[U-Boot] [PATCH] i2c: intel_i2c: SMBus driver PCI addition (e.g. BayTrail)

[Stefan Roese]

On 05.08.2016 09:10, Heiko Schocher wrote:
> Hello Bin,
>
> Am 05.08.2016 um 07:46 schrieb Bin Meng:
>> Simon, Stefan,
>>
>> On Tue, Jul 26, 2016 at 8:13 PM, Stefan Roese <sr at denx.de> wrote:
>>> Hi Simon,
>>>
>>> On 25.07.2016 04:07, Simon Glass wrote:
>>>>
>>>> On 28 June 2016 at 07:44, Stefan Roese <sr at denx.de> wrote:
>>>>>
>>>>> This patch adds support for the SMBus block read/write functionality.
>>>>> Other protocols like the SMBus quick command need to get added
>>>>> if this is needed.
>>>>>
>>>>> This patch also removed the SMBus related defines from the Ivybridge
>>>>> pch.h header. As they are integrated in this driver and should be
>>>>> used from here. This change is added in this patch to avoid compile
>>>>> breakage to keep the source git bisectable.
>>>>>
>>>>> Tested on a congatec BayTrail board to configure the SMSC2513 USB
>>>>> hub.
>>>>>
>>>>> Signed-off-by: Stefan Roese <sr at denx.de>
>>>>> Cc: Bin Meng <bmeng.cn at gmail.com>
>>>>> Cc: Simon Glass <sjg at chromium.org>
>>>>> Cc: Heiko Schocher <hs at denx.de>
>>>>> ---
>>>>> Simon, I'm not sure if this change breaks your Ivybridge targets
>>>>> using the probe part of this driver. Could you please let me
>>>>> know if this works? Or let me know what needs changes here?
>>>>
>>>>
>>>> Yes this breaks booting on link. Something odd is going on because the
>>>> call to set up I2C in ivybridge's print_cpuinfo() returns a very
>>>> strange error -726376.
>>>
>>>
>>> Hmmm, very strange.
>>>
>>>> But I then enabled CONFIG_CMD_I2C and it boots. However 'i2c probe'
>>>> produces a lot of errors like this:
>>>>
>>>> ERROR: len=0 on read
>>>> smbus_block_read (107): dev=0x3b offs=0x0 len=0x1
>>>> smbus_block_read (136): count=0 (len=1)
>>>
>>>
>>> A general question:
>>>
>>> Is the SMBus controller on Ivybridge also exported as PCI device? If
>>> yes, can't we just use the PCI code as done for BayTrail for this
>>> platform as well? And get rid of the platform specific stuff this
>>> way?
>>>
>>> Could you send me the output of "pci 0 long" on this platform?
>>>
>>
>> Do you plan to get this I2C merged in this release? If so, please work
>> this out .. I don't feel comfortable to apply this at present.
>
> Full Ack.

I really would like to see this SMBus support upstream. As other
patches depend on this. Unfortunately I can't test on the Ivybridge
platform. I talked with Simon on #irc some days ago and he
"volunteered" (thanks again) to fix / debug this Ivybridge problem
on his board - perhaps by moving to a PCI based probing there as
well.

Simon, did you find the time to dig into this? Please let me know if
there is something that I can do to help / assist you here.

Thanks,
Stefan