gpu: nvgpu: Correctly plumb -EAGAIN from vidmem allocations

Userspace can and should retry vidmem allocations if there are pending
clears still to be executed by the GPU. But this requires the -EAGAIN
to properly propagate back to userspace.

Bug 200378648

Change-Id: Ib930711270439843e043d65c2e87b60612a76239
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1669099
Reviewed-by: svc-mobile-coverity <svc-mobile-coverity@nvidia.com>
GVS: Gerrit_Virtual_Submit
Reviewed-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-by: Vijayakumar Subbu <vsubbu@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

1 files changed, 4 insertions, 2 deletions

diff --git a/drivers/gpu/nvgpu/common/linux/vidmem.c b/drivers/gpu/nvgpu/common/linux/vidmem.c
index 1d9fea71..684262ba 100644
--- a/drivers/gpu/nvgpu/common/linux/vidmem.c
+++ b/drivers/gpu/nvgpu/common/linux/vidmem.c
@@ -173,6 +173,8 @@ int nvgpu_vidmem_export_linux(struct gk20a *g, size_t bytes)
         if (!gk20a_get(g))
                 return -ENODEV;
 
+        vidmem_dbg(g, "Allocating vidmem buf: %zu bytes", bytes);
+
         priv = nvgpu_kzalloc(g, sizeof(*priv));
         if (!priv) {
                 err = -ENOMEM;
@@ -180,8 +182,8 @@ int nvgpu_vidmem_export_linux(struct gk20a *g, size_t bytes)
         }
 
         buf = nvgpu_vidmem_user_alloc(g, bytes);
-        if (!buf) {
-                err = -ENOMEM;
+        if (IS_ERR(buf)) {
+                err = PTR_ERR(buf);
                 goto fail;
         }