gpu: nvgpu: warn if cde fails to find addr to patch

Print the surface address for which we fail to resolve an iova address
when patcing cde parameters. This appears to happen extremely rarely for
yet unknown reasons.

Bug 2038362

Change-Id: I5ca300ea9b2f8c8867b7b43e37f51a50836129b7
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1748455
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 4 insertions, 1 deletions

diff --git a/drivers/gpu/nvgpu/common/linux/cde.c b/drivers/gpu/nvgpu/common/linux/cde.c
index 8a15b82a..1ada0793 100644
--- a/drivers/gpu/nvgpu/common/linux/cde.c
+++ b/drivers/gpu/nvgpu/common/linux/cde.c
@@ -442,8 +442,11 @@ static int gk20a_cde_patch_params(struct gk20a_cde_ctx *cde_ctx)
                 case TYPE_PARAM_SOURCE_SMMU_ADDR:
                         new_data = gpuva_to_iova_base(cde_ctx->vm,
                                                       cde_ctx->surf_vaddr);
-                        if (new_data == 0)
+                        if (new_data == 0) {
+                                nvgpu_warn(g, "cde: failed to find 0x%llx",
+                                                cde_ctx->surf_vaddr);
                                 return -EINVAL;
+                        }
                         break;
                 case TYPE_PARAM_BACKINGSTORE_BASE_HW:
                         new_data = g->gr.compbit_store.base_hw;