Re: [PATCH v4 2/3] drm/xe/vf: Introduce RESFIX start marker support

["K V P, Satyanarayana" <satyanarayana.k.v.p@intel.com>]

On 19-Nov-25 11:08 PM, Matthew Brost wrote:
> On Wed, Nov 19, 2025 at 06:24:45PM +0100, Michal Wajdeczko wrote:
>>
>> On 11/18/2025 12:41 PM, Satyanarayana K V P wrote:
>>> In scenarios involving double migration, the VF KMD may encounter
>>> situations where it is instructed to re-migrate before having the
>>> opportunity to send RESFIX_DONE for the initial migration. This can occur
>>> when the fix-up for the prior migration is still underway, but the VF KMD
>>> is migrated again.
>>>
>>> Consequently, this may lead to the possibility of sending two migration
>>> notifications (i.e., pending fix-up for the first migration and a second
>>> notification for the new migration). Upon receiving the first RES_FIX
>>> notification, the GuC will resume VF submission on the GPU, potentially
>>> resulting in undefined behavior, such as system hangs or crashes.
>>>
>>> To avoid this, post migration, a marker is sent to the GUC prior to the
>>> start of resource fixups to indicate start of resource fixups. The same
>>> marker is sent along with RESFIX_DONE notification so that GUC can avoid
>>> submitting jobs to HW in case of double migration.
>>>
>>> Signed-off-by: Satyanarayana K V P <satyanarayana.k.v.p@intel.com>
>>> Cc: Michal Wajdeczko <michal.wajdeczko@intel.com>
>>> Cc: Matthew Brost <matthew.brost@intel.com>
>>> Cc: Tomasz Lis <tomasz.lis@intel.com>
>>>
>>> ---
>>> V3 -> V4:
>>> - Updated RESFIX_DONE action name and documenation part. (Michal W)
>>> - Enable resfxi_start marked by default as sav/restore is gated on
>>> Guc version 70.54.0
>>>
>>> V2 -> V3:
>>> - Fixed review comments (Michal W).
>>> - Updated commit message.
>>> - Fixed CI.BAT issues.
>>> - Added helper function to assert on unsupported GUC versions.
>>> - Updated RESFIX_DONE action name and documenation part.
>>>
>>> V1 -> V2:
>>> - Squashed "Enable RESFIX start marker only on supported GUC
>>> versions" commit into a single commit. (Matt B)
>>> ---
>>>   .../gpu/drm/xe/abi/guc_actions_sriov_abi.h    | 60 +++++++++++---
>>>   drivers/gpu/drm/xe/xe_gt_sriov_vf.c           | 80 ++++++++++++++-----
>>>   drivers/gpu/drm/xe/xe_gt_sriov_vf_types.h     |  5 ++
>>>   drivers/gpu/drm/xe/xe_sriov_vf.c              | 16 +++-
>>>   4 files changed, 131 insertions(+), 30 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/xe/abi/guc_actions_sriov_abi.h b/drivers/gpu/drm/xe/abi/guc_actions_sriov_abi.h
>>> index 0b28659d94e9..1d84ce07b201 100644
>>> --- a/drivers/gpu/drm/xe/abi/guc_actions_sriov_abi.h
>>> +++ b/drivers/gpu/drm/xe/abi/guc_actions_sriov_abi.h
>>> @@ -502,13 +502,15 @@
>>>   #define VF2GUC_VF_RESET_RESPONSE_MSG_0_MBZ		GUC_HXG_RESPONSE_MSG_0_DATA0
>>>   
>>>   /**
>>> - * DOC: VF2GUC_NOTIFY_RESFIX_DONE
>>> + * DOC: VF2GUC_RESFIX_DONE
>>>    *
>>> - * This action is used by VF to notify the GuC that the VF KMD has completed
>>> + * This action is used by VF to inform the GuC that the VF KMD has completed
>>>    * post-migration recovery steps.
>> please mention that from 1.27 it shall only be sent after posting RESFIX_START
>> and that both @MARKER fields must match
>>
>>>    *
>>>    * This message must be sent as `MMIO HXG Message`_.
