Date: Fri, 20 Sep 2013 10:02:53 -0400 From: John Baldwin <jhb@freebsd.org> To: Gleb Smirnoff <glebius@freebsd.org> Cc: freebsd-current@freebsd.org Subject: Re: ipmi patch for review Message-ID: <201309201002.53480.jhb@freebsd.org> In-Reply-To: <20130920054452.GB4574@glebius.int.ru> References: <20130917102110.GK4574@glebius.int.ru> <201309191504.46986.jhb@freebsd.org> <20130920054452.GB4574@glebius.int.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Friday, September 20, 2013 1:44:52 am Gleb Smirnoff wrote: > John, > > On Thu, Sep 19, 2013 at 03:04:46PM -0400, John Baldwin wrote: > J> > When system is writing a kernel core dump, it issues watchdog > J> > pat wdog_kern_pat(WD_LASTVAL). If ipmi is in action, it registers > J> > ipmi_wd_event() as event for watchdog. Thus ipmi_wd_event() is > J> > called in dumping context. > J> > > J> > The problem is that ipmi_wd_event() calls into ipmi_set_watchdog(), > J> > that calls into ipmi_alloc_request(), which uses M_WAITOK and > J> > thus sleeps. This is a smaller problem, since can be converted to > J> > M_NOWAIT. But ipmi_set_watchdog() then calls into > J> > ipmi_submit_driver_request(), which calls msleep() any time. > J> > > J> > The attached patch allows me to successfully write cores in > J> > presence of IPMI. > J> > J> Of course, the watchdog might go off during your dump. :) > > Yes, I understand that :( > > But, imho patch improves situation, although is ugly. Yes, I think a temporary workaround is fine for now. > J> The real fix is more complicated, which is that we should not use > J> a worker thread for at least SMIC and KCS. > > -- > Totus tuus, Glebius. > -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201309201002.53480.jhb>