Enable core file dumps when an application crashes or segmentation faults
-
PR_SET_DUMPABLE ?Hello fellow Netgate community members, can you please help?
Where can we set up application based core file dumps? I can get crash data for system crashes however applications when they do segmentation faults are not generating a crash file anywhere. This I have learned is OS specific. Some applications we can set the folder where they should be dumped however it will not allow them to generate a core dump file anywhere I have set /var/crashes however nothing gets dumped....
-
Processes that segfault like that would usually dump to /root. What are you seeing segfaulting?
-
Thanks for the reply,
The application is 5.8 Squid when I am looking at facebook reels and pfSense GUI at the same time for over 15 mins. Squid development team wants the core file however I can't find it. I have set the file now to dump to /root with.
It doesn't matter if I have caching enabled disabled StoreID on or off. I am hoping that maybe someone in the Squid development team will look at the package and help. The Squid user email gets a lot of responses, they asked for the file to help. :) Weird right it shouldn't be Facebook specific like that.
coredump_dir /root -
Do you see the process coredumping in the system log?
-
No it starts swap and that file goes to cache folder weird right?
-
Not sure what you mean. pfSense starts to SWAP? What file goes to the cache folder?
-
@stephenw10 squid does a swap log squid does not pfSense, it generates a swap.log file or something in the cache folder. After I bound it to the loopback and not just the LAN I have not seen the segmentation fault. It occurred when I was viewing the squid status area the one showing active connections. So I think it needs to be also bound to loopback and not just lan. Again I still don’t feel comfortable that it didn’t creat a log, do I need to chmod, and or user group permissions for the folder?
-
Ah, hmm, no idea what that is then!
If you don't see the process segfaulting in the system log I wouldn't expect to see any core files.
-
n 11 10:29:15 kernel pid 4993 (squid), jid 0, uid 100: exited on signal 6
Jun 11 10:29:15 (squid-1) 4993 FATAL: Received Segment Violation...dying. listening port: 127.0.0.1:3128
Jun 11 10:23:13 kernel pid 43536 (squid), jid 0, uid 100: exited on signal 6
Jun 11 10:23:05 (squid-1) 43536 FATAL: Received Segment Violation...dying. connection: conn749025 local=192.168.1.1:3128 remote=192.168.1.5:59502 flags=1
Jun 11 10:19:00 sshguard 98282This is the error it is a segment fault right?
-
Oh, yes you need Squid to listen on localhost to see the stats from it. That's how it is queried.
-
@stephenw10 so that solved my error, normally would that generate a corefile? Squid developers said I can manually generate a core dump also to get the stack
For example, if you want to abort a Squid worker process that has OS process ID (PID) 12345, you may do something like this: sudo kill -SIGABRT 12345 You can use "ps" or "top" commands to learn PIDs of processes you want to signal.When I did this it also showed no core file.
-
@stephenw10 that swap file it creates is not readable it’s compiled information also. Just a side note
-
If a process aborts and creates a core file you should see something like:
Mar 19 23:14:22 kernel pid 76582 (php-fpm), jid 0, uid 0: exited on signal 11 (core dumped)May 7 15:15:47 fw1 kernel: pid 65129 (pftop), jid 0, uid 0: exited on signal 6 (core dumped) -
@stephenw10 do you know how or what commands are used to manually trigger it for squid?
-
Nope not something I've ever tried. Did you try the command they gave you?
-
@JonathanLee said in Enable core file dumps when an application crashes or segmentation faults:
coredump_dir /root
Where did you set that?
-
@stephenw10 yes it showed blank from command input in GUI it killed the process but no core file it didn’t generate one when I used that command
I am going to try this one recommend by a freeBSD forum user
killall -11 squid worked for me (as root) -
@stephenw10 advanced options on squid configure gui area, it takes when I run squid -k parse also
-
It tries to dump the core for both the main and child squid processes here but fails with:
Jun 12 19:02:08 kernel pid 80887 (squid), jid 0, uid 0: exited on signal 6 (no core dump - bad address) Jun 12 19:01:48 kernel pid 33418 (squid), jid 0, uid 100: exited on signal 6 (no core dump - bad address) -
@stephenw10 that looks like a permissions issue with Squid I wonder if we set it to squid logs location if that would work? Bad address? Does it mean path?