Some grub2 problems/questions

Some grub2 problems/questions

[walt]

Hi grub group.

Beginning with the commits of 2007-10-31 my two Athlon machines won't
work properly with grub2, and I'd like to find out why.

The first machine has two ordinary PATA drives on a VIA 8235 controller
and grub2 complains that the firmware is tainted.  Grub2 will use (ata1)
okay and lists all the partitions, but for the second drive grub2 will
print only (ata2) without listing any partitions.

The other Athlon machine has one PATA drive on a VIA 8235 controller
and one SATA drive on a Promise controller.  This machine makes it
through grub_ata_dumpinfo for both disks but then it hangs forever
without printing anything else.

As of today, 2007-12-26, should I expect grub2 to work properly in
these two machines?  Any debugging hints?

Thanks!

Re: Some grub2 problems/questions

[Robert Millan]

On Wed, Dec 26, 2007 at 12:26:03PM -0800, walt wrote:
> Hi grub group.
> 
> Beginning with the commits of 2007-10-31 my two Athlon machines won't
> work properly with grub2, and I'd like to find out why.

It's a known problem.  That's when we added the ATA driver.  If you don't
include ata.mod in your core.img, everything should be back to normal.

However....

> The first machine has two ordinary PATA drives on a VIA 8235 controller
> and grub2 complains that the firmware is tainted.  Grub2 will use (ata1)
> okay and lists all the partitions,

Nice.  Is that controller in the PCI bus?  It's strange that you got it
working on real hardware (i.e. not qemu) because PCI wasn't implemented
yet.

> but for the second drive grub2 will
> print only (ata2) without listing any partitions.

As expected.

> The other Athlon machine has one PATA drive on a VIA 8235 controller
> and one SATA drive on a Promise controller.  This machine makes it
> through grub_ata_dumpinfo for both disks but then it hangs forever
> without printing anything else.

Could you re-run this part with `debug' variable set to `ata'
("set debug=ata", then "insmod ata") and see what's the last thing
printed ?

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call, if you are unable to speak?
(as seen on /.)

Re: Some grub2 problems/questions

[Robert Millan]

On Thu, Dec 27, 2007 at 11:05:53AM +0100, Robert Millan wrote:
> On Wed, Dec 26, 2007 at 12:26:03PM -0800, walt wrote:
> > Hi grub group.
> > 
> > Beginning with the commits of 2007-10-31 my two Athlon machines won't
> > work properly with grub2, and I'd like to find out why.
> 
> It's a known problem.  That's when we added the ATA driver.  If you don't
> include ata.mod in your core.img, everything should be back to normal.

Btw, how come you enabled ata.mod.  Did you explicitly request it?

It's a problem if our installer tools are misleading about this.  We don't
want people to use ata.mod unless they understand the implications (e.g. you
won't be able to boot BIOS-based OSes like OpenBSD).

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call, if you are unable to speak?
(as seen on /.)

Re: Some grub2 problems/questions

[walt]

Robert Millan wrote:
> On Wed, Dec 26, 2007 at 12:26:03PM -0800, walt wrote:
...
>> The first machine has two ordinary PATA drives on a VIA 8235 controller
>> and grub2 complains that the firmware is tainted.  Grub2 will use (ata1)
>> okay and lists all the partitions,
> 
> Nice.  Is that controller in the PCI bus?  It's strange that you got it
> working on real hardware (i.e. not qemu) because PCI wasn't implemented
> yet.

Yes.  However, I'm saying that grub2 works as of today, not 10-31.

>> The other Athlon machine has one PATA drive on a VIA 8235 controller
>> and one SATA drive on a Promise controller.  This machine makes it
>> through grub_ata_dumpinfo for both disks but then it hangs forever
>> without printing anything else.

> Could you re-run this part with `debug' variable set to `ata'
> ("set debug=ata", then "insmod ata") and see what's the last thing
> printed ?

Ah, I didn't notice before that grub2 only recognizes the PATA drive
and the cd-burner on the PATA controller.  Grub2 stops dumping
info after the cd-burner and never gets to the SATA disk on the
Promise controller.  The Promise controller is built into the mobo
and the BIOS recognizes it and will boot from it, but the ata module
in grub2 will not.  Would this explain why grub2 hangs?

I should mention that I've never installed grub2 because I don't
really want it on my boot record yet.  I build the grub2 image using
./grub-mkimage -d . -o mygrub *.mod (as I read on the wiki) and then
use old grub to load grub2.  I loaded ata.mod by accident because I'm
too lazy to type all of those module names :o)  I certainly don't
mind testing and debugging the ata module, though, if I can be of
any help to you.

ata.mod hangs (Re: Some grub2 problems/questions)

[Robert Millan]

On Thu, Dec 27, 2007 at 09:33:58AM -0800, walt wrote:
> >>The other Athlon machine has one PATA drive on a VIA 8235 controller
> >>and one SATA drive on a Promise controller.  This machine makes it
> >>through grub_ata_dumpinfo for both disks but then it hangs forever
> >>without printing anything else.
> 
> >Could you re-run this part with `debug' variable set to `ata'
> >("set debug=ata", then "insmod ata") and see what's the last thing
> >printed ?
> 
> Ah, I didn't notice before that grub2 only recognizes the PATA drive
> and the cd-burner on the PATA controller.  Grub2 stops dumping
> info after the cd-burner and never gets to the SATA disk on the
> Promise controller.  The Promise controller is built into the mobo
> and the BIOS recognizes it and will boot from it, but the ata module
> in grub2 will not.  Would this explain why grub2 hangs?
> 
> I should mention that I've never installed grub2 because I don't
> really want it on my boot record yet.  I build the grub2 image using
> ./grub-mkimage -d . -o mygrub *.mod (as I read on the wiki) and then
> use old grub to load grub2.  I loaded ata.mod by accident because I'm
> too lazy to type all of those module names :o)  I certainly don't
> mind testing and debugging the ata module, though, if I can be of
> any help to you.

