Re: KMSAN: uninit-value in mpol_rebind_mm

[Vlastimil Babka]

On 12/31/18 8:51 AM, syzbot wrote:
> Hello,
> 
> syzbot found the following crash on:
> 
> HEAD commit:    79fc24ff6184 kmsan: highmem: use kmsan_clear_page() in cop..
> git tree:       kmsan
> console output: https://syzkaller.appspot.com/x/log.txt?x=13c48b67400000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=901dd030b2cc57e7
> dashboard link: https://syzkaller.appspot.com/bug?extid=b19c2dc2c990ea657a71
> compiler:       clang version 8.0.0 (trunk 349734)
> 
> Unfortunately, I don't have any reproducer for this crash yet.
> 
> IMPORTANT: if you fix the bug, please add the following tag to the commit:
> Reported-by: syzbot+b19c2dc2c990ea657a71@xxxxxxxxxxxxxxxxxxxxxxxxx
> 
> ==================================================================
> BUG: KMSAN: uninit-value in mpol_rebind_policy mm/mempolicy.c:353 [inline]
> BUG: KMSAN: uninit-value in mpol_rebind_mm+0x249/0x370 mm/mempolicy.c:384

The report doesn't seem to indicate where the uninit value resides in
the mempolicy object. I'll have to guess. mm/mempolicy.c:353 contains:

        if (!mpol_store_user_nodemask(pol) &&
            nodes_equal(pol->w.cpuset_mems_allowed, *newmask))

"mpol_store_user_nodemask(pol)" is testing pol->flags, which I couldn't
see being uninitialized after leaving mpol_new(). So I'll guess it's
actually about accessing pol->w.cpuset_mems_allowed on line 354.

For w.cpuset_mems_allowed to be not initialized and the nodes_equal()
reachable for a mempolicy where mpol_set_nodemask() is called in
do_mbind(), it seems the only possibility is a MPOL_PREFERRED policy
with empty set of nodes, i.e. MPOL_LOCAL equivalent. Let's see if the
patch below helps. This code is a maze to me. Note the uninit access
should be benign, rebinding this kind of policy is always a no-op.

----8<----