>>>    *
>>> + * Available since GuC VF compatibility 1.27.0.
>> hmm, actually RESFIX_DONE is also available prior 1.27,
>> just a meaning of the DATA0 has changed
>>
>> maybe:
>>
>>      * Updated since GuC VF compatibility 1.27.0.
>>
>>> + *
>>>    *  +---+-------+--------------------------------------------------------------+
>>>    *  |   | Bits  | Description                                                  |
>>>    *  +===+=======+==============================================================+
>>> @@ -516,9 +518,9 @@
>>>    *  |   +-------+--------------------------------------------------------------+
>>>    *  |   | 30:28 | TYPE = GUC_HXG_TYPE_REQUEST_                                 |
>>>    *  |   +-------+--------------------------------------------------------------+
>>> - *  |   | 27:16 | DATA0 = MBZ                                                  |
>>> + *  |   | 27:16 | DATA0 = MARKER - can't be zero                               |
>> and then we can keep legacy definition for the record:
>>
>>      *  |   +-------+--------------------------------------------------------------+
>>    - *  |   | 27:16 | DATA0 = MBZ                                                  |
>>    + *  |   | 27:16 | DATA0 = MARKER = MBZ (only prior 1.27.0)                     |
>>      *  |   +-------+--------------------------------------------------------------+
>>    + *  |   | 27:16 | DATA0 = MARKER - can't be zero (1.27.0+)                     |
>>    + *  |   +-------+--------------------------------------------------------------+
>>
>>
>>>    *  |   +-------+--------------------------------------------------------------+
>>> - *  |   |  15:0 | ACTION = _`GUC_ACTION_VF2GUC_NOTIFY_RESFIX_DONE` = 0x5508    |
>>> + *  |   |  15:0 | ACTION = _`GUC_ACTION_VF2GUC_RESFIX_DONE` = 0x5508           |
>>>    *  +---+-------+--------------------------------------------------------------+
>>>    *
>>>    *  +---+-------+--------------------------------------------------------------+
>>> @@ -531,13 +533,13 @@
>>>    *  |   |  27:0 | DATA0 = MBZ                                                  |
>>>    *  +---+-------+--------------------------------------------------------------+
>>>    */
>>> -#define GUC_ACTION_VF2GUC_NOTIFY_RESFIX_DONE		0x5508u
>>> +#define GUC_ACTION_VF2GUC_RESFIX_DONE			0x5508u
>>>   
>>> -#define VF2GUC_NOTIFY_RESFIX_DONE_REQUEST_MSG_LEN	GUC_HXG_REQUEST_MSG_MIN_LEN
>>> -#define VF2GUC_NOTIFY_RESFIX_DONE_REQUEST_MSG_0_MBZ	GUC_HXG_REQUEST_MSG_0_DATA0
>>> +#define VF2GUC_RESFIX_DONE_REQUEST_MSG_LEN		GUC_HXG_REQUEST_MSG_MIN_LEN
>>> +#define VF2GUC_RESFIX_DONE_REQUEST_MSG_0_MARKER		GUC_HXG_REQUEST_MSG_0_DATA0
>>>   
>>> -#define VF2GUC_NOTIFY_RESFIX_DONE_RESPONSE_MSG_LEN	GUC_HXG_RESPONSE_MSG_MIN_LEN
>>> -#define VF2GUC_NOTIFY_RESFIX_DONE_RESPONSE_MSG_0_MBZ	GUC_HXG_RESPONSE_MSG_0_DATA0
>>> +#define VF2GUC_RESFIX_DONE_RESPONSE_MSG_LEN		GUC_HXG_RESPONSE_MSG_MIN_LEN
>>> +#define VF2GUC_RESFIX_DONE_RESPONSE_MSG_0_MBZ		GUC_HXG_RESPONSE_MSG_0_DATA0
>>>   
>>>   /**
>>>    * DOC: VF2GUC_QUERY_SINGLE_KLV
>>> @@ -656,4 +658,44 @@
>>>   #define PF2GUC_SAVE_RESTORE_VF_RESPONSE_MSG_LEN		GUC_HXG_RESPONSE_MSG_MIN_LEN
>>>   #define PF2GUC_SAVE_RESTORE_VF_RESPONSE_MSG_0_USED	GUC_HXG_RESPONSE_MSG_0_DATA0
>>>   
>>> +/**
>>> + * DOC: VF2GUC_RESFIX_START
>>> + *
>>> + * This action is used by VF to inform the GuC that the VF KMD will be starting
>>> + * post-migration recovery fixups.
