Skip to contentSkip to conversation

FAILED: patch "[PATCH] drm/xe/guc_ads: use uncached mapping for UM queue BO" failed to apply to 6.12-stable tree

PATCH New [email protected] 2 participants 5 messages Aug 20 – Aug 24

Subsystems (3)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkout FETCH_HEAD
git cherry-pick -x 8d5134ae4177fa4f5a9bc8e71e6656cfc2852882
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026082055-unhinge-monstrous-7eeb@gregkh' --subject-prefix 'PATCH 6.12.y' 'HEAD^..'

Possible dependencies:

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

From 8d5134ae4177fa4f5a9bc8e71e6656cfc2852882 Mon Sep 17 00:00:00 2001
From: Jia Yao <[email protected]>
Date: Tue, 4 Aug 2026 16:50:57 +0000
Subject: [PATCH] drm/xe/guc_ads: use uncached mapping for UM queue BO
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

On Pre-Xe3p platform, the GAM write the UM queue through DPA using UC. if GuC reads the queue via GGTT (WB), stale data may be observed when the cacheline has been polluted by another agent.

To match the GAM's UC writes, configure the GuC mapping as UC as well.

Fixes 9c57bc08652a ("drm/xe/lnl: Drop force_probe requirement")
Cc Gwan-gyeong Mun <[email protected]>
Cc Matthew Auld <[email protected]>
Cc <[email protected]> # v6.12+

Reviewed by: Matthew Auld <[email protected]>

(cherry picked from commit 9daa302a82590eeee7bdc68023ddad302df4b88c)
drivers/gpu/drm/xe/xe_guc_ads.c +2 -1
index b6d07f2e73bf..886bafd31f45 100644
@@ -406,7 +406,8 @@ int xe_guc_ads_init(struct xe_guc_ads *ads)
 		u32 um_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile) |
 			       XE_BO_FLAG_GGTT |
 			       XE_BO_FLAG_GGTT_INVALIDATE |
-			       XE_BO_FLAG_PINNED_NORESTORE;
+			       XE_BO_FLAG_PINNED_NORESTORE |
+			       XE_BO_FLAG_NEEDS_UC;
 
 		bo = xe_managed_bo_create_pin_map(xe, tile, um_size, um_flags);
 		if (IS_ERR(bo))
#1 lore raw

PATCH 1/1 drm/xe/guc_ads: use uncached mapping for UM queue BO

On Pre-Xe3p platform, the GAM write the UM queue through DPA using UC. if GuC reads the queue via GGTT (WB), stale data may be observed when the cacheline has been polluted by another agent.

To match the GAM's UC writes, configure the GuC mapping as UC as well.

Fixes 9c57bc08652a ("drm/xe/lnl: Drop force_probe requirement")
Cc Gwan-gyeong Mun <[email protected]>
Cc Matthew Auld <[email protected]>
Cc <[email protected]> # v6.12+

Reviewed by: Matthew Auld <[email protected]>

(cherry picked from commit 9daa302a82590eeee7bdc68023ddad302df4b88c)
(cherry picked from commit 8d5134ae4177fa4f5a9bc8e71e6656cfc2852882)
---
 drivers/gpu/drm/xe/xe_guc_ads.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_guc_ads.c +2 -1
index 35168447a90b..a4932a45120b 100644
@@ -413,7 +413,8 @@ int xe_guc_ads_init(struct xe_guc_ads *ads)
 		u32 um_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile) |
 			       XE_BO_FLAG_GGTT |
 			       XE_BO_FLAG_GGTT_INVALIDATE |
-			       XE_BO_FLAG_PINNED_NORESTORE;
+			       XE_BO_FLAG_PINNED_NORESTORE |
+			       XE_BO_FLAG_NEEDS_UC;
 
 		bo = xe_managed_bo_create_pin_map(xe, tile, um_size, um_flags);
 		if (IS_ERR(bo))
Signature · 1 line
2.43.0
#2 lore raw

PATCH 1/3 drm/xe/guc_ads: allocate UM queues in a separate BO

In reply to [email protected]

Move the UM queues into a dedicated BO (ads->um_queue_bo) and avoid CPU memset operations on it, which eliminates the CPU as a potential cacheline-polluting agent and helps maintain consistency between GAM writes and GuC reads.

