Port-sparc64 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Upgrading from 5.1_STABLE to 5.99.46 on Sun Fire V100



On Sun, Feb 27, 2011 at 04:52:18PM -0500, Michael wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hello,
> 
> On Feb 27, 2011, at 3:32 PM, Martin Husemann wrote:
> 
> >On Sun, Feb 27, 2011 at 07:31:51PM +0100, Jan-Hinrich Fessel wrote:
> >>What happened?
> >>And, what is fan-control?
> >
> >I have no idea what fan control chip is used on that machines, if  
> >anyone
> >finds docs (or a driver for another OS) adding support is probably  
> >simple.
> 
> Apparently I missed half of this - what else is on the fan-control  
> node? If you're lucky there's something useful in  the 'compatible'  
> property. Otherwise the i2c address would be a good hint.

Actually, the interesting thing here is that fan-control is a peer of
the i2c bus, rather than a child of the bus.  The reason we don't find
the spdmem / admtemp / ... is that device_register() appears to get the
wrong (or incomplete?) set of devices to stuff into the i2c-child-devices
devprop.

Here's the device_register() that matters (with some added comments and
printfs by me):

    993         /*
    994          * Check for I2C busses and add data for their direct configurat
ion.
    995          */
    996         if (device_is_a(dev, "iic")) {
    997                 /*
    998                  * XXXrkb: I think if there are no i2c-child-devices, we
    999                  * should be entering i2c devs on the device's node, not
   1000                  * the busnode (which is the parent of the i2c bus).
   1001                  */
   1002                 int busnode = device_ofnode(busdev);
   1003                 printf("%s: adding iic bus at %d\n", __func__, busnode);
   1004
   1005                 if (busnode) {
   1006                         prop_dictionary_t props = device_properties(busd
ev);
   1007                         prop_object_t cfg = prop_dictionary_get(props,
   1008                                 "i2c-child-devices");
   1009                         if (!cfg) {
   1010                                 printf("%s: no i2c child devices\n", __f
unc__);
   1011                                 of_enter_i2c_devs(props, busnode,
   1012                                     sizeof(cell_t));
   1013                         } else {
   1014                                 printf("%s: found i2c child devices\n",
__func__);
   1015                         }
   1016                 }

and the OF device tree looks like so:

        f0069d48: /pci
        [...]
        f007b6bc: /pci/pmu@3
        f007be84: /pci/pmu@3/i2c@0,0
        f007d31c: /pci/pmu@3/i2c@0,0/temperature@0,30 ("i2c-max1617")
        f007d48c: /pci/pmu@3/i2c@0,0/dimm@0,a8        ("i2c-at34c02")
        (more dimms)
        f007d6b4: /pci/pmu@3/i2c@0,0/i2c-nvram@0,a0   ("i2c-at24c64")
        f007e1cc: /pci/pmu@3/i2c@0,0/i2c-nvram@0,a0/idprom@1fd8
        f007e480: /pci/pmu@3/i2c@0,0/motherboard-fru@0,a2 ("i2c-at24c64")
        f007f018: /pci/pmu@3/ppm@0,b3                 ("SUNW,smbus-ppm")
        f007f28c: /pci/pmu@3/beep@0,b2                ("SUNW,smbus-beep")
        f007f3a4: /pci/pmu@3/fan-control@0,c8         ("SUNW,smbus-fan-control")
        [...]

Note that what the i2c controller sees as children on my machine are the
'ppm', 'beep' and 'fan-control' instead of (what I would expect):
 "temparature", N instances of "dimm", "i2c-nvram", "motherboard-fru".

What I'm not 100% sure of is whether the peers of the i2c bus should be
passed as well or not -- it doesn't make sense to me from the OF tree,
but I'm also relatively clueless about OF, so maybe I am missing some-
thing.

--rafal

-- 
  Time is an illusion; lunchtime, doubly so.     |/\/\|           Rafal Boni
                   -- Ford Prefect               |\/\/|      
rafal%pobox.com@localhost


Home | Main Index | Thread Index | Old Index