Nut upsdev - Sep 2012 - nut-scanner pthread issue

Hi,

the terminating part of nut-scanner can crash in pthread code, because 
some threads won't exist. For example snmp part:

=============================	if( allow_snmp && nutscan_avail_snmp ) {
		if( start_ip == NULL ) {
			printq(quiet,"No start IP, skipping SNMP\n");
		}
		else {
			printq(quiet,"Scanning SNMP bus.\n");
#ifdef HAVE_PTHREAD
			if( pthread_create(&thread[TYPE_SNMP],NULL,run_snmp,&snmp_sec)) {
				nutscan_avail_snmp = 0;
			}
#else
			dev[TYPE_SNMP] = nutscan_scan_snmp(start_ip,end_ip,timeout,&snmp_sec);
#endif /* HAVE_PTHREAD */
		}
	}
=============================if start_ip==NULL, pthread is not created, but
later...

	if( allow_snmp && nutscan_avail_snmp && thread[TYPE_SNMP]) {
		pthread_join(thread[TYPE_SNMP],NULL);
	}

it calls pthread_join with threan[TYPE_SNMP]==0,
pthread_join(0, ?) behaviour is undefined, on my system undefined=crashes.

I used attached patch to get rid of it.

Cheers,
Michal

/me is back from vacation and going to another one tmrw, so I won't be 
reachable for some time (again)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: nut-2.6.5-pthreadfix.patch
Type: text/x-patch
Size: 1279 bytes
Desc: not available
URL:
<http://lists.alioth.debian.org/pipermail/nut-upsdev/attachments/20120906/989800fb/attachment.bin>

?Hi Michal,

There is indeed a bug, thanks for reporting it.

I am not sure your patch is the right way to fix. I guess it's crashing
because pthread_t is some kind of pointer on your system. But after a quick
search I did not find something saying that an implementation of pthread forbid
thre thread ID to be 0 (correct me if I am wrong). It might even break the
portability of the code. So I guess, setting the appropriate nutscan_avail_*
variable to 0 might be a better way.

Let me know what you think of the attached patch.

Regards,
Fred

--
Eaton Opensource Team - http://opensource.eaton.com



-----------------------------

-----------------------------

________________________________________
From: nut-upsdev-bounces+fredericbohe=eaton.com at lists.alioth.debian.org
[nut-upsdev-bounces+fredericbohe=eaton.com at lists.alioth.debian.org] on behalf
of Michal Hlavinka [mhlavink at redhat.com]
Sent: Thursday, September 06, 2012 3:24 PM
To: NUT Developers
Subject: [Nut-upsdev] nut-scanner pthread issue

Hi,

the terminating part of nut-scanner can crash in pthread code, because
some threads won't exist. For example snmp part:

=============================        if( allow_snmp &&
nutscan_avail_snmp ) {
                if( start_ip == NULL ) {
                        printq(quiet,"No start IP, skipping SNMP\n");
                }
                else {
                        printq(quiet,"Scanning SNMP bus.\n");
#ifdef HAVE_PTHREAD
                        if(
pthread_create(&thread[TYPE_SNMP],NULL,run_snmp,&snmp_sec)) {
                                nutscan_avail_snmp = 0;
                        }
#else
                        dev[TYPE_SNMP] =
nutscan_scan_snmp(start_ip,end_ip,timeout,&snmp_sec);
#endif /* HAVE_PTHREAD */
                }
        }
=============================if start_ip==NULL, pthread is not created, but
later...

        if( allow_snmp && nutscan_avail_snmp &&
thread[TYPE_SNMP]) {
                pthread_join(thread[TYPE_SNMP],NULL);
        }

it calls pthread_join with threan[TYPE_SNMP]==0,
pthread_join(0, ?) behaviour is undefined, on my system undefined=crashes.

I used attached patch to get rid of it.

Cheers,
Michal

/me is back from vacation and going to another one tmrw, so I won't be
reachable for some time (again)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: pthread.patch
Type: text/x-diff
Size: 651 bytes
Desc: pthread.patch
URL:
<http://lists.alioth.debian.org/pipermail/nut-upsdev/attachments/20120906/3776aa5e/attachment.patch>