gpu: nvgpu: Support semaphore sync when aborting jobs

When aborting jobs on channel error situations, we manually set
the channel syncpoint's min == max in gk20a_disable_channel_no_update.
Nvhost will notice this manual syncpoint increment, and will call back
to gk20a_channel_update, which will clean up the job.

With semaphore synchronization, we don't have anybody calling back to
gk20a_channel_update, so we need to call it ourselves. Release job
semaphores (the equivalent of set_min_eq_max) on
gk20a_disable_channel_no_update, and if any semaphores were released,
call gk20a_channel_update afterwards.

Because we are actually calling gk20a_channel_update in some situations,
gk20a_disable_channel_no_update is no longer an appropriate name for the
function. Rename it to gk20a_channel_abort.

Bug 1450122

Change-Id: I1267b099a5778041cbc8e91b7184844812145b93
Signed-off-by: Lauri Peltonen <lpeltonen@nvidia.com>
Reviewed-on: http://git-master/r/422161
GVS: Gerrit_Virtual_Submit
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>

drivers/gpu/nvgpu/gk20a/channel_gk20a.c

@@ -406,17 +406,36 @@ static int channel_gk20a_update_runlist(struct channel_gk20a *c, bool add)
 	return gk20a_fifo_update_runlist(c->g, 0, c->hw_chid, add, true);
 }
 
-void gk20a_disable_channel_no_update(struct channel_gk20a *ch)
+void gk20a_channel_abort(struct channel_gk20a *ch)
 {
+	struct channel_gk20a_job *job, *n;
+	bool released_job_semaphore = false;
+
 	/* ensure no fences are pending */
 	if (ch->sync)
 		ch->sync->set_min_eq_max(ch->sync);
 
+	/* release all job semaphores (applies only to jobs that use
+	   semaphore synchronization) */
+	mutex_lock(&ch->jobs_lock);
+	list_for_each_entry_safe(job, n, &ch->jobs, list) {
+		if (job->post_fence.semaphore) {
+			gk20a_semaphore_release(job->post_fence.semaphore);
+			released_job_semaphore = true;
+		}
+	}
+	mutex_unlock(&ch->jobs_lock);
+
 	/* disable channel */
 	gk20a_writel(ch->g, ccsr_channel_r(ch->hw_chid),
 		     gk20a_readl(ch->g,
 		     ccsr_channel_r(ch->hw_chid)) |
 		     ccsr_channel_enable_clr_true_f());
+
+	if (released_job_semaphore) {
+		wake_up_interruptible_all(&ch->semaphore_wq);
+		gk20a_channel_update(ch, 0);
+	}
 }
 
 int gk20a_wait_channel_idle(struct channel_gk20a *ch)
@@ -455,7 +474,7 @@ void gk20a_disable_channel(struct channel_gk20a *ch,
 	}
 
 	/* disable the channel from hw and increment syncpoints */
-	gk20a_disable_channel_no_update(ch);
+	gk20a_channel_abort(ch);
 
 	gk20a_wait_channel_idle(ch);
 

drivers/gpu/nvgpu/gk20a/channel_gk20a.h

@@ -153,7 +153,7 @@ bool gk20a_channel_update_and_check_timeout(struct channel_gk20a *ch,
 void gk20a_disable_channel(struct channel_gk20a *ch,
 			   bool wait_for_finish,
 			   unsigned long finish_timeout);
-void gk20a_disable_channel_no_update(struct channel_gk20a *ch);
+void gk20a_channel_abort(struct channel_gk20a *ch);
 int gk20a_channel_finish(struct channel_gk20a *ch, unsigned long timeout);
 void gk20a_set_error_notifier(struct channel_gk20a *ch, __u32 error);
 void gk20a_channel_semaphore_wakeup(struct gk20a *g);

drivers/gpu/nvgpu/gk20a/fifo_gk20a.c

@@ -1043,7 +1043,7 @@ static bool gk20a_fifo_handle_mmu_fault(struct gk20a *g)
 			if (ch->in_use) {
 				/* disable the channel from hw and increment
 				 * syncpoints */
-				gk20a_disable_channel_no_update(ch);
+				gk20a_channel_abort(ch);
 
 				/* remove the channel from runlist */
 				clear_bit(ch->hw_chid,
@@ -1180,7 +1180,7 @@ void gk20a_fifo_recover_ch(struct gk20a *g, u32 hw_chid, bool verbose)
 		struct channel_gk20a *ch =
 			g->fifo.channel + hw_chid;
 
-		gk20a_disable_channel_no_update(ch);
+		gk20a_channel_abort(ch);
 		for (i = 0; i < g->fifo.max_runlists; i++)
 			gk20a_fifo_update_runlist(g, i,
 					hw_chid, false, false);