Date: Wed, 28 Apr 2004 22:09:54 +0200
From: Jens Axboe <axboe@suse.de>
To: Kenneth Johansson <ken@kenjo.org>
Cc: Moritz Muehlenhoff <jmm@informatik.uni-bremen.de>,
       linux-kernel@vger.kernel.org, cw@f00f.org
Subject: Re: [BUG] DVD writing in 2.6.6-rc2
Message-ID: <20040428200953.GA3470@suse.de>
References: <1083088772.2679.11.camel@tiger> <20040427183607.GA3011@suse.de> <8n23m1-g22.ln1@legolas.mmuehlenhoff.de> <20040428113056.GA2150@suse.de> <1083176956.2679.19.camel@tiger>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1083176956.2679.19.camel@tiger>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2045
Lines: 47

On Wed, Apr 28 2004, Kenneth Johansson wrote:
> On Wed, 2004-04-28 at 13:30, Jens Axboe wrote:
> > On Wed, Apr 28 2004, Moritz Muehlenhoff wrote:
> > > Jens Axboe wrote:
> > > >> I have a problem when using growisofs version 5.19.
> > > >> 
> > > >> The problem is that in the very end when gowisofs tries to flush the
> > > >> cache. When stracing the process I can see it sits in a call to poll
> > > >> that never returns. 
> > > >
> > > > I noted the same thing yesterday with cdrdao, so yours is not an
> > > > isolated incident. I'll debug it tomorrow.
> > > 
> > > I can confirm this bug for kernel 2.6.3-rc4 as well, so it's not too recent.
> > > I found the following on my console after quitting k3b ungracefully, maybe
> > 
> > This is really strange, I haven't been able to locate a kernel problem.
> > Looking at command traces, cdrdao issues two FLUSH_CACHE commands when
> > ending the writes. The last commands are:
> > 
> > 0x2a (last real write)
> > 0x35 (first flush cache) (takes about 10 seconds to complete)
> > 0x00 (test unit ready)
> > 0x51 (read disc info, length 4)
> > 0x35 (2nd flush cache) -> never completes
> > 
> > That last sync cache never generates an interrupt, so cdrdao gets stuck
> > forever waiting on it. I cannot even reproduce this with a test case
> 
> My problem was in growisofs itself. They use poll to wait instead of
> using nanosleep and totally confusing me. It did not help that strace do
> not print out the arguments until the syscall return so I never saw the
> input to poll. 

So it just helped me unearth a different problem :-). There certainly is
a bug here, it looks like it's hardware though (see the above
description). ide-cd just needs to have it's SYNC_CACHE retries
limited, then the kernel should at least recover.

-- 
Jens Axboe

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/