We also need to ensure the base_dpa for the queue is contiguous on hw where this is used instead of a GGTT address. Another good reason to split this out to a separate BO.

Fixes 9c57bc08652a ("drm/xe/lnl: Drop force_probe requirement")
Cc Gwan-gyeong Mun <[email protected]>
Cc Matthew Auld <[email protected]>
Cc <[email protected]> # v6.12+
(cherry picked from commit 6af05de0dc19bbf3aaeec2020fe48b37c834b811)
(cherry picked from commit 99b01815957bd7d848420cb697f79ed11f7f215c)
---
 drivers/gpu/drm/xe/xe_bo.h            | 14 +++++
 drivers/gpu/drm/xe/xe_guc.c           |  6 ++
 drivers/gpu/drm/xe/xe_guc_ads.c       | 90 +++++++++++++++++----------
 drivers/gpu/drm/xe/xe_guc_ads_types.h |  5 ++
 4 files changed, 83 insertions(+), 32 deletions(-)
drivers/gpu/drm/xe/xe_bo.h +14 -0
index 556ce5390009..4640d311b5c5 100644
@@ -6,6 +6,7 @@
 #ifndef _XE_BO_H_
 #define _XE_BO_H_
 
+#include <drm/drm_prime.h>
 #include <drm/ttm/ttm_tt.h>
 
 #include "xe_bo_types.h"
@@ -316,6 +317,19 @@ void xe_bo_put_commit(struct llist_head *deferred);
 
 struct sg_table *xe_bo_sg(struct xe_bo *bo);
 
+/**
+ * xe_bo_sg_is_contiguous() - Check if a BO's DMA address space is contiguous.
+ * @bo: the BO to check (must have a valid sg table, i.e. !xe_bo_is_vram())
+ * @len: required contiguous length in bytes
+ *
+ * Returns true if the first @len bytes of the BO are mapped to a contiguous
+ * DMA address range.
+ */
+static inline bool xe_bo_sg_is_contiguous(struct xe_bo *bo, size_t len)
+{
+	return drm_prime_get_contiguous_size(xe_bo_sg(bo)) >= len;
+}
+
 /*
  * xe_sg_segment_size() - Provides upper limit for sg segment size.
  * @dev: device pointer
drivers/gpu/drm/xe/xe_guc.c +6 -0
index bedd9894e182..75ab0730c6cc 100644
@@ -293,6 +293,12 @@ static int xe_guc_realloc_post_hwconfig(struct xe_guc *guc)
 	if (ret)
 		return ret;
 
+	if (guc->ads.um_queue_bo) {
+		ret = xe_managed_bo_reinit_in_vram(xe, tile, &guc->ads.um_queue_bo);
+		if (ret)
+			return ret;
+	}
+
 	return 0;
 }
 
drivers/gpu/drm/xe/xe_guc_ads.c +58 -32

135 lines not on this page · raw message

drivers/gpu/drm/xe/xe_guc_ads_types.h +5 -0
index 2de5decfe0fd..99ad22bda325 100644
@@ -16,6 +16,11 @@ struct xe_bo;
 struct xe_guc_ads {
 	/** @bo: XE BO for GuC ads blob */
 	struct xe_bo *bo;
+	/**
+	 * @um_queue_bo: Dedicated BO for the HW fault ring (UM queues).
+	 * NULL if the platform does not support USM.
+	 */
+	struct xe_bo *um_queue_bo;
 	/** @golden_lrc_size: golden LRC size */
 	size_t golden_lrc_size;
 	/** @regset_size: size of register set passed to GuC for save/restore */
Signature · 1 line
2.43.0
#3 lore raw

PATCH 2/3 drm/xe/guc_ads: allocate UM queues in VRAM on dGFX

On iGPU, the UM queue BO is allocated in system memory. On dGFX, the BO was previously created in system memory and later reallocated in xe_guc_realloc_post_hwconfig(). Allocate the UM queue BO directly in VRAM on dGFX, where it is ultimately required.

