gpu: nvgpu: fix race between do_idle() and unrailgate()

While we are executing do_idle() API, it is possible that
unrailgate() gets invoked in midst of idling the GPU and
this can result in failure of do_idle()

To prevent simultaneous execution of these methods,
add a mutex railgate_lock and acquire it during
do_idle() and unrailgate() APIs

Also, keep this lock held if do_idle() is successful.
In success, lock will be released in do_unidle(),
otherwise release this lock before returning

Note that this lock should not be held in railgate() API
since we do not want it to be blocked during do_idle()

bug 1529160

Change-Id: I87114b5367eaa217376455a2699c0d21c451c889
Signed-off-by: Deepak Nibade <dnibade@nvidia.com>
Reviewed-on: http://git-master/r/434190
(cherry picked from commit 561dc8e0933ff2d72573292968b893a52f5f783a)
Reviewed-on: http://git-master/r/435131
Reviewed-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-by: Sachin Nikam <snikam@nvidia.com>

2 files changed, 21 insertions, 3 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.c b/drivers/gpu/nvgpu/gk20a/gk20a.c
index 629423c0..64be9398 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.c
@@ -1281,8 +1281,13 @@ static int _gk20a_pm_unrailgate(struct platform_device *pdev)
 {
         struct gk20a_platform *platform = platform_get_drvdata(pdev);
         int ret = 0;
-        if (platform->unrailgate)
+
+        if (platform->unrailgate) {
+                mutex_lock(&platform->railgate_lock);
                 ret = platform->unrailgate(pdev);
+                mutex_unlock(&platform->railgate_lock);
+        }
+
         return ret;
 }
 
@@ -1362,6 +1367,8 @@ static int gk20a_pm_init(struct platform_device *dev)
 
         gk20a_dbg_fn("");
 
+        mutex_init(&platform->railgate_lock);
+
         /* Initialise pm runtime */
         if (platform->clockgate_delay) {
                 pm_runtime_set_autosuspend_delay(&dev->dev,
@@ -1739,6 +1746,13 @@ int gk20a_do_idle(void)
         /* acquire busy lock to block other busy() calls */
         down_write(&g->busy_lock);
 
+        /* acquire railgate lock to prevent unrailgate in midst of do_idle() */
+        mutex_lock(&platform->railgate_lock);
+
+        /* check if it is already railgated ? */
+        if (platform->is_railgated(pdev))
+                return 0;
+
         /* prevent suspend by incrementing usage counter */
         pm_runtime_get_noresume(&pdev->dev);
 
@@ -1776,11 +1790,12 @@ int gk20a_do_idle(void)
         }
 
         /* GPU is not rail gated by now, return error */
-        up_write(&g->busy_lock);
-        return -EBUSY;
+        goto fail_timeout;
 
 fail:
         pm_runtime_put_noidle(&pdev->dev);
+fail_timeout:
+        mutex_unlock(&platform->railgate_lock);
         up_write(&g->busy_lock);
         return -EBUSY;
 }
@@ -1794,8 +1809,10 @@ int gk20a_do_unidle(void)
                 bus_find_device_by_name(&platform_bus_type,
                 NULL, "gk20a.0"));
         struct gk20a *g = get_gk20a(pdev);
+        struct gk20a_platform *platform = dev_get_drvdata(&pdev->dev);
 
         /* release the lock and open up all other busy() calls */
+        mutex_unlock(&platform->railgate_lock);
         up_write(&g->busy_lock);
 
         return 0;
diff --git a/drivers/gpu/nvgpu/gk20a/platform_gk20a.h b/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
index 231f97b3..76e9cf6c 100644
--- a/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/platform_gk20a.h
@@ -118,6 +118,7 @@ struct gk20a_platform {
 
         /* Called to turn on the device */
         int (*unrailgate)(struct platform_device *dev);
+        struct mutex railgate_lock;
 
         /* Called to check state of device */
         bool (*is_railgated)(struct platform_device *dev);