IFF_RUNNING and SIOCGMIIPHY support

[David Goodenough]

On Monday 01 March 2004 18:23, Jean Tourrilhes wrote:
> On Sun, Feb 29, 2004 at 01:01:57PM +0000, David Goodenough wrote:
> > Mii-tool does no appear to be supported by hostap, and I was wondering
> > why.  Obviously the mode makes a difference, so saying that a node in
> > master mode is connected makes little sense, but for a managed or
> > ad-hoc mode node the link speed and the presence or absence of an
> > active connection would be useful.
>
> 	Actually, ad-hoc mode is like master mode, you can't test the
> presence/absence of an active connection (i.e. the connection is
> always valid). This is why you would use iwspy and TXDROP in ad-hoc
> mode.
> 	So, this would work only in managed mode.
>
> > As traditionally used yes it is, but a single command line interface that
> > means that in a mixed wired/wireless environment I can find out if all
> > my links are up would be really useful.  I was thinking of this as a
> > simple human interface for diagnosing problems
>
> 	This is irrealistic, wireless does not behave like wired. Your
> wireless link may be up, but if you are associate with the wrong
> Access Point, it will do you no good. Or, the link may be up but
> barely usable because there is too many retries.
> 	Most people setting up long-haul wireless links have their own
> definition of goodness, especially when using ad-hoc mode.
>

Ad-hoc mode is is the one I am particularly interested in, and I suppose that
my definition is that traffic can flow.  It would of course also be useful
to know what speed the link was running at as some of the Quagga 
protocols are bandwidth dependant, and then one of the speeds would
be zero of course being a dead link.

> Jouni wrote :
> > Isn't mii-tool quite Ethernet-specific? What would be the purpose of
> > using it with a wireless driver? I did not notice SIOCGMIIPHY support in
> > any of the kernel tree driver for wireless devices.
>
> 	I even suspect that mii-tool is deprecated in favor of
> ethtool. Some wireless drivers implement the "information" part for
> "ethtool -i".

Well ethtool is new for me, I will have a look at it.

>
> > IFF_RUNNING was always set when the interface was up. Anyway, I have
> > been thinking of adding netif_carrier_{on,off}() calls to the driver
> > based on the association state and ended up doing it now. The
> > development snapshot from the CVS has now code for setting the carrier
> > off when in station mode and not associated. This clears the IFF_RUNNING
> > flag for user space programs and generates a linkwatch event in the
> > kernel.
>
> 	As in ad-hoc mode the card is always successfully associated,
> IFF_RUNNING would always be set, which is what we want. I believe the
> same would apply to master mode. You may want to check what's
> happening in monitor mode and WDS.
>
> 	Have fun...
>
> 	Jean
> _______________________________________________
> HostAP mailing list
> HostAP at shmoo.com
> http://lists.shmoo.com/mailman/listinfo/hostap

David