Re: dhclient wedged

2006-02-03 Thread Brooks Davis
On Thu, Feb 02, 2006 at 10:05:44PM -0800, John T. Yocum wrote:
 I've been having a same sort of issue that others have reported. After 
 dhclient has been running for a random length of time, it suddenly 
 starts consuming a large amount of CPU. Unfortunately for me, the 
 longest it tends to run is 48 hours before acting up.
 
 I did get it rebuilt with debug symbols, and got a stack trace from it, 
 while it was acting up. Hopefully, it will be useful.

I really need a core dump.  A single stack trace just tells me the
packet processing routing isn't working (or at least everything is
getting rejected).  Unfortunatly, I haven't had any luck finding bugs
there by inspection.

-- Brooks

-- 
Any statement of the form X is the one, true Y is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgp6aMMERprXM.pgp
Description: PGP signature


dhclient wedged

2006-02-02 Thread John T. Yocum
I've been having a same sort of issue that others have reported. After 
dhclient has been running for a random length of time, it suddenly 
starts consuming a large amount of CPU. Unfortunately for me, the 
longest it tends to run is 48 hours before acting up.


I did get it rebuilt with debug symbols, and got a stack trace from it, 
while it was acting up. Hopefully, it will be useful.


Attaching to program: /sbin/dhclient, process 2248
Reading symbols from /lib/libc.so.6...done.
Loaded symbols for /lib/libc.so.6
Reading symbols from /libexec/ld-elf.so.1...done.
Loaded symbols for /libexec/ld-elf.so.1
0x08051a0d in receive_packet (interface=0x8071000,
buf=0xbfbfdc20 \002\001\006, len=4096, from=0xbfbfec70, 
hfrom=0xbfbfec50)

at /usr/src/sbin/dhclient/bpf.c:308
308 memcpy(hdr, 
interface-rbuf[interface-rbuf_offset],

(gdb) bt
#0  0x08051a0d in receive_packet (interface=0x8071000,
buf=0xbfbfdc20 \002\001\006, len=4096, from=0xbfbfec70, 
hfrom=0xbfbfec50)

at /usr/src/sbin/dhclient/bpf.c:308
#1  0x08050dc1 in got_one (l=0x80741b0)
at /usr/src/sbin/dhclient/dispatch.c:256
#2  0x08050d58 in dispatch () at /usr/src/sbin/dhclient/dispatch.c:228
#3  0x0804a0d0 in main (argc=1, argv=0xbfbfed3c)
at /usr/src/sbin/dhclient/dhclient.c:429

BTW: For reference I'm running 6.0-RELEASE-p4.

Thanks,
John
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-02-01 Thread Peter Jeremy
On Tue, 2006-Jan-31 15:38:53 -0500, Lodewijk Vge wrote:
On 31-jan-2006, at 14:13, Brooks Davis wrote:

At the very least I need a coredump and your executable so I can  
look at variables
in receive_packet.

I accidentally killed it with my attempts to make it dump a core  
file. so, what should I do the next time it happens to make it dump  
core?

kill -QUIT ... is the generic answer.  sigaction(2) provides the
definitive list of which signals default to dumping core.

-- 
Peter Jeremy
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-01-31 Thread Lodewijk Vöge

On 28-jan-2006, at 20:33, Kelly D. Grills wrote:


Just a quick update. I've recompiled and am waiting for a failure.
If I recall correctly my problem occured when the ip actually
changed, so it may be a while before I post any debugging info.


it just wedged on my machine, also on comcast. this is the backtrace  
when I attach gdb:


(gdb) bt
#0  0x2813929a in memcpy () from /lib/libc.so.6
#1  0x0805199d in receive_packet (interface=0x807,
buf=0xbfbfdb60 \002\001\006, len=4096, from=0xbfbfebb0,
hfrom=0xbfbfeb90) at /usr/src/sbin/dhclient/bpf.c:306
#2  0x08050d51 in got_one (l=0x8073230)
at /usr/src/sbin/dhclient/dispatch.c:256
#3  0x08050ce8 in dispatch () at /usr/src/sbin/dhclient/dispatch.c:228
#4  0x0804a01c in main (argc=1, argv=0xbfbfec98)
at /usr/src/sbin/dhclient/dhclient.c:429

