[PATCH v2 00/10] blk-mq: move debugging information from sysfs to debugfs

[PATCH v2 00/10] blk-mq: move debugging information from sysfs to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

Changes from v1:

- Make the sbitmap seq_file helpers take a (struct sbitmap *) instead of a
  (void *), since it's not possible to use them directly as the seq_file
  show helper, anyways
- Fix a crash when reading ctx_map because it was attempting to do
  exactly that ^

Cover letter from v1:

This series ends our abuse of sysfs and puts all of the debugging information
in debugfs instead. This has a few benefits:

1. Removes the possibility of userspace being stupid and relying on something
   in sysfs that we only exposed for debugging.
2. Lifts the limitations of sysfs, including the one-value-per-file convention
   and maximum of one page of output.
3. Allows us to add more debugging information that we often want but don't
   have when debugging a live system.

Thanks!
Omar

Omar Sandoval (10):
  blk-mq: create debugfs directory tree
  blk-mq: add hctx->{state,flags} to debugfs
  blk-mq: move hctx->dispatch and ctx->rq_list from sysfs to debugfs
  blk-mq: add extra request information to debugfs
  sbitmap: add helpers for dumping to a seq_file
  blk-mq: export software queue pending map to debugfs
  blk-mq: move tags and sched_tags info from sysfs to debugfs
  blk-mq: add tags and sched_tags bitmaps to debugfs
  blk-mq: move hctx io_poll, stats, and dispatched from sysfs to debugfs
  blk-mq: move hctx and ctx counters from sysfs to debugfs

 block/Makefile          |   1 +
 block/blk-mq-debugfs.c  | 756 ++++++++++++++++++++++++++++++++++++++++++++++++
 block/blk-mq-sysfs.c    | 248 ++--------------
 block/blk-mq-tag.c      |  27 --
 block/blk-mq-tag.h      |   1 -
 block/blk-mq.c          |   2 +
 block/blk-mq.h          |  33 +++
 include/linux/blkdev.h  |   5 +
 include/linux/sbitmap.h |  28 ++
 lib/sbitmap.c           |  74 +++++
 10 files changed, 920 insertions(+), 255 deletions(-)
 create mode 100644 block/blk-mq-debugfs.c

-- 
2.11.0

[PATCH v2 01/10] blk-mq: create debugfs directory tree

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

In preparation for putting blk-mq debugging information in debugfs,
create a directory tree mirroring the one in sysfs:

    # tree -d /sys/kernel/debug/block
    /sys/kernel/debug/block
    |-- nvme0n1
    |   `-- mq
    |       |-- 0
    |       |   `-- cpu0
    |       |-- 1
    |       |   `-- cpu1
    |       |-- 2
    |       |   `-- cpu2
    |       `-- 3
    |           `-- cpu3
    `-- vda
        `-- mq
            `-- 0
                |-- cpu0
                |-- cpu1
                |-- cpu2
                `-- cpu3

Also add the scaffolding for the actual files that will go in here,
either under the hardware queue or software queue directories.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/Makefile         |   1 +
 block/blk-mq-debugfs.c | 152 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/blk-mq-sysfs.c   |   8 +++
 block/blk-mq.c         |   2 +
 block/blk-mq.h         |  33 +++++++++++
 include/linux/blkdev.h |   5 ++
 6 files changed, 201 insertions(+)
 create mode 100644 block/blk-mq-debugfs.c

diff --git a/block/Makefile b/block/Makefile
index 3ee0abd7205a..6cabe6bd2882 100644
--- a/block/Makefile
+++ b/block/Makefile
@@ -26,3 +26,4 @@ obj-$(CONFIG_BLK_DEV_INTEGRITY) += bio-integrity.o blk-integrity.o t10-pi.o
 obj-$(CONFIG_BLK_MQ_PCI)	+= blk-mq-pci.o
 obj-$(CONFIG_BLK_DEV_ZONED)	+= blk-zoned.o
 obj-$(CONFIG_BLK_WBT)		+= blk-wbt.o
+obj-$(CONFIG_DEBUG_FS)		+= blk-mq-debugfs.o
diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
new file mode 100644
index 000000000000..01711bbf5ade
--- /dev/null
+++ b/block/blk-mq-debugfs.c
@@ -0,0 +1,152 @@
+/*
+ * Copyright (C) 2017 Facebook
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <linux/kernel.h>
+#include <linux/blkdev.h>
+#include <linux/debugfs.h>
+
+#include <linux/blk-mq.h>
+#include "blk-mq.h"
+
+struct blk_mq_debugfs_attr {
+	const char *name;
+	umode_t mode;
+	const struct file_operations *fops;
+};
+
+static struct dentry *block_debugfs_root;
+
+static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
+};
+
+static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
+};
+
+int blk_mq_debugfs_register(struct request_queue *q, const char *name)
+{
+	if (!block_debugfs_root)
+		return -ENOENT;
+
+	q->debugfs_dir = debugfs_create_dir(name, block_debugfs_root);
+	if (!q->debugfs_dir)
+		goto err;
+
+	if (blk_mq_debugfs_register_hctxs(q))
+		goto err;
+
+	return 0;
+
+err:
+	blk_mq_debugfs_unregister(q);
+	return -ENOMEM;
+}
+
+void blk_mq_debugfs_unregister(struct request_queue *q)
+{
+	debugfs_remove_recursive(q->debugfs_dir);
+	q->mq_debugfs_dir = NULL;
+	q->debugfs_dir = NULL;
+}
+
+static int blk_mq_debugfs_register_ctx(struct request_queue *q,
+				       struct blk_mq_ctx *ctx,
+				       struct dentry *hctx_dir)
+{
+	struct dentry *ctx_dir;
+	char name[20];
+	int i;
+
+	snprintf(name, sizeof(name), "cpu%u", ctx->cpu);
+	ctx_dir = debugfs_create_dir(name, hctx_dir);
+	if (!ctx_dir)
+		return -ENOMEM;
+
+	for (i = 0; i < ARRAY_SIZE(blk_mq_debugfs_ctx_attrs); i++) {
+		const struct blk_mq_debugfs_attr *attr;
+
+		attr = &blk_mq_debugfs_ctx_attrs[i];
+		if (!debugfs_create_file(attr->name, attr->mode, ctx_dir, ctx,
+					 attr->fops))
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int blk_mq_debugfs_register_hctx(struct request_queue *q,
+					struct blk_mq_hw_ctx *hctx)
+{
+	struct blk_mq_ctx *ctx;
+	struct dentry *hctx_dir;
+	char name[20];
+	int i;
+
+	snprintf(name, sizeof(name), "%u", hctx->queue_num);
+	hctx_dir = debugfs_create_dir(name, q->mq_debugfs_dir);
+	if (!hctx_dir)
+		return -ENOMEM;
+
+	for (i = 0; i < ARRAY_SIZE(blk_mq_debugfs_hctx_attrs); i++) {
+		const struct blk_mq_debugfs_attr *attr;
+
+		attr = &blk_mq_debugfs_hctx_attrs[i];
+		if (!debugfs_create_file(attr->name, attr->mode, hctx_dir, hctx,
+					 attr->fops))
+			return -ENOMEM;
+	}
+
+	hctx_for_each_ctx(hctx, ctx, i) {
+		if (blk_mq_debugfs_register_ctx(q, ctx, hctx_dir))
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
+int blk_mq_debugfs_register_hctxs(struct request_queue *q)
+{
+	struct blk_mq_hw_ctx *hctx;
+	int i;
+
+	if (!q->debugfs_dir)
+		return -ENOENT;
+
+	q->mq_debugfs_dir = debugfs_create_dir("mq", q->debugfs_dir);
+	if (!q->mq_debugfs_dir)
+		goto err;
+
+	queue_for_each_hw_ctx(q, hctx, i) {
+		if (blk_mq_debugfs_register_hctx(q, hctx))
+			goto err;
+	}
+
+	return 0;
+
+err:
+	blk_mq_debugfs_unregister_hctxs(q);
+	return -ENOMEM;
+}
+
+void blk_mq_debugfs_unregister_hctxs(struct request_queue *q)
+{
+	debugfs_remove_recursive(q->mq_debugfs_dir);
+	q->mq_debugfs_dir = NULL;
+}
+
+void blk_mq_debugfs_init(void)
+{
+	block_debugfs_root = debugfs_create_dir("block", NULL);
+}
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index 2caecaa98e40..91b93ca1c1e0 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -468,6 +468,8 @@ static void __blk_mq_unregister_dev(struct device *dev, struct request_queue *q)
 		kobject_put(&hctx->kobj);
 	}
 
+	blk_mq_debugfs_unregister(q);
+
 	kobject_uevent(&q->mq_kobj, KOBJ_REMOVE);
 	kobject_del(&q->mq_kobj);
 	kobject_put(&q->mq_kobj);
@@ -517,6 +519,8 @@ int blk_mq_register_dev(struct device *dev, struct request_queue *q)
 
 	kobject_uevent(&q->mq_kobj, KOBJ_ADD);
 
+	blk_mq_debugfs_register(q, kobject_name(&dev->kobj));
+
 	queue_for_each_hw_ctx(q, hctx, i) {
 		ret = blk_mq_register_hctx(hctx);
 		if (ret)
@@ -542,6 +546,8 @@ void blk_mq_sysfs_unregister(struct request_queue *q)
 	if (!q->mq_sysfs_init_done)
 		return;
 
