Hello,
Can you compile without IDE TCQ and tell whats the difference?
Regards,
--
Bartlomiej
On Fri, 23 May 2003, Julien Oster wrote:
> Hello,
>
> I already mentioned it in another thread, but here again, more
> complete.
>
> Booting my workstation with 2.5.69-mm8 works, but I get the following
> message many times per second:
>
> May 22 23:34:01 frodo kernel: ide_dmaq_intr: stat=42, not expected
> May 22 23:34:01 frodo kernel: ide_dmaq_intr: stat=40, not expected
> May 22 23:34:01 frodo last message repeated 34 times
>
> It's mostly stat=40, sometimes stat=42. Look at the time: it's really
> very often. (more often than 34 times, there are other normal messages
> in between)
>
> The harddisks are attached on a Promise PDC20276 onboard RAID
> controller, but it's only used as an IDE controller. However, I have
> Linux RAID Partitions on the disks (all mounted filesystems are). Most
> are RAID 1, one is RAID 0, but the latter isn't accessed very often.
>
> Even more interesting: after a while, at least /var/log/kern.log
> (where the kernel messages are logged to) gets jammed. Here's
> something pasted right away from "less /var/log/kern.log":
>
> May 22 23:34:32 frodo kernel: ide_dmaq_intr: stat=40, not expected
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> ^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@
> [...]
> <A0><81><CA>^Y^A<9A>|!#n:<C8><F6><F5>^E<AF><EE><B5>@^D<A5>]^L<90><C0><98>=f<F3>A
> `<C7>^Y^F<A0>%H;4<9A>^L<E9><FA><BE><AD><DD><AA>~<B5>^U<FB>[^GB<83><AC>^@w<B5>eMa
> y 22 23:41:23 frodo kernel: klogd 1.4.1#11, log source = /proc/kmsg started.
>
> I cutted it down to a few lines, however the binary garbage is quite a
> lot. Obviously, something got jammed for good, because as you can see,
> the next readable line is from the (fine running) 2.4.21-rc2 kernel,
> which I booted immediately afterwards.
>
> Note however, that I didn't do a clean shutdown, since I forgot to
> include my input devices into the kernel (ouch). I just pressed
> reset. The fsck done by the working 2.4.21-rc2 kernel afterwards also
> ate up a configuration file from tomcat (unused inode, cleaned).
>
> All mirrored arrays were reconstructing right after booting
> 2.5.69-mm8, since they weren't clean before. I pressed reset before
> the reconstruction got finished (I had no input device and the
> ide_dmaq_intr-message actually scared me).
>
> So, taking the unclean shutdown and the reconstructing RAID into
> account, the jammed files may not be a cause of the strange message
> the kernel gives me.
>
> I included my .config and an "lspci -v" output with this mail. Please
> note that the latter one was created with my currently running
> 2.4.21-rc2 kernel, but that shouldn't matter I believe.
>
> Regards,
> Julien
Bartlomiej Zolnierkiewicz <[email protected]> writes:
> Hello,
Hello Bartlomiej,
> Can you compile without IDE TCQ and tell whats the difference?
Uh, well.
The message disappeared. However, since keyboard and mouse still
didn't work (although all input devices are compiled in - did I miss
something?), I had to press reset again.
Now my system seems quite fucked up. (or, better: "fscked up", since
the problems appeared there...)
After rebooting (again my stable 2.4.21-rc1 kernel), fsck ate a lot of
files on the root partition, all with "unused
inode... CLEARED". Strangely, ONLY on the root filesystem. All other
filesystems (all on md devices, like the root filesystem) are perfect.
I don't know if that's still an issue of IDE TCQ, but I think I'll
quit trying it out, since I already lost X right now and have to
restore quite a few things.
Well, of course I have backups, I wouldn't install a development
kernel without expecting things that are even much worse, but all
those recompile, reboot, retry, reset and restore backup cycles are
currently a bit too time consuming :)
However, if you need additional information to track the TCQ-problem
down, I see what I can give.
Regards,
Julien
Just so you know, 0x40 status indicates "queued command received, not ready
for service yet". However, some drives we've tested internally will stay
0x40 status instead of switching back to 0x50 status once they've serviced
the last queued tag.
They will go 0x50 following a successful non-queued command however.
--eric
-----Original Message-----
From: Julien Oster [mailto:[email protected]]
Sent: Thursday, May 22, 2003 5:26 PM
To: Bartlomiej Zolnierkiewicz
Cc: [email protected]
Subject: Re: [2.5.69-mm8] ide_dmaq_intr: stat=40, not expected
Bartlomiej Zolnierkiewicz <[email protected]> writes:
> Hello,
Hello Bartlomiej,
> Can you compile without IDE TCQ and tell whats the difference?
Uh, well.
The message disappeared. However, since keyboard and mouse still
didn't work (although all input devices are compiled in - did I miss
something?), I had to press reset again.
Now my system seems quite fucked up. (or, better: "fscked up", since
the problems appeared there...)
After rebooting (again my stable 2.4.21-rc1 kernel), fsck ate a lot of
files on the root partition, all with "unused
inode... CLEARED". Strangely, ONLY on the root filesystem. All other
filesystems (all on md devices, like the root filesystem) are perfect.
I don't know if that's still an issue of IDE TCQ, but I think I'll
quit trying it out, since I already lost X right now and have to
restore quite a few things.
Well, of course I have backups, I wouldn't install a development
kernel without expecting things that are even much worse, but all
those recompile, reboot, retry, reset and restore backup cycles are
currently a bit too time consuming :)
However, if you need additional information to track the TCQ-problem
down, I see what I can give.
Regards,
Julien