Date: Sun, 12 Feb 2006 16:52:58 +0100
From: Matthias Andree <matthias.andree@gmx.de>
To: axboe@suse.de, linux-kernel@vger.kernel.org
Subject: 2.6.15.4 ide-cd causes 'local unit communication failure' on NEC ND-4550A?
Message-ID: <20060212155258.GA8860@merlin.emma.line.org>
Mail-Followup-To: axboe@suse.de, linux-kernel@vger.kernel.org
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
User-Agent: Mutt/1.5.11
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2768
Lines: 71

Greetings,

catchy subject, here's the story:

SUSE Linux kernels 2.6.13-15.7,
                   2.6.13-15.8 (both SUSE Linux 10.0 i586) and
Vanilla    kernel  2.6.15.4

cause "local unit communication failure" when running "readcd -c2scan"
on my NEC ND-4550A CD/DVD drive (VIA KT8237, Athlon XP 2500+, 1 GB RAM)
after at most two seconds of reading.

I am aware of three ways to trigger the problem currently:

1. as little as switching to a different tab in GNOME-Terminal.

2. hald-addon-storage's poll with CDROM_MEDIA_CHANGED and
   CDROM_DRIVE_STATUS ioctls cause the same problem (hald-addon-storage
   doesn't do anything else except opening and closing the device and
   sleeping for two seconds).

3. running two readcd -c2scan dev=/dev/hdc commands on the same drive
   triggers the problem, too.  One might argue this isn't very sensible,
   but it shouldn't confuse the drive either.

The problem is invariant to hdparm -d0 (or -d1) /dev/hdc, and hdparm -u0
or -u1 make no difference either.

A Plextor PX-W4824TA (/dev/hdd) on the same cable as the NEC works like
a charm, and with FreeBSD 6-STABLE, both work like a charm with (3.).
(This is a dualboot machine, so it's the very same hardware.)

readcd uses commands like these:

Executing 'read_cd' command on Bus 1 Target 0, Lun 0 timeout 40s
CDB:  BE 00 00 00 02 7D 00 00 31 FA 00 00
ioctl(3, SG_IO, 0xbfa5b2ac)             = 0

In a second run -- I needed to capture the output ;-) -- I get this:

CDB:  BE 00 00 00 00 31 00 00 31 FA 00 00
status: 0x2 (CHECK CONDITION)
Sense Bytes: 70 00 04 00 00 00 00 0A 00 00 00 00 08 00 00 00
Sense Key: 0x4 Hardware Error, Segment 0
Sense Code: 0x08 Qual 0x00 (logical unit communication failure) Fru 0x0
Sense flags: Blk 0 (not valid)
resid: 123748
cmd finished after 6.408s timeout 40s
readcd: Success. read_cd: scsi sendcmd: no error

I'd been using "sudo" to run this with ruid == euid == 0 to rule out
command filtering issues. I hope that was the right approach.

What else do you need to debug this? Is there some low-level debugging
output I could enable to see what the kernel sends and gets to/from the
drive?

J?rg Schilling suspected (on cdwrite(#)other,debian!org - typos are
mine) this might be a DMA problem (but what does he know of Linux...
little compared to his Solaris-fu), but I think we can rule that out,
the problem persists after hdparm -d0.  J?rg also insisted I tell him if
I had a 40 or 80 wire cable, _after_ I'd posted that FreeBSD worked OK
on the same hardware.

-- 
Matthias Andree
-
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/