Promise ATA/133 Errors With 2.6.10+

Promise ATA/133 Errors With 2.6.10+

[Justin Piszcz]

I have two different machines with the 7200.8 Seagate 8MB 400GB drives.

Both have ATA/133 controllers, the error is the same on both:

Jun 24 15:24:18 localhost kernel: hde: no DRQ after issuing MULTWRITE_EXT

I put the drive on an (older) Promise ATA/100 controller = works great!
I put the drive on the second box on the motherboard IDE interface = works 
great!

What happened > 2.6.10 to the promise driver?

??

Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
Jun 24 15:24:18 localhost kernel: hde: timeout waiting for DMA
Jun 24 15:24:18 localhost kernel: hde: status error: status=0x58 { 
DriveReady SeekComplete DataRequest }
Jun 24 15:24:18 localhost kernel:
Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
Jun 24 15:24:18 localhost kernel: hde: drive not ready for command
Jun 24 15:24:18 localhost kernel: hde: status timeout: status=0xd0 { Busy 
}
Jun 24 15:24:18 localhost kernel:
Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
Jun 24 15:24:18 localhost kernel: hde: no DRQ after issuing MULTWRITE_EXT
Jun 24 15:24:18 localhost kernel: ide2: reset: success

Re: Promise ATA/133 Errors With 2.6.10+

[Erik Slagter]

On Fri, 2005-06-24 at 16:55 -0400, Justin Piszcz wrote:

> Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
> Jun 24 15:24:18 localhost kernel: hde: timeout waiting for DMA
> Jun 24 15:24:18 localhost kernel: hde: status error: status=0x58 { 
> DriveReady SeekComplete DataRequest }
> Jun 24 15:24:18 localhost kernel:
> Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
> Jun 24 15:24:18 localhost kernel: hde: drive not ready for command
> Jun 24 15:24:18 localhost kernel: hde: status timeout: status=0xd0 { Busy 
> }
> Jun 24 15:24:18 localhost kernel:
> Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
> Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
> Jun 24 15:24:18 localhost kernel: hde: no DRQ after issuing MULTWRITE_EXT
> Jun 24 15:24:18 localhost kernel: ide2: reset: success

I have exactly this (these messages) but then with a amd/via ata driver
on tyan/amd motherboard with an IBM/Hitachi harddisk.

It looks like the drive cpu locks up every now and then, notably when
the environmental temperature and/or the drive's temperature are high
and there is much activity on the drive.

A reset (either from the driver or manually using hdparm) helps
(temporarily).

I was going to buy new drives for this fact (from another brand) but it
looks that won't necessarily mean my problem will be solved :-(

BTW I have another, exactly identical harddisk in the same computer
(well, ok, 1 year younger) and that one doesn't show the problem. 

BTW2 could it be that somewhere a timeout has been lowered in recent
kernels? That must have been pre-2.6.11 then.

Re: Promise ATA/133 Errors With 2.6.10+

[Justin Piszcz]

On Sat, 25 Jun 2005, Erik Slagter wrote:

> On Fri, 2005-06-24 at 16:55 -0400, Justin Piszcz wrote:
>
>> Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
>> Jun 24 15:24:18 localhost kernel: hde: timeout waiting for DMA
>> Jun 24 15:24:18 localhost kernel: hde: status error: status=0x58 {
>> DriveReady SeekComplete DataRequest }
>> Jun 24 15:24:18 localhost kernel:
>> Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
>> Jun 24 15:24:18 localhost kernel: hde: drive not ready for command
>> Jun 24 15:24:18 localhost kernel: hde: status timeout: status=0xd0 { Busy
>> }
>> Jun 24 15:24:18 localhost kernel:
>> Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
>> Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
>> Jun 24 15:24:18 localhost kernel: hde: no DRQ after issuing MULTWRITE_EXT
>> Jun 24 15:24:18 localhost kernel: ide2: reset: success
>
> I have exactly this (these messages) but then with a amd/via ata driver
> on tyan/amd motherboard with an IBM/Hitachi harddisk.
>
> It looks like the drive cpu locks up every now and then, notably when
> the environmental temperature and/or the drive's temperature are high
> and there is much activity on the drive.
>
> A reset (either from the driver or manually using hdparm) helps
> (temporarily).
>
> I was going to buy new drives for this fact (from another brand) but it
> looks that won't necessarily mean my problem will be solved :-(
>
> BTW I have another, exactly identical harddisk in the same computer
> (well, ok, 1 year younger) and that one doesn't show the problem.
>
> BTW2 could it be that somewhere a timeout has been lowered in recent
> kernels? That must have been pre-2.6.11 then.
>

^^

I think so!

The box has worked for the past 6 months and started to have hiccups when
I upgraded it to 2.6.11.x or 2.6.12.x.

Re: Promise ATA/133 Errors With 2.6.10+

[Alan Cox]

On Sad, 2005-06-25 at 09:35, Justin Piszcz wrote:
> > BTW2 could it be that somewhere a timeout has been lowered in recent
> > kernels? That must have been pre-2.6.11 then.

Timeouts have not changed or have increased in fact.

