[PATCH v1 5/7] drm/xe/vf: Rebase HWSP of all contexts after migration

Lis, Tomasz tomasz.lis at intel.com
Thu May 15 22:07:22 UTC 2025


On 14.05.2025 20:37, Michal Wajdeczko wrote:
>
> On 14.05.2025 00:49, Tomasz Lis wrote:
>> All contexts require an update due to GGTT range shift, as that
>> affects their HWSP.
>>
>> The HW status page of a context contains GGTT references, which
>> need to be shifted to a new range (or re-computed using the
>> previously updated vma nodes). The references include ring start
>> address and indirect state address.
>>
>> Signed-off-by: Tomasz Lis <tomasz.lis at intel.com>
>> ---
>>   drivers/gpu/drm/xe/xe_lrc.c      | 18 ++++++++++++++++
>>   drivers/gpu/drm/xe/xe_lrc.h      |  1 +
>>   drivers/gpu/drm/xe/xe_sriov_vf.c | 35 ++++++++++++++++++++++++++++++--
>>   3 files changed, 52 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c
>> index 61a2e87990a9..43e1c18e1769 100644
>> --- a/drivers/gpu/drm/xe/xe_lrc.c
>> +++ b/drivers/gpu/drm/xe/xe_lrc.c
>> @@ -1184,6 +1184,24 @@ void xe_lrc_destroy(struct kref *ref)
>>   	kfree(lrc);
>>   }
>>   
>> +/**
>> + * xe_lrc_update_hwctx_regs_with_address - Re-compute GGTT references within given LRC.
>> + * @lrc: the &xe_lrc struct instance
>> + */
>> +void xe_lrc_update_hwctx_regs_with_address(struct xe_lrc *lrc)
>> +{
>> +	struct xe_gt *gt = lrc->fence_ctx.gt;
>> +
>> +	if (xe_gt_has_indirect_ring_state(gt)) {
> xe_lrc_has_indirect_ring_state(lrc)
ok, these are equivalent at this point of lrc life.
>> +		xe_lrc_write_ctx_reg(lrc, CTX_INDIRECT_RING_STATE,
>> +				     __xe_lrc_indirect_ring_ggtt_addr(lrc));
>> +
>> +		xe_lrc_write_indirect_ctx_reg(lrc, INDIRECT_CTX_RING_START,
>> +					      __xe_lrc_ring_ggtt_addr(lrc));
>> +	} else
>> +		xe_lrc_write_ctx_reg(lrc, CTX_RING_START, __xe_lrc_ring_ggtt_addr(lrc));
> did you really run a checkpatch.pl as promised?
Yes. No issues here. "check" is not an issue. Single line should not get 
a code block.
>> +}
>> +
>>   void xe_lrc_set_ring_tail(struct xe_lrc *lrc, u32 tail)
>>   {
>>   	if (xe_lrc_has_indirect_ring_state(lrc))
>> diff --git a/drivers/gpu/drm/xe/xe_lrc.h b/drivers/gpu/drm/xe/xe_lrc.h
>> index eb6e8de8c939..e7a99cfd0abe 100644
>> --- a/drivers/gpu/drm/xe/xe_lrc.h
>> +++ b/drivers/gpu/drm/xe/xe_lrc.h
>> @@ -88,6 +88,7 @@ bool xe_lrc_ring_is_idle(struct xe_lrc *lrc);
>>   u32 xe_lrc_indirect_ring_ggtt_addr(struct xe_lrc *lrc);
>>   u32 xe_lrc_ggtt_addr(struct xe_lrc *lrc);
>>   u32 *xe_lrc_regs(struct xe_lrc *lrc);
>> +void xe_lrc_update_hwctx_regs_with_address(struct xe_lrc *lrc);
>>   
>>   u32 xe_lrc_read_ctx_reg(struct xe_lrc *lrc, int reg_nr);
>>   void xe_lrc_write_ctx_reg(struct xe_lrc *lrc, int reg_nr, u32 val);
>> diff --git a/drivers/gpu/drm/xe/xe_sriov_vf.c b/drivers/gpu/drm/xe/xe_sriov_vf.c
>> index ab91ac68ef5f..016faa29cddd 100644
>> --- a/drivers/gpu/drm/xe/xe_sriov_vf.c
>> +++ b/drivers/gpu/drm/xe/xe_sriov_vf.c
>> @@ -7,12 +7,14 @@
>>   
>>   #include "xe_assert.h"
>>   #include "xe_device.h"
>> +#include "xe_exec_queue_types.h"
>>   #include "xe_gt.h"
>>   #include "xe_gt_sriov_printk.h"
>>   #include "xe_gt_sriov_vf.h"
>>   #include "xe_guc_ct.h"
>>   #include "xe_guc_submit.h"
>>   #include "xe_irq.h"
>> +#include "xe_lrc.h"
>>   #include "xe_pm.h"
>>   #include "xe_sriov.h"
>>   #include "xe_sriov_printk.h"
>> @@ -223,6 +225,34 @@ static int vf_post_migration_requery_guc(struct xe_device *xe)
>>   	return ret;
>>   }
>>   
>> +static void xe_exec_queue_contexts_hwsp_rebase(struct xe_exec_queue *eq)
> this function would better fit in xe_exec_queue.c
will move.
>> +{
>> +	int i;
>> +
>> +	for (i = 0; i < eq->width; ++i)
>> +		xe_lrc_update_hwctx_regs_with_address(eq->lrc[i]);
>> +}
>> +
>> +static void xe_guc_contexts_hwsp_rebase(struct xe_guc *guc)
> and this one in xe_guc_submit.c

will move.

-Tomasz

>> +{
>> +	struct xe_exec_queue *eq;
>> +	unsigned long index;
>> +
>> +	mutex_lock(&guc->submission_state.lock);
>> +	xa_for_each(&guc->submission_state.exec_queue_lookup, index, eq)
>> +		xe_exec_queue_contexts_hwsp_rebase(eq);
>> +	mutex_unlock(&guc->submission_state.lock);
>> +}
>> +
>> +static void vf_post_migration_fixup_contexts(struct xe_device *xe)
>> +{
>> +	struct xe_gt *gt;
>> +	unsigned int id;
>> +
>> +	for_each_gt(gt, xe, id)
>> +		xe_guc_contexts_hwsp_rebase(&gt->uc.guc);
>> +}
>> +
>>   static void vf_post_migration_fixup_ctb(struct xe_device *xe)
>>   {
>>   	struct xe_gt *gt;
>> @@ -303,9 +333,10 @@ static void vf_post_migration_recovery(struct xe_device *xe)
>>   		goto fail;
>>   
>>   	need_fixups = vf_post_migration_fixup_ggtt_nodes(xe);
>> -	/* FIXME: add the recovery steps */
>> -	if (need_fixups)
>> +	if (need_fixups) {
>> +		vf_post_migration_fixup_contexts(xe);
>>   		vf_post_migration_fixup_ctb(xe);
>> +	}
>>   
>>   	vf_post_migration_kickstart(xe);
>>   	vf_post_migration_notify_resfix_done(xe);


More information about the Intel-xe mailing list