recvmsg/pflocal

[Marcus Brinkmann]

Hi,

syslogd doesn't work correctly.  I verified that it works under Linux.
Again, it seems to have something to do with interrupted messages or so.
Maybe it is pflocal/libpipes fault.

I do 'while /bin/true; do logger hi; logger bye; echo -n . ; done'
and 'while /bin/true; do sudo /bin/echo -n . ; done'

(sudo uses syslog()).  Those commands hang quickly (after a couple of
dozens of iterations), and as it seems after a fixed number of iterations.

I attach backtraces of pflocal and syslogd.  The output in syslogd
looks especially bogus:

#0  0x010a48cc in __mach_msg_trap () from /lib/debug/libc.so.0.2
#1  0x010a4f79 in __mach_msg (msg=0x1262f74, option=3, send_size=32,
    rcv_size=4096, rcv_name=38, timeout=0, notify=0) at msg.c:47
#2  0x010a553d in __mach_msg_server_timeout (demux=0x10b0e40 <msgport_server>,
    max_size=4096, rcv_name=38, option=0, timeout=0) at msgserver.c:169
#3  0x010a55a4 in __mach_msg_server (demux=0x10b0e40 <msgport_server>,
    max_size=4096, rcv_name=38) at msgserver.c:209
#4  0x010b0f4b in _hurd_msgport_receive () at msgportdemux.c:68
#5  0x464c457f in ?? ()
#6  0x010a553d in __mach_msg_server_timeout (demux=0x10b0e40 <msgport_server>,
    max_size=4096, rcv_name=38, option=0, timeout=0) at msgserver.c:169
#7  0x010a55a4 in __mach_msg_server (demux=0x10b0e40 <msgport_server>,
    max_size=4096, rcv_name=38) at msgserver.c:209
#8  0x010b0f4b in _hurd_msgport_receive () at msgportdemux.c:68
#9  0x464c457f in ?? ()
...
frames #6 to #9 repeated hundreds of times.
Note that this is not the recvmsg thread, which shows a different
backtrace.  Please see the attached files for more info.
pflocal.gdb is with logger, pflocal.gdb2 with sudo.

Thanks,
Marcus

pflocal.dbg.gz
Description: Text document

pflocal.dbg2.gz
Description: Binary data