Re: snapper does not work on PowerBook5,6

[Michael <macallan%netbsd.org@localhost>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Oct 14, 2010, at 2:50 PM, Frank Wille wrote:

> Michael wrote:
>
> > Do you see interrupts when reading or writing to /dev/audio?
>
> Not at all. I have enabled SNAPPER_DEBUG and this is what happens when
> reading:
>
> trigger_input 0xd0057000 0xd0067000 0x1000
> precision: 16
> I2SSetDataWordSizeReg 0x00000000 -> 0x02000200
> reg: 1, 1 96
> I2SSetSerialFormatReg 0x6c01c193 -> 0x82190000
>
> And when writing:
>
> trigger_output 0xd0047000 0xd0057000 0x1000
> precision: 16
> I2SSetDataWordSizeReg 0x02000200 -> 0x02000200
>
> Curiously the system freezes not immediately. I can even log in over  
> the
> network and inspect everything. But as soon as I press a key on the  
> console
> where I typed "cat file > /dev/audio" it freezes.
>
> No such problem when reading from /dev/audio, BTW.
>
>
> > The cardbus problem looks a bit like missing interrupts as well - you
> > should at the very least see an interrupt when inserting a card, even
> > if it doesn't attach.
>
> No interrupt there as well.

Ok, so we either don't enable some interrupts or we enable the wrong  
ones for whatever reason.
Which interrupts does snapper use on your machine? Please check if the  
interrupt the cardbus bridge uses matches the interrupt register and  
the OF node. It wouldn't be the first time that one of them is lying.

> > Leaves the question what could cause /that/ - as far as I can see you
> > have a bog standard UniNorth OpenPIC which works just fine in
> > countless other machines. Wouldn't be the first time that Apple at
> > some point introduced variants that need special handling somewhere.
>
> Maybe there is still a problem with the interrupts, but I think I  
> found
> another problem in snapper(4) through SNAPPER_DEBUG.
> On initialization it prints:
>
> audio0 at snapper0: full duplex, playback, capture
> FCR(0x3c) 0x0x1803440<I2S0EN,I2S0CLKEN>
> /gpio 0xff981cd0
> 0xff982290 modem-reset
> 0xff9823e8 modem-power
> 0xff982540 accelerometer-1
> 0xff982700 accelerometer-2
> 0xff9828c0 headphone-mute
> 0xff982a18 amp-mute
> 0xff982b68 hw-reset
> 0xff982cb8 linein-detect
> 0xff982e78 headphone-detect
> 0xff983048 cpu-vcore-select
> 0xff9831b0 extint-gpio1
> 0xff983390 programmer-switch
> 0xff983528 gpio4
> headphone-mute 0x0
> amp-mute 0x0
> headphone-detect 0x0
> headphone-detect active 0
> headphone-detect intr ffffffff
> audio-hw-reset 0x0
> [...]
>
> The addresses of the GPIOs are not detected. snapper_init() looks for
> properties like "audio-gpio" and "AAPL,address" which do not exist  
> here.
> And to make it worse, it doesn't even check the result! So snapper  
> will
> access address 0 with gpio_read/write... :)
>
> Looking at the OF tree I should use the "name" or "compatible"  
> property for
> "audio-gpio" and "reg" for "AAPL,address". But the contents of "reg"  
> is not
> the final address and has probably to be added to some base address of
> obio?

It /should/ be relative to obio/gpio although sometimes it's relative  
to obio. The whole thing is a mess thanks to Apple being inconsistent  
even withing a single OF variant - my iBook has some gpios with  
addresses relative to obio and others - on the same parent node -  
relative to obio/gpio and the code tries to guess which one it is. So  
please leave existing functionality intact no matter how insane it  
looks, it's there for a reason.
On my iBook the gpio node has an address of 0x50 which is relative to  
obio. Some of its child nodes have addresses < 0x50 which we treat as  
relative to gpio, others have offsets > 0x50 which we treat as  
relative to obio. This works on my iBook G4, it may or may not work on  
other machines with different variants of this insanity.

> > Something completely different - a fairly recent ofwboot.xcf ( from
> > 5.99.29 ) Just Works(tm) on my beige G3 ( with OF patches from an OSX
> > CD it would happily boot a NetBSD install CD ), to my utter surprise.
> > These machines had to be installed via floppy or network thanks to OF
> > bugs.
>
> That's great! I hope I didn't break anything now. Can you also test  
> with my
> patched ofwboot?

I will. Do you have a beige G3 to confirm?
For the record, it's:
boot ide1/@0:,\OFWBOOT.XCF;1 ide1/@0/netbsd
the kernel file name doesn't need to fit OF's insane syntax ( it's  
listed as NETBSD.;1 ) thanks to ofwboot being a little more clever.


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBTLerU8pnzkX8Yg2nAQJd2Qf/TUlxmA8ZxR3ZI0SyIB1d2Bdqb1Q+Qhyh
4GjuqDoFPaOnFyU+uGGjQPoEkNry3mC9jUiRXwig9/LvIo3hXONyPPYYaGxWn+dL
zMub5ma9EWgD8nJm+FDqV4N+NblDQresoMZ0ESj3BtIJJwjD0IZsq/eCxlPJRFO0
NTnYshRJ5cCTr/ITl3pKoueynhuBgB0avxSbDMid5zC0BS1ji4asancA5Lv0Y5k5
90kNhO37Dbmhy065AW3lc03RTgfos2Y0V9icwcWb7PXerNEJZXgzxOUr6O1OTC5l
jdMDX3Rpjx6bvxBmANhnlOwl67BGZTCUNOP68D6adRoDOkfyI9pROA==
=FR0V
-----END PGP SIGNATURE-----