Skip to contentSkip to conversation

FAILED: patch "[PATCH] drm/xe/pt: Reset current_op in xe_pt_update_ops_init()" failed to apply to 6.18-stable tree

PATCH New [email protected] 2 participants 5 messages Jul 29 – Aug 2

Subsystems (3)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 6.18-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-6.18.y
git checkout FETCH_HEAD
git cherry-pick -x 6384271ac1ac0099198d15df79212a19ebdb929d
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2026072945-jaybird-vaguely-739f@gregkh' --subject-prefix 'PATCH 6.18.y' 'HEAD^..'

Possible dependencies:

thanks,

greg k-h

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

From 6384271ac1ac0099198d15df79212a19ebdb929d Mon Sep 17 00:00:00 2001
From: Zongyao Bai <[email protected]>
Date: Tue, 14 Jul 2026 23:24:32 +0000
Subject: [PATCH] drm/xe/pt: Reset current_op in xe_pt_update_ops_init()
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

xe_pt_update_ops_init() fails to reset current_op to 0. On the vm_bind path, ops_execute() calls xe_pt_update_ops_prepare() inside the xe_validation_guard() / drm_exec_until_all_locked() loop. When that loop retries due to lock contention or OOM eviction (drm_exec_retry_on_contention() / xe_validation_retry_on_oom()), xe_pt_update_ops_prepare() runs again on the same vops, and each call to bind_op_prepare() increments current_op without resetting it.

After N retries current_op exceeds the array size allocated by xe_vma_ops_alloc(), causing an out-of-bounds write into SLUB-poisoned memory and a subsequent UAF crash in xe_migrate_update_pgtables_cpu() when reading the corrupted pt_op->bind.

Also reset needs_svm_lock and needs_invalidation which are derived in the same prepare pass and would otherwise cause wrong migrate ops selection and redundant TLB invalidation on retry.

Fix this by resetting current_op, needs_svm_lock and needs_invalidation in xe_pt_update_ops_init().

v2 (Matt):
- Add details in commit message.
- Add Fixes tag and Cc to [email protected]

Fixes e8babb280b5e ("drm/xe: Convert multiple bind ops into single job")
Assisted-by GitHub-Copilot:claude-sonnet-4.6
(cherry picked from commit 046045543e530605c441063535e7dca0075369a6)
drivers/gpu/drm/xe/xe_pt.c +3 -0
index e787c0c27c42..39c9c8f0ea2d 100644
@@ -2365,8 +2365,11 @@ static void
 xe_pt_update_ops_init(struct xe_vm_pgtable_update_ops *pt_update_ops)
 {
 	init_llist_head(&pt_update_ops->deferred);
+	pt_update_ops->current_op = 0;
 	pt_update_ops->start = ~0x0ull;
 	pt_update_ops->last = 0x0ull;
+	pt_update_ops->needs_svm_lock = false;
+	pt_update_ops->needs_invalidation = false;
 	xe_page_reclaim_list_init(&pt_update_ops->prl);
 }
 
maintainer
#1 lore raw

PATCH 1/4 drm/xe: Use SVM range helpers in PT layer

From: Matthew Brost <[email protected]>

[ Upstream commit 9ea9b45701ab50049a722450abc28346d1121e6e ]

We have helpers SVM range start, end, and size. Use them in the PT layer rather than directly looking at the struct.

Stable-dep-of: 6384271ac1ac ("drm/xe/pt: Reset current_op in xe_pt_update_ops_init()")

---
 drivers/gpu/drm/xe/xe_pt.c | 28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)
drivers/gpu/drm/xe/xe_pt.c +14 -14
index 1f0262adc3ae..a32752c57ea4 100644
@@ -715,7 +715,7 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
 		.vm = vm,
 		.tile = tile,
 		.curs = &curs,
-		.va_curs_start = range ? range->base.itree.start :
+		.va_curs_start = range ? xe_svm_range_start(range) :
 			xe_vma_start(vma),
 		.vma = vma,
 		.wupd.entries = entries,
@@ -734,7 +734,7 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
 		}
 		if (xe_svm_range_has_dma_mapping(range)) {
 			xe_res_first_dma(range->base.pages.dma_addr, 0,
-					 range->base.itree.last + 1 - range->base.itree.start,
+					 xe_svm_range_size(range),
 					 &curs);
 			xe_svm_range_debug(range, "BIND PREPARE - MIXED");
 		} else {
@@ -778,8 +778,8 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
 