Lodewijk
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-01-31 Thread Brooks Davis
On Tue, Jan 31, 2006 at 12:29:48PM -0500, Lodewijk Vöge wrote:
 On 28-jan-2006, at 20:33, Kelly D. Grills wrote:
 
 Just a quick update. I've recompiled and am waiting for a failure.
 If I recall correctly my problem occured when the ip actually
 changed, so it may be a while before I post any debugging info.
 
 it just wedged on my machine, also on comcast. this is the backtrace  
 when I attach gdb:
 
 (gdb) bt
 #0  0x2813929a in memcpy () from /lib/libc.so.6
 #1  0x0805199d in receive_packet (interface=0x807,
 buf=0xbfbfdb60 \002\001\006, len=4096, from=0xbfbfebb0,
 hfrom=0xbfbfeb90) at /usr/src/sbin/dhclient/bpf.c:306
 #2  0x08050d51 in got_one (l=0x8073230)
 at /usr/src/sbin/dhclient/dispatch.c:256
 #3  0x08050ce8 in dispatch () at /usr/src/sbin/dhclient/dispatch.c:228
 #4  0x0804a01c in main (argc=1, argv=0xbfbfec98)
 at /usr/src/sbin/dhclient/dhclient.c:429

The backtrace doesn't tell me much at first glance.  I'm quite certain
it's not hung in memcpy since I'm highly confident memcpy doesn't
have an undiscovered bugs. :)  Instead it's presumably looping in
receive_packet or at a higher level, possibaly in dispatch.  At the very
least I need a coredump and your executable so I can look at variables
in receive_packet.

Thanks for your report.

-- Brooks

-- 
Any statement of the form X is the one, true Y is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgp3EEkAqmRuE.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-31 Thread Lodewijk Vöge

On 31-jan-2006, at 14:13, Brooks Davis wrote:

At the very least I need a coredump and your executable so I can  
look at variables

in receive_packet.


I accidentally killed it with my attempts to make it dump a core  
file. so, what should I do the next time it happens to make it dump  
core?


Lodewijk
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-01-28 Thread Kelly D. Grills
On Sun, Jan 22, 2006 at 06:40:19PM -0600, Kelly D. Grills wrote:
 On Sun, Jan 22, 2006 at 02:18:28PM -0800, Brooks Davis wrote:
  
  This definitly sounds like something particular to your dhcp servers.
  It would be nice if we could fix it, but without some debugging help
  that's going to be pretty much impossible.  If you can recompile
  dhclient with debugging symbols and when it goes into a spin, use the
  debugger to figure out where it's spinning, that would be useful.
  See this post for details:
  
  http://lists.freebsd.org/pipermail/freebsd-current/2005-August/054224.html
  
 I'm recompiling at the moment. Will follow up with debugging results
 when it fails.
 

Just a quick update. I've recompiled and am waiting for a failure.
If I recall correctly my problem occured when the ip actually
changed, so it may be a while before I post any debugging info.

-- 
Kelly D. Grills
[EMAIL PROTECTED]


pgpC24knsJv7G.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-27 Thread Will Froning
Kelly,

On Sun, 22 Jan 2006, Kelly D. Grills wrote:
=On Sun, Jan 22, 2006 at 02:52:44AM -0500, Forrest Aldrich wrote:
= 
= I reported this problem a few OS versions back... pre-6.  I'm on 
= FreeBSD-6-STABLE, and I've found the dhclient once again wedged in a 
= mode that was eating a lot of CPU.  
= 
= The solution is to kill it, and restart.
= 
= I'm on Comcast's network, so I don't really know if their DHCP server is 
= doing something that FreeBSD's stock dhclient doesn't like; however, I 
= wonder if someone else has noticed this problem, etc.
= 
=Yes, I've also had the same problem (6.0-RELEASE), on Comcast also.
=I installed isc-dhcp3-client-3.0.3_1 from ports and haven't had a 
=problem since. (crossing fingers ;=)

not sure if it's the same problem for you, but Comcast in my area 
supplies a bad entry for DNS Suffix/Search order

