drm/msm: print offender task name on hangcheck recovery

Track the pid per submit, so we can print the name of the task which
submitted the batch that caused the gpu to hang.

Signed-off-by: Rob Clark <robdclark@gmail.com>

diff --git a/drivers/gpu/drm/msm/msm_gem.h b/drivers/gpu/drm/msm/msm_gem.h
index 2771d155587afe486b1c42552e2ede4aa4ed0b3d..9facd4b6ffd9366a5a8105bf4a0c3d33f4a439f8 100644 (file)
--- a/drivers/gpu/drm/msm/msm_gem.h
+++ b/drivers/gpu/drm/msm/msm_gem.h
@@ -86,6 +86,7 @@ struct msm_gem_submit {
        struct list_head bo_list;
        struct ww_acquire_ctx ticket;
        struct fence *fence;
+       struct pid *pid;    /* submitting process */
        bool valid;         /* true if no cmdstream patching needed */
        unsigned int nr_cmds;
        unsigned int nr_bos;

diff --git a/drivers/gpu/drm/msm/msm_gem_submit.c b/drivers/gpu/drm/msm/msm_gem_submit.c
index 711a658cf1ea8772df99b2ce05ce5ad2e296980b..a9a001150b9edae73b4a588873003997fdc447f0 100644 (file)
--- a/drivers/gpu/drm/msm/msm_gem_submit.c
+++ b/drivers/gpu/drm/msm/msm_gem_submit.c
@@ -45,6 +45,7 @@ static struct msm_gem_submit *submit_create(struct drm_device *dev,
 
        submit->dev = dev;
        submit->gpu = gpu;
+       submit->pid = get_pid(task_pid(current));
 
        /* initially, until copy_from_user() and bo lookup succeeds: */
        submit->nr_bos = 0;
@@ -60,6 +61,7 @@ void msm_gem_submit_free(struct msm_gem_submit *submit)
 {
        fence_put(submit->fence);
        list_del(&submit->node);
+       put_pid(submit->pid);
        kfree(submit);
 }
 

diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
index 5aed93a87e4c1b20dc61539785196048faecab64..36ed53e661fe81f86a403a3ff30ee895983510ac 100644 (file)
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -272,16 +272,30 @@ static void recover_worker(struct work_struct *work)
 {
        struct msm_gpu *gpu = container_of(work, struct msm_gpu, recover_work);
        struct drm_device *dev = gpu->dev;
+       struct msm_gem_submit *submit;
        uint32_t fence = gpu->funcs->last_fence(gpu);
 
-       dev_err(dev->dev, "%s: hangcheck recover!\n", gpu->name);
-
        msm_update_fence(gpu->fctx, fence + 1);
 
        mutex_lock(&dev->struct_mutex);
-       if (msm_gpu_active(gpu)) {
-               struct msm_gem_submit *submit;
 
+       dev_err(dev->dev, "%s: hangcheck recover!\n", gpu->name);
+       list_for_each_entry(submit, &gpu->submit_list, node) {
+               if (submit->fence->seqno == (fence + 1)) {
+                       struct task_struct *task;
+
+                       rcu_read_lock();
+                       task = pid_task(submit->pid, PIDTYPE_PID);
+                       if (task) {
+                               dev_err(dev->dev, "%s: offending task: %s\n",
+                                               gpu->name, task->comm);
+                       }
+                       rcu_read_unlock();
+                       break;
+               }
+       }
+
+       if (msm_gpu_active(gpu)) {
                /* retire completed submits, plus the one that hung: */
                retire_submits(gpu);
 
@@ -293,6 +307,7 @@ static void recover_worker(struct work_struct *work)
                        gpu->funcs->submit(gpu, submit, NULL);
                }
        }
+
        mutex_unlock(&dev->struct_mutex);
 
        msm_gpu_retire(gpu);