+	blk_mq_debugfs_unregister_hctxs(q);
+
 	queue_for_each_hw_ctx(q, hctx, i)
 		blk_mq_unregister_hctx(hctx);
 }
@@ -554,6 +560,8 @@ int blk_mq_sysfs_register(struct request_queue *q)
 	if (!q->mq_sysfs_init_done)
 		return ret;
 
+	blk_mq_debugfs_register_hctxs(q);
+
 	queue_for_each_hw_ctx(q, hctx, i) {
 		ret = blk_mq_register_hctx(hctx);
 		if (ret)
diff --git a/block/blk-mq.c b/block/blk-mq.c
index dcb567642db7..4c3e667b2ab9 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -2799,6 +2799,8 @@ void blk_mq_enable_hotplug(void)
 
 static int __init blk_mq_init(void)
 {
+	blk_mq_debugfs_init();
+
 	cpuhp_setup_state_multi(CPUHP_BLK_MQ_DEAD, "block/mq:dead", NULL,
 				blk_mq_hctx_notify_dead);
 
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 0c7c034d9ddd..6c24b901acd7 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -78,6 +78,39 @@ extern int blk_mq_sysfs_register(struct request_queue *q);
 extern void blk_mq_sysfs_unregister(struct request_queue *q);
 extern void blk_mq_hctx_kobj_init(struct blk_mq_hw_ctx *hctx);
 
+/*
+ * debugfs helpers
+ */
+#ifdef CONFIG_DEBUG_FS
+void blk_mq_debugfs_init(void);
+int blk_mq_debugfs_register(struct request_queue *q, const char *name);
+void blk_mq_debugfs_unregister(struct request_queue *q);
+int blk_mq_debugfs_register_hctxs(struct request_queue *q);
+void blk_mq_debugfs_unregister_hctxs(struct request_queue *q);
+#else
+static inline void blk_mq_debugfs_init(void)
+{
+}
+
+int blk_mq_debugfs_register(struct request_queue *q, const char *name);
+{
+	return 0;
+}
+
+void blk_mq_debugfs_unregister(struct request_queue *q)
+{
+}
+
+int blk_mq_debugfs_register_hctxs(struct request_queue *q)
+{
+	return 0;
+}
+
+void blk_mq_debugfs_unregister_hctxs(struct request_queue *q)
+{
+}
+#endif
+
 extern void blk_mq_rq_timed_out(struct request *req, bool reserved);
 
 void blk_mq_release(struct request_queue *q);
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 25564857f5f8..0ee283f3cffe 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -571,6 +571,11 @@ struct request_queue {
 	struct list_head	tag_set_list;
 	struct bio_set		*bio_split;
 
+#ifdef CONFIG_DEBUG_FS
+	struct dentry		*debugfs_dir;
+	struct dentry		*mq_debugfs_dir;
+#endif
+
 	bool			mq_sysfs_init_done;
 };
 
-- 
2.11.0

[PATCH v2 02/10] blk-mq: add hctx->{state,flags} to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

hctx->state could come in handy for bugs where the hardware queue gets
stuck in the stopped state, and hctx->flags is just useful to know.

Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 42 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 42 insertions(+)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 01711bbf5ade..0c14511fa9e0 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -29,7 +29,49 @@ struct blk_mq_debugfs_attr {
 
 static struct dentry *block_debugfs_root;
 
+static int hctx_state_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	seq_printf(m, "0x%lx\n", hctx->state);
+	return 0;
+}
+
+static int hctx_state_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_state_show, inode->i_private);
+}
+
+static const struct file_operations hctx_state_fops = {
+	.open		= hctx_state_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int hctx_flags_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	seq_printf(m, "0x%lx\n", hctx->flags);
+	return 0;
+}
+
+static int hctx_flags_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_flags_show, inode->i_private);
+}
+
+static const struct file_operations hctx_flags_fops = {
+	.open		= hctx_flags_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
+	{"state", 0400, &hctx_state_fops},
+	{"flags", 0400, &hctx_flags_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
-- 
2.11.0

[PATCH v2 03/10] blk-mq: move hctx->dispatch and ctx->rq_list from sysfs to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

These lists are only useful for debugging; they definitely don't belong
in sysfs. Putting them in debugfs also removes the limitation of a
single page of output.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 106 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/blk-mq-sysfs.c   |  57 --------------------------
 2 files changed, 106 insertions(+), 57 deletions(-)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 0c14511fa9e0..1967c06d80e0 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -29,6 +29,20 @@ struct blk_mq_debugfs_attr {
 
 static struct dentry *block_debugfs_root;
 
+static int blk_mq_debugfs_seq_open(struct inode *inode, struct file *file,
+				   const struct seq_operations *ops)
+{
+	struct seq_file *m;
+	int ret;
+
+	ret = seq_open(file, ops);
+	if (!ret) {
+		m = file->private_data;
+		m->private = inode->i_private;
+	}
+	return ret;
+}
+
 static int hctx_state_show(struct seq_file *m, void *v)
 {
 	struct blk_mq_hw_ctx *hctx = m->private;
@@ -69,12 +83,104 @@ static const struct file_operations hctx_flags_fops = {
 	.release	= single_release,
 };
 
+static int blk_mq_debugfs_rq_show(struct seq_file *m, void *v)
+{
+	struct request *rq = list_entry_rq(v);
+
+	seq_printf(m, "%p\n", rq);
+	return 0;
+}
+
+static void *hctx_dispatch_start(struct seq_file *m, loff_t *pos)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	spin_lock(&hctx->lock);
+	return seq_list_start(&hctx->dispatch, *pos);
+}
+
+static void *hctx_dispatch_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	return seq_list_next(v, &hctx->dispatch, pos);
+}
+
+static void hctx_dispatch_stop(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	spin_unlock(&hctx->lock);
+}
+
+static const struct seq_operations hctx_dispatch_seq_ops = {
+	.start	= hctx_dispatch_start,
+	.next	= hctx_dispatch_next,
+	.stop	= hctx_dispatch_stop,
+	.show	= blk_mq_debugfs_rq_show,
+};
+
+static int hctx_dispatch_open(struct inode *inode, struct file *file)
+{
+	return blk_mq_debugfs_seq_open(inode, file, &hctx_dispatch_seq_ops);
+}
+
+static const struct file_operations hctx_dispatch_fops = {
+	.open		= hctx_dispatch_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
+static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
+{
+	struct blk_mq_ctx *ctx = m->private;
+
+	spin_lock(&ctx->lock);
+	return seq_list_start(&ctx->rq_list, *pos);
+}
+
+static void *ctx_rq_list_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	struct blk_mq_ctx *ctx = m->private;
+
+	return seq_list_next(v, &ctx->rq_list, pos);
+}
+
+static void ctx_rq_list_stop(struct seq_file *m, void *v)
+{
+	struct blk_mq_ctx *ctx = m->private;
+
+	spin_unlock(&ctx->lock);
+}
+
+static const struct seq_operations ctx_rq_list_seq_ops = {
+	.start	= ctx_rq_list_start,
+	.next	= ctx_rq_list_next,
+	.stop	= ctx_rq_list_stop,
+	.show	= blk_mq_debugfs_rq_show,
+};
+
+static int ctx_rq_list_open(struct inode *inode, struct file *file)
+{
+	return blk_mq_debugfs_seq_open(inode, file, &ctx_rq_list_seq_ops);
+}
+
+static const struct file_operations ctx_rq_list_fops = {
+	.open		= ctx_rq_list_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= seq_release,
+};
+
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"state", 0400, &hctx_state_fops},
 	{"flags", 0400, &hctx_flags_fops},
+	{"dispatch", 0400, &hctx_dispatch_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
+	{"rq_list", 0400, &ctx_rq_list_fops},
 };
 
 int blk_mq_debugfs_register(struct request_queue *q, const char *name)
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index 91b93ca1c1e0..ee3694d8c4ee 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -139,41 +139,6 @@ static ssize_t blk_mq_sysfs_completed_show(struct blk_mq_ctx *ctx, char *page)
 				ctx->rq_completed[0]);
 }
 
-static ssize_t sysfs_list_show(char *page, struct list_head *list, char *msg)
-{
-	struct request *rq;
-	int len = snprintf(page, PAGE_SIZE - 1, "%s:\n", msg);
-
-	list_for_each_entry(rq, list, queuelist) {
-		const int rq_len = 2 * sizeof(rq) + 2;
-
-		/* if the output will be truncated */
-		if (PAGE_SIZE - 1 < len + rq_len) {
-			/* backspacing if it can't hold '\t...\n' */
-			if (PAGE_SIZE - 1 < len + 5)
-				len -= rq_len;
-			len += snprintf(page + len, PAGE_SIZE - 1 - len,
-					"\t...\n");
-			break;
-		}
-		len += snprintf(page + len, PAGE_SIZE - 1 - len,
-				"\t%p\n", rq);
-	}
-
-	return len;
-}
-
-static ssize_t blk_mq_sysfs_rq_list_show(struct blk_mq_ctx *ctx, char *page)
-{
-	ssize_t ret;
-
-	spin_lock(&ctx->lock);
-	ret = sysfs_list_show(page, &ctx->rq_list, "CTX pending");
-	spin_unlock(&ctx->lock);
-
-	return ret;
-}
-
 static ssize_t blk_mq_hw_sysfs_poll_show(struct blk_mq_hw_ctx *hctx, char *page)
 {
 	return sprintf(page, "considered=%lu, invoked=%lu, success=%lu\n",
@@ -219,18 +184,6 @@ static ssize_t blk_mq_hw_sysfs_dispatched_show(struct blk_mq_hw_ctx *hctx,
 	return page - start_page;
 }
 
