drm/amdkfd: pause autosuspend when creating pdd

[ Upstream commit 438b39ac74e2a9dc0a5c9d653b7d8066877e86b1 ]

When using MES creating a pdd will require talking to the GPU to
setup the relevant context. The code here forgot to wake up the GPU
in case it was in suspend, this causes KVM to EFAULT for passthrough
GPU for example. This issue can be masked if the GPU was woken up by
other things (e.g. opening the KMS node) first and have not yet gone to sleep.

v4: do the allocation of proc_ctx_bo in a lazy fashion
when the first queue is created in a process (Felix)

Signed-off-by: Jesse Zhang <jesse.zhang@amd.com>
Reviewed-by: Yunxiang Li <Yunxiang.Li@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
index 4d9a406925e18952eadd46d90395366c9e7ee948..43fa260ddbcea056ce813537f61e1a0ecb5c8f6c 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c
@@ -197,6 +197,21 @@ static int add_queue_mes(struct device_queue_manager *dqm, struct queue *q,
        if (dqm->is_hws_hang)
                return -EIO;
 
+       if (!pdd->proc_ctx_cpu_ptr) {
+               r = amdgpu_amdkfd_alloc_gtt_mem(adev,
+                               AMDGPU_MES_PROC_CTX_SIZE,
+                               &pdd->proc_ctx_bo,
+                               &pdd->proc_ctx_gpu_addr,
+                               &pdd->proc_ctx_cpu_ptr,
+                               false);
+               if (r) {
+                       dev_err(adev->dev,
+                               "failed to allocate process context bo\n");
+                       return r;
+               }
+               memset(pdd->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
+       }
+
        memset(&queue_input, 0x0, sizeof(struct mes_add_queue_input));
        queue_input.process_id = qpd->pqm->process->pasid;
        queue_input.page_table_base_addr =  qpd->page_table_base;

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
index 577bdb6a9640a94b3553cdfd84c636c316af122a..64346c71c62a3001ed98863e5e0dc3052965f258 100644 (file)
--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c
@@ -1046,7 +1046,8 @@ static void kfd_process_destroy_pdds(struct kfd_process *p)
 
                kfd_free_process_doorbells(pdd->dev->kfd, pdd);
 
-               if (pdd->dev->kfd->shared_resources.enable_mes)
+               if (pdd->dev->kfd->shared_resources.enable_mes &&
+                       pdd->proc_ctx_cpu_ptr)
                        amdgpu_amdkfd_free_gtt_mem(pdd->dev->adev,
                                                   &pdd->proc_ctx_bo);
                /*
@@ -1572,7 +1573,6 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev,
                                                        struct kfd_process *p)
 {
        struct kfd_process_device *pdd = NULL;
-       int retval = 0;
 
        if (WARN_ON_ONCE(p->n_pdds >= MAX_GPU_INSTANCE))
                return NULL;
@@ -1596,21 +1596,6 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev,
        pdd->user_gpu_id = dev->id;
        atomic64_set(&pdd->evict_duration_counter, 0);
 
-       if (dev->kfd->shared_resources.enable_mes) {
-               retval = amdgpu_amdkfd_alloc_gtt_mem(dev->adev,
-                                               AMDGPU_MES_PROC_CTX_SIZE,
-                                               &pdd->proc_ctx_bo,
-                                               &pdd->proc_ctx_gpu_addr,
-                                               &pdd->proc_ctx_cpu_ptr,
-                                               false);
-               if (retval) {
-                       dev_err(dev->adev->dev,
-                               "failed to allocate process context bo\n");
-                       goto err_free_pdd;
-               }
-               memset(pdd->proc_ctx_cpu_ptr, 0, AMDGPU_MES_PROC_CTX_SIZE);
-       }
-
        p->pdds[p->n_pdds++] = pdd;
        if (kfd_dbg_is_per_vmid_supported(pdd->dev))
                pdd->spi_dbg_override = pdd->dev->kfd2kgd->disable_debug_trap(
@@ -1622,10 +1607,6 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_node *dev,
        idr_init(&pdd->alloc_idr);
 
        return pdd;
-
-err_free_pdd:
-       kfree(pdd);
-       return NULL;
 }
 
 /**