Re: 2.0.31p2 locks up
Dan Hollis (goemon@sasami.anime.net)
Thu, 7 Aug 1997 11:18:18 -0700 (PDT)
On Thu, 7 Aug 1997, Jon Lewis wrote:
> Strike the above :) While ksymoopsing an oops I just got on the above
> mentioned system today, I realized it's not running a pre-patch
> kernel...it's just 2.0.30 plus some fixes. And of course, after bragging
> about its stability, it spit this out this afternoon.
> >>EIP: 1460ac <tcp_recvmsg+170/40c>
> Trace: 150a7e <inet_recvmsg+72/88>
> Trace: 1369cb <sock_read+ab/c0>
> Trace: 1224ea <sys_read+8a/b0>
> Trace: 10a585 <system_call+55/80>
>
> Code: 1460ac <tcp_recvmsg+170/40c> movb 0xd(%eax),%al
> Code: 1460af <tcp_recvmsg+173/40c> testb $0x2,%al
> Code: 1460b1 <tcp_recvmsg+175/40c> je 1460b7 <tcp_recvmsg+17b/40c>
> Code: 1460b3 <tcp_recvmsg+177/40c> decl 0x10(%esp,1)
> Code: 1460b7 <tcp_recvmsg+17b/40c> movl 0x10(%esp,1),%edi
> Code: 1460bb <tcp_recvmsg+17f/40c> cmpl %edi,0x30(%ebx)
> Code: 1460be <tcp_recvmsg+182/40c> ja 90909018 <_EIP+90909018>
Ahh, this bug. I reported it several months ago, it seems more and more
people are duplicating it now. Basically it seems to be generated by
anything that turns over lots of connections (http server, proxy server,
etc).
I think someone needs to add some debugging code to pre-3 so that this
bug can get squashed.
-Dan