dns-oarc/dnsperf
1
0
Fork 0
Code Issues Activity

Adding upstream version 2.5.0+debian.

Browse source 
Signed-off-by: Daniel Baumann <daniel@debian.org>
This commit is contained in:
Daniel Baumann 2025-02-09 08:54:49 +01:00
parent 153471ed4b
commit 5d84f21420
Signed by: daniel
GPG key ID: FBB4F0E80A80222F
24 changed files with 2037 additions and 1001 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
.lgtm.yml
View file

@ -9,6 +9,7 @@ extraction:
- pkg-config - pkg-config
- libssl-dev - libssl-dev
- libldns-dev - libldns-dev
- libck-dev
configure: configure:
command: command:
- ./autogen.sh - ./autogen.sh

56
CHANGES
View file

@ -1,3 +1,59 @@
2021-03-12 Jerry Lundström
Release 2.5.0
This release adds re-connection support for TCP and DoT protocol,
new options to `resperf` and fixes a few bugs.
`dnsperf` and `resperf` will now try to re-connect when they lose a TCP
or DoT connection, and with that comes a few new statistics metrics.
For `dnsperf`, if a connection oriented protocol is used, it will now
show the total number of re-connections made and the connection latency.
For `resperf` it also shows the total number of re-connections made and
the gnuplot data now contains the total number of connections made and
the connection latency for each interval.
Beside re-connection support, improvements have been made when it comes
to tracking socket readiness while connections are established which
should generate less warnings about "socket not ready".
New `resperf` options:
- `-R`: Reopen the datafile if it runs out of data before the testing
is completed. This allows for long running tests on very small and
simple query datafile.
- `-F <fall_behind>`: Sets the maximum number of queries that can fall
behind being sent. `resperf` will stop when this many queries should
have been sent and it can be relative easy to hit if `-m <max_qps>`
is set too high.
The default is 1000 and setting it to zero (0) disables the check.
Bugfixes:
- Fixed port handling for host/network format when setting client side
port with `-x`
- Fix support for quoted characters, `\000` and `\.`, in domain names,
this was lost when removing BIND's internal development libraries
- Fix issue in `dnsperf`, it would loop forever if no connection could
be established
- Fix potential buffer overrun in `resperf` when using response id
for `queries[]`
- DoT: Fix bug when sending from buffer
Other changes:
- Always use `IPV6_V6ONLY` socket option for IPv6
- Add man-page on `-W` option added in v2.4.0
- Reformat man-pages
- `resperf`:
- Try and process more request each run to hopefully not hit max
outstanding so easy when high QPS
- Add default value to `-C` so it shows in help
9308361 man-page format
0e52fb4 man-page, opts, tuneups
e36211d stats
d9b9ba3 Response qid to index
2b2c37e fixes, reconnection
acd31e5 dname quote
de8f049 net
2021-02-23 Jerry Lundström 2021-02-23 Jerry Lundström
Release 2.4.2 Release 2.4.2

9
README.md
View file

@ -1,6 +1,6 @@
# dnsperf # dnsperf
[![Build Status](https://travis-ci.com/DNS-OARC/dnsperf.svg?branch=develop)](https://travis-ci.com/DNS-OARC/dnsperf) [![Total alerts](https://img.shields.io/lgtm/alerts/g/DNS-OARC/dnsperf.svg?logo=lgtm&logoWidth=18)](https://lgtm.com/projects/g/DNS-OARC/dnsperf/alerts/) [![Bugs](https://sonarcloud.io/api/project_badges/measure?project=dns-oarc%3Adnsperf&metric=bugs)](https://sonarcloud.io/dashboard?id=dns-oarc%3Adnsperf) [![Security Rating](https://sonarcloud.io/api/project_badges/measure?project=dns-oarc%3Adnsperf&metric=security_rating)](https://sonarcloud.io/dashboard?id=dns-oarc%3Adnsperf) [![Total alerts](https://img.shields.io/lgtm/alerts/g/DNS-OARC/dnsperf.svg?logo=lgtm&logoWidth=18)](https://lgtm.com/projects/g/DNS-OARC/dnsperf/alerts/) [![Bugs](https://sonarcloud.io/api/project_badges/measure?project=dns-oarc%3Adnsperf&metric=bugs)](https://sonarcloud.io/dashboard?id=dns-oarc%3Adnsperf) [![Security Rating](https://sonarcloud.io/api/project_badges/measure?project=dns-oarc%3Adnsperf&metric=security_rating)](https://sonarcloud.io/dashboard?id=dns-oarc%3Adnsperf)
`dnsperf` and `resperf` are free tools developed by Nominum/Akamai (2006-2018) `dnsperf` and `resperf` are free tools developed by Nominum/Akamai (2006-2018)
and DNS-OARC (since 2019) that make it simple to gather accurate latency and and DNS-OARC (since 2019) that make it simple to gather accurate latency and
@ -44,21 +44,22 @@ those queries to DNS servers to measure performance.
environment with autoconf, automake, libtool and pkgconfig. environment with autoconf, automake, libtool and pkgconfig.
- [OpenSSL](https://www.openssl.org/) - for TSIG support - [OpenSSL](https://www.openssl.org/) - for TSIG support
- [Concurrency Kit](http://concurrencykit.org/) - for atomic operations
- [LDNS](https://nlnetlabs.nl/projects/ldns/about/) - optional for dynamic update support - [LDNS](https://nlnetlabs.nl/projects/ldns/about/) - optional for dynamic update support
To install the dependencies under Debian/Ubuntu: To install the dependencies under Debian/Ubuntu:
``` ```
apt-get install -y libssl-dev libldns-dev apt-get install -y libssl-dev libldns-dev libck-dev
``` ```
To install the dependencies under CentOS (with EPEL enabled): To install the dependencies under CentOS (with EPEL enabled):
``` ```
yum install -y openssl-devel ldns-devel yum install -y openssl-devel ldns-devel ck-devel
``` ```
To install the dependencies under FreeBSD 12+ using `pkg`: To install the dependencies under FreeBSD 12+ using `pkg`:
``` ```
pkg install -y openssl ldns pkg install -y openssl ldns concurrencykit
``` ```
To install the dependencies under OpenBSD 6+ using `pkg_add`: To install the dependencies under OpenBSD 6+ using `pkg_add`:

9
configure.ac
View file

@ -16,7 +16,7 @@
# limitations under the License. # limitations under the License.
AC_PREREQ(2.64) AC_PREREQ(2.64)
AC_INIT([dnsperf], [2.4.2], [admin@dns-oarc.net], [dnsperf], [https://github.com/DNS-OARC/dnsperf/issues]) AC_INIT([dnsperf], [2.5.0], [admin@dns-oarc.net], [dnsperf], [https://github.com/DNS-OARC/dnsperf/issues])
AM_INIT_AUTOMAKE([-Wall -Werror foreign subdir-objects]) AM_INIT_AUTOMAKE([-Wall -Werror foreign subdir-objects])
AC_CONFIG_SRCDIR([src/dnsperf.c]) AC_CONFIG_SRCDIR([src/dnsperf.c])
AC_CONFIG_HEADER([src/config.h]) AC_CONFIG_HEADER([src/config.h])
@ -62,6 +62,13 @@ PKG_CHECK_MODULES([libssl], [libssl])
PKG_CHECK_MODULES([libcrypto], [libcrypto]) PKG_CHECK_MODULES([libcrypto], [libcrypto])
AC_CHECK_LIB([ssl], [TLS_method], AC_CHECK_LIB([ssl], [TLS_method],
[AC_DEFINE([HAVE_TLS_METHOD], [1], [Define to 1 if you have the 'TLS_method' function])]) [AC_DEFINE([HAVE_TLS_METHOD], [1], [Define to 1 if you have the 'TLS_method' function])])
PKG_CHECK_MODULES([ck], [ck >= 0], [
AS_VAR_APPEND([CFLAGS], [" $ck_CFLAGS"])
AS_VAR_APPEND([LIBS], [" $ck_LIBS"])
], [
AC_CHECK_HEADERS([ck_ring.h ck_pr.h],, [AC_MSG_ERROR([libck headers not found])])
AC_CHECK_LIB([ck], [ck_array_init],, [AC_MSG_ERROR([libck not found])])
])
# Check for LDNS # Check for LDNS
PKG_CHECK_MODULES([libldns], [libldns >= 1.7.0], [AC_DEFINE([HAVE_LDNS], [1], [Define to 1 if you have the LDNS library])], [ PKG_CHECK_MODULES([libldns], [libldns >= 1.7.0], [AC_DEFINE([HAVE_LDNS], [1], [Define to 1 if you have the LDNS library])], [

