-			pci_dev_unlock(bridge);
-	}
-
-out_restore_state:
-	pci_restore_state(pdev);
-out:
+	vfio_pci_core_try_reset(vdev);
 	pci_disable_device(pdev);
 
 	vfio_pci_dev_set_try_reset(vdev->vdev.dev_set);
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 02/20] vfio: Export various helpers from VFIO

[Vipin Sharma]

The upcoming VFIO live update feature needs to access various helper
functions for its logic which are not exported by default.

Export these functions and variable for other modules to use them.
Remove static from their declaration and add them in
include/linux/vfio.h and drivers/vfio/pci/vfio_pci_priv.h header files.

No functional change intended.

Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/pci/vfio_pci.c      | 2 +-
 drivers/vfio/pci/vfio_pci_core.c | 6 ++++--
 drivers/vfio/pci/vfio_pci_priv.h | 5 +++++
 drivers/vfio/vfio_main.c         | 3 ++-
 include/linux/vfio.h             | 1 +
 5 files changed, 13 insertions(+), 4 deletions(-)

diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
index 0c771064c0b8..1fb4b276afe7 100644
--- a/drivers/vfio/pci/vfio_pci.c
+++ b/drivers/vfio/pci/vfio_pci.c
@@ -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,
diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index 3c8b990ee92f..f31e59e5badb 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -623,7 +623,7 @@ int vfio_pci_core_enable(struct vfio_pci_core_device *vdev)
 }
 EXPORT_SYMBOL_GPL(vfio_pci_core_enable);
 
-static void vfio_pci_core_try_reset(struct vfio_pci_core_device *vdev)
+void vfio_pci_core_try_reset(struct vfio_pci_core_device *vdev)
 {
 	struct pci_dev *pdev = vdev->pdev;
 	struct pci_dev *bridge;
@@ -674,6 +674,7 @@ static void vfio_pci_core_try_reset(struct vfio_pci_core_device *vdev)
 out_restore_state:
 	pci_restore_state(pdev);
 }
+EXPORT_SYMBOL_GPL(vfio_pci_core_try_reset);
 
 void vfio_pci_core_disable(struct vfio_pci_core_device *vdev)
 {
@@ -1657,7 +1658,7 @@ ssize_t vfio_pci_core_write(struct vfio_device *core_vdev, const char __user *bu
 }
 EXPORT_SYMBOL_GPL(vfio_pci_core_write);
 
-static void vfio_pci_zap_bars(struct vfio_pci_core_device *vdev)
+void vfio_pci_zap_bars(struct vfio_pci_core_device *vdev)
 {
 	struct vfio_device *core_vdev = &vdev->vdev;
 	loff_t start = VFIO_PCI_INDEX_TO_OFFSET(VFIO_PCI_BAR0_REGION_INDEX);
@@ -1666,6 +1667,7 @@ static void vfio_pci_zap_bars(struct vfio_pci_core_device *vdev)
 
 	unmap_mapping_range(core_vdev->inode->i_mapping, start, len, true);
 }
+EXPORT_SYMBOL_GPL(vfio_pci_zap_bars);
 
 void vfio_pci_zap_and_down_write_memory_lock(struct vfio_pci_core_device *vdev)
 {
diff --git a/drivers/vfio/pci/vfio_pci_priv.h b/drivers/vfio/pci/vfio_pci_priv.h
index fca9d0dfac90..85bc1666cb0e 100644
--- a/drivers/vfio/pci/vfio_pci_priv.h
+++ b/drivers/vfio/pci/vfio_pci_priv.h
@@ -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;
@@ -73,6 +75,9 @@ u16 vfio_pci_memory_lock_and_enable(struct vfio_pci_core_device *vdev);
 void vfio_pci_memory_unlock_and_restore(struct vfio_pci_core_device *vdev,
 					u16 cmd);
 
+void vfio_pci_zap_bars(struct vfio_pci_core_device *vdev);
+void vfio_pci_core_try_reset(struct vfio_pci_core_device *vdev);
+
 #ifdef CONFIG_VFIO_PCI_IGD
 bool vfio_pci_is_intel_display(struct pci_dev *pdev);
 int vfio_pci_igd_init(struct vfio_pci_core_device *vdev);
diff --git a/drivers/vfio/vfio_main.c b/drivers/vfio/vfio_main.c
index 5e0422014523..2437af031169 100644
--- a/drivers/vfio/vfio_main.c
+++ b/drivers/vfio/vfio_main.c
@@ -1467,7 +1467,7 @@ const struct file_operations vfio_device_fops = {
 #endif
 };
 
-static struct vfio_device *vfio_device_from_file(struct file *file)
+struct vfio_device *vfio_device_from_file(struct file *file)
 {
 	struct vfio_device_file *df = file->private_data;
 
@@ -1475,6 +1475,7 @@ static struct vfio_device *vfio_device_from_file(struct file *file)
 		return NULL;
 	return df->device;
 }
+EXPORT_SYMBOL_GPL(vfio_device_from_file);
 
 /**
  * vfio_file_is_valid - True if the file is valid vfio file
diff --git a/include/linux/vfio.h b/include/linux/vfio.h
index 31b826efba00..cf0d941d0410 100644
--- a/include/linux/vfio.h
+++ b/include/linux/vfio.h
@@ -378,6 +378,7 @@ static inline bool vfio_file_has_dev(struct file *file, struct vfio_device *devi
 bool vfio_file_is_valid(struct file *file);
 bool vfio_file_enforced_coherent(struct file *file);
 void vfio_file_set_kvm(struct file *file, struct kvm *kvm);
+struct vfio_device *vfio_device_from_file(struct file *file);
 
 #define VFIO_PIN_PAGES_MAX_ENTRIES	(PAGE_SIZE/sizeof(unsigned long))
 
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 02/20] vfio: Export various helpers from VFIO

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] Missing EXPORT_SYMBOL_GPL for vfio_pci_ops, violating the stated intent to export the variable for other modules.
--

--- Patch [2]: [PATCH v5 02/20] vfio: Export various helpers from VFIO ---
commit fcee4889ab47813a22592eb00c453edea425e0f0
Author: Vipin Sharma <vipinsh@google.com>

vfio: Export various helpers from VFIO

This commit makes several vfio helpers and the vfio_pci_ops variable globally
visible by removing their static declarations and adding them to headers. The
stated intent is to export these symbols for use by other modules, such as
the upcoming live update feature.

> diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
> index 0c771064c0b84..1fb4b276afe7c 100644
> --- a/drivers/vfio/pci/vfio_pci.c
> +++ b/drivers/vfio/pci/vfio_pci.c
> @@ -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",

[Severity: Low]
The commit message states the intent is to "Export these functions and
variable for other modules to use them", but is there a missing
EXPORT_SYMBOL_GPL(vfio_pci_ops) here?

The other helpers modified in this patch received an EXPORT_SYMBOL_GPL() tag,
but vfio_pci_ops did not. Without it, external modules attempting to reference
vfio_pci_ops will fail with unresolved symbol errors during the build.

>  	.init		= vfio_pci_core_init_dev,
>  	.release	= vfio_pci_core_release_dev,

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=2

[PATCH v5 03/20] vfio/pci: Export vfio_pci_dma_buf_move for vfio-pci module

[Vipin Sharma]

VFIO Live update during freeze needs to reset the device before handing
over to next kernel.

Export DMA revoke function which VFIO Live Update will call in its
freeze before reset so that P2P DMA doesn't error out or cause some
other issue.

Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/pci/vfio_pci_dmabuf.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/vfio/pci/vfio_pci_dmabuf.c b/drivers/vfio/pci/vfio_pci_dmabuf.c
index c16f460c01d6..2c05fb5a171a 100644
--- a/drivers/vfio/pci/vfio_pci_dmabuf.c
+++ b/drivers/vfio/pci/vfio_pci_dmabuf.c
@@ -376,6 +376,7 @@ void vfio_pci_dma_buf_move(struct vfio_pci_core_device *vdev, bool revoked)
 		fput(priv->dmabuf->file);
 	}
 }
+EXPORT_SYMBOL_GPL(vfio_pci_dma_buf_move);
 
 void vfio_pci_dma_buf_cleanup(struct vfio_pci_core_device *vdev)
 {
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 04/20] liveupdate: Export symbols needed by modules

[Vipin Sharma]

Export liveupdate_register_file_handler() and
liveupdate_unregister_file_handler(). All of these will be used by
vfio-pci in a subsequent commit, which can be built as a module.

Reviewed-by: Samiullah Khawaja <skhawaja@google.com>
Reviewed-by: Pranjal Shrivastava <praan@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 kernel/liveupdate/luo_file.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/kernel/liveupdate/luo_file.c b/kernel/liveupdate/luo_file.c
index c39f96961a85..dd4e26ac34f0 100644
--- a/kernel/liveupdate/luo_file.c
+++ b/kernel/liveupdate/luo_file.c
@@ -909,6 +909,7 @@ int liveupdate_register_file_handler(struct liveupdate_file_handler *fh)
 	up_write(&luo_register_rwlock);
 	return err;
 }
+EXPORT_SYMBOL_GPL(liveupdate_register_file_handler);
 
 /**
  * liveupdate_unregister_file_handler - Unregister a liveupdate file handler
@@ -926,3 +927,4 @@ void liveupdate_unregister_file_handler(struct liveupdate_file_handler *fh)
 	luo_flb_unregister_all(fh);
 	list_del(&ACCESS_PRIVATE(fh, list));
 }
+EXPORT_SYMBOL_GPL(liveupdate_unregister_file_handler);
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 05/20] vfio/pci: Register a file handler with Live Update Orchestrator

[Vipin Sharma]

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). Register it with both live
update framework and PCI subsystem.

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.

Put this support behind a new Kconfig VFIO_PCI_LIVEUPDATE that is marked
experimental and default-disabled until more of the device preservation
support has landed in the kernel.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 MAINTAINERS                            |  1 +
 drivers/vfio/pci/Kconfig               | 12 ++++
 drivers/vfio/pci/Makefile              |  1 +
 drivers/vfio/pci/vfio_pci.c            | 11 +++-
 drivers/vfio/pci/vfio_pci_liveupdate.c | 80 ++++++++++++++++++++++++++
 drivers/vfio/pci/vfio_pci_priv.h       | 14 +++++
 include/linux/kho/abi/vfio_pci.h       | 28 +++++++++
 7 files changed, 146 insertions(+), 1 deletion(-)
 create mode 100644 drivers/vfio/pci/vfio_pci_liveupdate.c
 create mode 100644 include/linux/kho/abi/vfio_pci.h

diff --git a/MAINTAINERS b/MAINTAINERS
index 347c435ca404..77a35742de0a 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -28329,6 +28329,7 @@ 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
diff --git a/drivers/vfio/pci/Kconfig b/drivers/vfio/pci/Kconfig
index 296bf01e185e..c12d614fc6c4 100644
--- a/drivers/vfio/pci/Kconfig
+++ b/drivers/vfio/pci/Kconfig
@@ -42,6 +42,18 @@ config VFIO_PCI_IGD
 	  and LPC bridge config space.
 
