Skip to contentSkip to conversation

f2fs: use f2fs_{down,up}_(read,write}_trace() for nat_tree_lock

PATCH Applied Chao Yu 2 participants 2 messages Aug 17 – Aug 21

Subsystems (2)
lore
Share & tools
lore

Chao Yu

author maintainer
#0 lore raw

PATCH 1/1 f2fs: use f2fs_{down,up}_(read,write}_trace() for nat_tree_lock

Under heavy workloads or during background GC/fallocate operations, nat_tree_lock can experience high lock contention between background readers (e.g. f2fs_get_node_info() in gc_data_segment) and writers (e.g. flush_nat_entries, set_node_addr, shrinker).

[375067.327986][T13777]  schedule+0x4c/0x114
[375067.327997][T13777]  f2fs_get_node_info+0x438/0x5c4
[375067.328002][T13777]  f2fs_get_inode_page+0x1e0/0x3f0
[375067.328013][T13777]  f2fs_iget+0x88/0x1180
[375067.328024][T13777]  f2fs_lookup+0x168/0x3a8
[375067.328035][T13777]  path_openat+0xa28/0x1b04
[375067.328046][T13777]  do_filp_open+0xac/0x130
[375067.328056][T13777]  do_sys_openat2+0x140/0x21c
[375067.328066][T13777]  __arm64_sys_openat+0x70/0x9c
[375067.330299][T13777]  schedule+0x4c/0x114
[375067.330310][T13777]  schedule_preempt_disabled+0x24/0x40
[375067.330321][T13777]  rwsem_down_write_slowpath+0x3b4/0x9d0
[375067.330332][T13777]  down_write+0x98/0x170
[375067.330343][T13777]  set_node_addr+0x74/0x4b4
[375067.330354][T13777]  f2fs_new_node_page+0xb0/0x280
[375067.330444][T13777]  f2fs_new_inode_page+0x3c/0x64
[375067.330455][T13777]  f2fs_init_inode_metadata+0x4c/0x47c
[375067.330461][T13777]  f2fs_add_regular_entry+0x258/0x5b8
[375067.330471][T13777]  f2fs_add_dentry+0x100/0x158
[375067.330476][T13777]  f2fs_do_add_link+0x84/0x140
[375067.330487][T13777]  f2fs_create+0xec/0x250
[375067.331759][T13777]  schedule+0x4c/0x114
[375067.331770][T13777]  f2fs_down_read+0x9c/0xc4
[375067.331781][T13777]  f2fs_need_inode_block_update+0x20/0x10c
[375067.331792][T13777]  f2fs_do_sync_file+0x478/0x830
[375067.331802][T13777]  f2fs_sync_file+0x2c/0x40

This patch converts nat_tree_lock to use the f2fs_{down,up}_{read,write}_trace infrastructure.

---
 Documentation/ABI/testing/sysfs-fs-f2fs |  1 +
 fs/f2fs/checkpoint.c                    |  1 +
 fs/f2fs/f2fs.h                          |  1 +
 fs/f2fs/node.c                          | 71 ++++++++++++++-----------
 include/trace/events/f2fs.h             |  3 +-
 5 files changed, 46 insertions(+), 31 deletions(-)
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
index f4e6a7415cde..85194e4c7f01 100644
@@ -986,6 +986,7 @@ Description:	This sysfs entry can be used to enable/disable to adjust priority f
 		0x00000008     gc_lock
 		0x00000010     cp_global
 		0x00000020     io_rwsem
+		0x00000040     nat_tree_lock
 		==========     ==================
 
 What:		/sys/fs/f2fs/<disk>/lock_duration_priority
fs/f2fs/checkpoint.c +1 -0
index f81ba8cc861a..a7415fa0de7f 100644
@@ -107,6 +107,7 @@ static bool need_uplift_priority(struct f2fs_rwsem *sem, bool is_write)
 	case LOCK_NAME_GC_LOCK:
 	case LOCK_NAME_CP_GLOBAL:
 	case LOCK_NAME_IO_RWSEM:
+	case LOCK_NAME_NAT_TREE_LOCK:
 		return true;
 	default:
 		f2fs_bug_on(sem->sbi, 1);
fs/f2fs/f2fs.h +1 -0
index 1b96d8718c5c..a1f5f375045a 100644
@@ -193,6 +193,7 @@ enum f2fs_lock_name {
 	LOCK_NAME_GC_LOCK,
 	LOCK_NAME_CP_GLOBAL,
 	LOCK_NAME_IO_RWSEM,
+	LOCK_NAME_NAT_TREE_LOCK,
 	LOCK_NAME_MAX,
 };
 
fs/f2fs/node.c +41 -30

269 lines not on this page · raw message

include/trace/events/f2fs.h +2 -1
index 1dd9fc5afc46..d53be932df01 100644
@@ -191,7 +191,8 @@ TRACE_DEFINE_ENUM(CP_PHASE_FINISH_CHECKPOINT);
 		{ LOCK_NAME_NODE_WRITE,		"node_write" },		\
 		{ LOCK_NAME_GC_LOCK,		"gc_lock" },		\
 		{ LOCK_NAME_CP_GLOBAL,		"cp_global" },		\
-		{ LOCK_NAME_IO_RWSEM,		"io_rwsem" })
+		{ LOCK_NAME_IO_RWSEM,		"io_rwsem" },		\
+		{ LOCK_NAME_NAT_TREE_LOCK,	"nat_tree_lock" })
 
 struct f2fs_sb_info;
 struct f2fs_io_info;
Signature · 1 line
2.49.0
#1 lore raw

Re: [f2fs-dev] [PATCH] f2fs: use f2fs_{down, up}_(read, write}_trace() for nat_tree_lock

This patch was applied to jaegeuk/f2fs.git (dev) by Jaegeuk Kim <[email protected]>:

Hello:

This patch was applied to jaegeuk/f2fs.git (dev)
by Jaegeuk Kim <[email protected]>:

Quoted 15 lines · On Mon, 17 Aug 2026 11:19:16 +0000 you wrote:

Under heavy workloads or during background GC/fallocate operations, nat_tree_lock can experience high lock contention between background readers (e.g. f2fs_get_node_info() in gc_data_segment) and writers (e.g. flush_nat_entries, set_node_addr, shrinker).

[375067.327986][T13777]  schedule+0x4c/0x114
[375067.327997][T13777]  f2fs_get_node_info+0x438/0x5c4
[375067.328002][T13777]  f2fs_get_inode_page+0x1e0/0x3f0
[375067.328013][T13777]  f2fs_iget+0x88/0x1180
[375067.328024][T13777]  f2fs_lookup+0x168/0x3a8
[375067.328035][T13777]  path_openat+0xa28/0x1b04
[375067.328046][T13777]  do_filp_open+0xac/0x130
[375067.328056][T13777]  do_sys_openat2+0x140/0x21c
[375067.328066][T13777]  __arm64_sys_openat+0x70/0x9c
Quoted 1 line

[...]

Here is the summary with links:
- [f2fs-dev] f2fs: use f2fs_{down, up}_(read, write}_trace() for nat_tree_lock

    https://git.kernel.org/jaegeuk/f2fs/c/f9c899ec50ab

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html