ypbind hangs as of current from midday yesterday (kern+user)

Thread: ypbind hangs as of current from midday yesterday (kern+user)

Search this list only Search all lists
ypbind hangs as of current from midday yesterday (kern+user)
	2007-06-25 22:08:37
Hi all, Since upgrading to yesterdays current my NIS server (xen domu) won't start ypbind. I have tried reinitializing the yp database and files with no change. All the other NIS relevant processes are starting fine. NFS is working better than ever. What can I do to provide more details? I've not had to debug a process before so any information that can enable me to help would be appreciated. Sarton

Re: ypbind hangs as of current from midday yesterday (kern+user)
France	2007-06-26 11:17:26
On Tue, Jun 26, 2007 at 01:08:37PM +1000, Sarton O'Brien wrote: > Hi all, > > Since upgrading to yesterdays current my NIS server (xen domu) won't start > ypbind. > > I have tried reinitializing the yp database and files with no change. All the > other NIS relevant processes are starting fine. NFS is working better than > ever. > > What can I do to provide more details? I've not had to debug a process before > so any information that can enable me to help would be appreciated. Can you ping your ypbind client when this happens ? I would first start with tcpdump on both lo0 and the network interface ... -- Manuel Bouyer <bouyerantioche.eu.org> NetBSD: 26 ans d'experience feront toujours la difference --

Re: ypbind hangs as of current from midday yesterday (kern+user)
	2007-06-26 18:54:22
On Wed, 27 Jun 2007 02:17:26 am Manuel Bouyer wrote: > On Tue, Jun 26, 2007 at 01:08:37PM +1000, Sarton O'Brien wrote: > > Since upgrading to yesterdays current my NIS server (xen domu) won't > > start ypbind. > > > > I have tried reinitializing the yp database and files with no change. All > > the other NIS relevant processes are starting fine. NFS is working better > > than ever. > > > > What can I do to provide more details? I've not had to debug a process > > before so any information that can enable me to help would be > > appreciated. > > Can you ping your ypbind client when this happens ? > I would first start with tcpdump on both lo0 and the network interface ... It is the server I am trying to start ypbind on. When initiating ypbind via an ssh session, I can ping the server and the ssh session I am in is fine but I can't ssh in from another console. As soon as I ^C the ssh client in the other console logs in. Loopback is seeing: 09:44:50.049092 IP (tos 0x0, ttl 64, id 21366, offset 0, flags [none], length: 164, bad cksum 0 (->28d1)!) localhost.65458 > localhost.sunrpc: UDP, length: 136 09:44:50.049290 IP (tos 0x0, ttl 64, id 21367, offset 0, flags [none], length: 148, bad cksum 0 (->28e0)!) localhost.exp1 > localhost.1020: UDP, length: 120 09:44:50.049429 IP (tos 0x0, ttl 64, id 21368, offset 0, flags [none], length: 56, bad cksum 0 (->293b)!) localhost.1020 > localhost.exp1: [bad udp cksum b3df!] UDP, length: 28 09:44:50.049453 IP (tos 0x0, ttl 64, id 21369, offset 0, flags [none], length: 164, bad cksum 0 (->28ce)!) localhost.65458 > localhost.sunrpc: UDP, length: 136 09:44:50.049591 IP (tos 0x0, ttl 64, id 21370, offset 0, flags [none], length: 148, bad cksum 0 (->28dd)!) localhost.exp1 > localhost.1020: UDP, length: 120 09:44:50.049716 IP (tos 0x0, ttl 64, id 21371, offset 0, flags [none], length: 56, bad cksum 0 (->2938)!) localhost.1020 > localhost.exp1: [bad udp cksum 73df!] UDP, length: 28 09:44:56.099756 IP (tos 0x0, ttl 64, id 21488, offset 0, flags [none], length: 164, bad cksum 0 (->2857)!) localhost.65458 > localhost.sunrpc: UDP, length: 136 09:44:56.099934 IP (tos 0x0, ttl 64, id 21489, offset 0, flags [none], length: 148, bad cksum 0 (->2866)!) localhost.exp1 > localhost.1020: UDP, length: 120 09:44:56.100066 IP (tos 0x0, ttl 64, id 21490, offset 0, flags [none], length: 56, bad cksum 0 (->28c1)!) localhost.1020 > localhost.exp1: [bad udp cksum 33df!] UDP, length: 28 09:44:56.100089 IP (tos 0x0, ttl 64, id 21491, offset 0, flags [none], length: 164, bad cksum 0 (->2854)!) localhost.65458 > localhost.sunrpc: UDP, length: 136 09:44:56.100221 IP (tos 0x0, ttl 64, id 21492, offset 0, flags [none], length: 148, bad cksum 0 (->2863)!) localhost.exp1 > localhost.1020: UDP, length: 120 09:44:56.100320 IP (tos 0x0, ttl 64, id 21493, offset 0, flags [none], length: 56, bad cksum 0 (->28be)!) localhost.1020 > localhost.exp1: [bad udp cksum f3de!] UDP, length: 28 09:45:02.159762 IP (tos 0x0, ttl 64, id 22181, offset 0, flags [none], length: 164, bad cksum 0 (->25a2)!) localhost.65458 > localhost.sunrpc: UDP, length: 136 09:45:02.159939 IP (tos 0x0, ttl 64, id 22182, offset 0, flags [none], length: 148, bad cksum 0 (->25b1)!) localhost.exp1 > localhost.1020: UDP, length: 120 09:45:02.160071 IP (tos 0x0, ttl 64, id 22183, offset 0, flags [none], length: 56, bad cksum 0 (->260c)!) localhost.1020 > localhost.exp1: [bad udp cksum b3de!] UDP, length: 28 And with daily output telling me this in dom0 (daily from domu is fine): network: netstat: kvm_read: Bad address Name Ipkts Ierrs Opkts Oerrs Colls I'm getting the impression it's network card related. A bit more info: uname -a&&pkg_info\|grep xen NetBSD gogeta.internal 4.99.21 NetBSD 4.99.21 (XEN3_DOM0) #4: Mon Jun 25 05:04:37 EST 2007 rootspike.internal:/usr/obj/sys/arch/i386/compile/XEN3_DO M0 i386 xenkernel3-3.1.0 Xen3 Kernel xentools3-3.1.0 Userland Tools for Xen ifconfig -a bge0: flags=8b43<UP,BROADCAST,RUNNING,PROMISC,ALLMULTI,SIMPLEX, MULTICAST> mtu 1500 capabilities=3f80<TSO4,IP4CSUM_Rx,IP4CSUM_Tx,TCP4CSUM_Rx, TCP4CSUM_Tx,UDP4CSUM_Rx,UDP4CSUM_Tx> enabled=0 address: 00:13:72:18:02:ad media: Ethernet autoselect (100baseTX full-duplex,flowcontrol,rxpause,txpause) status: active inet 192.168.210.10 netmask 0xffffff00 broadcast 192.168.210.255 inet6 fe80::213:72ff:fe18:2ad%bge0 prefixlen 64 scopeid 0x1 I could start poking but this probably looks obvious to someone else. Should I disable hardware csums? Thanks, Sarton