 walk_pt:
 	ret = xe_pt_walk_range(&pt->base, pt->level,
-			       range ? range->base.itree.start : xe_vma_start(vma),
-			       range ? range->base.itree.last + 1 : xe_vma_end(vma),
+			       range ? xe_svm_range_start(range) : xe_vma_start(vma),
+			       range ? xe_svm_range_end(range) : xe_vma_end(vma),
 			       &xe_walk.base);
 
 	*num_entries = xe_walk.wupd.num_used_entries;
@@ -983,8 +983,8 @@ bool xe_pt_zap_ptes_range(struct xe_tile *tile, struct xe_vm *vm,
 	if (!(pt_mask & BIT(tile->id)))
 		return false;
 
-	(void)xe_pt_walk_shared(&pt->base, pt->level, range->base.itree.start,
-				range->base.itree.last + 1, &xe_walk.base);
+	(void)xe_pt_walk_shared(&pt->base, pt->level, xe_svm_range_start(range),
+				xe_svm_range_end(range), &xe_walk.base);
 
 	return xe_walk.needs_invalidate;
 }
@@ -1701,8 +1701,8 @@ static unsigned int xe_pt_stage_unbind(struct xe_tile *tile,
 				       struct xe_svm_range *range,
 				       struct xe_vm_pgtable_update *entries)
 {
-	u64 start = range ? range->base.itree.start : xe_vma_start(vma);
-	u64 end = range ? range->base.itree.last + 1 : xe_vma_end(vma);
+	u64 start = range ? xe_svm_range_start(range) : xe_vma_start(vma);
+	u64 end = range ? xe_svm_range_end(range) : xe_vma_end(vma);
 	struct xe_pt_stage_unbind_walk xe_walk = {
 		.base = {
 			.ops = &xe_pt_stage_unbind_ops,
@@ -1912,7 +1912,7 @@ static int bind_range_prepare(struct xe_vm *vm, struct xe_tile *tile,
 
 	vm_dbg(&xe_vma_vm(vma)->xe->drm,
 	       "Preparing bind, with range [%lx...%lx)\n",
-	       range->base.itree.start, range->base.itree.last);
+	       xe_svm_range_start(range), xe_svm_range_end(range) - 1);
 
 	pt_op->vma = NULL;
 	pt_op->bind = true;
@@ -1927,8 +1927,8 @@ static int bind_range_prepare(struct xe_vm *vm, struct xe_tile *tile,
 					pt_op->num_entries, true);
 
 		xe_pt_update_ops_rfence_interval(pt_update_ops,
-						 range->base.itree.start,
-						 range->base.itree.last + 1);
+						 xe_svm_range_start(range),
+						 xe_svm_range_end(range));
 		++pt_update_ops->current_op;
 		pt_update_ops->needs_svm_lock = true;
 
@@ -2023,7 +2023,7 @@ static int unbind_range_prepare(struct xe_vm *vm,
 
 	vm_dbg(&vm->xe->drm,
 	       "Preparing unbind, with range [%lx...%lx)\n",
-	       range->base.itree.start, range->base.itree.last);
+	       xe_svm_range_start(range), xe_svm_range_end(range) - 1);
 
 	pt_op->vma = XE_INVALID_VMA;
 	pt_op->bind = false;
@@ -2034,8 +2034,8 @@ static int unbind_range_prepare(struct xe_vm *vm,
 
 	xe_vm_dbg_print_entries(tile_to_xe(tile), pt_op->entries,
 				pt_op->num_entries, false);
-	xe_pt_update_ops_rfence_interval(pt_update_ops, range->base.itree.start,
-					 range->base.itree.last + 1);
+	xe_pt_update_ops_rfence_interval(pt_update_ops, xe_svm_range_start(range),
+					 xe_svm_range_end(range));
 	++pt_update_ops->current_op;
 	pt_update_ops->needs_svm_lock = true;
 	pt_update_ops->needs_invalidation |= xe_vm_has_scratch(vm) ||
Signature · 1 line
2.53.0
maintainer
#2 lore raw

