[lxc-devel] Centos template: symlink for /dev/fd missing? - And from Oracle too...

[Michael H. Warfield]

On Thu, 2014-02-13 at 15:22 -0500, Michael H. Warfield wrote:
> On Thu, 2014-02-13 at 14:44 -0500, Dwight Engen wrote:
> > On Thu, 13 Feb 2014 13:32:44 -0500
> > "Michael H. Warfield" <mhw at WittsEnd.com> wrote:
> > 
> > > On Thu, 2014-02-13 at 12:18 -0500, Michael H. Warfield wrote:
> > > > On Thu, 2014-02-13 at 14:02 +0100, Harald Dunkel wrote:
> > > > > Hi folks,
> > > > 
> > > > > Problem in LXC (beta4) running a Centos 6.5 client:
> > > > 
> > > > > 	# cat <(echo hello)
> > > > > 	cat: /dev/fd/63: No such file or directory
> > > > 
> > > > > On a "real" host /dev/fd is a symlink pointing to /proc/self/fd.
> > > > > AFAICS only the altlinux template creates this symlink. Debian
> > > > > seems to provide the link on its own.
> > > 
> > > > > lxc.autodev = 1 does not help. I tried, of course.
> > > > 
> > > > > Question: Should this symlink be added to the lxc-centos template,
> > > > > as it was done for altlinux?
> > > 
> > > > Probably.  I'll have to look at it.  If the CentOS template needs
> > > > it, then the Fedora template would also probably need it as well,
> > > > especially in the autodev case.  The Oracle template also doesn't
> > > > do anything with /dev/fd and it should be very similar to the
> > > > CentOS layout.  I'm not sure what the deal is.  Maybe they're
> > > > providing it in their startup.
> > > 
> > > Oh crud.  Isn't this a fine kettle of fish.
> > > 
> > > I've confirmed that /dev/fd is missing from CentOS and Oracle
> > > containers but not from Fedora (systemd w/ autodev) containers.  Not
> > > exactly what I was expecting.  I haven't tested it with earlier
> > > deprecated upstart or sysvinit versions of Fedora.  I don't see where
> > > it's being created by any of the init scripts.
> 
> > Okay, the symlink is being created in the init script in the initramfs
> > image. We're obviously not going to be running that in the container
> > case, so we'll have to do something in lxc. I'd vote for your
> > suggestion of just having lxc do it where its creating the ptys unless
> > someone has a case where it shouldn't exist (which I doubt since its
> > documented in the kernel's Documentation/devices.txt under "Compulsory
> > links").

> Lovely.  That explains why I couldn't find it then.  Good catch finding
> that documented.  If it's mandatory, that's gets my vote too.

Oh, crud...  This just keeps getting better.

Looks like we're missing a number of symlinks including core, stdin,
stderr, and stdout.  Gonna do one, probably should do them all.  Couple
of those compulsory links probably should be optional or recommended.
X0R -> null?  Really?

I haven't had my nose in that devices.txt file for AGES.  This is almost
embarrassing.  My old Computone Intelliport II serial device driver is
still listed in there with 3 entire character majors assigned to it (71,
72, 73).  What a waste.  We just finally deprecated that thing last
year.  Guess the doco hasn't caught up yet.  :-P

> > > I can add creating that symlink into the CentOS template, there's just
> > > one problem.
> > > 
> > > As you noticed, it also doesn't work if you set autodev = 1.  Adding
> > > the symlink into the CentOS and Oracle templates would fix the static
> > > fs case but the autodev case (which should work with both, even with
> > > sysvinit or upstart) will still be broken.  Maybe when CentOS 7, RHEL
> > > 7, Oracle 7 comes out with systemd (which will require autodev), that
> > > will fix itself.  With devtmpfs, that /dev/fd symlink has to be
> > > created dynamically when the system inits and mounts devtmpfs on /dev.
> > > 
> > > In my mind, that symlink needs to be created at run-time if it doesn't
> > > exist.  A workaround, right now, would be to add that to a hook (post
> > > mount hook maybe?).  Longer term, I think that may need to be in the
> > > lxc-start code itself, right where we're populating the /dev directory
> > > with things like the pty's (is there any case where we would NOT want
> > > to do this?).  We could also add a start-up script into the container
> > > itself to do this as it inits but I would prefer keeping those kinds
> > > of interventions to a minimum.
> > > 
> > > I'd like to hear Dwight's thoughts on this wrt the Oracle template
> > > before we go down either road or another.
> > > 
> > > > > Every helpful comment would be highly appreciated
> > > > > Harri

Regards,
Mike
-- 
Michael H. Warfield (AI4NB) | (770) 978-7061 |  mhw at WittsEnd.com
   /\/\|=mhw=|\/\/          | (678) 463-0932 |  http://www.wittsend.com/mhw/
   NIC whois: MHW9          | An optimist believes we live in the best of all
 PGP Key: 0x674627FF        | possible worlds.  A pessimist is sure of it!

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 465 bytes
Desc: This is a digitally signed message part
URL: <http://lists.linuxcontainers.org/pipermail/lxc-devel/attachments/20140213/5d7ed5ec/attachment.pgp>