>> please mention that @MARKER sent here must later match the MARKER posted in the
>> VF2GUC_RESFIX_DONE_ message
>>
>>> + *
>>> + * This message must be sent as `MMIO HXG Message`_.
>>> + *
>>> + * Available since GuC VF compatibility 1.27.0.
>>> + *
>>> + *  +---+-------+--------------------------------------------------------------+
>>> + *  |   | Bits  | Description                                                  |
>>> + *  +===+=======+==============================================================+
>>> + *  | 0 |    31 | ORIGIN = GUC_HXG_ORIGIN_HOST_                                |
>>> + *  |   +-------+--------------------------------------------------------------+
>>> + *  |   | 30:28 | TYPE = GUC_HXG_TYPE_REQUEST_                                 |
>>> + *  |   +-------+--------------------------------------------------------------+
>>> + *  |   | 27:16 | DATA0 = MARKER - can't be zero                               |
>>> + *  |   +-------+--------------------------------------------------------------+
>>> + *  |   |  15:0 | ACTION = _`GUC_ACTION_VF2GUC_RESFIX_START` = 0x550F          |
>>> + *  +---+-------+--------------------------------------------------------------+
>>> + *
>>> + *  +---+-------+--------------------------------------------------------------+
>>> + *  |   | Bits  | Description                                                  |
>>> + *  +===+=======+==============================================================+
>>> + *  | 0 |    31 | ORIGIN = GUC_HXG_ORIGIN_GUC_                                 |
>>> + *  |   +-------+--------------------------------------------------------------+
>>> + *  |   | 30:28 | TYPE = GUC_HXG_TYPE_RESPONSE_SUCCESS_                        |
>>> + *  |   +-------+--------------------------------------------------------------+
>>> + *  |   |  27:0 | DATA0 = MBZ                                                  |
>>> + *  +---+-------+--------------------------------------------------------------+
>>> + */
>>> +#define GUC_ACTION_VF2GUC_RESFIX_START			0x550Fu
>>> +
>>> +#define VF2GUC_RESFIX_START_REQUEST_MSG_LEN		GUC_HXG_REQUEST_MSG_MIN_LEN
>>> +#define VF2GUC_RESFIX_START_REQUEST_MSG_0_MARKER	GUC_HXG_REQUEST_MSG_0_DATA0
>>> +
>>> +#define VF2GUC_RESFIX_START_RESPONSE_MSG_LEN		GUC_HXG_RESPONSE_MSG_MIN_LEN
>>> +#define VF2GUC_RESFIX_START_RESPONSE_MSG_0_MBZ		GUC_HXG_RESPONSE_MSG_0_DATA0
>>> +
>>>   #endif
>>> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_vf.c b/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
>>> index 4c73a077d314..08c00b773a13 100644
>>> --- a/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
>>> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_vf.c
>>> @@ -299,12 +299,13 @@ void xe_gt_sriov_vf_guc_versions(struct xe_gt *gt,
>>>   		*found = gt->sriov.vf.guc_version;
>>>   }
>>>   
>>> -static int guc_action_vf_notify_resfix_done(struct xe_guc *guc)
>>> +static int guc_action_vf_notify_resfix_start(struct xe_guc *guc, u16 marker)
>>>   {
>>>   	u32 request[GUC_HXG_REQUEST_MSG_MIN_LEN] = {
>>>   		FIELD_PREP(GUC_HXG_MSG_0_ORIGIN, GUC_HXG_ORIGIN_HOST) |
>>>   		FIELD_PREP(GUC_HXG_MSG_0_TYPE, GUC_HXG_TYPE_REQUEST) |
>>> -		FIELD_PREP(GUC_HXG_REQUEST_MSG_0_ACTION, GUC_ACTION_VF2GUC_NOTIFY_RESFIX_DONE),
>>> +		FIELD_PREP(GUC_HXG_REQUEST_MSG_0_ACTION, GUC_ACTION_VF2GUC_RESFIX_START) |
>>> +		FIELD_PREP(VF2GUC_RESFIX_START_REQUEST_MSG_0_MARKER, marker),
>>>   	};
>>>   	int ret;
>>>   
>>> @@ -313,30 +314,54 @@ static int guc_action_vf_notify_resfix_done(struct xe_guc *guc)
>>>   	return ret > 0 ? -EPROTO : ret;
>>>   }
>>>   
>>> -/**
>>> - * vf_notify_resfix_done - Notify GuC about resource fixups apply completed.
