On Thu, Jul 24, 2014 at 02:21:03PM +0200, Gregory CLEMENT wrote: > On 24/07/2014 01:15, Jason Cooper wrote: > > On Thu, Jul 24, 2014 at 01:11:12AM +0200, Benoit Masson wrote: > >> Le 24 juil. 2014 à 00:58, Andrew Lunn <andrew@xxxxxxx> a écrit : > >> > >>>> For the marvell,mv78230-a0-i2c unfortunately i've spend 3 hours > >>>> trying to understand this but it only works with this on the > >>>> ix4-300d :(. There was multiple patch around this and maybe one > >>>> broke the auto-detect part of this, I've tried compiling with some > >>>> 3.10 or lower kernel but no luck here I still have to put this a0 > >>>> option. > >>> > >>> Lets first confirm you have an a0 SoC. > >>> > >>> At boot time, it should print: > >>> > >>> pr_info("MVEBU SoC ID=0x%X, Rev=0x%X\n", soc_dev_id, soc_rev); > >>> > >>> What revision do you have? > >>> > >>> If the auto detect code really is broken, Gregory will likely take a > >>> look. > >> > >> I sure do, > >> > >> confirmed by u-boot output below: > >> > >> U-Boot 2009.08 (Mar 04 2013 - 11:13:04) Marvell version: 2.3.2 PQ > >> U-Boot Addressing: > >> Code:..00600000:006BFFF0 > >> BSS:..00708EC0 > >> Stack:..0x5fff70 > >> PageTable:.0x8e0000 > >> Heap address:.0x900000:0xe00000 > >> Board: DB-78230-BP rev 2.0 Wistron > >> SoC: MV78230 A0 > >> > >> From kernel I get: > >> > >> mvebu-soc-id: MVEBU SoC ID=0x7823, Rev=0x1 > > > > Well, isn't that a peach? :) Gregory? > > A peach?? For me it is either a fruit or a princess, so I am puzzled! Doc Holiday quote from the movie Tombstone. The full quote was "Well, isn't that a peach of a hand?" while sitting at the poker table. > Well about the issue, we patch the device tree for the i2c quirk only for > the openblock AX3. Ahhh, that's right. I need to slow down and dig a bit more. :( > If I remember well it was because the device tree was already written > for this board before we found there was an issue with the A0 version > of the CPU. No, it's because we didn't think any manufacturers would be silly enough to use the a0 in production. That assumption worked out well. :-P > The rule was that for new boards then they have to set the marvell,mv78230-a0-i2c > compatible string. I also didn't expect that we found "new" product using the A0 version. > > We have 3 options now: > > - remove the check on the openblock AX3 board and always try to apply the quirck for A0 version > - add a check for this new board in the mvebu_dt_init function > - let the compatible string marvell,mv78230-a0-i2c in this dts > > I would prefer the option 1 but I fear that Arnd would prefer the 2 other options. I like option 1 and 3. Option 3 is a *correct* description of the hardware, and should be done. Option 1 makes Linux user-friendly for boards that are exactly the same, but changed out SoC stepping mid-production. Option 2 needs to be undone. We shouldn't need to change compiled code for every new board that comes along. Which was the whole point of DT, right? thx, Jason. -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html