52
rpm/dnsperf.spec
View file

@ -1,5 +1,5 @@
Name: dnsperf Name: dnsperf
Version: 2.4.2 Version: 2.5.0
Release: 1%{?dist} Release: 1%{?dist}
Summary: DNS Performance Testing Tool Summary: DNS Performance Testing Tool
Group: Productivity/Networking/DNS/Utilities Group: Productivity/Networking/DNS/Utilities
@ -16,6 +16,7 @@ BuildRequires: libtool
BuildRequires: openssl-devel BuildRequires: openssl-devel
BuildRequires: pkgconfig BuildRequires: pkgconfig
BuildRequires: ldns-devel BuildRequires: ldns-devel
BuildRequires: ck-devel
%description %description
dnsperf and resperf are free tools developed by Nominum/Akamai (2006-2018) dnsperf and resperf are free tools developed by Nominum/Akamai (2006-2018)
@ -93,6 +94,55 @@ rm -rf $RPM_BUILD_ROOT
%changelog %changelog
* Fri Mar 12 2021 Jerry Lundström <lundstrom.jerry@gmail.com> 2.5.0-1
- Release 2.5.0
* This release adds re-connection support for TCP and DoT protocol,
new options to `resperf` and fixes a few bugs.
* `dnsperf` and `resperf` will now try to re-connect when they lose a TCP
or DoT connection, and with that comes a few new statistics metrics.
For `dnsperf`, if a connection oriented protocol is used, it will now
show the total number of re-connections made and the connection latency.
For `resperf` it also shows the total number of re-connections made and
the gnuplot data now contains the total number of connections made and
the connection latency for each interval.
Beside re-connection support, improvements have been made when it comes
to tracking socket readiness while connections are established which
should generate less warnings about "socket not ready".
* New `resperf` options:
- `-R`: Reopen the datafile if it runs out of data before the testing
is completed. This allows for long running tests on very small and
simple query datafile.
- `-F <fall_behind>`: Sets the maximum number of queries that can fall
behind being sent. `resperf` will stop when this many queries should
have been sent and it can be relative easy to hit if `-m <max_qps>`
is set too high.
The default is 1000 and setting it to zero (0) disables the check.
* Bugfixes:
- Fixed port handling for host/network format when setting client side
port with `-x`
- Fix support for quoted characters, `\000` and `\.`, in domain names,
this was lost when removing BIND's internal development libraries
- Fix issue in `dnsperf`, it would loop forever if no connection could
be established
- Fix potential buffer overrun in `resperf` when using response id
for `queries[]`
- DoT: Fix bug when sending from buffer
* Other changes:
- Always use `IPV6_V6ONLY` socket option for IPv6
- Add man-page on `-W` option added in v2.4.0
- Reformat man-pages
- `resperf`:
- Try and process more request each run to hopefully not hit max
outstanding so easy when high QPS
- Add default value to `-C` so it shows in help
* Commits:
9308361 man-page format
0e52fb4 man-page, opts, tuneups
e36211d stats
d9b9ba3 Response qid to index
2b2c37e fixes, reconnection
acd31e5 dname quote
de8f049 net
* Tue Feb 23 2021 Jerry Lundström <lundstrom.jerry@gmail.com> 2.4.2-1 * Tue Feb 23 2021 Jerry Lundström <lundstrom.jerry@gmail.com> 2.4.2-1
- Release 2.4.2 - Release 2.4.2
* This release fixes a few issues with reading of the datafile which * This release fixes a few issues with reading of the datafile which

2
src/Makefile.am
View file

@ -30,7 +30,7 @@ bin_PROGRAMS = dnsperf resperf
dist_bin_SCRIPTS = resperf-report dist_bin_SCRIPTS = resperf-report
_libperf_sources = datafile.c dns.c log.c net.c opt.c os.c strerror.c qtype.c \ _libperf_sources = datafile.c dns.c log.c net.c opt.c os.c strerror.c qtype.c \
edns.c tsig.c edns.c tsig.c net_udp.c net_tcp.c net_dot.c
_libperf_headers = datafile.h dns.h log.h net.h opt.h os.h util.h strerror.h \ _libperf_headers = datafile.h dns.h log.h net.h opt.h os.h util.h strerror.h \
list.h result.h buffer.h qtype.h edns.h tsig.h list.h result.h buffer.h qtype.h edns.h tsig.h

51
src/dns.c
View file

@ -52,18 +52,39 @@ const char* perf_dns_rcode_strings[] = {
perf_result_t perf_dname_fromstring(const char* str, size_t len, perf_buffer_t* target) perf_result_t perf_dname_fromstring(const char* str, size_t len, perf_buffer_t* target)
{ {
size_t label_len; size_t label_len, at;
const char* orig_str = str; const char* orig_str = str;
bool is_quoted;
if (perf_buffer_availablelength(target) < len) { if (perf_buffer_availablelength(target) < len) {
return PERF_R_NOSPACE; return PERF_R_NOSPACE;
} }
while (len) { while (len) {
for (label_len = 0; label_len < len; label_len++) { is_quoted = false;
if (*(str + label_len) == '.') { for (label_len = 0, at = 0; at < len;) {
if (*(str + at) == '\\') {
is_quoted = true;
at++;
if (at >= len)
return PERF_R_FAILURE;
if (*(str + at) >= '0' && *(str + at) <= '9') {
at++;
if (at >= len)
return PERF_R_FAILURE;
if (*(str + at) < '0' || *(str + at) > '9')
return PERF_R_FAILURE;
at++;
if (at >= len)
return PERF_R_FAILURE;
if (*(str + at) < '0' || *(str + at) > '9')
return PERF_R_FAILURE;
}
} else if (*(str + at) == '.') {
break; break;
} }
label_len++;
at++;
} }
if (!label_len) { if (!label_len) {
// Just a dot // Just a dot
@ -81,9 +102,33 @@ perf_result_t perf_dname_fromstring(const char* str, size_t len, perf_buffer_t*
return PERF_R_FAILURE; return PERF_R_FAILURE;
} }
perf_buffer_putuint8(target, label_len); perf_buffer_putuint8(target, label_len);
if (is_quoted) {
for (at = 0; at < len; at++) {
if (*(str + at) == '\\') {
at++;
if (at >= len)
return PERF_R_FAILURE;
if (*(str + at) >= '0' && *(str + at) <= '9') {
char b[4];
long v;
memcpy(b, str, 3);
b[3] = 0;
v = strtol(b, 0, 7);
if (v < 0 || v > 255)
return PERF_R_FAILURE;
perf_buffer_putuint8(target, (uint8_t)v);
continue;
}
}
perf_buffer_putmem(target, str + at, 1);
}
str += at;
len -= at;
} else {
perf_buffer_putmem(target, str, label_len); perf_buffer_putmem(target, str, label_len);
str += label_len; str += label_len;
len -= label_len; len -= label_len;
}
if (len < 2) { if (len < 2) {
// Last label/dot // Last label/dot
perf_buffer_putuint8(target, 0); perf_buffer_putuint8(target, 0);