((hsd1.ca.comcast.net)(hsd2.ca.comcast.net))

And my little netgear router doesn't like that entry and _seems_ to 
only supply hsd1.ca.comcast.net)) via dhcp.  My Mac and Winbloze 
boxes do fine, but FBSD rejects the dhoffer unless isc's dhclient is 
used.  I haven't tried plugging directly into the Comcast router, so I 
don't know if FBSD will accept the original offer with parens...

Thanks,
Will

-- 
Will Froning
Unix Sys. Admin.
[EMAIL PROTECTED]
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-01-27 Thread Brooks Davis
On Fri, Jan 27, 2006 at 06:30:36AM -0800, Will Froning wrote:
 Kelly,
 
 On Sun, 22 Jan 2006, Kelly D. Grills wrote:
 =On Sun, Jan 22, 2006 at 02:52:44AM -0500, Forrest Aldrich wrote:
 = 
 = I reported this problem a few OS versions back... pre-6.  I'm on 
 = FreeBSD-6-STABLE, and I've found the dhclient once again wedged in a 
 = mode that was eating a lot of CPU.  
 = 
 = The solution is to kill it, and restart.
 = 
 = I'm on Comcast's network, so I don't really know if their DHCP server is 
 = doing something that FreeBSD's stock dhclient doesn't like; however, I 
 = wonder if someone else has noticed this problem, etc.
 = 
 =Yes, I've also had the same problem (6.0-RELEASE), on Comcast also.
 =I installed isc-dhcp3-client-3.0.3_1 from ports and haven't had a 
 =problem since. (crossing fingers ;=)
 
 not sure if it's the same problem for you, but Comcast in my area 
 supplies a bad entry for DNS Suffix/Search order
 
 ((hsd1.ca.comcast.net)(hsd2.ca.comcast.net))

I must say that's a really creative one.  I wonder if it even works for
Windows.

 And my little netgear router doesn't like that entry and _seems_ to 
 only supply hsd1.ca.comcast.net)) via dhcp.  My Mac and Winbloze 
 boxes do fine, but FBSD rejects the dhoffer unless isc's dhclient is 
 used.  I haven't tried plugging directly into the Comcast router, so I 
 don't know if FBSD will accept the original offer with parens...

If you upgrade to 6.0-STABLE the offer won't be rejected, it will just
delete the invalid entry.

-- Brooks

-- 
Any statement of the form X is the one, true Y is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgpRBidrMG5pi.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-23 Thread Krzysztof Kowalik
Brooks Davis [EMAIL PROTECTED] wrote:
 This definitly sounds like something particular to your dhcp servers.
 It would be nice if we could fix it, but without some debugging help
 that's going to be pretty much impossible.  [...]

It happens to me quite often, too. The only thing related in the
non-debug messages is:

Jan  8 12:02:19 moneypenny dhclient[68091]: 5 bad IP checksums seen in 5 packets
Jan  8 12:02:49 moneypenny last message repeated 743054 times
Jan  8 12:04:50 moneypenny last message repeated 2951866 times
Jan  8 12:14:51 moneypenny last message repeated 14457921 times
Jan  8 12:24:52 moneypenny last message repeated 14812032 times
Jan  8 12:34:53 moneypenny last message repeated 14770327 times
Jan  8 12:44:55 moneypenny last message repeated 14748300 times
Jan  8 12:51:44 moneypenny last message repeated 10037074 times

... which accounts for the CPU usage, I guess. I killed the bad IP
checksums messages, so it doesn't annoy my syslog anymore, but it of
course didn't fix the underlaying issue. 

I was looking at those packets with tcpdump once and didn't see anything
obvious/bad there.

