Bug 7507 - pata_it821x: system freezes
Summary: pata_it821x: system freezes
Status: REJECTED UNREPRODUCIBLE
Alias: None
Product: IO/Storage
Classification: Unclassified
Component: Serial ATA (show other bugs)
Hardware: i386 Linux
: P2 normal
Assignee: Alan
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2006-11-12 23:47 UTC by Martin von Gagern
Modified: 2009-03-17 09:03 UTC (History)
6 users (show)

See Also:
Kernel Version: 2.6.20-rc5
Subsystem:
Regression: ---
Bisected commit-id:


Attachments
manual screen copy after modprobe pata_it821x noraid=1 (1.03 KB, text/plain)
2006-11-12 23:48 UTC, Martin von Gagern
Details
kernel messages from 2.6.22 for boot and modprobe pata_it821x noraid=1 (20.49 KB, text/plain)
2007-07-09 06:00 UTC, Martin von Gagern
Details
console log of 2.6.23-rc2 for insmod pata_it821x.ko noraid=1 (1.12 KB, text/plain)
2007-08-09 04:40 UTC, Martin von Gagern
Details
dmesg after booting with "init=/bin/bash" and pata_it821x (21.21 KB, text/plain)
2007-09-04 06:38 UTC, Dmitry Ivanov
Details
2.6.23-rc6 with libata.pata_dma=0 for pata_it821x (1.74 KB, text/plain)
2007-09-13 01:31 UTC, Martin von Gagern
Details
2.6.23-rc6 with debug output added to pata_it821x (52.64 KB, text/plain)
2007-09-13 03:53 UTC, Martin von Gagern
Details

Description Martin von Gagern 2006-11-12 23:47:39 UTC
Most recent kernel where this bug did not occur: none
Distribution: Gentoo (with sys-kernel/vanilla-sources-2.6.19_rc3)
Hardware Environment: ASUS P5GDC-V MB
                      root FS on SATA HDD
                      IT8212 onboard in IDE (not RAID) mode
                      2 x FUJITSU MPG3409AT E as masters for this
Software Environment: Grub, Bash

Problem Description:
"modprobe pata_it821x noraid=1" freezes system.
Note that I have trouble with the old it821x driver as well, bug #7506. The two
issues might be related.

Steps to reproduce:
1. boot with "init=/bin/bash"
2. mount / -o remount,rw
3. modprobe pata_it821x noraid=1
4. take pen and paper to write down log messages
5. reboot
Comment 1 Martin von Gagern 2006-11-12 23:48:54 UTC
Created attachment 9481 [details]
manual screen copy after modprobe pata_it821x noraid=1

I wrote down the messages on a sheet of paper and just typed them into a file.
There might be minor errors, but basically this should be an (almost) exact
screen dump.
Comment 2 Martin von Gagern 2007-01-22 16:54:29 UTC
Still freezes for 2.6.20-rc5
Comment 3 Natalie Protasevich 2007-07-08 13:18:10 UTC
Martin, have you tried with 2.6.22-rc7 or current git?
Thanks.
Comment 4 Martin von Gagern 2007-07-09 06:00:41 UTC
Created attachment 11978 [details]
kernel messages from 2.6.22 for boot and modprobe pata_it821x noraid=1

(In reply to comment #3)
> Martin, have you tried with 2.6.22-rc7 or current git?

Just tried 2.6.22, problem still exists. Attaching a dmesg and console copy.
Timestamps 161.051733 through 161.714039 are printed to console directly after
the call to "modprobe pata_it821x noraid=1", with the cursor stying in the last line just after the colon. Using a serial console allowed me to copy the output this time.
Comment 5 Tejun Heo 2007-08-08 20:15:10 UTC
pata_it821x and it821x received related updates recently.  Can you please give a shot at 2.6.23-rc2?

Thanks.
Comment 6 Martin von Gagern 2007-08-09 04:40:06 UTC
Created attachment 12332 [details]
console log of 2.6.23-rc2 for insmod pata_it821x.ko noraid=1

There are some changes in my setup for this run. The kernel is pretty much a "make defconfig". I configured an initramfs containing a busybox, so I could make my PIIX drivers modules and run these tests without loading them.

(In reply to comment #5)
> Can you please give a shot at 2.6.23-rc2?

Done that, but the output sadly looks pretty much the same, except for some changed numbers due to the missing driver for my PIIX SATA controller. Not even the warning message I got in bug 7506 comment 9 now.
Comment 7 Dmitry Ivanov 2007-09-04 06:36:15 UTC
It's very likely that I ran into this bug.

