Re: [PATCH 2/8] drm/xe/xe_pci_error: Implement PCI error recovery callbacks

[Riana Tauro <riana.tauro@intel.com>]

On 2/16/2026 2:23 PM, Mallesh, Koujalagi wrote:
> Hi Riana,
> 
> On 22-01-2026 03:36 pm, Riana Tauro wrote:
>> Add error_detected, mmio_enabled, slot_reset and resume
>> recovery callbacks to handle PCIe Advanced Error Reporting
>> (AER) errors.
>>
>> For fatal errors, the device is wedged and becomes
>> inaccessible. Return PCI_ERS_RESULT_SLOT_RESET from
>> error_detected to request a Secondary Bus Reset (SBR).
>>
>> For non-fatal errors, return PCI_ERS_RESULT_CAN_RECOVER from
>> error_detected to trigger the mmio_enabled callback. In this callback,
>> the device is queried to determine the error cause and attempt
>> recovery based on the error type.
>>
>> Once the secondary bus reset(SBR) is completed the slot_reset callback
>> cleanly removes and reprobe the device to restore functionality.
>>
>> Signed-off-by: Riana Tauro<riana.tauro@intel.com>
>> ---
>>   drivers/gpu/drm/xe/Makefile          |  1 +
>>   drivers/gpu/drm/xe/xe_device.h       | 15 +++++
>>   drivers/gpu/drm/xe/xe_device_types.h |  3 +
>>   drivers/gpu/drm/xe/xe_pci.c          |  3 +
>>   drivers/gpu/drm/xe/xe_pci_error.c    | 85 ++++++++++++++++++++++++++++
>>   5 files changed, 107 insertions(+)
>>   create mode 100644 drivers/gpu/drm/xe/xe_pci_error.c
>>
>> diff --git a/drivers/gpu/drm/xe/Makefile b/drivers/gpu/drm/xe/Makefile
>> index f6650ec3ab42..5581f2180b5c 100644
>> --- a/drivers/gpu/drm/xe/Makefile
>> +++ b/drivers/gpu/drm/xe/Makefile
>> @@ -98,6 +98,7 @@ xe-y += xe_bb.o \
>>       xe_page_reclaim.o \
>>       xe_pat.o \
>>       xe_pci.o \
>> +    xe_pci_error.o \
>>       xe_pci_rebar.o \
>>       xe_pcode.o \
>>       xe_pm.o \
>> diff --git a/drivers/gpu/drm/xe/xe_device.h b/drivers/gpu/drm/xe/ 
>> xe_device.h
>> index 58d7d8b2fea3..81480248eeff 100644
>> --- a/drivers/gpu/drm/xe/xe_device.h
>> +++ b/drivers/gpu/drm/xe/xe_device.h
>> @@ -43,6 +43,21 @@ static inline struct xe_device 
>> *ttm_to_xe_device(struct ttm_device *ttm)
>>       return container_of(ttm, struct xe_device, ttm);
>>   }
>> +static inline bool xe_device_is_in_recovery(struct xe_device *xe)
>> +{
>> +    return atomic_read(&xe->in_recovery);
>> +}
>> +
>> +static inline void xe_device_set_in_recovery(struct xe_device *xe)
>> +{
>> +    atomic_set(&xe->in_recovery, 1);
>> +}
>> +
>> +static inline void xe_device_clear_in_recovery(struct xe_device *xe)
>> +{
>> +     atomic_set(&xe->in_recovery, 0);
>> +}
>> +
>>   struct xe_device *xe_device_create(struct pci_dev *pdev,
>>                      const struct pci_device_id *ent);
>>   int xe_device_probe_early(struct xe_device *xe);
>> diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/ 
>> xe/xe_device_types.h
>> index 944f909a86ad..2d140463dc5e 100644
>> --- a/drivers/gpu/drm/xe/xe_device_types.h
>> +++ b/drivers/gpu/drm/xe/xe_device_types.h
>> @@ -669,6 +669,9 @@ struct xe_device {
>>           bool inconsistent_reset;
>>       } wedged;
>> +    /** @in_recovery: Indicates if device is in recovery */
>> +    atomic_t in_recovery;
>> +
>>       /** @bo_device: Struct to control async free of BOs */
>>       struct xe_bo_dev {
>>           /** @bo_device.async_free: Free worker */
>> diff --git a/drivers/gpu/drm/xe/xe_pci.c b/drivers/gpu/drm/xe/xe_pci.c
>> index c92cc176f669..e1ee393b7461 100644
>> --- a/drivers/gpu/drm/xe/xe_pci.c
>> +++ b/drivers/gpu/drm/xe/xe_pci.c
>> @@ -1255,6 +1255,8 @@ static const struct dev_pm_ops xe_pm_ops = {
>>   };
>>   #endif
>> +extern const struct pci_error_handlers xe_pci_error_handlers;
>> +
>>   static struct pci_driver xe_pci_driver = {
>>       .name = DRIVER_NAME,
>>       .id_table = pciidlist,
>> @@ -1262,6 +1264,7 @@ static struct pci_driver xe_pci_driver = {
>>       .remove = xe_pci_remove,
>>       .shutdown = xe_pci_shutdown,
>>       .sriov_configure = xe_pci_sriov_configure,
>> +    .err_handler = &xe_pci_error_handlers,
>>   #ifdef CONFIG_PM_SLEEP
>>       .driver.pm = &xe_pm_ops,
>>   #endif
>> diff --git a/drivers/gpu/drm/xe/xe_pci_error.c b/drivers/gpu/drm/xe/ 
>> xe_pci_error.c
>> new file mode 100644
>> index 000000000000..a3cc01afa179
>> --- /dev/null
>> +++ b/drivers/gpu/drm/xe/xe_pci_error.c
>> @@ -0,0 +1,85 @@
>> +// SPDX-License-Identifier: MIT
>> +/*
>> + * Copyright © 2026 Intel Corporation
>> + */
>> +#include <drm/drm_drv.h>
>> +#include <linux/pci.h>
>> +
>> +#include "xe_device.h"
>> +#include "xe_gt.h"
>> +#include "xe_pci.h"
>> +#include "xe_uc.h"
>> +
>> +static void xe_pci_error_handling(struct pci_dev *pdev)
>> +{
>> +    struct xe_device *xe = pdev_to_xe_device(pdev);
>> +
>> +    xe_device_set_in_recovery(xe);
>> +    xe_device_declare_wedged(xe);
>> +
>> +    pci_disable_device(pdev);
>> +}
>> +
>> +static pci_ers_result_t xe_pci_error_detected(struct pci_dev *pdev, 
>> pci_channel_state_t state)
>> +{
>> +    dev_err(&pdev->dev, "PCI error detected, state %d\n", state);
>> +
> We need to set recovery flag here right?