Re: Promise ATA/133 Errors With 2.6.10+

[Erik Slagter]

On Sun, 2005-06-26 at 18:59 +0100, Alan Cox wrote:
> On Sad, 2005-06-25 at 09:35, Justin Piszcz wrote:
> > > BTW2 could it be that somewhere a timeout has been lowered in recent
> > > kernels? That must have been pre-2.6.11 then.
> 
> Timeouts have not changed or have increased in fact.

Never mind, the offending harddisk has ceased to be yesterday, it is no
more.

What really bothers me, though, is that until the very last moment it
was alive, it didn't report any smart error, nor did any self test fail.
I guess IBM is to blame here :-(

Re: Promise ATA/133 Errors With 2.6.10+

[Vojtech Pavlik]

On Mon, Jun 27, 2005 at 12:11:36PM +0200, Erik Slagter wrote:

> On Sun, 2005-06-26 at 18:59 +0100, Alan Cox wrote:
> > On Sad, 2005-06-25 at 09:35, Justin Piszcz wrote:
> > > > BTW2 could it be that somewhere a timeout has been lowered in recent
> > > > kernels? That must have been pre-2.6.11 then.
> > 
> > Timeouts have not changed or have increased in fact.
> 
> Never mind, the offending harddisk has ceased to be yesterday, it is no
> more.
> 
> What really bothers me, though, is that until the very last moment it
> was alive, it didn't report any smart error, nor did any self test fail.
> I guess IBM is to blame here :-(
 
Most drives report no SMART problems until they die. I've seen several
drives who weren't able to read/write or at least remap bad sectors, and
still their SMART statistics were almost perfect. The SMART event log
included the errors, though.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR

Re: Promise ATA/133 Errors With 2.6.10+

[Erik Slagter]

On Mon, 2005-06-27 at 13:33 +0200, Vojtech Pavlik wrote:
> On Mon, Jun 27, 2005 at 12:11:36PM +0200, Erik Slagter wrote:
> 
> > On Sun, 2005-06-26 at 18:59 +0100, Alan Cox wrote:
> > > On Sad, 2005-06-25 at 09:35, Justin Piszcz wrote:
> > > > > BTW2 could it be that somewhere a timeout has been lowered in recent
> > > > > kernels? That must have been pre-2.6.11 then.
> > > 
> > > Timeouts have not changed or have increased in fact.
> > 
> > Never mind, the offending harddisk has ceased to be yesterday, it is no
> > more.
> > 
> > What really bothers me, though, is that until the very last moment it
> > was alive, it didn't report any smart error, nor did any self test fail.
> > I guess IBM is to blame here :-(
>  
> Most drives report no SMART problems until they die. I've seen several
> drives who weren't able to read/write or at least remap bad sectors, and
> still their SMART statistics were almost perfect. The SMART event log
> included the errors, though.

In this case NONE of the smartctl -a output revealed any problem, not
the attribute values, not the event log, and not the selftest log
(performed nightly).

I guess it's a problem like a minimal bad contact on the pcb, because
when the drive has properly cooled down, it does function for a while,
then at a certain point (smartctl reports ~35 C) it stops spinning and
hangs the complete bus.

I'm done with it, it is going to be replaced asap.

Re: Promise ATA/133 Errors With 2.6.10+

[Justin Piszcz]

It appears that 2.6.13-rc7 has fixed the bug.
I would like to know *What* changed, but I'll probably never find out :(

On Thu, 28 Jul 2005, Andrew Morton wrote:

> Justin Piszcz <jpiszcz@lucidpixels.com> wrote:
>>
>> I have two different machines with the 7200.8 Seagate 8MB 400GB drives.
>>
>> Both have ATA/133 controllers, the error is the same on both:
>>
>> Jun 24 15:24:18 localhost kernel: hde: no DRQ after issuing MULTWRITE_EXT
>>
>> I put the drive on an (older) Promise ATA/100 controller = works great!
>> I put the drive on the second box on the motherboard IDE interface = works
>> great!
>>
>> What happened > 2.6.10 to the promise driver?
>>
>> ??
>>
>> Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
>> Jun 24 15:24:18 localhost kernel: hde: timeout waiting for DMA
>> Jun 24 15:24:18 localhost kernel: hde: status error: status=0x58 {
>> DriveReady SeekComplete DataRequest }
>> Jun 24 15:24:18 localhost kernel:
>> Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
>> Jun 24 15:24:18 localhost kernel: hde: drive not ready for command
>> Jun 24 15:24:18 localhost kernel: hde: status timeout: status=0xd0 { Busy
>> }
>> Jun 24 15:24:18 localhost kernel:
>> Jun 24 15:24:18 localhost kernel: ide: failed opcode was: unknown
>> Jun 24 15:24:18 localhost kernel: PDC202XX: Primary channel reset.
>> Jun 24 15:24:18 localhost kernel: hde: no DRQ after issuing MULTWRITE_EXT
>> Jun 24 15:24:18 localhost kernel: ide2: reset: success
>
> Is this still happening in 2.6.13-rc4?
>
> If so, can you please cc linux-kernel on the reply?  Thanks.
>