Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Max Staudt]

Hi Geert,

(CC linux-ide)

As for the Amiga defconfig, how about moving from IDE drivers to ATA?

The old IDE stack is slated for removal in less than 2 years, and people should probably move over to libata instead.

How about the following changes?
Is there any Amiga IDE controller left without a libata equivalent?

CONFIG_IDE=n
CONFIG_IDE_GD_ATAPI=n
CONFIG_BLK_DEV_IDECD=n
CONFIG_BLK_DEV_GAYLE=n
CONFIG_BLK_DEV_BUDDHA=n

CONFIG_ATA=y
CONFIG_ATA_VERBOSE_ERROR=y
CONFIG_PATA_GAYLE=y
CONFIG_PATA_BUDDHA=y


Thanks,
Max

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Geert Uytterhoeven]

Hi Max,

On Mon, Oct 7, 2019 at 4:12 PM Max Staudt <max@enpas.org> wrote:
> (CC linux-ide)
>
> As for the Amiga defconfig, how about moving from IDE drivers to ATA?
>
> The old IDE stack is slated for removal in less than 2 years, and people should probably move over to libata instead.
>
> How about the following changes?
> Is there any Amiga IDE controller left without a libata equivalent?
>
> CONFIG_IDE=n
> CONFIG_IDE_GD_ATAPI=n
> CONFIG_BLK_DEV_IDECD=n
> CONFIG_BLK_DEV_GAYLE=n
> CONFIG_BLK_DEV_BUDDHA=n
>
> CONFIG_ATA=y
> CONFIG_ATA_VERBOSE_ERROR=y
> CONFIG_PATA_GAYLE=y
> CONFIG_PATA_BUDDHA=y

Amiga is fine.

Mac and Q40 are not, apparently.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Max Staudt]

On 11/04/2019 12:06 PM, Geert Uytterhoeven wrote:
>> As for the Amiga defconfig, how about moving from IDE drivers to ATA?
>>
>> The old IDE stack is slated for removal in less than 2 years, and people should probably move over to libata instead.
>>
>> How about the following changes?
>> Is there any Amiga IDE controller left without a libata equivalent?
>>
>> CONFIG_IDE=n
>> CONFIG_IDE_GD_ATAPI=n
>> CONFIG_BLK_DEV_IDECD=n
>> CONFIG_BLK_DEV_GAYLE=n
>> CONFIG_BLK_DEV_BUDDHA=n
>>
>> CONFIG_ATA=y
>> CONFIG_ATA_VERBOSE_ERROR=y
>> CONFIG_PATA_GAYLE=y
>> CONFIG_PATA_BUDDHA=y
> 
> Amiga is fine.
> 
> Mac and Q40 are not, apparently.

Sigh. I guess we have to keep CONFIG_IDE then. Thank you for checking!

How about:

CONFIG_BLK_DEV_GAYLE=n
CONFIG_BLK_DEV_BUDDHA=n

CONFIG_ATA=y
CONFIG_ATA_VERBOSE_ERROR=y
CONFIG_PATA_GAYLE=y
CONFIG_PATA_BUDDHA=y

for the multi_defconfig, and the following *only* for amiga_defconfig (as it's useless without drivers on Amiga):

CONFIG_IDE=n
CONFIG_IDE_GD_ATAPI=n
CONFIG_BLK_DEV_IDECD=n


Max

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Christoph Hellwig]

On Mon, Nov 04, 2019 at 12:06:05PM +0100, Geert Uytterhoeven wrote:
> Amiga is fine.
> 
> Mac and Q40 are not, apparently.

Can any users of those legacy ide drivers please get in contact
with Bartlomiej Zolnierkiewicz to ensure we have a tested libata
driver quickly?

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[John Paul Adrian Glaubitz]

On 11/4/19 12:06 PM, Geert Uytterhoeven wrote:
> Amiga is fine.
> 
> Mac and Q40 are not, apparently.

