Do we still need the scsi IPR driver ?

Do we still need the scsi IPR driver ?

[Damien Le Moal]

Polling people here: Do we still need the scsi IPR driver for IBM Power
Linux RAID adapters (IBM iSeries: 5702, 5703, 2780, 5709, 570A, 570B
adapters) ?

The reason I am asking is because this driver is the *only* libsas/ata
driver that does not define a ->error_handler port operation. If this
driver is removed, or if it is modified to use a ->error_handler operation
to handle failed commands, then a lot of code simplification can be done
in libata, which I am trying to do to facilitate the processing of some
special error completion for commands using a command duration limit.

Thoughts ?

-- 
Damien Le Moal
Western Digital Research

Re: Do we still need the scsi IPR driver ?

[Hannes Reinecke]

On 6/20/22 06:48, Damien Le Moal wrote:
> 
> Polling people here: Do we still need the scsi IPR driver for IBM Power
> Linux RAID adapters (IBM iSeries: 5702, 5703, 2780, 5709, 570A, 570B
> adapters) ?
> 
> The reason I am asking is because this driver is the *only* libsas/ata
> driver that does not define a ->error_handler port operation. If this
> driver is removed, or if it is modified to use a ->error_handler operation
> to handle failed commands, then a lot of code simplification can be done
> in libata, which I am trying to do to facilitate the processing of some
> special error completion for commands using a command duration limit.
> 
> Thoughts ?
> 
We certainly don't, I've looked through our pool for any machines and 
draw a blank.
And I'm reasonably certain that none of our customers is using one of 
them old machines, neither; IBM tends to be very particular about 
obsolete machines :-)

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		           Kernel Storage Architect
hare@suse.de			                  +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer

Re: Do we still need the scsi IPR driver ?

[Brian King]

On 6/19/22 11:48 PM, Damien Le Moal wrote:
> 
> Polling people here: Do we still need the scsi IPR driver for IBM Power
> Linux RAID adapters (IBM iSeries: 5702, 5703, 2780, 5709, 570A, 570B
> adapters) ?
> 
> The reason I am asking is because this driver is the *only* libsas/ata
> driver that does not define a ->error_handler port operation. If this
> driver is removed, or if it is modified to use a ->error_handler operation
> to handle failed commands, then a lot of code simplification can be done
> in libata, which I am trying to do to facilitate the processing of some
> special error completion for commands using a command duration limit.

We still need it around for now. IBM still sells these adapters
and they can still be ordered even on our latest Power 10 systems.
 
Thanks,

Brian

-- 
Brian King
Power Linux I/O
IBM Linux Technology Center

Re: Do we still need the scsi IPR driver ?

[Brian King]

On 6/21/22 3:36 PM, Brian King wrote:
> On 6/19/22 11:48 PM, Damien Le Moal wrote:
>>
>> Polling people here: Do we still need the scsi IPR driver for IBM Power
>> Linux RAID adapters (IBM iSeries: 5702, 5703, 2780, 5709, 570A, 570B
>> adapters) ?
>>
>> The reason I am asking is because this driver is the *only* libsas/ata
>> driver that does not define a ->error_handler port operation. If this
>> driver is removed, or if it is modified to use a ->error_handler operation
>> to handle failed commands, then a lot of code simplification can be done
>> in libata, which I am trying to do to facilitate the processing of some
>> special error completion for commands using a command duration limit.
> 
> We still need it around for now. IBM still sells these adapters
> and they can still be ordered even on our latest Power 10 systems.

At one point I did look into modifying ipr to use an ->error_handler.
I recall I ran into some issues that resulted in this getting put
on the shelf, but its been a while. I'll go dig that code up and
see what it looks like.

Thanks,

Brian


-- 
Brian King
Power Linux I/O
IBM Linux Technology Center

Re: Do we still need the scsi IPR driver ?

