[pptp-devel] [PATCH] fix segfault on callmgr error handling path

[Avi K. <av...@ar...>]

pptp sometimes crashes in pptp_fd_set and leaves a core dump on my root. 
I think this is due to a signal being caught and longjmp'ed to 
callmgr_main's shutdown label after we are out of that function, causing 
garbage local variables to be used.

RCS file: /cvsroot/pptpclient/pptp-linux/pptp_callmgr.c,v
retrieving revision 1.16
diff -u -w -a -u -p -r1.16 pptp_callmgr.c
--- pptp_callmgr.c      21 Jul 2004 09:05:07 -0000      1.16
+++ pptp_callmgr.c      30 Sep 2004 02:27:08 -0000
@@ -239,6 +239,8 @@ skip_accept: /* Step 5c: Handle socket c
     } while (vector_size(call_list) > 0 || first);
 shutdown:
     {
+       signal(SIGINT, callmgr_do_nothing);
+       signal(SIGTERM, callmgr_do_nothing);
         fd_set read_set, write_set;
         struct timeval tv;
         /* warn("Shutdown"); */
@@ -281,6 +283,8 @@ shutdown:
         vector_destroy(call_list);
     }
 cleanup:
+    signal(SIGINT, callmgr_do_nothing);
+    signal(SIGTERM, callmgr_do_nothing);
     close_inetsock(inet_sock, inetaddr);
     close_unixsock(unix_sock, inetaddr);
     return 0;


(setjmp/longjmp? why, oh why?)

not subscribed.

--
Do not meddle in the internals of kernels, for they are subtle and quick 
to panic.