Date: Sun, 4 Oct 1998 17:52:16 -0700 From: "Morgan Davis" <mdavis@cts.com> To: <dg@root.com> Cc: <freebsd-stable@FreeBSD.ORG> Subject: RE: KERNEL PANICS in 2.2.7-RELEASE server Message-ID: <NCBBLNPFBLAJJOFDILMLGEKACAAA.mdavis@cts.com> In-Reply-To: <199810042045.NAA00224@implode.root.com>
next in thread | previous in thread | raw e-mail | index | archive | help
David: Below is a dump of the kernel symbols around that address per your request. Look like the offending function is _sosend. You may be right about mbuf clusters, as the kernel is simply GENERIC with all the non-essentials commented out. What kernel tweaks would you recommend to accomodate a high-bandwidth packet-passing system? It's an audio server that simply forwards encoded audio streams to potentially thousands of UDP and TCP client connections. It doesn't host users or do much else. I wonder if the tweaks we've used for some of our BSDI news servers would work. They use: maxusers 512 options "CHILD_MAX=512" options "OPEN_MAX=1024" options "NMBCLUSTERS=16384" options "TCBHASHSIZE=2048" options AHC_TAGENABLE --Morgan f011da70 t _sysctl___kern_somaxconn f011da98 t ___set_sysctl__kern_sym_sysctl___kern_somaxconn f011da9c T _socreate f011db8c T _sobind f011dbd8 T _solisten f011dc84 T _sofree f011dd50 T _soclose f011de70 T _soabort f011de98 T _soaccept f011df00 T _soconnect f011df88 T _soconnect2 f011dfd4 T _sodisconnect f011e040 T _sosend f011e700 T _soreceive f011f07c T _soshutdown f011f0b8 T _sorflush f011f184 T _sosetopt f011f44c T _sogetopt f011f690 T _sohasoutofband f011f6e0 F uipc_socket2.o f011f6e0 t _sysctl___kern_maxsockbuf f011f70c t ___set_sysctl__kern_sym_sysctl___kern_maxsockbuf > -----Original Message----- > From: root@implode.root.com [mailto:root@implode.root.com]On Behalf Of > David Greenman > Sent: Sunday, October 04, 1998 1:46 PM > To: Morgan Davis > Cc: freebsd-stable@FreeBSD.ORG > Subject: Re: KERNEL PANICS in 2.2.7-RELEASE server > > > >The machine is a Pentium II 266MHz with 256MB RAM, and pumps out 2.5 MB/s > >(that's mega_bytes_) via an Intel EtherExpress in 100BaseT mode. This > >server was just upgraded from 2.2.6-RELEASE to 2.2.7-RELEASE (including a > >newly built kernel based chiefly on GENERIC), and ever since it exhibits > >these intermittent panics: > > > >Fatal trap 12: page fault while in kernel mode > >fault virtual address = 0x18 > >fault code = supervisor write, page not present > >instruction pointer = 0x8:0xf011e3d1 > >stack pointer = 0x10:0xefbffe9c > >frame pointer = 0x10:0xefbffec0 > >code segment = base 0x0, limit 0xfffff, type 0x1b > > = DPL 0, pres 1, def32 1, gran 1 > >processor eflags = interrupt enabled, resume , IOPL = 0 > >current process = 501 (pnserver) > >interrupt mask = > >panic: page fault > > > >The panics seem to occur during periods of high activity after > running for a > >couple of hours. Otherwise, it appears to operate fine. Prior > to the Intel > >card, it was running an SMC Rev C (de driver) card that was even more > >sensitive and paniced frequently. I'm considering reverting > back to 2.2.6 > >unless there is a known issue with 2.2.7-RELEASE that has been resolved > >in -stable. Or maybe there is no some new kernel tweaks > required in 2.2.7 > >to avoid this? > > > >Any assistance is much appreciated. > > It would help a bunch if you could tell us what functions are around > the instruction pointer above (f011e3d1)...use nm /kernel | sort | more. > It kind of looks like you might be running out of mbuf clusters. You might > try increasing them and see if it helps. > > -DG > > David Greenman > Co-founder/Principal Architect, The FreeBSD Project > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?NCBBLNPFBLAJJOFDILMLGEKACAAA.mdavis>