And yes, I didn't have this problem with ISC client. And I surely use
different cable provider, than the original poster ;)

-- 
Krzysztof Kowalik   |  () ASCII Ribbon Campaign
Computer Center, AGH UST|  /\ Support plain text e-mail
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-01-23 Thread Brooks Davis
On Mon, Jan 23, 2006 at 10:06:56AM +0100, Krzysztof Kowalik wrote:
 Brooks Davis [EMAIL PROTECTED] wrote:
  This definitly sounds like something particular to your dhcp servers.
  It would be nice if we could fix it, but without some debugging help
  that's going to be pretty much impossible.  [...]
 
 It happens to me quite often, too. The only thing related in the
 non-debug messages is:
 
 Jan  8 12:02:19 moneypenny dhclient[68091]: 5 bad IP checksums seen in 5 
 packets
 Jan  8 12:02:49 moneypenny last message repeated 743054 times
 Jan  8 12:04:50 moneypenny last message repeated 2951866 times
 Jan  8 12:14:51 moneypenny last message repeated 14457921 times
 Jan  8 12:24:52 moneypenny last message repeated 14812032 times
 Jan  8 12:34:53 moneypenny last message repeated 14770327 times
 Jan  8 12:44:55 moneypenny last message repeated 14748300 times
 Jan  8 12:51:44 moneypenny last message repeated 10037074 times
 
 ... which accounts for the CPU usage, I guess. I killed the bad IP
 checksums messages, so it doesn't annoy my syslog anymore, but it of
 course didn't fix the underlaying issue. 
 
 I was looking at those packets with tcpdump once and didn't see anything
 obvious/bad there.
 
 And yes, I didn't have this problem with ISC client. And I surely use
 different cable provider, than the original poster ;)

What NIC are you using?  This particular issue sounds like a NIC
returning corrupt packets for some reason.  Alternatively, the sending
server could be producing corrupt packets.  Some tcpdump traces
(preferably raw dumps) could be useful.

-- Brooks

-- 
Any statement of the form X is the one, true Y is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgprpjH6da2ZF.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-23 Thread Krzysztof Kowalik
Brooks Davis [EMAIL PROTECTED] wrote:
 What NIC are you using?

fxp0: Intel 82559 Pro/100 Ethernet port 0xc000-0xc03f mem
0xdb121000-0xdb121fff,0xdb00-0xdb0f irq 11 at device 1.0 on pci2

 This particular issue sounds like a NIC returning corrupt packets for
 some reason. Alternatively, the sending server could be producing corrupt
 packets.  Some tcpdump traces (preferably raw dumps) could be useful.

As I said, I didn't see anything bad in the tcpdump. And since I really
didn't have such issues before, and I don't see anything weird in any
other place... I recompiled dhclient with -g and will tcpdump the
traffic again, if I notice it eating 99% of CPU time again.

-- 
Krzysztof Kowalik   |  () ASCII Ribbon Campaign
Computer Center, AGH UST|  /\ Support plain text e-mail
___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]


Re: dhclient wedged

2006-01-23 Thread Brooks Davis
On Mon, Jan 23, 2006 at 11:25:25PM +0100, Krzysztof Kowalik wrote:
 Brooks Davis [EMAIL PROTECTED] wrote:
  What NIC are you using?
 
 fxp0: Intel 82559 Pro/100 Ethernet port 0xc000-0xc03f mem
 0xdb121000-0xdb121fff,0xdb00-0xdb0f irq 11 at device 1.0 on pci2

What's ifconfig say?

  This particular issue sounds like a NIC returning corrupt packets for
  some reason. Alternatively, the sending server could be producing corrupt
  packets.  Some tcpdump traces (preferably raw dumps) could be useful.
 
 As I said, I didn't see anything bad in the tcpdump. And since I really
 didn't have such issues before, and I don't see anything weird in any
 other place... I recompiled dhclient with -g and will tcpdump the
 traffic again, if I notice it eating 99% of CPU time again.