>>> - * @gt: the &xe_gt struct instance linked to target GuC
>>> - *
>>> - * Returns: 0 if the operation completed successfully, or a negative error
>>> - * code otherwise.
>>> - */
>>> -static int vf_notify_resfix_done(struct xe_gt *gt)
>>> +static int vf_notify_resfix_start(struct xe_gt *gt, u16 marker)
>>>   {
>>>   	struct xe_guc *guc = &gt->uc.guc;
>>>   	int err;
>>>   
>>>   	xe_gt_assert(gt, IS_SRIOV_VF(gt_to_xe(gt)));
>>>   
>>> -	err = guc_action_vf_notify_resfix_done(guc);
>>> +	xe_gt_sriov_dbg(guc_to_gt(guc), "Sending resfix start marker %u\n", marker);
>> shouldn't this be xe_gt_sriov_dbg_verbose() instead?
>>
>>> +
>>> +	err = guc_action_vf_notify_resfix_start(guc, marker);
>>>   	if (unlikely(err))
>>> -		xe_gt_sriov_err(gt, "Failed to notify GuC about resource fixup done (%pe)\n",
>>> +		xe_gt_sriov_err(gt, "Failed to notify GuC about resource fixup start(%pe)\n",
>> add space between "start" and "(%pe)"
>>
>>>   				ERR_PTR(err));
>>> -	else
>>> -		xe_gt_sriov_dbg_verbose(gt, "sent GuC resource fixup done\n");
>>>   
>>>   	return err;
>>>   }
>>>   
>>> +static int guc_action_vf_notify_resfix_done(struct xe_guc *guc, u16 marker)
>>> +{
>>> +	u32 request[GUC_HXG_REQUEST_MSG_MIN_LEN] = {
>>> +		FIELD_PREP(GUC_HXG_MSG_0_ORIGIN, GUC_HXG_ORIGIN_HOST) |
>>> +		FIELD_PREP(GUC_HXG_MSG_0_TYPE, GUC_HXG_TYPE_REQUEST) |
>>> +		FIELD_PREP(GUC_HXG_REQUEST_MSG_0_ACTION, GUC_ACTION_VF2GUC_RESFIX_DONE) |
>>> +		FIELD_PREP(VF2GUC_RESFIX_DONE_REQUEST_MSG_0_MARKER, marker),
>>> +	};
>>> +	int ret;
>>> +
>>> +	ret = xe_guc_mmio_send(guc, request, ARRAY_SIZE(request));
>>> +
>>> +	return ret > 0 ? -EPROTO : ret;
>>> +}
>>> +
>>> +static int vf_notify_resfix_done(struct xe_gt *gt, u16 marker)
>>> +{
>>> +	struct xe_guc *guc = &gt->uc.guc;
>>> +	int err;
>>> +
>>> +	xe_gt_assert(gt, IS_SRIOV_VF(gt_to_xe(gt)));
>>> +
>>> +	xe_gt_sriov_dbg(guc_to_gt(guc), "Sending resfix done marker %u\n", marker);
>> dbg_verbose ?
