mv64xxx: I2C bus locked when scanning absent devices on Armada XP

[Gregory CLEMENT]

Hi,

I write this email mainly to let you know that there are some issues
on i2c on the Armada XP (rev A0 and B0) based boards.

What we observed was that if the i2c driver try to access an address
where the device is absent, then the bus is locked. After the timeout
the driver give up, but if we have a lot of i2c client registered,
then the kernel spend a lot of time trying to scan all the
addresses. It is noticeable when using the multiv7_defconfig where
many i2c clients are registered, whereas with mvebu_defconfig we have
fewer i2c clients, and most of them are present on the boards.

Here is an extract of what you can see during a boot:

[    4.127648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[    4.137338] rtc-s35390a 1-0030: rtc core: registered rtc-s35390a as rtc0
[    6.137649] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[    8.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   10.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   12.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   14.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   16.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   18.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   20.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   22.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   24.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   26.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   28.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   30.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   32.137649] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   34.137648] i2c i2c-0: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   36.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   38.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   40.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   42.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   44.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   46.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   48.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   50.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   52.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   54.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   56.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   58.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   60.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   62.137648] i2c i2c-1: mv64xxx: I2C bus locked, block: 1, time_left: 0
[   62.144403] sdhci: Secure Digital Host Controller Interface driver

Then the kernel continue to boot. and except this very annoying delay,
everything else is working on i2c.

If anyone have an idea of the cause of this issue, I would be glad to
have any input.

I continue to investigate it, and for the record v3.12 was not
affected but v3.13 was. So now I am going to bisect.

Gregory
-- 
Gregory Clement, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com