 	  To enable Intel IGD assignment through vfio-pci, say Y.
+
+config VFIO_PCI_LIVEUPDATE
+	bool "VFIO PCI support for Live Update (EXPERIMENTAL)"
+	depends on PCI_LIVEUPDATE
+	help
+	  Support for preserving devices bound to vfio-pci across a Live
+	  Update. This option should only be enabled by developers working on
+	  implementing this support. Once enough support has landed in the
+	  kernel, this option will no longer be marked EXPERIMENTAL.
+
+	  If you don't know what to do here, say N.
+
 endif
 
 config VFIO_PCI_ZDEV_KVM
diff --git a/drivers/vfio/pci/Makefile b/drivers/vfio/pci/Makefile
index 6138f1bf241d..bf296ab1c967 100644
--- a/drivers/vfio/pci/Makefile
+++ b/drivers/vfio/pci/Makefile
@@ -7,6 +7,7 @@ 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_VFIO_PCI_LIVEUPDATE) += vfio_pci_liveupdate.o
 obj-$(CONFIG_VFIO_PCI) += vfio-pci.o
 
 obj-$(CONFIG_MLX5_VFIO_PCI)           += mlx5/
diff --git a/drivers/vfio/pci/vfio_pci.c b/drivers/vfio/pci/vfio_pci.c
index 1fb4b276afe7..439d14d79d8f 100644
--- a/drivers/vfio/pci/vfio_pci.c
+++ b/drivers/vfio/pci/vfio_pci.c
@@ -264,10 +264,14 @@ static int __init vfio_pci_init(void)
 
 	vfio_pci_core_set_params(nointxmask, is_disable_vga, disable_idle_d3);
 
+	ret = vfio_pci_liveupdate_init();
+	if (ret)
+		return ret;
+
 	/* Register and scan for devices */
 	ret = pci_register_driver(&vfio_pci_driver);
 	if (ret)
-		return ret;
+		goto err_liveupdate_cleanup;
 
 	vfio_pci_fill_ids();
 
@@ -275,12 +279,17 @@ static int __init vfio_pci_init(void)
 		pr_warn("device denylist disabled.\n");
 
 	return 0;
+
+err_liveupdate_cleanup:
+	vfio_pci_liveupdate_cleanup();
+	return ret;
 }
 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);
 
diff --git a/drivers/vfio/pci/vfio_pci_liveupdate.c b/drivers/vfio/pci/vfio_pci_liveupdate.c
new file mode 100644
index 000000000000..803528c9f1b2
--- /dev/null
+++ b/drivers/vfio/pci/vfio_pci_liveupdate.c
@@ -0,0 +1,80 @@
+// SPDX-License-Identifier: GPL-2.0
+
+/*
+ * Copyright (c) 2026, Google LLC.
+ * Vipin Sharma <vipinsh@google.com>
+ * David Matlack <dmatlack@google.com>
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/errno.h>
+#include <linux/kho/abi/vfio_pci.h>
+#include <linux/liveupdate.h>
+#include <linux/module.h>
+
+#include "vfio_pci_priv.h"
+
+static bool vfio_pci_liveupdate_can_preserve(struct liveupdate_file_handler *handler,
+					     struct file *file)
+{
+	return false;
+}
+
+static int vfio_pci_liveupdate_preserve(struct liveupdate_file_op_args *args)
+{
+	return -EOPNOTSUPP;
+}
+
+static void vfio_pci_liveupdate_unpreserve(struct liveupdate_file_op_args *args)
+{
+}
+
+static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args)
+{
+	return -EOPNOTSUPP;
+}
+
+static void vfio_pci_liveupdate_finish(struct liveupdate_file_op_args *args)
+{
+}
+
+static const struct liveupdate_file_ops vfio_pci_liveupdate_file_ops = {
+	.can_preserve = vfio_pci_liveupdate_can_preserve,
+	.preserve = vfio_pci_liveupdate_preserve,
+	.unpreserve = vfio_pci_liveupdate_unpreserve,
+	.retrieve = vfio_pci_liveupdate_retrieve,
+	.finish = vfio_pci_liveupdate_finish,
+	.owner = THIS_MODULE,
+};
+
+static struct liveupdate_file_handler vfio_pci_liveupdate_fh = {
+	.ops = &vfio_pci_liveupdate_file_ops,
+	.compatible = VFIO_PCI_LUO_FH_COMPATIBLE,
+};
+
+int __init vfio_pci_liveupdate_init(void)
+{
+	int ret;
+
+	ret = liveupdate_register_file_handler(&vfio_pci_liveupdate_fh);
+	if (ret)
+		goto err_return;
+
+	ret = pci_liveupdate_register_flb(&vfio_pci_liveupdate_fh);
+	if (ret)
+		goto err_unregister;
+
+	return 0;
+
+err_unregister:
+	liveupdate_unregister_file_handler(&vfio_pci_liveupdate_fh);
+err_return:
+	return (ret == -EOPNOTSUPP) ? 0 : ret;
+}
+
+void vfio_pci_liveupdate_cleanup(void)
+{
+	pci_liveupdate_unregister_flb(&vfio_pci_liveupdate_fh);
+	liveupdate_unregister_file_handler(&vfio_pci_liveupdate_fh);
+}
diff --git a/drivers/vfio/pci/vfio_pci_priv.h b/drivers/vfio/pci/vfio_pci_priv.h
index 85bc1666cb0e..5d2ad128523d 100644
--- a/drivers/vfio/pci/vfio_pci_priv.h
+++ b/drivers/vfio/pci/vfio_pci_priv.h
@@ -142,4 +142,18 @@ static inline void vfio_pci_dma_buf_move(struct vfio_pci_core_device *vdev,
 }
 #endif
 
+#ifdef CONFIG_VFIO_PCI_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_VFIO_PCI_LIVEUPDATE */
+
 #endif
diff --git a/include/linux/kho/abi/vfio_pci.h b/include/linux/kho/abi/vfio_pci.h
new file mode 100644
index 000000000000..67e8199eec59
--- /dev/null
+++ b/include/linux/kho/abi/vfio_pci.h
@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+/*
+ * Copyright (c) 2026, Google LLC.
+ * Vipin Sharma <vipinsh@google.com>
+ * David Matlack <dmatlack@google.com>
+ */
+
+#ifndef _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H
+#define _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H
+
+/**
+ * DOC: VFIO PCI Live Update ABI
+ *
+ * VFIO uses the ABI defined below for preserving 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 */
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 05/20] vfio/pci: Register a file handler with Live Update Orchestrator

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [High] The Kconfig option VFIO_PCI_LIVEUPDATE depends on a non-existent configuration symbol PCI_LIVEUPDATE.
- [High] Calls to undeclared and undefined functions pci_liveupdate_register_flb() and pci_liveupdate_unregister_flb().
- [High] Masking -EOPNOTSUPP on FLB registration failure leads to a double unregistration of the file handler and potential kernel panic.
--

--- Patch [5]: [PATCH v5 05/20] vfio/pci: Register a file handler with Live Update Orchestrator ---
commit a3c1a8c69f4b2c27d2e063cca84eb003906e3bbf
Author: Vipin Sharma <vipinsh@google.com>

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

This commit adds a live update file handler for vfio-pci device files
with stub implementations. It registers the handler with the live update
framework and PCI subsystem, gated behind a new VFIO_PCI_LIVEUPDATE option.

> diff --git a/drivers/vfio/pci/Kconfig b/drivers/vfio/pci/Kconfig
> index 296bf01e185ec..c12d614fc6c4a 100644
> --- a/drivers/vfio/pci/Kconfig
> +++ b/drivers/vfio/pci/Kconfig
[ ... ]
> @@ -42,6 +42,18 @@ config VFIO_PCI_IGD
>  	  and LPC bridge config space.
>  
>  	  To enable Intel IGD assignment through vfio-pci, say Y.
> +
> +config VFIO_PCI_LIVEUPDATE
> +	bool "VFIO PCI support for Live Update (EXPERIMENTAL)"
> +	depends on PCI_LIVEUPDATE

[Severity: High]
Does the PCI_LIVEUPDATE configuration symbol exist anywhere in the kernel
tree? It appears this dependency cannot be met, preventing the new
VFIO_PCI_LIVEUPDATE option from being selected.

> +	help
> +	  Support for preserving devices bound to vfio-pci across a Live
[ ... ]
> diff --git a/drivers/vfio/pci/vfio_pci_liveupdate.c b/drivers/vfio/pci/vfio_pci_liveupdate.c
> new file mode 100644
> index 0000000000000..803528c9f1b27
> --- /dev/null
> +++ b/drivers/vfio/pci/vfio_pci_liveupdate.c
[ ... ]
> +int __init vfio_pci_liveupdate_init(void)
> +{
> +	int ret;
> +
> +	ret = liveupdate_register_file_handler(&vfio_pci_liveupdate_fh);
> +	if (ret)
> +		goto err_return;
> +
> +	ret = pci_liveupdate_register_flb(&vfio_pci_liveupdate_fh);

[Severity: High]
Are pci_liveupdate_register_flb() and pci_liveupdate_unregister_flb()
declared in any header or defined in any source file? Building with
this configuration enabled might result in implicit declaration compiler
errors and fatal linker errors.

> +	if (ret)
> +		goto err_unregister;
> +
> +	return 0;
> +
> +err_unregister:
> +	liveupdate_unregister_file_handler(&vfio_pci_liveupdate_fh);
> +err_return:
> +	return (ret == -EOPNOTSUPP) ? 0 : ret;

[Severity: High]
If pci_liveupdate_register_flb() returns -EOPNOTSUPP, this error path
unregisters the file handler but returns 0. Could this lead to a double
unregistration when the module is unloaded and vfio_pci_liveupdate_cleanup()
is called?

Calling liveupdate_unregister_file_handler() a second time on the already
removed list node might cause list corruption and a kernel panic.

> +}
> +
> +void vfio_pci_liveupdate_cleanup(void)
> +{
> +	pci_liveupdate_unregister_flb(&vfio_pci_liveupdate_fh);
> +	liveupdate_unregister_file_handler(&vfio_pci_liveupdate_fh);
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=5

[PATCH v5 06/20] vfio/pci: Preserve vfio-pci device files across Live Update

[Vipin Sharma]

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.

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 are the
device's PCI segment number and BDF, ensuring the device can be uniquely
identified after the Live Update.

Key design notes and constraints:

 * Device state before kexec: Userspace is required to disable
   interrupts on the device and ensure it is in the D0 power state.
   This ensures the device does not send any interrupts until new
   handlers are set up by the next kernel.

 * Device reset during freeze: The device is reset and its state
   is restored in the freeze() callback, ensuring the next kernel
   receives it in a consistent state. Eventually, this will be
   dropped so the device can be preserved in a running state, but
   that requires further work in VFIO and the core PCI layer.

 * FD usability after freeze: Once the device is frozen, the VFIO
   device FD becomes unusable (i.e., ioctls and memory accesses will
   return errors) because the device may have already been reset.
   If the freeze operation fails, userspace must close the FD and
   reinitialize the device.

