Libata PATA for 2.6.16

Libata PATA for 2.6.16

[Alan Cox]

Can be found at the usual location

	http://zeniv.linux.org.uk/~alan/IDE

Some further small changes and updates, in particular the use of
platform device class for VLB/ISA/legacy IDE ports and the removal of
the "no device" special cases from the core.

Alan

Re: Libata PATA for 2.6.16

[Matthieu CASTET]

Hi Alan,

Le Mon, 20 Mar 2006 15:38:15 +0000, Alan Cox a écrit :

> Can be found at the usual location
> 
> 	http://zeniv.linux.org.uk/~alan/IDE
> 
> Some further small changes and updates, in particular the use of
> platform device class for VLB/ISA/legacy IDE ports and the removal of
> the "no device" special cases from the core.
> 
I still got the same problems on via :
 - hardisk used at UDMA33 instead of UDMA100 (buggy cable select ??? for
the primary CS value was 0xf1f1e6e6 and for the secondary 0x2727e6e6)
 - a cdrw drive lost interrupt when setting xfermode and make failed a
 port initialisation (if I comment the disable code, everythings seems to
 work)

Matthieu

Re: Libata PATA for 2.6.16

[Chris Boot]

Hi Alan,

Alan Cox wrote:
> Can be found at the usual location
> 
> 	http://zeniv.linux.org.uk/~alan/IDE
> 
> Some further small changes and updates, in particular the use of
> platform device class for VLB/ISA/legacy IDE ports and the removal of
> the "no device" special cases from the core.
> 
> Alan

If you're still looking for success reports, I just burned a DVD on 
2.6.16-ide1 just fine. At the final stages of burning I got a few 
messages about the burner not being ready but it kept trying and 
eventually succeeded writing the lead-out. DVD works fine.

Still not tried the pesky tape drive, those tapes I ordered weeks ago 
are still on back order! It is detected properly and claimed by 'st' 
through.

HTH,
Chris

-- 
Chris Boot
bootc@bootc.net
http://www.bootc.net/

Schedule for adding pata to kernel?

[Harald Dunkel]

[-- Attachment #1: Type: text/plain, Size: 1092 bytes --]

Hi Alan,

Works for me, too. Is there a plan to add this to the source tree?

0000:00:1f.2 IDE interface: Intel Corporation 82801FB/FW (ICH6/ICH6W) SATA Controller (rev 04)
0000:00:1f.2 0101: 8086:2651 (rev 04)


Many thanx

Harri
======================================================================
Chris Boot wrote:
> Hi Alan,
> 
> Alan Cox wrote:
> 
>> Can be found at the usual location
>>
>>     http://zeniv.linux.org.uk/~alan/IDE
>>
>> Some further small changes and updates, in particular the use of
>> platform device class for VLB/ISA/legacy IDE ports and the removal of
>> the "no device" special cases from the core.
>>
>> Alan
> 
> 
> If you're still looking for success reports, I just burned a DVD on
> 2.6.16-ide1 just fine. At the final stages of burning I got a few
> messages about the burner not being ready but it kept trying and
> eventually succeeded writing the lead-out. DVD works fine.
> 
> Still not tried the pesky tape drive, those tapes I ordered weeks ago
> are still on back order! It is detected properly and claimed by 'st'
> through.
> 
> HTH,
> Chris
> 

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 258 bytes --]

Re: Schedule for adding pata to kernel?

[Alan Cox]

On Sul, 2006-03-26 at 11:53 +0200, Harald Dunkel wrote:
> Hi Alan,
> 
> Works for me, too. Is there a plan to add this to the source tree?

Yes

There are several chunks of work needed. Firstly was to resync with the
libata development tree which is now done. The second to push changes to
the core code neccessary to support the PATA drivers. This is about
2/3rd done (Tejun Ho has been doing lots of work on mode changes which
also meant chunks of it were not needed).

Of the other bits that need merging some need resubmits/tidying and
oddments noted/requested by Jeff fixing. At that point it should be
possible just to drop in drivers during 2.6.17 development.

The final hard bit will be the PIIX/ICH driver as this is a major update
to a critical driver used by very many SATA people rather than a new
driver, so has the highest risk of breaking stuff.