272
src/dnsperf.1.in
View file

@ -20,67 +20,72 @@ dnsperf \- test the performance of a DNS server
.SH SYNOPSIS .SH SYNOPSIS
.hy 0 .hy 0
.ad l .ad l
\fBdnsperf\fR\ [\fB\-a\ \fIlocal_addr\fB\fR] \fBdnsperf\fR\ [\fB\-a\ \fIlocal_addr\fR]
[\fB\-b\ \fIbufsize\fB\fR] [\fB\-b\ \fIbufsize\fR]
[\fB\-c\ \fIclients\fB\fR] [\fB\-c\ \fIclients\fR]
[\fB\-d\ \fIdatafile\fB\fR] [\fB\-d\ \fIdatafile\fR]
[\fB\-D\fR] [\fB\-D\fR]
[\fB\-e\fR] [\fB\-e\fR]
[\fB\-E\ \fIcode:secret\fB\fR] [\fB\-E\ \fIcode:secret\fR]
[\fB\-f\ \fIfamily\fB\fR] [\fB\-f\ \fIfamily\fR]
[\fB\-h\fR] [\fB\-h\fR]
[\fB\-l\ \fIlimit\fB\fR] [\fB\-l\ \fIlimit\fR]
[\fB\-m\ \fImode\fB\fR] [\fB\-m\ \fImode\fR]
[\fB\-n\ \fIruns_through_file\fB\fR] [\fB\-n\ \fIruns_through_file\fR]
[\fB\-p\ \fIport\fB\fR] [\fB\-p\ \fIport\fR]
[\fB\-q\ \fInum_queries\fB\fR] [\fB\-q\ \fInum_queries\fR]
[\fB\-Q\ \fImax_qps\fB\fR] [\fB\-Q\ \fImax_qps\fR]
[\fB\-s\ \fIserver_addr\fB\fR] [\fB\-s\ \fIserver_addr\fR]
[\fB\-S\ \fIstats_interval\fB\fR] [\fB\-S\ \fIstats_interval\fR]
[\fB\-t\ \fItimeout\fB\fR] [\fB\-t\ \fItimeout\fR]
[\fB\-T\ \fIthreads\fB\fR] [\fB\-T\ \fIthreads\fR]
[\fB\-u\fR] [\fB\-u\fR]
[\fB\-v\fR] [\fB\-v\fR]
[\fB\-x\ \fIlocal_port\fB\fR] [\fB\-W\fR]
[\fB\-y\ \fI[alg:]name:secret\fB\fR] [\fB\-x\ \fIlocal_port\fR]
[\fB\-y\ \fI[alg:]name:secret\fR]
.ad .ad
.hy .hy
.SH DESCRIPTION .SH DESCRIPTION
\fBdnsperf\fR is a DNS server performance testing tool. It is primarily \fBdnsperf\fR is a DNS server performance testing tool.
intended for measuring the performance of authoritative DNS servers, but it It is primarily intended for measuring the performance of authoritative DNS
can also be used for measuring caching server performance in a closed servers, but it can also be used for measuring caching server performance in
laboratory environment. For testing caching servers resolving against the a closed laboratory environment.
live Internet, the \fBresperf\fR program is preferred. For testing caching servers resolving against the live Internet, the
\fBresperf\fR program is preferred.
It is recommended that \fBdnsperf\fR and the name server under test be run It is recommended that \fBdnsperf\fR and the name server under test be run
on separate machines, so that the CPU usage of \fBdnsperf\fR itself does not on separate machines, so that the CPU usage of \fBdnsperf\fR itself does not
slow down the name server. The two machines should be connected with a fast slow down the name server.
network, preferably a dedicated Gigabit Ethernet segment. Testing through a The two machines should be connected with a fast network, preferably a
router or firewall is not advisable. dedicated Gigabit Ethernet segment.
Testing through a router or firewall is not advisable.
.SS "Configuring the name server" .SS "Configuring the name server"
If using \fBdnsperf\fR to test an authoritative server, the name server If using \fBdnsperf\fR to test an authoritative server, the name server
under test should be set up to serve one or more zones similar in size and under test should be set up to serve one or more zones similar in size and
number to what the server is expected to serve in production. number to what the server is expected to serve in production.
Also, be sure to turn off recursion in the server's configuration (in BIND Also, be sure to turn off recursion in the server's configuration (in BIND
8/9, specify "recursion no;" in the options block). In BIND 8, you should 8/9, specify "recursion no;" in the options block).
also specify "fetch-glue no;"; otherwise the server may attempt to retrieve In BIND 8, you should also specify "fetch-glue no;"; otherwise the server
glue information from the Internet during the test, slowing it down by an may attempt to retrieve glue information from the Internet during the test,
unpredictable factor. slowing it down by an unpredictable factor.
.SS "Constructing a query input file" .SS "Constructing a query input file"
A \fBdnsperf\fR input file should contain a large and realistic set of A \fBdnsperf\fR input file should contain a large and realistic set of
queries, on the order of ten thousand to a million. The input file contains queries, on the order of ten thousand to a million.
one line per query, consisting of a domain name and an RR type name The input file contains one line per query, consisting of a domain name and
separated by a space. The class of the query is implicitly IN. an RR type name separated by a space.
The class of the query is implicitly IN.
When measuring the performance serving non-terminal zones such as the root When measuring the performance serving non-terminal zones such as the root
zone or TLDs, note that such servers spend most of their time providing zone or TLDs, note that such servers spend most of their time providing
referral responses, not authoritative answers. Therefore, a realistic input referral responses, not authoritative answers.
file might consist mostly of queries for type A for names *below*, not at, Therefore, a realistic input file might consist mostly of queries for type
the delegations present in the zone. For example, when testing the A for names *below*, not at, the delegations present in the zone.
performance of a server configured to be authoritative for the top-level For example, when testing the performance of a server configured to be
domain "fi.", which contains delegations for domains like "helsinki.fi" and authoritative for the top-level domain "fi.", which contains delegations
"turku.fi", the input file could contain lines like for domains like "helsinki.fi" and "turku.fi", the input file could contain
lines like
.RS .RS
.hy 0 .hy 0
@ -98,7 +103,8 @@ should be in a random order.
.SS "Constructing a dynamic update input file" .SS "Constructing a dynamic update input file"
To test dynamic update performance, \fBdnsperf\fR is run with the \fB\-u\fR To test dynamic update performance, \fBdnsperf\fR is run with the \fB\-u\fR
option, and the input file is constructed of blocks of lines describing option, and the input file is constructed of blocks of lines describing
dynamic update messages. The first line in a block contains the zone name: dynamic update messages.
The first line in a block contains the zone name:
.RS .RS
.hy 0 .hy 0
@ -108,12 +114,14 @@ example.com
.hy .hy
.RE .RE
Subsequent lines contain prerequisites, if there are any. Prerequisites can Subsequent lines contain prerequisites, if there are any.
specify that a name may or may not exist, an rrset may or may not exist, or Prerequisites can specify that a name may or may not exist, an rrset may or
an rrset exists and its rdata matches all specified rdata for that name and may not exist, or an rrset exists and its rdata matches all specified rdata
type. The keywords "require" and "prohibit" are followed by the appropriate for that name and type.
information. All relative names are considered to be relative to the zone The keywords "require" and "prohibit" are followed by the appropriate
name. The following lines show the 5 types of prerequisites. information.
All relative names are considered to be relative to the zone name.
The following lines show the 5 types of prerequisites.
.RS .RS
.hy 0 .hy 0
@ -128,10 +136,10 @@ prohibit x A
.RE .RE
Subsequent lines contain records to be added, records to be deleted, rrsets Subsequent lines contain records to be added, records to be deleted, rrsets
to be deleted, or names to be deleted. The keywords "add" or "delete" are to be deleted, or names to be deleted.
followed by the appropriate information. All relative names are considered The keywords "add" or "delete" are followed by the appropriate information.
to be relative to the zone name. The following lines show the 4 types of All relative names are considered to be relative to the zone name.
updates. The following lines show the 4 types of updates.
.RS .RS
.hy 0 .hy 0
@ -155,47 +163,49 @@ send
.RE .RE
.SS "Running the tests" .SS "Running the tests"
When running \fBdnsperf\fR, a data file (the \fB\-d\fR option) and server When running \fBdnsperf\fR, a data file (the \fB\-d\fR option) and server
(the \fB\-s\fR option) will normally be specified. The output of dnsperf is (the \fB\-s\fR option) will normally be specified.
mostly self-explanatory. Pay attention to the number of dropped packets The output of dnsperf is mostly self-explanatory.
reported - when running the test over a local Ethernet connection, it should Pay attention to the number of dropped packets reported - when running the
be zero. If one or more packets has been dropped, there may be a problem test over a local Ethernet connection, it should be zero.
with the network connection. In that case, the results should be considered If one or more packets has been dropped, there may be a problem with the
suspect and the test repeated. network connection.
In that case, the results should be considered suspect and the test repeated.
.SH OPTIONS .SH OPTIONS
