]> git.hungrycats.org Git - linux/commit
udf: validate sparing table length as an entry count, not a byte count
authorBryam Vargas <hexlabsecurity@proton.me>
Fri, 12 Jun 2026 06:40:01 +0000 (01:40 -0500)
committerJan Kara <jack@suse.cz>
Fri, 12 Jun 2026 10:49:27 +0000 (12:49 +0200)
commit3ec997bd5508e9b25210b5bbec89031629cdb093
tree523e6d8cc9c34e3b7e86f440532b5dd8de8e6c7e
parent5fa1d6a5ec2356d2107dead614437c66fa7138b1
udf: validate sparing table length as an entry count, not a byte count

udf_load_sparable_map() accepts a sparing table when

sizeof(*st) + le16_to_cpu(st->reallocationTableLen) > sb->s_blocksize

is false, i.e. it treats reallocationTableLen as a number of BYTES that
must fit in the block.  But the table is walked as an array of 8-byte
sparingEntry elements:

for (i = 0; i < le16_to_cpu(st->reallocationTableLen); i++) {
struct sparingEntry *entry = &st->mapEntry[i];
... entry->origLocation ...
}

in udf_get_pblock_spar15() and udf_relocate_blocks().  A
reallocationTableLen of N therefore passes the check whenever
sizeof(*st) + N <= blocksize, yet the consumers index
sizeof(*st) + N * sizeof(struct sparingEntry) bytes -- up to ~8x the
block.  On a crafted UDF image this is an out-of-bounds read in
udf_get_pblock_spar15(); udf_relocate_blocks() additionally feeds the
same length to udf_update_tag(), whose crc_itu_t() reads far past the
block, and its memmove() through st->mapEntry[] is an out-of-bounds
write.

Validate reallocationTableLen as the entry count it is, with
struct_size().

Fixes: 1df2ae31c724 ("udf: Fortify loading of sparing table")
Cc: stable@vger.kernel.org
Signed-off-by: Bryam Vargas <hexlabsecurity@proton.me>
Link: https://patch.msgid.link/20260612-b4-disp-91780c4e-v1-1-f15112ff6882@proton.me
Signed-off-by: Jan Kara <jack@suse.cz>
fs/udf/super.c