| 1 |
Jeremy would like a side-by-side comparison... |
| 2 |
> and I wrote that |
| 3 |
> when running my stress-test on vanilla 2.6.6, the OOM-killer starts |
| 4 |
> killing things off 'correctly' and it doesn't seem to crash |
| 5 |
> -- at least I haven't run it into the ground yet. I suppose I can |
| 6 |
> try again. |
| 7 |
> |
| 8 |
> gds-2.6.7-r16 dies abruptly without ever calling the OOM-killer. |
| 9 |
> |
| 10 |
> > If that's the case, would you be willing to test out 2.6.7-rc1 and |
| 11 |
> > 2.6.7-rc2 and help narrow down when the lockup was introduced... |
| 12 |
> |
| 13 |
> I assume you're speaking of the vanilla sources? |
| 14 |
> I'll download and compile them.... |
| 15 |
|
| 16 |
|
| 17 |
Oh, yeah.. Fun stuff. |
| 18 |
|
| 19 |
make -j all |
| 20 |
|
| 21 |
start at 15:14 |
| 22 |
first OOM message at 15:15:48 |
| 23 |
first call trace at: 15:18:27 |
| 24 |
|
| 25 |
apache: page allocation failure. order:0, mode:0x1d2 |
| 26 |
Call trace: |
| 27 |
filemap_nopage |
| 28 |
do_no_page |
| 29 |
handle_mm_fault |
| 30 |
do_sparc64_fault |
| 31 |
sparc64_realfault_common |
| 32 |
|
| 33 |
(I can send the full logs of this if you're really interested) |
| 34 |
|
| 35 |
but the system kept trying to compile. load average was about 273. heh |
| 36 |
And sure, gcc reported lots of internal errors because of the OOM killer. |
| 37 |
|
| 38 |
I could still swap consoles around, and then decided to killall -9 make & |
| 39 |
cc1 |
| 40 |
the system fully recovered, and I was able to continue compiling with a bit |
| 41 |
nicer-on-the-system "make -j 5 all" |
| 42 |
|
| 43 |
So from that respect, I couldn't make 2.6.6 crash. |
| 44 |
Unlike gds-2.6.7-r16, which crashed hard. |
| 45 |
|
| 46 |
vanilla 267-rc[12] continues compiling... |
Archives