-static ssize_t blk_mq_hw_sysfs_rq_list_show(struct blk_mq_hw_ctx *hctx,
-					    char *page)
-{
-	ssize_t ret;
-
-	spin_lock(&hctx->lock);
-	ret = sysfs_list_show(page, &hctx->dispatch, "HCTX pending");
-	spin_unlock(&hctx->lock);
-
-	return ret;
-}
-
 static ssize_t blk_mq_hw_sysfs_sched_tags_show(struct blk_mq_hw_ctx *hctx, char *page)
 {
 	if (hctx->sched_tags)
@@ -320,16 +273,11 @@ static struct blk_mq_ctx_sysfs_entry blk_mq_sysfs_completed = {
 	.attr = {.name = "completed", .mode = S_IRUGO },
 	.show = blk_mq_sysfs_completed_show,
 };
-static struct blk_mq_ctx_sysfs_entry blk_mq_sysfs_rq_list = {
-	.attr = {.name = "rq_list", .mode = S_IRUGO },
-	.show = blk_mq_sysfs_rq_list_show,
-};
 
 static struct attribute *default_ctx_attrs[] = {
 	&blk_mq_sysfs_dispatched.attr,
 	&blk_mq_sysfs_merged.attr,
 	&blk_mq_sysfs_completed.attr,
-	&blk_mq_sysfs_rq_list.attr,
 	NULL,
 };
 
@@ -349,10 +297,6 @@ static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_active = {
 	.attr = {.name = "active", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_active_show,
 };
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_pending = {
-	.attr = {.name = "pending", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_rq_list_show,
-};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_sched_tags = {
 	.attr = {.name = "sched_tags", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_sched_tags_show,
@@ -380,7 +324,6 @@ static struct attribute *default_hw_ctx_attrs[] = {
 	&blk_mq_hw_sysfs_queued.attr,
 	&blk_mq_hw_sysfs_run.attr,
 	&blk_mq_hw_sysfs_dispatched.attr,
-	&blk_mq_hw_sysfs_pending.attr,
 	&blk_mq_hw_sysfs_tags.attr,
 	&blk_mq_hw_sysfs_sched_tags.attr,
 	&blk_mq_hw_sysfs_cpus.attr,
-- 
2.11.0

[PATCH v2 04/10] blk-mq: add extra request information to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

The request pointers by themselves aren't super useful.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 1967c06d80e0..ee947f2f605b 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -87,7 +87,9 @@ static int blk_mq_debugfs_rq_show(struct seq_file *m, void *v)
 {
 	struct request *rq = list_entry_rq(v);
 
-	seq_printf(m, "%p\n", rq);
+	seq_printf(m, "%p {.cmd_type=%u, .cmd_flags=0x%x, .rq_flags=0x%x, .tag=%d, .internal_tag=%d}\n",
+		   rq, rq->cmd_type, rq->cmd_flags, (unsigned int)rq->rq_flags,
+		   rq->tag, rq->internal_tag);
 	return 0;
 }
 
-- 
2.11.0

[PATCH v2 05/10] sbitmap: add helpers for dumping to a seq_file

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

This is useful debugging information that will be used in the blk-mq
debugfs directory.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 include/linux/sbitmap.h | 28 +++++++++++++++++++
 lib/sbitmap.c           | 74 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 102 insertions(+)

diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h
index f017fd6e69c4..4a7998355f2b 100644
--- a/include/linux/sbitmap.h
+++ b/include/linux/sbitmap.h
@@ -259,6 +259,25 @@ static inline int sbitmap_test_bit(struct sbitmap *sb, unsigned int bitnr)
 unsigned int sbitmap_weight(const struct sbitmap *sb);
 
 /**
+ * sbitmap_show() - Dump bitmap information to a struct seq_file.
+ * @sb: Bitmap to show.
+ * @m: struct seq_file to write to.
+ *
+ * This is intended for debugging. The format may change at any time.
+ */
+void sbitmap_show(struct sbitmap *sb, struct seq_file *m);
+
+/**
+ * sbitmap_bitmap_show() - Dump the raw bitmap to a struct seq_file.
+ * @sb: Bitmap to show.
+ * @m: struct seq_file to write to.
+ *
+ * This is intended for debugging. The output isn't guaranteed to be internally
+ * consistent.
+ */
+void sbitmap_bitmap_show(struct sbitmap *sb, struct seq_file *m);
+
+/**
  * sbitmap_queue_init_node() - Initialize a &struct sbitmap_queue on a specific
  * memory node.
  * @sbq: Bitmap queue to initialize.
@@ -370,4 +389,13 @@ static inline struct sbq_wait_state *sbq_wait_ptr(struct sbitmap_queue *sbq,
  */
 void sbitmap_queue_wake_all(struct sbitmap_queue *sbq);
 
+/**
+ * sbitmap_queue_show() - Dump bitmap queue information to a struct seq_file.
+ * @sbq: Bitmap queue to show.
+ * @m: struct seq_file to write to.
+ *
+ * This is intended for debugging. The format may change at any time.
+ */
+void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m);
+
 #endif /* __LINUX_SCALE_BITMAP_H */
diff --git a/lib/sbitmap.c b/lib/sbitmap.c
index 8f5c3b268c77..48cb6b5a8e7e 100644
--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -17,6 +17,7 @@
 
 #include <linux/random.h>
 #include <linux/sbitmap.h>
+#include <linux/seq_file.h>
 
 int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
 		      gfp_t flags, int node)
@@ -180,6 +181,45 @@ unsigned int sbitmap_weight(const struct sbitmap *sb)
 }
 EXPORT_SYMBOL_GPL(sbitmap_weight);
 
+void sbitmap_show(struct sbitmap *sb, struct seq_file *m)
+{
+	seq_printf(m, "depth=%u\n", sb->depth);
+	seq_printf(m, "weight=%u\n", sbitmap_weight(sb));
+	seq_printf(m, "bits_per_word=%u\n", 1U << sb->shift);
+	seq_printf(m, "map_nr=%u\n", sb->map_nr);
+}
+EXPORT_SYMBOL_GPL(sbitmap_show);
+
+void sbitmap_bitmap_show(struct sbitmap *sb, struct seq_file *m)
+{
+	u8 byte = 0;
+	unsigned int byte_bits = 0;
+	int i;
+
+	for (i = 0; i < sb->map_nr; i++) {
+		unsigned long word = READ_ONCE(sb->map[i].word);
+		unsigned int word_bits = READ_ONCE(sb->map[i].depth);
+
+		while (word_bits > 0) {
+			unsigned int bits = min(8 - byte_bits, word_bits);
+
+			byte |= (word & (BIT(bits) - 1)) << byte_bits;
+			byte_bits += bits;
+			if (byte_bits == 8) {
+				seq_write(m, &byte, sizeof(byte));
+				byte = 0;
+				byte_bits = 0;
+			}
+			word >>= bits;
+			word_bits -= bits;
+		}
+	}
+
+	if (byte_bits)
+		seq_write(m, &byte, sizeof(byte));
+}
+EXPORT_SYMBOL_GPL(sbitmap_bitmap_show);
+
 static unsigned int sbq_calc_wake_batch(unsigned int depth)
 {
 	unsigned int wake_batch;
@@ -377,3 +417,37 @@ void sbitmap_queue_wake_all(struct sbitmap_queue *sbq)
 	}
 }
 EXPORT_SYMBOL_GPL(sbitmap_queue_wake_all);
+
+void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m)
+{
+	bool first;
+	int i;
+
+	sbitmap_show(&sbq->sb, m);
+
+	seq_puts(m, "alloc_hint={");
+	first = true;
+	for_each_possible_cpu(i) {
+		if (!first)
+			seq_puts(m, ", ");
+		first = false;
+		seq_printf(m, "%u", *per_cpu_ptr(sbq->alloc_hint, i));
+	}
+	seq_puts(m, "}\n");
+
+	seq_printf(m, "wake_batch=%u\n", sbq->wake_batch);
+	seq_printf(m, "wake_index=%d\n", atomic_read(&sbq->wake_index));
+
+	seq_puts(m, "ws={\n");
+	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
+		struct sbq_wait_state *ws = &sbq->ws[i];
+
+		seq_printf(m, "\t{.wait_cnt=%d, .wait=%s},\n",
+			   atomic_read(&ws->wait_cnt),
+			   waitqueue_active(&ws->wait) ? "active" : "inactive");
+	}
+	seq_puts(m, "}\n");
+
+	seq_printf(m, "round_robin=%d\n", sbq->round_robin);
+}
+EXPORT_SYMBOL_GPL(sbitmap_queue_show);
-- 
2.11.0

[PATCH v2 06/10] blk-mq: export software queue pending map to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

