IDE on LongTrail

IDE on LongTrail

[Geert Uytterhoeven]

	Hi,

Anyone using a LongTrail with IDE devices?

I just tried adding an IDE disk to mine. The disk is identified correctly, but
I always get `hda: lost interrupt' for any other operation. According to lspci
the sl82c105 uses IRQ 32 (OpenPIC), but /proc/interrupt shows no interrupts
coming in there.

However, if I hardcode the IRQ to be 14 in drivers/ide/sl82c105.c, IDE works
fine (and is even reasonably fast: when using DMA, I get ca. 17 MB from a new
low-end 5400 RPM disk, about the same as the old expensive U2W Viking II on the
Sym53c875).

Apparently other PPC platforms with a Winbond 83C553 (e.g. PowerPlus and PReP)
have some fixup code to disable legacy IRQ mode and route the IDE interrupt to
IRQ 14. But wouldn't it be better to route it directly to the OpenPIC, like the
PCI config space indicates? Guess I'll have to dive in the 83C553 docs...

I'm using 2.4.20-rc1 (from linuxppc_2_4).

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


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: IDE on LongTrail

[Tom Rini]

On Sun, Nov 03, 2002 at 10:40:42PM +0100, Geert Uytterhoeven wrote:

> I just tried adding an IDE disk to mine. The disk is identified correctly, but
> I always get `hda: lost interrupt' for any other operation. According to lspci
> the sl82c105 uses IRQ 32 (OpenPIC), but /proc/interrupt shows no interrupts
> coming in there.

WHich kernel?  It worked on mine as of 2.4.18'ish I believe.  2.5 has
known problems with that driver in general, which RMK has on his to fix
list.

> I'm using 2.4.20-rc1 (from linuxppc_2_4).

I see.  Uh-oh.  I'll have to compile up a new kernel and give it a shot
this week..

--
Tom Rini (TR1265)
http://gate.crashing.org/~trini/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: IDE on LongTrail

[Gabriel Paubert]

Geert Uytterhoeven wrote:
> 	Hi,
>
> Anyone using a LongTrail with IDE devices?
>
> I just tried adding an IDE disk to mine. The disk is identified correctly, but
> I always get `hda: lost interrupt' for any other operation. According to lspci
> the sl82c105 uses IRQ 32 (OpenPIC), but /proc/interrupt shows no interrupts
> coming in there.
>
> However, if I hardcode the IRQ to be 14 in drivers/ide/sl82c105.c, IDE works
> fine (and is even reasonably fast: when using DMA, I get ca. 17 MB from a new
> low-end 5400 RPM disk, about the same as the old expensive U2W Viking II on the
> Sym53c875).
>
> Apparently other PPC platforms with a Winbond 83C553 (e.g. PowerPlus and PReP)
> have some fixup code to disable legacy IRQ mode and route the IDE interrupt to
> IRQ 14. But wouldn't it be better to route it directly to the OpenPIC, like the
> PCI config space indicates? Guess I'll have to dive in the 83C553 docs...
>

AFAIR, the 83C553 IDE interrupts documentation is muddy at best. To be
able to route directly to OpenPIC you would have to be able to connect
interrupt outputs from the 83C553 to the OpenPIC. The only hint that it
may be possible in the 83C553 documentation is that pins IRQC/IRQD are
documented as Input/OD (open drain outputs), while IRQA/IRQB are shown
as inputs only.

However to use this it would be necessary to have a connection from the
IRQC/IRQD to some OpenPIC inputs, which I believe is very unlikely. In
practice, you will have to go through the 8259 cascade. AFAIU the
datasheet and barring typos, native mode will use IRQ14/IRQ15, while
legacy mode gives you more choices. Look strange, but hey, it's based
on PC's so called architecture, so it's baroque by definition ;-)

However, take what I say with a pinch of salt, I don't know much about
the IDE interface since is not used at all on the MVME boards, only the
ISA bridge part oh the W83C553 (or 554).

> I'm using 2.4.20-rc1 (from linuxppc_2_4).

