Skip to contentSkip to conversation

drm/amdgpu: implement per-process MES context

PATCH In review Zhu Lingshan 2 participants 3 messages Jun 1 – Jun 9

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 drm/amdgpu: implement per-process MES context

MES process context is a process-level page where process specific context is saved for MES scheduler.

However, current user-queue code path assigns fw_obj of a queue to MES process_context_addr when adding the queue to MES.

This means every new queue from the same process would replace the previous process context address with that queue's fw_obj address.
What's worse is, when user space frees a queue, its fw_obj will be freed as well, causing MES working on a NULL page pointer.

This issue leads to inconsistency and crash in the scheduler.

This commit allocates a process-level page for MES process contexts for a process other than queue-level

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c  |  6 +++
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h  |  2 +
 drivers/gpu/drm/amd/amdgpu/mes_userqueue.c | 51 +++++++++++++++++-----
 3 files changed, 47 insertions(+), 12 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +6 -0
index 38e310a8694d..951d5da850be 100644
@@ -1172,6 +1172,7 @@ int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *f
 	xa_init_flags(&userq_mgr->userq_xa, XA_FLAGS_ALLOC);
 	userq_mgr->adev = adev;
 	userq_mgr->file = file_priv;
+	mutex_init(&userq_mgr->proc_ctx_lock);
 
 	INIT_DELAYED_WORK(&userq_mgr->resume_work, amdgpu_userq_restore_worker);
 	INIT_WORK(&userq_mgr->reset_work, amdgpu_userq_mgr_reset_work);
@@ -1225,6 +1226,11 @@ void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
 	 */
 	cancel_work_sync(&userq_mgr->reset_work);
 
+	amdgpu_bo_free_kernel(&userq_mgr->proc_ctx_obj.obj,
+			      &userq_mgr->proc_ctx_obj.gpu_addr,
+			      &userq_mgr->proc_ctx_obj.cpu_ptr);
+
+	mutex_destroy(&userq_mgr->proc_ctx_lock);
 	mutex_destroy(&userq_mgr->userq_mutex);
 }
 
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h +2 -0
index 28cfc6682333..a5867ffe6988 100644
@@ -127,6 +127,8 @@ struct amdgpu_userq_mgr {
 	struct amdgpu_device		*adev;
 	struct delayed_work		resume_work;
 	struct drm_file			*file;
+	struct mutex			proc_ctx_lock;
+	struct amdgpu_userq_obj		proc_ctx_obj;
 
 	/**
 	 * @reset_work:
drivers/gpu/drm/amd/amdgpu/mes_userqueue.c +39 -12

93 lines not on this page · raw message

Signature · 1 line
2.54.0
#1 lore raw

AMD General

Gentle ping, Christian and Alex

-----Original Message-----
From: Zhu, Lingshan <[email protected]>
Sent: Monday, June 1, 2026 2:41 PM
To: Deucher, Alexander <[email protected]>; Koenig, Christian <[email protected]>
Cc: [email protected]; Huang, Ray <[email protected]>; Zhu, Lingshan <[email protected]>
Subject: [PATCH V3] drm/amdgpu: implement per-process MES context

MES process context is a process-level page where process specific context is saved for MES scheduler.

However, current user-queue code path assigns fw_obj of a queue to MES process_context_addr when adding the queue to MES.

This means every new queue from the same process would replace the previous process context address with that queue's fw_obj address. What's worse is, when user space frees a queue, its fw_obj will be freed as well, causing MES working on a NULL page pointer.

This issue leads to inconsistency and crash in the scheduler.

This commit allocates a process-level page for MES process contexts for a process other than queue-level

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c  |  6 +++  drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h  |  2 +  drivers/gpu/drm/amd/amdgpu/mes_userqueue.c | 51 +++++++++++++++++-----
 3 files changed, 47 insertions(+), 12 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +1 -0
index 38e310a8694d..951d5da850be 100644
@@ -1172,6 +1172,7 @@ int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *f
        xa_init_flags(&userq_mgr->userq_xa, XA_FLAGS_ALLOC);
        userq_mgr->adev = adev;
        userq_mgr->file = file_priv;
+       mutex_init(&userq_mgr->proc_ctx_lock);
 
        INIT_DELAYED_WORK(&userq_mgr->resume_work, amdgpu_userq_restore_worker);
        INIT_WORK(&userq_mgr->reset_work, amdgpu_userq_mgr_reset_work); @@ -1225,6 +1226,11 @@ void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
         */
        cancel_work_sync(&userq_mgr->reset_work);

