pcp
[Top] [All Lists]

Re: [pcp] Collateral damage in non-root changes

To: Nathan Scott <nathans@xxxxxxxxxx>
Subject: Re: [pcp] Collateral damage in non-root changes
From: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Fri, 23 Nov 2012 20:58:13 +1100
Cc: pcp@xxxxxxxxxxx
In-reply-to: <1325320647.29899954.1353572267882.JavaMail.root@xxxxxxxxxx>
References: <1325320647.29899954.1353572267882.JavaMail.root@xxxxxxxxxx>
On Thu, 2012-11-22 at 03:17 -0500, Nathan Scott wrote:
> ----- Original Message -----
> > On Wed, 2012-11-21 at 21:45 -0500, Nathan Scott wrote:
> > ...
> > > That doesn't make sense to me (nor do I see how chown on the
> > > pmlogger directory changes this behaviour) - the permissions
> > > of a symlink are irrelevant - as described here...
> > > http://superuser.com/questions/303040/how-do-file-permissions-apply-to-symlinks
> > > 
> > > Which suggests its the directory... hmmm... the only thing I
> > > can think of is a kernel issue. ...
> > 
> > Nod.
> > 
> 
> Also "lsattr -d <dir>" might show some unexpected attribute?
> Can't think what attribute might cause this behaviour though.
> 
> Otherwise maybe some selinux or other security module coming
> into play?  (guessing wildly)

Explanation is here: http://lwn.net/Articles/390323/

Since the "primary" symlink can only ever be owned by "pcp", the fix is
to make $PCP_TMP_DIR/pmlogger owned by "pcp".

Not sure how this all works if the PCP user is something other than pcp,
as per your mail on another topic ... I suspect this is outside the
scope of what can be done in the packaging, so $PCP_TMP_DIR/pmlogger
would have to be chown'd by a packing exitop or by the pmlogger control
scripts.

<Prev in Thread] Current Thread [Next in Thread>