Re: Regression in inv_mpu6050: 4.6.0-rc5

From: Jonathan Cameron
Date: Wed May 04 2016 - 03:49:18 EST

On 03/05/16 19:54, Crestez Dan Leonard wrote:
> On 05/01/2016 10:58 PM, Jonathan Cameron wrote:
>> On 27/04/16 16:56, One Thousand Gnomes wrote:
>>> On Tue, 26 Apr 2016 18:07:55 -0500
>>> Michael Welling <mwelling@xxxxxxxx> wrote:
>>>> On Tue, Apr 26, 2016 at 11:26:51PM +0100, One Thousand Gnomes wrote:
>>>>> This now causes us to crash and burn on the ASUS T100TA Baytrail/T
>>>>> platforms
>>>> I believe this regression has already been patched.
>>>> Check the latest commits in linux-next.
>>>> See if the latest patches fix your issue.
>>> It does - as this is a regression can we please get those fixes into the
>>> next -rc ?
>> I'm afraid I'm lost in this one - which patch caused the regression and
>> which one fixed it? The only patches I can immediately see in next
>> both introduce and then squish a similar bug, but neither of them
>> has hit Linus' tree yet.
>> Or are we dealing with what was fixed in:
>> c816d9e7 iio: imu: mpu6050: fix possible NULL dereferences
>> I had understood that one as more hypothetical than real...
>> Unfortunately I'm travelling and I suspect that means this will only get
>> in just after the release (so for 4.6.1) once I've confirmed which fixes
>> we actually need to backport.
> Commit
> c816d9e7: iio: imu: mpu6050: fix possible NULL dereferences
> Fixes:
> 33da559f: iio: imu: mpu6050: add mpu6500 register settings
> As far as I can tell this crash will always happen when the device is
> probed via ACPI.

Hi Greg,

A quick heads up.

Unfortunately this regression has come up whilst I'm travelling and
don't have appropriate signing keys with me to do a pull request.
Should be able to do one tomorrow evening as I'll back home.

Turns out the 'possible' is quite common and causing a mess.
Even better the fix actually has a fix as well...

Fastest option is probably a cherry pick of:

c816d9e7: iio: imu: mpu6050: fix possible NULL dereferences
718ba46e: iio: imu: mpu6050: Fix name/chip_id when using ACPI

I'll send you a pull request of my
togreg-in-a-hurry branch tomorrow.

Sorry for these being so late in the cycle.

Anyhow, run for train time.