>>
>>> +
>>> +	err = guc_action_vf_notify_resfix_done(guc, marker);
>>> +	if (unlikely(err))
>>> +		xe_gt_sriov_err(gt, "Failed to notify GuC about resource fixup done (%pe)\n",
>> hmm, it's not only about that _we_ failed, it could be that _GuC_
>> encountered some errors, as there is ERROR_RESFIX_FAILED, so maybe:
>>
>> 	"Recovery failed at GuC FIXUP_DONE step (%pe)"
>>
>>> +				ERR_PTR(err));
>>> +	return err;
>>> +}
>>> +
>>>   static int guc_action_query_single_klv(struct xe_guc *guc, u32 key,
>>>   				       u32 *value, u32 value_len)
>>>   {
>>> @@ -1183,7 +1208,7 @@ static void vf_post_migration_abort(struct xe_gt *gt)
>>>   	xe_guc_submit_pause_abort(&gt->uc.guc);
>>>   }
>>>   
>>> -static int vf_post_migration_notify_resfix_done(struct xe_gt *gt)
>>> +static int vf_post_migration_notify_resfix_done(struct xe_gt *gt, u16 marker)
>>>   {
>>>   	bool skip_resfix = false;
>>>   
>>> @@ -1206,14 +1231,21 @@ static int vf_post_migration_notify_resfix_done(struct xe_gt *gt)
>>>   	 */
>>>   	xe_irq_resume(gt_to_xe(gt));
>>>   
>>> -	return vf_notify_resfix_done(gt);
>>> +	return vf_notify_resfix_done(gt, marker);
>>> +}
>>> +
>>> +static u16 vf_post_migration_resfix_start_marker(struct xe_gt *gt)
>>> +{
>>> +	xe_gt_assert(gt, IS_SRIOV_VF(gt_to_xe(gt)));
>>> +	return ++gt->sriov.vf.migration.resfix_marker;
>> should we protect that with lock?
>>
> No lock required - this code runs a per-GT ordered workqueue which has
> built in mutual exclusion.
>
>> also see below
>>
>>>   }
>>>   
>>>   static void vf_post_migration_recovery(struct xe_gt *gt)
>>>   {
>>>   	struct xe_device *xe = gt_to_xe(gt);
>>> -	int err;
>>> +	u16 marker;
>>>   	bool retry;
>>> +	int err;
>>>   
>>>   	xe_gt_sriov_dbg(gt, "migration recovery in progress\n");
>>>   
>>> @@ -1227,13 +1259,25 @@ static void vf_post_migration_recovery(struct xe_gt *gt)
>>>   		goto fail;
>>>   	}
>>>   
>>> +	/*
>>> +	 * Increment the startup marker again if it overflows, since GUC
>>> +	 * requires a non-zero marker to be set.
>>> +	 */
>>> +	marker = vf_post_migration_resfix_start_marker(gt);
>>> +	if (!marker)
>>> +		marker = vf_post_migration_resfix_start_marker(gt);
>> this "overflow" logic shall be in vf_post_migration_resfix_start_marker()
>>
> I think I suggested the above as well or at least thought about it.
>
>> OTOH by looking at the expected flow, maybe we don't need to track this
>> marker at all, as it should be sufficient to always pass the same const
>> non-zero value, GuC will just compare it with 0 anyway
>>
>> and we send RESFIX_START/DONE only from within this worker, so we will
>> never have two parallel recovery sequences which would warrant different
>> markers
> Yes, I think a const marker probably works. A marker that moves does
> maybe is better for debug logging though?
>
> Matt

Yes. I do agree to have a different marker which will be useful for 
debugging.

Fixed all other review comments in the new revision.

-Satya.

>>> +
>>> +	err = vf_notify_resfix_start(gt, marker);
>>> +	if (err)
>>> +		goto fail;
>>> +
>>>   	err = vf_post_migration_fixups(gt);
>>>   	if (err)
>>>   		goto fail;
>>>   
>>>   	vf_post_migration_rearm(gt);
>>>   
>>> -	err = vf_post_migration_notify_resfix_done(gt);
>>> +	err = vf_post_migration_notify_resfix_done(gt, marker);
>>>   	if (err && err != -EAGAIN)
>>>   		goto fail;
>>>   
>>> diff --git a/drivers/gpu/drm/xe/xe_gt_sriov_vf_types.h b/drivers/gpu/drm/xe/xe_gt_sriov_vf_types.h
>>> index 420b0e6089de..66c0062a42c6 100644
>>> --- a/drivers/gpu/drm/xe/xe_gt_sriov_vf_types.h
>>> +++ b/drivers/gpu/drm/xe/xe_gt_sriov_vf_types.h
>>> @@ -52,6 +52,11 @@ struct xe_gt_sriov_vf_migration {
>>>   	wait_queue_head_t wq;
>>>   	/** @scratch: Scratch memory for VF recovery */
>>>   	void *scratch;
>>> +	/**
>>> +	 * @resfix_marker: Marker sent on start and on end of post-migration
>>> +	 * steps.
