Merge tag 'drm-msm-next-2017-11-01' of git://people.freedesktop.org/~robclark/linux into drm-next

 + preemption support for a5xx[1][2]

 + display fixes for 8x96 (snapdragon 820) including fixes for 4k scanout
   (hwpipe assignment re-work to handle multiple hwpipe assigned to plane
   for wide scanout)

 + async cursor plane updates and fixes

 + refactor adreno_bind/hwinit.. still defer fw loading until device open,
   but move clk/irq/etc to probe/bind time to fix issues when fw isn't
   present in filesys

 + clk/dt bindings cleanups w/ backward compat via msm_clk_get() (dt docs
   part ack'ed by Rob Herring)

 + fw loading re-work with helper to handle either /lib/firmware/qcom/$fw
   or /lib/firmware/$fw.. background, we've started landing fw for some of
   generations in linux-firmware, but there is a preference to put fw files
   under 'qcom' subdirectory, which is not what was done on android or for
   people who copied fw from android.  So now we first look in qcom subdir
   and then fallback to the original location.

 + bunch of GPU debugging enhancements, to dump full cmdline of processes
   that trigger faults, and to add a new debugfs to capture cmdstream of
   just submits that triggered faults.. both quite useful for piglit ;-)

* tag 'drm-msm-next-2017-11-01' of git://people.freedesktop.org/~robclark/linux: (38 commits)
  drm/msm: use %z format modifier for printing size_t
  drm/msm/mdp5: Don't use async plane update path if plane visibility changes
  drm/msm/mdp5: mdp5_crtc: Restore cursor state only if LM cursors are enabled
  drm/msm/mdp5: Update mdp5_pipe_assign to spit out both planes
  drm/msm/mdp5: Prepare mdp5_pipe_assign for some rework
  drm/msm: remove mdp5_cursor_plane_funcs
  drm/msm: update cursors asynchronously through atomic
  drm/msm/atomic: switch to drm_atomic_helper_check
  drm/msm/mdp5: restore cursor state when enabling crtc
  drm/msm/mdp5: don't use autosuspend
  drm/msm/mdp5: ignore planes that are not visible
  drm/msm: dump submits which triggered gpu hang
  drm/msm: preserve IOVAs in submit's bo table
  drm/msm/rd: allow adding addition msg to top of dump
  drm/msm: split rd debugfs file
  drm/msm: add special _get_vaddr_active() for cmdstream dumps
  drm/msm: show task cmdline in gpu recovery messages
  drm/msm: dump a rd GPUADDR header for all buffers in the command
  drm/msm: Removed unused struct_mutex_task
  drm/msm: Implement preemption for A5XX targets
  ...

1 files changed, 101 insertions, 41 deletions

diff --git a/drivers/gpu/drm/msm/msm_rd.c b/drivers/gpu/drm/msm/msm_rd.c
index 0366b80..96d678b 100644
--- a/drivers/gpu/drm/msm/msm_rd.c
+++ b/drivers/gpu/drm/msm/msm_rd.c
@@ -19,11 +19,17 @@
  *
  *   tail -f /sys/kernel/debug/dri/<minor>/rd > logfile.rd
  *
- * To log the cmdstream in a format that is understood by freedreno/cffdump
+ * to log the cmdstream in a format that is understood by freedreno/cffdump
  * utility.  By comparing the last successfully completed fence #, to the
  * cmdstream for the next fence, you can narrow down which process and submit
  * caused the gpu crash/lockup.
  *
+ * Additionally:
+ *
+ *   tail -f /sys/kernel/debug/dri/<minor>/hangrd > logfile.rd
+ *
+ * will capture just the cmdstream from submits which triggered a GPU hang.
+ *
  * This bypasses drm_debugfs_create_files() mainly because we need to use
  * our own fops for a bit more control.  In particular, we don't want to
  * do anything if userspace doesn't have the debugfs file open.
@@ -212,53 +218,89 @@ static const struct file_operations rd_debugfs_fops = {
 	.release = rd_release,
 };
 
-int msm_rd_debugfs_init(struct drm_minor *minor)
+
+static void rd_cleanup(struct msm_rd_state *rd)
+{
+	if (!rd)
+		return;
+
+	mutex_destroy(&rd->read_lock);
+	kfree(rd);
+}
+
+static struct msm_rd_state *rd_init(struct drm_minor *minor, const char *name)
 {
-	struct msm_drm_private *priv = minor->dev->dev_private;
 	struct msm_rd_state *rd;
 	struct dentry *ent;
-
-	/* only create on first minor: */
-	if (priv->rd)
-		return 0;
+	int ret = 0;
 
 	rd = kzalloc(sizeof(*rd), GFP_KERNEL);
 	if (!rd)
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 
 	rd->dev = minor->dev;
 	rd->fifo.buf = rd->buf;
 
 	mutex_init(&rd->read_lock);
-	priv->rd = rd;
 
 	init_waitqueue_head(&rd->fifo_event);
 
-	ent = debugfs_create_file("rd", S_IFREG | S_IRUGO,
+	ent = debugfs_create_file(name, S_IFREG | S_IRUGO,
 			minor->debugfs_root, rd, &rd_debugfs_fops);
 	if (!ent) {
-		DRM_ERROR("Cannot create /sys/kernel/debug/dri/%pd/rd\n",
-				minor->debugfs_root);
+		DRM_ERROR("Cannot create /sys/kernel/debug/dri/%pd/%s\n",
+				minor->debugfs_root, name);
+		ret = -ENOMEM;
 		goto fail;
 	}
 
