Description: Don't call schedule_timeout_killable() with oom_lock held
References:
 https://patchwork.kernel.org/patch/10395909/
 https://patchwork.kernel.org/patch/9842889/
 https://www.spinics.net/lists/linux-mm/msg117896.html
 https://www.spinics.net/lists/linux-mm/msg117960.html
Notes:
 carnil> Commit 9bfe5ded054b ("mm, oom: remove sleep from under oom_lock")
 carnil> is a mitigation for CVE-2016-10723.
 carnil> https://lore.kernel.org/lkml/cb2d635c-c14d-c2cc-868a-d4c447364f0d@i-love.sakura.ne.jp/
 bwh> On 3.16 the OOM killer usually kills the reproducer fairly quickly,
 bwh> but not always.  It still spams the kernel log and in some cases
 bwh> it seemed to cause a filesystem error causing / to go read-only.
 bwh> I assume 4.9 is also affected.
Bugs:
upstream: ignored "Negligible security impact, long-standing limitation"
4.19-upstream-stable: ignored "Negligible security impact, long-standing limitation"
4.9-upstream-stable: ignored "Negligible security impact, long-standing limitation"
3.16-upstream-stable: ignored "Negligible security impact, long-standing limitation"
sid: ignored "Negligible security impact, long-standing limitation"
4.19-buster-security: ignored "Negligible security impact, long-standing limitation"
4.9-stretch-security: ignored "Negligible security impact, long-standing limitation"
3.16-jessie-security: ignored "Negligible security impact, long-standing limitation"