]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commit
ksm: fix endless loop on oom
authorHugh Dickins <hugh.dickins@tiscali.co.uk>
Tue, 22 Sep 2009 00:02:16 +0000 (17:02 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 22 Sep 2009 14:17:32 +0000 (07:17 -0700)
commitd952b79136a6c32a3f97e0628ca78340f1d5c6f9
treed46b096fa097c39faa21c89f329d0c84bd700062
parent81464e30609cdbd3d96d8dd6991e7481195a89a1
ksm: fix endless loop on oom

break_ksm has been looping endlessly ignoring VM_FAULT_OOM: that should
only be a problem for ksmd when a memory control group imposes limits
(normally the OOM killer will kill others with an mm until it succeeds);
but in general (especially for MADV_UNMERGEABLE and KSM_RUN_UNMERGE) we
do need to route the error (or kill) back to the caller (or sighandling).

Test signal_pending in unmerge_ksm_pages, which could be a lengthy
procedure if it has to spill into swap: returning -ERESTARTSYS so that
trivial signals will restart but fatals will terminate (is that right?
we do different things in different places in mm, none exactly this).

unmerge_and_remove_all_rmap_items was forgetting to lock when going
down the mm_list: fix that.  Whether it's successful or not, reset
ksm_scan cursor to head; but only if it's successful, reset seqnr
(shown in full_scans) - page counts will have gone down to zero.

This patch leaves a significant OOM deadlock, but it's a good step
on the way, and that deadlock is fixed in a subsequent patch.

Signed-off-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Acked-by: Izik Eidus <ieidus@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/ksm.c