mptspi: Unexpected doorbell active!

mptspi: Unexpected doorbell active!

[Douglas Gilbert]

Cute error message but it appears just before my machine
locks solid. It looks like it is coming from a LSI 53c1030
SPI controller which lspci says is PCI-X but is actually
PCIe.

I get that with 2 different tape drives (LTO-1 and LTO-3)
whose true identities are hidden behind badge engineering,
Tandberg data perhaps? Anyway both seem to have a major
problem with the LOG SENSE command and the Tape Alert
page in particular.

The lockup occurs with both a Debian Lenny standard kernel
(lk 2.6.26-2-686) and a kernel.org lk 2.6.30 .

There is nothing other than the 53c1030 and a single tape
drive on the SCSI bus. So there is no good reason to lock
up the machine for any SCSI related error.

Doug Gilbert

Re: mptspi: Unexpected doorbell active!

[James Bottomley]

On Wed, 2009-07-08 at 17:06 +0200, Douglas Gilbert wrote:
> Cute error message but it appears just before my machine
> locks solid. It looks like it is coming from a LSI 53c1030
> SPI controller which lspci says is PCI-X but is actually
> PCIe.
> 
> I get that with 2 different tape drives (LTO-1 and LTO-3)
> whose true identities are hidden behind badge engineering,
> Tandberg data perhaps? Anyway both seem to have a major
> problem with the LOG SENSE command and the Tape Alert
> page in particular.
> 
> The lockup occurs with both a Debian Lenny standard kernel
> (lk 2.6.26-2-686) and a kernel.org lk 2.6.30 .

So the problem has been around for a while ...

> There is nothing other than the 53c1030 and a single tape
> drive on the SCSI bus. So there is no good reason to lock
> up the machine for any SCSI related error.

LSI just reworked the code that does this doorbell handling for the
2.6.31 merge window could you see if 2.6.31-rc2 is still exhibiting the
problem?

Thanks,

James

Re: mptspi: Unexpected doorbell active!

[Douglas Gilbert]

James Bottomley wrote:
> On Wed, 2009-07-08 at 17:06 +0200, Douglas Gilbert wrote:
>> Cute error message but it appears just before my machine
>> locks solid. It looks like it is coming from a LSI 53c1030
>> SPI controller which lspci says is PCI-X but is actually
>> PCIe.
>>
>> I get that with 2 different tape drives (LTO-1 and LTO-3)
>> whose true identities are hidden behind badge engineering,
>> Tandberg data perhaps? Anyway both seem to have a major
>> problem with the LOG SENSE command and the Tape Alert
>> page in particular.
>>
>> The lockup occurs with both a Debian Lenny standard kernel
>> (lk 2.6.26-2-686) and a kernel.org lk 2.6.30 .
> 
> So the problem has been around for a while ...
> 
>> There is nothing other than the 53c1030 and a single tape
>> drive on the SCSI bus. So there is no good reason to lock
>> up the machine for any SCSI related error.
> 
> LSI just reworked the code that does this doorbell handling for the
> 2.6.31 merge window could you see if 2.6.31-rc2 is still exhibiting the
> problem?

Yes it is.

The controller is a LSI20320IE and the main chip appears
to be a LSI53C1020A A1.

Doug Gilbert

RE: mptspi: Unexpected doorbell active!

[Desai, Kashyap]

Can you provide dmesg log with log level enable to 0x8188.

"echo 0x8188 > /sys/class/scsi_host/hostX/debug_level"

Thanks,
Kashyap

-----Original Message-----
From: linux-scsi-owner@vger.kernel.org [mailto:linux-scsi-owner@vger.kernel.org] On Behalf Of Douglas Gilbert
Sent: Friday, July 10, 2009 2:12 AM
To: James Bottomley
Cc: linux-scsi@vger.kernel.org
Subject: Re: mptspi: Unexpected doorbell active!

James Bottomley wrote:
> On Wed, 2009-07-08 at 17:06 +0200, Douglas Gilbert wrote:
>> Cute error message but it appears just before my machine
>> locks solid. It looks like it is coming from a LSI 53c1030
>> SPI controller which lspci says is PCI-X but is actually
>> PCIe.
>>
>> I get that with 2 different tape drives (LTO-1 and LTO-3)
>> whose true identities are hidden behind badge engineering,
>> Tandberg data perhaps? Anyway both seem to have a major
>> problem with the LOG SENSE command and the Tape Alert
>> page in particular.
>>
>> The lockup occurs with both a Debian Lenny standard kernel
>> (lk 2.6.26-2-686) and a kernel.org lk 2.6.30 .
> 
> So the problem has been around for a while ...
> 
>> There is nothing other than the 53c1030 and a single tape
>> drive on the SCSI bus. So there is no good reason to lock
>> up the machine for any SCSI related error.
> 
> LSI just reworked the code that does this doorbell handling for the
> 2.6.31 merge window could you see if 2.6.31-rc2 is still exhibiting the
> problem?

Yes it is.

The controller is a LSI20320IE and the main chip appears
to be a LSI53C1020A A1.

Doug Gilbert


--
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: mptspi: Unexpected doorbell active!

[Grant Grundler]

On Wed, Jul 8, 2009 at 8:06 AM, Douglas Gilbert<dgilbert@interlog.com> wrote:
> Cute error message but it appears just before my machine
> locks solid. It looks like it is coming from a LSI 53c1030
> SPI controller which lspci says is PCI-X but is actually
> PCIe.

It's probably not relevant to the problem, but I'm pretty sure 53c1030
SPI is a PCI-X device. I supported this driver for HP's IA64 (e.g
rx2600) machines starting around 2001 or 2002 (before PCIe was
available).  Can you dump the lspci output that claims this is a PCIe
device?

> I get that with 2 different tape drives (LTO-1 and LTO-3)
> whose true identities are hidden behind badge engineering,
> Tandberg data perhaps? Anyway both seem to have a major
> problem with the LOG SENSE command and the Tape Alert
> page in particular.
>
> The lockup occurs with both a Debian Lenny standard kernel
> (lk 2.6.26-2-686) and a kernel.org lk 2.6.30 .
>
> There is nothing other than the 53c1030 and a single tape
> drive on the SCSI bus. So there is no good reason to lock
> up the machine for any SCSI related error.

Other people have reported similar problems before:
    http://www.mail-archive.com/linux-scsi@vger.kernel.org/msg14431.html

hth,
grant

Re: mptspi: Unexpected doorbell active!

[Martin K. Petersen]

>>>>> "Grant" == Grant Grundler <grundler@google.com> writes:

>> Cute error message but it appears just before my machine locks
>> solid. It looks like it is coming from a LSI 53c1030 SPI controller
>> which lspci says is PCI-X but is actually PCIe.

Grant> It's probably not relevant to the problem, but I'm pretty sure
Grant> 53c1030 SPI is a PCI-X device. I supported this driver for HP's
Grant> IA64 (e.g rx2600) machines starting around 2001 or 2002 (before
Grant> PCIe was available).  Can you dump the lspci output that claims
Grant> this is a PCIe device?

I believe there is a 3rd party that makes PCIe 1030 boards.  They are
not directly marketed by LSI.

The boards consist of a PCIe to PCI-X bridge and a regular 1030 chip.

-- 
Martin K. Petersen	Oracle Linux Engineering