[LTP] [PATCH v2 8/8] Rewrite userns08.c using new LTP API

[Petr Vorel]

Hi Andrea,

> Hi Petr,

> I can check it, but it will probably take some time because I'm not familiar
> with that particular scenario. Can we go ahead with the patch-set anyway?

OK, I'll implement the fix in the library.
And for sure this is not a blocker for your patchset.
I've just haven't finishing reviewing it.

Kind regards,
Petr

> Andrea

> On 3/23/22 13:54, Petr Vorel wrote:
> > > Hi!
> > > > ./userns08 -i50
> > > > tst_kconfig.c:82: TINFO: Parsing kernel config '/proc/config.gz'
> > > > tst_test.c:1456: TINFO: Timeout per run is 0h 05m 00s
> > > > userns08.c:65: TPASS: Denied write access to ./restricted : EACCES (13)
> > > > userns08.c:65: TPASS: Denied write access to ./restricted : EACCES (13)
> > > > userns08.c:65: TPASS: Denied write access to ./restricted : EACCES (13)
> > > > userns08.c:65: TPASS: Denied write access to ./restricted : EACCES (13)
> > > > userns08.c:65: TPASS: Denied write access to ./restricted : EACCES (13)
> > > > userns08.c:36: TBROK: clone3 failed: ENOSPC (28)
> > > > Something needs to be closed after each run.
> > > ENOSPC means that we created too many user namespaces. The problem is
> > > likely that we are creating the namespaces faster than they are being
> > > asynchronously cleaned up in the kernel. Adding sleep(1) to the
> > > clone_newuser() function gives kernel enough time to clean the
> > > namespaces and the test works with any -i. Also note that we get the
> > > exact same failure if we execute the test a few times in a row, i.e.
> > > for i in `seq 10`; do
> > > 	./userns08
> > > done
> > +1

> > > The original test fails in the same way, so while it should be fixed,
> > > it's not really reason to block this patchset.
> > Agree (I forget to write I suspected the problem wasn't new in this patchset).

> > > And the only correct fix would be retrying the clone() on ENOSPC in the
> > > SAFE_CLONE().
> > +1. I suppose Andrea will have look into it (otherwise I'll do it).

> > Kind regards,
> > Petr