Yeah we have to set this for all errors. Will fix this

>> +    switch (state) {
>> +    case pci_channel_io_normal:
>> +        return PCI_ERS_RESULT_CAN_RECOVER;
>> +    case pci_channel_io_frozen:
>> +        xe_pci_error_handling(pdev);
>> +        return PCI_ERS_RESULT_NEED_RESET;
>> +    case pci_channel_io_perm_failure:
>> +        return PCI_ERS_RESULT_DISCONNECT;
>> +    }
>> +
>> +    return PCI_ERS_RESULT_NEED_RESET;
> Please make default case where we see "Unknown channel state" as dev_err 
> and return PCI_ERS_RESULT_NEED_RESET.

Sure

>> +}
>> +
>> +static pci_ers_result_t xe_pci_error_mmio_enabled(struct pci_dev *pdev)
>> +{
>> +    dev_err(&pdev->dev, "PCI mmio enabled\n");
>> +
>> +    return PCI_ERS_RESULT_NEED_RESET;
>> +}
>> +
>> +static pci_ers_result_t xe_pci_error_slot_reset(struct pci_dev *pdev)
>> +{
>> +    const struct pci_device_id *ent = pci_match_id(pdev->driver- 
>> >id_table, pdev);
>> +    struct xe_device *xe = pdev_to_xe_device(pdev);
>> +
> Check xe is null or not? if null then return PCI_ERS_RESULT_DISCONNECT.
>> +    dev_err(&pdev->dev, "PCI slot reset\n");
>> +
>> +    pci_restore_state(pdev);
>> +
>> +    if (pci_enable_device(pdev)) {
>> +        dev_err(&pdev->dev,
>> +            "Cannot re-enable PCI device after reset\n");
>> +        return PCI_ERS_RESULT_DISCONNECT;
>> +    }
>> +
>> +    /*
>> +     * Secondary Bus Reset wipes out all device memory
>> +     * requiring XE KMD to perform a device removal and reprobe.
>> +     */
>> +    pdev->driver->remove(pdev);
>> +    xe_device_clear_in_recovery(xe);
>> +
>> +    if (!pdev->driver->probe(pdev, ent))
>> +        return PCI_ERS_RESULT_RECOVERED;
>> +
>> +    return PCI_ERS_RESULT_RECOVERED;
>> +}
>> +
>> +static void xe_pci_error_resume(struct pci_dev *pdev)
>> +{
>> +    dev_info(&pdev->dev, "PCI error resume\n");
> 
> We need to clear recovery flag (if not already cleared), for normal 
> operations.

Yeah if i add the in_recovery for all errors and we don't trigger a 
reset. We need to clear it here.

Will add this

Thanks
Riana


> 
> Thanks
> 
> -/Mallesh
> 
>> +}
>> +
>> +const struct pci_error_handlers xe_pci_error_handlers = {
>> +    .error_detected    = xe_pci_error_detected,
>> +    .mmio_enabled    = xe_pci_error_mmio_enabled,
>> +    .slot_reset    = xe_pci_error_slot_reset,
>> +    .resume        = xe_pci_error_resume,
>> +};