View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0001751 | unreal | ircd | public | 2004-04-21 01:39 | 2004-04-23 19:06 |
| Reporter | Ron2K | Assigned To | |||
| Priority | normal | Severity | major | Reproducibility | always |
| Status | closed | Resolution | open | ||
| Product Version | 3.2-RC2 | ||||
| Summary | 0001751: remote includes hanging ircd if remote server is down | ||||
| Description | OK. I was playing around with the IRCd last night (I'm using remote includes) and wondered: "What happens if you rehash when the remote server is down?" (I was mainly wondering this because I've got an idea for a remote include feature, which I'll post just now.) So, I switched off the remote server (which runs on the same machine) and rehashed. It gave me "Ron2K is rehashing config file"... and then just sat there. And sat there. I came back two hours later and it was _still_ sitting there. It wasn't responding to ANY commands - in short, the IRCd had completely hung. I eventually had to shut it down via the task manager. Incidentally, /restart doesn't have the same problem, as service.log reports "unable to download file XXXX" and the ircd aborts in the usual manner. Seems to only be /rehash. The above was tested with Unreal3.2-RC2fix running on Windows XP Pro. The remote server used was Apache 1.3.28. | ||||
| Steps To Reproduce | Just switch off the remote server and rehash the ircd. | ||||
| Attached Files | remoteinc.diff (4,302 bytes) | ||||
| 3rd party modules | |||||
|
|
We use cache and we have a whole asynchronous-/rehash system, however... it indeed doesn't seem to be working correctly... * I booted up with an 'include "http://192.168.5.221/blah.conf";' (with just a ban nick block), went fine... * rehashed, went fine. * Then I firewalled the host and /rehash'ed * Now it seems to 'hang'. (for codemastr:) it seems to "hang" in url_do_transfers_async(), in the while(cont) loop to be exact: Breakpoint 1, url_do_transfers_async () at url.c:281 281 CURLMsg *msg; (gdb) n 282 while(CURLM_CALL_MULTI_PERFORM == curl_multi_perform(multihandle, &cont)) (gdb) 285 while(cont) { (gdb) 291 FD_ZERO(&fdread); (gdb) 292 FD_ZERO(&fdwrite); (gdb) 293 FD_ZERO(&fdexcep); (gdb) 294 timeout.tv_sec = 1; (gdb) 295 timeout.tv_usec = 0; (gdb) 297 curl_multi_fdset(multihandle, &fdread, &fdwrite, &fdexcep, &maxfd); (gdb) 299 rc = select(maxfd+1, &fdread, &fdwrite, &fdexcep, &timeout); [*1 sec delay*] (gdb) 301 switch(rc) { (gdb) 304 break; (gdb) p rc $1 = 0 (gdb) n 285 while(cont) { (gdb) 291 FD_ZERO(&fdread); (gdb) 292 FD_ZERO(&fdwrite); (gdb) 293 FD_ZERO(&fdexcep); (gdb) 294 timeout.tv_sec = 1; (gdb) 295 timeout.tv_usec = 0; (gdb) 297 curl_multi_fdset(multihandle, &fdread, &fdwrite, &fdexcep, &maxfd); (gdb) 299 rc = select(maxfd+1, &fdread, &fdwrite, &fdexcep, &timeout); [*1 sec delay*] (gdb) 301 switch(rc) { (gdb) 304 break; (gdb) [and this continues forever] I'm using curl 7.11.1 w/c-ares 1.2.0. |
|
|
(we are looking at it at the moment / tomorrow..) |
|
|
This should be fixed in current CVS. You can also get 'pre2' from www.vulnscan.org/UnrealIrcd/Unreal3.2-pre2.tar.gz If you find any issues/bugs let us know ASAP (since we are preparing for release). |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2004-04-21 01:39 | Ron2K | New Issue | |
| 2004-04-21 11:40 | syzop | Note Added: 0005903 | |
| 2004-04-21 20:28 | syzop | Note Added: 0005906 | |
| 2004-04-21 20:28 | syzop | Status | new => confirmed |
| 2004-04-23 17:39 | syzop | File Added: remoteinc.diff | |
| 2004-04-23 19:06 | syzop | Status | confirmed => closed |
| 2004-04-23 19:06 | syzop | Note Added: 0005919 |