This is useful for debugging problems where we've gotten stuck with
requests in the software queues.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index ee947f2f605b..857bdadc6b1f 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -134,6 +134,26 @@ static const struct file_operations hctx_dispatch_fops = {
 	.release	= seq_release,
 };
 
+static int hctx_ctx_map_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	sbitmap_bitmap_show(&hctx->ctx_map, m);
+	return 0;
+}
+
+static int hctx_ctx_map_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_ctx_map_show, inode->i_private);
+}
+
+static const struct file_operations hctx_ctx_map_fops = {
+	.open		= hctx_ctx_map_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
 {
 	struct blk_mq_ctx *ctx = m->private;
@@ -179,6 +199,7 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"state", 0400, &hctx_state_fops},
 	{"flags", 0400, &hctx_flags_fops},
 	{"dispatch", 0400, &hctx_dispatch_fops},
+	{"ctx_map", 0400, &hctx_ctx_map_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
-- 
2.11.0

[PATCH v2 07/10] blk-mq: move tags and sched_tags info from sysfs to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

These are very tied to the blk-mq tag implementation, so exposing them
to sysfs isn't a great idea. Move the debugging information to debugfs
and add basic entries for the number of tags and the number of reserved
tags to sysfs.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 70 ++++++++++++++++++++++++++++++++++++++++++++++++++
 block/blk-mq-sysfs.c   | 33 ++++++++++++------------
 block/blk-mq-tag.c     | 27 -------------------
 block/blk-mq-tag.h     |  1 -
 4 files changed, 86 insertions(+), 45 deletions(-)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 857bdadc6b1f..9f811007cf4f 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -20,6 +20,7 @@
 
 #include <linux/blk-mq.h>
 #include "blk-mq.h"
+#include "blk-mq-tag.h"
 
 struct blk_mq_debugfs_attr {
 	const char *name;
@@ -154,6 +155,73 @@ static const struct file_operations hctx_ctx_map_fops = {
 	.release	= single_release,
 };
 
+static void blk_mq_debugfs_tags_show(struct seq_file *m,
+				     struct blk_mq_tags *tags)
+{
+	seq_printf(m, "nr_tags=%u\n", tags->nr_tags);
+	seq_printf(m, "nr_reserved_tags=%u\n", tags->nr_reserved_tags);
+	seq_printf(m, "active_queues=%d\n",
+		   atomic_read(&tags->active_queues));
+
+	seq_puts(m, "\nbitmap_tags:\n");
+	sbitmap_queue_show(&tags->bitmap_tags, m);
+
+	if (tags->nr_reserved_tags) {
+		seq_puts(m, "\nbreserved_tags:\n");
+		sbitmap_queue_show(&tags->breserved_tags, m);
+	}
+}
+
+static int hctx_tags_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+	struct request_queue *q = hctx->queue;
+
+	mutex_lock(&q->sysfs_lock);
+	if (hctx->tags)
+		blk_mq_debugfs_tags_show(m, hctx->tags);
+	mutex_unlock(&q->sysfs_lock);
+
+	return 0;
+}
+
+static int hctx_tags_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_tags_show, inode->i_private);
+}
+
+static const struct file_operations hctx_tags_fops = {
+	.open		= hctx_tags_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int hctx_sched_tags_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+	struct request_queue *q = hctx->queue;
+
+	mutex_lock(&q->sysfs_lock);
+	if (hctx->sched_tags)
+		blk_mq_debugfs_tags_show(m, hctx->sched_tags);
+	mutex_unlock(&q->sysfs_lock);
+
+	return 0;
+}
+
+static int hctx_sched_tags_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_sched_tags_show, inode->i_private);
+}
+
+static const struct file_operations hctx_sched_tags_fops = {
+	.open		= hctx_sched_tags_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
 {
 	struct blk_mq_ctx *ctx = m->private;
@@ -200,6 +268,8 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"flags", 0400, &hctx_flags_fops},
 	{"dispatch", 0400, &hctx_dispatch_fops},
 	{"ctx_map", 0400, &hctx_ctx_map_fops},
+	{"tags", 0400, &hctx_tags_fops},
+	{"sched_tags", 0400, &hctx_sched_tags_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index ee3694d8c4ee..a0ae1f536ed0 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -184,17 +184,16 @@ static ssize_t blk_mq_hw_sysfs_dispatched_show(struct blk_mq_hw_ctx *hctx,
 	return page - start_page;
 }
 
-static ssize_t blk_mq_hw_sysfs_sched_tags_show(struct blk_mq_hw_ctx *hctx, char *page)
+static ssize_t blk_mq_hw_sysfs_nr_tags_show(struct blk_mq_hw_ctx *hctx,
+					    char *page)
 {
-	if (hctx->sched_tags)
-		return blk_mq_tag_sysfs_show(hctx->sched_tags, page);
-
-	return 0;
+	return sprintf(page, "%u\n", hctx->tags->nr_tags);
 }
 
-static ssize_t blk_mq_hw_sysfs_tags_show(struct blk_mq_hw_ctx *hctx, char *page)
+static ssize_t blk_mq_hw_sysfs_nr_reserved_tags_show(struct blk_mq_hw_ctx *hctx,
+						     char *page)
 {
-	return blk_mq_tag_sysfs_show(hctx->tags, page);
+	return sprintf(page, "%u\n", hctx->tags->nr_reserved_tags);
 }
 
 static ssize_t blk_mq_hw_sysfs_active_show(struct blk_mq_hw_ctx *hctx, char *page)
@@ -293,18 +292,18 @@ static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_dispatched = {
 	.attr = {.name = "dispatched", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_dispatched_show,
 };
+static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_nr_tags = {
+	.attr = {.name = "nr_tags", .mode = S_IRUGO },
+	.show = blk_mq_hw_sysfs_nr_tags_show,
+};
+static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_nr_reserved_tags = {
+	.attr = {.name = "nr_reserved_tags", .mode = S_IRUGO },
+	.show = blk_mq_hw_sysfs_nr_reserved_tags_show,
+};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_active = {
 	.attr = {.name = "active", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_active_show,
 };
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_sched_tags = {
-	.attr = {.name = "sched_tags", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_sched_tags_show,
-};
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_tags = {
-	.attr = {.name = "tags", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_tags_show,
-};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_cpus = {
 	.attr = {.name = "cpu_list", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_cpus_show,
@@ -324,8 +323,8 @@ static struct attribute *default_hw_ctx_attrs[] = {
 	&blk_mq_hw_sysfs_queued.attr,
 	&blk_mq_hw_sysfs_run.attr,
 	&blk_mq_hw_sysfs_dispatched.attr,
-	&blk_mq_hw_sysfs_tags.attr,
-	&blk_mq_hw_sysfs_sched_tags.attr,
+	&blk_mq_hw_sysfs_nr_tags.attr,
+	&blk_mq_hw_sysfs_nr_reserved_tags.attr,
 	&blk_mq_hw_sysfs_cpus.attr,
 	&blk_mq_hw_sysfs_active.attr,
 	&blk_mq_hw_sysfs_poll.attr,
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index 1b156ca79af6..f8de2dbbb29f 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -329,11 +329,6 @@ void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
 
 }
 
-static unsigned int bt_unused_tags(const struct sbitmap_queue *bt)
-{
-	return bt->sb.depth - sbitmap_weight(&bt->sb);
-}
-
 static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
 		    bool round_robin, int node)
 {
@@ -469,25 +464,3 @@ u32 blk_mq_unique_tag(struct request *rq)
 		(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
 }
 EXPORT_SYMBOL(blk_mq_unique_tag);
-
-ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags *tags, char *page)
-{
-	char *orig_page = page;
-	unsigned int free, res;
-
-	if (!tags)
-		return 0;
-
-	page += sprintf(page, "nr_tags=%u, reserved_tags=%u, "
-			"bits_per_word=%u\n",
-			tags->nr_tags, tags->nr_reserved_tags,
-			1U << tags->bitmap_tags.sb.shift);
-
-	free = bt_unused_tags(&tags->bitmap_tags);
-	res = bt_unused_tags(&tags->breserved_tags);
-
-	page += sprintf(page, "nr_free=%u, nr_reserved=%u\n", free, res);
-	page += sprintf(page, "active_queues=%u\n", atomic_read(&tags->active_queues));
-
-	return page - orig_page;
-}
diff --git a/block/blk-mq-tag.h b/block/blk-mq-tag.h
index ac22878462e7..63497423c5cd 100644
--- a/block/blk-mq-tag.h
+++ b/block/blk-mq-tag.h
@@ -28,7 +28,6 @@ extern unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data);
 extern void blk_mq_put_tag(struct blk_mq_hw_ctx *hctx, struct blk_mq_tags *tags,
 			   struct blk_mq_ctx *ctx, unsigned int tag);
 extern bool blk_mq_has_free_tags(struct blk_mq_tags *tags);
-extern ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags *tags, char *page);
 extern int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
 					struct blk_mq_tags **tags,
 					unsigned int depth, bool can_grow);
-- 
2.11.0

[PATCH v2 08/10] blk-mq: add tags and sched_tags bitmaps to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

These can be used to debug issues like tag leaks and stuck requests.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 50 insertions(+)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 9f811007cf4f..9b87b15619f1 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -197,6 +197,30 @@ static const struct file_operations hctx_tags_fops = {
 	.release	= single_release,
 };
 
