Date: Thu, 7 Nov 1996 16:29:51 -0800 (PST) From: asami@FreeBSD.ORG (Satoshi Asami) To: wpaul@skynet.ctr.columbia.edu Cc: current@FreeBSD.ORG Subject: Re: yp_next failure Message-ID: <199611080029.QAA07049@vader.cs.berkeley.edu> In-Reply-To: <199611071653.LAA07481@skynet.ctr.columbia.edu> (message from Bill Paul on Thu, 7 Nov 1996 11:53:13 -0500 (EST))
next in thread | previous in thread | raw e-mail | index | archive | help
I was wrong about the patch making no difference, it did indeed turn
off the messages. (So instead of seeing "oh my gosh...yp_next isn't
feeling well...segmentation fault", it's simply seg faults.)
By the way, I recompiled sendmail with -g and this is what I got.
-------
(gdb) run -q
Starting program: /usr/obj/a/src/usr.sbin/sendmail/src/sendmail -q
Program received signal SIGSEGV, Segmentation fault.
0x2008ebc6 in _yp_dobind ()
(gdb) bt
#0 0x2008ebc6 in _yp_dobind ()
#1 0x2008f458 in yp_first ()
#2 0x2007ac62 in endpwent ()
#3 0x2007a48f in endpwent ()
#4 0x2007a21a in getpwuid ()
#5 0x6e80 in sm_getpwuid (uid=0) at /a/src/usr.sbin/sendmail/src/conf.c:4081
#6 0x17fbe in main (argc=2, argv=0xefbfc2ac, envp=0xefbfc2b8)
at /a/src/usr.sbin/sendmail/src/main.c:284
-------
So you were right about yp_dobind() getting confused. I recompiled
the yp part of libc with -g too, and here is a more detailed report:
-------
(gdb) run -q
Starting program: /usr/obj/a/src/usr.sbin/sendmail/src/sendmail -q
Program received signal SIGSEGV, Segmentation fault.
0x2008f036 in _yp_dobind (dom=0x200c88b0 "mammoth", ypdb=0xefbf9e60)
at /a/src/lib/libc/yp/yplib.c:269
269 clnt_destroy(ysd->dom_client);
(gdb) p ysd
$1 = (struct dom_binding *) 0x5c000
(gdb) p *ysd
$2 = {dom_pnext = 0x0, dom_domain = "mammoth", '\000' <repeats 57 times>,
dom_server_addr = {sin_len = 16 '\020', sin_family = 2 '\002',
sin_port = 6659, sin_addr = {s_addr = 421929088},
sin_zero = "\000\000\000\000\000\000\000"}, dom_server_port = 6659,
dom_socket = -1, dom_client = 0x0, dom_local_port = 14083, dom_vers = -1}
-------
Since clnt_destroy is a macro that takes a pointer and deferences it,
I guess that's the problem.
Line 269 is this:
-------
sock = ysd->dom_socket;
save = dup(ysd->dom_socket);
clnt_destroy(ysd->dom_client); <===
ysd->dom_vers = 0;
ysd->dom_client = NULL;
sock = dup2(save, sock);
-------
in case the patch has shifted things around a bit.
So, I changed this to:
-------
sock = ysd->dom_socket;
save = dup(ysd->dom_socket);
if (ysd->dom_client)
clnt_destroy(ysd->dom_client);
ysd->dom_vers = 0;
ysd->dom_client = NULL;
sock = dup2(save, sock);
-------
(hey, stop laughing). Now things seem to have "calmed down", at least
sendmail/ssh/from/mailq don't seg fault any more. I'll have to beat
on it a little more, I'm now going to reboot the machine with the new
libc and see how it goes.
Satoshi
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611080029.QAA07049>