 * Lifecycle: LUO holds a reference to this file while it is preserved.
   Therefore, VFIO is guaranteed that vfio_df_device_last_close() will
   not be called on this device until userspace has unpreserved the
   file via LUO.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/pci/Kconfig               |   2 +-
 drivers/vfio/pci/vfio_pci_config.c     |   3 +
 drivers/vfio/pci/vfio_pci_core.c       |  15 ++
 drivers/vfio/pci/vfio_pci_liveupdate.c | 198 ++++++++++++++++++++++++-
 drivers/vfio/pci/vfio_pci_priv.h       |   9 ++
 include/linux/kho/abi/vfio_pci.h       |  15 ++
 include/linux/vfio_pci_core.h          |   3 +
 7 files changed, 242 insertions(+), 3 deletions(-)

diff --git a/drivers/vfio/pci/Kconfig b/drivers/vfio/pci/Kconfig
index c12d614fc6c4..d97c809599fa 100644
--- a/drivers/vfio/pci/Kconfig
+++ b/drivers/vfio/pci/Kconfig
@@ -45,7 +45,7 @@ config VFIO_PCI_IGD
 
 config VFIO_PCI_LIVEUPDATE
 	bool "VFIO PCI support for Live Update (EXPERIMENTAL)"
-	depends on PCI_LIVEUPDATE
+	depends on PCI_LIVEUPDATE && VFIO_DEVICE_CDEV
 	help
 	  Support for preserving devices bound to vfio-pci across a Live
 	  Update. This option should only be enabled by developers working on
diff --git a/drivers/vfio/pci/vfio_pci_config.c b/drivers/vfio/pci/vfio_pci_config.c
index a10ed733f0e3..a51805515d30 100644
--- a/drivers/vfio/pci/vfio_pci_config.c
+++ b/drivers/vfio/pci/vfio_pci_config.c
@@ -406,6 +406,9 @@ bool __vfio_pci_memory_enabled(struct vfio_pci_core_device *vdev)
 	struct pci_dev *pdev = vdev->pdev;
 	u16 cmd = le16_to_cpu(*(__le16 *)&vdev->vconfig[PCI_COMMAND]);
 
+	if (vfio_pci_dev_is_frozen(vdev))
+		return false;
+
 	/*
 	 * Memory region cannot be accessed if device power state is D3.
 	 *
diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index f31e59e5badb..c312b0eaf7d0 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -1087,6 +1087,9 @@ int vfio_pci_ioctl_get_region_info(struct vfio_device *core_vdev,
 	struct pci_dev *pdev = vdev->pdev;
 	int i, ret;
 
+	if (vfio_pci_dev_is_frozen(vdev))
+		return -EIO;
+
 	switch (info->index) {
 	case VFIO_PCI_CONFIG_REGION_INDEX:
 		info->offset = VFIO_PCI_INDEX_TO_OFFSET(info->index);
@@ -1513,6 +1516,9 @@ long vfio_pci_core_ioctl(struct vfio_device *core_vdev, unsigned int cmd,
 		container_of(core_vdev, struct vfio_pci_core_device, vdev);
 	void __user *uarg = (void __user *)arg;
 
+	if (vfio_pci_dev_is_frozen(vdev))
+		return -EIO;
+
 	switch (cmd) {
 	case VFIO_DEVICE_GET_INFO:
 		return vfio_pci_ioctl_get_info(vdev, uarg);
@@ -1567,6 +1573,9 @@ int vfio_pci_core_ioctl_feature(struct vfio_device *device, u32 flags,
 	struct vfio_pci_core_device *vdev =
 		container_of(device, struct vfio_pci_core_device, vdev);
 
+	if (vfio_pci_dev_is_frozen(vdev))
+		return -EIO;
+
 	switch (flags & VFIO_DEVICE_FEATURE_MASK) {
 	case VFIO_DEVICE_FEATURE_LOW_POWER_ENTRY:
 		return vfio_pci_core_pm_entry(vdev, flags, arg, argsz);
@@ -1591,6 +1600,9 @@ static ssize_t vfio_pci_rw(struct vfio_pci_core_device *vdev, char __user *buf,
 	unsigned int index = VFIO_PCI_OFFSET_TO_INDEX(*ppos);
 	int ret;
 
+	if (vfio_pci_dev_is_frozen(vdev))
+		return -EIO;
+
 	if (index >= VFIO_PCI_NUM_REGIONS + vdev->num_regions)
 		return -EINVAL;
 
@@ -1775,6 +1787,9 @@ int vfio_pci_core_mmap(struct vfio_device *core_vdev, struct vm_area_struct *vma
 	u64 phys_len, req_len, pgoff, req_start;
 	void __iomem *bar_io;
 
+	if (vfio_pci_dev_is_frozen(vdev))
+		return -EIO;
+
 	index = vma->vm_pgoff >> (VFIO_PCI_OFFSET_SHIFT - PAGE_SHIFT);
 
 	if (index >= VFIO_PCI_NUM_REGIONS + vdev->num_regions)
diff --git a/drivers/vfio/pci/vfio_pci_liveupdate.c b/drivers/vfio/pci/vfio_pci_liveupdate.c
index 803528c9f1b2..e6eb7f6b53d7 100644
--- a/drivers/vfio/pci/vfio_pci_liveupdate.c
+++ b/drivers/vfio/pci/vfio_pci_liveupdate.c
@@ -6,28 +6,221 @@
  * David Matlack <dmatlack@google.com>
  */
 
+/**
+ * DOC: VFIO PCI Preservation via LUO
+ *
+ * VFIO PCI devices can be preserved over a kexec using the Live Update
+ * Orchestrator (LUO) file preservation. This allows userspace (such as a VMM)
+ * to transfer an in-use device to the next kernel.
+ *
+ * .. note::
+ *    The support for preserving VFIO PCI devices is currently *partial* and
+ *    should be considered *experimental*. It should only be used by developers
+ *    working on expanding the support for the time being.
+ *
+ *    To avoid accidental usage while the support is still experimental, this
+ *    support is hidden behind a default-disable config option
+ *    ``CONFIG_VFIO_PCI_LIVEUPDATE``. Once the kernel support has stabilized and
+ *    become complete, this option will be enabled by default when
+ *    ``CONFIG_VFIO_PCI`` and ``CONFIG_PCI_LIVEUPDATE`` are enabled.
+ *
+ * Usage Example
+ * =============
+ *
+ * VFIO PCI devices can be preserved across a kexec by preserving the file
+ * associated with the device in a LUO session::
+ *
+ *   device_fd = open("/dev/vfio/devices/vfioX");
+ *   ...
+ *   ioctl(session_fd, LIVEUPDATE_SESSION_PRESERVE_FD, { ..., device_fd, ...});
+ *
+ * .. note::
+ *    LUO will hold an extra reference to the device file for as long as it is
+ *    preserved, so there is no way for the file to be destroyed or the device
+ *    to be unbound from the vfio-pci driver while it is preserved.
+ *
+ * Retrieving the file after kexec is not yet supported.
+ *
+ * Restrictions
+ * ============
+ *
+ * The kernel imposes the following restrictions when preserving VFIO devices:
+ *
+ *  * The device must be bound to the ``vfio-pci`` driver.
+ *
+ *  * ``CONFIG_VFIO_PCI_ZDEV_KVM`` must not be enabled. This may be relaxed in
+ *    the future.
+ *
+ *  * The device must not be an Intel display device. This may be relaxed in
+ *    the future.
+ *
+ *  * The device file descriptor must be obtained by opening the VFIO character
+ *    device (``/dev/vfio/devices/vfioX``) and not via
+ *    ``VFIO_GROUP_GET_DEVICE_FD``.
+ *
+ *  * The device must meet following conditions prior to kexec without which the
+ *    ``reboot(LINUX_REBOOT_CMD_KEXEC)`` syscall (to initiate the kexec) will
+ *    fail.
+ *
+ *    - The device must have interrupt disabled.
+ *    - The device must be in D0 state.
+ *
+ * In addition, the device must meet all of the restrictions imposed by the
+ * core PCI layer documented at :doc:`/PCI/liveupdate`.
+ *
+ * Preservation Behavior
+ * =====================
+ *
+ * The eventual goal of this support is to avoid disrupting the workload, state,
+ * or configuration of each preserved device during a Live Update. This would
+ * include allowing the device to perform DMA to preserved memory buffers and
+ * perform P2P DMA to other preserved devices. However, there are many pieces
+ * that still need to land in the kernel.
+ *
+ * For now, VFIO only preserves the following state for devices:
+ *
+ *  * The PCI Segment, Bus, Device, and Function numbers of the device. The
+ *    kernel guarantees the these will not change across a kexec when a device
+ *    is preserved.
+ *
+ * Since the kernel is not yet prepared to preserve all parts of the device and
+ * its dependencies (such as DMA mappings), VFIO currently resets and restores
+ * preserved devices back into an idle state during kexec, before handing off
+ * control to the next kernel. This will be relaxed in future versions of the
+ * kernel once it is safe to allow the device to keep running across kexec.
+ */
+
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/errno.h>
+#include <linux/kexec_handover.h>
 #include <linux/kho/abi/vfio_pci.h>
 #include <linux/liveupdate.h>
 #include <linux/module.h>
+#include <linux/vfio.h>
 
 #include "vfio_pci_priv.h"
 
 static bool vfio_pci_liveupdate_can_preserve(struct liveupdate_file_handler *handler,
 					     struct file *file)
 {
-	return false;
+	struct vfio_device *device = vfio_device_from_file(file);
+	struct vfio_pci_core_device *vdev;
+	struct pci_dev *pdev;
+
+	if (!device)
+		return false;
+
+	/* Live Update support is limited to cdev files. */
+	if (!vfio_device_cdev_opened(device))
+		return false;
+
+	if (device->ops != &vfio_pci_ops)
+		return false;
+
+	vdev = container_of(device, struct vfio_pci_core_device, vdev);
+	pdev = vdev->pdev;
+
+	/*
+	 * Don't support specialized vfio-pci devices for now since they haven't
+	 * been tested.
+	 */
+	if (IS_ENABLED(CONFIG_VFIO_PCI_ZDEV_KVM) || vfio_pci_is_intel_display(pdev))
+		return false;
+
+	/* Only non-VFs are supported for now. */
+	if (pdev->is_virtfn)
+		return false;
+
+	return true;
 }
 
 static int vfio_pci_liveupdate_preserve(struct liveupdate_file_op_args *args)
 {
-	return -EOPNOTSUPP;
+	struct vfio_device *device = vfio_device_from_file(args->file);
+	struct vfio_pci_core_device_ser *ser;
+	struct vfio_pci_core_device *vdev;
+	struct pci_dev *pdev;
+	int ret;
+
+	vdev = container_of(device, struct vfio_pci_core_device, vdev);
+	pdev = vdev->pdev;
+
+	ret = pci_liveupdate_preserve(pdev);
+	if (ret)
+		return ret;
+
+	ser = kho_alloc_preserve(sizeof(*ser));
+	if (IS_ERR(ser)) {
+		ret = PTR_ERR(ser);
+		goto err_unpreserve;
+	}
+
+	ser->bdf = pci_dev_id(pdev);
+	ser->domain = pci_domain_nr(pdev->bus);
+
+	args->serialized_data = virt_to_phys(ser);
+	return 0;
+
+err_unpreserve:
+	pci_liveupdate_unpreserve(pdev);
+	return ret;
 }
 
 static void vfio_pci_liveupdate_unpreserve(struct liveupdate_file_op_args *args)
 {
+	struct vfio_device *device = vfio_device_from_file(args->file);
+
+	pci_liveupdate_unpreserve(to_pci_dev(device->dev));
+	kho_unpreserve_free(phys_to_virt(args->serialized_data));
+}
+
+static int vfio_pci_liveupdate_freeze(struct liveupdate_file_op_args *args)
+{
+	struct vfio_device *device = vfio_device_from_file(args->file);
+	struct vfio_pci_core_device *vdev;
+	struct pci_dev *pdev;
+
+	vdev = container_of(device, struct vfio_pci_core_device, vdev);
+	pdev = vdev->pdev;
+
+	guard(mutex)(&device->dev_set->lock);
+	guard(mutex)(&vdev->igate);
+	guard(rwsem_write)(&vdev->memory_lock);
+
+	/*
+	 * Userspace must disable interrupts on the device prior to freeze so
+	 * that the device does not send any interrupts until new interrupt
+	 * handlers have been established by the next kernel.
+	 */
+	if (vdev->irq_type != VFIO_PCI_NUM_IRQS) {
+		pci_err(pdev, "Freeze failed! Interrupts are still enabled.\n");
+		return -EINVAL;
+	}
+
+	if (pdev->current_state != PCI_D0) {
+		pci_err(pdev, "Freeze failed! Device not in D0 state.\n");
+		return -EINVAL;
+	}
+
+	/*
+	 * Reset is a temporary measure to provide kernel after kexec a clean
+	 * device while VFIO live update work is under development and not
+	 * fully supported. It will go away once continuous DMA support is
+	 * added to device preservation.
+	 */
+	vfio_pci_zap_bars(vdev);
+	vfio_pci_dma_buf_move(vdev, true);
+	vfio_pci_core_try_reset(vdev);
+	pci_write_config_word(pdev, PCI_COMMAND, PCI_COMMAND_INTX_DISABLE);
+	/*
+	 * Userspace cannot use the FD correctly now irrespective of liveupdate
+	 * freeze failing or succeeding. They will have to reinitialize the VFIO
+	 * device to continue using it as reset might have loaded default PCI
+	 * state. Disable ioctl, read, write and mmap access to the device.
+	 */
+	smp_store_release(&vdev->liveupdate_frozen, true);
+	return 0;
 }
 