+static int hctx_tags_bitmap_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+	struct request_queue *q = hctx->queue;
+
+	mutex_lock(&q->sysfs_lock);
+	if (hctx->tags)
+		sbitmap_bitmap_show(&hctx->tags->bitmap_tags.sb, m);
+	mutex_unlock(&q->sysfs_lock);
+	return 0;
+}
+
+static int hctx_tags_bitmap_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_tags_bitmap_show, inode->i_private);
+}
+
+static const struct file_operations hctx_tags_bitmap_fops = {
+	.open		= hctx_tags_bitmap_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static int hctx_sched_tags_show(struct seq_file *m, void *v)
 {
 	struct blk_mq_hw_ctx *hctx = m->private;
@@ -222,6 +246,30 @@ static const struct file_operations hctx_sched_tags_fops = {
 	.release	= single_release,
 };
 
+static int hctx_sched_tags_bitmap_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+	struct request_queue *q = hctx->queue;
+
+	mutex_lock(&q->sysfs_lock);
+	if (hctx->sched_tags)
+		sbitmap_bitmap_show(&hctx->sched_tags->bitmap_tags.sb, m);
+	mutex_unlock(&q->sysfs_lock);
+	return 0;
+}
+
+static int hctx_sched_tags_bitmap_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_sched_tags_bitmap_show, inode->i_private);
+}
+
+static const struct file_operations hctx_sched_tags_bitmap_fops = {
+	.open		= hctx_sched_tags_bitmap_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
 {
 	struct blk_mq_ctx *ctx = m->private;
@@ -269,7 +317,9 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"dispatch", 0400, &hctx_dispatch_fops},
 	{"ctx_map", 0400, &hctx_ctx_map_fops},
 	{"tags", 0400, &hctx_tags_fops},
+	{"tags_bitmap", 0400, &hctx_tags_bitmap_fops},
 	{"sched_tags", 0400, &hctx_sched_tags_fops},
+	{"sched_tags_bitmap", 0400, &hctx_sched_tags_bitmap_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
-- 
2.11.0

[PATCH v2 09/10] blk-mq: move hctx io_poll, stats, and dispatched from sysfs to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

These statistics _might_ be useful to userspace, but it's better not to
commit to an ABI for these yet. Also, the dispatched file in sysfs
couldn't be cleared, so make it clearable like the others in debugfs.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 132 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/blk-mq-sysfs.c   |  92 ----------------------------------
 2 files changed, 132 insertions(+), 92 deletions(-)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 9b87b15619f1..28ca56dd2c4e 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -270,6 +270,135 @@ static const struct file_operations hctx_sched_tags_bitmap_fops = {
 	.release	= single_release,
 };
 
+static int hctx_io_poll_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	seq_printf(m, "considered=%lu\n", hctx->poll_considered);
+	seq_printf(m, "invoked=%lu\n", hctx->poll_invoked);
+	seq_printf(m, "success=%lu\n", hctx->poll_success);
+	return 0;
+}
+
+static int hctx_io_poll_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_io_poll_show, inode->i_private);
+}
+
+static ssize_t hctx_io_poll_write(struct file *file, const char __user *buf,
+				  size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	hctx->poll_considered = hctx->poll_invoked = hctx->poll_success = 0;
+	return count;
+}
+
+static const struct file_operations hctx_io_poll_fops = {
+	.open		= hctx_io_poll_open,
+	.read		= seq_read,
+	.write		= hctx_io_poll_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static void print_stat(struct seq_file *m, struct blk_rq_stat *stat)
+{
+	seq_printf(m, "samples=%d, mean=%lld, min=%llu, max=%llu",
+		   stat->nr_samples, stat->mean, stat->min, stat->max);
+}
+
+static int hctx_stats_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+	struct blk_rq_stat stat[2];
+
+	blk_stat_init(&stat[BLK_STAT_READ]);
+	blk_stat_init(&stat[BLK_STAT_WRITE]);
+
+	blk_hctx_stat_get(hctx, stat);
+
+	seq_puts(m, "read: ");
+	print_stat(m, &stat[BLK_STAT_READ]);
+	seq_puts(m, "\n");
+
+	seq_puts(m, "write: ");
+	print_stat(m, &stat[BLK_STAT_WRITE]);
+	seq_puts(m, "\n");
+	return 0;
+}
+
+static int hctx_stats_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_stats_show, inode->i_private);
+}
+
+static ssize_t hctx_stats_write(struct file *file, const char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_hw_ctx *hctx = m->private;
+	struct blk_mq_ctx *ctx;
+	int i;
+
+	hctx_for_each_ctx(hctx, ctx, i) {
+		blk_stat_init(&ctx->stat[BLK_STAT_READ]);
+		blk_stat_init(&ctx->stat[BLK_STAT_WRITE]);
+	}
+	return count;
+}
+
+static const struct file_operations hctx_stats_fops = {
+	.open		= hctx_stats_open,
+	.read		= seq_read,
+	.write		= hctx_stats_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int hctx_dispatched_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+	int i;
+
+	seq_printf(m, "%8u\t%lu\n", 0U, hctx->dispatched[0]);
+
+	for (i = 1; i < BLK_MQ_MAX_DISPATCH_ORDER - 1; i++) {
+		unsigned int d = 1U << (i - 1);
+
+		seq_printf(m, "%8u\t%lu\n", d, hctx->dispatched[i]);
+	}
+
+	seq_printf(m, "%8u+\t%lu\n", 1U << (i - 1), hctx->dispatched[i]);
+	return 0;
+}
+
+static int hctx_dispatched_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_dispatched_show, inode->i_private);
+}
+
+static ssize_t hctx_dispatched_write(struct file *file, const char __user *buf,
+				     size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_hw_ctx *hctx = m->private;
+	int i;
+
+	for (i = 0; i < BLK_MQ_MAX_DISPATCH_ORDER; i++)
+		hctx->dispatched[i] = 0;
+	return count;
+}
+
+static const struct file_operations hctx_dispatched_fops = {
+	.open		= hctx_dispatched_open,
+	.read		= seq_read,
+	.write		= hctx_dispatched_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
 {
 	struct blk_mq_ctx *ctx = m->private;
@@ -320,6 +449,9 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"tags_bitmap", 0400, &hctx_tags_bitmap_fops},
 	{"sched_tags", 0400, &hctx_sched_tags_fops},
 	{"sched_tags_bitmap", 0400, &hctx_sched_tags_bitmap_fops},
+	{"io_poll", 0600, &hctx_io_poll_fops},
+	{"stats", 0600, &hctx_stats_fops},
+	{"dispatched", 0600, &hctx_dispatched_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index a0ae1f536ed0..f505da2c542e 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -139,21 +139,6 @@ static ssize_t blk_mq_sysfs_completed_show(struct blk_mq_ctx *ctx, char *page)
 				ctx->rq_completed[0]);
 }
 
-static ssize_t blk_mq_hw_sysfs_poll_show(struct blk_mq_hw_ctx *hctx, char *page)
-{
-	return sprintf(page, "considered=%lu, invoked=%lu, success=%lu\n",
-		       hctx->poll_considered, hctx->poll_invoked,
-		       hctx->poll_success);
-}
-
-static ssize_t blk_mq_hw_sysfs_poll_store(struct blk_mq_hw_ctx *hctx,
-					  const char *page, size_t size)
-{
-	hctx->poll_considered = hctx->poll_invoked = hctx->poll_success = 0;
-
-	return size;
-}
-
 static ssize_t blk_mq_hw_sysfs_queued_show(struct blk_mq_hw_ctx *hctx,
 					   char *page)
 {
@@ -165,25 +150,6 @@ static ssize_t blk_mq_hw_sysfs_run_show(struct blk_mq_hw_ctx *hctx, char *page)
 	return sprintf(page, "%lu\n", hctx->run);
 }
 
-static ssize_t blk_mq_hw_sysfs_dispatched_show(struct blk_mq_hw_ctx *hctx,
-					       char *page)
-{
-	char *start_page = page;
-	int i;
-
-	page += sprintf(page, "%8u\t%lu\n", 0U, hctx->dispatched[0]);
-
-	for (i = 1; i < BLK_MQ_MAX_DISPATCH_ORDER - 1; i++) {
-		unsigned int d = 1U << (i - 1);
-
-		page += sprintf(page, "%8u\t%lu\n", d, hctx->dispatched[i]);
-	}
-
-	page += sprintf(page, "%8u+\t%lu\n", 1U << (i - 1),
-						hctx->dispatched[i]);
-	return page - start_page;
-}
-
 static ssize_t blk_mq_hw_sysfs_nr_tags_show(struct blk_mq_hw_ctx *hctx,
 					    char *page)
 {
@@ -219,47 +185,6 @@ static ssize_t blk_mq_hw_sysfs_cpus_show(struct blk_mq_hw_ctx *hctx, char *page)
 	return ret;
 }
 
