gpu: nvgpu: Fix the GPFIFO submit condition

GPFIFO HW with fifo size N can actually accept
N-1 entries. Also, kernel can insert two extra
entries, before and after the user GPFIFOs.

So, GPFIFO with N size can accept max N-3 user
gpfifos.

Bug 1613125

Change-Id: I173526afb70dddf1b2b9ec0a99890335c81f0e02
Signed-off-by: Gagan Grover <ggrover@nvidia.com>
Reviewed-on: http://git-master/r/725380
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>

1 files changed, 5 insertions, 2 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
index 7fa8a552..328f49b1 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_gk20a.c
@@ -1460,8 +1460,11 @@ int gk20a_submit_channel_gpfifo(struct channel_gk20a *c,
         if (c->has_timedout)
                 return -ETIMEDOUT;
 
-        /* fifo not large enough for request. Return error immediately */
-        if (c->gpfifo.entry_num < num_entries) {
+        /* fifo not large enough for request. Return error immediately.
+         * Kernel can insert gpfifo entries before and after user gpfifos.
+         * So, add extra_entries in user request. Also, HW with fifo size N
+         * can accept only N-1 entreis and so the below condition */
+        if (c->gpfifo.entry_num - 1 < num_entries + extra_entries) {
                 gk20a_err(d, "not enough gpfifo space allocated");
                 return -ENOMEM;
         }