Bug#622325: linux-image-2.6.38-2-orion5x: Problem With I2C

[arno@natisbad.org (Arnaud Ebalard)]

Hi guys,

While debugging a similar mv64xxx i2c bus lock on a Netgear ReadyNAS 102
NAS (based on a Marvell Armada 370 SoC), I ended up on #622325 bug
report. Reverting eda6bee6c7 makes the issue disappear for me and also
for Simon (in Cc:) for another Marvell-based NAS with the same i2c
controller (its issue [1] appears different but is fixed in the same way
by the revert). Both on current 3.10-rc kernels (not tested previous
one though). 

[1] http://www.spinics.net/lists/arm-kernel/msg254423.html

Note that I also tested d295a86e but it does not fix things for me.

I reported the issue to i2c kernel people and got the following
suggestion from Wolfram in (CC:) [2]: 

  > does nothing fancy either. While debugging I ended up on Debian bug
  > #622325 (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=622325)
  > which made the driver a little less suspect when a revert of commit
  > eda6bee6c7 fixed the issue for me.
  
  Either wait for 3.11 or pick 4243fa0bad551b8c8d4ff7104e8fd557ae848845 
  "(I2C: mv64xxx: fix race between FSM/interrupt and process context")
  from my for-next branch. (Note the other patches from Russell,
  probably all worth to pick). Would be nice if you could pass the info
  to the Debian bug as well.
  
  [2]: http://www.spinics.net/lists/arm-kernel/msg254677.html

I can confirm that 4243fa0b is a valid fix for the bus i2c lock problem
with marvell mv64xxx i2c controller driver (note that I also applied
other patches from Russel, as this commit depends on some of those).
Maybe you should give it a try and consider it a long term solution for
#622325.

Cheers,

a+