gpu: host1x: Rewrite job opcode sequence

For new (Tegra186+) SoCs, use a new ('full-featured') job opcode
sequence that is compatible with virtualization. In particular,
the Host1x hardware in Tegra234 is more strict regarding the sequence,
requiring ACQUIRE_MLOCK-SETCLASS-SETSTREAMID opcodes to occur in
that sequence without gaps (except for SETPAYLOAD), so let's do it
properly in one go now.
Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>

drivers/gpu/host1x/hw/channel_hw.c

@@ -47,10 +47,41 @@ static void trace_write_gather(struct host1x_cdma *cdma, struct host1x_bo *bo,
 	}
 }
 
-static void submit_wait(struct host1x_cdma *cdma, u32 id, u32 threshold,
+static void submit_wait(struct host1x_job *job, u32 id, u32 threshold,
 			u32 next_class)
 {
-#if HOST1X_HW >= 2
+	struct host1x_cdma *cdma = &job->channel->cdma;
+
+#if HOST1X_HW >= 6
+	u32 stream_id;
+
+	/*
+	 * If a memory context has been set, use it. Otherwise
+	 * (if context isolation is disabled) use the engine's
+	 * firmware stream ID.
+	 */
+	if (job->memory_context)
+		stream_id = job->memory_context->stream_id;
+	else
+		stream_id = job->engine_fallback_streamid;
+
+	host1x_cdma_push_wide(cdma,
+		host1x_opcode_setclass(
+			HOST1X_CLASS_HOST1X,
+			HOST1X_UCLASS_LOAD_SYNCPT_PAYLOAD_32,
+			/* WAIT_SYNCPT_32 is at SYNCPT_PAYLOAD_32+2 */
+			BIT(0) | BIT(2)
+		),
+		threshold,
+		id,
+		HOST1X_OPCODE_NOP
+	);
+	host1x_cdma_push_wide(&job->channel->cdma,
+		host1x_opcode_setclass(job->class, 0, 0),
+		host1x_opcode_setpayload(stream_id),
+		host1x_opcode_setstreamid(job->engine_streamid_offset / 4),
+		HOST1X_OPCODE_NOP);
+#elif HOST1X_HW >= 2
 	host1x_cdma_push_wide(cdma,
 		host1x_opcode_setclass(
 			HOST1X_CLASS_HOST1X,
@@ -97,7 +128,7 @@ static void submit_gathers(struct host1x_job *job, u32 job_syncpt_base)
 			else
 				threshold = cmd->wait.threshold;
 
-			submit_wait(cdma, cmd->wait.id, threshold, cmd->wait.next_class);
+			submit_wait(job, cmd->wait.id, threshold, cmd->wait.next_class);
 		} else {
 			struct host1x_job_gather *g = &cmd->gather;
 
@@ -180,42 +211,70 @@ static void host1x_enable_gather_filter(struct host1x_channel *ch)
 #endif
 }
 
-static void host1x_channel_program_engine_streamid(struct host1x_job *job)
+static void channel_program_cdma(struct host1x_job *job)
 {
+	struct host1x_cdma *cdma = &job->channel->cdma;
+	struct host1x_syncpt *sp = job->syncpt;
+
 #if HOST1X_HW >= 6
 	u32 fence;
 
-	if (!job->memory_context)
-		return;
-
-	fence = host1x_syncpt_incr_max(job->syncpt, 1);
-
-	/* First, increment a syncpoint on OP_DONE condition.. */
+	/* Enter engine class with invalid stream ID. */
+	host1x_cdma_push_wide(cdma,
+		host1x_opcode_acquire_mlock(job->class),
+		host1x_opcode_setclass(job->class, 0, 0),
+		host1x_opcode_setpayload(0),
+		host1x_opcode_setstreamid(job->engine_streamid_offset / 4));
 
+	/* Before switching stream ID to real stream ID, ensure engine is idle. */
+	fence = host1x_syncpt_incr_max(sp, 1);
 	host1x_cdma_push(&job->channel->cdma,
 		host1x_opcode_nonincr(HOST1X_UCLASS_INCR_SYNCPT, 1),
 		HOST1X_UCLASS_INCR_SYNCPT_INDX_F(job->syncpt->id) |
-			HOST1X_UCLASS_INCR_SYNCPT_COND_F(1));
+			HOST1X_UCLASS_INCR_SYNCPT_COND_F(4));
+	submit_wait(job, job->syncpt->id, fence, job->class);
 
-	/* Wait for syncpoint to increment */
+	/* Submit work. */
+	job->syncpt_end = host1x_syncpt_incr_max(sp, job->syncpt_incrs);
+	submit_gathers(job, job->syncpt_end - job->syncpt_incrs);
 