\fB-a \fIlocal_addr\fB\fR \fB-a \fIlocal_addr\fR
.br .br
.RS .RS
Specifies the local address from which to send requests. The default is the Specifies the local address from which to send requests.
wildcard address. The default is the wildcard address.
.RE .RE
\fB-b \fIbufsize\fB\fR \fB-b \fIbufsize\fR
.br .br
.RS .RS
Sets the size of the socket's send and receive buffers, in kilobytes. If not Sets the size of the socket's send and receive buffers, in kilobytes.
specified, the operating system's default is used. If not specified, the operating system's default is used.
.RE .RE
\fB-c \fIclients\fB\fR \fB-c \fIclients\fR
.br .br
.RS .RS
Act as multiple clients. Requests are sent from multiple sockets. The Act as multiple clients.
default is to act as 1 client. Requests are sent from multiple sockets.
The default is to act as 1 client.
.RE .RE
\fB-d \fIdatafile\fB\fR \fB-d \fIdatafile\fR
.br .br
.RS .RS
Specifies the input data file. If not specified, \fBdnsperf\fR will read Specifies the input data file.
from standard input. If not specified, \fBdnsperf\fR will read from standard input.
.RE .RE
\fB-D\fR \fB-D\fR
.br .br
.RS .RS
Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent. This also enables Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent.
EDNS0, which is required for DNSSEC. This also enables EDNS0, which is required for DNSSEC.
.RE .RE
\fB-e\fR \fB-e\fR
@ -204,21 +214,21 @@ EDNS0, which is required for DNSSEC.
Enables EDNS0 [RFC2671], by adding an OPT record to all packets sent. Enables EDNS0 [RFC2671], by adding an OPT record to all packets sent.
.RE .RE
\fB-E \fIcode:value\fB\fR \fB-E \fIcode:value\fR
.br .br
.RS .RS
Add an EDNS [RFC2671] option to all packets sent, using the specified Add an EDNS [RFC2671] option to all packets sent, using the specified
numeric option code and value expressed as a a hex-encoded string. This also numeric option code and value expressed as a a hex-encoded string.
enables EDNS0. This also enables EDNS0.
.RE .RE
\fB-f \fIfamily\fB\fR \fB-f \fIfamily\fR
.br .br
.RS .RS
Specifies the address family used for sending DNS packets. The possible Specifies the address family used for sending DNS packets.
values are "inet", "inet6", or "any". If "any" (the default value) is The possible values are "inet", "inet6", or "any".
specified, \fBdnsperf\fR will use whichever address family is appropriate If "any" (the default value) is specified, \fBdnsperf\fR will use whichever
for the server it is sending packets to. address family is appropriate for the server it is sending packets to.
.RE .RE
\fB-h\fR \fB-h\fR
@ -227,111 +237,129 @@ for the server it is sending packets to.
Print a usage statement and exit. Print a usage statement and exit.
.RE .RE
\fB-l \fIlimit\fB\fR \fB-l \fIlimit\fR
.br .br
.RS .RS
Specifies a time limit for the run, in seconds. This may cause the input to Specifies a time limit for the run, in seconds.
be read multiple times, or only some of the input to be read. The default This may cause the input to be read multiple times, or only some of the
behavior is to read the input once, and have no specific time limit. input to be read.
The default behavior is to read the input once, and have no specific time
limit.
.RE .RE
\fB-n \fIruns_through_file\fB\fR \fB-n \fIruns_through_file\fR
.br .br
.RS .RS
Run through the input file at most this many times. If no time limit is set, Run through the input file at most this many times.
the file will be read exactly this number of times; if a time limit is set, If no time limit is set, the file will be read exactly this number of
the file may be read fewer times. times; if a time limit is set, the file may be read fewer times.
.RE .RE
\fB-p \fIport\fB\fR \fB-p \fIport\fR
.br .br
.RS .RS
Sets the port on which the DNS packets are sent. If not specified, the Sets the port on which the DNS packets are sent.
standard DNS port (udp/tcp 53, dot/tls 853) is used. If not specified, the standard DNS port (udp/tcp 53, DoT 853) is used.
.RE .RE
\fB-q \fInum_queries\fB\fR \fB-q \fInum_queries\fR
.br .br
.RS .RS
Sets the maximum number of outstanding requests. When this value is reached, Sets the maximum number of outstanding requests.
\fBdnsperf\fR will not send any more requests until either responses are When this value is reached, \fBdnsperf\fR will not send any more requests
received or requests time out. The default value is 100. until either responses are received or requests time out.
The default value is 100.
.RE .RE
\fB-Q \fImax_qps\fB\fR \fB-Q \fImax_qps\fR
.br .br
.RS .RS
Limits the number of requests per second. There is no default limit. Limits the number of requests per second.
There is no default limit.
.RE .RE
\fB-m \fImode\fB\fR \fB-m \fImode\fR
.br .br
.RS .RS
Specifies the transport mode to use, "udp", "tcp" or "dot"/"tls". Specifies the transport mode to use, "udp", "tcp" or "dot".
Default is "udp". Default is "udp".
.RE .RE
\fB-s \fIserver_addr\fB\fR \fB-s \fIserver_addr\fR
.br .br
.RS .RS
Specifies the name or address of the server to which requests will be sent. Specifies the name or address of the server to which requests will be sent.
The default is the loopback address, 127.0.0.1. The default is the loopback address, 127.0.0.1.
.RE .RE
\fB-S \fIstats_interval\fB\fR \fB-S \fIstats_interval\fR
.br .br
.RS .RS
If this parameter is specified, a count of the number of queries per second If this parameter is specified, a count of the number of queries per second
during the interval will be printed out every stats_interval seconds. during the interval will be printed out every stats_interval seconds.
.RE .RE
\fB-t \fItimeout\fB\fR \fB-t \fItimeout\fR
.br .br
.RS .RS
Specifies the request timeout value, in seconds. \fBdnsperf\fR will no Specifies the request timeout value, in seconds.
longer wait for a response to a particular request after this many seconds \fBdnsperf\fR will no longer wait for a response to a particular request
have elapsed. The default is 5 seconds. after this many seconds have elapsed.
The default is 5 seconds.
.RE .RE
\fB-T \fIthreads\fB\fR \fB-T \fIthreads\fR
.br .br
.RS .RS
Run multiple client threads. By default, \fBdnsperf\fR uses one thread for Run multiple client threads.
sending requests and one thread for receiving responses. If this option is By default, \fBdnsperf\fR uses one thread for sending requests and one
specified, \fBdnsperf\fR will instead use N pairs of send/receive threads. thread for receiving responses.
If this option is specified, \fBdnsperf\fR will instead use N pairs of
send/receive threads.
.RE .RE
\fB-u\fR \fB-u\fR
.br .br
.RS .RS
Instructs \fBdnsperf\fR to send DNS dynamic update messages, rather than Instructs \fBdnsperf\fR to send DNS dynamic update messages, rather than
queries. The format of the input file is different in this case; see the queries.
The format of the input file is different in this case; see the
"Constructing a dynamic update input file" section for more details. "Constructing a dynamic update input file" section for more details.
.RE .RE
\fB-v\fR \fB-v\fR
.br .br
.RS .RS
Enables verbose mode. The DNS RCODE of each response will be reported to Enables verbose mode.
standard output when the response is received, as will the latency. If a The DNS RCODE of each response will be reported to standard output when
query times out, it will be reported with the special string "T" instead of the response is received, as will the latency.
a normal DNS RCODE. If a query is interrupted, it will be reported with the If a query times out, it will be reported with the special string "T"
special string "I". Additional information regarding network readiness and instead of a normal DNS RCODE.
congestion will also be reported. If a query is interrupted, it will be reported with the special string "I".
Additional information regarding network readiness and congestion will
also be reported.
.RE .RE
\fB-x \fIlocal_port\fB\fR \fB-W\fR
.br .br
.RS .RS
Specifies the local port from which to send requests. The default is the Log warnings and errors to standard output instead of standard error making
wildcard port (0). it easier for script, test and automation to capture all output.
If acting as multiple clients and the wildcard port is used, each client
will use a different random port. If a port is specified, the clients will
use a range of ports starting with the specified one.
.RE .RE
\fB-y \fI[alg:]name:secret\fB\fR \fB-x \fIlocal_port\fR
.br
.RS
Specifies the local port from which to send requests.
The default is the wildcard port (0).
If acting as multiple clients and the wildcard port is used, each client
will use a different random port.
If a port is specified, the clients will use a range of ports starting with
the specified one.
.RE
\fB-y \fI[alg:]name:secret\fR
.br .br
.RS .RS
Add a TSIG record [RFC2845] to all packets sent, using the specified TSIG Add a TSIG record [RFC2845] to all packets sent, using the specified TSIG