Alan

Re: Schedule for adding pata to kernel?

[James Cloos]

>>>>> "Alan" == Alan Cox <alan@lxorguk.ukuu.org.uk> writes:

Alan> The final hard bit will be the PIIX/ICH driver as this is a major update
Alan> to a critical driver used by very many SATA people rather than a new
Alan> driver, so has the highest risk of breaking stuff.

Incidently, on that front, what is the magic to make it work?

I've not yet tried with 2.6.16 final, but I didn't have any luck with
earlier versions.

I have a:

,----[lspci]
| 00:1f.1 0101: 8086:244a (rev 03)
| 00:1f.1 IDE interface: Intel Corporation 82801BAM IDE U100 (rev 03)
`----

but never managed to determine the CONFIG that used ata_piix rather
than the old ide drivers.  Each attempt left me with a kernel which
could not mount its root....

-JimC

Re: Schedule for adding pata to kernel?

[Robert Hancock]

James Cloos wrote:
> Incidently, on that front, what is the magic to make it work?
> 
> I've not yet tried with 2.6.16 final, but I didn't have any luck with
> earlier versions.
> 
> I have a:
> 
> ,----[lspci]
> | 00:1f.1 0101: 8086:244a (rev 03)
> | 00:1f.1 IDE interface: Intel Corporation 82801BAM IDE U100 (rev 03)
> `----
> 
> but never managed to determine the CONFIG that used ata_piix rather
> than the old ide drivers.  Each attempt left me with a kernel which
> could not mount its root....

What distribution? If you're building ata_piix and the other drivers 
(scsi, sd, etc.) as modules you'll have to play with your initrd to get 
it to load the correct modules on startup. If you build them into the 
kernel, it should just work..

-- 
Robert Hancock      Saskatoon, SK, Canada
To email, remove "nospam" from hancockr@nospamshaw.ca
Home Page: http://www.roberthancock.com/

Re: Schedule for adding pata to kernel?

[Jeff Garzik]

Robert Hancock wrote:
> James Cloos wrote:
>> Incidently, on that front, what is the magic to make it work?
>>
>> I've not yet tried with 2.6.16 final, but I didn't have any luck with
>> earlier versions.
>>
>> I have a:
>>
>> ,----[lspci]
>> | 00:1f.1 0101: 8086:244a (rev 03)
>> | 00:1f.1 IDE interface: Intel Corporation 82801BAM IDE U100 (rev 03)
>> `----
>>
>> but never managed to determine the CONFIG that used ata_piix rather
>> than the old ide drivers.  Each attempt left me with a kernel which
>> could not mount its root....
> 
> What distribution? If you're building ata_piix and the other drivers 
> (scsi, sd, etc.) as modules you'll have to play with your initrd to get 
> it to load the correct modules on startup. If you build them into the 
> kernel, it should just work..

Normally you add the requisite drivers to /etc/modprobe.conf, and 
mkinitrd should select the correct modules from there.

	Jeff

Re: Schedule for adding pata to kernel?

[James Cloos]

>>>>> "Robert" == Robert Hancock <hancockr@shaw.ca> writes:

JimC> but [I] never managed to determine the CONFIG that used ata_piix rather
JimC> than the old ide drivers.  Each attempt left me with a kernel which
JimC> could not mount its root....

Robert> If you build them into the kernel, it should just work..

Agreed.  It should.

But if I leave in CONFIG_IDE=y, CONFIG_BLK_DEV_PIIX=y I get root on
0x0302 rather than 0x0802 even when I also have CONFIG_SCSI_SATA=y
and CONFIG_SCSI_ATA_PIIX=y.

If the first two are not set then the kernel cannot find anything.

My understanding is that with Alan's patch everything should be
using major 8, and that CONFIG_IDE is no longer required, yes?

-JimC
-- 
James H. Cloos, Jr. <cloos@jhcloos.com>

Re: Schedule for adding pata to kernel?

