Date: Sat, 07 Nov 1998 12:42:13 +0800 From: Peter Wemm <peter@netplex.com.au> To: Bill Paul <wpaul@skynet.ctr.columbia.edu> Cc: phk@critter.freebsd.dk (Poul-Henning Kamp), current@FreeBSD.ORG Subject: Re: Grrr... calcru: negative time blah blah blah Message-ID: <199811070442.MAA18044@spinner.netplex.com.au> In-Reply-To: Your message of "Fri, 06 Nov 1998 14:05:07 EST." <199811061905.OAA15554@skynet.ctr.columbia.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
Bill Paul wrote: [..] > Actually, looking at the patch closer, I realize now that none of the > warning messages actually triggered. However each time, the calcru > messages did appear and the system console response became sluggish. > > I'm starting to think the problem in this case is an interrupt storm, > but I'm not sure how to debug it. If I set up a second system to do > a remote gdb of the first one, can I single step through things like > interrupt handlers without Weird Things (tm) happening? Just a thought that might be worth checking into.. Is the kstack growing down into struct pstats, the sigacts, and perhaps pcb? This would be highly dependent on interrupt handlers, machine load (amount of nesting) etc and could explain why it hits some more than others. I've talked with a couple of people about moving the pcb, sigacts, pstats etc to the top of the two upages and have the kstack grow down from underneath it, and possibly allow for an unmapped page below the first UPAGE as a diagnostic tool (that will force a double panic on a kstack overflow). The top of stack can be tested on full stack unwind and return to user mode if we need to. Cheers, -Peter To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199811070442.MAA18044>