My hardware is similar: ASUS P5WD2 motherboard + onboard ITE8211F controller.
Just tested 2.6.23-rc5 and the system locked up at boot. After booting with "init=/bin/bash" I was able to grap dmesg (attached below). The system freezed after ~30 seconds but that was enough to save dmesg.
Comment 8 Dmitry Ivanov 2007-09-04 06:38:31 UTC
Created attachment 12693 [details]
dmesg after booting with "init=/bin/bash" and pata_it821x
Comment 9 Martin von Gagern 2007-09-04 06:51:59 UTC
(In reply to comment #7)
> It's very likely that I ran into this bug.

Comparing your log to mine I see it recognizes the disk partitions and even mounts one file system for you. This is a lot farther than I ever got.

Even more interesting for me might be the indication where mine stopped: while recognizing the partitions. My disks were part of a software raid once, and probably haven't been repartitioned since then, so the partition table won't work well for a single disk. I guess I'll repartition them using the old it821x driver and see whether this fixes the problem for pata_it821x. I'll let you know the outcome once I can reboot.
Comment 10 Martin von Gagern 2007-09-04 09:30:07 UTC
(In reply to comment #9)
> I guess I'll repartition them using the old it821x driver and see whether
> this fixes the problem for pata_it821x.

No luck, sadly.
Comment 11 Dmitry Ivanov 2007-09-04 10:13:18 UTC
(In reply to comment #9)

> Comparing your log to mine I see it recognizes the disk partitions and even
> mounts one file system for you. This is a lot farther than I ever got.

Actually, no. The hard drive is connected to the onboard Intel ICH7 family IDE controller which is handled by the ata_piix driver - see the lines 47.502632 - 47.905235. It works fine.

My DVD drive is attached to the ITE8211F controller (lines 48.233996 - 48.930191). When I enable the controller in BIOS _and_ attach a CD or DVD drive to it I get lockups (note: even without mounting CD/DVD disks). With no devices attached to it the system works fine.

Unfortunately Ubuntu is shipped with pata_it821x enabled by default so I get all sorts of headaches.
Comment 12 Dmitry Ivanov 2007-09-12 01:48:59 UTC
Good news!

I just tried 2.6.23-rc6 and it works. I guess the patch from Jeff Norden (found in 'git log -p drivers/ata/pata_it821x.c') fixed it. Also there is a thread at https://bugzilla.redhat.com/show_bug.cgi?id=242229 where all debugging goes.

Martin, could you test 2.6.23-rc6?
Comment 13 Martin von Gagern 2007-09-12 10:58:32 UTC
(In reply to comment #12)
> Martin, could you test 2.6.23-rc6?

Tried that, same result than before. :(

> https://bugzilla.redhat.com/show_bug.cgi?id=242229

Looks pretty FC-specific most of the time. I'll try this pata_dma param patch tomorrow if I find the time, to see whether DMA is the problem for me here as well. I would think so, as bug #7506 seems to be related to DMA as well.
Comment 14 Martin von Gagern 2007-09-13 01:31:01 UTC
Created attachment 12811 [details]
2.6.23-rc6 with libata.pata_dma=0 for pata_it821x

OK, it definitely is a DMA issue, although not fixed in vanilla rc6.
I applied linux-2.6-libata-pata_dma-param.patch to my kernel and booted with libata.pata_dma=0. With this setup I could load pata_it821x successfully, console log attached.
Comment 15 Martin von Gagern 2007-09-13 03:53:28 UTC
Created attachment 12816 [details]
2.6.23-rc6 with debug output added to pata_it821x

I read some more of that Red Hat bugzilla and added some debugging code to the module. See https://bugzilla.redhat.com/show_bug.cgi?id=242229#c47 for my post of this log and the following comment for the source modifications I applied.
What can be seen from this log here is that it821x_passthru_bmdma_start gets called and returns, then the system hangs for 30 seconds, then the freeze routine gets called, issues iowrite8, and then the system hangs for good.
Hopefully more experienced people can see even more. Tell me when any added debugging code could give you further insights here.
Comment 16 Natalie Protasevich 2008-04-05 02:19:30 UTC
Any updates on this problem please.
Thanks.
Comment 17 Martin von Gagern 2008-06-20 11:40:03 UTC
As mentioned in bug 7506 comment 15, I figured out that one of the two drives I was using in this setup was probably defective. So while at last test (around 2.6.25.0 iirc) the system would still crash when the driver was loaded with this defective device installed, this might say nothing about working hardware.
Sorry for not performing tests with single drives before.

Note You need to log in before you can comment on or make changes to this bug.