+ amdgpu_bo_free_kernel(&userq_mgr->proc_ctx_obj.obj,
+ &userq_mgr->proc_ctx_obj.gpu_addr,
+ &userq_mgr->proc_ctx_obj.cpu_ptr);
+

+       mutex_destroy(&userq_mgr->proc_ctx_lock);
        mutex_destroy(&userq_mgr->userq_mutex);

}

drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h +2 -0
index 28cfc6682333..a5867ffe6988 100644
@@ -127,6 +127,8 @@ struct amdgpu_userq_mgr {
        struct amdgpu_device            *adev;
        struct delayed_work             resume_work;
        struct drm_file                 *file;
+       struct mutex                    proc_ctx_lock;
+       struct amdgpu_userq_obj         proc_ctx_obj;
 
        /**
         * @reset_work:
drivers/gpu/drm/amd/amdgpu/mes_userqueue.c +3 -3
index e9189f07c6dc..6c8a44cee34b 100644
@@ -133,8 +133,8 @@ static int mes_userq_map(struct amdgpu_usermode_queue *queue)
        queue_input.gang_quantum = 10000;
        queue_input.paging = false;
 
-       queue_input.process_context_addr = ctx->gpu_addr;
-       queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+       queue_input.process_context_addr = uq_mgr->proc_ctx_obj.gpu_addr;
+       queue_input.gang_context_addr = ctx->gpu_addr;
        queue_input.inprocess_gang_priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
        queue_input.gang_global_priority_level = convert_to_mes_priority(queue->priority);
 
@@ -169,7 +169,7 @@ static int mes_userq_unmap(struct amdgpu_usermode_queue *queue)
 
        memset(&queue_input, 0x0, sizeof(struct mes_remove_queue_input));
        queue_input.doorbell_offset = queue->doorbell_index;
-       queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+       queue_input.gang_context_addr = ctx->gpu_addr;
 
        amdgpu_mes_lock(&adev->mes);
        r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input); @@ -186,12 +186,8 @@ static int mes_userq_create_ctx_space(struct amdgpu_userq_mgr *uq_mgr,
        struct amdgpu_userq_obj *ctx = &queue->fw_obj;
        int r, size;

- /*

-        * The FW expects at least one page space allocated for
-        * process ctx and gang ctx each. Create an object
-        * for the same.
-        */
-       size = AMDGPU_USERQ_PROC_CTX_SZ + AMDGPU_USERQ_GANG_CTX_SZ;
+       /* The FW expects at least one page space allocated for gang ctx. */
+       size = AMDGPU_USERQ_GANG_CTX_SZ;
        r = amdgpu_bo_create_kernel(uq_mgr->adev, size, 0,
                                    AMDGPU_GEM_DOMAIN_GTT,
                                    &ctx->obj, &ctx->gpu_addr,
@@ -257,6 +253,30 @@ static int mes_userq_detect_and_reset(struct amdgpu_device *adev,
        return r;

}

+static int mes_userq_create_proc_ctx_space(struct amdgpu_userq_mgr +*uq_mgr) {
+ int r = 0;
+
+ mutex_lock(&uq_mgr->proc_ctx_lock);
+ /* This check is a necessary because amdgpu_bo_create_kernel()
+ * calls helpers like amdgpu_bo_pin() and memset() unconditionally
+ */
+ if (!uq_mgr->proc_ctx_obj.obj) {
+ r = amdgpu_bo_create_kernel(uq_mgr->adev, AMDGPU_USERQ_PROC_CTX_SZ,

+                                           0, AMDGPU_GEM_DOMAIN_GTT,
+                                           &uq_mgr->proc_ctx_obj.obj,
+                                           &uq_mgr->proc_ctx_obj.gpu_addr,
+                                           &uq_mgr->proc_ctx_obj.cpu_ptr);

+
+ if (!r)
+ memset(uq_mgr->proc_ctx_obj.cpu_ptr, 0, AMDGPU_USERQ_PROC_CTX_SZ);
+ }
+
+ mutex_unlock(&uq_mgr->proc_ctx_lock);
+
+ return r;
+}
+
static int mes_userq_mqd_create(struct amdgpu_usermode_queue *queue,

                                struct drm_amdgpu_userq_in *args_in)  { @@ -429,7 +449,14 @@ static int mes_userq_mqd_create(struct amdgpu_usermode_queue *queue,
                goto free_mqd;
        }
-       /* Create BO for FW operations */
+       /* Create per-process MES process context BO */
+       r = mes_userq_create_proc_ctx_space(uq_mgr);
+       if (r) {

