From: Yu Kuai <yukuai3@xxxxxxxxxx>
After a disk is created, debugfs inode and dentry will be created
together, and the memory used for debugfs can't be freed until disk
removal.
The number of debugfs inode and dentry is based on how many cpus and
hctxs. For example, testing on a 128-core environemt, with default
module parameters, each loop device will cost 1679KB memory, and debugfs
will cost 336KB(20%).
The memory cost for debugfs for a disk seems little, but if a big machine
contains thousands of disks, the cost will be xxGB. This memory overhead
can be avoided by disabling CONFIG_BLK_DEBUG_FS.
This patch add a disk level switch that can enable/disable debugfs
dynamically, so that user can disable debugfs if they care about the
memory overhead, in the meantime, debugfs can be enabled again in demand.
Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>
---
block/blk-mq-debugfs.c | 58 ++++++++++++++++++++++++++++++++++++++----
block/blk-mq-debugfs.h | 1 +
block/blk-sysfs.c | 41 +++++++++++++++++++++++++++++
include/linux/blkdev.h | 2 ++
4 files changed, 97 insertions(+), 5 deletions(-)
diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 212a7f301e73..3ffc27fd4d07 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -657,6 +657,9 @@ void blk_mq_debugfs_register(struct request_queue *q)
struct blk_mq_hw_ctx *hctx;
unsigned long i;
+ if (!test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags))
+ return;
+
debugfs_create_files(q->debugfs_dir, q, blk_mq_debugfs_queue_attrs);
/*
@@ -685,6 +688,47 @@ void blk_mq_debugfs_register(struct request_queue *q)
}
}
+static void debugfs_remove_files(struct dentry *parent,
+ const struct blk_mq_debugfs_attr *attr)
+{
+ if (IS_ERR_OR_NULL(parent))
+ return;
+
+ for (; attr->name; attr++)
+ debugfs_lookup_and_remove(attr->name, parent);
+}
+
+void blk_mq_debugfs_unregister(struct request_queue *q)
+{
+ struct blk_mq_hw_ctx *hctx;
+ unsigned long i;
+
+ if (q->rq_qos) {
+ struct rq_qos *rqos = q->rq_qos;
+
+ while (rqos) {
+ if (rqos->debugfs_dir)
+ blk_mq_debugfs_unregister_rqos(rqos);
+ rqos = rqos->next;
+ }
+ }
+
+ debugfs_remove_recursive(q->rqos_debugfs_dir);
+ q->rqos_debugfs_dir = NULL;
+
+ queue_for_each_hw_ctx(q, hctx, i) {
+ if (hctx->debugfs_dir)
+ blk_mq_debugfs_unregister_hctx(hctx);
+ if (hctx->sched_debugfs_dir)
+ blk_mq_debugfs_unregister_sched_hctx(hctx);
+ }
+
+ if (q->sched_debugfs_dir)
+ blk_mq_debugfs_unregister_sched(q);
+
+ debugfs_remove_files(q->debugfs_dir, blk_mq_debugfs_queue_attrs);
+}
+
static void blk_mq_debugfs_register_ctx(struct blk_mq_hw_ctx *hctx,
struct blk_mq_ctx *ctx)
{
@@ -704,7 +748,7 @@ void blk_mq_debugfs_register_hctx(struct request_queue *q,
char name[20];
int i;
- if (!q->debugfs_dir)
+ if (!q->debugfs_dir || !test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags))
return;
snprintf(name, sizeof(name), "hctx%u", hctx->queue_num);
@@ -718,7 +762,8 @@ void blk_mq_debugfs_register_hctx(struct request_queue *q,
void blk_mq_debugfs_unregister_hctx(struct blk_mq_hw_ctx *hctx)
{
- if (!hctx->queue->debugfs_dir)
+ if (!hctx->queue->debugfs_dir ||
+ !test_bit(QUEUE_FLAG_DEBUGFS, &hctx->queue->queue_flags))
return;
debugfs_remove_recursive(hctx->debugfs_dir);
hctx->sched_debugfs_dir = NULL;
@@ -756,7 +801,8 @@ void blk_mq_debugfs_register_sched(struct request_queue *q)
if (!q->debugfs_dir)
return;
- if (!e->queue_debugfs_attrs)
+ if (!e->queue_debugfs_attrs ||
+ !test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags))
return;
q->sched_debugfs_dir = debugfs_create_dir("sched", q->debugfs_dir);