[Damien Le Moal]

On 6/22/22 05:52, Brian King wrote:
> On 6/21/22 3:36 PM, Brian King wrote:
>> On 6/19/22 11:48 PM, Damien Le Moal wrote:
>>>
>>> Polling people here: Do we still need the scsi IPR driver for IBM Power
>>> Linux RAID adapters (IBM iSeries: 5702, 5703, 2780, 5709, 570A, 570B
>>> adapters) ?
>>>
>>> The reason I am asking is because this driver is the *only* libsas/ata
>>> driver that does not define a ->error_handler port operation. If this
>>> driver is removed, or if it is modified to use a ->error_handler operation
>>> to handle failed commands, then a lot of code simplification can be done
>>> in libata, which I am trying to do to facilitate the processing of some
>>> special error completion for commands using a command duration limit.
>>
>> We still need it around for now. IBM still sells these adapters
>> and they can still be ordered even on our latest Power 10 systems.
> 
> At one point I did look into modifying ipr to use an ->error_handler.
> I recall I ran into some issues that resulted in this getting put
> on the shelf, but its been a while. I'll go dig that code up and
> see what it looks like.

Thanks. It would be really great if you can convert to using
error_handler. This is really the last ata/libsas driver that does not use
this.

> 
> Thanks,
> 
> Brian
> 
> 


-- 
Damien Le Moal
Western Digital Research

Re: Do we still need the scsi IPR driver ?

[John Garry]

On 21/06/2022 23:12, Damien Le Moal wrote:
>>> We still need it around for now. IBM still sells these adapters
>>> and they can still be ordered even on our latest Power 10 systems.
>> At one point I did look into modifying ipr to use an ->error_handler.
>> I recall I ran into some issues that resulted in this getting put
>> on the shelf, but its been a while. I'll go dig that code up and
>> see what it looks like.
> Thanks. It would be really great if you can convert to using
> error_handler. This is really the last ata/libsas driver that does not use
> this.
> 

Hi Brian,

I am wondering if there is any update here?

As you may have seen in [0], I think that we need to make progress on 
this topic first to keep the solution there a bit simpler.

[0] 
https://lore.kernel.org/linux-scsi/1663669630-21333-1-git-send-email-john.garry@huawei.com/T/#mf890cb4f1627112652831524dca62cbde4a0a637

Thanks,
John

Re: Do we still need the scsi IPR driver ?

[Brian King]

On 9/20/22 8:07 AM, John Garry wrote:
> On 21/06/2022 23:12, Damien Le Moal wrote:
>>>> We still need it around for now. IBM still sells these adapters
>>>> and they can still be ordered even on our latest Power 10 systems.
>>> At one point I did look into modifying ipr to use an ->error_handler.
>>> I recall I ran into some issues that resulted in this getting put
>>> on the shelf, but its been a while. I'll go dig that code up and
>>> see what it looks like.
>> Thanks. It would be really great if you can convert to using
>> error_handler. This is really the last ata/libsas driver that does not use
>> this.
>>
> 
> Hi Brian,
> 
> I am wondering if there is any update here?
> 
> As you may have seen in [0], I think that we need to make progress on this topic first to keep the solution there a bit simpler.
> 
> [0] https://lore.kernel.org/linux-scsi/1663669630-21333-1-git-send-email-john.garry@huawei.com/T/#mf890cb4f1627112652831524dca62cbde4a0a637

I've made some progress. I was able to dig up the code to move ipr to use error_handler
and have gotten it to compile, but haven't gotten to trying it in the lab yet.

-Brian


-- 
Brian King
Power Linux I/O
IBM Linux Technology Center

Re: Do we still need the scsi IPR driver ?