I'm fighting with 2.5 right now, and losing :-(

	Regards,
	Gabriel.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: IDE on LongTrail

[Geert Uytterhoeven]

On Tue, 5 Nov 2002, Gabriel Paubert wrote:
> Geert Uytterhoeven wrote:
> > Anyone using a LongTrail with IDE devices?
> >
> > I just tried adding an IDE disk to mine. The disk is identified correctly, but
> > I always get `hda: lost interrupt' for any other operation. According to lspci
> > the sl82c105 uses IRQ 32 (OpenPIC), but /proc/interrupt shows no interrupts
> > coming in there.
> >
> > However, if I hardcode the IRQ to be 14 in drivers/ide/sl82c105.c, IDE works
> > fine (and is even reasonably fast: when using DMA, I get ca. 17 MB from a new
> > low-end 5400 RPM disk, about the same as the old expensive U2W Viking II on the
> > Sym53c875).
> >
> > Apparently other PPC platforms with a Winbond 83C553 (e.g. PowerPlus and PReP)
> > have some fixup code to disable legacy IRQ mode and route the IDE interrupt to
> > IRQ 14. But wouldn't it be better to route it directly to the OpenPIC, like the
> > PCI config space indicates? Guess I'll have to dive in the 83C553 docs...
> >
>
> AFAIR, the 83C553 IDE interrupts documentation is muddy at best. To be
> able to route directly to OpenPIC you would have to be able to connect
> interrupt outputs from the 83C553 to the OpenPIC. The only hint that it
> may be possible in the 83C553 documentation is that pins IRQC/IRQD are
> documented as Input/OD (open drain outputs), while IRQA/IRQB are shown
> as inputs only.
>
> However to use this it would be necessary to have a connection from the
> IRQC/IRQD to some OpenPIC inputs, which I believe is very unlikely. In
> practice, you will have to go through the 8259 cascade. AFAIU the
> datasheet and barring typos, native mode will use IRQ14/IRQ15, while
> legacy mode gives you more choices. Look strange, but hey, it's based
> on PC's so called architecture, so it's baroque by definition ;-)

According to the IBM LongTrail schematics, PINT[CD] on the W83C553 are
connected to EXT_INT[56] on the Apple Hydra, which are OpenPIC inputs.
The connections are called `IDE_NM_INT_P/S' and 'IDE_NM_INT_S'. So it should be
possible, unless Tatung changed that.

Will do some more W83C553 doc parsing soon :-)

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


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: IDE on LongTrail

[Gabriel Paubert]

Geert Uytterhoeven wrote:

> According to the IBM LongTrail schematics, PINT[CD] on the W83C553 are
> connected to EXT_INT[56] on the Apple Hydra, which are OpenPIC inputs.
> The connections are called `IDE_NM_INT_P/S' and 'IDE_NM_INT_S'. So it should be
> possible, unless Tatung changed that.

Ok, in this case. look at page 125. You have to set bit 11 (LEGIRQ) to
route the interrupts to IRQC for both channels. You can only separately
route each channel (IRQC for channel 0, IRQD for channel1) if you channel
1 in legacy mode it seems. That's at least what I understand.

>
> Will do some more W83C553 doc parsing soon :-)

Parsing is nothing; you also have to interpret the oracle ;-)

	Gabriel.


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: IDE on LongTrail

[Geert Uytterhoeven]

On Tue, 5 Nov 2002, Gabriel Paubert wrote:
> Geert Uytterhoeven wrote:
> > According to the IBM LongTrail schematics, PINT[CD] on the W83C553 are
> > connected to EXT_INT[56] on the Apple Hydra, which are OpenPIC inputs.
> > The connections are called `IDE_NM_INT_P/S' and 'IDE_NM_INT_S'. So it should be
> > possible, unless Tatung changed that.
>
> Ok, in this case. look at page 125. You have to set bit 11 (LEGIRQ) to
> route the interrupts to IRQC for both channels. You can only separately
> route each channel (IRQC for channel 0, IRQD for channel1) if you channel
> 1 in legacy mode it seems. That's at least what I understand.
>
> > Will do some more W83C553 doc parsing soon :-)
>
> Parsing is nothing; you also have to interpret the oracle ;-)

Well, after exactly following the instructions at page 125 I made it work
(using setpci for manual register poking):
  1. Disable ISA interrupts by writing 0x00 to the IDE interrupt routing
     register on the PCI/ISA bridge (used to be 0xef, i.e. IRQs 14 and 15)
  2. Enable native mode for both channels by writing 0x33 to the IDE mode
     register
  3. Set LEGIRQ to route the interrupts for both channels to IRQC.

And `modprobe ide-probe-mod' gave me a working primary IDE channel! Haven't
tested the second channel yet.

The confusing part is that for (2) you need `native mode' (PCI native mode, I
think), while for (3) you need `legacy mode'...

Will cook a patch later...

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


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/