Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 3 Dec 2003 14:34:57 -0500
From:      Mathew Kanner <mat@cnd.mcgill.ca>
To:        Peter Hofer <peter.hofer2@liwest.at>
Cc:        freebsd-current@freebsd.org
Subject:   Re: Strange pcm/network problem
Message-ID:  <20031203193457.GM54011@cnd.mcgill.ca>
In-Reply-To: <3FCE33C2.50200@liwest.at>
References:  <3FCE33C2.50200@liwest.at>

next in thread | previous in thread | raw e-mail | index | archive | help
	Peter,
	I'm afraid I haven't a clue why this happens to you but please
see my notes below.

On Dec 03, Peter Hofer wrote:
> since I installed 5.2-BETA, I've got a very strange problem. Always
> when I'm listening to some music and start a network transfer, both
> the music and the transfer stop after some time. Nothing is being
> received or sent any more, and the following message appears a few
> times:
> 
> bfe0: watchdog timeout -- resetting
> 
> This goes on until I exit xmms. When I exit xmms, after a few seconds
> the network works again. There's no problem in receiving and sending
> data then.
> 
> I was able to reproduce this with mpg123, it just stops playing and
> exits when data is being sent over the network.

	And dmesg then says "pcm channel dead" ?

> 
> The problem also persists when I'm using my rl nic instead of the
> onboard bfe chip.
> 
> I'm using sources from yesterday (2003-11-02) now and this problem
> still occurs, no matter if sound support is loaded with kldload or
> compiled into the kernel.

	No need to try compiling PCM in the kernel, klds should work
exactly the same.

	I notice that your network cards, sound card and sio use IRQ
4, I also note that your aren't using acpi, what happens when you
enable it?

	A brief look at the ich code reveals that it's flagged as
mpsafe when it isn't, but since you are UP, I don't see this being the
cause.  I also vaguely remember that another OS had a note about
certain corner cases missing interrupts but I don't think that should
affect other irq handlers either.  I really am stumped.

	If we can't figure this out, you always do a binary search to
find the commit that caused it, though since I wouldn't likely do this
myself, I wouldn't expect you to.

	Input from others warmly received :)

	--mat

-- 
	Any idiot can face a crisis; it is this day-to-day living
	that wears you out.
			- Chekhov



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031203193457.GM54011>