-static void blk_mq_stat_clear(struct blk_mq_hw_ctx *hctx)
-{
-	struct blk_mq_ctx *ctx;
-	unsigned int i;
-
-	hctx_for_each_ctx(hctx, ctx, i) {
-		blk_stat_init(&ctx->stat[BLK_STAT_READ]);
-		blk_stat_init(&ctx->stat[BLK_STAT_WRITE]);
-	}
-}
-
-static ssize_t blk_mq_hw_sysfs_stat_store(struct blk_mq_hw_ctx *hctx,
-					  const char *page, size_t count)
-{
-	blk_mq_stat_clear(hctx);
-	return count;
-}
-
-static ssize_t print_stat(char *page, struct blk_rq_stat *stat, const char *pre)
-{
-	return sprintf(page, "%s samples=%llu, mean=%lld, min=%lld, max=%lld\n",
-			pre, (long long) stat->nr_samples,
-			(long long) stat->mean, (long long) stat->min,
-			(long long) stat->max);
-}
-
-static ssize_t blk_mq_hw_sysfs_stat_show(struct blk_mq_hw_ctx *hctx, char *page)
-{
-	struct blk_rq_stat stat[2];
-	ssize_t ret;
-
-	blk_stat_init(&stat[BLK_STAT_READ]);
-	blk_stat_init(&stat[BLK_STAT_WRITE]);
-
-	blk_hctx_stat_get(hctx, stat);
-
-	ret = print_stat(page, &stat[BLK_STAT_READ], "read :");
-	ret += print_stat(page + ret, &stat[BLK_STAT_WRITE], "write:");
-	return ret;
-}
-
 static struct blk_mq_ctx_sysfs_entry blk_mq_sysfs_dispatched = {
 	.attr = {.name = "dispatched", .mode = S_IRUGO },
 	.show = blk_mq_sysfs_dispatched_show,
@@ -288,10 +213,6 @@ static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_run = {
 	.attr = {.name = "run", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_run_show,
 };
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_dispatched = {
-	.attr = {.name = "dispatched", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_dispatched_show,
-};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_nr_tags = {
 	.attr = {.name = "nr_tags", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_nr_tags_show,
@@ -308,27 +229,14 @@ static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_cpus = {
 	.attr = {.name = "cpu_list", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_cpus_show,
 };
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_poll = {
-	.attr = {.name = "io_poll", .mode = S_IWUSR | S_IRUGO },
-	.show = blk_mq_hw_sysfs_poll_show,
-	.store = blk_mq_hw_sysfs_poll_store,
-};
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_stat = {
-	.attr = {.name = "stats", .mode = S_IRUGO | S_IWUSR },
-	.show = blk_mq_hw_sysfs_stat_show,
-	.store = blk_mq_hw_sysfs_stat_store,
-};
 
 static struct attribute *default_hw_ctx_attrs[] = {
 	&blk_mq_hw_sysfs_queued.attr,
 	&blk_mq_hw_sysfs_run.attr,
-	&blk_mq_hw_sysfs_dispatched.attr,
 	&blk_mq_hw_sysfs_nr_tags.attr,
 	&blk_mq_hw_sysfs_nr_reserved_tags.attr,
 	&blk_mq_hw_sysfs_cpus.attr,
 	&blk_mq_hw_sysfs_active.attr,
-	&blk_mq_hw_sysfs_poll.attr,
-	&blk_mq_hw_sysfs_stat.attr,
 	NULL,
 };
 
-- 
2.11.0

[PATCH v2 10/10] blk-mq: move hctx and ctx counters from sysfs to debugfs

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

These counters aren't as out-of-place in sysfs as the other stuff, but
debugfs is a slightly better home for them.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 block/blk-mq-debugfs.c | 181 +++++++++++++++++++++++++++++++++++++++++++++++++
 block/blk-mq-sysfs.c   |  64 -----------------
 2 files changed, 181 insertions(+), 64 deletions(-)

diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 28ca56dd2c4e..5cd2b435a9f5 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -399,6 +399,88 @@ static const struct file_operations hctx_dispatched_fops = {
 	.release	= single_release,
 };
 
+static int hctx_queued_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	seq_printf(m, "%lu\n", hctx->queued);
+	return 0;
+}
+
+static int hctx_queued_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_queued_show, inode->i_private);
+}
+
+static ssize_t hctx_queued_write(struct file *file, const char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	hctx->queued = 0;
+	return count;
+}
+
+static const struct file_operations hctx_queued_fops = {
+	.open		= hctx_queued_open,
+	.read		= seq_read,
+	.write		= hctx_queued_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int hctx_run_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	seq_printf(m, "%lu\n", hctx->run);
+	return 0;
+}
+
+static int hctx_run_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_run_show, inode->i_private);
+}
+
+static ssize_t hctx_run_write(struct file *file, const char __user *buf,
+				 size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	hctx->run = 0;
+	return count;
+}
+
+static const struct file_operations hctx_run_fops = {
+	.open		= hctx_run_open,
+	.read		= seq_read,
+	.write		= hctx_run_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int hctx_active_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_hw_ctx *hctx = m->private;
+
+	seq_printf(m, "%d\n", atomic_read(&hctx->nr_active));
+	return 0;
+}
+
+static int hctx_active_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, hctx_active_show, inode->i_private);
+}
+
+static const struct file_operations hctx_active_fops = {
+	.open		= hctx_active_open,
+	.read		= seq_read,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
 {
 	struct blk_mq_ctx *ctx = m->private;
@@ -440,6 +522,99 @@ static const struct file_operations ctx_rq_list_fops = {
 	.release	= seq_release,
 };
 
+static int ctx_dispatched_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_ctx *ctx = m->private;
+
+	seq_printf(m, "%lu %lu\n", ctx->rq_dispatched[1], ctx->rq_dispatched[0]);
+	return 0;
+}
+
+static int ctx_dispatched_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, ctx_dispatched_show, inode->i_private);
+}
+
+static ssize_t ctx_dispatched_write(struct file *file, const char __user *buf,
+				    size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_ctx *ctx = m->private;
+
+	ctx->rq_dispatched[0] = ctx->rq_dispatched[1] = 0;
+	return count;
+}
+
+static const struct file_operations ctx_dispatched_fops = {
+	.open		= ctx_dispatched_open,
+	.read		= seq_read,
+	.write		= ctx_dispatched_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int ctx_merged_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_ctx *ctx = m->private;
+
+	seq_printf(m, "%lu\n", ctx->rq_merged);
+	return 0;
+}
+
+static int ctx_merged_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, ctx_merged_show, inode->i_private);
+}
+
+static ssize_t ctx_merged_write(struct file *file, const char __user *buf,
+				    size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_ctx *ctx = m->private;
+
+	ctx->rq_merged = 0;
+	return count;
+}
+
+static const struct file_operations ctx_merged_fops = {
+	.open		= ctx_merged_open,
+	.read		= seq_read,
+	.write		= ctx_merged_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
+static int ctx_completed_show(struct seq_file *m, void *v)
+{
+	struct blk_mq_ctx *ctx = m->private;
+
+	seq_printf(m, "%lu %lu\n", ctx->rq_completed[1], ctx->rq_completed[0]);
+	return 0;
+}
+
+static int ctx_completed_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, ctx_completed_show, inode->i_private);
+}
+
+static ssize_t ctx_completed_write(struct file *file, const char __user *buf,
+				   size_t count, loff_t *ppos)
+{
+	struct seq_file *m = file->private_data;
+	struct blk_mq_ctx *ctx = m->private;
+
+	ctx->rq_completed[0] = ctx->rq_completed[1] = 0;
+	return count;
+}
+
+static const struct file_operations ctx_completed_fops = {
+	.open		= ctx_completed_open,
+	.read		= seq_read,
+	.write		= ctx_completed_write,
+	.llseek		= seq_lseek,
+	.release	= single_release,
+};
+
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"state", 0400, &hctx_state_fops},
 	{"flags", 0400, &hctx_flags_fops},
@@ -452,10 +627,16 @@ static const struct blk_mq_debugfs_attr blk_mq_debugfs_hctx_attrs[] = {
 	{"io_poll", 0600, &hctx_io_poll_fops},
 	{"stats", 0600, &hctx_stats_fops},
 	{"dispatched", 0600, &hctx_dispatched_fops},
+	{"queued", 0600, &hctx_queued_fops},
+	{"run", 0600, &hctx_run_fops},
+	{"active", 0400, &hctx_active_fops},
 };
 
 static const struct blk_mq_debugfs_attr blk_mq_debugfs_ctx_attrs[] = {
 	{"rq_list", 0400, &ctx_rq_list_fops},
+	{"dispatched", 0600, &ctx_dispatched_fops},
+	{"merged", 0600, &ctx_merged_fops},
+	{"completed", 0600, &ctx_completed_fops},
 };
 
 int blk_mq_debugfs_register(struct request_queue *q, const char *name)
diff --git a/block/blk-mq-sysfs.c b/block/blk-mq-sysfs.c
index f505da2c542e..308b3f4fc310 100644
--- a/block/blk-mq-sysfs.c
+++ b/block/blk-mq-sysfs.c
@@ -122,34 +122,6 @@ static ssize_t blk_mq_hw_sysfs_store(struct kobject *kobj,
 	return res;
 }
 