I have not been able to come by a Q40 or 68k-Mac with an
IDE controller, unfortunately.

If the Mac IDE controller is the same as on the PowerBook
3400c, I would be able to test a converted driver as I have
that PowerBook.

Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Brad Boyer]

On Tue, Nov 05, 2019 at 01:14:04AM +0100, John Paul Adrian Glaubitz wrote:
> On 11/4/19 12:06 PM, Geert Uytterhoeven wrote:
> > Amiga is fine.
> > 
> > Mac and Q40 are not, apparently.
> 
> I have not been able to come by a Q40 or 68k-Mac with an
> IDE controller, unfortunately.
> 
> If the Mac IDE controller is the same as on the PowerBook
> 3400c, I would be able to test a converted driver as I have
> that PowerBook.

I have a couple old macs with IDE. I have a PowerBook 190 and a
Performa (a 636? - it's buried away so I'm not 100% sure) both with
IDE drives. I'll try to find time to pull one of them out and see if
they still run. Can Linux run on a system with a 68LC040 these days?
I know there were issues with FPU emulation at various points. Both
of those would lack FPU due to using a 68LC040 chip.

The 3400c is PCI based and so probably isn't at all compatible. It
should be using the same driver as the other PCI models. There is
already a PATA_MACIO configuration option for that style. The 5300
is the one that's basically an upgraded 68k model (it's nearly
identical to the 190 internally).

	Brad Boyer
	flar@allandria.com

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Finn Thain]

Hi Brad,

On Mon, 4 Nov 2019, Brad Boyer wrote:

> 
> I have a couple old macs with IDE. I have a PowerBook 190 and a Performa 
> (a 636? - it's buried away so I'm not 100% sure) both with IDE drives. 
> I'll try to find time to pull one of them out and see if they still run. 
> Can Linux run on a system with a 68LC040 these days? I know there were 
> issues with FPU emulation at various points. Both of those would lack 
> FPU due to using a 68LC040 chip.
> 

All the PB 190 machines that I've come across have a late revision 68LC040 
CPU that is free of errata. Please see,
http://www.mac.linux-m68k.org/docs/faq.php#sec-4.5

The Performa 636 is highly likely to be affected, but you can run Linux 
from initramfs to avoid page faults in code sections. That would allow you 
to test the IDE driver.

Or you can just replace the CPU with a full 68040, since it is socketed. 

But watch out for leaking capacitors and batteries...

-- 

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Brad Boyer]

On Tue, Nov 05, 2019 at 01:33:00PM +1100, Finn Thain wrote:
> All the PB 190 machines that I've come across have a late revision 68LC040 
> CPU that is free of errata. Please see,
> http://www.mac.linux-m68k.org/docs/faq.php#sec-4.5
> 
> The Performa 636 is highly likely to be affected, but you can run Linux 
> from initramfs to avoid page faults in code sections. That would allow you 
> to test the IDE driver.
> 
> Or you can just replace the CPU with a full 68040, since it is socketed. 

I'll try the PB190 first anyway. It should be easier due to not needing
to setup a monitor. I'm not sure if I ever booted Linux on either of
them, since I acquired both about the time I started getting too busy
to spend time on such things. I just found the Performa, and it's actually
a Performa 630CD but I don't see any obvious difference based on the specs.

I just took a look at the macide driver, and it appears to do basically
nothing other than pass a list of addresses into the core ide code. It's
one of the smallest drivers I've ever seen.

> But watch out for leaking capacitors and batteries...

I should pull out every machine in my collection and look for those sorts
of issues. None of them have been checked in at least 5 or 6 years.

	Brad Boyer
	flar@allandria.com

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Finn Thain]

On Mon, 4 Nov 2019, Brad Boyer wrote:

> 
> I'll try the PB190 first anyway. It should be easier due to not needing 
> to setup a monitor. 