[Damien Le Moal]

On 10/6/22 02:20, Brian King wrote:
> On 9/20/22 8:07 AM, John Garry wrote:
>> On 21/06/2022 23:12, Damien Le Moal wrote:
>>>>> We still need it around for now. IBM still sells these adapters
>>>>> and they can still be ordered even on our latest Power 10 systems.
>>>> At one point I did look into modifying ipr to use an ->error_handler.
>>>> I recall I ran into some issues that resulted in this getting put
>>>> on the shelf, but its been a while. I'll go dig that code up and
>>>> see what it looks like.
>>> Thanks. It would be really great if you can convert to using
>>> error_handler. This is really the last ata/libsas driver that does not use
>>> this.
>>>
>>
>> Hi Brian,
>>
>> I am wondering if there is any update here?
>>
>> As you may have seen in [0], I think that we need to make progress on this topic first to keep the solution there a bit simpler.
>>
>> [0] https://lore.kernel.org/linux-scsi/1663669630-21333-1-git-send-email-john.garry@huawei.com/T/#mf890cb4f1627112652831524dca62cbde4a0a637
> 
> I've made some progress. I was able to dig up the code to move ipr to use error_handler
> and have gotten it to compile, but haven't gotten to trying it in the lab yet.

Great ! Thanks for doing that.

> 
> -Brian
> 
> 

-- 
Damien Le Moal
Western Digital Research

Re: Do we still need the scsi IPR driver ?

[Hannes Reinecke]

On 10/5/22 19:20, Brian King wrote:
> On 9/20/22 8:07 AM, John Garry wrote:
>> On 21/06/2022 23:12, Damien Le Moal wrote:
>>>>> We still need it around for now. IBM still sells these adapters
>>>>> and they can still be ordered even on our latest Power 10 systems.
>>>> At one point I did look into modifying ipr to use an ->error_handler.
>>>> I recall I ran into some issues that resulted in this getting put
>>>> on the shelf, but its been a while. I'll go dig that code up and
>>>> see what it looks like.
>>> Thanks. It would be really great if you can convert to using
>>> error_handler. This is really the last ata/libsas driver that does not use
>>> this.
>>>
>>
>> Hi Brian,
>>
>> I am wondering if there is any update here?
>>
>> As you may have seen in [0], I think that we need to make progress on this topic first to keep the solution there a bit simpler.
>>
>> [0] https://lore.kernel.org/linux-scsi/1663669630-21333-1-git-send-email-john.garry@huawei.com/T/#mf890cb4f1627112652831524dca62cbde4a0a637
> 
> I've made some progress. I was able to dig up the code to move ipr to use error_handler
> and have gotten it to compile, but haven't gotten to trying it in the lab yet.
> 
Hmm. In which machines can I find an IPR installed? I could go hunting 
in our lab, maybe I can locate one and aid testing/development ...

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                Kernel Storage Architect
hare@suse.de                              +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Ivo Totev, Andrew
Myers, Andrew McDonald, Martje Boudien Moerman

RE: Do we still need the scsi IPR driver ?

[Brian King]

On 10/6/22 2:35 AM, Hannes Reinecke wrote:
> On 10/5/22 19:20, Brian King wrote:
>> On 9/20/22 8:07 AM, John Garry wrote:
>>> On 21/06/2022 23:12, Damien Le Moal wrote:
>>>>>> We still need it around for now. IBM still sells these adapters
>>>>>> and they can still be ordered even on our latest Power 10 systems.
>>>>> At one point I did look into modifying ipr to use an ->error_handler.
>>>>> I recall I ran into some issues that resulted in this getting put
>>>>> on the shelf, but its been a while. I'll go dig that code up and
>>>>> see what it looks like.
>>>> Thanks. It would be really great if you can convert to using
>>>> error_handler. This is really the last ata/libsas driver that does not use
>>>> this.
>>>>
>>>
>>> Hi Brian,
>>>
>>> I am wondering if there is any update here?
>>>
>>> As you may have seen in [0], I think that we need to make progress on this topic first to keep the solution there a bit simpler.
>>>
>>> [0] https://lore.kernel.org/linux-scsi/1663669630-21333-1-git-send-email-john.garry@huawei.com/T/#mf890cb4f1627112652831524dca62cbde4a0a637  
>>
>> I've made some progress. I was able to dig up the code to move ipr to use error_handler
>> and have gotten it to compile, but haven't gotten to trying it in the lab yet.
>>
> Hmm. In which machines can I find an IPR installed? I could go hunting in our lab, maybe I can locate one and aid testing/development ...