+ DRM_ERROR("Failed to allocate MES process context space bo, error: %d\n", r);
+ goto free_mqd;
+ }
+
+ /* Create BO of a gang for FW operations */

        r = mes_userq_create_ctx_space(uq_mgr, queue, mqd_user);
        if (r) {
                DRM_ERROR("Failed to allocate BO for userqueue (%d)", r); @@ -492,7 +519,7 @@ static int mes_userq_preempt(struct amdgpu_usermode_queue *queue)
        *fence_ptr = 0;

        memset(&queue_input, 0x0, sizeof(struct mes_suspend_gang_input));
-       queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+       queue_input.gang_context_addr = ctx->gpu_addr;
        queue_input.suspend_fence_addr = fence_gpu_addr;
        queue_input.suspend_fence_value = 1;
        amdgpu_mes_lock(&adev->mes);
@@ -529,7 +556,7 @@ static int mes_userq_restore(struct amdgpu_usermode_queue *queue)
                return 0;

        memset(&queue_input, 0x0, sizeof(struct mes_resume_gang_input));

- queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+ queue_input.gang_context_addr = ctx->gpu_addr;

        amdgpu_mes_lock(&adev->mes);
        r = adev->mes.funcs->resume_gang(&adev->mes, &queue_input);

--
2.54.0

maintainer Reviewed-by on v3 1/1
#2 lore raw
In reply to Zhu Lingshan
Quoted 15 lines · Zhu Lingshan

MES process context is a process-level page where process specific context is saved for MES scheduler.

However, current user-queue code path assigns fw_obj of a queue to MES process_context_addr when adding the queue to MES.

This means every new queue from the same process would replace the previous process context address with that queue's fw_obj address.
What's worse is, when user space frees a queue, its fw_obj will be freed as well, causing MES working on a NULL page pointer.

Quoted 7 lines

This issue leads to inconsistency and crash in the scheduler.

This commit allocates a process-level page for MES process contexts for a process other than queue-level

Quoted 147 lines
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c  |  6 +++
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h  |  2 +
 drivers/gpu/drm/amd/amdgpu/mes_userqueue.c | 51 +++++++++++++++++-----
 3 files changed, 47 insertions(+), 12 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c +6 -0
index 38e310a8694d..951d5da850be 100644
@@ -1172,6 +1172,7 @@ int amdgpu_userq_mgr_init(struct amdgpu_userq_mgr *userq_mgr, struct drm_file *f
 	xa_init_flags(&userq_mgr->userq_xa, XA_FLAGS_ALLOC);
 	userq_mgr->adev = adev;
 	userq_mgr->file = file_priv;
+	mutex_init(&userq_mgr->proc_ctx_lock);
 
 	INIT_DELAYED_WORK(&userq_mgr->resume_work, amdgpu_userq_restore_worker);
 	INIT_WORK(&userq_mgr->reset_work, amdgpu_userq_mgr_reset_work);
@@ -1225,6 +1226,11 @@ void amdgpu_userq_mgr_fini(struct amdgpu_userq_mgr *userq_mgr)
 	 */
 	cancel_work_sync(&userq_mgr->reset_work);
 
+	amdgpu_bo_free_kernel(&userq_mgr->proc_ctx_obj.obj,
+			      &userq_mgr->proc_ctx_obj.gpu_addr,
+			      &userq_mgr->proc_ctx_obj.cpu_ptr);
+
+	mutex_destroy(&userq_mgr->proc_ctx_lock);
 	mutex_destroy(&userq_mgr->userq_mutex);
 }
 
drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h +2 -0
index 28cfc6682333..a5867ffe6988 100644
@@ -127,6 +127,8 @@ struct amdgpu_userq_mgr {
 	struct amdgpu_device		*adev;
 	struct delayed_work		resume_work;
 	struct drm_file			*file;
+	struct mutex			proc_ctx_lock;
+	struct amdgpu_userq_obj		proc_ctx_obj;
 
 	/**
 	 * @reset_work:
drivers/gpu/drm/amd/amdgpu/mes_userqueue.c +39 -12
index e9189f07c6dc..6c8a44cee34b 100644
@@ -133,8 +133,8 @@ static int mes_userq_map(struct amdgpu_usermode_queue *queue)
 	queue_input.gang_quantum = 10000;
 	queue_input.paging = false;
 
-	queue_input.process_context_addr = ctx->gpu_addr;
-	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+	queue_input.process_context_addr = uq_mgr->proc_ctx_obj.gpu_addr;
+	queue_input.gang_context_addr = ctx->gpu_addr;
 	queue_input.inprocess_gang_priority = AMDGPU_MES_PRIORITY_LEVEL_NORMAL;
 	queue_input.gang_global_priority_level = convert_to_mes_priority(queue->priority);
 
