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

Hello,

On Sep 26, 2006, at 19:27, David Young wrote:

> In machdep, I see variables cachelinesize, timebase, printfataltraps,
> altivec, model, and powersave, of which only machdep.powersave=1 looks
> promising.

That's for putting the CPU into power saving mode when idle. Works 
wonders on my G3 but doesn't do a lot on my iBook - I guess the latter 
has a few more heat sources in a smaller case.

>> If not please have a look at your OF device tree, there should be a
>> 'fan' node somewhere, its 'compatible' property should tell us what
>> kind of temperature sensor / fan controller you have ( likely some
>> Analog Devices i2c part, data sheets should be available and writing a
>> driver should be more or less trivial )
>
> I cannot find the 'fan' node.  I have probably missed something.
> See attached dmesg, ofdump, and ofdump -p output.  I am using ofdump
> from <http://mickey.usask.ca/dyoung/ofdump.c>.  BTW, I tried Matt 
> Thomas'
> ofdump2.c; it yields the same output.

Can't find a node for /any/ fan controller. So they apparently didn't 
bother adding one which makes it impossible to autodetect them right 
away. We can however still use the <driver> at iic? addr <whatever> 
attachment and cross fingers. This isn't exactly well tested though and 
rather error-prone since i2c-addresses aren't unique ( since there are 
only 7 resp. 10 significant bits ) so just tossing all possible 
<whatever> at iic? into the kernel config may or may not work since not 
all i2c devices can be properly detected ( we can detect /something/ in 
those cases but can't really make sure it is exactly what we're looking 
for, so when adding <something> at iic? you better know what's there ). 
The current code doesn't know how to handle the i2c-bus at the PMU - 
I'm going to change that though. I already have half-assed support for 
iic* at cuda?. The other problem is that there are so many i2c-buses in 
UniNorth-based *Books ( two ki2c which can each control at least two 
buses, then i2c at pmu... )

Maybe I should add a debug option to the iic code which makes it scan 
the bus when attaching and report all addresses where something 
responds.

> BTW, is there specific support for the ATI Rage Mobility framebuffer?
> Just curious.  ofb is tolerable with OFB_ENABLE_CACHE.

There is no unified 'ATI Rage Mobility' driver. ATI used this name for 
both mach64 and Rage 128 based chipsets, machfb supports some of the 
mach64-based ones ( notable not the Rage LT ) but there is currently no 
specific kernel support for anything Rage 128 based. Shouldn't be hard 
to do, what's missing is - as usual - the right hardware in the right 
hands.
You have a Rage Mobility M3 which is a Rage 128.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iQEVAwUBRWx7aMpnzkX8Yg2nAQIKDwgAszX1PQa0eMFHyKSz6PJUGrEUvXP4nDne
WBprnox2451U8E30AdIuyRGy0Yn+5kYfecGuolA/jzkf1nXiiHH40OuPS5cmpvvq
EZUO8lDBfJgeRYB6Api+KhSVyZZWCjDCOI+lPRS33Gggd9mdZTXQxWcFoirlCMGJ
0aMmrJxaL/QwpegSBuXGeRAz6+rYf+O+n05Qq3DkC9mDcFCtSB6hVdlpL1l/8mlg
ygxsnZaAtRFJ+f2sMUBcg1hxXIJwYix+b8WbpelXT6EkupBC5ZnTpQuwxamBZwtz
U5RXWbtOgofWfGibHH4++88COm1aF9bi+AJIGY5iQnY9siQyhpgDaQ==
=0Ace
-----END PGP SIGNATURE-----