Merge tag 'drm-xe-fixes-2026-04-02' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes

uAPI Fix:
 - Accept canonical GPU addresses in xe_vm_madvise_ioctl (Arvind)

Driver Fixes:
 - Disallow writes to read-only VMAs (Jonathan)
 - PXP fixes (Daniele)
 - Disable garbage collector work item on SVM clos (Brost)
 - void memory allocations in xe_device_declare_wedged (Brost)

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patch.msgid.link/ac5mDHs-McR5cJSV@intel.com

5 files changed, 48 insertions, 26 deletions

diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
index 52ee24e9cd3a..3eb06b27db7e 100644
--- a/drivers/gpu/drm/xe/xe_device.c
+++ b/drivers/gpu/drm/xe/xe_device.c
@@ -837,6 +837,14 @@ static void detect_preproduction_hw(struct xe_device *xe)
 	}
 }
 
+static void xe_device_wedged_fini(struct drm_device *drm, void *arg)
+{
+	struct xe_device *xe = arg;
+
+	if (atomic_read(&xe->wedged.flag))
+		xe_pm_runtime_put(xe);
+}
+
 int xe_device_probe(struct xe_device *xe)
 {
 	struct xe_tile *tile;
@@ -1013,6 +1021,10 @@ int xe_device_probe(struct xe_device *xe)
 
 	detect_preproduction_hw(xe);
 
+	err = drmm_add_action_or_reset(&xe->drm, xe_device_wedged_fini, xe);
+	if (err)
+		goto err_unregister_display;
+
 	return devm_add_action_or_reset(xe->drm.dev, xe_device_sanitize, xe);
 
 err_unregister_display:
@@ -1216,13 +1228,6 @@ u64 xe_device_uncanonicalize_addr(struct xe_device *xe, u64 address)
 	return address & GENMASK_ULL(xe->info.va_bits - 1, 0);
 }
 
-static void xe_device_wedged_fini(struct drm_device *drm, void *arg)
-{
-	struct xe_device *xe = arg;
-
-	xe_pm_runtime_put(xe);
-}
-
 /**
  * DOC: Xe Device Wedging
  *
@@ -1300,15 +1305,9 @@ void xe_device_declare_wedged(struct xe_device *xe)
 		return;
 	}
 
-	xe_pm_runtime_get_noresume(xe);
-
-	if (drmm_add_action_or_reset(&xe->drm, xe_device_wedged_fini, xe)) {
-		drm_err(&xe->drm, "Failed to register xe_device_wedged_fini clean-up. Although device is wedged.\n");
-		return;
-	}
-
 	if (!atomic_xchg(&xe->wedged.flag, 1)) {
 		xe->needs_flr_on_fini = true;
+		xe_pm_runtime_get_noresume(xe);
 		drm_err(&xe->drm,
 			"CRITICAL: Xe has declared device %s as wedged.\n"
 			"IOCTLs and executions are blocked. Only a rebind may clear the failure\n"
diff --git a/drivers/gpu/drm/xe/xe_pagefault.c b/drivers/gpu/drm/xe/xe_pagefault.c
index 6bee53d6ffc3..922a4f3344b1 100644
--- a/drivers/gpu/drm/xe/xe_pagefault.c
+++ b/drivers/gpu/drm/xe/xe_pagefault.c
@@ -187,6 +187,12 @@ static int xe_pagefault_service(struct xe_pagefault *pf)
 		goto unlock_vm;
 	}
 
+	if (xe_vma_read_only(vma) &&
+	    pf->consumer.access_type != XE_PAGEFAULT_ACCESS_TYPE_READ) {
+		err = -EPERM;
+		goto unlock_vm;
+	}
+
 	atomic = xe_pagefault_access_is_atomic(pf->consumer.access_type);
 
 	if (xe_vma_is_cpu_addr_mirror(vma))
diff --git a/drivers/gpu/drm/xe/xe_pxp.c b/drivers/gpu/drm/xe/xe_pxp.c
index d61446bf9c19..872217f30375 100644
--- a/drivers/gpu/drm/xe/xe_pxp.c
+++ b/drivers/gpu/drm/xe/xe_pxp.c
@@ -380,6 +380,18 @@ int xe_pxp_init(struct xe_device *xe)
 		return 0;
 	}
 
