Re: [Familiar] 0.8.4 udev rules - was: Re: v0.8.4-rc3 - wifi problem which doesn't exist underkernel 2.6.15-hh0

Hi Rene,
When I reviewed the /etc/pcmcia/hostap_cs.conf file - I saw no 
delineation between any of the cards there - They were all set to 
ignore_cis_vcc=1

I understand your point about it potentially damaging a card, but this 
is an issue that effects most if not all hostap_cs users - it is one of 
the most common questions on the listservs, since the 0.8.4-rc2.3 
release and certainly previous images seem to have taken a 'peanut 
butter' approach to solving it - yes, the cards were expressly named, 
but otherwise, it effected all of them. (as a note, we could expressly 
name all the cards via udev rules, it would just be a bit of work to 
transfer that over and from a maintenance perspective it is not as clean)

if modprobe.d is the proper place for it, that's fine - but the familiar 
images did not seem to have fully switched over to modprobe.d -- 
/etc/modutils is no longer parsed, from all the testing I have done.

I still don't understand why this worked in Matt Reimer's corrected 
images of 0.8.4-rc1 - there were certainly no udev rules present, I can 
only assume that it was a slightly different option set in the kernel 
that allowed /etc/pcmcia/hostap_cs.conf to be parsed, but I do not know 
what that would be.

If at all possible, I think we should have a solution for this that 
allows end users (who may not know the first thing about writing udev 
rules) to be able to use their cards with familiar out-of-the-box.  If 
that is not possible, then we need to communicate clearly the option - 
wiki, I guess... maybe post-install documentation for the udev systems?

j.


Michal Panczyk wrote:
> Hi Rene.
> I haven't thought about that - I do agree that is is better to make is
> "use at own risk" than to know that "familiar linux fired somebodies
> card":).
>
> Just one thing I found while trying to make my card working.
> In /etc/pcmcia/hostap (or something like that - I don't remember the
> exact path and file name) there is a list of cards handled by the
> hostap module. As far I understood that list it also has distinction
> of the 3.3 v and 5 v cards - there is a line there with something like
> opts="ignore_cis_vcc=1". So I my guess is that we have the list
> already so I it easly to convert it to a "safe config". With the udev
> rule solution it is not the clearest/easiest solution but it may work.
> I don't if that would be possible with the modutils.d way - I have not
> been going too deep into that.
>
> Sorry for generalizing but I don't have my ipaq anywhere near me....
>
> Michal
>
> On 8/12/06, Rene Wagner <rw-CN5wO63fgwogsBAKwltoeQ@xxxxxxxxxxxxxxxx> wrote:
> > Hi Joshua,
> >
> > On Thu, 2006-08-10 at 13:52 -0700, Joshua Layne wrote:
> > > ACTION=="add", DEVICE=="hostap_cs", SYSFS{func_id}=="0x06", 
> > RUN+="/sbin/modprobe hostap_cs ignore_cis_vcc=1"
> > >
> > > That is all my rule consists of and it works for the Netgear MA701
> > >
> > > We are trying to not enter every single card in the udev rules (as was
> > > done for /etc/pcmcia/hostap_cs.conf in previous versions) - hopefully
> > > the flexibility of udev will allow us a bit more abstraction.
> > >
> > > I would very much like to see this added to 0.8.4, but I have no 
> > control
> > > over that - I think Rene or Erik would have to include it.
> >
> > The problem with this is that the VCC checking the driver does is a
> > safety measure. Setting ignore_cis_vcc=1 may result in operating
> > a 3.3V only card at 5V which will likely fry the card.
> >
> > I'm fine with documenting this as a "use at your own risk" workaround
> > but I'm a bit reluctant to enable it by default. I'm also going to
> > remove similiar entries from other configuration files unless someone
> > comes up with a good explanation why those would be harmless.
> >
> > Don't get me wrong, I appreciate the work you've done on this and
> > would love to make things work out of the box, but not at the risk
> > of frying anyone's hardware.
> >
> > On a related note, I don't think a udev rule is the right place to
> > set options passed to kernel modules. This should probably go in
> > a file in /etc/modutils (or /etc/modprobe.d if the modprobe from
> > module-init-tools doesn't parse /etc/modutils/*conf any more).
> >
> > Regards,
> >
> > Rene
> >
> > --
> > Rene Wagner
> > rw at handhelds dot org
> > 4F33 7FD7 93B3 166B BADA  D6F8 71A1 FEA8 58B4 36D0
> >
> >
> > _______________________________________________
> > H2200-port mailing list
> > H2200-port-CN5wO63fgwogsBAKwltoeQ@xxxxxxxxxxxxxxxx
> > https://handhelds.org/mailman/listinfo/h2200-port
> _______________________________________________
> H2200-port mailing list
> H2200-port-CN5wO63fgwogsBAKwltoeQ@xxxxxxxxxxxxxxxx
> https://handhelds.org/mailman/listinfo/h2200-port