 static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args)
@@ -43,6 +236,7 @@ static const struct liveupdate_file_ops vfio_pci_liveupdate_file_ops = {
 	.can_preserve = vfio_pci_liveupdate_can_preserve,
 	.preserve = vfio_pci_liveupdate_preserve,
 	.unpreserve = vfio_pci_liveupdate_unpreserve,
+	.freeze = vfio_pci_liveupdate_freeze,
 	.retrieve = vfio_pci_liveupdate_retrieve,
 	.finish = vfio_pci_liveupdate_finish,
 	.owner = THIS_MODULE,
diff --git a/drivers/vfio/pci/vfio_pci_priv.h b/drivers/vfio/pci/vfio_pci_priv.h
index 5d2ad128523d..4b65753e8936 100644
--- a/drivers/vfio/pci/vfio_pci_priv.h
+++ b/drivers/vfio/pci/vfio_pci_priv.h
@@ -145,6 +145,11 @@ static inline void vfio_pci_dma_buf_move(struct vfio_pci_core_device *vdev,
 #ifdef CONFIG_VFIO_PCI_LIVEUPDATE
 int __init vfio_pci_liveupdate_init(void);
 void vfio_pci_liveupdate_cleanup(void);
+static inline bool vfio_pci_dev_is_frozen(struct vfio_pci_core_device *vdev)
+{
+	/* Written in VFIO PCI Liveupdate during freeze() event */
+	return smp_load_acquire(&vdev->liveupdate_frozen);
+}
 #else
 static inline int vfio_pci_liveupdate_init(void)
 {
@@ -154,6 +159,10 @@ static inline int vfio_pci_liveupdate_init(void)
 static inline void vfio_pci_liveupdate_cleanup(void)
 {
 }
+static inline bool vfio_pci_dev_is_frozen(struct vfio_pci_core_device *vdev)
+{
+	return false;
+}
 #endif /* CONFIG_VFIO_PCI_LIVEUPDATE */
 
 #endif
diff --git a/include/linux/kho/abi/vfio_pci.h b/include/linux/kho/abi/vfio_pci.h
index 67e8199eec59..229a13ac1e52 100644
--- a/include/linux/kho/abi/vfio_pci.h
+++ b/include/linux/kho/abi/vfio_pci.h
@@ -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.
+ *
+ * @domain: The device's PCI domain number (segment).
+ * @bdf: The device's PCI bus, device, and function number.
+ */
+struct vfio_pci_core_device_ser {
+	u32 domain;
+	u16 bdf;
+} __packed;
+
 #endif /* _LINUX_LIVEUPDATE_ABI_VFIO_PCI_H */
diff --git a/include/linux/vfio_pci_core.h b/include/linux/vfio_pci_core.h
index 5fc6ce4dd786..adc5d05456bd 100644
--- a/include/linux/vfio_pci_core.h
+++ b/include/linux/vfio_pci_core.h
@@ -143,6 +143,9 @@ struct vfio_pci_core_device {
 	struct notifier_block	nb;
 	struct rw_semaphore	memory_lock;
 	struct list_head	dmabufs;
+#ifdef CONFIG_VFIO_PCI_LIVEUPDATE
+	bool			liveupdate_frozen;
+#endif
 };
 
 enum vfio_pci_io_width {
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 06/20] vfio/pci: Preserve vfio-pci device files across Live Update

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] Lockless checks of `vfio_pci_dev_is_frozen()` in IOCTL and read/write handlers allow userspace to bypass the freeze state, breaking kexec handoff guarantees.
- [Medium] `vdev->liveupdate_frozen` is never cleared and `unfreeze` is unimplemented, permanently bricking the VFIO device if a kexec aborts.
--

--- Patch [6]: [PATCH v5 06/20] vfio/pci: Preserve vfio-pci device files across Live Update ---
commit 03e42ce3ade0f06aa286a1a8cacc4d3f5d2537c0
Author: Vipin Sharma <vipinsh@google.com>
vfio/pci: Preserve vfio-pci device files across Live Update

This commit implements the Live Update file handler callbacks to preserve a
vfio-pci device across a Live Update. It serializes device state into
vfio_pci_core_device_ser and resets the device during the freeze phase.

> diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
> index f31e59e5badb4..c312b0eaf7d00 100644
> --- a/drivers/vfio/pci/vfio_pci_core.c
> +++ b/drivers/vfio/pci/vfio_pci_core.c
[ ... ]
> @@ -1513,6 +1516,9 @@ long vfio_pci_core_ioctl(struct vfio_device *core_vdev, unsigned int cmd,
>  		container_of(core_vdev, struct vfio_pci_core_device, vdev);
>  	void __user *uarg = (void __user *)arg;
>  
> +	if (vfio_pci_dev_is_frozen(vdev))
> +		return -EIO;
> +

[Severity: High]
Can a userspace thread race with the freeze operation here?

If a thread locklessly evaluates vfio_pci_dev_is_frozen() as false and then
blocks waiting for vdev->igate or vdev->memory_lock, couldn't
vfio_pci_liveupdate_freeze() concurrently acquire those locks, freeze the
device, set liveupdate_frozen to true, and release the locks?

The ioctl thread would then wake up, acquire the lock, and modify the frozen
device's state.

> diff --git a/drivers/vfio/pci/vfio_pci_liveupdate.c b/drivers/vfio/pci/vfio_pci_liveupdate.c
> index 803528c9f1b27..e6eb7f6b53d7f 100644
> --- a/drivers/vfio/pci/vfio_pci_liveupdate.c
> +++ b/drivers/vfio/pci/vfio_pci_liveupdate.c
[ ... ]
> @@ -6,28 +6,221 @@
> +static int vfio_pci_liveupdate_freeze(struct liveupdate_file_op_args *args)
> +{
[ ... ]
> +	/*
> +	 * Userspace cannot use the FD correctly now irrespective of liveupdate
> +	 * freeze failing or succeeding. They will have to reinitialize the VFIO
> +	 * device to continue using it as reset might have loaded default PCI
> +	 * state. Disable ioctl, read, write and mmap access to the device.
> +	 */
> +	smp_store_release(&vdev->liveupdate_frozen, true);
> +	return 0;
> +}

[Severity: Medium]
Does this permanently brick the VFIO device if a kexec aborts?

The liveupdate_frozen flag is set to true here, but it looks like there is
no corresponding clear on device close/open or kexec abort.

Because the unfreeze operation callback in vfio_pci_liveupdate_file_ops is
unimplemented, the flag would remain set. Userspace attempting to recover
by closing and reopening the FD would continue to receive -EIO until the
driver is unbound.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=6

[PATCH v5 07/20] vfio: Introduce vfio_find_device() helper

[Vipin Sharma]

Introduce vfio_find_device() to discover a registered vfio_device from
vfio_device_class matching user-provided criteria.

To ensure safe lifecycle handling, vfio_find_device() acquires a
registration reference (device->refcount) via
vfio_device_try_get_registration() rather than relying solely on
device.kref taken in class_find_device(). Holding only kref allows
vendor modules to unload while external callers hold device pointers,
leading to a kernel panic on release. Callers must release the device
via vfio_device_put_registration().

Additionally, move refcount_set() prior to vfio_device_add() in
__vfio_register_dev() so device->refcount is initialized before the
device is published to class lookups, avoiding registration data races.

On vfio_device_add() failure, the error path drops the initial
registration reference and waits for completion to drain any reference
taken by a concurrent lookup before returning the error to the driver.

Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/vfio_main.c | 64 +++++++++++++++++++++++++++++++++++++---
 include/linux/vfio.h     |  2 ++
 2 files changed, 62 insertions(+), 4 deletions(-)

diff --git a/drivers/vfio/vfio_main.c b/drivers/vfio/vfio_main.c
index 2437af031169..e91fd83dc58d 100644
--- a/drivers/vfio/vfio_main.c
+++ b/drivers/vfio/vfio_main.c
@@ -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>
@@ -359,17 +360,20 @@ static int __vfio_register_dev(struct vfio_device *device,
 		goto err_out;
 	}
 
-	ret = vfio_device_add(device);
-	if (ret)
-		goto err_out;
-
 	/* Refcounting can't start until the driver calls register */
 	refcount_set(&device->refcount, 1);
 
+	ret = vfio_device_add(device);
+	if (ret)
+		goto err_add;
+
 	vfio_device_group_register(device);
 	vfio_device_debugfs_init(device);
 
 	return 0;
+err_add:
+	vfio_device_put_registration(device);
+	wait_for_completion(&device->comp);
 err_out:
 	vfio_device_remove_group(device);
 	return ret;
@@ -1797,6 +1801,58 @@ int vfio_dma_rw(struct vfio_device *device, dma_addr_t iova, void *data,
 }
 EXPORT_SYMBOL(vfio_dma_rw);
 
+struct vfio_device_match_data {
+	const void *data;
+	device_match_t match;
+};
+
+static int vfio_device_match_fn(struct device *dev, const void *data)
+{
+	struct vfio_device *vdev = container_of(dev, struct vfio_device, device);
+	const struct vfio_device_match_data *mdata = data;
+
+	if (mdata->match(dev, mdata->data)) {
+		if (vfio_device_try_get_registration(vdev))
+			return 1;
+	}
+	return 0;
+}
+
+/**
+ * vfio_find_device - Find a registered VFIO device matching search criteria
+ * @data: Context data passed to the @match function
+ * @match: Device match callback function
+ *
+ * Finds a registered VFIO device in the vfio_device_class matching the search
+ * criteria specified by @match.
+ *
+ * On success, this function acquires a registration reference on the matched
+ * device via vfio_device_try_get_registration(). This pins the device and
+ * prevents its vendor driver module from unloading or unbinding while the
+ * reference is held.
+ *
+ * Return: Pointer to the matching vfio_device on success, or NULL if no
+ * matching device is found. The caller MUST release the acquired registration
+ * reference using vfio_device_put_registration() when finished with the device.
+ */
+struct vfio_device *vfio_find_device(const void *data, device_match_t match)
+{
+	struct vfio_device_match_data mdata = { .data = data, .match = match };
+	struct device *device;
+
+	device = class_find_device(&vfio_device_class, NULL, &mdata, vfio_device_match_fn);
+	if (!device)
+		return NULL;
+
+	/*
+	 * Drop standard kref reference. Caller holds reference from
+	 * vfio_device_try_get_registration().
+	 */
+	put_device(device);
+	return container_of(device, struct vfio_device, device);
+}
+EXPORT_SYMBOL_GPL(vfio_find_device);
+
 /*
  * Module/class support
  */
diff --git a/include/linux/vfio.h b/include/linux/vfio.h
index cf0d941d0410..42c3a9ae3a85 100644
--- a/include/linux/vfio.h
+++ b/include/linux/vfio.h
@@ -430,4 +430,6 @@ 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);
 
+struct vfio_device *vfio_find_device(const void *data, device_match_t match);
+
 #endif /* VFIO_H */
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 08/20] vfio: Refactor vfio_device_fops_cdev_open()

[Vipin Sharma]

Refactor vfio_device_fops_cdev_open() to use a static helper
vfio_device_cdev_open(). This helper will be used in subsequent
patches to support opening the device file during Live Update restore.