123
src/dnsperf.c
View file

@ -51,8 +51,8 @@
#define DEFAULT_SERVER_NAME "127.0.0.1" #define DEFAULT_SERVER_NAME "127.0.0.1"
#define DEFAULT_SERVER_PORT 53 #define DEFAULT_SERVER_PORT 53
#define DEFAULT_SERVER_TLS_PORT 853 #define DEFAULT_SERVER_DOT_PORT 853
#define DEFAULT_SERVER_PORTS "udp/tcp 53 or dot/tls 853" #define DEFAULT_SERVER_PORTS "udp/tcp 53 or DoT 853"
#define DEFAULT_LOCAL_PORT 0 #define DEFAULT_LOCAL_PORT 0
#define DEFAULT_MAX_OUTSTANDING 100 #define DEFAULT_MAX_OUTSTANDING 100
#define DEFAULT_TIMEOUT 5 #define DEFAULT_TIMEOUT 5
@ -111,6 +111,14 @@ typedef struct {
uint64_t latency_sum_squares; uint64_t latency_sum_squares;
uint64_t latency_min; uint64_t latency_min;
uint64_t latency_max; uint64_t latency_max;
uint64_t num_conn_reconnect;
uint64_t num_conn_completed;
uint64_t conn_latency_sum;
uint64_t conn_latency_sum_squares;
uint64_t conn_latency_min;
uint64_t conn_latency_max;
} stats_t; } stats_t;
typedef perf_list(struct query_info) query_list; typedef perf_list(struct query_info) query_list;
@ -142,7 +150,7 @@ typedef struct {
unsigned int nsocks; unsigned int nsocks;
int current_sock; int current_sock;
struct perf_net_socket* socks; struct perf_net_socket** socks;
bool done_sending; bool done_sending;
uint64_t done_send_time; uint64_t done_send_time;
@ -194,8 +202,13 @@ print_initial_status(const config_t* config)
printf("\n"); printf("\n");
perf_sockaddr_format(&config->server_addr, buf, sizeof(buf)); perf_sockaddr_format(&config->server_addr, buf, sizeof(buf));
printf("[Status] Sending %s (to %s)\n", if (perf_sockaddr_isinet6(&config->server_addr)) {
config->updates ? "updates" : "queries", buf); printf("[Status] Sending %s (to [%s]:%d)\n",
config->updates ? "updates" : "queries", buf, perf_sockaddr_port(&config->server_addr));
} else {
printf("[Status] Sending %s (to %s:%d)\n",
config->updates ? "updates" : "queries", buf, perf_sockaddr_port(&config->server_addr));
}
now = time(NULL); now = time(NULL);
printf("[Status] Started at: %s", ctime_r(&now, ct)); printf("[Status] Started at: %s", ctime_r(&now, ct));
@ -310,6 +323,27 @@ print_statistics(const config_t* config, const times_t* times, stats_t* stats)
} }
printf("\n"); printf("\n");
if (!stats->num_conn_completed) {
return;
}
printf("Connection Statistics:\n\n");
printf(" Reconnections: %" PRIu64 "\n\n", stats->num_conn_reconnect);
latency_avg = PERF_SAFE_DIV(stats->conn_latency_sum, stats->num_conn_completed);
printf(" Average Latency (s): %u.%06u (min %u.%06u, max %u.%06u)\n",
(unsigned int)(latency_avg / MILLION),
(unsigned int)(latency_avg % MILLION),
(unsigned int)(stats->conn_latency_min / MILLION),
(unsigned int)(stats->conn_latency_min % MILLION),
(unsigned int)(stats->conn_latency_max / MILLION),
(unsigned int)(stats->conn_latency_max % MILLION));
if (stats->num_conn_completed > 1) {
printf(" Latency StdDev (s): %f\n",
stddev(stats->conn_latency_sum_squares, stats->conn_latency_sum, stats->num_conn_completed) / MILLION);
}
printf("\n");
} }
static void static void
@ -339,6 +373,16 @@ sum_stats(const config_t* config, stats_t* total)
total->latency_min = stats->latency_min; total->latency_min = stats->latency_min;
if (stats->latency_max > total->latency_max) if (stats->latency_max > total->latency_max)
total->latency_max = stats->latency_max; total->latency_max = stats->latency_max;
total->num_conn_completed += stats->num_conn_completed;
total->num_conn_reconnect += stats->num_conn_reconnect;
total->conn_latency_sum += stats->conn_latency_sum;
total->conn_latency_sum_squares += stats->conn_latency_sum_squares;
if (stats->conn_latency_min < total->conn_latency_min || i == 0)
total->conn_latency_min = stats->conn_latency_min;
if (stats->conn_latency_max > total->conn_latency_max)
total->conn_latency_max = stats->conn_latency_max;
} }
} }
@ -378,7 +422,7 @@ setup(int argc, char** argv, config_t* config)
perf_opt_add('f', perf_opt_string, "family", perf_opt_add('f', perf_opt_string, "family",
"address family of DNS transport, inet or inet6", "any", "address family of DNS transport, inet or inet6", "any",
&family); &family);
perf_opt_add('m', perf_opt_string, "mode", "set transport mode: udp, tcp or dot/tls", "udp", &mode); perf_opt_add('m', perf_opt_string, "mode", "set transport mode: udp, tcp or dot", "udp", &mode);
perf_opt_add('s', perf_opt_string, "server_addr", perf_opt_add('s', perf_opt_string, "server_addr",
"the server to query", DEFAULT_SERVER_NAME, &server_name); "the server to query", DEFAULT_SERVER_NAME, &server_name);
perf_opt_add('p', perf_opt_port, "port", perf_opt_add('p', perf_opt_port, "port",
@ -449,7 +493,7 @@ setup(int argc, char** argv, config_t* config)
config->mode = perf_net_parsemode(mode); config->mode = perf_net_parsemode(mode);
if (!server_port) { if (!server_port) {
server_port = config->mode == sock_tls ? DEFAULT_SERVER_TLS_PORT : DEFAULT_SERVER_PORT; server_port = config->mode == sock_dot ? DEFAULT_SERVER_DOT_PORT : DEFAULT_SERVER_PORT;
} }
if (family != NULL) if (family != NULL)
@ -572,6 +616,7 @@ do_send(void* arg)
unsigned int length; unsigned int length;
int n, i, any_inprogress = 0; int n, i, any_inprogress = 0;
perf_result_t result; perf_result_t result;
bool all_fail;
tinfo = (threadinfo_t*)arg; tinfo = (threadinfo_t*)arg;
config = tinfo->config; config = tinfo->config;
@ -619,14 +664,16 @@ do_send(void* arg)
q->timestamp = UINT64_MAX; q->timestamp = UINT64_MAX;
i = tinfo->nsocks * 2; i = tinfo->nsocks * 2;
all_fail = true;
while (i--) { while (i--) {
q->sock = &tinfo->socks[tinfo->current_sock++ % tinfo->nsocks]; q->sock = tinfo->socks[tinfo->current_sock++ % tinfo->nsocks];
switch (perf_net_sockready(q->sock, threadpipe[0], TIMEOUT_CHECK_TIME)) { switch (perf_net_sockready(q->sock, threadpipe[0], TIMEOUT_CHECK_TIME)) {
case 0: case 0:
if (config->verbose) { if (config->verbose) {
perf_log_warning("socket %p not ready", q->sock); perf_log_warning("socket %p not ready", q->sock);
} }
q->sock = 0; q->sock = 0;
all_fail = false;
continue; continue;
case -1: case -1:
if (errno == EINPROGRESS) { if (errno == EINPROGRESS) {
@ -637,12 +684,19 @@ do_send(void* arg)
if (config->verbose) { if (config->verbose) {
perf_log_warning("socket %p readiness check timed out", q->sock); perf_log_warning("socket %p readiness check timed out", q->sock);
} }
q->sock = 0;
continue;
default: default:
break; break;
} }
all_fail = false;
break; break;
}; };
if (all_fail) {
perf_log_fatal("all sockets reported failure, can not continue");
}
if (!q->sock) { if (!q->sock) {
query_move(tinfo, q, prepend_unused); query_move(tinfo, q, prepend_unused);
PERF_UNLOCK(&tinfo->lock); PERF_UNLOCK(&tinfo->lock);
@ -686,7 +740,7 @@ do_send(void* arg)
} }
q->timestamp = now; q->timestamp = now;
n = perf_net_sendto(q->sock, base, length, 0, &config->server_addr.sa.sa, n = perf_net_sendto(q->sock, qid, base, length, 0, &config->server_addr.sa.sa,
config->server_addr.length); config->server_addr.length);
if (n < 0) { if (n < 0) {
if (errno == EINPROGRESS) { if (errno == EINPROGRESS) {
@ -719,7 +773,7 @@ do_send(void* arg)
while (any_inprogress) { while (any_inprogress) {
any_inprogress = 0; any_inprogress = 0;
for (i = 0; i < tinfo->nsocks; i++) { for (i = 0; i < tinfo->nsocks; i++) {
if (perf_net_sockready(&tinfo->socks[i], threadpipe[0], TIMEOUT_CHECK_TIME) == -1 && errno == EINPROGRESS) { if (perf_net_sockready(tinfo->socks[i], threadpipe[0], TIMEOUT_CHECK_TIME) == -1 && errno == EINPROGRESS) {
any_inprogress = 1; any_inprogress = 1;
} }
} }
@ -788,7 +842,7 @@ recv_one(threadinfo_t* tinfo, int which_sock,
packet_header = (uint16_t*)packet_buffer; packet_header = (uint16_t*)packet_buffer;
n = perf_net_recv(&tinfo->socks[which_sock], packet_buffer, packet_size, 0); n = perf_net_recv(tinfo->socks[which_sock], packet_buffer, packet_size, 0);
now = perf_get_time(); now = perf_get_time();
if (n < 0) { if (n < 0) {
*saved_errnop = errno; *saved_errnop = errno;
@ -799,7 +853,7 @@ recv_one(threadinfo_t* tinfo, int which_sock,
*saved_errnop = EAGAIN; *saved_errnop = EAGAIN;
return false; return false;
} }
recvd->sock = &tinfo->socks[which_sock]; recvd->sock = tinfo->socks[which_sock];
recvd->qid = ntohs(packet_header[0]); recvd->qid = ntohs(packet_header[0]);
recvd->rcode = ntohs(packet_header[1]) & 0xF; recvd->rcode = ntohs(packet_header[1]) & 0xF;
recvd->size = n; recvd->size = n;
@ -1045,6 +1099,36 @@ per_thread(uint32_t total, uint32_t nthreads, unsigned int offset)
return value; return value;
} }
static void perf__net_sent(struct perf_net_socket* sock, uint16_t qid)
{
threadinfo_t* tinfo = (threadinfo_t*)sock->data;
query_info* q;
q = &tinfo->queries[qid];
if (q->timestamp != UINT64_MAX) {
q->timestamp = perf_get_time();
}
}
static void perf__net_event(struct perf_net_socket* sock, perf_socket_event_t event, uint64_t elapsed_time)
{
stats_t* stats = &((threadinfo_t*)sock->data)->stats;
switch (event) {
case perf_socket_event_reconnect:
stats->num_conn_reconnect++;
case perf_socket_event_connect:
stats->num_conn_completed++;
stats->conn_latency_sum += elapsed_time;
stats->conn_latency_sum_squares += (elapsed_time * elapsed_time);
if (elapsed_time < stats->conn_latency_min || stats->num_conn_completed == 1)
stats->conn_latency_min = elapsed_time;
if (elapsed_time > stats->conn_latency_max)
stats->conn_latency_max = elapsed_time;
}
}
static void static void
threadinfo_init(threadinfo_t* tinfo, const config_t* config, threadinfo_init(threadinfo_t* tinfo, const config_t* config,
const times_t* times) const times_t* times)
@ -1091,11 +1175,18 @@ threadinfo_init(threadinfo_t* tinfo, const config_t* config,
socket_offset = 0; socket_offset = 0;
for (i = 0; i < offset; i++) for (i = 0; i < offset; i++)
socket_offset += threads[i].nsocks; socket_offset += threads[i].nsocks;
for (i = 0; i < tinfo->nsocks; i++) for (i = 0; i < tinfo->nsocks; i++) {
tinfo->socks[i] = perf_net_opensocket(config->mode, &config->server_addr, tinfo->socks[i] = perf_net_opensocket(config->mode, &config->server_addr,
&config->local_addr, &config->local_addr,
socket_offset++, socket_offset++,
config->bufsize); config->bufsize);
if (!tinfo->socks[i]) {
perf_log_fatal("perf_net_opensocket(): no socket returned, out of memory?");
}
tinfo->socks[i]->data = tinfo;
tinfo->socks[i]->sent = perf__net_sent;
tinfo->socks[i]->event = perf__net_event;
}
tinfo->current_sock = 0; tinfo->current_sock = 0;
PERF_THREAD(&tinfo->receiver, do_recv, tinfo); PERF_THREAD(&tinfo->receiver, do_recv, tinfo);
@ -1118,7 +1209,7 @@ threadinfo_cleanup(threadinfo_t* tinfo, times_t* times)
if (interrupted) if (interrupted)
cancel_queries(tinfo); cancel_queries(tinfo);
for (i = 0; i < tinfo->nsocks; i++) for (i = 0; i < tinfo->nsocks; i++)
perf_net_close(&tinfo->socks[i]); perf_net_close(tinfo->socks[i]);
if (tinfo->last_recv > times->end_time) if (tinfo->last_recv > times->end_time)
times->end_time = tinfo->last_recv; times->end_time = tinfo->last_recv;
} }
@ -1152,8 +1243,8 @@ int main(int argc, char** argv)
perf_os_blocksignal(SIGINT, true); perf_os_blocksignal(SIGINT, true);
switch (config.mode) { switch (config.mode) {
case sock_tcp: case sock_tcp:
case sock_tls: case sock_dot:
// block SIGPIPE for TCP/TLS mode, if connection is closed it will generate a signal // block SIGPIPE for TCP/DOT mode, if connection is closed it will generate a signal
perf_os_blocksignal(SIGPIPE, true); perf_os_blocksignal(SIGPIPE, true);
break; break;
default: default:

