Re: [PATCH v2] f2fs: fix out-of-free problem caused by atomic write

From: Yunlong Song
Date: Fri Nov 03 2017 - 10:42:24 EST


Test:
Newest kernel source code from f2fs-dev
1G zram with f2fs
8 threads to atomic write one same file on zram
there are four kinds of atomic write at the same time:
1 no atomic start, with atomic commit
2 no atomic start, no atomic commit
3 atomic start, with atomic commit
4 atomic start, no atomic commit

And I add dump_stack after the check as following,
+ if ((sbi->user_block_count - valid_user_blocks(sbi)) <
+ fi->inmem_blocks) {
+ dump_stack();
+ err = -ENOSPC;
+ goto drop;
+ }

then we have:

[ 136.237247] F2FS-fs (zram1): Unexpected flush for atomic writes: ino=4, npages=8193
[ 136.952469] CPU: 1 PID: 1274 Comm: atomic_t2 Not tainted 4.14.0-rc4+ #109
[ 136.952947] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.8.2-0-g33fbe13 by qemu-project.org 04/01/2014
[ 136.953162] Call Trace:
[ 136.953162] dump_stack+0x4d/0x6e
[ 136.953162] commit_inmem_pages+0x258/0x270
[ 136.953162] ? __sb_start_write+0x48/0x80
[ 136.953162] ? __mnt_want_write_file+0x18/0x30
[ 136.953162] f2fs_ioctl+0x1025/0x1e30
[ 136.953162] ? up_write+0x25/0x30
[ 136.953162] ? f2fs_file_write_iter+0xf3/0x1e0
[ 136.953162] ? selinux_file_ioctl+0x114/0x1e0
[ 136.953162] do_vfs_ioctl+0x96/0x5a0
[ 136.953162] SyS_ioctl+0x79/0x90
[ 136.953162] ? SyS_lseek+0x87/0xb0
[ 136.953162] entry_SYSCALL_64_fastpath+0x13/0x94
[ 136.953162] RIP: 0033:0x434b97
[ 136.953162] RSP: 002b:00007ffc68859de8 EFLAGS: 00000202 ORIG_RAX: 0000000000000010
[ 136.953162] RAX: ffffffffffffffda RBX: 00000000006b78e0 RCX: 0000000000434b97
[ 136.953162] RDX: 00000000006b70e8 RSI: 000000000000f502 RDI: 0000000000000003
[ 136.953162] RBP: 0000000002000010 R08: 00000000006b70e8 R09: 00000000006b7160
[ 136.953162] R10: 0000000000000022 R11: 0000000000000202 R12: 00007f491a1c4010
[ 136.953162] R13: 0000000002001000 R14: 0000000002000000 R15: 00000000006b7938

So I think we should add the check code.

On 2017/11/3 12:48, Yunlong Song wrote:
Because I found that it will still lead to out-of-free problem with out that check.
I trace and find that it is possible that the committing date pages of the atomic
file is bigger than the sbi->user_block_count - valid_user_blocks(sbi), so I add
this check.

On 2017/11/3 11:46, Jaegeuk Kim wrote:
On 10/30, Yunlong Song wrote:
f2fs_balance_fs only actives once in the commit_inmem_pages, but there
are more than one page to commit, so all the other pages will miss the
check. This will lead to out-of-free problem when commit a very large
file. However, we cannot do f2fs_balance_fs for each inmem page, since
this will break atomicity. As a result, we should collect prefree
segments if needed and stop atomic commit when there are not enough
available blocks to write atomic pages.

