Sendmail rset command hangs socket on 1.6-Release
Matthew Dillon
dillon at apollo.backplane.com
Wed Aug 2 10:01:09 PDT 2006
:Based on the ktrace it would appear that sendmail sees it, but fails to
:write to the sendmail.st file:
:
: 50367 sendmail RET sendto 163/0xa3
:...
: 50367 sendmail RET fstat 0
: 50367 sendmail CALL flock(0x7,0x2)
This is great news! It's stuck on the flock() call, which is the
same problem that was reported with postfix.
Now do this. With sendmail stuck on the flock call, compile up
/usr/src/test/debug/vnodeinfo.c and run vnodeinfo -a out to a file.
Find the file the flock is stuck on and lets see if there is anyone
else holding the lock.
NOTE! You will have to use the vnodeinfo.c from HEAD. The one in
REL does not have the lock tracing code in it. And you may have
to make some modifications if there are #include file snafu's.
I feel confident that we can track down this problem and fix it. What
I need to know is whether that flock is being held on the vnode's blocked
list without any locks held on the locked list, which would indicate a
tsleep/wakeup race, or whether there is someone else holding the lock.
-Matt
More information about the Bugs
mailing list