[Robert Hancock]

James Cloos wrote:
>>>>>> "Robert" == Robert Hancock <hancockr@shaw.ca> writes:
> 
> JimC> but [I] never managed to determine the CONFIG that used ata_piix rather
> JimC> than the old ide drivers.  Each attempt left me with a kernel which
> JimC> could not mount its root....
> 
> Robert> If you build them into the kernel, it should just work..
> 
> Agreed.  It should.
> 
> But if I leave in CONFIG_IDE=y, CONFIG_BLK_DEV_PIIX=y I get root on
> 0x0302 rather than 0x0802 even when I also have CONFIG_SCSI_SATA=y
> and CONFIG_SCSI_ATA_PIIX=y.
> 
> If the first two are not set then the kernel cannot find anything.
> 
> My understanding is that with Alan's patch everything should be
> using major 8, and that CONFIG_IDE is no longer required, yes?

You should be able to set CONFIG_IDE=n and have it work. That PCI ID is 
listed in the ata_piix device table so I should think it would work.. 
can you post the output?

Re: Schedule for adding pata to kernel?

[Alan Cox]

On Mer, 2006-03-29 at 17:30 -0500, James Cloos wrote:
> My understanding is that with Alan's patch everything should be
> using major 8, and that CONFIG_IDE is no longer required, yes?

This is correct both if you are just using the existing SATA drivers or
if you are using the libata patch to drive older chips via libata.

You can mix and match but don't compile in two drivers for the same chip

Alan

Re: Schedule for adding pata to kernel?

[James Cloos]

I did try it again with 2.6.16.

And now it does work.

Even seems faster than 2.6.16 w/o Alan's patch. :>

But that is probably just the short uptime.

(The laptop has always slowed down after being up for a few days....)

-JimC
-- 
James H. Cloos, Jr. <cloos@jhcloos.com>

Re: Schedule for adding pata to kernel?

[Alan Cox]

On Maw, 2006-04-18 at 13:44 +0200, Fabio Comolli wrote:
> In case PIIX/ICH driver should not make it in 2.6.17, are you planning
> to release patches for 2.6.17-rc release cycle?

I've been on holiday and am now tied up in other work until the start of
May, at which point Jeff goes off and gets married so there may be some
delay.

2.6.17-rc actually has 95% of the stuff needed to drop the PATA drivers
in and I will try to do patches at least versus 2.6.17 final. The -rcs
will depend upon available time and also what gets integrated that
causes additional work (notably Tejun Ho's stuff will make much merge
work, although its work I'm very glad to do as the improvements and
hotplug support are all needed).

Alan

Re: Schedule for adding pata to kernel?

[Tejun Heo]

Hello, all.

[adding Jeff and linux-ide to cc list]

On Tue, Apr 18, 2006 at 01:00:13PM +0100, Alan Cox wrote:
> On Maw, 2006-04-18 at 13:44 +0200, Fabio Comolli wrote:
> > In case PIIX/ICH driver should not make it in 2.6.17, are you planning
> > to release patches for 2.6.17-rc release cycle?
> 
> I've been on holiday and am now tied up in other work until the start of
> May, at which point Jeff goes off and gets married so there may be some
> delay.
> 
> 2.6.17-rc actually has 95% of the stuff needed to drop the PATA drivers
> in and I will try to do patches at least versus 2.6.17 final. The -rcs
> will depend upon available time and also what gets integrated that
> causes additional work (notably Tejun Ho's stuff will make much merge

BTW, my name is Tejun Heo.  Tejun Ho sounds horrible in Korean.

> work, although its work I'm very glad to do as the improvements and
> hotplug support are all needed).

I'm currently working on port multiplier support.  My working tree now
successfully probes and attaches all devices over the PM and I'm
currently trying to get EH and hotplug to work with it nicely.
EH/hotplug are being changed to support PM.  Effects on LLDDs are
minimal but you can probably save some work by waiting for the next
round of patches before porting to new EH.

I think/hope this can be finished in this week and bombard Jeff with
patches (updated EH, NCQ, hotplug and PM) before the weekend, so that
Jeff can have some time to review and hopefully merge some of it into
#upstream before he goes off on honeymoon.  In some convoluted way,
the patches will be my marriage gift, heh heh.

Jeff, *BIG* congratulations.  I wish you a great marriage.

Thanks.