No functional change intended.

Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/device_cdev.c | 24 +++++++++++++++---------
 1 file changed, 15 insertions(+), 9 deletions(-)

diff --git a/drivers/vfio/device_cdev.c b/drivers/vfio/device_cdev.c
index 54abf312cf04..2231423fe56d 100644
--- a/drivers/vfio/device_cdev.c
+++ b/drivers/vfio/device_cdev.c
@@ -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)
+static int vfio_device_cdev_open(struct vfio_device *device, struct file *file)
 {
-	struct vfio_device *device = container_of(inode->i_cdev,
-						  struct vfio_device, cdev);
 	struct vfio_device_file *df;
 	int ret;
 
@@ -37,14 +31,14 @@ int vfio_device_fops_cdev_open(struct inode *inode, struct file *filep)
 		goto err_put_registration;
 	}
 
-	filep->private_data = df;
+	file->private_data = df;
 
 	/*
 	 * Use the pseudo fs inode on the device to link all mmaps
 	 * to the same address space, allowing us to unmap all vmas
 	 * associated to this device using unmap_mapping_range().
 	 */
-	filep->f_mapping = device->inode->i_mapping;
+	file->f_mapping = device->inode->i_mapping;
 
 	return 0;
 
@@ -53,6 +47,18 @@ int vfio_device_fops_cdev_open(struct inode *inode, struct file *filep)
 	return ret;
 }
 
+/*
+ * 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 *file)
+{
+	struct vfio_device *device = container_of(inode->i_cdev,
+						  struct vfio_device, cdev);
+
+	return vfio_device_cdev_open(device, file);
+}
+
 static void vfio_df_get_kvm_safe(struct vfio_device_file *df)
 {
 	spin_lock(&df->kvm_ref_lock);
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 09/20] vfio: Add API to open cdev device for Live Update restore

[Vipin Sharma]

During Live Update restore, VFIO PCI driver needs to recreate the device
file context that was preserved before kexec and return it to the user.

Provide an API for VFIO Live Update to get VFIO device file. 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.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/device_cdev.c | 39 ++++++++++++++++++++++++++++++++++++--
 include/linux/vfio.h       |  9 +++++++++
 2 files changed, 46 insertions(+), 2 deletions(-)

diff --git a/drivers/vfio/device_cdev.c b/drivers/vfio/device_cdev.c
index 2231423fe56d..970932f22ecb 100644
--- a/drivers/vfio/device_cdev.c
+++ b/drivers/vfio/device_cdev.c
@@ -2,6 +2,7 @@
 /*
  * Copyright (c) 2023 Intel Corporation.
  */
+#include <linux/anon_inodes.h>
 #include <linux/vfio.h>
 #include <linux/iommufd.h>
 
@@ -16,9 +17,10 @@ void vfio_init_device_cdev(struct vfio_device *device)
 	device->cdev.owner = THIS_MODULE;
 }
 
-static int vfio_device_cdev_open(struct vfio_device *device, struct file *file)
+static int vfio_device_cdev_open(struct vfio_device *device, struct file **filep)
 {
 	struct vfio_device_file *df;
+	struct file *file = *filep;
 	int ret;
 
 	/* Paired with the put in vfio_device_fops_release() */
@@ -31,6 +33,24 @@ static int vfio_device_cdev_open(struct vfio_device *device, struct file *file)
 		goto err_put_registration;
 	}
 
+	/*
+	 * Simulate opening the character device using an anonymous inode. The
+	 * returned file has the same properties as a cdev file (e.g. operations
+	 * are blocked until BIND_IOMMUFD is called).
+	 */
+	if (!file) {
+		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 err_free_device_file;
+		}
+
+		*filep = file;
+	}
+
 	file->private_data = df;
 
 	/*
@@ -42,11 +62,26 @@ static int vfio_device_cdev_open(struct vfio_device *device, struct file *file)
 
 	return 0;
 
+err_free_device_file:
+	kfree(df);
 err_put_registration:
 	vfio_device_put_registration(device);
 	return ret;
 }
 
+struct file *vfio_device_liveupdate_cdev_open(struct vfio_device *device)
+{
+	struct file *file = NULL;
+	int ret;
+
+	ret = vfio_device_cdev_open(device, &file);
+	if (ret)
+		return ERR_PTR(ret);
+
+	return file;
+}
+EXPORT_SYMBOL_GPL(vfio_device_liveupdate_cdev_open);
+
 /*
  * device access via the fd opened by this function is blocked until
  * .open_device() is called successfully during BIND_IOMMUFD.
@@ -56,7 +91,7 @@ int vfio_device_fops_cdev_open(struct inode *inode, struct file *file)
 	struct vfio_device *device = container_of(inode->i_cdev,
 						  struct vfio_device, cdev);
 
-	return vfio_device_cdev_open(device, file);
+	return vfio_device_cdev_open(device, &file);
 }
 
 static void vfio_df_get_kvm_safe(struct vfio_device_file *df)
diff --git a/include/linux/vfio.h b/include/linux/vfio.h
index 42c3a9ae3a85..b896c64b11da 100644
--- a/include/linux/vfio.h
+++ b/include/linux/vfio.h
@@ -430,6 +430,15 @@ 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)
+struct file *vfio_device_liveupdate_cdev_open(struct vfio_device *device);
+#else
+static inline struct file *vfio_device_liveupdate_cdev_open(struct vfio_device *device)
+{
+	return ERR_PTR(-EOPNOTSUPP);
+}
+#endif /* IS_ENABLED(CONFIG_VFIO_DEVICE_CDEV) */
+
 struct vfio_device *vfio_find_device(const void *data, device_match_t match);
 
 #endif /* VFIO_H */
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 10/20] vfio/pci: Retrieve preserved device files after Live Update

[Vipin Sharma]

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

During retrieve, find the VFIO device using the BDF and domain
information from the serialized state, and open it using the new
vfio_device_liveupdate_cdev_open() API.

Disallow Live Update to finish if userspace hasn't opened the file
successfully or any error occurred during retrieve.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/pci/vfio_pci_liveupdate.c | 64 +++++++++++++++++++++++++-
 1 file changed, 62 insertions(+), 2 deletions(-)

diff --git a/drivers/vfio/pci/vfio_pci_liveupdate.c b/drivers/vfio/pci/vfio_pci_liveupdate.c
index e6eb7f6b53d7..4baac80d03f0 100644
--- a/drivers/vfio/pci/vfio_pci_liveupdate.c
+++ b/drivers/vfio/pci/vfio_pci_liveupdate.c
@@ -39,7 +39,13 @@
  *    preserved, so there is no way for the file to be destroyed or the device
  *    to be unbound from the vfio-pci driver while it is preserved.
  *
- * Retrieving the file after kexec is not yet supported.
+ * After kexec, the preserved VFIO device file can be retrieved from the session
+ * just like any other preserved file::
+ *
+ *   ioctl(session_fd, LIVEUPDATE_SESSION_RETRIEVE_FD, &arg);
+ *   device_fd = arg.fd;
+ *   ...
+ *   ioctl(session_fd, LIVEUPDATE_SESSION_FINISH, ...);
  *
  * Restrictions
  * ============
@@ -93,6 +99,7 @@
 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
 #include <linux/errno.h>
+#include <linux/file.h>
 #include <linux/kexec_handover.h>
 #include <linux/kho/abi/vfio_pci.h>
 #include <linux/liveupdate.h>
@@ -223,13 +230,65 @@ static int vfio_pci_liveupdate_freeze(struct liveupdate_file_op_args *args)
 	return 0;
 }
 
+static int match_device(struct device *dev, const void *arg)
+{
+	struct vfio_device *device = container_of(dev, struct vfio_device, device);
+	const struct vfio_pci_core_device_ser *ser = arg;
+	struct pci_dev *pdev;
+
+	pdev = dev_is_pci(device->dev) ? to_pci_dev(device->dev) : NULL;
+	if (!pdev)
+		return false;
+
+	return ser->bdf == pci_dev_id(pdev) && ser->domain == pci_domain_nr(pdev->bus);
+}
+
 static int vfio_pci_liveupdate_retrieve(struct liveupdate_file_op_args *args)
 {
-	return -EOPNOTSUPP;
+	struct vfio_pci_core_device_ser *ser;
+	struct vfio_device *device;
+	struct file *file;
+	int ret = 0;
+
+	ser = phys_to_virt(args->serialized_data);
+
+	device = vfio_find_device(ser, match_device);
+	if (!device)
+		return -ENODEV;
+
+	file = vfio_device_liveupdate_cdev_open(device);
+	if (IS_ERR(file)) {
+		ret = PTR_ERR(file);
+		goto out;
+	}
+
+	args->file = file;
+out:
+	/* Drop the reference from vfio_find_device() */
+	vfio_device_put_registration(device);
+	return ret;
+
+}
+
+
+static bool vfio_pci_liveupdate_can_finish(struct liveupdate_file_op_args *args)
+{
+	struct vfio_device *device;
+
+	if (args->retrieve_status <= 0)
+		return false;
+
+	device = vfio_device_from_file(args->file);
+	guard(mutex)(&device->dev_set->lock);
+	return vfio_device_cdev_opened(device);
 }
 
 static void vfio_pci_liveupdate_finish(struct liveupdate_file_op_args *args)
 {
+	struct vfio_device *device = vfio_device_from_file(args->file);
+
+	pci_liveupdate_finish(to_pci_dev(device->dev));
+	kho_restore_free(phys_to_virt(args->serialized_data));
 }
 
 static const struct liveupdate_file_ops vfio_pci_liveupdate_file_ops = {
@@ -238,6 +297,7 @@ static const struct liveupdate_file_ops vfio_pci_liveupdate_file_ops = {
 	.unpreserve = vfio_pci_liveupdate_unpreserve,
 	.freeze = vfio_pci_liveupdate_freeze,
 	.retrieve = vfio_pci_liveupdate_retrieve,
+	.can_finish = vfio_pci_liveupdate_can_finish,
 	.finish = vfio_pci_liveupdate_finish,
 	.owner = THIS_MODULE,
 };
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 11/20] vfio: Enforce preserved devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

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.

