On Thu 07-01-21 17:41:16, Steve Magnani wrote:
> From: Steven J. Magnani <magn...@ieee.org>
>
> When extending a file, make sure that the pointer to the last written
> extent does not get adjusted into the header (tag) portion of an
> Allocation Extent Descriptor. Otherwise, a subsequent call to
> udf_update_exents() can clobber the AED's tagLocation field, replacing
> it with the logical block number of a file extent.
>
> Signed-off-by: Steven J. Magnani <magn...@ieee.org>
Thanks! It took me some time to understand what's the actual problem but
eventually I've understood - I've updated the changelog and a comment to
explain a bit more and merged the patch into my tree.
Honza
> ---
> --- a/fs/udf/inode.c 2020-12-28 20:51:29.000000000 -0600
> +++ b/fs/udf/inode.c 2021-01-01 19:20:37.163767576 -0600
> @@ -547,11 +547,14 @@ static int udf_do_extend_file(struct ino
>
> udf_write_aext(inode, last_pos, &last_ext->extLocation,
> last_ext->extLength, 1);
> - /*
> - * We've rewritten the last extent but there may be empty
> - * indirect extent after it - enter it.
> - */
> - udf_next_aext(inode, last_pos, &tmploc, &tmplen, 0);
> +
> + if (new_block_bytes || prealloc_len) {
> + /*
> + * We've rewritten the last extent but there may be
> empty
> + * indirect extent after it - enter it.
> + */
> + udf_next_aext(inode, last_pos, &tmploc, &tmplen, 0);
> + }
> }
>
> /* Managed to do everything necessary? */
>
--
Jan Kara <j...@suse.com>
SUSE Labs, CR
