gpu: host1x: debug: Dump only relevant parts of CDMA push buffer

Dumping the full CDMA push buffer takes a long time and isn't very
useful since most of the contents are not relevant. Instead only show
the CDMA push buffer entries associated with current jobs.

While at it, tweak the indentation a bit to make the output more
readable.

Signed-off-by: Thierry Reding <[e-mail]>

drivers/gpu/host1x/hw/debug_hw.c

···
 		u32 val = *(map_addr + offset / 4 + i);
 
 		if (!data_count) {
-			host1x_debug_output(o, "%pad: %08x: ", &addr, val);
+			host1x_debug_output(o, "    %pad: %08x: ", &addr, val);
 			data_count = show_channel_command(o, val, &payload);
 		} else {
 			host1x_debug_cont(o, "%08x%s", val,
···
 	struct push_buffer *pb = &cdma->push_buffer;
 	struct host1x_job *job;
 
-	host1x_debug_output(o, "PUSHBUF at %pad, %u words\n",
-			    &pb->dma, pb->size / 4);
-
-	show_gather(o, pb->dma, pb->size / 4, cdma, pb->dma, pb->mapped);
-
 	list_for_each_entry(job, &cdma->sync_queue, list) {
 		unsigned int i;
 
-		host1x_debug_output(o, "\n%p: JOB, syncpt_id=%d, syncpt_val=%d, first_get=%08x, timeout=%d num_slots=%d, num_handles=%d\n",
-				    job, job->syncpt->id, job->syncpt_end,
-				    job->first_get, job->timeout,
+		host1x_debug_output(o, "JOB, syncpt %u: %u timeout: %u num_slots: %u num_handles: %u\n",
+				    job->syncpt->id, job->syncpt_end, job->timeout,
 				    job->num_slots, job->num_unpins);
+
+		show_gather(o, pb->dma + job->first_get, job->num_slots * 2, cdma,
+			    pb->dma + job->first_get, pb->mapped + job->first_get);
 
 		for (i = 0; i < job->num_cmds; i++) {
 			struct host1x_job_gather *g;
···
 				continue;
 			}
 
-			host1x_debug_output(o, "    GATHER at %pad+%#x, %d words\n",
+			host1x_debug_output(o, "  GATHER at %pad+%#x, %d words\n",
 					    &g->base, g->offset, g->words);
 
 			show_gather(o, g->base + g->offset, g->words, cdma,