The related code is unchanged, so the double free is present here as well.
+++ This bug was initially created as a clone of Bug #1258846 +++
Description of problem:
pmdamemcache crashes with following backtrace:
Core was generated by `perl /var/lib/pcp/pmdas/memcache/pmdamemcache.pl'.
Program terminated with signal 6, Aborted.
#0 0x00007f9464552625 in raise (sig=<value optimized out>) at ../nptl/sysdeps/unix/sysv/linux/raise.c:64
#1 0x00007f9464553e05 in abort () at abort.c:92
#2 0x00007f9464590537 in __libc_message (do_abort=2, fmt=0x7f9464678940 "*** glibc detected *** %s: %s: 0x%s ***\n")
at ../sysdeps/unix/sysv/linux/libc_fatal.c:198
#3 0x00007f9464595e66 in malloc_printerr (action=3, str=0x7f9464676a2e "free(): invalid pointer", ptr=<value optimized out>)
at malloc.c:6336
#4 0x00007f946459897a in _int_free (av=0x7f94648afe80, p=0x7f94648afef8, have_lock=0) at malloc.c:4832
#5 0x00007f945de06d25 in __pmHostEntFree (hostent=0x15cb740) at auxconnect.c:39
#6 0x00007f945e265517 in local_sock (host=0x15cb780 "127.0.0.1", port=11211, callback=0x16ba050, cookie=0) at local.c:209
#7 0x00007f945e273eaa in XS_PCP__PMDA_add_sock (my_perl=0x7633, cv=0x1693560) at PMDA.xs:1099
#8 0x00007f946588e815 in Perl_pp_entersub (my_perl=0x159f010) at pp_hot.c:2888
#9 0x00007f946588cb06 in Perl_runops_standard (my_perl=0x159f010) at run.c:40
#10 0x00007f94658350d8 in S_run_body (my_perl=0x159f010) at perl.c:2435
#11 perl_run (my_perl=0x159f010) at perl.c:2353
#12 0x0000000000400e74 in main (argc=2, argv=0x7ffcf217ad98, env=0x7ffcf217adb0) at perlmain.c:117
Version-Release number of selected component (if applicable):
pcp-3.10.3-3.el6.x86_64
perl-PCP-PMDA-3.10.3-3.el6.x86_64
How reproducible:
No reproducer known so far, abrt reports generated often in customer's environment
Steps to Reproduce:
1.
2.
3.
Actual results:
Program terminated with signal 6, Aborted.
Expected results:
No abort
Additional info:
The same double free bug seem to be present in RHEL 7's, fedora master's pcp as well, upstream propagation needed.
Hypothesis:
pcp-3.10.3/src/perl/PMDA/local.c:
int
local_sock(char *host, int port, scalar_t *callback, int cookie)
{
__pmSockAddr *myaddr;
__pmHostEnt *servinfo = NULL;
...
if ((servinfo = __pmGetAddrInfo(host)) == NULL) {
__pmNotifyErr(LOG_ERR, "__pmGetAddrInfo (%s): %s", host, netstrerror());
goto error; <------------ local.c:158
...
__pmHostEntFree(servinfo); < first free()
...
if (sts < 0) {
__pmNotifyErr(LOG_ERR, "__pmConnect (%s): %s", host, netstrerror());
goto error; <------------ local.c:196
}
...
error:
if (fd >= 0)
__pmCloseSocket(fd);
if (servinfo)
__pmHostEntFree(servinfo); <------------ local.c:209 second free()
We can get to line 209 only from line 196 (just after __pmHostEntFree(), so servinfo is already freed). We cannot get there from line 158 because servinfo would be NULL.
Thanks for the detailed problem report and diagnosis Filip, this is now fixed upstream and will make its way into RHEL7 at the first available opportunity.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://rhn.redhat.com/errata/RHBA-2016-2344.html