Re: [Regression] 6.8 - i2c_hid_acpi: probe of i2c-XXX0001:00 failed with error -110

[Hans de Goede <hdegoede@redhat.com>]

Hi Philip,

On 3/19/24 5:52 AM, Philip Müller wrote:
> On 18/03/2024 17:58, Philip Müller wrote:
>> I'm currently developing on the OrangePi Neo-01, which ships with two similar touchpads using the Synaptics driver. On 6.7.10 those two devices get detected normally. On 6.8.1 it seems to error out.
>>
>> I either get none, or at best only one of those two devices.
>>
>> i2c_hid_acpi: probe of i2c-XXX0001:00 failed with error -110
>> i2c_hid_acpi: probe of i2c-XXX0002:00 failed with error -110
>>
>> what would be the best way to debug this?
>>
> 
> I found the regression in commit aa69d6974185e9f7a552ba982540a38e34f69690
> HID: i2c-hid: Switch i2c_hid_parse() to goto style error handling

I just checked that patch and I don't see anyway how that can create
this regression. I assume you did a git bisect ? 

Did you try the last commit in the tree before that commit got added
and verified that that one works where as building a kernel from commit
aa69d6974185e9f itself does not work ?

> When I use the commit before I can rmmod and modprobe in a batch script using a loop without erroring out to -110. Attached the testing script and dmesg log snippets
> 
> #!/bin/bash
> for ((n=0;n<5;n++))
> do
> sudo rmmod i2c_hid_acpi
> sleep 1
> sudo modprobe i2c_hid_acpi --force-vermagic
> sleep 2
> done

Ok, so you did try the commit before and that did work. Are you
sure that aa69d6974185e9f was not actually the last working
commit ?

AFAICT aa69d6974185e9f makes no functional changes, except for
actually propagating the error from i2c_hid_read_register()
rather then hardcoding -EIO. But that should not matter...

Note that commit aa69d6974185e9f is part of a series and
I would not be surprised if some other commit in that series
is causing your problem, but aa69d6974185e9f itself seems
rather harmless.

Regards,

Hans