+	/*
+	 * On PTL, older GSC FWs have a bug that can cause them to crash during
+	 * PXP invalidation events, which leads to a complete loss of power
+	 * management on the media GT. Therefore, we can't use PXP on FWs that
+	 * have this bug, which was fixed in PTL GSC build 1396.
+	 */
+	if (xe->info.platform == XE_PANTHERLAKE &&
+	    gt->uc.gsc.fw.versions.found[XE_UC_FW_VER_RELEASE].build < 1396) {
+		drm_info(&xe->drm, "PXP requires PTL GSC build 1396 or newer\n");
+		return 0;
+	}
+
 	pxp = drmm_kzalloc(&xe->drm, sizeof(struct xe_pxp), GFP_KERNEL);
 	if (!pxp) {
 		err = -ENOMEM;
@@ -512,7 +524,7 @@ static int __exec_queue_add(struct xe_pxp *pxp, struct xe_exec_queue *q)
 static int pxp_start(struct xe_pxp *pxp, u8 type)
 {
 	int ret = 0;
-	bool restart = false;
+	bool restart;
 
 	if (!xe_pxp_is_enabled(pxp))
 		return -ENODEV;
@@ -541,6 +553,8 @@ wait_for_idle:
 					 msecs_to_jiffies(PXP_ACTIVATION_TIMEOUT_MS)))
 		return -ETIMEDOUT;
 
+	restart = false;
+
 	mutex_lock(&pxp->mutex);
 
 	/* If PXP is not already active, turn it on */
@@ -583,6 +597,7 @@ wait_for_idle:
 			drm_err(&pxp->xe->drm, "PXP termination failed before start\n");
 			mutex_lock(&pxp->mutex);
 			pxp->status = XE_PXP_ERROR;
+			complete_all(&pxp->termination);
 
 			goto out_unlock;
 		}
@@ -870,11 +885,6 @@ wait_for_activation:
 		pxp->key_instance++;
 		needs_queue_inval = true;
 		break;
-	default:
-		drm_err(&pxp->xe->drm, "unexpected state during PXP suspend: %u",
-			pxp->status);
-		ret = -EIO;
-		goto out;
 	}
 
 	/*
@@ -899,7 +909,6 @@ wait_for_activation:
 
 	pxp->last_suspend_key_instance = pxp->key_instance;
 
-out:
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/xe/xe_svm.c b/drivers/gpu/drm/xe/xe_svm.c
index ca67d0bdbfac..c6b92d4ea6f4 100644
--- a/drivers/gpu/drm/xe/xe_svm.c
+++ b/drivers/gpu/drm/xe/xe_svm.c
@@ -903,7 +903,7 @@ int xe_svm_init(struct xe_vm *vm)
 void xe_svm_close(struct xe_vm *vm)
 {
 	xe_assert(vm->xe, xe_vm_is_closed(vm));
-	flush_work(&vm->svm.garbage_collector.work);
+	disable_work_sync(&vm->svm.garbage_collector.work);
 	xe_svm_put_pagemaps(vm);
 	drm_pagemap_release_owner(&vm->svm.peer);
 }
diff --git a/drivers/gpu/drm/xe/xe_vm_madvise.c b/drivers/gpu/drm/xe/xe_vm_madvise.c
index bc39a9a9790c..b4086129a364 100644
--- a/drivers/gpu/drm/xe/xe_vm_madvise.c
+++ b/drivers/gpu/drm/xe/xe_vm_madvise.c
@@ -408,8 +408,15 @@ int xe_vm_madvise_ioctl(struct drm_device *dev, void *data, struct drm_file *fil
 	struct xe_device *xe = to_xe_device(dev);
 	struct xe_file *xef = to_xe_file(file);
 	struct drm_xe_madvise *args = data;
-	struct xe_vmas_in_madvise_range madvise_range = {.addr = args->start,
-							 .range =  args->range, };
+	struct xe_vmas_in_madvise_range madvise_range = {
+		/*
+		 * Userspace may pass canonical (sign-extended) addresses.
+		 * Strip the sign extension to get the internal non-canonical
+		 * form used by the GPUVM, matching xe_vm_bind_ioctl() behavior.
+		 */
+		.addr = xe_device_uncanonicalize_addr(xe, args->start),
+		.range = args->range,
+	};
 	struct xe_madvise_details details;
 	struct xe_vm *vm;
 	struct drm_exec exec;
@@ -439,7 +446,7 @@ int xe_vm_madvise_ioctl(struct drm_device *dev, void *data, struct drm_file *fil
 	if (err)
 		goto unlock_vm;
 
-	err = xe_vm_alloc_madvise_vma(vm, args->start, args->range);
+	err = xe_vm_alloc_madvise_vma(vm, madvise_range.addr, args->range);
 	if (err)
 		goto madv_fini;
 
@@ -482,7 +489,8 @@ int xe_vm_madvise_ioctl(struct drm_device *dev, void *data, struct drm_file *fil
 	madvise_funcs[attr_type](xe, vm, madvise_range.vmas, madvise_range.num_vmas, args,
 				 &details);
 
-	err = xe_vm_invalidate_madvise_range(vm, args->start, args->start + args->range);
+	err = xe_vm_invalidate_madvise_range(vm, madvise_range.addr,
+					     madvise_range.addr + args->range);
 
 	if (madvise_range.has_svm_userptr_vmas)
 		xe_svm_notifier_unlock(vm);