[RFC v2 5/6] drm/xe: Add replay_offset and replay_length lines to LRC HWCTX snapshot

Carlos Santa carlos.santa at intel.corp-partner.google.com
Tue Feb 11 02:20:05 UTC 2025


From: Matthew Brost <matthew.brost at intel.com>

Add replay_offset and replay_length lines to LRC HWCTX snapshot with the
idea being this information can be used extract the data which needs to
be pass to exec queue extension DRM_XE_EXEC_QUEUE_SET_HANG_REPLAY_STATE
so GPU hang can be replayed via a Mesa tool.

The additional lines look like:

[HWCTX].replay_offset: 0x%x
[HWCTX].replay_length: 0x%x

Cc: José Roberto de Souza <jose.souza at intel.com>
Signed-off-by: Matthew Brost <matthew.brost at intel.com>
---
 drivers/gpu/drm/xe/xe_lrc.c       | 8 ++++++++
 drivers/gpu/drm/xe/xe_lrc.h       | 1 +
 drivers/gpu/drm/xe/xe_lrc_types.h | 3 +++
 3 files changed, 12 insertions(+)

diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c
index df3ceddede07..a77fbce523e6 100644
--- a/drivers/gpu/drm/xe/xe_lrc.c
+++ b/drivers/gpu/drm/xe/xe_lrc.c
@@ -897,6 +897,9 @@ static int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe,
 
 	kref_init(&lrc->refcount);
 	lrc->flags = 0;
+	lrc->replay_size = xe_gt_lrc_size(gt, hwe->class);
+	if (xe_gt_has_indirect_ring_state(gt))
+		lrc->replay_size -= LRC_INDIRECT_RING_STATE_SIZE;
 	lrc_size = ring_size + xe_gt_lrc_size(gt, hwe->class);
 	if (xe_gt_has_indirect_ring_state(gt))
 		lrc->flags |= XE_LRC_FLAG_INDIRECT_RING_STATE;
@@ -1683,6 +1686,8 @@ struct xe_lrc_snapshot *xe_lrc_snapshot_capture(struct xe_lrc *lrc)
 	snapshot->lrc_bo = xe_bo_get(lrc->bo);
 	snapshot->lrc_offset = xe_lrc_pphwsp_offset(lrc);
 	snapshot->lrc_size = lrc->bo->size - snapshot->lrc_offset;
+	snapshot->replay_offset = 0;
+	snapshot->replay_size = lrc->replay_size;
 	snapshot->lrc_snapshot = NULL;
 	snapshot->ctx_timestamp = xe_lrc_ctx_timestamp(lrc);
 	snapshot->ctx_job_timestamp = xe_lrc_ctx_job_timestamp(lrc);
@@ -1757,6 +1762,9 @@ void xe_lrc_snapshot_print(struct xe_lrc_snapshot *snapshot, struct drm_printer
 	}
 
 	drm_printf(p, "\n\t[HWCTX].length: 0x%lx\n", snapshot->lrc_size - LRC_PPHWSP_SIZE);
+	drm_printf(p, "\n\t[HWCTX].replay_offset: 0x%lx\n", snapshot->replay_offset);
+	drm_printf(p, "\n\t[HWCTX].replay_length: 0x%lx\n", snapshot->replay_size);
+
 	drm_puts(p, "\t[HWCTX].data: ");
 	for (; i < snapshot->lrc_size; i += sizeof(u32)) {
 		u32 *val = snapshot->lrc_snapshot + i;
diff --git a/drivers/gpu/drm/xe/xe_lrc.h b/drivers/gpu/drm/xe/xe_lrc.h
index 0b40f349ab95..2d6838645858 100644
--- a/drivers/gpu/drm/xe/xe_lrc.h
+++ b/drivers/gpu/drm/xe/xe_lrc.h
@@ -23,6 +23,7 @@ struct xe_lrc_snapshot {
 	struct xe_bo *lrc_bo;
 	void *lrc_snapshot;
 	unsigned long lrc_size, lrc_offset;
+	unsigned long replay_size, replay_offset;
 
 	u32 context_desc;
 	u32 ring_addr;
diff --git a/drivers/gpu/drm/xe/xe_lrc_types.h b/drivers/gpu/drm/xe/xe_lrc_types.h
index 71ecb453f811..c2766719ff77 100644
--- a/drivers/gpu/drm/xe/xe_lrc_types.h
+++ b/drivers/gpu/drm/xe/xe_lrc_types.h
@@ -25,6 +25,9 @@ struct xe_lrc {
 	/** @size: size of lrc including any indirect ring state page */
 	u32 size;
 
+	/** @replay_size: Size LRC needed for replaying a hang */
+	u32 replay_size;
+
 	/** @tile: tile which this LRC belongs to */
 	struct xe_tile *tile;
 
-- 
2.43.0



More information about the Intel-xe mailing list