RE: AACRAID driver broken in 2.6.22.x (and beyond?) [WAS: Re: 2.6.22.16 MD raid1 doesn't mark removed disk faulty, MD thread goes UN]

At which version of the kernel did the aacraid driver allegedly first go broken? At which version did it get fixed? (Since 1.1.5-2451 is older than latest represented on kernel.org)

How is the SATA disk'd arrayed on the aacraid controller? The controller is limited to generating 24 arrays and since /dev/sdac is the 29th target, it would appear we need more details on your array's topology inside the aacraid controller. If you are using the driver with aacraid.physical=1 and thus using the physical drives directly (in the case of a SATA disk, a SATr0.9 translation in the Firmware), this is not a supported configuration and was added only to enable limited experimentation. If there is a problem in that path in the driver, I will glad to fix it, but still unsupported.

You may need to acquire a diagnostic dump from the controller (Adaptec technical support can advise, it will depend on your application suite) and a report of any error recovery actions reported by the driver in the system log as initiated by the SCSI subsystem.

There are no changes in the I/O path for the aacraid driver. Due to the simplicity of the I/O path to the processor based controller, it is unlikely to be an issue in this path. There have been several changes in the driver to deal with error recovery actions initiated by the SCSI subsystem. One likely candidate was to extend the default SCSI layer timeout because it was shorter than the adapter's firmware timeout. You can check if this is the issue by manually increasing the timeout for the target(s) via sysfs. There were recent patches to deal with orphaned commands resulting from devices being taken offline by the SCSI layer. There has been changes in the driver to reset the controller should it go into a BlinkLED (Firmware Assert) state. The symptom also acts like a condition in the ol
 der drivers (pre 08/08/2006 on scsi-misc-2.6, showing up in 2.6.20.4) which did not reset the adapter when it entered the BlinkLED state and merely allowed the system to lock, but alas you are working with a driver with this reset fix in the version you report. A BlinkLED condition generally indicates a serious hardware problem or target incompatibility; and is generally rare as they are a result of corner case conditions within the Adapter Firmware. The diagnostic dump reported by the Adaptec utilities should be able to point to the fault you are experiencing if these appear to be the root causes.

Sincerely -- Mark Salyzyn

-----Original Message-----
From: Mike Snitzer [mailto:snitzer@gmail.com]
Sent: Tuesday, January 22, 2008 7:10 PM
To: linux-raid@vger.kernel.org; NeilBrown
Cc: linux-kernel@vger.kernel.org; K. Tanaka; AACRAID;
linux-scsi@vger.kernel.org
Subject: AACRAID driver broken in 2.6.22.x (and beyond?)
[WAS: Re: 2.6.22.16 MD raid1 doesn't mark removed disk
faulty, MD thread goes UN]

On Jan 22, 2008 12:29 AM, Mike Snitzer [off-list ref] wrote:

quoted

cc'ing Tanaka-san given his recent raid1 BUG report:
http://lkml.org/lkml/2008/1/14/515


On Jan 21, 2008 6:04 PM, Mike Snitzer [off-list ref] wrote:

quoted

Under 2.6.22.16, I physically pulled a SATA disk

(/dev/sdac, connected to

quoted

quoted

an aacraid controller) that was acting as the local raid1

member of

quoted

quoted

/dev/md30.

Linux MD didn't see an /dev/sdac1 error until I tried

forcing the issue by

quoted

quoted

doing a read (with dd) from /dev/md30:

....

quoted

The raid1d thread is locked at line 720 in raid1.c

(raid1d+2437); aka

quoted

freeze_array:

(gdb) l *0x0000000000002539
0x2539 is in raid1d (drivers/md/raid1.c:720).
715              * wait until barrier+nr_pending match nr_queued+2
716              */
717             spin_lock_irq(&conf->resync_lock);
718             conf->barrier++;
719             conf->nr_waiting++;
720             wait_event_lock_irq(conf->wait_barrier,
721

conf->barrier+conf->nr_pending ==

quoted

conf->nr_queued+2,
722                                 conf->resync_lock,
723

raid1_unplug(conf->mddev->queue));

quoted

724             spin_unlock_irq(&conf->resync_lock);

Given Tanaka-san's report against 2.6.23 and me hitting

what seems to

quoted

be the same deadlock in 2.6.22.16; it stands to reason this affects
raid1 in 2.6.24-rcX too.

Turns out that the aacraid driver in 2.6.22.x is HORRIBLY BROKEN (when
you pull a drive); it responds to MD's write requests with uptodate=1
(in raid1_end_write_request) for the drive that was pulled!  I've not
looked to see if aacraid has been fixed in newer kernels... are others
aware of any crucial aacraid fixes in 2.6.23.x or 2.6.24?

After the drive was physically pulled, and small periodic writes
continued to the associated MD device, the raid1 MD driver did _NOT_
detect the pulled drive's writes as having failed (verified this with
systemtap).  MD happily thought the write completed to both members
(so MD had no reason to mark the pulled drive "faulty"; or mark the
raid "degraded").

Installing an Adaptec-provided 1.1-5[2451] driver enabled raid1 to
work as expected.

That said, I now have a recipe for hitting the raid1 deadlock that
Tanaka first reported over a week ago.  I'm still surprised that all
of this chatter about that BUG hasn't drawn interest/scrutiny from
others!?

regards,
Mike