@@ -169,7 +169,7 @@ static int mes_userq_unmap(struct amdgpu_usermode_queue *queue)
 
 	memset(&queue_input, 0x0, sizeof(struct mes_remove_queue_input));
 	queue_input.doorbell_offset = queue->doorbell_index;
-	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+	queue_input.gang_context_addr = ctx->gpu_addr;
 
 	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->remove_hw_queue(&adev->mes, &queue_input);
@@ -186,12 +186,8 @@ static int mes_userq_create_ctx_space(struct amdgpu_userq_mgr *uq_mgr,
 	struct amdgpu_userq_obj *ctx = &queue->fw_obj;
 	int r, size;
 
-	/*
-	 * The FW expects at least one page space allocated for
-	 * process ctx and gang ctx each. Create an object
-	 * for the same.
-	 */
-	size = AMDGPU_USERQ_PROC_CTX_SZ + AMDGPU_USERQ_GANG_CTX_SZ;
+	/* The FW expects at least one page space allocated for gang ctx. */
+	size = AMDGPU_USERQ_GANG_CTX_SZ;
 	r = amdgpu_bo_create_kernel(uq_mgr->adev, size, 0,
 				    AMDGPU_GEM_DOMAIN_GTT,
 				    &ctx->obj, &ctx->gpu_addr,
@@ -257,6 +253,30 @@ static int mes_userq_detect_and_reset(struct amdgpu_device *adev,
 	return r;
 }
 
+static int mes_userq_create_proc_ctx_space(struct amdgpu_userq_mgr *uq_mgr)
+{
+	int r = 0;
+
+	mutex_lock(&uq_mgr->proc_ctx_lock);
+	/* This check is a necessary because amdgpu_bo_create_kernel()
+	 * calls helpers like amdgpu_bo_pin() and memset() unconditionally
+	 */
+	if (!uq_mgr->proc_ctx_obj.obj) {
+		r = amdgpu_bo_create_kernel(uq_mgr->adev, AMDGPU_USERQ_PROC_CTX_SZ,
+					    0, AMDGPU_GEM_DOMAIN_GTT,
+					    &uq_mgr->proc_ctx_obj.obj,
+					    &uq_mgr->proc_ctx_obj.gpu_addr,
+					    &uq_mgr->proc_ctx_obj.cpu_ptr);
+
+		if (!r)
+			memset(uq_mgr->proc_ctx_obj.cpu_ptr, 0, AMDGPU_USERQ_PROC_CTX_SZ);
+	}
+
+	mutex_unlock(&uq_mgr->proc_ctx_lock);
+
+	return r;
+}
+
 static int mes_userq_mqd_create(struct amdgpu_usermode_queue *queue,
 				struct drm_amdgpu_userq_in *args_in)
 {
@@ -429,7 +449,14 @@ static int mes_userq_mqd_create(struct amdgpu_usermode_queue *queue,
 		goto free_mqd;
 	}
 
-	/* Create BO for FW operations */
+	/* Create per-process MES process context BO */
+	r = mes_userq_create_proc_ctx_space(uq_mgr);
+	if (r) {
+		DRM_ERROR("Failed to allocate MES process context space bo, error: %d\n", r);
+		goto free_mqd;
+	}
+
+	/* Create BO of a gang for FW operations */
 	r = mes_userq_create_ctx_space(uq_mgr, queue, mqd_user);
 	if (r) {
 		DRM_ERROR("Failed to allocate BO for userqueue (%d)", r);
@@ -492,7 +519,7 @@ static int mes_userq_preempt(struct amdgpu_usermode_queue *queue)
 	*fence_ptr = 0;
 
 	memset(&queue_input, 0x0, sizeof(struct mes_suspend_gang_input));
-	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+	queue_input.gang_context_addr = ctx->gpu_addr;
 	queue_input.suspend_fence_addr = fence_gpu_addr;
 	queue_input.suspend_fence_value = 1;
 	amdgpu_mes_lock(&adev->mes);
@@ -529,7 +556,7 @@ static int mes_userq_restore(struct amdgpu_usermode_queue *queue)
 		return 0;
 
 	memset(&queue_input, 0x0, sizeof(struct mes_resume_gang_input));
-	queue_input.gang_context_addr = ctx->gpu_addr + AMDGPU_USERQ_PROC_CTX_SZ;
+	queue_input.gang_context_addr = ctx->gpu_addr;
 
 	amdgpu_mes_lock(&adev->mes);
 	r = adev->mes.funcs->resume_gang(&adev->mes, &queue_input);