Re: ypbind hangs as of current from midday yesterday (kern+user)
France	2007-06-27 04:08:06
On Wed, Jun 27, 2007 at 09:54:22AM +1000, Sarton O'Brien wrote: > > It is the server I am trying to start ypbind on. When initiating ypbind via an > ssh session, I can ping the server and the ssh session I am in is fine but I > can't ssh in from another console. As soon as I ^C the ssh client in the > other console logs in. It's probably because it's waiting on ypbind. Once ypbind is killed nsswitch only uses local files again. > > Loopback is seeing: > > 09:44:50.049092 IP (tos 0x0, ttl 64, id 21366, offset 0, flags [none], > length: 164, bad cksum 0 (->28d1)!) localhost.65458 > localhost.sunrpc: UDP, > [...] it's normal on loopback I think, checksums are not computed on loopback. You can try: sysctl -w net.inet.ip.do_loopback_cksum=1 sysctl -w net.inet.tcp.do_loopback_cksum=1 sysctl -w net.inet.udp.do_loopback_cksum=1 > And with daily output telling me this in dom0 (daily from domu is fine): > > network: > netstat: kvm_read: Bad address > Name Ipkts Ierrs Opkts Oerrs Colls > > I'm getting the impression it's network card related. Maybe kernel/userland mismatch ? -- Manuel Bouyer, LIP6, Universite Paris VI. Manuel.Bouyerlip6.fr NetBSD: 26 ans d'experience feront toujours la difference --

Re: ypbind hangs as of current from midday yesterday (kern+user)
	2007-06-27 20:33:51
On Wed, 27 Jun 2007 07:08:06 pm Manuel Bouyer wrote: > As soon as I ^C the ssh > > client in the other console logs in. > > It's probably because it's waiting on ypbind. Once ypbind is killed > nsswitch only uses local files again. Does the ordering matter? I have files first then nis. I found this behaviour strange as it _is_ the server and the auth files reside there. > > 09:44:50.049092 IP (tos 0x0, ttl 64, id 21366, offset 0, flags [none], > > length: 164, bad cksum 0 (->28d1)!) localhost.65458 > localhost.sunrpc: > > UDP, [...] > it's normal on loopback I think, checksums are not computed on loopback. > You can try: > sysctl -w net.inet.ip.do_loopback_cksum=1 > sysctl -w net.inet.tcp.do_loopback_cksum=1 > sysctl -w net.inet.udp.do_loopback_cksum=1 Yeah no difference. I'm guessing this probably isn't related. > > network: > > netstat: kvm_read: Bad address > > Name Ipkts Ierrs Opkts Oerrs Colls > > > > I'm getting the impression it's network card related. > > Maybe kernel/userland mismatch ? I blew away the obj dir and recompiled with sources from last night and this is now working but ypbind is still hanging. Is ypbind suppose to hang indefinitely? Does anyone actually have a current system running nis? I was so stoked to see nfs working properly again, figures nis would break on me Thanks, Sarton

Re: ypbind hangs as of current from midday yesterday (kern+user)
	2007-06-27 22:49:22
On Thu, 28 Jun 2007 12:29:34 pm Christos Zoulas wrote: > Did you post a ktrace of it? Here goes ... first time for everything. If I've done something wrong or could do it better or properly, let me know Below is the output. This repeats indefinitely: 1447 1 ypbind RET read 500/0x1f4 1447 1 ypbind CALL read(9,0x8068000,0x4000) 1447 1 ypbind GIO fd 9 read 0 bytes "" 1447 1 ypbind RET read 0 1447 1 ypbind CALL close(9) 1447 1 ypbind RET close 0 1447 1 ypbind CALL sendto(6,0xbfbfd4b0,0x88,0,0xbfbfde88,0x10) 1447 1 ypbind GIO fd 6 wrote 136 bytes "^B^AM^F240^B ^E^ADFM^C^Y M-\0^Nspike.internalb 0^B^C ^D^E^T^_^B 0^AM^FM-$^B ^Bfbinternal" 1447 1 ypbind RET sendto 136/0x88 1447 1 ypbind CALL read(8,0x8064000,0x4000) 1447 1 ypbind GIO fd 8 read 0 bytes "" 1447 1 ypbind RET read 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL select(8,0xbfbfdfb0,0,0,0xbfbfdfe8) 1447 1 ypbind RET select 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL select(8,0xbfbfdfb0,0,0,0xbfbfdfe8) 1447 1 ypbind RET select 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL select(8,0xbfbfdfb0,0,0,0xbfbfdfe8) 1447 1 ypbind RET select 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL select(8,0xbfbfdfb0,0,0,0xbfbfdfe8) 1447 1 ypbind RET select 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL select(8,0xbfbfdfb0,0,0,0xbfbfdfe8) 1447 1 ypbind RET select 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL select(8,0xbfbfdfb0,0,0,0xbfbfdfe8) 1447 1 ypbind RET select 0 1447 1 ypbind CALL gettimeofday(0xbfbfcbb8,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL __sysctl(0xbfbfca48,2,0xbfbfcac7,0xbfbfca50,0,0) 1447 1 ypbind RET __sysctl 0 1447 1 ypbind CALL geteuid 1447 1 ypbind RET geteuid 0 1447 1 ypbind CALL getegid 1447 1 ypbind RET getegid 0 1447 1 ypbind CALL getgroups(0x10,0xbfbfca84) 1447 1 ypbind RET getgroups 8 1447 1 ypbind CALL gettimeofday(0xbfbfc8a0,0) 1447 1 ypbind RET gettimeofday 0 1447 1 ypbind CALL lseek(8,0,0,0,0) 1447 1 ypbind RET lseek 0 1447 1 ypbind CALL read(8,0x8064000,0x4000) 1447 1 ypbind GIO fd 8 read 15 bytes "spike.internal " 1447 1 ypbind RET read 15/0xf 1447 1 ypbind CALL __stat30(0xbbb9d6a0,0xbfbfca68) 1447 1 ypbind NAMI "/etc/nsswitch.conf" 1447 1 ypbind RET __stat30 0 1447 1 ypbind CALL open(0xbbb9b9e3,0,0x1b6) 1447 1 ypbind NAMI "/etc/hosts" 1447 1 ypbind RET open 9 1447 1 ypbind CALL __fstat30(9,0xbfbfc908) 1447 1 ypbind RET __fstat30 0 1447 1 ypbind CALL read(9,0x8068000,0x4000) 1447 1 ypbind GIO fd 9 read 500 bytes "::1 localhost localhost.internal 127.0.0.1 localhost localhost.internal # Entries to allow offline-DNS NFS mounts 192.168.8.10 gogeta gogeta.internal 192.168.8.9 symbiote symbiote.internal 192.168.8.8 spike spike.internal 192.168.8.7 sammy sammy.internal 192.168.8.20 cre cre.internal 192.168.8.6 babylon babylon.internal 192.168.8.102 portal-wired portal-wired.internal 192.168.8.123 portal portal.internal # NIS update hack (strips the last char) # 192.168.8.8 spike spike.interna " I hope this helps ... Sarton

Re: ypbind hangs as of current from midday yesterday (kern+user)
United Kingdom	2007-06-29 12:29:56
On Thu, Jun 28, 2007 at 01:49:22PM +1000, Sarton O'Brien wrote: > On Thu, 28 Jun 2007 12:29:34 pm Christos Zoulas wrote: > > Did you post a ktrace of it? > > Below is the output. This repeats indefinitely: > > 1447 1 ypbind RET read 500/0x1f4 > 1447 1 ypbind CALL read(9,0x8068000,0x4000) > 1447 1 ypbind GIO fd 9 read 0 bytes > "" > 1447 1 ypbind RET read 0 > 1447 1 ypbind CALL close(9) > 1447 1 ypbind RET close 0 > 1447 1 ypbind CALL sendto(6,0xbfbfd4b0,0x88,0,0xbfbfde88,0x10) > 1447 1 ypbind GIO fd 6 wrote 136 bytes > "^B^AM^F240^B ^E^ADFM^C^Y > > M-\0^Nspike.internalb 0^B^C > > ^D^E^T^_^B 0^AM^FM-$^B > ^Bfbinternal" > 1447 1 ypbind RET sendto 136/0x88 I did wonder if I'd managed to break sendto() - not much actually uses it. But traceroute(8) does, and seems to work on my system running a very recent kernel. But I'm not inside xen.... Unfortunately ktrace doesn't get the 'sockaddr' ... David -- David Laight: davidl8s.co.uk

Re: ypbind hangs as of current from midday yesterday (kern+user)
	2007-07-01 14:48:57
On Thu, Jun 28, 2007 at 11:33:51AM +1000, Sarton O'Brien wrote: > On Wed, 27 Jun 2007 07:08:06 pm Manuel Bouyer wrote: > > As soon as I ^C the ssh > > > client in the other console logs in. > > > > It's probably because it's waiting on ypbind. Once ypbind is killed > > nsswitch only uses local files again. > > Does the ordering matter? I have files first then nis. I found this behaviour > strange as it _is_ the server and the auth files reside there. When login, the sytems needs to look at all the groups to know in which groups the user is. As long as NIS is enabled for groups the system will look at it at login time, whichever order is specified in nsswitch.conf for group > Is ypbind suppose to hang indefinitely? I believe it'll wait for a server to reply, yes > Does anyone actually have a current > system running nis? I just set up a test nis domain on a domU. With default files in /etc it came up without troubles. Do you have NIS or DNS enabled for hosts in /etc/nsswitch.conf ? It could be the cause of the hang. Especially if NIS is enabled and the hosts involved in NIS startup aren't listed in /etc/hosts (with the proper names), there will be a deadlock. Are the hosts in the ypservers map and /var/yp/binding/<domain>.ypservers properly recorded in /etc/hosts ? -- Manuel Bouyer <bouyerantioche.eu.org> NetBSD: 26 ans d'experience feront toujours la difference --

[1-8]

about | contact Other archives ( Real Estate discussion Medical topics )