Re: [PATCH net 0/3] Fix broken TC_ACT_REDIRECT from qdiscs

[Daniel Borkmann <daniel@iogearbox.net>]

On 6/30/26 4:37 PM, Sebastian Andrzej Siewior wrote:
> On 2026-06-30 14:33:28 [+0200], Daniel Borkmann wrote:
>> This is an alternative fix to [0] in order to not uglify
>> __dev_queue_xmit() with sprinkled ifdefs given this can be
>> simplified and isolated through a simple test into the BPF
>> redirect helper itself.
>>
>> I've also added a proper BPF selftest, so there is no need
>> to check-in a binary BPF object into selftests given we do
>> have BPF infra for all of this.
> 
> 1/3 makes sense. Assuming we wouldn't have this per-task memory
> assignment, wouldn't then the state from one redirect leak into another?

For the normal/functional case out of tcx / cls_bpf via clsact
not given skb_do_redirect() is called right after return. (For
the full qdisc case it was never working in the first place.)

> For what it's worth:
> Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> 
> Sebastian