Any Power 9 or older generation PowerVM based system would have an IPR installed as the boot device.
Additionally, on Power 10 systems, ipr SAS controllers are available as an add in card. 

However, the SATA support in ipr was only used to attach the onboard SATA DVD. Power 8 systems were
the last generation of systems that had an onboard SATA DVD. So, to do any testing with a
SATA DVD, you'd need a Power 8 or older system. 

Right now I have a patch that removes the SATA support from ipr completely and a patch that changes
to use the error_handler libata support. The one that changes to use the error_handler libata API
adds a bit of complexity for a function that should have few or no users that would need this support
on a current upstream kernel, since only Power 8 and older systems use this support. I'm getting
a system setup to try out both patches, but at this point I'm leaning towards the patch that
removes the libata dependency from ipr.

Thanks,

Brian


-- 
Brian King
Power Linux I/O
IBM Linux Technology Center

Re: Do we still need the scsi IPR driver ?

[John Garry]

On 13/10/2022 04:36, Brian King wrote:
>>> I've made some progress. I was able to dig up the code to move ipr to use error_handler
>>> and have gotten it to compile, but haven't gotten to trying it in the lab yet.
>>>
>> Hmm. In which machines can I find an IPR installed? I could go hunting in our lab, maybe I can locate one and aid testing/development ...
> Any Power 9 or older generation PowerVM based system would have an IPR installed as the boot device.
> Additionally, on Power 10 systems, ipr SAS controllers are available as an add in card.
> 
> However, the SATA support in ipr was only used to attach the onboard SATA DVD. Power 8 systems were
> the last generation of systems that had an onboard SATA DVD. So, to do any testing with a
> SATA DVD, you'd need a Power 8 or older system.
> 
> Right now I have a patch that removes the SATA support from ipr completely and a patch that changes
> to use the error_handler libata support. The one that changes to use the error_handler libata API
> adds a bit of complexity for a function that should have few or no users that would need this support
> on a current upstream kernel, since only Power 8 and older systems use this support. I'm getting
> a system setup to try out both patches, but at this point I'm leaning towards the patch that
> removes the libata dependency from ipr.

Hi Brian,

I was just wondering did you ever get to test these patches you mention? 
Or any other update on this topic?

Thanks,
John

RE: Do we still need the scsi IPR driver ?

[Brian King]

On 3/10/23 3:56 AM, John Garry wrote:
> On 13/10/2022 04:36, Brian King wrote:
>>>> I've made some progress. I was able to dig up the code to move ipr to use error_handler
>>>> and have gotten it to compile, but haven't gotten to trying it in the lab yet.
>>>>
>>> Hmm. In which machines can I find an IPR installed? I could go hunting in our lab, maybe I can locate one and aid testing/development ...
>> Any Power 9 or older generation PowerVM based system would have an IPR installed as the boot device.
>> Additionally, on Power 10 systems, ipr SAS controllers are available as an add in card.
>>
>> However, the SATA support in ipr was only used to attach the onboard SATA DVD. Power 8 systems were
>> the last generation of systems that had an onboard SATA DVD. So, to do any testing with a
>> SATA DVD, you'd need a Power 8 or older system.
>>
>> Right now I have a patch that removes the SATA support from ipr completely and a patch that changes
>> to use the error_handler libata support. The one that changes to use the error_handler libata API
>> adds a bit of complexity for a function that should have few or no users that would need this support
>> on a current upstream kernel, since only Power 8 and older systems use this support. I'm getting
>> a system setup to try out both patches, but at this point I'm leaning towards the patch that
>> removes the libata dependency from ipr.
> 
> Hi Brian,
> 
> I was just wondering did you ever get to test these patches you mention? Or any other update on this topic?

Yes. I've been running both patches in the lab. I'm still leaning towards the patch that just
removes the libata dependency from ipr, since that is the simpler patch. I'll try to get something sent
out soon.

Thanks,

Brian


-- 
Brian King
Power Linux I/O
IBM Linux Technology Center