Re: [PATCH net-next V4 2/2] veth: apply qdisc backpressure on full ptr_ring to reduce TX drops

[Jesper Dangaard Brouer <hawk@kernel.org>]

On 16/04/2025 15.58, Toke Høiland-Jørgensen wrote:
> Jakub Kicinski <kuba@kernel.org> writes:
> 
>> On Tue, 15 Apr 2025 15:45:05 +0200 Jesper Dangaard Brouer wrote:
>>> In production, we're seeing TX drops on veth devices when the ptr_ring
>>> fills up. This can occur when NAPI mode is enabled, though it's
>>> relatively rare. However, with threaded NAPI - which we use in
>>> production - the drops become significantly more frequent.
>>
>> It splats:
>>
>> [ 5319.025772][ C1] dump_stack_lvl (lib/dump_stack.c:123)
>> [ 5319.025786][ C1] lockdep_rcu_suspicious (kernel/locking/lockdep.c:6866)
>> [ 5319.025797][ C1] veth_xdp_rcv (drivers/net/veth.c:907 (discriminator 9))
>> [ 5319.025850][ C1] veth_poll (drivers/net/veth.c:977)
> 
> I believe the way to silence this one is to use:
> 
> rcu_dereference_check(priv->peer, rcu_read_lock_bh_held());
> 
> instead of just rcu_dereference()
> 

Thanks for this suggestion.
Normally, this indicate I should add a rcu_read_lock() section around
for-loop in veth_xdp_rcv(), but this isn't necessary due to NAPI, right?

For background, this is because (1) NAPI is already running with RCU
read-lock held or is it because (2) BH is considered a RCU section?

--Jesper