Skip to contentSkip to conversation

f2fs: wait on atomic writes to count F2FS_CP_WB_DATA

PATCH In review Jaegeuk Kim 2 participants 5 messages Jan 4, 2019

Subsystems (2)
lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/2 f2fs: wait on atomic writes to count F2FS_CP_WB_DATA

Otherwise, we can get wrong counts incurring checkpoint hang.

IO_W (CP: -24, Data: 24, Flush: ( 0 0 1), Discard: ( 0 0))

---
 fs/f2fs/file.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)
fs/f2fs/file.c +7 -5
index bba56b39dcc5..ae2b45e75847 100644
@@ -1750,10 +1750,12 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
 
 	down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 
-	if (!get_dirty_pages(inode))
-		goto skip_flush;
-
-	f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
+	/*
+	 * Should wait end_io to count F2FS_WB_CP_DATA correctly by
+	 * f2fs_is_atomic_file.
+	 */
+	if (get_dirty_pages(inode))
+		f2fs_msg(F2FS_I_SB(inode)->sb, KERN_WARNING,
 		"Unexpected flush for atomic writes: ino=%lu, npages=%u",
 					inode->i_ino, get_dirty_pages(inode));
 	ret = filemap_write_and_wait_range(inode->i_mapping, 0, LLONG_MAX);
@@ -1761,7 +1763,7 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
 		up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
 		goto out;
 	}
-skip_flush:
+
 	set_inode_flag(inode, FI_ATOMIC_FILE);
 	clear_inode_flag(inode, FI_ATOMIC_REVOKE_REQUEST);
 	up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
Signature · 1 line
2.19.0.605.g01d371f741-goog
author maintainer
#1 lore raw

PATCH 2/2 f2fs: don't access node/meta inode mapping after iput

This fixes wrong access of address spaces of node and meta inodes after iput.

Fixes 60aa4d5536ab ("f2fs: fix use-after-free issue when accessing sbi->stat_info")
---
 fs/f2fs/debug.c | 19 ++++++++++++-------
 fs/f2fs/super.c |  5 +++++
 2 files changed, 17 insertions(+), 7 deletions(-)
fs/f2fs/debug.c +12 -7
index ebcc121920ba..503fde8349e6 100644
@@ -96,8 +96,10 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->free_secs = free_sections(sbi);
 	si->prefree_count = prefree_segments(sbi);
 	si->dirty_count = dirty_segments(sbi);
-	si->node_pages = NODE_MAPPING(sbi)->nrpages;
-	si->meta_pages = META_MAPPING(sbi)->nrpages;
+	if (sbi->node_inode)
+		si->node_pages = NODE_MAPPING(sbi)->nrpages;
+	if (sbi->meta_inode)
+		si->meta_pages = META_MAPPING(sbi)->nrpages;
 	si->nats = NM_I(sbi)->nat_cnt;
 	si->dirty_nats = NM_I(sbi)->dirty_nat_cnt;
 	si->sits = MAIN_SEGS(sbi);
@@ -175,7 +177,6 @@ static void update_sit_info(struct f2fs_sb_info *sbi)
 static void update_mem_info(struct f2fs_sb_info *sbi)
 {
 	struct f2fs_stat_info *si = F2FS_STAT(sbi);
-	unsigned npages;
 	int i;
 
 	if (si->base_mem)
@@ -258,10 +259,14 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
 						sizeof(struct extent_node);
 
 	si->page_mem = 0;
-	npages = NODE_MAPPING(sbi)->nrpages;
-	si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
-	npages = META_MAPPING(sbi)->nrpages;
-	si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
+	if (sbi->node_inode) {
+		unsigned npages = NODE_MAPPING(sbi)->nrpages;
+		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
+	}
+	if (sbi->meta_inode) {
+		unsigned npages = META_MAPPING(sbi)->nrpages;
+		si->page_mem += (unsigned long long)npages << PAGE_SHIFT;
+	}
 }
 
 static int stat_show(struct seq_file *s, void *v)
fs/f2fs/super.c +5 -0
index c46a1d4318d4..14f033e1ab42 100644
@@ -1075,7 +1075,10 @@ static void f2fs_put_super(struct super_block *sb)
 	f2fs_bug_on(sbi, sbi->fsync_node_num);
 
 	iput(sbi->node_inode);
+	sbi->node_inode = NULL;
+
 	iput(sbi->meta_inode);
+	sbi->meta_inode = NULL;
 
 	/*
 	 * iput() can update stat information, if f2fs_write_checkpoint()
@@ -3410,6 +3413,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 	f2fs_release_ino_entry(sbi, true);
 	truncate_inode_pages_final(NODE_MAPPING(sbi));
 	iput(sbi->node_inode);
+	sbi->node_inode = NULL;
 free_stats:
 	f2fs_destroy_stats(sbi);
 free_nm:
@@ -3422,6 +3426,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 free_meta_inode:
 	make_bad_inode(sbi->meta_inode);
 	iput(sbi->meta_inode);
+	sbi->meta_inode = NULL;
 free_io_dummy:
 	mempool_destroy(sbi->write_io_dummy);
 free_percpu:
Signature · 1 line
2.19.0.605.g01d371f741-goog
Reviewed-by on v1 1/2
#3 lore raw
In reply to Jaegeuk Kim
Quoted 6 lines · Jaegeuk Kim

Otherwise, we can get wrong counts incurring checkpoint hang.

IO_W (CP: -24, Data: 24, Flush: ( 0 0 1), Discard: ( 0 0))

Good catch! ;)

I can understand this condition, but for other new developer who reads this commit, it will be a little hard to understand situation here.

How about explaining a little more about problem here, maybe:

Thread A Thread B
- f2fs_write_data_pages
- __write_data_page
- f2fs_submit_page_write
- inc_page_count(F2FS_WB_DATA)
type is F2FS_WB_DATA due to file is non-atomic one
- f2fs_ioc_start_atomic_write
- set_inode_flag(FI_ATOMIC_FILE)
- f2fs_write_end_io
- dec_page_count(F2FS_WB_CP_DATA) type is F2FS_WB_DATA due to file becomes atomic one

Thanks,

author maintainer
#4 lore raw
In reply to Chao Yu
Quoted 28 lines · Chao Yu
Quoted 6 lines · Jaegeuk Kim

Otherwise, we can get wrong counts incurring checkpoint hang.

IO_W (CP: -24, Data: 24, Flush: ( 0 0 1), Discard: ( 0 0))

Good catch! ;)

I can understand this condition, but for other new developer who reads this commit, it will be a little hard to understand situation here.

How about explaining a little more about problem here, maybe:

Thread A Thread B
- f2fs_write_data_pages
- __write_data_page
- f2fs_submit_page_write
- inc_page_count(F2FS_WB_DATA)
type is F2FS_WB_DATA due to file is non-atomic one
- f2fs_ioc_start_atomic_write
- set_inode_flag(FI_ATOMIC_FILE)
- f2fs_write_end_io
- dec_page_count(F2FS_WB_CP_DATA) type is F2FS_WB_DATA due to file becomes atomic one

Quoted 1 line

Thanks, added the comment. :P

Quoted 3 lines

Thanks,