Some more data from our BSD/OS 4.2 box and the latest crash after 5 days of
uninterrupted operation.
I've instrumented counters on the box to count fragments on the two interfaces it has,
here's the outcome:
su-2.02# ipfstat -anhi -M bsdcore.4 -N bsd.4
pos=0x0 kmemcpy:read: Undefined error: 0
pos=0x0 kmemcpy:read: Undefined error: 0
pos=0x0 kmemcpy:read: Undefined error: 0
pos=0x0 kmemcpy:read: Undefined error: 0
0 0 @1 count in quick on exp0 proto icmp from any to any with frag
0 0 @2 count in quick on exp1 proto icmp from any to any with frag
14935 17441044 @3 count in quick on exp0 from any to any with frag
9 4143 @4 count in quick on exp1 from any to any with frag
It looks as though we've had a bunch of fragments incoming (mostly on exp0 which is
towards the Internet).
The fragments weren't icmp though. This would then point to this being a more general
fragment related issue ??
Alex
Alexander Latour-Henner wrote:
>
> We get the same type of panic on BSD/OS 4.2 as well (malloc after nat_new). The
> setup is quite similar to Greg's, i.e. only BIMAP for around 100 hosts. Our box
> isn't as regular though, it can stay up for a week or crash three times the same day.
> As BSD/OS 4.2 doesn't implement device polling, my impression is that this feature
> ought at most to be a contributing factor rather than the root cause of our problem.
>
> Alex
begin:vcard
n:Latour-Henner;Alexander
tel;fax:+46 8 180029
tel;work:+46 70 5954466
x-mozilla-html:FALSE
url:www.servicefactory.com
org:Service Factory AB
version:2.1
email;internet:[EMAIL PROTECTED]
adr;quoted-printable:;;=C5rsta=E4ngsv=E4gen 17, 6tr;Stockholm;;SE-117 43;Sweden
fn:Alex
end:vcard
