gpu: nvgpu: rename has_timedout and make it thread safe

Currently has_timedout variable is protected by wmb at places
where it is being set and there is no correspoding rmb whenever
has_timedout variable is read. This is prone to errors for
concurrent execution. This change is supposed to fix this issue.
Rename has_timedout variable of channel struct to ch_timedout.
Also to avoid rmb every time ch_timedout is read,
ch_timedout_spinlock is added to protect ch_timedout
variable for taking care of concurrent execution.

Bug 2404865
Bug 2092051

Change-Id: I0bee9f50af0a48720aa8b54cbc3af97ef9f6df00
Signed-off-by: Seema Khowala <seemaj@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1930935
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
(cherry picked from commit 1f54ea09e3445d9ca3cf7a69b4967849cc9defc8
in dev-kernel)
Reviewed-on: https://git-master.nvidia.com/r/2016975
GVS: Gerrit_Virtual_Submit
Reviewed-by: Alex Waterman <alexw@nvidia.com>
Reviewed-by: Bibek Basu <bbasu@nvidia.com>
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

3 files changed, 15 insertions, 9 deletions

diff --git a/drivers/gpu/nvgpu/os/linux/cde.c b/drivers/gpu/nvgpu/os/linux/cde.c
index a15ef560..7b2cba7d 100644
--- a/drivers/gpu/nvgpu/os/linux/cde.c
+++ b/drivers/gpu/nvgpu/os/linux/cde.c
@@ -1272,7 +1272,7 @@ __releases(&cde_app->mutex)
                 nvgpu_log_info(g, "double finish cde context %p on channel %p",
                                 cde_ctx, ch);
 
-        if (ch->has_timedout) {
+        if (gk20a_channel_check_timedout(ch)) {
                 if (cde_ctx->is_temporary) {
                         nvgpu_warn(g,
                                         "cde: channel had timed out"
@@ -1299,8 +1299,9 @@ __releases(&cde_app->mutex)
                         msecs_to_jiffies(CTX_DELETE_TIME));
         }
 
-        if (!ch->has_timedout)
+        if (!gk20a_channel_check_timedout(ch)) {
                 gk20a_cde_ctx_release(cde_ctx);
+        }
 }
 
 static int gk20a_cde_load(struct gk20a_cde_ctx *cde_ctx)
diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_channel.c b/drivers/gpu/nvgpu/os/linux/ioctl_channel.c
index 45d49474..d0d4b1af 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_channel.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_channel.c
@@ -636,8 +636,9 @@ static int gk20a_channel_wait_semaphore(struct channel_gk20a *ch,
         int ret = 0;
 
         /* do not wait if channel has timed out */
-        if (ch->has_timedout)
+        if (gk20a_channel_check_timedout(ch)) {
                 return -ETIMEDOUT;
+        }
 
         dmabuf = dma_buf_get(id);
         if (IS_ERR(dmabuf)) {
@@ -656,7 +657,8 @@ static int gk20a_channel_wait_semaphore(struct channel_gk20a *ch,
 
         ret = NVGPU_COND_WAIT_INTERRUPTIBLE(
                         &ch->semaphore_wq,
-                        *semaphore == payload || ch->has_timedout,
+                        *semaphore == payload ||
+                        gk20a_channel_check_timedout(ch),
                         timeout);
 
         dma_buf_kunmap(dmabuf, offset >> PAGE_SHIFT, data);
@@ -680,8 +682,9 @@ static int gk20a_channel_wait(struct channel_gk20a *ch,
 
         nvgpu_log_fn(g, " ");
 
-        if (ch->has_timedout)
+        if (gk20a_channel_check_timedout(ch)) {
                 return -ETIMEDOUT;
+        }
 
         switch (args->type) {
         case NVGPU_WAIT_TYPE_NOTIFIER:
@@ -716,7 +719,8 @@ static int gk20a_channel_wait(struct channel_gk20a *ch,
                  * calling this ioctl */
                 remain = NVGPU_COND_WAIT_INTERRUPTIBLE(
                                 &ch->notifier_wq,
-                                notif->status == 0 || ch->has_timedout,
+                                notif->status == 0 ||
+                                gk20a_channel_check_timedout(ch),
                                 args->timeout);
 
                 if (remain == 0 && notif->status != 0) {
@@ -786,8 +790,9 @@ static int gk20a_ioctl_channel_submit_gpfifo(
         profile = gk20a_fifo_profile_acquire(ch->g);
         gk20a_fifo_profile_snapshot(profile, PROFILE_IOCTL_ENTRY);
 
-        if (ch->has_timedout)
+        if (gk20a_channel_check_timedout(ch)) {
                 return -ETIMEDOUT;
+        }
 
         nvgpu_get_fence_args(&args->fence, &fence);
         submit_flags =
@@ -1249,7 +1254,7 @@ long gk20a_channel_ioctl(struct file *filp,
         }
         case NVGPU_IOCTL_CHANNEL_GET_TIMEDOUT:
                 ((struct nvgpu_get_param_args *)buf)->value =
-                        ch->has_timedout;
+                        gk20a_channel_check_timedout(ch);
                 break;
         case NVGPU_IOCTL_CHANNEL_ENABLE:
                 err = gk20a_busy(ch->g);
diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c b/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c
index b0cdf5e5..c5079bd6 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c
@@ -141,7 +141,7 @@ static int gk20a_tsg_unbind_channel_fd(struct tsg_gk20a *tsg, int ch_fd)
          * Mark the channel timedout since channel unbound from TSG
          * has no context of its own so it can't serve any job
          */
-        ch->has_timedout = true;
+        gk20a_channel_set_timedout(ch);
 
 out:
         gk20a_channel_put(ch);