Re: Regressions can bus with MCP2515 on sama5d3 connected over SPI stop working on kernel > 5.4.271

[Daniel Smolik <smolik@mydatex.cz>]

Hi ,
6.8  is affected   with this bug.  I try bisect the bug.

Regards
             Dan

Dne 12. 03. 24 v 11:39 Linux regression tracking (Thorsten Leemhuis) 
napsal(a):
> On 12.03.24 11:27, Dan Smolik wrote:
>> Dne 12. 03. 24 v 11:19 Linux regression tracking (Thorsten Leemhuis)
>> napsal(a):
>>> On 11.03.24 22:55, Daniel Smolik wrote:
>>>> I have found problem with kernel > 5.4.271 (long term) on sama5d3. When
>>>> I config can bus with:
>>>>
>>>> ip link set can0 type can bitrate 125000 triple-sampling on
>>>>
>>>> and type
>>>> ifconfig can0 up
>>>>
>>>> prompt never returns and cannot run eg. ip a (never returns) existing
>>>> ssh connection works but can´t do new one.
>>>>
>>>>
>>>>     I have my own daughter  board with can bus driver MCP2515 connected
>>>> over SPI bus. There is
>>>> snippet of my DTB:
>>>>
>>>>    can0: can@0 {
>>>>                                           compatible =
>>>> "microchip,mcp2515";
>>>>                                           reg = <0>;
>>>>                                           clocks = <&can0_osc_fixed>;
>>>>                                           interrupt-parent = <&pioD>;
>>>>                                           interrupts = <29
>>>> IRQ_TYPE_EDGE_RISING>;
>>>>                                           spi-max-frequency = <10000000>;
>>>>                                       };
>>>>
>>>> with this all working  perfect  on long term kernels 4.19.  I try switch
>>>> to newer kernel and latest I get working is 5.4.271.   Kernel 5.5.0
>>>> -5.5.6 I can´t  compile and 5.5.14  is affected with bug.
>>> The two important question here are:
>>>
>>> * Is mainline (e.g. 6.8) affected as well? That determines if the
>>> regular developers or the stable team has to look into this.
>> I mean yes,  I tested 6.6.21 and is affected.   6.8 I can test today
>> afternoon.
> Yes, please test 6.8.
>
>>> * Could you bisect the problem (e.g. I assume between 5.4.271 and
>>> 5.4.272)? I'm working on a guide that explains this:
>>> https://www.leemhuis.info/files/misc/How%20to%20bisect%20a%20Linux%20kernel%20regression%20%e2%80%94%20The%20Linux%20Kernel%20documentation.html
>> Yes I try this but it looks like that latest good is 5.4.271 and first
>> bad us 5.5.0  but this kernel fail when building to arm platform.  First
>> which I build for arm is 5.5.14.
>>
>> I try bisecting but when I can't build kernel what can I do ?
> Bisect between 5.4 and 5.5 and during the bisection use "git cherry-pick
> --no-commit <sha1sum>" to apply the fix(es) without committing them
> before you build; once the build is ready, use "git reset --hard" to
> reset things.
>
> But please check 6.8 first; if it is affected we can CC a few
> developers. If you are lucky they have an idea what might be wrong here
> and then no bisection might be needed.
>
> Ciao, Thorsten