+	return rd;
+
+fail:
+	rd_cleanup(rd);
+	return ERR_PTR(ret);
+}
+
+int msm_rd_debugfs_init(struct drm_minor *minor)
+{
+	struct msm_drm_private *priv = minor->dev->dev_private;
+	struct msm_rd_state *rd;
+	int ret;
+
+	/* only create on first minor: */
+	if (priv->rd)
+		return 0;
+
+	rd = rd_init(minor, "rd");
+	if (IS_ERR(rd)) {
+		ret = PTR_ERR(rd);
+		goto fail;
+	}
+
+	priv->rd = rd;
+
+	rd = rd_init(minor, "hangrd");
+	if (IS_ERR(rd)) {
+		ret = PTR_ERR(rd);
+		goto fail;
+	}
+
+	priv->hangrd = rd;
+
 	return 0;
 
 fail:
 	msm_rd_debugfs_cleanup(priv);
-	return -1;
+	return ret;
 }
 
 void msm_rd_debugfs_cleanup(struct msm_drm_private *priv)
 {
-	struct msm_rd_state *rd = priv->rd;
-
-	if (!rd)
-		return;
-
+	rd_cleanup(priv->rd);
 	priv->rd = NULL;
-	mutex_destroy(&rd->read_lock);
-	kfree(rd);
+
+	rd_cleanup(priv->hangrd);
+	priv->hangrd = NULL;
 }
 
 static void snapshot_buf(struct msm_rd_state *rd,
@@ -268,10 +310,6 @@ static void snapshot_buf(struct msm_rd_state *rd,
 	struct msm_gem_object *obj = submit->bos[idx].obj;
 	const char *buf;
 
-	buf = msm_gem_get_vaddr(&obj->base);
-	if (IS_ERR(buf))
-		return;
-
 	if (iova) {
 		buf += iova - submit->bos[idx].iova;
 	} else {
@@ -279,20 +317,33 @@ static void snapshot_buf(struct msm_rd_state *rd,
 		size = obj->base.size;
 	}
 
+	/*
+	 * Always write the GPUADDR header so can get a complete list of all the
+	 * buffers in the cmd
+	 */
 	rd_write_section(rd, RD_GPUADDR,
 			(uint32_t[3]){ iova, size, iova >> 32 }, 12);
+
+	/* But only dump the contents of buffers marked READ */
+	if (!(submit->bos[idx].flags & MSM_SUBMIT_BO_READ))
+		return;
+
+	buf = msm_gem_get_vaddr_active(&obj->base);
+	if (IS_ERR(buf))
+		return;
+
 	rd_write_section(rd, RD_BUFFER_CONTENTS, buf, size);
 
 	msm_gem_put_vaddr(&obj->base);
 }
 
 /* called under struct_mutex */
-void msm_rd_dump_submit(struct msm_gem_submit *submit)
+void msm_rd_dump_submit(struct msm_rd_state *rd, struct msm_gem_submit *submit,
+		const char *fmt, ...)
 {
 	struct drm_device *dev = submit->dev;
-	struct msm_drm_private *priv = dev->dev_private;
-	struct msm_rd_state *rd = priv->rd;
-	char msg[128];
+	struct task_struct *task;
+	char msg[256];
 	int i, n;
 
 	if (!rd->open)
@@ -303,23 +354,32 @@ void msm_rd_dump_submit(struct msm_gem_submit *submit)
 	 */
 	WARN_ON(!mutex_is_locked(&dev->struct_mutex));
 
-	n = snprintf(msg, sizeof(msg), "%.*s/%d: fence=%u",
-			TASK_COMM_LEN, current->comm, task_pid_nr(current),
-			submit->fence->seqno);
+	if (fmt) {
+		va_list args;
 
-	rd_write_section(rd, RD_CMD, msg, ALIGN(n, 4));
+		va_start(args, fmt);
+		n = vsnprintf(msg, sizeof(msg), fmt, args);
+		va_end(args);
 
-	if (rd_full) {
-		for (i = 0; i < submit->nr_bos; i++) {
-			/* buffers that are written to probably don't start out
-			 * with anything interesting:
-			 */
-			if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
-				continue;
+		rd_write_section(rd, RD_CMD, msg, ALIGN(n, 4));
+	}
 
-			snapshot_buf(rd, submit, i, 0, 0);
-		}
+	rcu_read_lock();
+	task = pid_task(submit->pid, PIDTYPE_PID);
+	if (task) {
+		n = snprintf(msg, sizeof(msg), "%.*s/%d: fence=%u",
+				TASK_COMM_LEN, task->comm,
+				pid_nr(submit->pid), submit->seqno);
+	} else {
+		n = snprintf(msg, sizeof(msg), "???/%d: fence=%u",
+				pid_nr(submit->pid), submit->seqno);
 	}
+	rcu_read_unlock();
+
+	rd_write_section(rd, RD_CMD, msg, ALIGN(n, 4));
+
+	for (i = 0; rd_full && i < submit->nr_bos; i++)
+		snapshot_buf(rd, submit, i, 0, 0);
 
 	for (i = 0; i < submit->nr_cmds; i++) {
 		uint64_t iova = submit->cmd[i].iova;