-static ssize_t blk_mq_sysfs_dispatched_show(struct blk_mq_ctx *ctx, char *page)
-{
-	return sprintf(page, "%lu %lu\n", ctx->rq_dispatched[1],
-				ctx->rq_dispatched[0]);
-}
-
-static ssize_t blk_mq_sysfs_merged_show(struct blk_mq_ctx *ctx, char *page)
-{
-	return sprintf(page, "%lu\n", ctx->rq_merged);
-}
-
-static ssize_t blk_mq_sysfs_completed_show(struct blk_mq_ctx *ctx, char *page)
-{
-	return sprintf(page, "%lu %lu\n", ctx->rq_completed[1],
-				ctx->rq_completed[0]);
-}
-
-static ssize_t blk_mq_hw_sysfs_queued_show(struct blk_mq_hw_ctx *hctx,
-					   char *page)
-{
-	return sprintf(page, "%lu\n", hctx->queued);
-}
-
-static ssize_t blk_mq_hw_sysfs_run_show(struct blk_mq_hw_ctx *hctx, char *page)
-{
-	return sprintf(page, "%lu\n", hctx->run);
-}
-
 static ssize_t blk_mq_hw_sysfs_nr_tags_show(struct blk_mq_hw_ctx *hctx,
 					    char *page)
 {
@@ -162,11 +134,6 @@ static ssize_t blk_mq_hw_sysfs_nr_reserved_tags_show(struct blk_mq_hw_ctx *hctx,
 	return sprintf(page, "%u\n", hctx->tags->nr_reserved_tags);
 }
 
-static ssize_t blk_mq_hw_sysfs_active_show(struct blk_mq_hw_ctx *hctx, char *page)
-{
-	return sprintf(page, "%u\n", atomic_read(&hctx->nr_active));
-}
-
 static ssize_t blk_mq_hw_sysfs_cpus_show(struct blk_mq_hw_ctx *hctx, char *page)
 {
 	unsigned int i, first = 1;
@@ -185,34 +152,10 @@ static ssize_t blk_mq_hw_sysfs_cpus_show(struct blk_mq_hw_ctx *hctx, char *page)
 	return ret;
 }
 
-static struct blk_mq_ctx_sysfs_entry blk_mq_sysfs_dispatched = {
-	.attr = {.name = "dispatched", .mode = S_IRUGO },
-	.show = blk_mq_sysfs_dispatched_show,
-};
-static struct blk_mq_ctx_sysfs_entry blk_mq_sysfs_merged = {
-	.attr = {.name = "merged", .mode = S_IRUGO },
-	.show = blk_mq_sysfs_merged_show,
-};
-static struct blk_mq_ctx_sysfs_entry blk_mq_sysfs_completed = {
-	.attr = {.name = "completed", .mode = S_IRUGO },
-	.show = blk_mq_sysfs_completed_show,
-};
-
 static struct attribute *default_ctx_attrs[] = {
-	&blk_mq_sysfs_dispatched.attr,
-	&blk_mq_sysfs_merged.attr,
-	&blk_mq_sysfs_completed.attr,
 	NULL,
 };
 
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_queued = {
-	.attr = {.name = "queued", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_queued_show,
-};
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_run = {
-	.attr = {.name = "run", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_run_show,
-};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_nr_tags = {
 	.attr = {.name = "nr_tags", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_nr_tags_show,
@@ -221,22 +164,15 @@ static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_nr_reserved_tags = {
 	.attr = {.name = "nr_reserved_tags", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_nr_reserved_tags_show,
 };
-static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_active = {
-	.attr = {.name = "active", .mode = S_IRUGO },
-	.show = blk_mq_hw_sysfs_active_show,
-};
 static struct blk_mq_hw_ctx_sysfs_entry blk_mq_hw_sysfs_cpus = {
 	.attr = {.name = "cpu_list", .mode = S_IRUGO },
 	.show = blk_mq_hw_sysfs_cpus_show,
 };
 
 static struct attribute *default_hw_ctx_attrs[] = {
-	&blk_mq_hw_sysfs_queued.attr,
-	&blk_mq_hw_sysfs_run.attr,
 	&blk_mq_hw_sysfs_nr_tags.attr,
 	&blk_mq_hw_sysfs_nr_reserved_tags.attr,
 	&blk_mq_hw_sysfs_cpus.attr,
-	&blk_mq_hw_sysfs_active.attr,
 	NULL,
 };
 
-- 
2.11.0

[PATCH v3 05/10] sbitmap: add helpers for dumping to a seq_file

[Omar Sandoval]

From: Omar Sandoval <osandov@fb.com>

This is useful debugging information that will be used in the blk-mq
debugfs directory.

Reviewed-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
Jens took offense to me making the bitmap dumps binary, so this turns them into
nicely formatted hex dumps, like this one:

# cat /sys/kernel/debug/block/nvme0n1/mq/0/tags_bitmap
00000000: 0000 0000 0000 0000 0000 0000 0000 0000
00000010: 0000 2000 0000 0000 0000 0000 0000 0000
00000020: 0000 0000 0000 0000 0000 0000 0000 0000
00000030: 0000 0000 0000 0000 0000 0000 0000 0000
00000040: 0000 0000 0000 0000 0000 0000 0000 0000
00000050: 0000 0000 0000 0000 0000 0000 0000 0000
00000060: 0000 0000 0000 0000 0000 0000 0000 0000
00000070: 0000 0000 0000 0000 0000 0000 0000 0000

I decided to go with this format because 1) it's consistent with
print_hex_dump() already in the kernel, and 2) it's much easier to read
than a string of bits or hex values. In the example above, it's easy to
see that byte 0x12 is set to 0x20 (and do some arithmetic from there to
figure out that the stuck bit is 149).

 include/linux/sbitmap.h | 30 ++++++++++++++++
 lib/sbitmap.c           | 91 +++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 121 insertions(+)

diff --git a/include/linux/sbitmap.h b/include/linux/sbitmap.h
index f017fd6e69c4..d4e0a204c118 100644
--- a/include/linux/sbitmap.h
+++ b/include/linux/sbitmap.h
@@ -259,6 +259,26 @@ static inline int sbitmap_test_bit(struct sbitmap *sb, unsigned int bitnr)
 unsigned int sbitmap_weight(const struct sbitmap *sb);
 
 /**
+ * sbitmap_show() - Dump &struct sbitmap information to a &struct seq_file.
+ * @sb: Bitmap to show.
+ * @m: struct seq_file to write to.
+ *
+ * This is intended for debugging. The format may change at any time.
+ */
+void sbitmap_show(struct sbitmap *sb, struct seq_file *m);
+
+/**
+ * sbitmap_bitmap_show() - Write a hex dump of a &struct sbitmap to a &struct
+ * seq_file.
+ * @sb: Bitmap to show.
+ * @m: struct seq_file to write to.
+ *
+ * This is intended for debugging. The output isn't guaranteed to be internally
+ * consistent.
+ */
+void sbitmap_bitmap_show(struct sbitmap *sb, struct seq_file *m);
+
+/**
  * sbitmap_queue_init_node() - Initialize a &struct sbitmap_queue on a specific
  * memory node.
  * @sbq: Bitmap queue to initialize.
@@ -370,4 +390,14 @@ static inline struct sbq_wait_state *sbq_wait_ptr(struct sbitmap_queue *sbq,
  */
 void sbitmap_queue_wake_all(struct sbitmap_queue *sbq);
 
+/**
+ * sbitmap_queue_show() - Dump &struct sbitmap_queue information to a &struct
+ * seq_file.
+ * @sbq: Bitmap queue to show.
+ * @m: struct seq_file to write to.
+ *
+ * This is intended for debugging. The format may change at any time.
+ */
+void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m);
+
 #endif /* __LINUX_SCALE_BITMAP_H */
diff --git a/lib/sbitmap.c b/lib/sbitmap.c
index 8f5c3b268c77..014913b5407a 100644
--- a/lib/sbitmap.c
+++ b/lib/sbitmap.c
@@ -17,6 +17,7 @@
 
 #include <linux/random.h>
 #include <linux/sbitmap.h>
+#include <linux/seq_file.h>
 
 int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
 		      gfp_t flags, int node)
@@ -180,6 +181,62 @@ unsigned int sbitmap_weight(const struct sbitmap *sb)
 }
 EXPORT_SYMBOL_GPL(sbitmap_weight);
 
