Re: Permission for 'dev/pmu'

[Sven Luther <sven.luther@wanadoo.fr>]

On Tue, Jan 11, 2005 at 08:11:14PM +0200, Martin-Éric Racine wrote:
> On Tue, 11 Jan 2005, Sven Luther wrote:
> 
> > On Tue, Jan 11, 2005 at 01:21:02PM +0100, Frederic Peters wrote:
> > > Martin-Éric Racine wrote:
> > > 
> > > > > Sjoerd proposed to replace acme_error by g_warning; it *is* trivial.
> > > > > Patch attached.
> > > > 
> > > > This patch fails as-is, because another patch removes that code fragment.
> > >  
> > > Yes, it was done a bit too fast, Sjoerd pointed this in a later mail
> > > and announced he has a correct patch in his local package.
> > 
> > So let's solve this. So, as i understand the current prefered fix is : 
> > 
> >   1) if there is no /proc/device-tree/aliases/via-pmu -> we issue a warning.
> 
> No.  WThis means that we are on hardware not offering PMU so we issue
> nothing and we disable FB_LEVEL features completely.

Yes, that is what i meant. Silly me.

> >   2) if the /dev/pmu device is not existent and/or not writteable
> >      -> we issue a warning, and don't use the feature
> 
> If /proc/device-tree/aliases/via-pmu is not user-writable, we issue a
> warning to syslog to remind the user that appropriate permissions are
> needed to access the feature.

/dev/pmu you mean.

> >   3) if the /dev/pmu is existent and writteable -> we use the feature
> 
> Correct.
>  
> > By default, the /dev/pmu is not writteable, so the feature is disabled, so we
> > need to add a line in Debian.Readme explaining this, and also the relative
> > unsecurity of this approach.
> 
> Agreed.

Ok, this looks sensible, let's do it.

Friendly,

Sven Luther