Re: [PATCH -next RFC] f2fs: fix null-ptr-deref in f2fs_get_dnode_of_data

[Daeho Jeong]

On Sun, Jul 24, 2022 at 3:21 AM Chao Yu <chao@xxxxxxxxxx> wrote:
>
> On 2022/7/18 19:36, Ye Bin wrote:
> > There is issue as follows when test f2fs atomic write:
> > F2FS-fs (loop0): Can't find valid F2FS filesystem in 2th superblock
> > F2FS-fs (loop0): invalid crc_offset: 0
> > F2FS-fs (loop0): f2fs_check_nid_range: out-of-range nid=1, run fsck to fix.
> > F2FS-fs (loop0): f2fs_check_nid_range: out-of-range nid=2, run fsck to fix.
> > ==================================================================
> > BUG: KASAN: null-ptr-deref in f2fs_get_dnode_of_data+0xac/0x16d0
> > Read of size 8 at addr 0000000000000028 by task rep/1990
> >
> > CPU: 4 PID: 1990 Comm: rep Not tainted 5.19.0-rc6-next-20220715 #266
> > Call Trace:
> >   <TASK>
> >   dump_stack_lvl+0x6e/0x91
> >   print_report.cold+0x49a/0x6bb
> >   kasan_report+0xa8/0x130
> >   f2fs_get_dnode_of_data+0xac/0x16d0
> >   f2fs_do_write_data_page+0x2a5/0x1030
> >   move_data_page+0x3c5/0xdf0
> >   do_garbage_collect+0x2015/0x36c0
> >   f2fs_gc+0x554/0x1d30
> >   f2fs_balance_fs+0x7f5/0xda0
> >   f2fs_write_single_data_page+0xb66/0xdc0
> >   f2fs_write_cache_pages+0x716/0x1420
> >   f2fs_write_data_pages+0x84f/0x9a0
> >   do_writepages+0x130/0x3a0
> >   filemap_fdatawrite_wbc+0x87/0xa0
> >   file_write_and_wait_range+0x157/0x1c0
> >   f2fs_do_sync_file+0x206/0x12d0
> >   f2fs_sync_file+0x99/0xc0
> >   vfs_fsync_range+0x75/0x140
> >   f2fs_file_write_iter+0xd7b/0x1850
> >   vfs_write+0x645/0x780
> >   ksys_write+0xf1/0x1e0
> >   do_syscall_64+0x3b/0x90
> >   entry_SYSCALL_64_after_hwframe+0x63/0xcd
> >
> > As 3db1de0e582c commit changed atomic write way which new a cow_inode for
> > atomic write file, and also mark cow_inode as FI_ATOMIC_FILE.
> > When f2fs_do_write_data_page write cow_inode will use cow_inode's cow_inode
> > which is NULL. Then will trigger null-ptr-deref.
> > To solve above issue, don't mark cow_inode as FI_ATOMIC_FILE.
> >
> > Fiexes: 3db1de0e582c("f2fs: change the current atomic write way")
> > Signed-off-by: Ye Bin <yebin10@xxxxxxxxxx>
> > ---
> >   fs/f2fs/file.c | 1 -
> >   1 file changed, 1 deletion(-)
> >
> > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> > index 466eb00aa516..735802b6b754 100644
> > --- a/fs/f2fs/file.c
> > +++ b/fs/f2fs/file.c
> > @@ -2062,7 +2062,6 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
> >       spin_unlock(&sbi->inode_lock[ATOMIC_FILE]);
> >
> >       set_inode_flag(inode, FI_ATOMIC_FILE);
> > -     set_inode_flag(fi->cow_inode, FI_ATOMIC_FILE);
>
> It needs to remove clear_inode_flag(fi->cow_inode, FI_ATOMIC_FILE) in
> f2fs_abort_atomic_write()?
>
> To Daeho, could you please review this patch? IIUC, cow_inode doesn't
> need to be set as atomic file?
>

Sorry for the late reply, I was OOO.

I think I missed the FG GC case for COW inode. :(
I used the FI_ATOMIC_FILE flag for COW inode in
__get_segment_type_6(). Maybe we need another flag for the COW inode
case in the function, if we remove the flag from the COW inode.
Plus, we need to remove clear_inode_flag(fi->cow_inode,
FI_ATOMIC_FILE) too in f2fs_abort_atomic_write().

Thanks,

> Thanks,
>
> >       clear_inode_flag(fi->cow_inode, FI_INLINE_DATA);
> >       f2fs_up_write(&fi->i_gc_rwsem[WRITE]);
> >