Date: Tue, 31 Jul 2007 21:17:38 +0400 (MSD) From: Dmitry Morozovsky <marck@rinet.ru> To: Maksim Yevmenkin <maksim.yevmenkin@gmail.com> Cc: current@freebsd.org Subject: Re: possible showstopper: kbdmux hangs -current Message-ID: <20070731211230.B15561@woozle.rinet.ru> In-Reply-To: <bb4a86c70707311010h303fd8f5v69510157e08c5873@mail.gmail.com> References: <20070727141421.H42349@woozle.rinet.ru> <bb4a86c70707271018p45f3ccbap142bd6506c4dfc46@mail.gmail.com> <20070728173355.O42349@woozle.rinet.ru> <bb4a86c70707301518o5a6e3226j98e992cd4553e677@mail.gmail.com> <20070731123820.T29245@woozle.rinet.ru> <bb4a86c70707311010h303fd8f5v69510157e08c5873@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 31 Jul 2007, Maksim Yevmenkin wrote:
MY> > MY> > MY> > on some of my mobos -current hangs early (when starning init) if kbdmux is
MY> > MY> > MY> > included in kernel (both on i386 and amd64); this seems to be some race, as
MY> > MY> > MY> > hangs are not 100% reproducible. What info should I provide to debug?
MY> > MY> > MY>
MY> > MY> > MY> the usual suspects are
MY> > MY> > MY>
MY> > MY> > MY> 1) while (KBDMUX_CHECK_CHAR(kbd)) { ... } loop in kbdmux_kbd_event().
MY> > MY> > MY> could you please try to put some debug printf's into it and make sure
MY> > MY> > MY> it does not stuck there.
MY> > MY> >
MY> > MY> > errrm. Added two printfs with ppsratecheck - and, as usual, 'specialist
MY> > MY> > presense effect' is in place: no single hang since, both on stock GENERIC and
MY> > MY> > my own stripped down kernel so far (approx 20 reboots).
MY> > MY> >
MY> > MY> > Will try further.
MY> > MY>
MY> > MY> if you remove your debug code and recompile kbdmux(4), does your
MY> > MY> problem re-appear?
MY> >
MY> > Grrm. It does not (yet).
MY>
MY> ok. i have to ask.
MY>
MY> 1) are you using any non-standard compiler options?
MY>
MY> 2) how are you building new kernel? (or rather do you use NO_CLEAN or
MY> something like that?)
MY>
MY> kbdmux(4) has not been changed for awhile. while i completely accept
MY> possibility of bug in kbdmux(4), i find it somewhat strange that it
MY> does not manifest itself all the time.
1. CPUTYPE?= on i386 worlds, nothing significant (KERNCONF?=
MODULES_OVERRIDE?=) on amd64
2. make buildkernel; I do use NO_CLEAN most of the time, but in these strange
cases I blew out /usr/obj/sys completely to make sure there are no leftovers.
And remember: the problem did not manifest itself 100% from the beginning.
I'll continue to try to reproduce problem.
Thanks!
Sincerely,
D.Marck [DM5020, MCK-RIPE, DM3-RIPN]
------------------------------------------------------------------------
*** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck@rinet.ru ***
------------------------------------------------------------------------
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070731211230.B15561>