Date: Tue, 06 Mar 2001 13:17:28 -0800 (PST) From: John Baldwin <jhb@FreeBSD.org> To: David Malone <dwmalone@maths.tcd.ie> Cc: cvs-all@FreeBSD.org, cvs-committers@FreeBSD.org, Dag-Erling Smorgrav <des@ofug.org> Subject: Re: cvs commit: src/sys/i386/isa apic_vector.s icu_vector.s Message-ID: <XFMail.010306131728.jhb@FreeBSD.org> In-Reply-To: <20010306105246.A28173@walton.maths.tcd.ie>
next in thread | previous in thread | raw e-mail | index | archive | help
On 06-Mar-01 David Malone wrote: > On Mon, Mar 05, 2001 at 05:41:28PM +0100, Dag-Erling Smorgrav wrote: >> David Malone <dwmalone@maths.tcd.ie> writes: >> > On Mon, Mar 05, 2001 at 07:48:43AM -0800, John Baldwin wrote: >> > > Heavy load could generate enough interrupts to overflow the stack and >> > > tra >> > > pcb, so this very well might fix that. >> > I think I may have been seeing this. [...] >> >> Seems this commit fixed it, I've been doing port builds on a fresh >> kernel for the past two and a half hours with no hangs. > > This seems to have slightly improved the situation, now my tail > test program finishes, but a buildworld is still too much for the > machine. > > It prints a page fault in kernel mode message, with the following > details: > > fault virtual address 0xfffff037 > instruction pointer 0x8:0xfffff037 > stack pointer 0x10:0xc8346318 > frame pointer 0x10:0xc8346e8c > code base 0x0 limit 0xffff type 0x11 DPL 0 pres 1 def32 1 gran 1 > > before it prints a DDB prompt, there is another page fault, which is > identical apart from: > > instruction pointer 0x8:0xc0262fe8 > stack pointer 0x10:0xc8346c84 > frame pointer 0x10:0xc8346c88 > > then I get a DDB prompt. Asking for a stack trace just produces > another page fault. Yeah, cause the stack is hosed it looks like. Hrmmmm. Add the following options to your kernel: options KTR options KTR_EXTEND options KTR_COMPILE=0x1208 Then before you do a buildworld, run the followign command as root: # sysctl -w debug.ktr.mask=0x1200 Then run the buildworld and crash teh machine. :-P The type 'show ktr' to get a list of the various interrupts and process changes that occured up until things blew up. If you have a serial console and can knab that output that would be great. > David. -- John Baldwin <jhb@FreeBSD.org> -- http://www.FreeBSD.org/~jhb/ PGP Key: http://www.baldwin.cx/~john/pgpkey.asc "Power Users Use the Power to Serve!" - http://www.FreeBSD.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.010306131728.jhb>