-- 
tejun

Re: Schedule for adding pata to kernel?

[Jeff Garzik]

Tejun Heo wrote:
> Hello, all.
> 
> [adding Jeff and linux-ide to cc list]
> 
> On Tue, Apr 18, 2006 at 01:00:13PM +0100, Alan Cox wrote:
>> On Maw, 2006-04-18 at 13:44 +0200, Fabio Comolli wrote:
>>> In case PIIX/ICH driver should not make it in 2.6.17, are you planning
>>> to release patches for 2.6.17-rc release cycle?
>> I've been on holiday and am now tied up in other work until the start of
>> May, at which point Jeff goes off and gets married so there may be some
>> delay.
>>
>> 2.6.17-rc actually has 95% of the stuff needed to drop the PATA drivers
>> in and I will try to do patches at least versus 2.6.17 final. The -rcs
>> will depend upon available time and also what gets integrated that
>> causes additional work (notably Tejun Ho's stuff will make much merge
> 
> BTW, my name is Tejun Heo.  Tejun Ho sounds horrible in Korean.
> 
>> work, although its work I'm very glad to do as the improvements and
>> hotplug support are all needed).
> 
> I'm currently working on port multiplier support.  My working tree now
> successfully probes and attaches all devices over the PM and I'm
> currently trying to get EH and hotplug to work with it nicely.
> EH/hotplug are being changed to support PM.  Effects on LLDDs are
> minimal but you can probably save some work by waiting for the next
> round of patches before porting to new EH.
> 
> I think/hope this can be finished in this week and bombard Jeff with
> patches (updated EH, NCQ, hotplug and PM) before the weekend, so that
> Jeff can have some time to review and hopefully merge some of it into
> #upstream before he goes off on honeymoon.  In some convoluted way,
> the patches will be my marriage gift, heh heh.
> 
> Jeff, *BIG* congratulations.  I wish you a great marriage.

Thanks :)  And thanks for working on this stuff...  As you see, once the 
EH hurdle is cleared, it is much easier to add new features.  New 
features will start flooding in, from you and others.

BTW don't forget we want to push Albert's irq-pio branch into upstream 
sometime soon after your EH work settles.  I would put irq-pio in front 
of NCQ and PM, particularly.

	Jeff

Re: Schedule for adding pata to kernel?

[Tejun Heo]

On Thu, Apr 20, 2006 at 06:48:31PM -0400, Jeff Garzik wrote:
> Tejun Heo wrote:
> >Hello, all.
> >
> >[adding Jeff and linux-ide to cc list]
> >
> >On Tue, Apr 18, 2006 at 01:00:13PM +0100, Alan Cox wrote:
> >>On Maw, 2006-04-18 at 13:44 +0200, Fabio Comolli wrote:
> >>>In case PIIX/ICH driver should not make it in 2.6.17, are you planning
> >>>to release patches for 2.6.17-rc release cycle?
> >>I've been on holiday and am now tied up in other work until the start of
> >>May, at which point Jeff goes off and gets married so there may be some
> >>delay.
> >>
> >>2.6.17-rc actually has 95% of the stuff needed to drop the PATA drivers
> >>in and I will try to do patches at least versus 2.6.17 final. The -rcs
> >>will depend upon available time and also what gets integrated that
> >>causes additional work (notably Tejun Ho's stuff will make much merge
> >
> >BTW, my name is Tejun Heo.  Tejun Ho sounds horrible in Korean.
> >
> >>work, although its work I'm very glad to do as the improvements and
> >>hotplug support are all needed).
> >
> >I'm currently working on port multiplier support.  My working tree now
> >successfully probes and attaches all devices over the PM and I'm
> >currently trying to get EH and hotplug to work with it nicely.
> >EH/hotplug are being changed to support PM.  Effects on LLDDs are
> >minimal but you can probably save some work by waiting for the next
> >round of patches before porting to new EH.
> >
> >I think/hope this can be finished in this week and bombard Jeff with
> >patches (updated EH, NCQ, hotplug and PM) before the weekend, so that
> >Jeff can have some time to review and hopefully merge some of it into
> >#upstream before he goes off on honeymoon.  In some convoluted way,
> >the patches will be my marriage gift, heh heh.
> >
> >Jeff, *BIG* congratulations.  I wish you a great marriage.
> 
> Thanks :)  And thanks for working on this stuff...  As you see, once the 
> EH hurdle is cleared, it is much easier to add new features.  New 
> features will start flooding in, from you and others.
> 
> BTW don't forget we want to push Albert's irq-pio branch into upstream 
> sometime soon after your EH work settles.  I would put irq-pio in front 
> of NCQ and PM, particularly.

