Re: [PATCH iwl-net v2 2/3] idpf: improve locking around idpf_vc_xn_push_free()

["Tantilov, Emil S" <emil.s.tantilov@intel.com>]

On 3/20/2026 12:42 AM, Sebastian Andrzej Siewior wrote:
> On 2026-03-19 14:13:34 [-0700], Emil Tantilov wrote:
>>   drivers/net/ethernet/intel/idpf/idpf_virtchnl.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
>> index 582e0c8e9dc0..fbd5a15b015c 100644
>> --- a/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
>> +++ b/drivers/net/ethernet/intel/idpf/idpf_virtchnl.c
>> @@ -402,7 +402,9 @@ static void idpf_vc_xn_push_free(struct idpf_vc_xn_manager *vcxn_mngr,
>>   				 struct idpf_vc_xn *xn)
>>   {
>>   	idpf_vc_xn_release_bufs(xn);
>> +	spin_lock_bh(&vcxn_mngr->xn_bm_lock);
>>   	set_bit(xn->idx, vcxn_mngr->free_xn_bm);
> 
> If all of your bit manipulations happen under the same lock you could
> replace atomic set_bit()/ clear_bit() with their non-atomic counter
> parts __set_bit()/ __clear_bit().

We have taken similar approach in the refactor/move to libie:
https://lore.kernel.org/netdev/20251117134912.18566-7-larysa.zaremba@intel.com/

Thanks,
Emil

> 
> The lockless alternative would be find_first_bit() +
> test_and_set_bit() loop. Probably another atomic op for salt. Using the
> __ is free with this change.
> 
>> +	spin_unlock_bh(&vcxn_mngr->xn_bm_lock);
>>   }
> 
> Sebastian