Signed-off-by: Yunlong Song <yunlong.song@xxxxxxxxxx>
---
fs/f2fs/f2fs.h | 1 +
fs/f2fs/segment.c | 29 ++++++++++++++++++++++++++++-
2 files changed, 29 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 13a96b8..04ce48f 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -610,6 +610,7 @@ struct f2fs_inode_info {
struct list_head inmem_pages; /* inmemory pages managed by f2fs */
struct task_struct *inmem_task; /* store inmemory task */
struct mutex inmem_lock; /* lock for inmemory pages */
+ unsigned long inmem_blocks; /* inmemory blocks */
struct extent_tree *extent_tree; /* cached extent_tree entry */
struct rw_semaphore dio_rwsem[2];/* avoid racing between dio and gc */
struct rw_semaphore i_mmap_sem;
diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 46dfbca..813c110 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -210,6 +210,7 @@ void register_inmem_page(struct inode *inode, struct page *page)
list_add_tail(&fi->inmem_ilist, &sbi->inode_list[ATOMIC_FILE]);
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
inc_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
+ fi->inmem_blocks++;
mutex_unlock(&fi->inmem_lock);
trace_f2fs_register_inmem_page(page, INMEM);
@@ -221,6 +222,7 @@ static int __revoke_inmem_pages(struct inode *inode,
struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
struct inmem_pages *cur, *tmp;
int err = 0;
+ struct f2fs_inode_info *fi = F2FS_I(inode);
list_for_each_entry_safe(cur, tmp, head, list) {
struct page *page = cur->page;
@@ -263,6 +265,7 @@ static int __revoke_inmem_pages(struct inode *inode,
list_del(&cur->list);
kmem_cache_free(inmem_entry_slab, cur);
dec_page_count(F2FS_I_SB(inode), F2FS_INMEM_PAGES);
+ fi->inmem_blocks--;
}
return err;
}
@@ -302,6 +305,10 @@ void drop_inmem_pages(struct inode *inode)
if (!list_empty(&fi->inmem_ilist))
list_del_init(&fi->inmem_ilist);
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
+ if (fi->inmem_blocks) {
+ f2fs_bug_on(sbi, 1);
+ fi->inmem_blocks = 0;
+ }
mutex_unlock(&fi->inmem_lock);
clear_inode_flag(inode, FI_ATOMIC_FILE);
@@ -326,6 +333,7 @@ void drop_inmem_page(struct inode *inode, struct page *page)
f2fs_bug_on(sbi, !cur || cur->page != page);
list_del(&cur->list);
+ fi->inmem_blocks--;
mutex_unlock(&fi->inmem_lock);
dec_page_count(sbi, F2FS_INMEM_PAGES);
@@ -410,11 +418,26 @@ int commit_inmem_pages(struct inode *inode)
INIT_LIST_HEAD(&revoke_list);
f2fs_balance_fs(sbi, true);
+ if (prefree_segments(sbi)
+ && has_not_enough_free_secs(sbi, 0,
+ fi->inmem_blocks / BLKS_PER_SEC(sbi))) {
+ struct cp_control cpc;
+
+ cpc.reason = __get_cp_reason(sbi);
+ err = write_checkpoint(sbi, &cpc);
+ if (err)
+ goto drop;
+ }
f2fs_lock_op(sbi);
set_inode_flag(inode, FI_ATOMIC_COMMIT);
mutex_lock(&fi->inmem_lock);
+ if ((sbi->user_block_count - valid_user_blocks(sbi)) <
What does this mean? We already allocated blocks successfully?

+ fi->inmem_blocks) {
+ err = -ENOSPC;
+ goto drop;
+ }
err = __commit_inmem_pages(inode, &revoke_list);
if (err) {
int ret;
@@ -429,7 +452,7 @@ int commit_inmem_pages(struct inode *inode)
ret = __revoke_inmem_pages(inode, &revoke_list, false, true);
if (ret)
err = ret;
-
+drop:
/* drop all uncommitted pages */
__revoke_inmem_pages(inode, &fi->inmem_pages, true, false);
}
@@ -437,6 +460,10 @@ int commit_inmem_pages(struct inode *inode)
if (!list_empty(&fi->inmem_ilist))
list_del_init(&fi->inmem_ilist);
spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
+ if (fi->inmem_blocks) {
+ f2fs_bug_on(sbi, 1);
+ fi->inmem_blocks = 0;
+ }
mutex_unlock(&fi->inmem_lock);
clear_inode_flag(inode, FI_ATOMIC_COMMIT);
--
1.8.5.2
.



--
Thanks,
Yunlong Song