Fixes 9c57bc08652a ("drm/xe/lnl: Drop force_probe requirement")
Cc Gwan-gyeong Mun <[email protected]>
Cc Matthew Auld <[email protected]>
Cc <[email protected]> # v6.12+
(cherry picked from commit ace076ef0a854ab5940bacc539bf66afd61d118c)
(cherry picked from commit f342810a141f8a7e8b3786a6e4b6c0695a078a74)
---
 drivers/gpu/drm/xe/xe_guc.c     | 6 ------
 drivers/gpu/drm/xe/xe_guc_ads.c | 8 +++++---
 2 files changed, 5 insertions(+), 9 deletions(-)
drivers/gpu/drm/xe/xe_guc.c +0 -6
index 75ab0730c6cc..bedd9894e182 100644
@@ -293,12 +293,6 @@ static int xe_guc_realloc_post_hwconfig(struct xe_guc *guc)
 	if (ret)
 		return ret;
 
-	if (guc->ads.um_queue_bo) {
-		ret = xe_managed_bo_reinit_in_vram(xe, tile, &guc->ads.um_queue_bo);
-		if (ret)
-			return ret;
-	}
-
 	return 0;
 }
 
drivers/gpu/drm/xe/xe_guc_ads.c +5 -3
index 1bb74e04ed77..35168447a90b 100644
@@ -400,15 +400,17 @@ int xe_guc_ads_init(struct xe_guc_ads *ads)
 		/*
 		 * Allocate a separate BO for the HW fault ring (UM queues).
 		 *
-		 * Round the size up to the next power of two so that
+		 * Round the size up to the next power of two so that on iGPU
 		 * (system memory, no IOMMU) the TTM pool issues a single
 		 * alloc_pages(order=N) call, maximising the chance of getting
 		 * a physically contiguous block.  GuC requires contiguous DPA.
 		 */
-		size_t um_size = roundup_pow_of_two(GUC_UM_QUEUE_SIZE *
+		size_t um_size = IS_DGFX(xe) ?
+				 GUC_UM_QUEUE_SIZE * GUC_UM_HW_QUEUE_MAX :
+				 roundup_pow_of_two(GUC_UM_QUEUE_SIZE *
 						    GUC_UM_HW_QUEUE_MAX);
 
-		u32 um_flags = XE_BO_FLAG_SYSTEM |
+		u32 um_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile) |
 			       XE_BO_FLAG_GGTT |
 			       XE_BO_FLAG_GGTT_INVALIDATE |
 			       XE_BO_FLAG_PINNED_NORESTORE;
Signature · 1 line
2.43.0
#4 lore raw

PATCH 3/3 drm/xe/guc_ads: use uncached mapping for UM queue BO

In reply to Jia Yao

On Pre-Xe3p platform, the GAM write the UM queue through DPA using UC. if GuC reads the queue via GGTT (WB), stale data may be observed when the cacheline has been polluted by another agent.

To match the GAM's UC writes, configure the GuC mapping as UC as well.

Fixes 9c57bc08652a ("drm/xe/lnl: Drop force_probe requirement")
Cc Gwan-gyeong Mun <[email protected]>
Cc Matthew Auld <[email protected]>
Cc <[email protected]> # v6.12+

Reviewed by: Matthew Auld <[email protected]>

(cherry picked from commit 9daa302a82590eeee7bdc68023ddad302df4b88c)
(cherry picked from commit 8d5134ae4177fa4f5a9bc8e71e6656cfc2852882)
---
 drivers/gpu/drm/xe/xe_guc_ads.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_guc_ads.c +2 -1
index 35168447a90b..a4932a45120b 100644
@@ -413,7 +413,8 @@ int xe_guc_ads_init(struct xe_guc_ads *ads)
 		u32 um_flags = XE_BO_FLAG_VRAM_IF_DGFX(tile) |
 			       XE_BO_FLAG_GGTT |
 			       XE_BO_FLAG_GGTT_INVALIDATE |
-			       XE_BO_FLAG_PINNED_NORESTORE;
+			       XE_BO_FLAG_PINNED_NORESTORE |
+			       XE_BO_FLAG_NEEDS_UC;
 
 		bo = xe_managed_bo_create_pin_map(xe, tile, um_size, um_flags);
 		if (IS_ERR(bo))
Signature · 1 line
2.43.0