Messages in this thread | | | Date | Sat, 19 Nov 2022 13:24:32 +0800 | Subject | Re: [PATCH] nilfs2: Fix nilfs_sufile_mark_dirty() not set segment usage as dirty | From | Chen Zhongjin <> |
| |
On 2022/11/19 6:11, Andrew Morton wrote: > On Fri, 18 Nov 2022 14:33:04 +0800 Chen Zhongjin <chenzhongjin@huawei.com> wrote: > >> In nilfs_sufile_mark_dirty(), the buffer and inode are set dirty, but >> nilfs_segment_usage is not set dirty, which makes it can be found by >> nilfs_sufile_alloc() because it checks nilfs_segment_usage_clean(su). >> >> This will cause the problem reported by syzkaller: >> https://syzkaller.appspot.com/bug?id=c7c4748e11ffcc367cef04f76e02e931833cbd24 >> >> It's because the case starts with segbuf1.segnum = 3, nextnum = 4, and >> nilfs_sufile_alloc() not called to allocate a new segment. >> >> The first time nilfs_segctor_extend_segments() allocated segment >> segbuf2.segnum = segbuf1.nextnum = 4, then nilfs_sufile_alloc() found >> nextnextnum = 4 segment because its su is not set dirty. >> So segbuf2.nextnum = 4, which causes next segbuf3.segnum = 4. >> >> sb_getblk() will get same bh for segbuf2 and segbuf3, and this bh is >> added to both buffer lists of two segbuf. >> It makes the list head of second list linked to the first one. When >> iterating the first one, it will access and deref the head of second, >> which causes NULL pointer dereference. >> >> Fixes: 9ff05123e3bf ("nilfs2: segment constructor") > Merged in 2009!
Yes, seems it is introduced at the beginning of this file and the function called nilfs_touch_segusage().
Wired that this problem is not discovered utill now. So I'm wondering that whether this is a real-world problem or just a use case constructed maliciously by syzkaller. But according to the result of syzkaller bisection, this problem should have a history. >> --- a/fs/nilfs2/sufile.c >> +++ b/fs/nilfs2/sufile.c >> @@ -495,12 +495,18 @@ void nilfs_sufile_do_free(struct inode *sufile, __u64 segnum, >> int nilfs_sufile_mark_dirty(struct inode *sufile, __u64 segnum) >> { >> struct buffer_head *bh; >> + void *kaddr; >> + struct nilfs_segment_usage *su; >> int ret; >> >> ret = nilfs_sufile_get_segment_usage_block(sufile, segnum, 0, &bh); >> if (!ret) { >> mark_buffer_dirty(bh); >> nilfs_mdt_mark_dirty(sufile); >> + kaddr = kmap_atomic(bh->b_page); >> + su = nilfs_sufile_block_get_segment_usage(sufile, segnum, bh, kaddr); >> + nilfs_segment_usage_set_dirty(su); >> + kunmap_atomic(kaddr); >> brelse(bh); >> } >> return ret; > Do we feel that this fix should be backported into -stable kernels? Sorry that I'm not familiar with the specific use scenarios of nilfs2. So I can't offer a better advice. I think if it is a problem that not happen easily in normal situations there's no necessary to backport it to stable.
| |