498
src/net.c
View file

@ -23,25 +23,28 @@
#include "log.h" #include "log.h"
#include "opt.h" #include "opt.h"
#include "os.h"
#include "strerror.h"
#include <errno.h> #include <errno.h>
#include <fcntl.h> #include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h> #include <string.h>
#include <unistd.h>
#include <errno.h>
#include <poll.h> #include <poll.h>
#include <openssl/err.h>
#include <netdb.h> #include <netdb.h>
#include <arpa/inet.h> #include <arpa/inet.h>
#define TCP_RECV_BUF_SIZE (16 * 1024) enum perf_net_mode perf_net_parsemode(const char* mode)
#define TCP_SEND_BUF_SIZE (4 * 1024) {
if (!strcmp(mode, "udp")) {
return sock_udp;
} else if (!strcmp(mode, "tcp")) {
return sock_tcp;
} else if (!strcmp(mode, "tls") || !strcmp(mode, "dot")) {
return sock_dot;
}
static SSL_CTX* ssl_ctx = 0; perf_log_warning("invalid socket mode");
perf_opt_usage();
exit(1);
}
int perf_net_parsefamily(const char* family) int perf_net_parsefamily(const char* family)
{ {
@ -49,10 +52,8 @@ int perf_net_parsefamily(const char* family)
return AF_UNSPEC; return AF_UNSPEC;
else if (strcmp(family, "inet") == 0) else if (strcmp(family, "inet") == 0)
return AF_INET; return AF_INET;
#ifdef AF_INET6
else if (strcmp(family, "inet6") == 0) else if (strcmp(family, "inet6") == 0)
return AF_INET6; return AF_INET6;
#endif
else { else {
fprintf(stderr, "invalid family %s\n", family); fprintf(stderr, "invalid family %s\n", family);
perf_opt_usage(); perf_opt_usage();
@ -82,9 +83,9 @@ in_port_t perf_sockaddr_port(const perf_sockaddr_t* sockaddr)
{ {
switch (sockaddr->sa.sa.sa_family) { switch (sockaddr->sa.sa.sa_family) {
case AF_INET: case AF_INET:
return sockaddr->sa.sin.sin_port; return ntohs(sockaddr->sa.sin.sin_port);
case AF_INET6: case AF_INET6:
return sockaddr->sa.sin6.sin6_port; return ntohs(sockaddr->sa.sin6.sin6_port);
default: default:
break; break;
} }
@ -95,10 +96,10 @@ void perf_sockaddr_setport(perf_sockaddr_t* sockaddr, in_port_t port)
{ {
switch (sockaddr->sa.sa.sa_family) { switch (sockaddr->sa.sa.sa_family) {
case AF_INET: case AF_INET:
sockaddr->sa.sin.sin_port = port; sockaddr->sa.sin.sin_port = htons(port);
break; break;
case AF_INET6: case AF_INET6:
sockaddr->sa.sin6.sin6_port = port; sockaddr->sa.sin6.sin6_port = htons(port);
break; break;
default: default:
break; break;
@ -188,76 +189,15 @@ void perf_net_parselocal(int family, const char* name, unsigned int port,
exit(1); exit(1);
} }
struct perf_net_socket perf_net_opensocket(enum perf_net_mode mode, const perf_sockaddr_t* server, const perf_sockaddr_t* local, unsigned int offset, int bufsize) struct perf_net_socket* perf_net_opensocket(enum perf_net_mode mode, const perf_sockaddr_t* server, const perf_sockaddr_t* local, unsigned int offset, size_t bufsize)
{ {
int family;
perf_sockaddr_t tmp;
int port; int port;
int ret; perf_sockaddr_t tmp;
int flags;
struct perf_net_socket sock = { .mode = mode, .is_ready = 1 };
family = server->sa.sa.sa_family; if (server->sa.sa.sa_family != local->sa.sa.sa_family) {
if (local->sa.sa.sa_family != family) {
perf_log_fatal("server and local addresses have different families"); perf_log_fatal("server and local addresses have different families");
} }
switch (mode) {
case sock_udp:
sock.fd = socket(family, SOCK_DGRAM, 0);
break;
case sock_tls:
if (pthread_mutex_init(&sock.lock, 0)) {
perf_log_fatal("pthread_mutex_init() failed");
}
if ((sock.fd = socket(family, SOCK_STREAM, 0)) < 0) {
char __s[256];
perf_log_fatal("socket: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
if (!ssl_ctx) {
#ifdef HAVE_TLS_METHOD
if (!(ssl_ctx = SSL_CTX_new(TLS_method()))) {
perf_log_fatal("SSL_CTX_new(): %s", ERR_error_string(ERR_get_error(), 0));
}
if (!SSL_CTX_set_min_proto_version(ssl_ctx, TLS1_2_VERSION)) {
perf_log_fatal("SSL_CTX_set_min_proto_version(TLS1_2_VERSION): %s", ERR_error_string(ERR_get_error(), 0));
}
#else
if (!(ssl_ctx = SSL_CTX_new(SSLv23_client_method()))) {
perf_log_fatal("SSL_CTX_new(): %s", ERR_error_string(ERR_get_error(), 0));
}
#endif
}
if (!(sock.ssl = SSL_new(ssl_ctx))) {
perf_log_fatal("SSL_new(): %s", ERR_error_string(ERR_get_error(), 0));
}
if (!(ret = SSL_set_fd(sock.ssl, sock.fd))) {
perf_log_fatal("SSL_set_fd(): %s", ERR_error_string(SSL_get_error(sock.ssl, ret), 0));
}
break;
case sock_tcp:
sock.fd = socket(family, SOCK_STREAM, 0);
break;
default:
perf_log_fatal("perf_net_opensocket(): invalid mode");
}
if (sock.fd == -1) {
char __s[256];
perf_log_fatal("socket: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
#if defined(AF_INET6) && defined(IPV6_V6ONLY)
if (family == AF_INET6) {
int on = 1;
if (setsockopt(sock.fd, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) == -1) {
perf_log_warning("setsockopt(IPV6_V6ONLY) failed");
}
}
#endif
tmp = *local; tmp = *local;
port = perf_sockaddr_port(&tmp); port = perf_sockaddr_port(&tmp);
if (port != 0 && offset != 0) { if (port != 0 && offset != 0) {
@ -267,398 +207,16 @@ struct perf_net_socket perf_net_opensocket(enum perf_net_mode mode, const perf_s
perf_sockaddr_setport(&tmp, port); perf_sockaddr_setport(&tmp, port);
} }
if (bind(sock.fd, &tmp.sa.sa, tmp.length) == -1) { switch (mode) {
char __s[256]; case sock_udp:
perf_log_fatal("bind: %s", perf_strerror_r(errno, __s, sizeof(__s))); return perf_net_udp_opensocket(server, &tmp, bufsize);
}
if (bufsize > 0) {
bufsize *= 1024;
ret = setsockopt(sock.fd, SOL_SOCKET, SO_RCVBUF,
&bufsize, sizeof(bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_RCVBUF) failed");
ret = setsockopt(sock.fd, SOL_SOCKET, SO_SNDBUF,
&bufsize, sizeof(bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_SNDBUF) failed");
}
flags = fcntl(sock.fd, F_GETFL, 0);
if (flags < 0)
perf_log_fatal("fcntl(F_GETFL)");
ret = fcntl(sock.fd, F_SETFL, flags | O_NONBLOCK);
if (ret < 0)
perf_log_fatal("fcntl(F_SETFL)");
if (mode == sock_tcp || mode == sock_tls) {
if (connect(sock.fd, &server->sa.sa, server->length)) {
if (errno == EINPROGRESS) {
sock.is_ready = 0;
} else {
char __s[256];
perf_log_fatal("connect() failed: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
}
sock.recvbuf = malloc(TCP_RECV_BUF_SIZE);
sock.at = 0;
sock.have_more = 0;
sock.sendbuf = malloc(TCP_SEND_BUF_SIZE);
if (!sock.recvbuf || !sock.sendbuf) {
perf_log_fatal("perf_net_opensocket() failed: unable to allocate buffers");
}
}
return sock;
}
ssize_t perf_net_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags)
{
switch (sock->mode) {
case sock_tls: {
ssize_t n;
uint16_t dnslen, dnslen2;
if (!sock->have_more) {
if (pthread_mutex_lock(&sock->lock)) {
perf_log_fatal("pthread_mutex_lock() failed");
}
if (!sock->is_ready) {
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
errno = EAGAIN;
return -1;
}
n = SSL_read(sock->ssl, sock->recvbuf + sock->at, TCP_RECV_BUF_SIZE - sock->at);
if (n < 0) {
int err = SSL_get_error(sock->ssl, n);
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
if (err == SSL_ERROR_WANT_READ) {
errno = EAGAIN;
} else {
errno = EBADF;
}
return -1;
}
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
sock->at += n;
if (sock->at < 3) {
errno = EAGAIN;
return -1;
}
}
memcpy(&dnslen, sock->recvbuf, 2);
dnslen = ntohs(dnslen);
if (sock->at < dnslen + 2) {
errno = EAGAIN;
return -1;
}
memcpy(buf, sock->recvbuf + 2, len < dnslen ? len : dnslen);
memmove(sock->recvbuf, sock->recvbuf + 2 + dnslen, sock->at - 2 - dnslen);
sock->at -= 2 + dnslen;
if (sock->at > 2) {
memcpy(&dnslen2, sock->recvbuf, 2);
dnslen2 = ntohs(dnslen2);
if (sock->at >= dnslen2 + 2) {
sock->have_more = 1;
return dnslen;
}
}
sock->have_more = 0;
return dnslen;
}
case sock_tcp: {
ssize_t n;
uint16_t dnslen, dnslen2;
if (!sock->have_more) {
n = recv(sock->fd, sock->recvbuf + sock->at, TCP_RECV_BUF_SIZE - sock->at, flags);
if (n < 0) {
if (errno == ECONNRESET) {
// Treat connection reset like try again until reconnection features are in
errno = EAGAIN;
}
return n;
}
sock->at += n;
if (sock->at < 3) {
errno = EAGAIN;
return -1;
}
}
memcpy(&dnslen, sock->recvbuf, 2);
dnslen = ntohs(dnslen);
if (sock->at < dnslen + 2) {
errno = EAGAIN;
return -1;
}
memcpy(buf, sock->recvbuf + 2, len < dnslen ? len : dnslen);
memmove(sock->recvbuf, sock->recvbuf + 2 + dnslen, sock->at - 2 - dnslen);
sock->at -= 2 + dnslen;
if (sock->at > 2) {
memcpy(&dnslen2, sock->recvbuf, 2);
dnslen2 = ntohs(dnslen2);
if (sock->at >= dnslen2 + 2) {
sock->have_more = 1;
return dnslen;
}
}
sock->have_more = 0;
return dnslen;
}
default:
break;
}
return recv(sock->fd, buf, len, flags);
}
ssize_t perf_net_sendto(struct perf_net_socket* sock, const void* buf, size_t len, int flags,
const struct sockaddr* dest_addr, socklen_t addrlen)
{
switch (sock->mode) {
case sock_tls: {
size_t send = len < TCP_SEND_BUF_SIZE - 2 ? len : (TCP_SEND_BUF_SIZE - 2);
// TODO: We only send what we can send, because we can't continue sending
uint16_t dnslen = htons(send);
ssize_t n;
memcpy(sock->sendbuf, &dnslen, 2);
memcpy(sock->sendbuf + 2, buf, send);
if (pthread_mutex_lock(&sock->lock)) {
perf_log_fatal("pthread_mutex_lock() failed");
}
n = SSL_write(sock->ssl, sock->sendbuf, send + 2);
if (n < 0) {
perf_log_warning("SSL_write(): %s", ERR_error_string(SSL_get_error(sock->ssl, n), 0));
errno = EBADF;
}
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
if (n > 0 && n < send + 2) {
sock->sending = n;
sock->flags = flags;
memcpy(&sock->dest_addr, dest_addr, addrlen);
sock->addrlen = addrlen;
sock->is_ready = 0;
errno = EINPROGRESS;
return -1;
}
return n > 0 ? n - 2 : n;
}
case sock_tcp: {
size_t send = len < TCP_SEND_BUF_SIZE - 2 ? len : (TCP_SEND_BUF_SIZE - 2);
// TODO: We only send what we can send, because we can't continue sending
uint16_t dnslen = htons(send);
ssize_t n;
memcpy(sock->sendbuf, &dnslen, 2);
memcpy(sock->sendbuf + 2, buf, send);
n = sendto(sock->fd, sock->sendbuf, send + 2, flags, dest_addr, addrlen);
if (n > 0 && n < send + 2) {
sock->sending = n;
sock->flags = flags;
memcpy(&sock->dest_addr, dest_addr, addrlen);
sock->addrlen = addrlen;
sock->is_ready = 0;
errno = EINPROGRESS;
return -1;
}
return n > 0 ? n - 2 : n;
}
default:
break;
}
return sendto(sock->fd, buf, len, flags, dest_addr, addrlen);
}
int perf_net_close(struct perf_net_socket* sock)
{
return close(sock->fd);
}
int perf_net_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b)
{
return sock_a->fd == sock_b->fd;
}
enum perf_net_mode perf_net_parsemode(const char* mode)
{
if (!strcmp(mode, "udp")) {
return sock_udp;
} else if (!strcmp(mode, "tcp")) {
return sock_tcp;
} else if (!strcmp(mode, "tls") || !strcmp(mode, "dot")) {
return sock_tls;
}
perf_log_warning("invalid socket mode");
perf_opt_usage();
exit(1);
}
int perf_net_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{
if (sock->is_ready) {
return 1;
}
switch (sock->mode) {
case sock_tls: {
int ret;
if (sock->sending) {
uint16_t dnslen;
ssize_t n;
memcpy(&dnslen, sock->sendbuf, 2);
dnslen = ntohs(dnslen);
if (pthread_mutex_lock(&sock->lock)) {
perf_log_fatal("pthread_mutex_lock() failed");
}
n = SSL_write(sock->ssl, sock->sendbuf + sock->sending, dnslen + 2 - sock->sending);
if (n < 1) {
if (n < 0) {
perf_log_warning("SSL_write(): %s", ERR_error_string(SSL_get_error(sock->ssl, n), 0));
errno = EBADF;
}
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
return -1;
}
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
sock->sending += n;
if (sock->sending < dnslen + 2) {
errno = EINPROGRESS;
return -1;
}
sock->sending = 0;
sock->is_ready = 1;
return 1;
}
if (!sock->is_ssl_ready) {
switch (perf_os_waituntilanywritable(sock, 1, pipe_fd, timeout)) {
case PERF_R_TIMEDOUT:
return -1;
case PERF_R_SUCCESS: {
int error = 0;
socklen_t len = (socklen_t)sizeof(error);
getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void*)&error, &len);
if (error != 0) {
if (error == EINPROGRESS
#if EWOULDBLOCK != EAGAIN
|| error == EWOULDBLOCK
#endif
|| error == EAGAIN) {
return 0;
}
return -1;
}
}
}
sock->is_ssl_ready = 1;
}
if (pthread_mutex_lock(&sock->lock)) {
perf_log_fatal("pthread_mutex_lock() failed");
}
ret = SSL_connect(sock->ssl);
if (!ret) {
perf_log_warning("SSL_connect(): %s", ERR_error_string(SSL_get_error(sock->ssl, ret), 0));
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
return -1;
}
if (ret < 0) {
int err = SSL_get_error(sock->ssl, ret);
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
if (err == SSL_ERROR_WANT_READ || err == SSL_ERROR_WANT_WRITE) {
return 0;
}
perf_log_warning("SSL_connect(): %s", ERR_error_string(err, 0));
return -1;
}
sock->is_ready = 1;
if (pthread_mutex_unlock(&sock->lock)) {
perf_log_fatal("pthread_mutex_unlock() failed");
}
return 1;
}
case sock_tcp: case sock_tcp:
if (sock->sending) { return perf_net_tcp_opensocket(server, &tmp, bufsize);
uint16_t dnslen; case sock_dot:
ssize_t n; return perf_net_dot_opensocket(server, &tmp, bufsize);
memcpy(&dnslen, sock->sendbuf, 2);
dnslen = ntohs(dnslen);
n = sendto(sock->fd, sock->sendbuf + sock->sending, dnslen + 2 - sock->sending, sock->flags, (struct sockaddr*)&sock->dest_addr, sock->addrlen);
if (n < 1) {
return -1;
}
sock->sending += n;
if (sock->sending < dnslen + 2) {
errno = EINPROGRESS;
return -1;
}
sock->sending = 0;
sock->is_ready = 1;
return 1;
}
switch (perf_os_waituntilanywritable(sock, 1, pipe_fd, timeout)) {
case PERF_R_TIMEDOUT:
return -1;
case PERF_R_SUCCESS: {
int error = 0;
socklen_t len = (socklen_t)sizeof(error);
getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void*)&error, &len);
if (error != 0) {
if (error == EINPROGRESS
#if EWOULDBLOCK != EAGAIN
|| error == EWOULDBLOCK
#endif
|| error == EAGAIN) {
return 0;
}
return -1;
}
sock->is_ready = 1;
return 1;
}
}
break;
default: default:
break; perf_log_fatal("perf_net_opensocket(): invalid mode");
} }
return -1; return 0;
} }