@@ -802,7 +848,8 @@ void blk_mq_debugfs_register_rqos(struct rq_qos *rqos)
lockdep_assert_held(&q->debugfs_mutex);
- if (rqos->debugfs_dir || !rqos->ops->debugfs_attrs)
+ if (rqos->debugfs_dir || !rqos->ops->debugfs_attrs ||
+ !test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags))
return;
if (!q->rqos_debugfs_dir)
@@ -828,7 +875,8 @@ void blk_mq_debugfs_register_sched_hctx(struct request_queue *q,
if (!hctx->debugfs_dir)
return;
- if (!e->hctx_debugfs_attrs)
+ if (!e->hctx_debugfs_attrs ||
+ !test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags))
return;
hctx->sched_debugfs_dir = debugfs_create_dir("sched",
diff --git a/block/blk-mq-debugfs.h b/block/blk-mq-debugfs.h
index 9c7d4b6117d4..c8dd03f73d8c 100644
--- a/block/blk-mq-debugfs.h
+++ b/block/blk-mq-debugfs.h
@@ -21,6 +21,7 @@ int __blk_mq_debugfs_rq_show(struct seq_file *m, struct request *rq);
int blk_mq_debugfs_rq_show(struct seq_file *m, void *v);
void blk_mq_debugfs_register(struct request_queue *q);
+void blk_mq_debugfs_unregister(struct request_queue *q);
void blk_mq_debugfs_register_hctx(struct request_queue *q,
struct blk_mq_hw_ctx *hctx);
void blk_mq_debugfs_unregister_hctx(struct blk_mq_hw_ctx *hctx);
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 1a743b4f2958..450acea23f21 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -618,6 +618,44 @@ QUEUE_RW_ENTRY(queue_iostats, "iostats");
QUEUE_RW_ENTRY(queue_random, "add_random");
QUEUE_RW_ENTRY(queue_stable_writes, "stable_writes");
+#ifdef CONFIG_BLK_DEBUG_FS
+static ssize_t queue_debugfs_show(struct request_queue *q, char *page)
+{
+ return queue_var_show(test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags),
+ page);
+}
+
+static ssize_t queue_debugfs_store(struct request_queue *q, const char *page,
+ size_t count)
+{
+ unsigned long val;
+ bool enabled;
+ ssize_t ret = queue_var_store(&val, page, count);
+
+ if (ret < 0)
+ return ret;
+
+ mutex_lock(&q->debugfs_mutex);
+ enabled = test_bit(QUEUE_FLAG_DEBUGFS, &q->queue_flags);
+ if (val) {
+ if (!enabled && queue_is_mq(q)) {
+ blk_queue_flag_set(QUEUE_FLAG_DEBUGFS, q);
+ blk_mq_debugfs_register(q);
+ }
+ } else {
+ if (enabled) {
+ blk_mq_debugfs_unregister(q);
+ blk_queue_flag_clear(QUEUE_FLAG_DEBUGFS, q);
+ }
+ }
+ mutex_unlock(&q->debugfs_mutex);
+
+ return ret;
+}
+
+QUEUE_RW_ENTRY(queue_debugfs, "debugfs");
+#endif
+
static struct attribute *queue_attrs[] = {
&queue_requests_entry.attr,
&queue_ra_entry.attr,
@@ -664,6 +702,9 @@ static struct attribute *queue_attrs[] = {
#endif
&queue_virt_boundary_mask_entry.attr,
&queue_dma_alignment_entry.attr,
+#ifdef CONFIG_BLK_DEBUG_FS
+ &queue_debugfs_entry.attr,
+#endif
NULL,
};
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index e3242e67a8e3..be51592751b2 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -544,6 +544,7 @@ struct request_queue {
#define QUEUE_FLAG_NONROT 6 /* non-rotational device (SSD) */
#define QUEUE_FLAG_VIRT QUEUE_FLAG_NONROT /* paravirt device */
#define QUEUE_FLAG_IO_STAT 7 /* do disk/partitions IO accounting */
+#define QUEUE_FLAG_DEBUGFS 8 /* supports debugfs */
#define QUEUE_FLAG_NOXMERGES 9 /* No extended merges */
#define QUEUE_FLAG_ADD_RANDOM 10 /* Contributes to random pool */
#define QUEUE_FLAG_SYNCHRONOUS 11 /* always completes in submit context */
@@ -566,6 +567,7 @@ struct request_queue {
#define QUEUE_FLAG_SKIP_TAGSET_QUIESCE 31 /* quiesce_tagset skip the queue*/
#define QUEUE_FLAG_MQ_DEFAULT ((1UL << QUEUE_FLAG_IO_STAT) | \
+ (1UL << QUEUE_FLAG_DEBUGFS) | \
(1UL << QUEUE_FLAG_SAME_COMP) | \
(1UL << QUEUE_FLAG_NOWAIT))
--
2.39.2
