OpenSolaris

Printable Version Enter a New Search
Bug ID 4986151
Synopsis "ifconfig qfe0 inet6 up" hung during ipmp test
State 10-Fix Delivered (Fix available in build)
Category:Subcategory kernel:tcp-ip
Keywords fireengine
Responsible Engineer Ashish Mehta
Reported Against 5.10
Duplicate Of
Introduced In solaris_10
Commit to Fix s10_57
Fixed In s10_57
Release Fixed solaris_10(s10_57)
Related Bugs 4898645
Submit Date 30-January-2004
Last Update Date 27-April-2007
Description
The "ifconfig qfe0 inet6 up" hung without apparent reason for several hours
during ipmp testsuite execution. I eventually introduced the core dump so that
the cause can be looked for.

The ifconfig hang happened after more than 30 executions of ipmp on s10_52. The
test was ran on two separate machines (Ultra2 and SunBlade-1000), of which the
Ultra2 showed the failure. We have never experienced an "ifconfig up" failure
during ipmp on previous builds of s10 but given the low probability of occurence
on s10_52 this may be due to a chance.

For the core dump and additional details please see below!


Symptoms:

03:21:36: ti2standby6: PASS
03:21:36: ti2standby6: Clean up
03:21:36: ti2standby6: Bring down and unplumb interface qfe0 (v6)
03:21:36: ti2standby6: Bring down and unplumb interface qfe1 (v6)
03:21:37: ti2standby6: Kill in.mpathd.
03:21:37: ti2standby6: ends (returning 0)
03:21:38: ti3bothfover6: begins
03:21:38: ti3bothfover6: Kill in.mpathd.
03:21:38: ti3bothfover6: Set up 3 interfaces: 1 2 3
03:21:38: ti3bothfover6: Plumb and configure qfe0 inet6
<hang here for more than 5 hours after which the machine was brought down>



In the machine:

# ps -ef | grep ifconfig
    root 116954 116924   0 03:21:39 console     0:00 ifconfig qfe0 inet6 up
    root 117036 117032   0 08:59:47 pts/1       0:00 grep ifconfig
# ptree 116954
100408 -sh
  100444 /bin/ksh ./wrapper
    100656 /bin/ksh /earth/domain_scripts/bin/runit ipmp 2 64+64+-/64+64+-/32+32
      103118 /bin/sh /net/diablo.ireland/gates/Production/S10/links/ipmp/1.10.X/
        103519 make results
          104049 sh -ce cd stc_files; make --e execute
            104050 make --e execute
              104054 /bin/ksh -p ./stc_runtest
                116924 /bin/ksh -p ./ti3bothfover6
                  116954 ifconfig qfe0 inet6 up
# truss -p 116954
ioctl(5, SIOCSLIFFLAGS, 0xFFBFE4C8) (sleeping...)
^C# truss -p 116924
read(65, 0xFFBFBE18, 1024)      (sleeping...)
^C# ifconfig -a
lo0: flags=1000849<UP,LOOPBACK,RUNNING,MULTICAST,IPv4> mtu 8232 index 1
        inet 127.0.0.1 netmask ff000000 
hme0: flags=1000843<UP,BROADCAST,RUNNING,MULTICAST,IPv4> mtu 1500 index 2
        inet 129.156.229.19 netmask ffffff00 broadcast 129.156.229.255
        ether 8:0:20:9b:e9:20 
^C# 
(had to ctrl-c to get out)


There were no console messages for ifconfig (The ifconfig process started at 03:21:39):

Jan 30 03:21:11 delenn in.ndpd[103464]: Interface qfe0 has been removed from ker
nel. in.ndpd will no longer use it??
Jan 30 03:21:11 delenn in.ndpd[103464]: Interface qfe1 has been removed from ker
nel. in.ndpd will no longer use it??
Jan 30 03:21:36 delenn in.ndpd[103464]: Interface qfe0 has been removed from ker
nel. in.ndpd will no longer use it??
Jan 30 03:21:36 delenn in.ndpd[103464]: phyint_init_from_k: ioctl (get flags) (i
nterface qfe1): No such device or address??
Jan 30 03:21:36 delenn in.ndpd[103464]: Interface qfe1 has been removed from ker
nel. in.ndpd will no longer use it??
<Console messages stopped here>

The routing setup is as follows and looks normal:
# netstat -nr

Routing Table: IPv4
  Destination           Gateway           Flags  Ref   Use   Interface
-------------------- -------------------- ----- ----- ------ ---------
129.156.229.0        129.156.229.19       U         1    109  hme0
224.0.0.0            129.156.229.19       U         1      0  hme0
default              129.156.229.1        UG        1      0  hme0
127.0.0.1            127.0.0.1            UH        5    441  lo0


I took the machine down to ok prompt and dumped the kernel. It is availble in
   ****
   /net/diablo.ireland/export/crash/s10_52/ipmp
   ****

mdb shows that ifconfig was indeed still running at the time of dump so finding
reason for the hang should be possible:

# mdb unix.0 vmcore.0
Loading modules: [ unix krtld genunix sd ip nca nfs random logindmux ptm cpc ]
> ::ptree
000000000185d1d0  sched
     00000300010bc008  fsflush
     00000300010bca88  pageout
     00000300010bd508  init
          0000030003d2e070  in.ndpd
          00000300023a8ac8  sshd
          0000030003d48068  snmpd
          0000030003d48ae8  sh
               0000030003a14ad8  wrapper
                    0000030003d49568  runit
                         00000300023a8048  doipmp.client
                              00000300042c8080  make
                                   00000300041ceaf8  sh
                                        00000300042c8b00  make
                                             000003000b970a88  stc_runtest
                                                  00000300041ce078  tee
                                                  00000300042c1500  egrep
                                                  00000300042c0000  tee
                                                  0000030003d2eaf0  
                                                  ti3bothfover6
                                                       000003000b971508  
                                                       ifconfig
Work Around
N/A
Comments
N/A