svn commit: r218347 - stable/8/sys/dev/ata/chipsets - still fails

[Bob Willcox]

On Sun, Mar 20, 2011 at 10:06:08AM -0700, Charlie Kester wrote:
> On Sun 20 Mar 2011 at 09:30:20 PDT Bob Willcox wrote:
> >On Sun, Mar 20, 2011 at 04:18:02PM +0200, Alexander Motin wrote:
> >> Bob Willcox wrote:
> >> > On Thu, Mar 17, 2011 at 01:41:30AM +0200, Alexander Motin wrote:
> >> >> On 16.03.2011 15:51, Bob Willcox wrote:
> >> >>> This change has broken SATA disk support on my Intel Atom D525 ITX system. By
> >> >>> reverting this change 8.2-STABLE works again on this system.
> >> >>>
> >> >>> My 'uname -a' output is:
> >> >>>
> >> >>> FreeBSD maul.immure.com 8.2-STABLE FreeBSD 8.2-STABLE #6: Wed Mar 16 08:15:43 CDT 2011     bob at maul.immure.com:/usr/obj/usr/src/sys/MAUL  amd64
> >> >>>
> >> >>> When booting the system I get tons of these messages:
> >> >>>
> >> >>> Mar 16 07:38:19 maul kernel: ad4: WARNING - SETFEATURES ENABLE RCACHE requeued due to channel reset
> >> >>> Mar 16 07:38:19 maul kernel: ad4: interrupt on idle channel ignored
> >> >> As I can see, it means that channel has some active request, but it is 
> >> >> in IDLE state. It is strange, but I won't be surprised much if it is the 
> >> >> result of some locking problem in ata(4) in non-CAM mode.
> >> >>
> >> >>> repeated over and over, and then lots of these:
> >> >>>
> >> >>> Mar 16 07:38:21 maul kernel: ad4: WARNING - READ_DMA48 requeued due to channel reset LBA=617964479
> >> >>> Mar 16 07:38:21 maul kernel: ata2: FAILURE - already active DMA on this device
> >> >>> Mar 16 07:38:21 maul kernel: ata2: setting up DMA failed
> >> >>>
> >> >>> for different LBA values.
> >> >>>
> >> >>> As one might expect, I then start seeing I/O errors on the disk and programs
> >> >>> failing
> >> >>>
> >> >>> I've attached the 'pciconf -lv' output.
> >> >> Send me please full verbose log, if you can save it. I am especially 
> >> >> interested in place around first errors.
> >> >>
> >> >> You may try to build kernel with `options ATA_CAM` to see if it helps. 
> >> >> I've mostly tested this patch in that mode.
> >> > 
> >> > I tried setting `options ATA_CAM` but that didn't fix the problem. I still got
> >> > continuous ATA error messages spewed out while probing the ATA devices. I'm
> >> > not able to capture the verbose output as the system never successfully boots
> >> > and I don't have a serial console attached to the system.
> >> 
> >> It's strange. Can you at least show errors you received in that case? I
> >> suppose they should be different from the original.
> >
> >They were similar, though this time there was mention of CAM in them (which I
> >assumed was the result of using the CAM interface). The trouble is they roll
> >by quite fast and I'm unable to stop them from scrolling.
> >
> >If this wasn't my most critical system I'd try debugging it more, but I really
> >can't afford for this machine to be out of service. I was hoping that someone
> >with one of these same motherboards:
> >
> >http://www.intel.com/Products/Desktop/Motherboards/db-d525mw/D525MW-overview.htm
> >
> >in a less critical situation might have also seen the problem and been able to
> >do more debugging of it than I am inclined to.
> >
> >Bob
> 
> FWIW, this problem isn't confined to the D525's.  I saw the same problem
> on my D510MO sometime between 8.2-RC3 and -RELEASE.  My mobo has the
> NM10 chipset, which is also used on the D525.
> 
> I've since reverted to RC3 and have been postponing an upgrade to
> -STABLE until this issue is resolved.   Like Bob, I can't afford to have
> this machine disabled.
> 
> Nor am I able to gather any logs, etc., because I don't have a serial
> terminal and the problem makes any disk io unreliable.

Against my better judgement, I decided to give the latest version of
ata-intel.c another try on my gateway system yesterday. Unfortunately, the
problem still persists. To complicate matters upon installing the new kernel
for some reason my old (working) kernel was NOT saved in /boot/kernel.old. I
have no idea why, and didn't notice this till after attempting to reboot the
old kernel (when the new kernel failed).

Several hours of frantic recovery later I finally got my system back up with a
new kernel and the old version of ata-intel.c that still works (had some minor
adjustments to make on it due to the removal of a symbol that it used,
ATA_I82801IB_AH2).

I do find it hard to believe that Charlie and I are the only ones out there
with systems that fail with this newer version of the driver. I wish I could
help debug it more than I've been able to, but this is the only system I have
with this hardware and it's my most critical one to be up and running.

Bob

> _______________________________________________
> svn-src-all at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/svn-src-all
> To unsubscribe, send any mail to "svn-src-all-unsubscribe at freebsd.org"

-- 
Bob Willcox          "In the age of the internet attaching a famous name to
bob at immure.com       your personal opinion to give more weight to it is a very
Austin, TX           valid strategy."       -- Benjamin Franklin