Subject: 1.5ZC GENERIC, ipmon not logging correctly
To: None <port-sparc64@netbsd.org>
From: Tomi Nylund <wizard@in.finland.invalid>
List: port-sparc64
Date: 05/15/2002 16:17:15
Hello,

tried the new snapshot, and after getting it installed, seems much
more stable that 1.5ZA. However, ipmon logging does not seem to
work correctly. If you start ipfilter and ipmon normally from
/etc/rc.conf,
it does not log dropped packets, although it logs state changes, but
they seem to be corrupt also, like this:

01/01/1970 02:00:00.210052 STATE:CLOSE 192.168.20.10,65531 ->
132.227.74.11,22 PR tcp Pkts 22 Bytes 2752
01/01/1970 02:00:00.210055 STATE:CLOSE 192.168.20.10,65529 ->
132.227.74.11,22 PR 
tcp Pkts 24 Bytes 3100

Logging blocked packets does not work through syslog nor normal files.
When running ipmon with ktrace, it shows this:

First ipmon initializes..

<snip>
 280 ipmon    CALL  open(0x1047a8,0x601,0x1a4)
   280 ipmon    NAMI  "/var/run/ipmon.pid"
   280 ipmon    RET   open 5
   280 ipmon    CALL  fcntl(0x5,0x3,0)
   280 ipmon    RET   fcntl 1
   280 ipmon    CALL  getpid
   280 ipmon    RET   getpid 280/0x118
   280 ipmon    CALL  __fstat13(0x5,0xffffffffffffaff0)
   280 ipmon    RET   __fstat13 0
   280 ipmon    CALL  write(0x5,0x20e000,0x4)
   280 ipmon    GIO   fd 5 wrote 4 bytes
       "280
       "
   280 ipmon    RET   write 4
   280 ipmon    CALL  close(0x5)
   280 ipmon    RET   close 0
   280 ipmon    CALL  close(0x5)
-->   280 ipmon    RET   close -1 errno 9 Bad file descriptor
   280 ipmon    CALL 
__sigaction14(0x1,0xffffffffffffb720,0xffffffffffffb700)
   280 ipmon    RET   __sigaction14 0
   280 ipmon    CALL  ioctl(0x4,FIONREAD,0xffffffffffffb80c)
   280 ipmon    RET   ioctl 0
   280 ipmon    CALL  nanosleep(0xffffffffffffb730,0xffffffffffffb720)

..but then something goes wrong, and it drops to nanosleep.

/dev/ipl seems to work, as cat /dev/ipl gives you a constant stream of
data when you ping the box, for example. The kernel for installation
was taken from binary/kernel-subdirectory in 1.5ZC, as the one in
binary/sets was mysteriously curupted.

Any help appreciated,


Tomi

PS: Please reply to the mailing list, as my e-mail address is
obfuscated to repel spammers.