gpu: nvgpu: fix inst block leak for vidmem

Test for size, not cpu_va, to check for buffer validity before
attempting to free.

Jira DNVGPU-22

Change-Id: I416c0963bf4e1819aa2f8d200c69a2d989524f83
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: http://git-master/r/1215575
(cherry picked from commit ce0077feca55bfb5665c82972598a075abd8f2a0)
Reviewed-on: http://git-master/r/1219702
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 1 insertions, 1 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
index 54d85267..70bdbc04 100644
--- a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
@@ -4680,7 +4680,7 @@ int gk20a_alloc_inst_block(struct gk20a *g, struct mem_desc *inst_block)
 
 void gk20a_free_inst_block(struct gk20a *g, struct mem_desc *inst_block)
 {
-        if (inst_block->cpu_va)
+        if (inst_block->size)
                 gk20a_gmmu_free(g, inst_block);
 }