Re: OOM killer doesn't kill the right task....

[David Chinner]

On Mon, Apr 21, 2008 at 05:24:04PM +0900, KOSAKI Motohiro wrote:
> Hi David
> 
> > Running in a 512MB UML system without swap, XFSQA test 084 reliably
> > kills the kernel completely as the OOM killer is unable to find a
> > task to kill. log output is below.
> > 
> > I don't know when it started failing - ISTR this working just fine
> > on 2.6.24 kernels.
> 
> Can you reproduce it on non UML box?

Not exactly. On a 64k page size ia64 box it kills my ssh session and all
it's children which includes the errant process (log below).

It doesn't kill the machine, but if I cared enough I'd argue that
even that is killing the wrong process because it's pretty damn
clear that the only process on the box using more than a couple of
MB of memory is the resvtest program....

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

resvtest invoked oom-killer: gfp_mask=0x1280d2, order=0, oomkilladj=0

Call Trace:
 [<a0000001000125e0>] show_stack+0x40/0xa0
                                sp=e000003022f0fb20 bsp=e000003022f01168
 [<a000000100012670>] dump_stack+0x30/0x60
                                sp=e000003022f0fcf0 bsp=e000003022f01150
 [<a000000100104060>] oom_kill_process+0x80/0x3a0
                                sp=e000003022f0fcf0 bsp=e000003022f010f8
 [<a000000100104ce0>] out_of_memory+0x4e0/0x660
                                sp=e000003022f0fd00 bsp=e000003022f010b0
 [<a00000010010bf00>] __alloc_pages+0x500/0x620
                                sp=e000003022f0fd90 bsp=e000003022f01040
 [<a000000100145020>] alloc_page_vma+0x1c0/0x200
                                sp=e000003022f0fda0 bsp=e000003022f01008
 [<a000000100122d00>] handle_mm_fault+0x3a0/0xe60
                                sp=e000003022f0fda0 bsp=e000003022f00f88
 [<a00000010085b700>] ia64_do_page_fault+0x2a0/0xaa0
                                sp=e000003022f0fda0 bsp=e000003022f00f30
 [<a000000100009e20>] ia64_leave_kernel+0x0/0x270
                                sp=e000003022f0fe30 bsp=e000003022f00f30
Mem-info:
Node 0 Normal per-cpu:
CPU    0: hi:    0, btch:   1 usd:   0
CPU    1: hi:    0, btch:   1 usd:   0
CPU    2: hi:    0, btch:   1 usd:   0
CPU    3: hi:    0, btch:   1 usd:   0
Active:5044 inactive:64 dirty:0 writeback:0 unstable:0
 free:34 slab:1110 mapped:6 pagetables:104 bounce:0
Node 0 Normal free:2176kB min:2880kB low:3584kB high:4288kB active:322816kB inactive:4096kB present:523776kB pages_scanned:32980 all_unreclaimable? yes
lowmem_reserve[]: 0 0
Node 0 Normal: 2*64kB 2*128kB 0*256kB 0*512kB 0*1024kB 1*2048kB 0*4096kB 0*8192kB 0*16384kB 0*32768kB 0*65536kB 0*131072kB 0*262144kB 0*524288kB 0*1048576kB 0*2097152kB 0*4194304kB = 2432kB
76 total pagecache pages
Swap cache: add 0, delete 0, find 0/0
Free swap  = 0kB
Total swap = 0kB
Free swap:            0kB
Node memory in pages:
Node    0:  RAM:        8192, rsvd:     1327, shrd:        129, swpd:          0
Node    1:  RAM:           0, rsvd:        0, shrd:          0, swpd:          0
8192 pages of RAM
1327 reserved pages
129 pages shared
0 pages swap cached
Total of 45 pages in page table cache
8117 free buffer pages
Out of memory: kill process 2908 (sshd) score 379 or a child
Killed process 2909 (bash)

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/