[EMAIL PROTECTED] wrote:
>
> /usr/local/apache2.0.42/corefiles/httpd.core.2
>
> #0 bndm (n=0x281f7351 "<!--#", nl=5,
> h=0x28249000 <Address 0x28249000 out of bounds>, hl=8533, t=0x80e28f4)
> at mod_include.c:299
> #1 0x281f1d50 in find_start_sequence (dptr=0x8144038, ctx=0x813f728,
> bb=0x814d8c0, do_cleanup=0xbfbff5ac) at mod_include.c:480
Looks like mod_include is an innocent victim this time.
[gregames@daedalus apache]$ ls -l corefiles/httpd.core.2
-rw-r--r-- 1 nobody wheel 1736704 Sep 25 09:43 corefiles/httpd.core.2
[gregames@daedalus apache]$ uptime
9:57AM up 1 day, 6 mins, 5 users, load averages: 0.90, 3.09, 3.08
The dump happened shortly before daedalus was rebooted yesterday.
(gdb) p r->filename
$6 = 0x814ead8 "/www/httpd.apache.org/index.html"
not exactly an unusual url.
(gdb) dump_brigade bb
dump of brigade 0x814d8c0
| type (address) | length | data addr | contents | rc
--------------------------------------------------------------------------------
0 | MMAP (0x08144038) | 8533 | 0x08144248 | [
Cannot access memory at address 0x28249000.
(gdb) dump_bucket bb->list.next
bucket=MMAP (0x08144038) length=8533 data=0x08144248
contents=[Cannot access memory at address 0x28249000.
(gdb) dump_bucket bb->list.next.link.next
bucket=EOS (0x081441f0) length=0 data=0x00000000
contents= rc=n/a
[gregames@daedalus apache]$ ls -l /www/httpd.apache.org/index.html
-rw-rw-r-- 1 striker httpd 8533 Sep 24 15:55 /www/httpd.apache.org/index.html
So we had an MMAP bucket with a bad virtual address and a good length. In case
you don't follow the infrastructure list, Brian B posted a bunch of kernel
messages including a number of vm faults that affected httpd. Nothing in the
/www directory was accessable for a while.
Greg