>>> +	 */
>>> +	u16 resfix_marker;
>>>   	/** @recovery_teardown: VF post migration recovery is being torn down */
>>>   	bool recovery_teardown;
>>>   	/** @recovery_queued: VF post migration recovery in queued */
>>> diff --git a/drivers/gpu/drm/xe/xe_sriov_vf.c b/drivers/gpu/drm/xe/xe_sriov_vf.c
>>> index b73498097df5..64b2ddabd3f9 100644
>>> --- a/drivers/gpu/drm/xe/xe_sriov_vf.c
>>> +++ b/drivers/gpu/drm/xe/xe_sriov_vf.c
>>> @@ -49,11 +49,13 @@
>>>    *
>>>    * As soon as Virtual GPU of the VM starts, the VF driver within receives
>>>    * the MIGRATED interrupt and schedules post-migration recovery worker.
>>> - * That worker queries GuC for new provisioning (using MMIO communication),
>>> + * That worker sends `VF2GUC_NOTIFY_RESFIX_START` action along with non-zero
>> drop NOTIFY tag and use trailing _ to create a link:
>>
>> 	VF2GUC_RESFIX_START_
>>
>>> + * marker, queries GuC for new provisioning (using MMIO communication),
>>>    * and applies fixups to any non-virtualized resources used by the VF.
>>>    *
>>>    * When the VF driver is ready to continue operation on the newly connected
>>> - * hardware, it sends `VF2GUC_NOTIFY_RESFIX_DONE` which causes it to
>>> + * hardware, it sends `VF2GUC_NOTIFY_RESFIX_DONE` action along with the same
>>> + * marker which was sent with `VF2GUC_NOTIFY_RESFIX_START` which causes it to
>> ditto
>>
>>>    * enter the long awaited `VF_RUNNING` state, and therefore start handling
>>>    * CTB messages and scheduling workloads from the VF::
>>>    *
>>> @@ -102,6 +104,11 @@
>>>    *      |                              [ ]        new VF provisioning  [ ]
>>>    *      |                              [ ]---------------------------> [ ]
>>>    *      |                               |                              [ ]
>>> + *      |                               |   VF2GUC_NOTIFY_RESFIX_START [ ]
>> ditto, drop NOTIFY
>>
>>> + *      |                              [ ] <---------------------------[ ]
>>> + *      |                              [ ]                             [ ]
>>> + *      |                              [ ]                     success [ ]
>>> + *      |                              [ ]---------------------------> [ ]
>>>    *      |                               |       VF driver applies post [ ]
>>>    *      |                               |      migration fixups -------[ ]
>>>    *      |                               |                       |      [ ]
>>> @@ -114,7 +121,10 @@
>>>    *      |                              [ ]------- VF_RUNNING           [ ]
>>>    *      |                              [ ]      |                      [ ]
>>>    *      |                              [ ] <-----                      [ ]
>>> - *      |                              [ ]                     success [ ]
>>> + *      |                              [ ]   success (on marker match) [ ]
>>> + *      |                              [ ]---------------------------> [ ]
>>> + *      |                              [ ]   error (on marker match)   [ ]
>>> + *      |                              [ ] ERROR_RESFIX_MARKER_MISMATCH[ ]
>> this error is about bad programming, not worth mentioning here
>>
>> for the double-migration case, we expect STATUS_VF_MIGRATED instead
>>
>> and in case of error/double migration, VF will not be moved to RUNNING state
>>
>>>    *      |                              [ ]---------------------------> [ ]
>>>    *      |                               |                               |
>>>    *      |                               |                               |