Under certain circumstances (it appears, high I/O load), PPP userland/client unpredictably throws a core. That didn't happen with 3.2-RELEASE and, as I remember, neither with 3.3-RC. (I have updated the "Modem"'s firmware in the meantime, but that doesn't appear to be the real reason, since it works under 3.2-REL like it did before.) Fix: Unknown. Brian has been informed, but I lack of the ability to supply the data he wants. How-To-Repeat: Under certain circumstances (empty cache, empty proxy), browsing http://www.jpc.de/ has seemed to be a reliable method to produce a PPP's core dump, while there were several I/O-intensive processes in the background (load above 1.8). It's a site with very many objects are to be loaded at the same time.
On Tue, Oct 05, 1999 at 07:37:02PM +0200, Klaus-Juergen Wolf wrote: > [...] > Under certain circumstances (it appears, high I/O load), PPP > userland/client unpredictably throws a core. That didn't happen > with 3.2-RELEASE and, as I remember, neither with 3.3-RC. (I have > updated the "Modem"'s firmware in the meantime, but that doesn't > appear to be the real reason, since it works under 3.2-REL like > it did before.) > [...] > > Unknown. Brian has been informed, but I lack of the ability to supply > the data he wants. > Yeah, this happened to me too: Oct 5 10:28:08 relay /kernel: pid 23005 (ppp), uid 0: exited on signal 10 (core dumped) Oct 5 20:07:28 relay /kernel: pid 77580 (ppp), uid 0: exited on signal 10 (core dumped) I have compiled `ppp' with debug symbols, and will send a backtrace on the next core. Anything else, Brian? -- Ruslan Ermilov Sysadmin and DBA of the ru@ucb.crimea.ua United Commercial Bank, ru@FreeBSD.org FreeBSD committer, +380.652.247.647 Simferopol, Ukraine http://www.FreeBSD.org The Power To Serve http://www.oracle.com Enabling The Information Age
[.....] > Yeah, this happened to me too: > > Oct 5 10:28:08 relay /kernel: pid 23005 (ppp), uid 0: exited on signal 10 (core dumped) > Oct 5 20:07:28 relay /kernel: pid 77580 (ppp), uid 0: exited on signal 10 (core dumped) > > I have compiled `ppp' with debug symbols, and will send a backtrace on the > next core. > > Anything else, Brian? Some patches ? ;^1 Seriously, I believe there's a bug in the way VJ packets are handled where ppp ends up scribbling over the return address on the stack. Once the deed has happened, there's very little information left.... What I think I really need is a way of actually reproducing the problem. It's never happened to me, but it's been happening to people for years. I think the only way to catch something like this is to get the compiler to put the function return address in read-only memory so that a stack-scribble will produce a core when it happens rather than after the fact.... Do you know if gcc is capable of doing this ? Do you know of any better ways of tackling the problem ? > -- > Ruslan Ermilov Sysadmin and DBA of the > ru@ucb.crimea.ua United Commercial Bank, > ru@FreeBSD.org FreeBSD committer, > +380.652.247.647 Simferopol, Ukraine > > http://www.FreeBSD.org The Power To Serve > http://www.oracle.com Enabling The Information Age Cheers. -- Brian <brian@Awfulhak.org> <brian@FreeBSD.org> <http://www.Awfulhak.org> <brian@OpenBSD.org> Don't _EVER_ lose your sense of humour ! <brian@FreeBSD.org.uk>
State Changed From-To: open->closed Fixed in -current. I'll MFC as soon as Klaus-Juergen confirms it's working ok.
State Changed From-To: closed->open Doh! Wrong PR ! I was thinking ``that isn't ru@'s name'' !
State Changed From-To: open->closed Fixed in -current with *lots* of help from ru. Will MFC soon.