Re: [PATCH] net: l3mdev: Set OIF flag only when an L3 master exists

[David Ahern <dsahern@kernel.org>]

On 10/26/26 2:30 AM, Daniel Machon wrote:
> Hi Haon,
> 
>> l3mdev_master_ifindex_rcu() may return 0 when the device is not
>> associated with an L3 master. In that case, l3mdev_update_flow()
>> currently sets FLOWI_FLAG_L3MDEV_OIF even though flowi_l3mdev
>> remains 0.
>>
>> Set FLOWI_FLAG_L3MDEV_OIF only when a non-zero L3 master ifindex
>> is returned, keeping the flag consistent with flowi_l3mdev.
>>
>> Signed-off-by: Haon KIM <akrn112@naver.com>
>> ---
>>  net/l3mdev/l3mdev.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/l3mdev/l3mdev.c b/net/l3mdev/l3mdev.c
>> index 5432a5f2dfc8..92b48d6a4301 100644
>> --- a/net/l3mdev/l3mdev.c
>> +++ b/net/l3mdev/l3mdev.c
>> @@ -279,7 +279,8 @@ void l3mdev_update_flow(struct net *net, struct flowi *fl)
>>  		if (dev) {
>>  			if (!fl->flowi_l3mdev) {
>>  				fl->flowi_l3mdev = l3mdev_master_ifindex_rcu(dev);
>> -				fl->flowi_flags |= FLOWI_FLAG_L3MDEV_OIF;
>> +				if (fl->flowi_l3mdev)
>> +					fl->flowi_flags |= FLOWI_FLAG_L3MDEV_OIF;
> 
> Looking at the readers of this flag, it does not look like this does much, so if
> anything this is a no-functional clean-up for net-next, I guess. The commit
> description could be more specific about this. Also, the target tree is missing.
> 

Example of a failing use case?