[PATCH] tests: check decoding of NETLINK_SOCK_DIAG AF_UNIX messages

[Dmitry V. Levin]

On Wed, Jun 14, 2017 at 11:53:02AM +0800, JingPiao Chen wrote:
> On Wed, Jun 14, 2017 at 05:06:06AM +0300, Dmitry V. Levin wrote:
> > On Wed, Jun 14, 2017 at 03:19:46AM +0300, Dmitry V. Levin wrote:
> > > On Wed, Jun 14, 2017 at 07:49:34AM +0800, JingPiao Chen wrote:
> > > > On Tue, Jun 13, 2017 at 07:10:13PM +0300, Dmitry V. Levin wrote:
> > > > > On Tue, Jun 13, 2017 at 07:04:21PM +0800, JingPiao Chen wrote:
> > > > > > +static void
> > > > > > +test_unix_diag_req(const int fd)
> > > > > > +{
> > > > > > ...
> > > > > > + /* short read of unix_diag_req */
> > > > > > + nlh = nlh0 - (sizeof(*req) - 1);
> > > > > > + memmove(nlh, nlh0 - sizeof(*req), NLMSG_HDRLEN + sizeof(*req) - 1);
> > > > > > +
> > > > > > + rc = sendto(fd, nlh, NLMSG_HDRLEN + sizeof(*req), MSG_DONTWAIT,
> > > > > > +    NULL, 0);
> > > > > > +
> > > > > > + printf("sendto(%d, {{len=%u, type=SOCK_DIAG_BY_FAMILY"
> > > > > > +       ", flags=NLM_F_REQUEST, seq=0, pid=0}, {family=AF_UNIX, %p}}"
> > > > > > +       ", %u, MSG_DONTWAIT, NULL, 0) = %s\n",
> > > > > > +       fd, NLMSG_HDRLEN + (unsigned int) sizeof(*req),
> > > > > > +       NLMSG_DATA(nlh) + 1,
> > > > > > +       NLMSG_HDRLEN + (unsigned int) sizeof(*req),
> > > > > > +       sprintrc(rc));
> > > > > > +}
> > > > > >
> > > > > > When (sizeof(*req) - 1 - sizeof(*family) > DEFAULT_STRLEN), the test
> > > > will
> > > > > > fail.
> > > > >
> > > > > Why?  In this part of the test strace is expected to print a pointer,
> > > > > not a string, so it shouldn't be affected by DEFAULT_STRLEN.
> > > > This fault appear when call test_inet_diag_req.
> > > > 
> > > > Call stack is decode_inet_diag_req -->> decode_family -->> printstrn
> > > > len = sizeof(*req) - 1 - sizeof(*family),
> > > > When sizeof(*req) - 1 - sizeof(*family) > DEFAULT_STRLEN, umoven will
> > > > return true, because: (util.c: 770)
> > > > size = max_strlen + 1;
> > > > if (size > len)
> > > > size = len;
> > > 
> > > I see.  Would it be better to print the address instead of the string
> > > in case of umove failure, e.g.
> > > 
> > > 	if (len < sizeof(msg))
> > > 		return false;
> > > 
> > > 	if (umove_or_printaddr(tcp, addr, &msg))
> > > 		return true;
> > 
> > OK, I've changed this parser a bit and pushed it again to ldv/netlink
> > branch, please have a look.
> Ok, thank you.

... and again, amending the last commit to simplify subsequent
netlink_sock_diag test changes.


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20170614/96ec3764/attachment.bin>