Hi Carlos,
Cc block list given it is related with interface between fs and block layer.
On Fri, Feb 22, 2019 at 10:14 PM Carlos Maiolino <cmaiolino@xxxxxxxxxx> wrote:
>
> guard_bio_eod() can truncate a segment in bio to allow it to do IO on
> odd last sectors of a device.
>
> It already checks if the IO starts past EOD, but it does not consider
> the possibility of an IO request starting within device boundaries can
> contain more than one segment past EOD.
>
> In such cases, truncated_bytes can be bigger than PAGE_SIZE, and will
> underflow bvec->bv_len.
It can cause memory corruption even for < PAGE_SIZE, also it can be correct
to see > PAGE_SIZE truncated_bytes:
- xfs is going to support big block size which may be 64k
- multi-page bvec patches have been in block tree, and it is planned to land
v5.1, so > PAGE_SIZE truncating may come because .bv_len can be much
bigger than PAGE_SIZE
- suppose fs block size is 4k, bio sector is 1022 and size is 4k, and
disk size is
1024, last bvec is (0, 1024), then 'truncated_bytes' will be 3k, so
the check can't
capture this case, and memory corruption still may be caused.
>
> Fix this by checking if truncated_bytes is lower than PAGE_SIZE.
>
> This situation has been found on filesystems such as isofs and vfat,
> which doesn't check the device size before mount, if the device is
> smaller than the filesystem itself, a readahead on such filesystem,
> which spans EOD, can trigger this situation, leading a call to
> zero_user() with a wrong size possibly corrupting memory.
>
> I didn't see any crash, or didn't let the system run long enough to
> check if memory corruption will be hit somewhere, but adding
> instrumentation to guard_bio_end() to check truncated_bytes size, was
> enough to see the error.
>
> The following script can trigger the error.
>
> MNT=/mnt
> IMG=./DISK.img
> DEV=/dev/loop0
>
> mkfs.vfat $IMG
> mount $IMG $MNT
> cp -R /etc $MNT &> /dev/null
> umount $MNT
>
> losetup -D
>
> losetup --find --show --sizelimit 16247280 $IMG
> mount $DEV $MNT
>
> find $MNT -type f -exec cat {} + >/dev/null
BTW, which kernel is for this issue? Did you investigate why the bad bio
is made?
>
> Kudos to Eric Sandeen for coming up with the reproducer above
>
> Signed-off-by: Carlos Maiolino <cmaiolino@xxxxxxxxxx>
> ---
>
> P.S. I'm not 100% proficient in bio internals, so I'm not sure if this is the
> right fix, so comments are much appreciated.
> Thanks
>
> fs/buffer.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/fs/buffer.c b/fs/buffer.c
> index 784de3dbcf28..32dc5cd2f6ba 100644
> --- a/fs/buffer.c
> +++ b/fs/buffer.c
> @@ -3063,6 +3063,13 @@ void guard_bio_eod(int op, struct bio *bio)
> /* Uhhuh. We've got a bio that straddles the device size! */
> truncated_bytes = bio->bi_iter.bi_size - (maxsector << 9);
>
> + /*
> + * The bio contains more than one segment which spans EOD, just return
> + * and let IO layer turn it into an EIO
> + */
> + if (truncated_bytes > PAGE_SIZE)
> + return;
> +
The correct check should be:
+ if (truncated_bytes > bvec->bv_len)
+ return;
Also it should be helpful to warn on this bad bio, but it is fine to not do it
here cause block layer logs this bad bio too.
BTW, this is just a workaround, not one real fix on the reported problem.
Thanks,
Ming Lei
