[CRIU] [PATCH 05/13] parasite-syscall: Add parasite_[init|fini]_threads_seized helpers

[Cyrill Gorcunov]

On Wed, Nov 07, 2012 at 02:07:33PM +0400, Pavel Emelyanov wrote:
> On 11/06/2012 08:53 PM, Cyrill Gorcunov wrote:
> > 
> > We will use them from crtools code to save and restore blocked
> > signals mask of threads.
> > 
> > Signed-off-by: Cyrill Gorcunov <gorcunov at openvz.org>
> > ---
> >  include/parasite-syscall.h |    4 ++++
> >  parasite-syscall.c         |   44 ++++++++++++++++++++++++++++++++++++++++++++
> >  2 files changed, 48 insertions(+), 0 deletions(-)
> > 
> 
> 
> > +int parasite_fini_threads_seized(struct parasite_ctl *ctl, struct pstree_item *item)
> > +{
> > +	struct parasite_thread_args *args = parasite_args(ctl, *args);
> > +	int ret = 0, i;
> > +
> > +	for (i = 0; i < item->nr_threads; i++) {
> > +		if (item->pid.real == item->threads[i].real)
> > +			continue;
> > +
> > +		args->tid = item->threads[i].real;
> > +		ret = parasite_execute_by_pid(PARASITE_CMD_FINI_THREAD, ctl,
> > +					      item->threads[i].real);
> > +		if (ret && ret != -ENOENT) {
> 
> 1. Why -ENOENT check?
> 2. For the last thread reporting ENOENT this error will be reported up w/o log message

Imagine you have 5 threads, then when you've been init'ing them the procedure
failed at 3rd thread, thus only 3 threads will be found on fini call for
4 and 5 there will be -ENOENT, so we should not print any error here
since we're on "curing" path and main error already reported when we
were trying to init the threads.

	Cyrill