Coverity: do_migrate_range(): Error handling issues

Coverity: do_migrate_range(): Error handling issues

[coverity-bot]

Hello!

This is an experimental semi-automated report about issues detected by
Coverity from a scan of next-20210608 as part of the linux-next scan project:
https://scan.coverity.com/projects/linux-next-weekly-scan

You're getting this email because you were associated with the identified
lines of code (noted below) that were touched by commits:

  Tue Dec 15 12:13:39 2020 -0800
    013339df116c ("mm/rmap: always do TTU_IGNORE_ACCESS")

Coverity reported the following:

*** CID 1505146:  Error handling issues  (CHECKED_RETURN)
/mm/memory_hotplug.c: 1422 in do_migrate_range()
1416     		 * the unmap as the catch all safety net).
1417     		 */
1418     		if (PageHWPoison(page)) {
1419     			if (WARN_ON(PageLRU(page)))
1420     				isolate_lru_page(page);
1421     			if (page_mapped(page))
vvv     CID 1505146:  Error handling issues  (CHECKED_RETURN)
vvv     Calling "try_to_unmap" without checking return value (as is done elsewhere 4 out of 5 times).
1422     				try_to_unmap(page, TTU_IGNORE_MLOCK);
1423     			continue;
1424     		}
1425
1426     		if (!get_page_unless_zero(page))
1427     			continue;

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@chromium.org>
Addresses-Coverity-ID: 1505146 ("Error handling issues")
Fixes: 013339df116c ("mm/rmap: always do TTU_IGNORE_ACCESS")

Thanks for your attention!

-- 
Coverity-bot

Re: Coverity: do_migrate_range(): Error handling issues

[Linus Torvalds]

On Tue, Jun 8, 2021 at 11:05 AM coverity-bot <keescook@chromium.org> wrote:
>
> You're getting this email because you were associated with the identified
> lines of code (noted below) that were touched by commits:
>
>   Tue Dec 15 12:13:39 2020 -0800
>     013339df116c ("mm/rmap: always do TTU_IGNORE_ACCESS")

Heh. Entirely independent change, but whatever..

That does mean that the suggested

> Fixes: 013339df116c ("mm/rmap: always do TTU_IGNORE_ACCESS")

tag would have been entirely bogus. That commit didn't actually change
the coverity issue in question anyway, it just happened to change that
line in other ways.

Maybe the coverity-bot logic could actually try to see if a coverity
report changes, which it wouldn't have done here.

The report itself is:

> 1421                            if (page_mapped(page))
> vvv     CID 1505146:  Error handling issues  (CHECKED_RETURN)
> vvv     Calling "try_to_unmap" without checking return value (as is done elsewhere 4 out of 5 times).
> 1422                                    try_to_unmap(page, TTU_IGNORE_MLOCK);

Hmm. That "4 out of 5 times" seems misleading. It implies that 80% of
callers check the return value.

But when I actually grep for it, I see 4 call-sites that don't check
the return value, and 5 that do. So it's more like 55% of callers that
do check.

Anyway, this is a false positive. You don't have to check the return
value.  The main example of doing so is the page-out code in
mm/vmscan.c, and it does so just to keep track of success/fail
statistics.

             Linus

Re: Coverity: do_migrate_range(): Error handling issues

[Shakeel Butt]

On Tue, Jun 8, 2021 at 12:35 PM Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
[...]
>
> Anyway, this is a false positive. You don't have to check the return
> value.  The main example of doing so is the page-out code in
> mm/vmscan.c, and it does so just to keep track of success/fail
> statistics.

BTW there is an in-fly patch to make the try_to_umap() void function.

https://lore.kernel.org/lkml/alpine.LSU.2.11.2106031527070.11704@eggly.anvils/

Re: Coverity: do_migrate_range(): Error handling issues

[Kees Cook]

On Tue, Jun 08, 2021 at 12:35:34PM -0700, Linus Torvalds wrote:
> But when I actually grep for it, I see 4 call-sites that don't check
> the return value, and 5 that do. So it's more like 55% of callers that
> do check.
> 
> Anyway, this is a false positive. You don't have to check the return
> value.  The main example of doing so is the page-out code in
> mm/vmscan.c, and it does so just to keep track of success/fail
> statistics.

Okay, cool; thanks for taking a closer look at it.

-- 
Kees Cook