Reviewed-by: Pranjal Shrivastava <praan@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Co-developed-by: Vipin Sharma <vipinsh@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 drivers/vfio/device_cdev.c             |  8 ++++++++
 drivers/vfio/group.c                   |  9 +++++++++
 drivers/vfio/pci/vfio_pci_liveupdate.c |  6 ++++++
 drivers/vfio/vfio.h                    | 18 ++++++++++++++++++
 4 files changed, 41 insertions(+)

diff --git a/drivers/vfio/device_cdev.c b/drivers/vfio/device_cdev.c
index 970932f22ecb..0050f0fe456b 100644
--- a/drivers/vfio/device_cdev.c
+++ b/drivers/vfio/device_cdev.c
@@ -49,6 +49,14 @@ static int vfio_device_cdev_open(struct vfio_device *device, struct file **filep
 		}
 
 		*filep = file;
+	} else if (vfio_liveupdate_incoming_is_preserved(device)) {
+		/*
+		 * Since it is live update preserved device, it must be
+		 * retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD instead of
+		 * opening /dev/vfio/devices/vfioX.
+		 */
+		ret = -EBUSY;
+		goto err_free_device_file;
 	}
 
 	file->private_data = df;
diff --git a/drivers/vfio/group.c b/drivers/vfio/group.c
index b2299e5bc6df..62b4eaabc829 100644
--- a/drivers/vfio/group.c
+++ b/drivers/vfio/group.c
@@ -316,6 +316,15 @@ 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)) {
+		vfio_device_put_registration(device);
+		return -EBUSY;
+	}
+
 	fd = FD_ADD(O_CLOEXEC, vfio_device_open_file(device));
 	if (fd < 0)
 		vfio_device_put_registration(device);
diff --git a/drivers/vfio/pci/vfio_pci_liveupdate.c b/drivers/vfio/pci/vfio_pci_liveupdate.c
index 4baac80d03f0..f0ea37d98696 100644
--- a/drivers/vfio/pci/vfio_pci_liveupdate.c
+++ b/drivers/vfio/pci/vfio_pci_liveupdate.c
@@ -47,6 +47,12 @@
  *   ...
  *   ioctl(session_fd, LIVEUPDATE_SESSION_FINISH, ...);
  *
+ * .. note::
+ *    After kexec, if a device was preserved by the previous kernel, attempting
+ *    to open a new file for the device via its character device
+ *    (``/dev/vfio/devices/X``) or via ``VFIO_GROUP_GET_DEVICE_FD`` will fail
+ *    with ``-EBUSY``.
+ *
  * Restrictions
  * ============
  *
diff --git a/drivers/vfio/vfio.h b/drivers/vfio/vfio.h
index e4b72e79b7e3..496180950eb6 100644
--- a/drivers/vfio/vfio.h
+++ b/drivers/vfio/vfio.h
@@ -11,6 +11,7 @@
 #include <linux/cdev.h>
 #include <linux/module.h>
 #include <linux/vfio.h>
+#include <linux/pci.h>
 
 struct iommufd_ctx;
 struct iommu_group;
@@ -461,4 +462,21 @@ static inline void vfio_device_debugfs_init(struct vfio_device *vdev) { }
 static inline void vfio_device_debugfs_exit(struct vfio_device *vdev) { }
 #endif /* CONFIG_VFIO_DEBUGFS */
 
+#ifdef CONFIG_VFIO_PCI_LIVEUPDATE
+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_is_incoming(to_pci_dev(d));
+
+	return false;
+}
+#else
+static inline bool vfio_liveupdate_incoming_is_preserved(struct vfio_device *device)
+{
+	return false;
+}
+#endif /* CONFIG_VFIO_PCI_LIVEUPDATE */
+
 #endif
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 11/20] vfio: Enforce preserved devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] The function `pci_liveupdate_is_incoming()` is called but is neither declared nor defined anywhere in the kernel tree.
--

--- Patch [11]: [PATCH v5 11/20] vfio: Enforce preserved devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD ---
commit 27d87a41631434a0adb6626f478ee97f35b6c917
Author: David Matlack <dmatlack@google.com>

vfio: Enforce preserved devices are retrieved via LIVEUPDATE_SESSION_RETRIEVE_FD

This commit enforces that incoming VFIO devices preserved by a previous
kernel are retrieved using LIVEUPDATE_SESSION_RETRIEVE_FD. It adds checks
during device open and group device fd retrieval to return -EBUSY if
an attempt is made to bypass the preserved state.

> diff --git a/drivers/vfio/vfio.h b/drivers/vfio/vfio.h
> index e4b72e79b7e3e..496180950eb6d 100644
> --- a/drivers/vfio/vfio.h
> +++ b/drivers/vfio/vfio.h

[ ... ]

> @@ -461,4 +462,21 @@ static inline void vfio_device_debugfs_init(struct vfio_device *vdev) { }
>  static inline void vfio_device_debugfs_exit(struct vfio_device *vdev) { }
>  #endif /* CONFIG_VFIO_DEBUGFS */
>  
> +#ifdef CONFIG_VFIO_PCI_LIVEUPDATE
> +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_is_incoming(to_pci_dev(d));

[Severity: Low]
Does this code compile successfully when CONFIG_VFIO_PCI_LIVEUPDATE is enabled?
It appears that pci_liveupdate_is_incoming is neither declared nor defined
anywhere in the kernel tree.

Could a prerequisite patch adding this functionality to the PCI subsystem
have been left out of this series?

> +
> +	return false;
> +}

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=11

[PATCH v5 12/20] docs: liveupdate: Add documentation for VFIO PCI

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

Add documentation for preserving VFIO device files across a Live Update,
generated from the kernel-doc comments in the code.

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 Documentation/core-api/liveupdate.rst         |  1 +
 .../driver-api/vfio_pci_liveupdate.rst        | 23 +++++++++++++++++++
 MAINTAINERS                                   |  1 +
 3 files changed, 25 insertions(+)
 create mode 100644 Documentation/driver-api/vfio_pci_liveupdate.rst

diff --git a/Documentation/core-api/liveupdate.rst b/Documentation/core-api/liveupdate.rst
index 2bce2644eba2..20f1d2e20508 100644
--- a/Documentation/core-api/liveupdate.rst
+++ b/Documentation/core-api/liveupdate.rst
@@ -38,6 +38,7 @@ The following types of file descriptors can be preserved
    :maxdepth: 1
 
    ../mm/memfd_preservation
+   ../driver-api/vfio_pci_liveupdate
 
 Public API
 ==========
diff --git a/Documentation/driver-api/vfio_pci_liveupdate.rst b/Documentation/driver-api/vfio_pci_liveupdate.rst
new file mode 100644
index 000000000000..1098b84e5ecd
--- /dev/null
+++ b/Documentation/driver-api/vfio_pci_liveupdate.rst
@@ -0,0 +1,23 @@
+.. SPDX-License-Identifier: GPL-2.0-or-later
+
+====================================
+VFIO PCI Device Preservation via LUO
+====================================
+
+.. kernel-doc:: drivers/vfio/pci/vfio_pci_liveupdate.c
+   :doc: VFIO PCI Preservation via LUO
+
+VFIO PCI Preservation ABI
+=========================
+
+.. kernel-doc:: include/linux/kho/abi/vfio_pci.h
+   :doc: VFIO PCI Live Update ABI
+
+.. kernel-doc:: include/linux/kho/abi/vfio_pci.h
+   :internal:
+
+See Also
+========
+
+- :doc:`/core-api/liveupdate`
+- :doc:`/core-api/kho/index`
diff --git a/MAINTAINERS b/MAINTAINERS
index 77a35742de0a..43b3165d1c68 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -28328,6 +28328,7 @@ T:	git https://github.com/awilliam/linux-vfio.git
 F:	Documentation/ABI/testing/debugfs-vfio
 F:	Documentation/ABI/testing/sysfs-devices-vfio-dev
 F:	Documentation/driver-api/vfio.rst
+F:	Documentation/driver-api/vfio_pci_liveupdate.rst
 F:	drivers/vfio/
 F:	include/linux/kho/abi/vfio_pci.h
 F:	include/linux/vfio.h
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 13/20] vfio: selftests: Build liveupdate library in VFIO selftests

[Vipin Sharma]

Import and build liveupdate selftest library in VFIO selftests.

It allows using liveupdate ioctls in VFIO selftests

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 tools/testing/selftests/vfio/Makefile | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/vfio/Makefile b/tools/testing/selftests/vfio/Makefile
index e6e8cb52ab03..4c8a30f950c4 100644
--- a/tools/testing/selftests/vfio/Makefile
+++ b/tools/testing/selftests/vfio/Makefile
@@ -21,6 +21,7 @@ TEST_FILES += scripts/setup.sh
 
 include ../lib.mk
 include lib/libvfio.mk
+include ../liveupdate/lib/libliveupdate.mk
 
 CFLAGS += -I$(top_srcdir)/tools/include
 CFLAGS += -MD
@@ -31,14 +32,17 @@ LDFLAGS += -pthread
 
 LDLIBS += -luuid
 
-$(TEST_GEN_PROGS): $(OUTPUT)/%: $(OUTPUT)/%.o $(LIBVFIO_O)
-	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $< $(LIBVFIO_O) $(LDLIBS) -o $@
+LIBS_O := $(LIBVFIO_O)
+LIBS_O += $(LIBLIVEUPDATE_O)
+
+$(TEST_GEN_PROGS): $(OUTPUT)/%: $(OUTPUT)/%.o $(LIBS_O)
+	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $< $(LIBS_O) $(LDLIBS) -o $@
 
 TEST_GEN_PROGS_O = $(patsubst %, %.o, $(TEST_GEN_PROGS))
 $(TEST_GEN_PROGS_O): $(OUTPUT)/%.o: %.c
 	$(CC) $(CFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c $< -o $@
 
-TEST_DEP_FILES = $(patsubst %.o, %.d, $(TEST_GEN_PROGS_O) $(LIBVFIO_O))
+TEST_DEP_FILES = $(patsubst %.o, %.d, $(TEST_GEN_PROGS_O) $(LIBS_O))
 -include $(TEST_DEP_FILES)
 
 EXTRA_CLEAN += $(TEST_GEN_PROGS_O) $(TEST_DEP_FILES)
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 13/20] vfio: selftests: Build liveupdate library in VFIO selftests

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] The VFIO selftests Makefile includes a non-existent file `../liveupdate/lib/libliveupdate.mk`, which breaks the build.
--

--- Patch [13]: [PATCH v5 13/20] vfio: selftests: Build liveupdate library in VFIO selftests ---
commit 8a41a1e6305bdbe73648566ef65764a964e68dbb
Author: Vipin Sharma <vipinsh@google.com>

vfio: selftests: Build liveupdate library in VFIO selftests

This commit imports and builds the liveupdate selftest library within the
VFIO selftests to allow using liveupdate ioctls.

> diff --git a/tools/testing/selftests/vfio/Makefile b/tools/testing/selftests/vfio/Makefile
> --- a/tools/testing/selftests/vfio/Makefile
> +++ b/tools/testing/selftests/vfio/Makefile
> @@ -21,6 +21,7 @@ TEST_FILES += scripts/setup.sh
>  
>  include ../lib.mk
>  include lib/libvfio.mk
> +include ../liveupdate/lib/libliveupdate.mk

[Severity: Low]
Does this file exist in the tree? Including this missing file causes a fatal
build error when running make in the vfio selftests directory because
../liveupdate/lib/libliveupdate.mk cannot be found. Was the liveupdate
library intended to be added in a prerequisite patch that was missed?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=13

[PATCH v5 14/20] vfio: selftests: Add vfio_pci_liveupdate_uapi_test

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

Add a selftest to exercise preserving 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.