I think you can put the 630 into a standard VESA video mode (with MacOS or 
Linux) given the right adapter/cable. I have a pin-out somewhere.

> I'm not sure if I ever booted Linux on either of them, since I acquired 
> both about the time I started getting too busy to spend time on such 
> things. I just found the Performa, and it's actually a Performa 630CD 
> but I don't see any obvious difference based on the specs.
> 
> I just took a look at the macide driver, and it appears to do basically
> nothing other than pass a list of addresses into the core ide code. It's
> one of the smallest drivers I've ever seen.
> 

The fly in the ointment is interrupt handling. There is a theoretical bug. 
(Though it doesn't seem to hurt in practice.)

AFAIK the hardware is publicly undocumented and so we need to do 
experiments like this:
https://github.com/fthain/linux/commit/01405199e8d05500bf458df690027655f526a7fd

My suspicion is that macide_clear_irq() does nothing useful. It's not 
called on a Powerbook 190. Maybe it is needed on a PowerBook 150 and 
Performa 630, maybe not...

> > But watch out for leaking capacitors and batteries...
> 
> I should pull out every machine in my collection and look for those 
> sorts of issues. None of them have been checked in at least 5 or 6 
> years.
> 

None of the machines in my collection have any batteries now. Desoldering 
the Ni-Cd PRAM battery from a Powerbook 14x/16x/170/180 is difficult but 
necessary. Powerbook 150 and 190 are easier (no desoldering needed).

-- 

> 	Brad Boyer
> 	flar@allandria.com
> 
> 

Re: [PATCH] m68k: defconfig: Update defconfigs for v5.4-rc1

[Brad Boyer]

On Tue, Nov 05, 2019 at 03:13:31PM +1100, Finn Thain wrote:
> On Mon, 4 Nov 2019, Brad Boyer wrote:
> > I'll try the PB190 first anyway. It should be easier due to not needing 
> > to setup a monitor. 
> 
> I think you can put the 630 into a standard VESA video mode (with MacOS or 
> Linux) given the right adapter/cable. I have a pin-out somewhere.

I have an adapter. I also own several official Apple monitors of various
styles and capabilities. The problem is desk space. I need to clean before
I have enough room to setup anything bigger than a laptop.

> > I just took a look at the macide driver, and it appears to do basically
> > nothing other than pass a list of addresses into the core ide code. It's
> > one of the smallest drivers I've ever seen.
> > 
> 
> The fly in the ointment is interrupt handling. There is a theoretical bug. 
> (Though it doesn't seem to hurt in practice.)
> 
> AFAIK the hardware is publicly undocumented and so we need to do 
> experiments like this:
> https://github.com/fthain/linux/commit/01405199e8d05500bf458df690027655f526a7fd
> 
> My suspicion is that macide_clear_irq() does nothing useful. It's not 
> called on a Powerbook 190. Maybe it is needed on a PowerBook 150 and 
> Performa 630, maybe not...

My understanding has always been that Apple was really sloppy with how
the hardware handled interrupts. On many models, you can't actually
block interrupts in a sane fashion. Because of that, you usually have
to shut up the interrupts by talking to the actual device.

I would not be surprised in the slightest to see that register write
do absolutely nothing. It might be more interesting to just disable
the clear function for all models and see if it still works. I don't
have a PB 150 to test that style, but I do have the other two.

> > > But watch out for leaking capacitors and batteries...
> > 
> > I should pull out every machine in my collection and look for those 
> > sorts of issues. None of them have been checked in at least 5 or 6 
> > years.
> > 
> 
> None of the machines in my collection have any batteries now. Desoldering 
> the Ni-Cd PRAM battery from a Powerbook 14x/16x/170/180 is difficult but 
> necessary. Powerbook 150 and 190 are easier (no desoldering needed).

That makes sense. I need to do the same. It's just a pain since a few
models won't boot without a battery.

	Brad Boyer
	flar@allandria.com