PATCH 2/4 drm/xe: Stub out new pagefault layer

From: Matthew Brost <[email protected]>

[ Upstream commit 620a09fb0bddf387f418663478b48ca4ba62b6d6 ]

Stub out the new page fault layer and add kernel documentation. This is intended as a replacement for the GT page fault layer, enabling multiple producers to hook into a shared page fault consumer interface.

v2:
- Fix kernel doc typo (checkpatch)
- Remove comment around GT (Stuart)
- Add explaination around reclaim (Francois)
- Add comment around u8 vs enum (Francois)
- Include engine instance (Stuart)
v3:
- Fix XE_PAGEFAULT_TYPE_ATOMIC_ACCESS_VIOLATION kernel doc (Stuart)

Stable-dep-of: 6384271ac1ac ("drm/xe/pt: Reset current_op in xe_pt_update_ops_init()")

---
 drivers/gpu/drm/xe/Makefile             |   1 +
 drivers/gpu/drm/xe/xe_pagefault.c       |  65 +++++++++++
 drivers/gpu/drm/xe/xe_pagefault.h       |  19 ++++
 drivers/gpu/drm/xe/xe_pagefault_types.h | 136 ++++++++++++++++++++++++
 4 files changed, 221 insertions(+)
 create mode 100644 drivers/gpu/drm/xe/xe_pagefault.c
 create mode 100644 drivers/gpu/drm/xe/xe_pagefault.h
 create mode 100644 drivers/gpu/drm/xe/xe_pagefault_types.h
drivers/gpu/drm/xe/Makefile +1 -0
index ee52b751623f..c5fc8cc71149 100644
@@ -94,6 +94,7 @@ xe-y += xe_bb.o \
 	xe_nvm.o \
 	xe_oa.o \
 	xe_observation.o \
+	xe_pagefault.o \
 	xe_pat.o \
 	xe_pci.o \
 	xe_pcode.o \
drivers/gpu/drm/xe/xe_pagefault.c new file +65 -0

65 lines not on this page · raw message

drivers/gpu/drm/xe/xe_pagefault.h new file +19 -0
new file mode 100644
index 000000000000..bd0cdf9ed37f
@@ -0,0 +1,19 @@
+/* SPDX-License-Identifier: MIT */
+/*
+ * Copyright © 2025 Intel Corporation
+ */
+
+#ifndef _XE_PAGEFAULT_H_
+#define _XE_PAGEFAULT_H_
+
+struct xe_device;
+struct xe_gt;
+struct xe_pagefault;
+
+int xe_pagefault_init(struct xe_device *xe);
+
+void xe_pagefault_reset(struct xe_device *xe, struct xe_gt *gt);
+
+int xe_pagefault_handler(struct xe_device *xe, struct xe_pagefault *pf);
+
+#endif
drivers/gpu/drm/xe/xe_pagefault_types.h new file +136 -0

136 lines not on this page · raw message

Signature · 1 line
2.53.0
maintainer
#3 lore raw

PATCH 3/4 drm/xe: Add page reclamation info to device info

In reply to Sasha Levin

From: Oak Zeng <[email protected]>

[ Upstream commit 9b1a0e0a15c97987fdf56a615f3d13995bafd042 ]

Starting from Xe3p, HW adds a feature assisting range based page reclamation. Introduce a bit in device info to indicate whether device has such capability.

