Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/xe: Use managed BO in memirq
@ 2024-12-04 11:16 Ilia Levi
  2024-12-04 11:50 ` ✓ CI.Patch_applied: success for " Patchwork
                   ` (8 more replies)
  0 siblings, 9 replies; 10+ messages in thread
From: Ilia Levi @ 2024-12-04 11:16 UTC (permalink / raw)
  To: intel-xe
  Cc: ilia.levi, piotr.piorkowski, michal.wajdeczko, koby.elbaz,
	yaron.avizrat

When probe has an error, the current scheme causes a page fault during
resource unwinding. This happens because GTTMMADR_BAR gets unmapped
before the allocated BO is released via drmm. Switching to a managed BO
that is released via devm solves this ordering issue.

Signed-off-by: Ilia Levi <ilia.levi@intel.com>
---
 drivers/gpu/drm/xe/xe_memirq.c | 23 +++++++----------------
 1 file changed, 7 insertions(+), 16 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_memirq.c b/drivers/gpu/drm/xe/xe_memirq.c
index 51dc90906003..404fa2a456d5 100644
--- a/drivers/gpu/drm/xe/xe_memirq.c
+++ b/drivers/gpu/drm/xe/xe_memirq.c
@@ -155,13 +155,6 @@ static const char *guc_name(struct xe_guc *guc)
  *
  */
 
-static void __release_xe_bo(struct drm_device *drm, void *arg)
-{
-	struct xe_bo *bo = arg;
-
-	xe_bo_unpin_map_no_vm(bo);
-}
-
 static inline bool hw_reports_to_instance_zero(struct xe_memirq *memirq)
 {
 	/*
@@ -184,14 +177,12 @@ static int memirq_alloc_pages(struct xe_memirq *memirq)
 	BUILD_BUG_ON(!IS_ALIGNED(XE_MEMIRQ_SOURCE_OFFSET(0), SZ_64));
 	BUILD_BUG_ON(!IS_ALIGNED(XE_MEMIRQ_STATUS_OFFSET(0), SZ_4K));
 
-	/* XXX: convert to managed bo */
-	bo = xe_bo_create_pin_map(xe, tile, NULL, bo_size,
-				  ttm_bo_type_kernel,
-				  XE_BO_FLAG_SYSTEM |
-				  XE_BO_FLAG_GGTT |
-				  XE_BO_FLAG_GGTT_INVALIDATE |
-				  XE_BO_FLAG_NEEDS_UC |
-				  XE_BO_FLAG_NEEDS_CPU_ACCESS);
+	bo = xe_managed_bo_create_pin_map(xe, tile, bo_size,
+					  XE_BO_FLAG_SYSTEM |
+					  XE_BO_FLAG_GGTT |
+					  XE_BO_FLAG_GGTT_INVALIDATE |
+					  XE_BO_FLAG_NEEDS_UC |
+					  XE_BO_FLAG_NEEDS_CPU_ACCESS);
 	if (IS_ERR(bo)) {
 		err = PTR_ERR(bo);
 		goto out;
@@ -215,7 +206,7 @@ static int memirq_alloc_pages(struct xe_memirq *memirq)
 		     xe_bo_ggtt_addr(bo), bo_size, XE_MEMIRQ_SOURCE_OFFSET(0),
 		     XE_MEMIRQ_STATUS_OFFSET(0));
 
-	return drmm_add_action_or_reset(&xe->drm, __release_xe_bo, memirq->bo);
+	return 0;
 
 out:
 	memirq_err(memirq, "Failed to allocate memirq page (%pe)\n", ERR_PTR(err));
-- 
2.43.2


^ permalink raw reply related	[flat|nested] 10+ messages in thread
end of thread, other threads:[~2024-12-04 19:32 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-12-04 11:16 [PATCH] drm/xe: Use managed BO in memirq Ilia Levi
2024-12-04 11:50 ` ✓ CI.Patch_applied: success for " Patchwork
2024-12-04 11:50 ` ✓ CI.checkpatch: " Patchwork
2024-12-04 11:51 ` ✓ CI.KUnit: " Patchwork
2024-12-04 12:09 ` ✓ CI.Build: " Patchwork
2024-12-04 12:12 ` ✓ CI.Hooks: " Patchwork
2024-12-04 12:13 ` ✓ CI.checksparse: " Patchwork
2024-12-04 12:34 ` ✓ Xe.CI.BAT: " Patchwork
2024-12-04 14:12 ` ✗ Xe.CI.Full: failure " Patchwork
2024-12-04 19:32 ` [PATCH] " Lucas De Marchi

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox