This morning I synced my portage and decided to do a -u world, portage wanted to update glibc to 20040808. Everything was fine until I decided to rebuild SSH with skey support after the glibc update. Now whenever i try to ssh into the system, I get: sshd: root@pts/0: relocation error: /lib/libnss_dns.so.2: symbol __res_maybe_init, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference Connection to xxx closed. Reproducible: Didn't try Steps to Reproduce: I'd post my system info but I cannot get to my system at the moment, I'll update this bug with the necessary information tomorrow when I get back to work. Off the top of my head, my cflags are -O2 -mcpu=pentium3 -msse -mmmx -pipe -fomit-frame-pointer and I'm not using nptl.
Is also upgraded my glibc this morning and I didn't recompile ssh but now I can't connect to my server... : djfunk@dgouju djfunk $ ssh rei Connection closed by XXX.XXX.XXX.XXX The exchange of public keys work, but it closes connection after...
Same problem here.
I had the same problem -- I emerged glibc-2.3.4.20040808-r1 and openssh stopped working. When I downgraded to glibc-2.3.3-20040420-r2, ssh started working again.
same exact problem here...will downgrade glibc when i get home and wait for a fix.
ok, well chalk this up to stupidity on my part. restarting sshd fixed it.
It looks like after recompiling ssh, you should restart ssh so ssh uses the new libraries. I believe that fixed it for me. Never hurts to try, but that is if you still have access to the box :P hehe
It also crashed my apache 1.3.31 and mod_php 4.3.9: /usr/sbin/apache: relocation error: /lib/libnss_dns.so.2: symbol __res_maybe_init, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference Syntax error on line 57 of /etc/apache/conf/apache.conf: Cannot load /etc/apache/modules/libphp4.so into server: /lib/libresolv.so.2: symbol __res_maybe_init, version GLIBC_PRIVATE not defined in file libc.so.6 with link time reference I'm downgrading my glibc...
djfunk: i had the exact problem and errors you have. try restarting apache first. that fixed it for me.
the fix for res_init on amd64+ia64 with newer binutils is now amd64+ia64-only. gmsoft made this change earlier today. can everyone having this problem sync up and re-install glibc?
Hi there, I think I found the cause and this is not (may be) exactly glibc bug... you decide. So I did upgrade glibc on a few machines and then lost contact with them (i.e. cannot ssh into them). I was trying to isolate the bug and found out that sshd needs restarting after (this) glibc upgrade. Why? After doing a strace on a misbehaving sshd: Process 13111 attached - interrupt to quit select(6, [3], NULL, NULL, NULL) = 1 (in [3]) accept(3, {sa_family=AF_INET6, sin6_port=htons(32772), inet_pton(AF_INET6, "2001:240:517:1::3", &sin6_addr), sin6_flowinfo=0, sin6_scope_id=0}, [28]) = 4 fcntl64(4, F_SETFL, O_RDONLY) = 0 pipe([5, 6]) = 0 fork(Process 4269 attached ) = 4269 [pid 13111] close(6) = 0 [pid 13111] getpid() = 13111 [pid 13111] getpid() = 13111 [pid 13111] close(4) = 0 [pid 13111] select(6, [3 5], NULL, NULL, NULL <unfinished ...> [pid 4269] close(5) = 0 [pid 4269] close(3) = 0 [pid 4269] setsid() = 4269 [pid 4269] alarm(0) = 0 [pid 4269] rt_sigaction(SIGALRM, NULL, {SIG_DFL}, 8) = 0 ...<snip>... [pid 4269] close(3) = 0 [pid 4269] open("/lib/libnss_dns.so.2", O_RDONLY) = 3 [pid 4269] read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\200\16"..., 512) = 512 [pid 4269] fstat64(3, {st_mode=S_IFREG|0755, st_size=14464, ...