OK, thanks.  My other guess would be that we're getting misaligned
relative to the buffer somehow and that's causing all the checksums to
fail.  If you can get into gdb when dhclient is spinning and check the
values of interface-rbuf_offset, interface-rbuf_len, and length in
receive_packet() that might tell us something intresting.

-- Brooks

-- 
Any statement of the form X is the one, true Y is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgpEH1LOU5INk.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-22 Thread Kelly D. Grills
On Sun, Jan 22, 2006 at 02:52:44AM -0500, Forrest Aldrich wrote:
 
 I reported this problem a few OS versions back... pre-6.  I'm on 
 FreeBSD-6-STABLE, and I've found the dhclient once again wedged in a 
 mode that was eating a lot of CPU.  
 
 The solution is to kill it, and restart.
 
 I'm on Comcast's network, so I don't really know if their DHCP server is 
 doing something that FreeBSD's stock dhclient doesn't like; however, I 
 wonder if someone else has noticed this problem, etc.
 
Yes, I've also had the same problem (6.0-RELEASE), on Comcast also.
I installed isc-dhcp3-client-3.0.3_1 from ports and haven't had a 
problem since. (crossing fingers ;=)

-- 
Kelly D. Grills
[EMAIL PROTECTED]



pgpDlQTBKxIUa.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-22 Thread Brooks Davis
On Sun, Jan 22, 2006 at 09:51:43AM -0600, Kelly D. Grills wrote:
 On Sun, Jan 22, 2006 at 02:52:44AM -0500, Forrest Aldrich wrote:
  
  I reported this problem a few OS versions back... pre-6.  I'm on 
  FreeBSD-6-STABLE, and I've found the dhclient once again wedged in a 
  mode that was eating a lot of CPU.  
  
  The solution is to kill it, and restart.
  
  I'm on Comcast's network, so I don't really know if their DHCP server is 
  doing something that FreeBSD's stock dhclient doesn't like; however, I 
  wonder if someone else has noticed this problem, etc.
  
 Yes, I've also had the same problem (6.0-RELEASE), on Comcast also.
 I installed isc-dhcp3-client-3.0.3_1 from ports and haven't had a 
 problem since. (crossing fingers ;=)

This definitly sounds like something particular to your dhcp servers.
It would be nice if we could fix it, but without some debugging help
that's going to be pretty much impossible.  If you can recompile
dhclient with debugging symbols and when it goes into a spin, use the
debugger to figure out where it's spinning, that would be useful.
See this post for details:

http://lists.freebsd.org/pipermail/freebsd-current/2005-August/054224.html

-- Brooks

-- 
Any statement of the form X is the one, true Y is FALSE.
PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4


pgp5Iv4wgDxY6.pgp
Description: PGP signature


Re: dhclient wedged

2006-01-22 Thread Kelly D. Grills
On Sun, Jan 22, 2006 at 02:18:28PM -0800, Brooks Davis wrote:
 
 This definitly sounds like something particular to your dhcp servers.
 It would be nice if we could fix it, but without some debugging help
 that's going to be pretty much impossible.  If you can recompile
 dhclient with debugging symbols and when it goes into a spin, use the
 debugger to figure out where it's spinning, that would be useful.
 See this post for details:
 
 http://lists.freebsd.org/pipermail/freebsd-current/2005-August/054224.html
 
I'm recompiling at the moment. Will follow up with debugging results
when it fails.

-- 
Kelly D. Grills
[EMAIL PROTECTED]


pgpT2QVOSEkvS.pgp
Description: PGP signature


dhclient wedged

2006-01-21 Thread Forrest Aldrich
I reported this problem a few OS versions back... pre-6.  I'm on 
FreeBSD-6-STABLE, and I've found the dhclient once again wedged in a 
mode that was eating a lot of CPU.  


The solution is to kill it, and restart.

I'm on Comcast's network, so I don't really know if their DHCP server is 
doing something that FreeBSD's stock dhclient doesn't like; however, I 
wonder if someone else has noticed this problem, etc.



_F

___
freebsd-stable@freebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to [EMAIL PROTECTED]