Re: [PATCH 6.18.y] can: isotp: fix timer drain order, wakeup handling and tx_gen ordering

On 06.08.26 19:40, Sasha Levin wrote:
> On Thu, Aug 06, 2026 at 02:09:36PM +0200, Oliver Hartkopp wrote:
>>
>>
>> On 06.08.26 13:36, Sasha Levin wrote:
>>> The skb->hash echo tag doesn't work here.  isotp_set_skb_uid() is only
>>> called from the TX paths, never from isotp_send_fc(), so FC loopback
>>> echoes arrive with skb->hash == 0 and alias the so->cfecho == 0 "nothing
>>> pending" sentinel.  Upstream is fine because d4fb6514ff8e stamps the UID
>>> in can_receive(), but that commit isn't on 6.18.
>>>
>>
>> Ouch! Thanks for catching this!!
>>
>> I just focused on isotp.c - my bad.
>>
>> What would be your suggestion to solve this issue?
>>
>> 1. create a 2-patch series which applies d4fb6514ff8e first?
>> 2. turn it back to the original if (so->cfecho != *(u32 *)cf->data) code?
>>
>> The latter was tagged by sashiko-bot to be not safe enough.
> 
> I think that just a 2 patch series would make the most sense. Thanks!
> 

I just posted the v2 patch series for 6.6.y , 6.12.y and 6.18.y

Btw. I double checked that the former integration would have been 
working too, as we have a special case for those echo skbs, that get 
their uid already in isotp.c and not in can_receive() to detect the echo 
of self generated skbs.

Anyway the solution with two patches is much better as it makes the 
isotp.c code more look like the mainline HEAD for ongoing maintenance.

I'll start the work for 5.10.y 5.15.y 6.1.y too - looking forward to the 
FAIL email to get the --in-reply-to Tag ;-)

Many thanks and best regards,
Oliver