Stable-dep-of: 6384271ac1ac ("drm/xe/pt: Reset current_op in xe_pt_update_ops_init()")

---
 drivers/gpu/drm/xe/xe_device_types.h | 2 ++
 drivers/gpu/drm/xe/xe_pci.c          | 1 +
 drivers/gpu/drm/xe/xe_pci_types.h    | 1 +
 3 files changed, 4 insertions(+)
drivers/gpu/drm/xe/xe_device_types.h +2 -0
index 0e80f2940c99..2ed03fd074ff 100644
@@ -305,6 +305,8 @@ struct xe_device {
 		 * pcode mailbox commands.
 		 */
 		u8 has_mbx_power_limits:1;
+		/** @info.has_page_reclaim_hw_assist: Device supports page reclamation feature */
+		u8 has_page_reclaim_hw_assist:1;
 		/** @info.has_pxp: Device has PXP support */
 		u8 has_pxp:1;
 		/** @info.has_range_tlb_inval: Has range based TLB invalidations */
drivers/gpu/drm/xe/xe_pci.c +1 -0
index 89cc6d32f041..b78729211790 100644
@@ -591,6 +591,7 @@ static int xe_info_init_early(struct xe_device *xe,
 	xe->info.has_heci_cscfi = desc->has_heci_cscfi;
 	xe->info.has_late_bind = desc->has_late_bind;
 	xe->info.has_llc = desc->has_llc;
+	xe->info.has_page_reclaim_hw_assist = desc->has_page_reclaim_hw_assist;
 	xe->info.has_pxp = desc->has_pxp;
 	xe->info.has_sriov = desc->has_sriov;
 	xe->info.skip_guc_pc = desc->skip_guc_pc;
drivers/gpu/drm/xe/xe_pci_types.h +1 -0
index 9b9766a3baa3..a7b6576725ed 100644
@@ -42,6 +42,7 @@ struct xe_device_desc {
 	u8 has_late_bind:1;
 	u8 has_llc:1;
 	u8 has_mbx_power_limits:1;
+	u8 has_page_reclaim_hw_assist:1;
 	u8 has_pxp:1;
 	u8 has_sriov:1;
 	u8 needs_scratch:1;
Signature · 1 line
2.53.0
maintainer
#4 lore raw

PATCH 4/4 drm/xe/pt: Reset current_op in xe_pt_update_ops_init()

In reply to Sasha Levin

From: Zongyao Bai <[email protected]>

[ Upstream commit 6384271ac1ac0099198d15df79212a19ebdb929d ]

xe_pt_update_ops_init() fails to reset current_op to 0. On the vm_bind path, ops_execute() calls xe_pt_update_ops_prepare() inside the xe_validation_guard() / drm_exec_until_all_locked() loop. When that loop retries due to lock contention or OOM eviction (drm_exec_retry_on_contention() / xe_validation_retry_on_oom()), xe_pt_update_ops_prepare() runs again on the same vops, and each call to bind_op_prepare() increments current_op without resetting it.

After N retries current_op exceeds the array size allocated by xe_vma_ops_alloc(), causing an out-of-bounds write into SLUB-poisoned memory and a subsequent UAF crash in xe_migrate_update_pgtables_cpu() when reading the corrupted pt_op->bind.

Also reset needs_svm_lock and needs_invalidation which are derived in the same prepare pass and would otherwise cause wrong migrate ops selection and redundant TLB invalidation on retry.

Fix this by resetting current_op, needs_svm_lock and needs_invalidation in xe_pt_update_ops_init().

v2 (Matt):
- Add details in commit message.
- Add Fixes tag and Cc to [email protected]

Fixes e8babb280b5e ("drm/xe: Convert multiple bind ops into single job")
Assisted-by GitHub-Copilot:claude-sonnet-4.6
(cherry picked from commit 046045543e530605c441063535e7dca0075369a6)
---
 drivers/gpu/drm/xe/xe_pt.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/gpu/drm/xe/xe_pt.c +3 -0
index a32752c57ea4..c21f46a56be3 100644
@@ -2143,8 +2143,11 @@ static void
 xe_pt_update_ops_init(struct xe_vm_pgtable_update_ops *pt_update_ops)
 {
 	init_llist_head(&pt_update_ops->deferred);
+	pt_update_ops->current_op = 0;
 	pt_update_ops->start = ~0x0ull;
 	pt_update_ops->last = 0x0ull;
+	pt_update_ops->needs_svm_lock = false;
+	pt_update_ops->needs_invalidation = false;
 }
 
 /**
Signature · 1 line
2.53.0