Re: [PATCH 0/1] Platform: x86: chromeos_laptop - convert to i2c_driver to handle i915 race

[Martin Nordholts <enselic@gmail.com>]

2013/8/5 Benson Leung <bleung@chromium.org>:
> Hi Martin,
>
> Thank you for looking into this problem. Obviously, this is a problem
> Olof and our team has been working on as well.

Hi Benson,

Thank you for spending time reviewing and testing my patch. I
understand your team is busy and am honored you took the time to look
at my code.

> Your driver would result in three failed probes in 2 different drivers
> on my board.

Yes that is one down-side; sometimes the detect() function will be
triggered even though the device is not supported (the return -ENODEV
code paths in my driver). I thought that maybe a simpler driver could
make up for unnecessary detect() calls. However ...

> "Detect" is method #3, but it comes with a
> stern warning : Once again, method 3 should be avoided wherever
> possible. Explicit device instantiation (methods 1 and 2) is much
> preferred for it is safer and faster.

... I must admit I did not see this. I realize that this makes my
patch a dead end. I wish I would have seen that earlier, that would
have saved time for everyone. My apologies.

Regarding the interaction with non-i2c devices in chromeos_laptop, I
was thinking that could perhaps have been solved by putting the i2c
stuff in a separate driver, say chromeos_laptop_i2c, but that point is
irrelevant now that you pointed out that the use of the
detect()-method is discouraged by the documentation.

Thank you again for reviewing the patch and for providing feedback.

BR,
Martin