+void sbitmap_show(struct sbitmap *sb, struct seq_file *m)
+{
+	seq_printf(m, "depth=%u\n", sb->depth);
+	seq_printf(m, "weight=%u\n", sbitmap_weight(sb));
+	seq_printf(m, "bits_per_word=%u\n", 1U << sb->shift);
+	seq_printf(m, "map_nr=%u\n", sb->map_nr);
+}
+EXPORT_SYMBOL_GPL(sbitmap_show);
+
+static inline void emit_byte(struct seq_file *m, unsigned int offset, u8 byte)
+{
+	if ((offset & 0xf) == 0) {
+		if (offset != 0)
+			seq_putc(m, '\n');
+		seq_printf(m, "%08x:", offset);
+	}
+	if ((offset & 0x1) == 0)
+		seq_putc(m, ' ');
+	seq_printf(m, "%02x", byte);
+}
+
+void sbitmap_bitmap_show(struct sbitmap *sb, struct seq_file *m)
+{
+	u8 byte = 0;
+	unsigned int byte_bits = 0;
+	unsigned int offset = 0;
+	int i;
+
+	for (i = 0; i < sb->map_nr; i++) {
+		unsigned long word = READ_ONCE(sb->map[i].word);
+		unsigned int word_bits = READ_ONCE(sb->map[i].depth);
+
+		while (word_bits > 0) {
+			unsigned int bits = min(8 - byte_bits, word_bits);
+
+			byte |= (word & (BIT(bits) - 1)) << byte_bits;
+			byte_bits += bits;
+			if (byte_bits == 8) {
+				emit_byte(m, offset, byte);
+				byte = 0;
+				byte_bits = 0;
+				offset++;
+			}
+			word >>= bits;
+			word_bits -= bits;
+		}
+	}
+	if (byte_bits) {
+		emit_byte(m, offset, byte);
+		offset++;
+	}
+	if (offset)
+		seq_putc(m, '\n');
+}
+EXPORT_SYMBOL_GPL(sbitmap_bitmap_show);
+
 static unsigned int sbq_calc_wake_batch(unsigned int depth)
 {
 	unsigned int wake_batch;
@@ -377,3 +434,37 @@ void sbitmap_queue_wake_all(struct sbitmap_queue *sbq)
 	}
 }
 EXPORT_SYMBOL_GPL(sbitmap_queue_wake_all);
+
+void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m)
+{
+	bool first;
+	int i;
+
+	sbitmap_show(&sbq->sb, m);
+
+	seq_puts(m, "alloc_hint={");
+	first = true;
+	for_each_possible_cpu(i) {
+		if (!first)
+			seq_puts(m, ", ");
+		first = false;
+		seq_printf(m, "%u", *per_cpu_ptr(sbq->alloc_hint, i));
+	}
+	seq_puts(m, "}\n");
+
+	seq_printf(m, "wake_batch=%u\n", sbq->wake_batch);
+	seq_printf(m, "wake_index=%d\n", atomic_read(&sbq->wake_index));
+
+	seq_puts(m, "ws={\n");
+	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
+		struct sbq_wait_state *ws = &sbq->ws[i];
+
+		seq_printf(m, "\t{.wait_cnt=%d, .wait=%s},\n",
+			   atomic_read(&ws->wait_cnt),
+			   waitqueue_active(&ws->wait) ? "active" : "inactive");
+	}
+	seq_puts(m, "}\n");
+
+	seq_printf(m, "round_robin=%d\n", sbq->round_robin);
+}
+EXPORT_SYMBOL_GPL(sbitmap_queue_show);
-- 
2.11.0

Re: [PATCH v2 00/10] blk-mq: move debugging information from sysfs to debugfs

[Jens Axboe]

On Wed, Jan 25 2017, Omar Sandoval wrote:
> From: Omar Sandoval <osandov@fb.com>
> 
> Changes from v1:
> 
> - Make the sbitmap seq_file helpers take a (struct sbitmap *) instead of a
>   (void *), since it's not possible to use them directly as the seq_file
>   show helper, anyways
> - Fix a crash when reading ctx_map because it was attempting to do
>   exactly that ^
> 
> Cover letter from v1:
> 
> This series ends our abuse of sysfs and puts all of the debugging information
> in debugfs instead. This has a few benefits:
> 
> 1. Removes the possibility of userspace being stupid and relying on something
>    in sysfs that we only exposed for debugging.
> 2. Lifts the limitations of sysfs, including the one-value-per-file convention
>    and maximum of one page of output.
> 3. Allows us to add more debugging information that we often want but don't
>    have when debugging a live system.

Thanks Omar, applied for 4.11. The series has already been helpful in
root causing a stall.

-- 
Jens Axboe

Re: [PATCH v2 01/10] blk-mq: create debugfs directory tree

[Bart Van Assche]

On Wed, 2017-01-25 at 08:06 -0800, Omar Sandoval wrote:
> +=A0=A0=A0=A0=A0=A0=A0for (i =3D 0; i < ARRAY_SIZE(blk_mq_debugfs_ctx_att=
rs); i++) {
> +=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0const struct blk_mq_debugfs=
_attr *attr;
> +
> +=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0attr =3D &blk_mq_debugfs_ct=
x_attrs[i];
> +=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0if (!debugfs_create_file(at=
tr->name, attr->mode, ctx_dir, ctx,
> +=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=
=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 attr->fops))
> +=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0ret=
urn -ENOMEM;
> +=A0=A0=A0=A0=A0=A0=A0}

This loop occurs two times. Please consider introducing a helper function,
e.g. debugfs_create_files(). If you follow the same convention as for sysfs=
,
namely to use a sentinel at the end of an attribute array then that helper
function will only need a pointer to the attribute array and won't need the
size of that array.

Bart.=

Re: [PATCH v2 03/10] blk-mq: move hctx->dispatch and ctx->rq_list from sysfs to debugfs

[Bart Van Assche]

On Wed, 2017-01-25 at 08:06 -0800, Omar Sandoval wrote:
> +static void *hctx_dispatch_start(struct seq_file *m, loff_t *pos)
> +{
> +	struct blk_mq_hw_ctx *hctx =3D m->private;
> +
> +	spin_lock(&hctx->lock);
> +	return seq_list_start(&hctx->dispatch, *pos);
> +}
> [ ... ]
> +static void hctx_dispatch_stop(struct seq_file *m, void *v)
> +{
> +	struct blk_mq_hw_ctx *hctx =3D m->private;
> +
> +	spin_unlock(&hctx->lock);
> +}
> [ ... ]
> +static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
> +{
> +	struct blk_mq_ctx *ctx =3D m->private;
> +
> +	spin_lock(&ctx->lock);
> +	return seq_list_start(&ctx->rq_list, *pos);
> +}
> [ ... ]
> +static void ctx_rq_list_stop(struct seq_file *m, void *v)
> +{
> +	struct blk_mq_ctx *ctx =3D m->private;
> +
> +	spin_unlock(&ctx->lock);
> +}

Please add __acquires() / __releases() annotations to these functions as
appropriate to keep sparse happy.

Thanks,

Bart.=

Re: [PATCH v2 03/10] blk-mq: move hctx->dispatch and ctx->rq_list from sysfs to debugfs

[Omar Sandoval]

On Fri, Jan 27, 2017 at 10:00:19PM +0000, Bart Van Assche wrote:
> On Wed, 2017-01-25 at 08:06 -0800, Omar Sandoval wrote:
> > +static void *hctx_dispatch_start(struct seq_file *m, loff_t *pos)
> > +{
> > +	struct blk_mq_hw_ctx *hctx = m->private;
> > +
> > +	spin_lock(&hctx->lock);
> > +	return seq_list_start(&hctx->dispatch, *pos);
> > +}
> > [ ... ]
> > +static void hctx_dispatch_stop(struct seq_file *m, void *v)
> > +{
> > +	struct blk_mq_hw_ctx *hctx = m->private;
> > +
> > +	spin_unlock(&hctx->lock);
> > +}
> > [ ... ]
> > +static void *ctx_rq_list_start(struct seq_file *m, loff_t *pos)
> > +{
> > +	struct blk_mq_ctx *ctx = m->private;
> > +
> > +	spin_lock(&ctx->lock);
> > +	return seq_list_start(&ctx->rq_list, *pos);
> > +}
> > [ ... ]
> > +static void ctx_rq_list_stop(struct seq_file *m, void *v)
> > +{
> > +	struct blk_mq_ctx *ctx = m->private;
> > +
> > +	spin_unlock(&ctx->lock);
> > +}
> 
> Please add __acquires() / __releases() annotations to these functions as
> appropriate to keep sparse happy.
> 
> Thanks,
> 
> Bart.

Hey, Bart,

Jens pulled these in to his for-4.11/block branch already, could you
send these comments as incremental patches?

Thanks for taking a look!
Omar

Re: [PATCH v3 05/10] sbitmap: add helpers for dumping to a seq_file

[Bart Van Assche]

On Wed, 2017-01-25 at 14:32 -0800, Omar Sandoval wrote:
> From: Omar Sandoval <osandov@fb.com>
>=20
> This is useful debugging information that will be used in the blk-mq
> debugfs directory.
>=20
> Reviewed-by: Hannes Reinecke <hare@suse.com>
> Signed-off-by: Omar Sandoval <osandov@fb.com>
> ---
> Jens took offense to me making the bitmap dumps binary, so this turns the=
m into
> nicely formatted hex dumps, like this one:
>=20
> # cat /sys/kernel/debug/block/nvme0n1/mq/0/tags_bitmap
> 00000000: 0000 0000 0000 0000 0000 0000 0000 0000
> 00000010: 0000 2000 0000 0000 0000 0000 0000 0000
> 00000020: 0000 0000 0000 0000 0000 0000 0000 0000
> 00000030: 0000 0000 0000 0000 0000 0000 0000 0000
> 00000040: 0000 0000 0000 0000 0000 0000 0000 0000
> 00000050: 0000 0000 0000 0000 0000 0000 0000 0000
> 00000060: 0000 0000 0000 0000 0000 0000 0000 0000
> 00000070: 0000 0000 0000 0000 0000 0000 0000 0000
>=20
> I decided to go with this format because 1) it's consistent with
> print_hex_dump() already in the kernel, and 2) it's much easier to read
> than a string of bits or hex values. In the example above, it's easy to
> see that byte 0x12 is set to 0x20 (and do some arithmetic from there to
> figure out that the stuck bit is 149).

Hello Omar,

Have you considered to convert the bitmap into a contiguous bitmap first
and to format that bitmap with snprintf(..., "%*pb", ...)? I think that
will result in even easier to read output.

Bart.=