gpu: nvgpu: Balance usage count on resume fail

If PM runtime resume fails, pm_runtime_get_sync() still increments
the usage count. Balance the usage count by decrementing it on
fail.

Bug 200003289

Change-Id: I127d2697ff2601d4884a4ecfdec8ad50894bf7d0
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: http://git-master/r/411285
(cherry picked from commit 736846ff999bb9d1ca3340fc02ab49f8c65c4145)
Reviewed-on: http://git-master/r/411473
Reviewed-by: Automatic_Commit_Validation_User
GVS: Gerrit_Virtual_Submit

1 files changed, 2 insertions, 0 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.c b/drivers/gpu/nvgpu/gk20a/gk20a.c
index f04e9187..2e406474 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.c
@@ -1514,6 +1514,8 @@ int gk20a_busy(struct platform_device *pdev)
 
 #ifdef CONFIG_PM_RUNTIME
         ret = pm_runtime_get_sync(&pdev->dev);
+        if (ret < 0)
+                pm_runtime_put_noidle(&pdev->dev);
 #endif
         gk20a_scale_notify_busy(pdev);