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 tools/testing/selftests/vfio/Makefile         |  1 +
 .../vfio/vfio_pci_liveupdate_uapi_test.c      | 97 +++++++++++++++++++
 2 files changed, 98 insertions(+)
 create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c

diff --git a/tools/testing/selftests/vfio/Makefile b/tools/testing/selftests/vfio/Makefile
index 4c8a30f950c4..dc06fddcf855 100644
--- a/tools/testing/selftests/vfio/Makefile
+++ b/tools/testing/selftests/vfio/Makefile
@@ -13,6 +13,7 @@ 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_sriov_uapi_test
+TEST_GEN_PROGS += vfio_pci_liveupdate_uapi_test
 
 TEST_FILES += scripts/cleanup.sh
 TEST_FILES += scripts/lib.sh
diff --git a/tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c b/tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c
new file mode 100644
index 000000000000..f81c8ab21db3
--- /dev/null
+++ b/tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c
@@ -0,0 +1,97 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <libliveupdate.h>
+#include <libvfio.h>
+#include <kselftest_harness.h>
+
+static const char *device_bdf;
+
+FIXTURE(vfio_pci_liveupdate_uapi_test) {
+	int luo_fd;
+	int session_fd;
+	struct iommu *iommu;
+	struct vfio_pci_device *device;
+};
+
+FIXTURE_VARIANT(vfio_pci_liveupdate_uapi_test) {
+	const char *iommu_mode;
+};
+
+#define FIXTURE_VARIANT_ADD_IOMMU_MODE(_iommu_mode)			\
+FIXTURE_VARIANT_ADD(vfio_pci_liveupdate_uapi_test, _iommu_mode) {	\
+	.iommu_mode = #_iommu_mode,					\
+}
+
+FIXTURE_VARIANT_ADD_ALL_IOMMU_MODES();
+#undef FIXTURE_VARIANT_ADD_IOMMU_MODE
+
+FIXTURE_SETUP(vfio_pci_liveupdate_uapi_test)
+{
+	self->luo_fd = luo_open_device();
+	ASSERT_GT(self->luo_fd, 0);
+
+	self->session_fd = luo_create_session(self->luo_fd, "session");
+	ASSERT_GT(self->session_fd, 0);
+
+	self->iommu = iommu_init(variant->iommu_mode);
+	self->device = vfio_pci_device_init(device_bdf, self->iommu);
+}
+
+FIXTURE_TEARDOWN(vfio_pci_liveupdate_uapi_test)
+{
+	if (self->device)
+		vfio_pci_device_cleanup(self->device);
+	if (self->iommu)
+		iommu_cleanup(self->iommu);
+	if (self->session_fd > 0)
+		close(self->session_fd);
+	if (self->luo_fd > 0)
+		close(self->luo_fd);
+}
+
+TEST_F(vfio_pci_liveupdate_uapi_test, preserve_device)
+{
+	int ret;
+
+	ret = luo_session_preserve_fd(self->session_fd, self->device->fd, 0);
+
+	/* Preservation should only be supported for VFIO cdev files. */
+	ASSERT_EQ(ret, self->iommu->iommufd ? 0 : -ENOENT);
+}
+
+TEST_F(vfio_pci_liveupdate_uapi_test, preserve_group_fails)
+{
+	int ret;
+
+	if (self->iommu->iommufd)
+		SKIP(return, "iommufd-mode does not have group files");
+
+	ret = luo_session_preserve_fd(self->session_fd, self->device->group_fd, 0);
+	ASSERT_EQ(ret, -ENOENT);
+}
+
+TEST_F(vfio_pci_liveupdate_uapi_test, preserve_container_fails)
+{
+	int ret;
+
+	if (self->iommu->iommufd)
+		SKIP(return, "iommufd-mode does not have container files");
+
+	ret = luo_session_preserve_fd(self->session_fd, self->iommu->container_fd, 0);
+	ASSERT_EQ(ret, -ENOENT);
+}
+
+int main(int argc, char *argv[])
+{
+	int fd;
+
+	fd = luo_open_device();
+	if (fd < 0)
+		ksft_exit_skip("open(%s) failed: %s, skipping\n",
+			       LUO_DEVICE, strerror(errno));
+
+	close(fd);
+
+	device_bdf = vfio_selftests_get_bdf(&argc, argv);
+	return test_harness_run(argc, argv);
+}
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 14/20] vfio: selftests: Add vfio_pci_liveupdate_uapi_test

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] The selftest `vfio_pci_liveupdate_uapi_test.c` fails to compile and link due to missing headers, libraries, and undefined functions.
--

--- Patch [14]: [PATCH v5 14/20] vfio: selftests: Add vfio_pci_liveupdate_uapi_test ---
commit 4ef2d85343eb23f4f743330b67aabe0e09e08f3d
Author: David Matlack <dmatlack@google.com>

vfio: selftests: Add vfio_pci_liveupdate_uapi_test

This commit adds a selftest to exercise preserving various VFIO files
through /dev/liveupdate. It verifies that VFIO cdev device files can be
preserved while group-based device files, group files, and container files
fail.

> diff --git a/tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c b/tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c
> new file mode 100644
> index 0000000000000..f81c8ab21db3f
> --- /dev/null
> +++ b/tools/testing/selftests/vfio/vfio_pci_liveupdate_uapi_test.c
> @@ -0,0 +1,97 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +
> +#include <libliveupdate.h>

[Severity: Low]
Does this header file exist in the tree? The build fails because
<libliveupdate.h> cannot be found. 

Additionally, is the Makefile missing an include for
../liveupdate/lib/libliveupdate.mk to ensure the dependencies are properly
pulled in for the build?

> +#include <libvfio.h>
> +#include <kselftest_harness.h>

[ ... ]

> +TEST_F(vfio_pci_liveupdate_uapi_test, preserve_device)
> +{
> +	int ret;
> +
> +	ret = luo_session_preserve_fd(self->session_fd, self->device->fd, 0);

[Severity: Low]
Is this calling a kernel-space function from a user-space selftest?

Looking at the code, luo_session_preserve_fd() appears to be defined in
kernel/liveupdate/luo_session.c. Since this is a user-space selftest,
calling it here results in an undefined reference during compilation. 

Should this be invoking an ioctl directly, or is there a user-space library
wrapper intended to be used instead?

> +
> +	/* Preservation should only be supported for VFIO cdev files. */
> +	ASSERT_EQ(ret, self->iommu->iommufd ? 0 : -ENOENT);
> +}

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=14

[PATCH v5 15/20] vfio: selftests: Initialize vfio_pci_device using a VFIO cdev FD

[Vipin Sharma]

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 writing selftests where VFIO device FD is preserved
using liveupdate and retrieved later using liveupdate ioctl after kexec.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 .../lib/include/libvfio/vfio_pci_device.h     |  3 +++
 .../selftests/vfio/lib/vfio_pci_device.c      | 25 ++++++++++++++-----
 2 files changed, 22 insertions(+), 6 deletions(-)

diff --git a/tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h b/tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h
index 3eabead717bb..66fc29662daa 100644
--- a/tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h
+++ b/tools/testing/selftests/vfio/lib/include/libvfio/vfio_pci_device.h
@@ -40,6 +40,9 @@ struct vfio_pci_device {
 
 struct vfio_pci_device *vfio_pci_device_alloc(const char *bdf, struct iommu *iommu);
 void vfio_pci_device_free(struct vfio_pci_device *device);
+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);
 
diff --git a/tools/testing/selftests/vfio/lib/vfio_pci_device.c b/tools/testing/selftests/vfio/lib/vfio_pci_device.c
index 94dc5fcecbeb..a6e0e7c6e18e 100644
--- a/tools/testing/selftests/vfio/lib/vfio_pci_device.c
+++ b/tools/testing/selftests/vfio/lib/vfio_pci_device.c
@@ -414,9 +414,13 @@ void vfio_pci_cdev_open(struct vfio_pci_device *device, const char *bdf)
 }
 
 static void vfio_pci_iommufd_setup(struct vfio_pci_device *device,
-				   const char *bdf, const char *vf_token)
+				   const char *bdf, int device_fd,
+				   const char *vf_token)
 {
-	vfio_pci_cdev_open(device, bdf);
+	if (device_fd >= 0)
+		device->fd = device_fd;
+	else
+		vfio_pci_cdev_open(device, bdf);
 	vfio_device_bind_iommufd(device->fd, device->iommu->iommufd, vf_token);
 	vfio_device_attach_iommufd_pt(device->fd, device->iommu->ioas_id);
 }
@@ -440,16 +444,20 @@ void vfio_pci_device_free(struct vfio_pci_device *device)
 	free(device);
 }
 
-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;
 
 	device = vfio_pci_device_alloc(bdf, iommu);
 
-	if (iommu->mode->container_path)
+	if (iommu->mode->container_path) {
+		VFIO_ASSERT_EQ(device_fd, -1);
 		vfio_pci_container_setup(device, bdf, NULL);
-	else
-		vfio_pci_iommufd_setup(device, bdf, NULL);
+	} else {
+		vfio_pci_iommufd_setup(device, bdf, device_fd, NULL);
+	}
 
 	vfio_pci_device_setup(device);
 	vfio_pci_driver_probe(device);
@@ -457,6 +465,11 @@ 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;
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 16/20] vfio: selftests: Add Makefile support for TEST_GEN_PROGS_EXTENDED

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

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.

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 tools/testing/selftests/vfio/Makefile | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/vfio/Makefile b/tools/testing/selftests/vfio/Makefile
index dc06fddcf855..591f1a52c310 100644
--- a/tools/testing/selftests/vfio/Makefile
+++ b/tools/testing/selftests/vfio/Makefile
@@ -36,16 +36,18 @@ LDLIBS += -luuid
 LIBS_O := $(LIBVFIO_O)
 LIBS_O += $(LIBLIVEUPDATE_O)
 
-$(TEST_GEN_PROGS): $(OUTPUT)/%: $(OUTPUT)/%.o $(LIBS_O)
+$(TEST_GEN_PROGS) $(TEST_GEN_PROGS_EXTENDED): $(OUTPUT)/%: $(OUTPUT)/%.o $(LIBS_O)
 	$(CC) $(CFLAGS) $(CPPFLAGS) $(LDFLAGS) $< $(LIBS_O) $(LDLIBS) -o $@
 
-TEST_GEN_PROGS_O = $(patsubst %, %.o, $(TEST_GEN_PROGS))
-$(TEST_GEN_PROGS_O): $(OUTPUT)/%.o: %.c
+TESTS_O := $(patsubst %, %.o, $(TEST_GEN_PROGS))
+TESTS_O += $(patsubst %, %.o, $(TEST_GEN_PROGS_EXTENDED))
+
+$(TESTS_O): $(OUTPUT)/%.o: %.c
 	$(CC) $(CFLAGS) $(CPPFLAGS) $(TARGET_ARCH) -c $< -o $@
 
-TEST_DEP_FILES = $(patsubst %.o, %.d, $(TEST_GEN_PROGS_O) $(LIBS_O))
+TEST_DEP_FILES = $(patsubst %.o, %.d, $(TESTS_O) $(LIBS_O))
 -include $(TEST_DEP_FILES)
 
-EXTRA_CLEAN += $(TEST_GEN_PROGS_O) $(TEST_DEP_FILES)
+EXTRA_CLEAN += $(TESTS_O) $(TEST_DEP_FILES)
 
 endif
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 17/20] vfio: selftests: Add vfio_pci_liveupdate_kexec_test

[Vipin Sharma]

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.

