On Fri, Nov 18, 2022 at 04:05:36PM +0300, Ananda Badmaev wrote:
> 18.11.2022 01:20, coverity-bot пишет:
> > Coverity reported the following:
> >
> > *** CID 1527352: Memory - illegal accesses (OVERRUN)
> > mm/zblock.c:320 in zblock_alloc()
> > 314 }
> > 315 list = &(pool->block_lists[block_type]);
> > 316
> > 317 check:
> > 318 spin_lock(&list->lock);
> > 319 /* check if there are free slots in cache */
> > vvv CID 1527352: Memory - illegal accesses (OVERRUN)
> > vvv Overrunning array of 10208 bytes at byte offset 10208 by dereferencing pointer "list".
> > 320 block = cache_find_block(list);
> > 321 if (block)
> > 322 goto found;
> > 323 spin_unlock(&list->lock);
> > 324
> > 325 /* not found block with free slots try to allocate new empty block */
> >
> > If this is a false positive, please let us know so we can mark it as
> > such, or teach the Coverity rules to be smarter. If not, please make
> > sure fixes get into linux-next. :) For patches fixing this, please
> > include these lines (but double-check the "Fixes" first):
> >
> > Reported-by: coverity-bot <keescook+coverity-bot@xxxxxxxxxxxx>
> > Addresses-Coverity-ID: 1527352 ("Memory - illegal accesses")
> > Fixes: 9097e28c25c8 ("mm: add zblock - new allocator for use via zpool API")
> >
> > It looks like block_type is not checked to be < ARRAY_SIZE(block_desc)
> > after exiting the earlier loop, so the access through "list" may be past
> > the end of pool->block_lists.
> >
>
> There is no need for this check because it is guaranteed that this code will
> be executed only if size <= PAGE_SIZE. Since slot_size for the last list
> even exceeds PAGE_SIZE, block_type will be always valid.
Ah-ha, understood. Well, if you do want to catch it if there is ever a
typo in the block_desc values (which are not obviously >4096 without
sitting down and calculating them), perhaps add:
if (WARN_ON(block_type >= ARRAY_SIZE(block_desc))
return -ENOSPC;
--
Kees Cook