+	/* Before releasing MLOCK, ensure engine is idle again. */
+	fence = host1x_syncpt_incr_max(sp, 1);
 	host1x_cdma_push(&job->channel->cdma,
-		host1x_opcode_setclass(HOST1X_CLASS_HOST1X,
-			host1x_uclass_wait_syncpt_r(), 1),
-		host1x_class_host_wait_syncpt(job->syncpt->id, fence));
+		host1x_opcode_nonincr(HOST1X_UCLASS_INCR_SYNCPT, 1),
+		HOST1X_UCLASS_INCR_SYNCPT_INDX_F(job->syncpt->id) |
+			HOST1X_UCLASS_INCR_SYNCPT_COND_F(4));
+	submit_wait(job, job->syncpt->id, fence, job->class);
 
+	/* Release MLOCK. */
+	host1x_cdma_push(cdma,
+		HOST1X_OPCODE_NOP, host1x_opcode_release_mlock(job->class));
+#else
+	if (job->serialize) {
 		/*
-	 * Now that we know the engine is idle, return to class and
-	 * change stream ID.
+		 * Force serialization by inserting a host wait for the
+		 * previous job to finish before this one can commence.
 		 */
+		host1x_cdma_push(cdma,
+				 host1x_opcode_setclass(HOST1X_CLASS_HOST1X,
+					host1x_uclass_wait_syncpt_r(), 1),
+				 host1x_class_host_wait_syncpt(job->syncpt->id,
+					host1x_syncpt_read_max(sp)));
+	}
 
-	host1x_cdma_push(&job->channel->cdma,
+	/* Synchronize base register to allow using it for relative waiting */
+	if (sp->base)
+		synchronize_syncpt_base(job);
+
+	/* add a setclass for modules that require it */
+	if (job->class)
+		host1x_cdma_push(cdma,
 				 host1x_opcode_setclass(job->class, 0, 0),
 				 HOST1X_OPCODE_NOP);
 
-	host1x_cdma_push(&job->channel->cdma,
-		host1x_opcode_setpayload(job->memory_context->stream_id),
-		host1x_opcode_setstreamid(job->engine_streamid_offset / 4));
+	job->syncpt_end = host1x_syncpt_incr_max(sp, job->syncpt_incrs);
+
+	submit_gathers(job, job->syncpt_end - job->syncpt_incrs);
 #endif
 }
 
@@ -223,7 +282,6 @@ static int channel_submit(struct host1x_job *job)
 {
 	struct host1x_channel *ch = job->channel;
 	struct host1x_syncpt *sp = job->syncpt;
-	u32 user_syncpt_incrs = job->syncpt_incrs;
 	u32 prev_max = 0;
 	u32 syncval;
 	int err;
@@ -251,6 +309,7 @@ static int channel_submit(struct host1x_job *job)
 
 	host1x_channel_set_streamid(ch);
 	host1x_enable_gather_filter(ch);
+	host1x_hw_syncpt_assign_to_channel(host, sp, ch);
 
 	/* begin a CDMA submit */
 	err = host1x_cdma_begin(&ch->cdma, job);
@@ -259,40 +318,7 @@ static int channel_submit(struct host1x_job *job)
 		goto error;
 	}
 
-	if (job->serialize) {
-		/*
-		 * Force serialization by inserting a host wait for the
-		 * previous job to finish before this one can commence.
-		 */
-		host1x_cdma_push(&ch->cdma,
-				 host1x_opcode_setclass(HOST1X_CLASS_HOST1X,
-					host1x_uclass_wait_syncpt_r(), 1),
-				 host1x_class_host_wait_syncpt(job->syncpt->id,
-					host1x_syncpt_read_max(sp)));
-	}
-
-	/* Synchronize base register to allow using it for relative waiting */
-	if (sp->base)
-		synchronize_syncpt_base(job);
-
-	host1x_hw_syncpt_assign_to_channel(host, sp, ch);
-
-	/* add a setclass for modules that require it */
-	if (job->class)
-		host1x_cdma_push(&ch->cdma,
-				 host1x_opcode_setclass(job->class, 0, 0),
-				 HOST1X_OPCODE_NOP);
-
-	/*
-	 * Ensure engine DMA is idle and set new stream ID. May increment
-	 * syncpt max.
-	 */
-	host1x_channel_program_engine_streamid(job);
-
-	syncval = host1x_syncpt_incr_max(sp, user_syncpt_incrs);
-	job->syncpt_end = syncval;
-
-	submit_gathers(job, syncval - user_syncpt_incrs);
+	channel_program_cdma(job);
 
 	/* end CDMA submit & stash pinned hMems into sync queue */
 	host1x_cdma_end(&ch->cdma, job);