Hmmm... EH, NCQ, PM stuff doesn't really interfere with irq-pio.  The
only thing which needs special attention is hotplug support for
sata_sil as the driver requires PIO HSM implementation but needs to
use its own interrupt handler.  So, we need to separate out PIO HSM
from ata_host_intr() and allow LLDD irq handlers to drive it.

Albert, what do you think?

-- 
tejun

Re: Schedule for adding pata to kernel?

[Albert Lee]

Tejun Heo wrote:
> On Thu, Apr 20, 2006 at 06:48:31PM -0400, Jeff Garzik wrote:
> 
>>Tejun Heo wrote:
>>
>>>Hello, all.
>>>
>>>[adding Jeff and linux-ide to cc list]
>>>
>>>On Tue, Apr 18, 2006 at 01:00:13PM +0100, Alan Cox wrote:
>>>
>>>>On Maw, 2006-04-18 at 13:44 +0200, Fabio Comolli wrote:
>>>>
>>>>>In case PIIX/ICH driver should not make it in 2.6.17, are you planning
>>>>>to release patches for 2.6.17-rc release cycle?
>>>>
>>>>I've been on holiday and am now tied up in other work until the start of
>>>>May, at which point Jeff goes off and gets married so there may be some
>>>>delay.
>>>>
>>>>2.6.17-rc actually has 95% of the stuff needed to drop the PATA drivers
>>>>in and I will try to do patches at least versus 2.6.17 final. The -rcs
>>>>will depend upon available time and also what gets integrated that
>>>>causes additional work (notably Tejun Ho's stuff will make much merge
>>>
>>>BTW, my name is Tejun Heo.  Tejun Ho sounds horrible in Korean.
>>>
>>>
>>>>work, although its work I'm very glad to do as the improvements and
>>>>hotplug support are all needed).
>>>
>>>I'm currently working on port multiplier support.  My working tree now
>>>successfully probes and attaches all devices over the PM and I'm
>>>currently trying to get EH and hotplug to work with it nicely.
>>>EH/hotplug are being changed to support PM.  Effects on LLDDs are
>>>minimal but you can probably save some work by waiting for the next
>>>round of patches before porting to new EH.
>>>
>>>I think/hope this can be finished in this week and bombard Jeff with
>>>patches (updated EH, NCQ, hotplug and PM) before the weekend, so that
>>>Jeff can have some time to review and hopefully merge some of it into
>>>#upstream before he goes off on honeymoon.  In some convoluted way,
>>>the patches will be my marriage gift, heh heh.
>>>
>>>Jeff, *BIG* congratulations.  I wish you a great marriage.
>>
>>Thanks :)  And thanks for working on this stuff...  As you see, once the 
>>EH hurdle is cleared, it is much easier to add new features.  New 
>>features will start flooding in, from you and others.
>>
>>BTW don't forget we want to push Albert's irq-pio branch into upstream 
>>sometime soon after your EH work settles.  I would put irq-pio in front 
>>of NCQ and PM, particularly.
> 
> 
> Hmmm... EH, NCQ, PM stuff doesn't really interfere with irq-pio.  The
> only thing which needs special attention is hotplug support for
> sata_sil as the driver requires PIO HSM implementation but needs to
> use its own interrupt handler.  So, we need to separate out PIO HSM
> from ata_host_intr() and allow LLDD irq handlers to drive it.

 In the irq-pio branch, the HSM code has been seperated out from
ata_host_intr() to ata_hsm_move() as static function currently.
Maybe we can export ata_hsm_move() as external?
I didn't look into how sil3611 hotplug interacts with PIO HSM.
Don't know whether this is good enough for the need of sata_sil.

--
albert