gpu: nvgpu: fix race condition in pmu_rpc_handler

There was a race condition between pmu_rpc_handler and
nvgpu_pmu_rpc_execute. The later could free the buffer before
pmu_rpc_handler could access related data.
Added explicit field in rpc_payload, so that nvgpu_pmu_rpc_execute
can wait until pmu_rpc_handler completes.

Bug 2331655

Change-Id: Ic2653524159eff10504b9c2625b5241610b5f5f0
Reviewed-on: https://git-master.nvidia.com/r/1811299
Signed-off-by: Vaikundanathan S <vaikuns@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1817582
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 1 insertions, 0 deletions

diff --git a/drivers/gpu/nvgpu/include/nvgpu/pmu.h b/drivers/gpu/nvgpu/include/nvgpu/pmu.h
index 81abdb24..7283755a 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/pmu.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/pmu.h
@@ -192,6 +192,7 @@ typedef void (*pmu_callback)(struct gk20a *, struct pmu_msg *, void *, u32,
 struct rpc_handler_payload {
         void *rpc_buff;
         bool is_mem_free_set;
+        bool complete;
 };
 
 struct pmu_rpc_desc {