Thank you.  Marco knows the ATA side of things better, although he's a bit
busy lately.  Let's see if he picks up on this...

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call, if you are unable to speak?
(as seen on /.)

Re: ata.mod hangs (Re: Some grub2 problems/questions)

[walt]

Robert Millan wrote:
> On Thu, Dec 27, 2007 at 09:33:58AM -0800, walt wrote:
>>>> The other Athlon machine has one PATA drive on a VIA 8235 controller
>>>> and one SATA drive on a Promise controller.  This machine makes it
>>>> through grub_ata_dumpinfo for both disks but then it hangs forever
>>>> without printing anything else.
>>> Could you re-run this part with `debug' variable set to `ata'
>>> ("set debug=ata", then "insmod ata") and see what's the last thing
>>> printed ?
>> Ah, I didn't notice before that grub2 only recognizes the PATA drive
>> and the cd-burner on the PATA controller.  Grub2 stops dumping
>> info after the cd-burner and never gets to the SATA disk on the
>> Promise controller.  The Promise controller is built into the mobo
>> and the BIOS recognizes it and will boot from it, but the ata module
>> in grub2 will not.  Would this explain why grub2 hangs?
>>
>> I should mention that I've never installed grub2 because I don't
>> really want it on my boot record yet.  I build the grub2 image using
>> ./grub-mkimage -d . -o mygrub *.mod (as I read on the wiki) and then
>> use old grub to load grub2.  I loaded ata.mod by accident because I'm
>> too lazy to type all of those module names :o)  I certainly don't
>> mind testing and debugging the ata module, though, if I can be of
>> any help to you.
>
> Thank you.  Marco knows the ATA side of things better, although he's a bit
> busy lately.  Let's see if he picks up on this...
>

I'm happy to report that the ata module no longer hangs as I described
above.  I'm guessing (just a guess) that this commit fixed it:

2008-01-10  Robert Millan  <rmh@aybabtu.com>
     * kern/device.c (grub_device_iterate): Do not abort device iteration
       when one of the devices cannot be opened.

Unfortunately, ata.mod still doesn't recognize the disk on the Promise
sata controller.  But, that's expected at this point?

Re: ata.mod hangs (Re: Some grub2 problems/questions)

[Robert Millan]

On Sat, Jan 12, 2008 at 04:14:38PM -0800, walt wrote:
> 
> I'm happy to report that the ata module no longer hangs as I described
> above.  I'm guessing (just a guess) that this commit fixed it:
> 
> 2008-01-10  Robert Millan  <rmh@aybabtu.com>
>     * kern/device.c (grub_device_iterate): Do not abort device iteration
>       when one of the devices cannot be opened.

That's strange.  Can you confirm it?  (by testing the revisions in CVS
inmediately before and after this commit)

> Unfortunately, ata.mod still doesn't recognize the disk on the Promise
> sata controller.  But, that's expected at this point?

Yes.  Some SATA controllers emulate PATA, but not all do.

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call, if you are unable to speak?
(as seen on /.)

Re: ata.mod hangs (Re: Some grub2 problems/questions)

[Vesa Jääskeläinen]

Robert Millan wrote:
> On Sat, Jan 12, 2008 at 04:14:38PM -0800, walt wrote:
>> Unfortunately, ata.mod still doesn't recognize the disk on the Promise
>> sata controller.  But, that's expected at this point?
> 
> Yes.  Some SATA controllers emulate PATA, but not all do.

I suppose there are multiple controller on the board and we need PCI to 
detect extra ones.

Re: ata.mod hangs (Re: Some grub2 problems/questions)

[Marco Gerards]

Vesa Jääskeläinen <chaac@nic.fi> writes:

> Robert Millan wrote:
>> On Sat, Jan 12, 2008 at 04:14:38PM -0800, walt wrote:
>>> Unfortunately, ata.mod still doesn't recognize the disk on the Promise
>>> sata controller.  But, that's expected at this point?
>>
>> Yes.  Some SATA controllers emulate PATA, but not all do.
>
> I suppose there are multiple controller on the board and we need PCI
> to detect extra ones.

Right, adding PCI support will solve most of our problems in ata.mod.
Currently, a reset is sent to each device.  If PCI is used, every ATA
device can be found (if they support the ATA/ATAPI standard).  Besides
that, the reset might cause problems.  Perhaps the ATA driver can even
coexist with biosdisk.mod in this case...

--
Marco

Re: ata.mod hangs (Re: Some grub2 problems/questions)

[walt]

Robert Millan wrote:
> On Sat, Jan 12, 2008 at 04:14:38PM -0800, walt wrote:
>> I'm happy to report that the ata module no longer hangs as I described
>> above.  I'm guessing (just a guess) that this commit fixed it:
>>
>> 2008-01-10  Robert Millan<rmh@aybabtu.com>
>>      * kern/device.c (grub_device_iterate): Do not abort device iteration
>>        when one of the devices cannot be opened.
>
> That's strange.  Can you confirm it?  (by testing the revisions in CVS
> inmediately before and after this commit)

You were right to be suspicious :o(  I have two almost identical machines
on a KVM switch and I was confused about which machine I was using.  When
I'm wide awake that doesn't happen.  Sorry for the false report.