[LTP] [PATCH v2 3/3] cve/cve-2018-1000001: Add Realpath Buffer Underflow test

[Petr Vorel]

Hi Cyril,

thanks for review. I'll send changes in v3 + see comments bellow.

> I'm not sure if it's a good idea to separate the CVE testcases like
> that, CVE is a CVE and as it looks this has been caused by a change in
> the Linux kernel anyways.

> So maybe we should put this into a syscalls/getcwd/ directory after all
> as the cve identifiers are not really human readable and getcwd05.c
> sounds better to me.
Agree, added there for v3.


> > +++ b/runtest/cve
> > @@ -33,3 +33,4 @@ cve-2017-17052 cve-2017-17052
> >  cve-2017-16939 cve-2017-16939
> >  cve-2017-17053 cve-2017-17053
> >  cve-2018-5803 sctp_big_chunk
> > +cve-2018-1000001_libc_realpath_buffer_underflow cve-2018-1000001

> This long name loos kind of ugly, I would vote for having only the cve
> identifier in the test id here.
OK. + I added it into these runfiles (these where getcwd0[1-4] are + into CVE
itself):
runtest/cve
 cve-2018-5803 sctp_big_chunk
+cve-2018-1000001 getcwd05

runtest/ltplite
runtest/stress.part3
runtest/syscalls
 getcwd04 getcwd04
+getcwd05 getcwd05

...
> > +	if (TST_RET_PTR != NULL) {
> > +		tst_res(TFAIL, "syscall didn't return NULL: '%s'",
> > +				(char *)TST_RET_PTR);

> I doubt that we need the cast to char* here.

We do:
getcwd05.c: In function ‘run’:
../../../../include/tst_test.h:53:40: warning: format ‘%s’ expects argument of type ‘char *’, but argument 5 has type ‘void *’ [-Wformat=]
  tst_res_(__FILE__, __LINE__, (ttype), (arg_fmt), ##__VA_ARGS__)

So is something wrong in my TESTPTR() implementation?


Kind regards,
Petr