Date: Wed, 23 Aug 2006 20:11:08 +0900 From: Pyun YongHyeon <pyunyh@gmail.com> To: Gleb Smirnoff <glebius@FreeBSD.org> Cc: freebsd-current@FreeBSD.org Subject: Re: call for bge(4) testers Message-ID: <20060823111108.GK17902@cdnetworks.co.kr> In-Reply-To: <20060823110619.GL96644@cell.sick.ru> References: <20060822042023.GC12848@cdnetworks.co.kr> <20060823093741.GF96644@FreeBSD.org> <20060823095504.GI17902@cdnetworks.co.kr> <20060823100420.GG96644@cell.sick.ru> <20060823105118.GJ17902@cdnetworks.co.kr> <20060823110619.GL96644@cell.sick.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Aug 23, 2006 at 03:06:19PM +0400, Gleb Smirnoff wrote: > On Wed, Aug 23, 2006 at 07:51:18PM +0900, Pyun YongHyeon wrote: > P> > P> Its sole purpose is to reinitialize hardware for real watchdog > P> > P> timeouts. It's not fix for general watchdog timeouts. As I said other > P> > P> mails, the fake watchdog timeout(losing Tx interrupts) for hardwares > P> > P> with Tx interrupt moderation capability could be normal thing. So I > P> > P> just want to know bge(4) also has the same feature(bug). > P> > > P> > According to several emails about em(4) fake watchdog timeouts, the > P> > problem can be fixed by setting debug.mpsafenet=0. This makes me think > P> > that the problem isn't caused by TX interrupt moderation, but some race > P> > in the kernel. Really, if_slowtimo() doesn't acquire driver lock before > P> > checking and modifying the if_timer field. > P> > > P> > P> Hmm... I didn't say the problem was caused by TX interrupt moderation. > P> I can't sure but I'm under the impression it has *two* different issues. > P> If you think fake watchdog timeout fix is not adequate one please > P> let me know. I'll backout the change if you want. > > I don't think you should backout it until we find a solution. > However, I'd ask you don't MFC it. > ok. Lets find out real cause of bug. > P> > Afaik, NIC drivers that can do interrupt moderation should set a timer > P> > to a sane value, based on interrupt moderation settings, so that the > P> > watchdog won't be ever called fakely. > P> > P> Yes. Normally it should. But I saw the issues on Marvell Yukon too. > > Does Marvell Yukon have interrupt moderation? > Yes. Check archives for sk(4) watchdog. -- Regards, Pyun YongHyeon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060823111108.GK17902>