Newsgroups: comp.parallel.pvm
From: singlis@cs.waikato.ac.nz (Stuart Inglis)
Reply-To: singlis@waikato.ac.nz
Subject: [BUG] pvmd3 stealing packets on LINUX
Organization: The University of Waikato
Date: 29 Apr 1997 22:13:31 GMT
Message-ID: <5k5rqb$s6o@netserv.waikato.ac.nz>


Hello all, (please preceed all comments with
"I think/it looks like" etc.)

pvmd3 is stealing YP packets and therefore causes
  "yp_match: clnt_call: RPC: Timed out"
errors when it steals YP messages. This cause the
machine to pause for 5-10 seconds everytime it happens.

##
## Setup
##

Local machine [tawa, linux/gnu 2.0.30, pvm 3.3.11]:
YP server [xena -- I replaced the IP numbers below with xena's name]:

##
## command trace
##

tawa% pvm
pvm> quit
pvmd still running.

tawa% ps
singlis  19035  0.2  0.4   928   548  p2 S    09:56   \
	0:00 /home/singlis/pvm3/lib/LINUX/pvmd3

tawa% strace -p 19035
select(10, [7 9], [], NULL, {30, 280000}) = 0 (Timeout)
gettimeofday({862351088, 763836}, NULL) = 0
select(10, [7 9], [], NULL, {60, 0})    = 0 (Timeout)
gettimeofday({862351148, 773878}, NULL) = 0
select(10, [7 9], [], NULL, {60, 0}

# in a second shell
#
tawa-2% ls -l
#
# ls -l cause a YP lookup to get usernames etc.

#
# output continues from strace ...
tawa%
select(10, [7 9], [], NULL, {60, 0} = 1 (in [7], left {3, 380000})
recvfrom(7, "4c\333\177\0\0\0\1\0\0\0\0\0\0\0"..., 4096, 0, {sin_family=AF_INET, sin_p\
ort=htons(607), sin_addr=inet_addr("xena")}, [16]) = 120
gettimeofday({862351205, 397518}, NULL) = 0
select(10, [7 9], [], NULL, {3, 376360}) = 0 (Timeout)
gettimeofday({862351208, 783862}, NULL) = 0
select(10, [7 9], [], NULL, {60, 0})    = 1 (in [7], left {58, 380000})
recvfrom(7, "4c\333\177\0\0\0\1\0\0\0\0\0\0\0"..., 4096, 0, {sin_family=AF_INET, sin_p\
ort=htons(607), sin_addr=inet_addr("xena")}, [16]) = 120
gettimeofday({862351210, 406717}, NULL) = 0
select(10, [7 9], [], NULL, {58, 377145}

## pvmd3 seems to have received something from xena? Why would it
## want to talk to the YP server? Surely if I'm not using pvm
## then pvmd3 should be quiet.

# the output of the ls -l
#
tawa-2% ls -l
yp_match: clnt_call: RPC: Timed out
-rw-r--r--   1 singlis  compsci     17705 Apr 30 09:15 Makefile
-rw-r--r--   1 singlis  compsci     32626 Apr 16 18:31 Makefile~
drwxr-xr-x   2 singlis  compsci      1024 Apr 15 11:26 arff/

## Approximately one in 10 'ls -l' commands have the RPC error
## and pvmd3 complains of bad packets.

## Can anyone help me? 

--
Stuart Inglis,
Department of Computer Science
University of Waikato, Hamilton, New Zealand