From mboxrd@z Thu Jan  1 00:00:00 1970
From: Jan Stancek <jstancek@redhat.com>
Date: Wed, 12 Jun 2019 10:25:49 -0400 (EDT)
Subject: [LTP] [PATCH v2] syscalls/ioctl_ns0[156]: align stack and wait
 for child
In-Reply-To: <20190612135929.GA3048@rei.lan>
References: <916c20b9a379badd37a85aa1e1339566c9807d23.1560248542.git.jstancek@redhat.com>
 <20190612135929.GA3048@rei.lan>
Message-ID: <278919491.27726511.1560349549390.JavaMail.zimbra@redhat.com>
List-Id: <ltp.lists.linux.it>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ltp@lists.linux.it


----- Original Message -----
> Hi!
> > Test crashes (SIGBUS) when using child stack have been observed for
> > ioctl_ns01. Align stack to 64 bytes for all testcases using clone,
> > which should work for any arch.
> 
> Looking at the rest of the test it seems that all of them use malloc()
> to allocate the child stack and depends on the libc to align the
> buffers, maybe it would be easier to change these tests to use malloc()
> as well.

Default alignment is not enough:
  Alignment:                              2 * sizeof(size_t) (default)
       (i.e., 8 byte alignment with 4byte size_t). This suffices for
       nearly all current machines and C compilers. However, you can
       define MALLOC_ALIGNMENT to be wider than this if necessary.

I'm guessing most of tests cross M_MMAP_THRESHOLD, and get page alignment
from mmap. But should we rely on that?

How about posix_memalign()?

> 
> > Add SIGCHLD to clone flags, so that LTP library can reap all children
> > and check their return code.  Also check ltp_clone() return value.
> > 
> > Suppress warning for unused *arg in child().
> 
> The rest is OK.
> 
> --
> Cyril Hrubis
> chrubis@suse.cz
>