- take maintainer
- new distsite
- changes since 2008
o Allow longer server info timeout for third party via --idle-data-timeout
o Added feature to specify source port with "-p#:#" and "-P#:#"
o Add "-g" option to specify multicast IP address to use
o Clean up really confused transmit code for IPv4/IPv6 SSM multicast
o Bug fix from Aristeu Rozanski:
Crash caused by closing TCP_ADV_WIN_SCALE file even if open failed
o Fix bug with server CPU affinity being parsed as %X instead of %d
For non-sinkmode insure complete network block is written to stdout
Above fixes nuttscp bug seen with --copy-dir getting premature EOF
o Not every system has ERESTART added in 7.1.4, wrapped in ifdef
o Fix non-Linux systems to properly count TCP retrans for multiple streams
o Detect EOD for non-sinkmode UDP transfers
o Suppress bogus warning when using maximum size UDP packet
o Terminate non-sinkmode after specified file size with "-n" option
o Allow multilink aggregation with "-N##m" option to work for receive
o Add "-sz" zero copy option for non-sinkmode when input is a regular file
o Remove "-Sf" forced server mode from Usage: statement
o Fix zeroing of clientaddr6 during server cleanup
o Fix freeaddrinfo() processing during cleanup
o Change manually started oneshot server to have parent process just exit
o Provide summary TCP retrans info for multi-stream TCP
o Fix bug with retrans interval info when -fparse
o Add "+stride" or "+n.n.n.n" syntax for multi-stream TCP (IPv4)
o Fix third party bug with "-xc" option adding extraneous 't' character
o Add optional client-side name resolution for third party host
o Add "-N##m" option for multilink aggregation for multiple streams
o Add "-xc#/#" and "-P#/#" options to Usage: statement
o Enable jitter measurements with "-j" option
o Enable one-way delay measurements with "-o" option
o Fix bug with RTT and -fparse
o Change ctl/data port checks to < 1024 instead of < 5000
o Fix "--idle-data-timeout" Usage: statement for new default minimum
o Improve transmit performance with "-i" by setting poll() timeout to 0
o Don't output interval retrans info if non-sinkmode (for nuttscp)
o Skip check for unACKed data at end of transfer if -DBROKEN_UNACKED
o Play nice with iperf (change default data port to 5101)
o Delay sending of server "OK" until after successful server bind()
o Client check for server errors before starting data transfer
o Continue checking for server output while draining client transmission
o Correct "server not ACKing data" error message (server -> receiver)
o Add "--packet-burst" option for Rob
o Fix "--idle-data-timeout" Usage: statement for client
o Improve accuracy of retrans info timing synchronization (client xmitter)
o Allow rate limit to be exceeded temporarily by n packets ("-Rixxx/n")
o Allow setting server CPU affinity from client via "-xcs" option
o Allow setting client & server CPU affinity via third party
o Allow passing of third party control port via "-Pctlport/ctlport3"
o Up default idle data minimum to 15 sec to better handle net transients
o Don't reset nstream until after last use (fix getaddrinfo() memory leak)
o Fix bug with simultaneous server connections to manually started server
o Fix Usage: statement: "--idle-data-timeout" both server & client option
o Don't reset priority on server cleanup
o Fix priority output for "-fparse"
o Fix bad third party bug causing >= 1 minute transfers to silently fail
o Fix Usage: statement: "--idle-data-timeout" not just a server option
o Added IPv6 and SSM MC support
Ported from Rob's 5.5.5 based code by Bill Fink
o Fix client lockup with third party when network problem (for scripts)
o Bugfix: set chk_idle_data on client (now also checks no data received)
o Use argv[0] instead of "nuttcp" for third party
o Bugfix: give error message again on error starting server
o Timeout client accept() too and give nice error message (for scripts)
PR: 197475
Submitted by: pi
Approved by: Andrew Clark <andrewclarkii@gmail.com> (old maintainer)
by network and system managers. Its most basic usage is to determine
the raw TCP (or UDP) network layer throughput by transferring memory
buffers from a source system across an interconnecting network to
a destination system, either transferring data for a specified time
interval, or alternatively transferring a specified number of bytes.
nuttcp is based on nttcp, but have several useful features such
as a server mode, rate limiting, multiple parallel streams, and
timer based usage.
WWW: http://www.lcp.nrl.navy.mil/nuttcp/
PR: ports/137876
Submitted by: Andy Clark <andrewclarkii at gmail.com>
Kurt Jaeger
Martin Wilke