rfkill read / write access needs to be enabled for normal users

rfkill read / write access needs to be enabled for normal users

[Andrea Veri]

hi,

while working on gnome-bluetooth for Debian, I noticed that rfkill's
read / write access needs to be enabled for normal users.

There is a patch for this already:

KERNEL=="rfkill", ENV{ACL_MANAGE}="1"

This is done through an udev rule, but I don't think this is the best
way to have it fixed.

I guess that having a fix included into bluez itself would make the
trick IMHO.

Is there someone working on this already? if not, someone can suggest
me another way to have this fixed without having to use an udev rule?

Thanks for your time,


Andrea

Re: rfkill read / write access needs to be enabled for normal users

[Marcel Holtmann]

Hi Andrea,

> while working on gnome-bluetooth for Debian, I noticed that rfkill's
> read / write access needs to be enabled for normal users.
> 
> There is a patch for this already:
> 
> KERNEL=="rfkill", ENV{ACL_MANAGE}="1"
> 
> This is done through an udev rule, but I don't think this is the best
> way to have it fixed.
> 
> I guess that having a fix included into bluez itself would make the
> trick IMHO.
> 
> Is there someone working on this already? if not, someone can suggest
> me another way to have this fixed without having to use an udev rule?

I have said this before and let me repeat it again, gnome-bluetooth
should not mess directly with the RFKILL subsystem. It is the wrong
approach and that is the reason why normal users don't have uncontrolled
access to /dev/rfkill.

Regards

Marcel

Re: rfkill read / write access needs to be enabled for normal users

[Bastien Nocera]

On Sun, 2009-10-25 at 01:24 +0900, Marcel Holtmann wrote:
> Hi Andrea,
> 
> > while working on gnome-bluetooth for Debian, I noticed that rfkill's
> > read / write access needs to be enabled for normal users.
> > 
> > There is a patch for this already:
> > 
> > KERNEL=="rfkill", ENV{ACL_MANAGE}="1"
> > 
> > This is done through an udev rule, but I don't think this is the best
> > way to have it fixed.
> > 
> > I guess that having a fix included into bluez itself would make the
> > trick IMHO.
> > 
> > Is there someone working on this already? if not, someone can suggest
> > me another way to have this fixed without having to use an udev rule?

This isn't a bug, it's mentioned in gnome-bluetooth's README, along with
the dependencies.

> I have said this before and let me repeat it again, gnome-bluetooth
> should not mess directly with the RFKILL subsystem. It is the wrong
> approach and that is the reason why normal users don't have uncontrolled
> access to /dev/rfkill.

Except that you have failed to come up with a better solution for _now_.
I don't think that we want to wait for you to write rfkilld, or to
change your mind again about how it should be implemented.

bluetoothd didn't even work properly with killswitches before I
mentioned it on the list and on IRC, so I'd appreciate a bit more
honesty in your blanket opinions...

Cheers

Re: rfkill read / write access needs to be enabled for normal users

[Andrea Veri]

Il giorno Sun, 25 Oct 2009 01:24:27 +0900
Marcel Holtmann <marcel@holtmann.org> ha scritto:


> I have said this before and let me repeat it again, gnome-bluetooth
> should not mess directly with the RFKILL subsystem. It is the wrong
> approach and that is the reason why normal users don't have
> uncontrolled access to /dev/rfkill.


yeah, I know. That's why I didnt apply the patch yet, but indeed this
needs to be fixed ASAP, so my question is what can I do in the meantime
to have it fixed? (e.g while upstream implement it directly into bluez)

is the patch proposed by Bastien, the only one available for *now*?

Thanks for your work,


Andrea