[RFC][PATCH 8/8] check files for checkpointability

Serge E. Hallyn serue at us.ibm.com
Mon Mar 2 08:28:57 PST 2009


Quoting Nathan Lynch (ntl at pobox.com):
> On Mon, 2 Mar 2009 07:37:54 -0600
> "Serge E. Hallyn" <serue at us.ibm.com> wrote:
> 
> > Quoting Dave Hansen (dave at linux.vnet.ibm.com):
> > > 
> > > Introduce a files_struct counter to indicate whether a particular
> > > file_struct has ever contained a file which can not be
> > > checkpointed.  This flag is a one-way trip; once it is set, it may
> > > not be unset.
> > > 
> > > We assume at allocation that a new files_struct is clean and may
> > > be checkpointed.  However, as soon as it has had its files filled
> > > from its parent's, we check it for real in __scan_files_for_cr().
> > > At that point, we mark it if it contained any uncheckpointable
> > > files.
> > > 
> > > We also check each 'struct file' when it is installed in a fd
> > > slot.  This way, if anyone open()s or managed to dup() an
> > > unsuppored file, we can catch it.
> > > 
> > > Signed-off-by: Dave Hansen <dave at linux.vnet.ibm.com>
> > 
> > So on a practical note, Ingo's scheme appears to be paying off.  In
> > order for any program's files_struct to be checkpointable right now,
> > it must be statically compiled, else ld.so (I assume) looks up
> > /proc/$$/status.  So since proc is not checkpointable, the result
> > is irreversibly non-checkpointable.
> > 
> > So...  does it make sense to mark proc as checkpointable?  Do we
> > reasonably assume that the same procfile will be available at
> > restart?
> 
> With respect to /proc/$x/* where $x is the pid the restarted task wants,
> is that not a chicken-and-egg problem?

I don't think so... the task will get the pid back (eventually :).  So
sure it won't really be supported yet but we can ignore that for now
imo.

The question is, do we worry about the fact that the procfile contents
might be different at restart (different kernel, etc).

-serge


More information about the Containers mailing list