[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Permission for 'dev/pmu'



On Sat, Jan 08, 2005 at 07:40:48PM +0100, Christoph Hellwig wrote:
> On Fri, Jan 07, 2005 at 11:48:41PM -0800, Brad Boyer wrote:
> > On Sat, Jan 08, 2005 at 07:32:32AM +0100, Sven Luther wrote:
> > > Do we already have a bug against udev to have the above line per default, or
> > > otherwise finally fix this issue ? 
> > 
> > There were some objections to it the last time it was mentioned. I
> > believe there were some questions about it being a bad idea to
> > make the permissions more open by default. Requesting sleep is
> > restricted to root anyway, but there are some other operations
> > that while less sensitive than sleep should still be restricted.
> 
> Actually marking it some group that users aren't added to by default
> is okay.  Last time people wanted to make it 666 which is definitly
> not okay.

Ok, we are advancing on a fix. We are going to check for :

/proc/device-tree/aliases/via-pmu, to see if the machine indeed supports a
pmu, which should kill this obnoxious error message on non-pmu using machines.

Now, the problem still remains that the acme thingy needs to set the
PMU_IOC_GRAB_BACKLIGHT ioctl on the /dev/pmu device, and thus evidently need
write access. The easy solution to this problem is to set it to a group, which
the console-users get added to when loggin in. Does anyone know of such a group ?

The other solution, would be to use pmud or pbuttonsd to get this done,
instead of doing it oneself, but even then i would say there is a fallback
needed for cases where neither is installed.

Friendly,

Sven Luther





Reply to: