set_multicast_list vs. set_rx_mode

set_multicast_list vs. set_rx_mode

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 467 bytes --]

Hey,

Is it intentional that in the case where set_rx_mode is assigned, you
still need to assign set_multicast_list even if it won't ever be called
as a flag for SIOCADDMULTI?

I was thinking of converting the wireless code to use set_rx_mode and
assign set_multicast_list only if the underlying hardware supports
multicast filtering, and it seems that is well-supported, but it does
seem a bit weird that set_multicast_list degrades to a flag.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: set_multicast_list vs. set_rx_mode

[Patrick McHardy]

Johannes Berg wrote:
> Is it intentional that in the case where set_rx_mode is assigned, you
> still need to assign set_multicast_list even if it won't ever be called
> as a flag for SIOCADDMULTI?
> 
> I was thinking of converting the wireless code to use set_rx_mode and
> assign set_multicast_list only if the underlying hardware supports
> multicast filtering, and it seems that is well-supported, but it does
> seem a bit weird that set_multicast_list degrades to a flag.


Indeed, I missed that. It should check for !dev->set_multicast_list &&
!dev->set_rx_mode before returning -EINVAL.

Re: set_multicast_list vs. set_rx_mode

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 849 bytes --]

On Wed, 2007-08-15 at 14:33 +0200, Patrick McHardy wrote:
> Johannes Berg wrote:
> > Is it intentional that in the case where set_rx_mode is assigned, you
> > still need to assign set_multicast_list even if it won't ever be called
> > as a flag for SIOCADDMULTI?
> > 
> > I was thinking of converting the wireless code to use set_rx_mode and
> > assign set_multicast_list only if the underlying hardware supports
> > multicast filtering, and it seems that is well-supported, but it does
> > seem a bit weird that set_multicast_list degrades to a flag.
> 
> 
> Indeed, I missed that. It should check for !dev->set_multicast_list &&
> !dev->set_rx_mode before returning -EINVAL.

Ok. Want me to send a patch?

And then the expected behaviour is that set_rx_mode will set
IFF_ALLMULTI if it can't honour the list, right?

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]

Re: set_multicast_list vs. set_rx_mode

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 1008 bytes --]

On Wed, 2007-08-15 at 14:33 +0200, Patrick McHardy wrote:
> Johannes Berg wrote:
> > Is it intentional that in the case where set_rx_mode is assigned, you
> > still need to assign set_multicast_list even if it won't ever be called
> > as a flag for SIOCADDMULTI?
> > 
> > I was thinking of converting the wireless code to use set_rx_mode and
> > assign set_multicast_list only if the underlying hardware supports
> > multicast filtering, and it seems that is well-supported, but it does
> > seem a bit weird that set_multicast_list degrades to a flag.
> 
> 
> Indeed, I missed that. It should check for !dev->set_multicast_list &&
> !dev->set_rx_mode before returning -EINVAL.

Hmm. We don't really support multiple unicast addresses so I should
probably not use set_rx_mode. What is the meaning of
dev->change_rx_flags? It seems to be called with IFF_ALLMULTI but if it
is assigned and set_multicast_list is not then you also cannot add
multicast addresses via SIOCADDMULTI.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 190 bytes --]