Co-developed-by: David Matlack <dmatlack@google.com>
Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 tools/testing/selftests/vfio/Makefile         |  4 +
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 89 +++++++++++++++++++
 2 files changed, 93 insertions(+)
 create mode 100644 tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c

diff --git a/tools/testing/selftests/vfio/Makefile b/tools/testing/selftests/vfio/Makefile
index 591f1a52c310..a2f0a459e1f4 100644
--- a/tools/testing/selftests/vfio/Makefile
+++ b/tools/testing/selftests/vfio/Makefile
@@ -15,6 +15,10 @@ TEST_GEN_PROGS += vfio_pci_driver_test
 TEST_GEN_PROGS += vfio_pci_sriov_uapi_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_FILES += scripts/cleanup.sh
 TEST_FILES += scripts/lib.sh
 TEST_FILES += scripts/run.sh
diff --git a/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
new file mode 100644
index 000000000000..15b3e3af91d1
--- /dev/null
+++ b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
@@ -0,0 +1,89 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+#include <libliveupdate.h>
+#include <libvfio.h>
+
+static const char *device_bdf;
+
+static char state_session[LIVEUPDATE_SESSION_NAME_LENGTH];
+static char device_session[LIVEUPDATE_SESSION_NAME_LENGTH];
+
+enum {
+	STATE_TOKEN,
+	DEVICE_TOKEN,
+};
+
+static void before_kexec(int luo_fd)
+{
+	struct vfio_pci_device *device;
+	struct iommu *iommu;
+	int session_fd;
+	int ret;
+
+	iommu = iommu_init("iommufd");
+	device = vfio_pci_device_init(device_bdf, iommu);
+
+	create_state_file(luo_fd, state_session, STATE_TOKEN, /*next_stage=*/2);
+
+	session_fd = luo_create_session(luo_fd, device_session);
+	VFIO_ASSERT_GE(session_fd, 0);
+
+	printf("Preserving device in session\n");
+	ret = luo_session_preserve_fd(session_fd, device->fd, DEVICE_TOKEN);
+	VFIO_ASSERT_EQ(ret, 0);
+
+	close(luo_fd);
+	daemonize_and_wait();
+}
+
+static void after_kexec(int luo_fd, int state_session_fd)
+{
+	struct vfio_pci_device *device;
+	struct iommu *iommu;
+	int session_fd;
+	int device_fd;
+	int stage;
+
+	restore_and_read_stage(state_session_fd, STATE_TOKEN, &stage);
+	VFIO_ASSERT_EQ(stage, 2);
+
+	session_fd = luo_retrieve_session(luo_fd, device_session);
+	VFIO_ASSERT_GE(session_fd, 0);
+
+	printf("Finishing the session before retrieving the device (should fail)\n");
+	VFIO_ASSERT_NE(luo_session_finish(session_fd), 0);
+
+	printf("Retrieving the device FD from LUO\n");
+	device_fd = luo_session_retrieve_fd(session_fd, DEVICE_TOKEN);
+	VFIO_ASSERT_GE(device_fd, 0);
+
+	printf("Finishing the session before binding to iommufd (should fail)\n");
+	VFIO_ASSERT_NE(luo_session_finish(session_fd), 0);
+
+	printf("Binding the device to an iommufd and setting it up\n");
+	iommu = iommu_init("iommufd");
+
+	/*
+	 * This will invoke various ioctls on device_fd such as
+	 * VFIO_DEVICE_GET_INFO. So this is a decent sanity test
+	 * that LUO actually handed us back a valid VFIO device
+	 * file and not something else.
+	 */
+	device = __vfio_pci_device_init(device_bdf, iommu, device_fd);
+
+	printf("Finishing the session\n");
+	VFIO_ASSERT_EQ(luo_session_finish(session_fd), 0);
+
+	vfio_pci_device_cleanup(device);
+	iommu_cleanup(iommu);
+}
+
+int main(int argc, char *argv[])
+{
+	device_bdf = vfio_selftests_get_bdf(&argc, argv);
+
+	sprintf(device_session, "device-%s", device_bdf);
+	sprintf(state_session, "state-%s", device_bdf);
+
+	return luo_test(argc, argv, state_session, before_kexec, after_kexec);
+}
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 17/20] vfio: selftests: Add vfio_pci_liveupdate_kexec_test

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] Missing header and Makefile dependencies for liveupdate library.
--

--- Patch [17]: [PATCH v5 17/20] vfio: selftests: Add vfio_pci_liveupdate_kexec_test ---
commit 5f1d9fcb84decd9a67e94d71a492e75ba0113c1b
Author: Vipin Sharma <vipinsh@google.com>

vfio: selftests: Add vfio_pci_liveupdate_kexec_test

This commit adds a new selftest to exercise preserving a vfio-pci device
across a Live Update. The test verifies that the device file can be
preserved before kexec and retrieved afterward.

> diff --git a/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
> new file mode 100644
> index 0000000000000..15b3e3af91d1e
> --- /dev/null
> +++ b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
> @@ -0,0 +1,89 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +
> +#include <libliveupdate.h>

[Severity: Low]
Does this header file exist in the tree? It looks like this could cause a
build failure because libliveupdate.h cannot be found. The other liveupdate
selftests appear to rely on luo_test_utils.h or direct ioctl calls instead
of a libliveupdate library.

[ ... ]

> +static void before_kexec(int luo_fd)
> +{
> +	struct vfio_pci_device *device;

[ ... ]

> +	printf("Preserving device in session\n");
> +	ret = luo_session_preserve_fd(session_fd, device->fd, DEVICE_TOKEN);

[Severity: Low]
Will this result in a linker error? It seems luo_session_preserve_fd is
undefined without the accompanying library implementation. Should this test
use direct ioctls or are there missing Makefile dependencies for a library?

[ ... ]

> +static void after_kexec(int luo_fd, int state_session_fd)
> +{
> +	struct vfio_pci_device *device;

[ ... ]

> +	printf("Retrieving the device FD from LUO\n");
> +	device_fd = luo_session_retrieve_fd(session_fd, DEVICE_TOKEN);

[Severity: Low]
Similarly here in after_kexec(), is luo_session_retrieve_fd implemented
elsewhere in the test suite? It looks like this will also fail to link.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=17

[PATCH v5 18/20] vfio: selftests: Expose iommu_modes to tests

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

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

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 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(-)

diff --git a/tools/testing/selftests/vfio/lib/include/libvfio/iommu.h b/tools/testing/selftests/vfio/lib/include/libvfio/iommu.h
index e9a3386a4719..4b9cbe262159 100644
--- a/tools/testing/selftests/vfio/lib/include/libvfio/iommu.h
+++ b/tools/testing/selftests/vfio/lib/include/libvfio/iommu.h
@@ -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 {
diff --git a/tools/testing/selftests/vfio/lib/iommu.c b/tools/testing/selftests/vfio/lib/iommu.c
index 035dac069d60..95a494f829d2 100644
--- a/tools/testing/selftests/vfio/lib/iommu.c
+++ b/tools/testing/selftests/vfio/lib/iommu.c
@@ -23,7 +23,7 @@
 const char *default_iommu_mode = 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 = MODE_VFIO_TYPE1_IOMMU,
 		.container_path = "/dev/vfio/vfio",
@@ -49,6 +49,8 @@ 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;
-- 
2.55.0.795.g602f6c329a-goog

[PATCH v5 19/20] vfio: selftests: Verify that opening VFIO device fails during Live Update

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

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 preserved across a
Live Update, the file must be retrieved from /dev/liveupdate, not from
VFIO directly.

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 38 +++++++++++++++++++
 1 file changed, 38 insertions(+)

diff --git a/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
index 15b3e3af91d1..ec3fe8f19502 100644
--- a/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
+++ b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
@@ -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, device_bdf);
+		vfio_container_set_iommu(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);
 
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 19/20] vfio: selftests: Verify that opening VFIO device fails during Live Update

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [Medium] Test assertion failure due to IOMMU group already being claimed by preserved cdev
- [Medium] Test assertion failure due to re-binding already bound preserved cdev
- [Low] Abstraction violation using free() instead of vfio_pci_device_free()
--

--- Patch [19]: [PATCH v5 19/20] vfio: selftests: Verify that opening VFIO device fails during Live Update ---
Note: Due to recitation restrictions, this report uses a free-form summary
of the findings rather than quoting the original patch diff.

commit b5a56d527567933e2f5fd3077aed99e8404f21c3
Author: David Matlack <dmatlack@google.com>
Subject: vfio: selftests: Verify that opening VFIO device fails during Live Update

This patch verifies 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.

[Severity: Medium]
In check_open_vfio_device_fails(), the test loops over legacy modes and
calls vfio_pci_group_setup(). Since the device is already opened via cdev
before kexec, the group's cdev_device_open_cnt is greater than zero. 

Does this cause the group open attempt inside vfio_pci_group_setup() to
fail with EBUSY? If so, it looks like this will trigger the VFIO_ASSERT_GE()
assertion on the group_fd, which would abort the test before it can even
reach the VFIO_GROUP_GET_DEVICE_FD ioctl it is intended to verify.

[Severity: Medium]
In after_kexec(), the test retrieves the preserved device_fd and attempts to
initialize it with a newly created iommufd by calling
__vfio_pci_device_init(), which then issues the VFIO_DEVICE_BIND_IOMMUFD
ioctl.

Because the device_fd is preserved across the Live Update, doesn't it still
retain its original binding state and reference to the original iommufd
context? It appears this might cause the bind ioctl to return -EINVAL,
triggering a fatal VFIO_ASSERT_EQ() in vfio_device_bind_iommufd() and
crashing the test.

[Severity: Low]
In check_open_vfio_device_fails(), the code directly calls free(device) on
the pointer returned by vfio_pci_device_alloc().

Would it be better to use vfio_pci_device_free() here instead? Even if it
only wraps free() currently, using the library's abstraction might help
prevent future memory leaks if internal allocations are added later.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=19

[PATCH v5 20/20] vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

[Vipin Sharma]

From: David Matlack <dmatlack@google.com>

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 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.

Signed-off-by: David Matlack <dmatlack@google.com>
Signed-off-by: Vipin Sharma <vipinsh@google.com>
---
 .../vfio/vfio_pci_liveupdate_kexec_test.c     | 129 ++++++++++++++++++
 1 file changed, 129 insertions(+)

diff --git a/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
index ec3fe8f19502..51b2f24f352c 100644
--- a/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
+++ b/tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c
@@ -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;
 
 static char state_session[LIVEUPDATE_SESSION_NAME_LENGTH];
@@ -11,8 +19,89 @@ static char device_session[LIVEUPDATE_SESSION_NAME_LENGTH];
 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_READ | 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 before_kexec(int luo_fd)
 {
 	struct vfio_pci_device *device;
@@ -32,6 +121,27 @@ 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();
 }
@@ -78,6 +188,7 @@ static void after_kexec(int luo_fd, int state_session_fd)
 	struct iommu *iommu;
 	int session_fd;
 	int device_fd;
+	int memfd;
 	int stage;
 
 	check_open_vfio_device_fails();
@@ -88,6 +199,10 @@ static void after_kexec(int luo_fd, int state_session_fd)
 	session_fd = luo_retrieve_session(luo_fd, device_session);
 	VFIO_ASSERT_GE(session_fd, 0);
 
+	printf("Retrieving memfd from LUO\n");
+	memfd = luo_session_retrieve_fd(session_fd, MEMFD_TOKEN);
+	VFIO_ASSERT_GE(memfd, 0);
+
 	printf("Finishing the session before retrieving the device (should fail)\n");
 	VFIO_ASSERT_NE(luo_session_finish(session_fd), 0);
 
@@ -109,9 +224,23 @@ static void after_kexec(int luo_fd, int state_session_fd)
 	 */
 	device = __vfio_pci_device_init(device_bdf, iommu, device_fd);
 
+	dma_memfd_map(device, memfd);
+
 	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);
 }
-- 
2.55.0.795.g602f6c329a-goog

Re: [PATCH v5 20/20] vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] Partial initialization of DMA buffers in dma_memcpy_start().
--

--- Patch [20]: [PATCH v5 20/20] vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test ---
Note: Due to system recitation restrictions, this report does not use the standard inline diff quoting format. The findings are instead summarized below without quoting the original patch code.

commit 123ec8cfeb386e8660107c771eabe7ddc5cba029
Author: David Matlack <dmatlack@google.com>

vfio: selftests: Add continuous DMA to vfio_pci_liveupdate_kexec_test

This commit adds a long-running DMA memcpy operation to the VFIO PCI live
update kexec test. It ensures that active device DMAs during a live update do
not cause system instability, even though the DMAs are currently expected to
be dropped.

[Severity: Medium]
In tools/testing/selftests/vfio/vfio_pci_liveupdate_kexec_test.c, looking at
the dma_memcpy_start() function, the buffer initialization seems incomplete.

The code calculates the transfer size and sets the destination pointer, but
then calls memset() on the source and destination buffers using 'size / 2'
instead of 'size'. Immediately after, vfio_pci_driver_memcpy_start() is
kicked off using the full 'size'.

Does this code unintentionally leave the second half of the DMA buffers
uninitialized? While the test does not currently verify the transferred data
since iommufd preservation is unsupported, will this partial initialization
cause false positives or negatives once the test is expanded to validate the
completed DMA operations?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260714151505.3466855-1-vipinsh@google.com?part=20