Re: [PATCH] btrfs: fix a out-of-boundary access for copy_compressed_data_to_page()

[David Sterba <dsterba@suse.cz>]

On Fri, Nov 12, 2021 at 12:41:37PM +0800, Qu Wenruo wrote:
> 
> 
> On 2021/11/12 12:17, Josef Bacik wrote:
> > On Fri, Nov 12, 2021 at 10:22:53AM +0800, Qu Wenruo wrote:
> >> [BUG]
> >> The following script can cause btrfs to crash:
> >>
> >>   mount -o compress-force=lzo $DEV /mnt
> >>   dd if=/dev/urandom of=/mnt/foo bs=4k count=1
> >>   sync
> >>
> >> The calltrace looks like this:
> >>
> >>   general protection fault, probably for non-canonical address 0xe04b37fccce3b000: 0000 [#1] PREEMPT SMP NOPTI
> >>   CPU: 5 PID: 164 Comm: kworker/u20:3 Not tainted 5.15.0-rc7-custom+ #4
> >>   Workqueue: btrfs-delalloc btrfs_work_helper [btrfs]
> >>   RIP: 0010:__memcpy+0x12/0x20
> >>   Call Trace:
> >>    lzo_compress_pages+0x236/0x540 [btrfs]
> >>    btrfs_compress_pages+0xaa/0xf0 [btrfs]
> >>    compress_file_range+0x431/0x8e0 [btrfs]
> >>    async_cow_start+0x12/0x30 [btrfs]
> >>    btrfs_work_helper+0xf6/0x3e0 [btrfs]
> >>    process_one_work+0x294/0x5d0
> >>    worker_thread+0x55/0x3c0
> >>    kthread+0x140/0x170
> >>    ret_from_fork+0x22/0x30
> >>   ---[ end trace 63c3c0f131e61982 ]---
> >>
> >> [CAUSE]
> >> In lzo_compress_pages(), parameter @out_pages is not only an output
> >> parameter (for the compressed pages), but also an input parameter, for
> >> the maximum amount of pages we can utilize.
> >>
> >> In commit d4088803f511 ("btrfs: subpage: make lzo_compress_pages()
> >> compatible"), the refactor doesn't take @out_pages as an input, thus
> >> completely ignoring the limit.
> >>
> >> And for compress-force case, we could hit incompressible data that
> >> compressed size would go beyond the page limit, and cause above crash.
> >>
> >> [FIX]
> >> Save @out_pages as @max_nr_page, and pass it to lzo_compress_pages(),
> >> and check if we're beyond the limit before accessing the pages.
> >>
> >> Reported-by: Omar Sandoval <osandov@fb.com>
> >> Fixes: d4088803f511 ("btrfs: subpage: make lzo_compress_pages() compatible")
> >> Signed-off-by: Qu Wenruo <wqu@suse.com>
> >> ---
> >>   fs/btrfs/lzo.c | 12 +++++++++++-
> >>   1 file changed, 11 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/fs/btrfs/lzo.c b/fs/btrfs/lzo.c
> >> index 00cffc183ec0..f410ceabcdbd 100644
> >> --- a/fs/btrfs/lzo.c
> >> +++ b/fs/btrfs/lzo.c
> >> @@ -125,6 +125,7 @@ static inline size_t read_compress_length(const char *buf)
> >>   static int copy_compressed_data_to_page(char *compressed_data,
> >>   					size_t compressed_size,
> >>   					struct page **out_pages,
> >> +					unsigned long max_nr_page,
> >
> > If you want to do const down below you should use const here probably?  Thanks,
> 
> Right, max_nr_page should also be const.

const for non-pointer parameters does not make much sense, it only
prevents reuse of the variable inside the function.