}) = 0 [pid 4269] mmap2(NULL, 12696, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40429000 [pid 4269] mmap2(0x4042c000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 0x3) = 0x4042c000 [pid 4269] close(3) = 0 [pid 4269] munmap(0x40417000, 69893) = 0 [pid 4269] writev(2, [{"sshd: kalin [priv]", 18}, {": ", 2}, {"relocation error", 16}, {": ", 2}, {"/lib/libnss_dns.so.2", 20}, {": ", 2}, {"symbol __res_maybe_init, version"..., 101}, {"", 0}, {"", 0}, {"\n", 1}], 10) = 162 [pid 4269] exit_group(127) = ? Process 4269 detached [pid 13111] <... select resumed> ) = ? ERESTARTNOHAND (To be restarted) [pid 13111] --- SIGCHLD (Child exited) @ 0 (0) --- [pid 13111] waitpid(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 127}], WNOHANG) = 4269 [pid 13111] waitpid(-1, 0xbfffe714, WNOHANG) = -1 ECHILD (No child processes) [pid 13111] rt_sigaction(SIGCHLD, NULL, {0x804f500, [], 0}, 8) = 0 [pid 13111] sigreturn() = ? (mask now []) [pid 13111] select(6, [3 5], NULL, NULL, NULL <unfinished ...> [pid 4270] <... write resumed> ) = 5 [pid 4270] write(3, "\0\0\0\27\3", 5) = -1 EPIPE (Broken pipe) [pid 4270] --- SIGPIPE (Broken pipe) @ 0 (0) --- [pid 4270] time([1097982279]) = 1097982279 [pid 4270] getpid() = 4270 [pid 4270] rt_sigaction(SIGPIPE, {0x4024fce0, [], 0}, {SIG_IGN}, 8) = 0 [pid 4270] socket(PF_FILE, SOCK_DGRAM, 0) = 5 [pid 4270] fcntl64(5, F_SETFD, FD_CLOEXEC) = 0 [pid 4270] connect(5, {sa_family=AF_FILE, path="/dev/log"}, 16) = -1 ENOENT (No such file or directory) [pid 4270] close(5) = 0 [pid 4270] rt_sigaction(SIGPIPE, {SIG_IGN}, NULL, 8) = 0 [pid 4270] exit_group(255) = ? Process 4270 detached <... select resumed> ) = 1 (in [5]) close(5) = 0 select(6, [3], NULL, NULL, NULL <unfinished ...> Process 13111 detached So, the culprit is /lib/libnss_dns.so.2 which is part of glibc (`etcat -b /lib/libnss_dns.so.2`). After a lot of Googling, I came up to this thread: http://sources.redhat.com/ml/libc-hacker/2004-08/msg00004.html So, I guess the only way (at the moment) is to restart things that use /lib/libnss_dns.so.2 after a glibc upgrade... Till now openssh and apache2 were found. Any other? How to handle that in portage? The very least is a patch to warn the user (sending after this) after glibc install. Is there anyway to check who uses /lib/libnss_dns.so.2 ?? ldd was of no help.
Created attachment 42004 [details, diff] glibc-2.3.4.20040808-r1.restart_warning.patch This is the least that can be done, a warning. *** Patch untested! ***
Created attachment 42007 [details, diff] glibc-2.3.4.20040808-r1.restart_warning.patch with ebeep and epause Just a quick update of the patch, using ebeep and epause (Thanks Georgi!)
hm, > 1098264395:
hm, > 1098264395: >>> emerge (5 of 13) sys-libs/glibc-2.3.4.20040808-r1 to / I have emerged glibs on 3 servers, but /etc/init.d/ssh restart does not work for me to restart ssh after glibc upgrade this works: > (killall sshd ; killall -9 sshd ; /etc/init.d/sshd zap ; /etc/init.d/sshd start) it means, that there must be big warning about SSH, and how to restart that... I had to call colocation to restart the server
Same here on 17 remote boxes (ugh!). Restart sshd fixes it.
i had this problem too and fixed it _remotely_ by ssh root@myserver "/etc/init.d/sshd restart"
I kept getting comments through my email so i thought i'd close this heh..