130
src/net.h
View file

@ -25,6 +25,11 @@
#include <openssl/ssl.h> #include <openssl/ssl.h>
#include <pthread.h> #include <pthread.h>
#include <netinet/in.h> #include <netinet/in.h>
#include <assert.h>
#include <stdbool.h>
#define TCP_RECV_BUF_SIZE (16 * 1024)
#define TCP_SEND_BUF_SIZE (4 * 1024)
struct perf_sockaddr { struct perf_sockaddr {
union { union {
@ -42,43 +47,120 @@ enum perf_net_mode {
sock_pipe, sock_pipe,
sock_udp, sock_udp,
sock_tcp, sock_tcp,
sock_tls sock_dot
}; };
struct perf_net_socket;
typedef ssize_t (*perf_net_recv_t)(struct perf_net_socket* sock, void* buf, size_t len, int flags);
typedef ssize_t (*perf_net_sendto_t)(struct perf_net_socket* sock, uint16_t qid, const void* buf, size_t len, int flags, const struct sockaddr* dest_addr, socklen_t addrlen);
typedef int (*perf_net_close_t)(struct perf_net_socket* sock);
typedef int (*perf_net_sockeq_t)(struct perf_net_socket* sock, struct perf_net_socket* other);
/* sockready return:
* -1: An error occurred, see errno
* - EINPROGRESS: socket is still sending
* 0: Socket is not ready, may still be connecting or negotiating
* 1: Socket is ready and can be used for sending to
*/
typedef int (*perf_net_sockready_t)(struct perf_net_socket* sock, int pipe_fd, int64_t timeout);
/* Indicates if there are more data to be read in buffers of the transport */
typedef bool (*perf_net_have_more_t)(struct perf_net_socket* sock);
/* Callback for when a query has been sent if it was delayed due to partily sent or reconnection */
typedef void (*perf_net_sent_cb_t)(struct perf_net_socket* sock, uint16_t qid);
typedef enum perf_socket_event {
perf_socket_event_connect,
perf_socket_event_reconnect
} perf_socket_event_t;
/* Callback for socket events related to connection oriented protocols, for statistics */
typedef void (*perf_net_event_cb_t)(struct perf_net_socket* sock, perf_socket_event_t event, uint64_t elapsed_time);
struct perf_net_socket { struct perf_net_socket {
void* data; /* user data */
enum perf_net_mode mode; enum perf_net_mode mode;
int fd, have_more, is_ready, flags, is_ssl_ready; perf_net_recv_t recv;
char* recvbuf; perf_net_sendto_t sendto;
size_t at, sending; perf_net_close_t close;
char* sendbuf; perf_net_sockeq_t sockeq;
struct sockaddr_storage dest_addr; perf_net_sockready_t sockready;
socklen_t addrlen; perf_net_have_more_t have_more;
SSL* ssl;
pthread_mutex_t lock; /*
* Not set by protocol, set by caller.
* May be 0 if caller don't care.
* MUST NOT be called from sendto(), only called if query is delayed in some way.
*/
perf_net_sent_cb_t sent;
/* Used if caller want info on connection oriented events */
perf_net_event_cb_t event;
/*
* The system file descriptor that is used for transport, this is used
* in os functions to poll/wait for read/write.
*/
int fd;
}; };
static inline ssize_t perf_net_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags)
{
return sock->recv(sock, buf, len, flags);
}
static inline ssize_t perf_net_sendto(struct perf_net_socket* sock, uint16_t qid, const void* buf, size_t len, int flags, const struct sockaddr* dest_addr, socklen_t addrlen)
{
return sock->sendto(sock, qid, buf, len, flags, dest_addr, addrlen);
}
static inline int perf_net_close(struct perf_net_socket* sock)
{
return sock->close(sock);
}
static inline int perf_net_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b)
{
assert(sock_a);
assert(sock_b);
assert(sock_a->mode == sock_b->mode);
return sock_a->sockeq(sock_a, sock_b);
}
static inline int perf_net_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{
return sock->sockready(sock, pipe_fd, timeout);
}
static inline int perf_net_have_more(struct perf_net_socket* sock)
{
return sock->have_more ? sock->have_more(sock) : false;
}
enum perf_net_mode perf_net_parsemode(const char* mode);
int perf_net_parsefamily(const char* family);
void perf_net_parseserver(int family, const char* name, unsigned int port, perf_sockaddr_t* addr);
void perf_net_parselocal(int family, const char* name, unsigned int port, perf_sockaddr_t* addr);
void perf_sockaddr_fromin(perf_sockaddr_t* sockaddr, const struct in_addr* in, in_port_t port); void perf_sockaddr_fromin(perf_sockaddr_t* sockaddr, const struct in_addr* in, in_port_t port);
void perf_sockaddr_fromin6(perf_sockaddr_t* sockaddr, const struct in6_addr* in, in_port_t port); void perf_sockaddr_fromin6(perf_sockaddr_t* sockaddr, const struct in6_addr* in, in_port_t port);
in_port_t perf_sockaddr_port(const perf_sockaddr_t* sockaddr); in_port_t perf_sockaddr_port(const perf_sockaddr_t* sockaddr);
void perf_sockaddr_setport(perf_sockaddr_t* sockaddr, in_port_t port); void perf_sockaddr_setport(perf_sockaddr_t* sockaddr, in_port_t port);
void perf_sockaddr_format(const perf_sockaddr_t* sockaddr, char* buf, size_t len); void perf_sockaddr_format(const perf_sockaddr_t* sockaddr, char* buf, size_t len);
ssize_t perf_net_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags); static inline int perf_sockaddr_isinet6(const perf_sockaddr_t* sockaddr)
ssize_t perf_net_sendto(struct perf_net_socket* sock, const void* buf, size_t len, int flags, {
const struct sockaddr* dest_addr, socklen_t addrlen); return sockaddr->sa.sa.sa_family == AF_INET6;
}
int perf_net_close(struct perf_net_socket* sock); struct perf_net_socket* perf_net_opensocket(enum perf_net_mode mode, const perf_sockaddr_t* server, const perf_sockaddr_t* local, unsigned int offset, size_t bufsize);
int perf_net_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b);
int perf_net_parsefamily(const char* family); struct perf_net_socket* perf_net_udp_opensocket(const perf_sockaddr_t*, const perf_sockaddr_t*, size_t);
struct perf_net_socket* perf_net_tcp_opensocket(const perf_sockaddr_t*, const perf_sockaddr_t*, size_t);
void perf_net_parseserver(int family, const char* name, unsigned int port, perf_sockaddr_t* addr); struct perf_net_socket* perf_net_dot_opensocket(const perf_sockaddr_t*, const perf_sockaddr_t*, size_t);
void perf_net_parselocal(int family, const char* name, unsigned int port, perf_sockaddr_t* addr);
struct perf_net_socket perf_net_opensocket(enum perf_net_mode mode, const perf_sockaddr_t* server, const perf_sockaddr_t* local, unsigned int offset, int bufsize);
enum perf_net_mode perf_net_parsemode(const char* mode);
int perf_net_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout);
#endif #endif

468
src/net_dot.c Normal file
View file

@ -0,0 +1,468 @@
/*
* Copyright 2019-2021 OARC, Inc.
* Copyright 2017-2018 Akamai Technologies
* Copyright 2006-2016 Nominum, Inc.
* All rights reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "config.h"
#include "net.h"
#include "log.h"
#include "strerror.h"
#include "util.h"
#include "os.h"
#include <errno.h>
#include <fcntl.h>
#include <unistd.h>
#include <openssl/err.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <ck_pr.h>
static SSL_CTX* ssl_ctx = 0;
#define self ((struct perf__dot_socket*)sock)
struct perf__dot_socket {
struct perf_net_socket base;
pthread_mutex_t lock;
SSL* ssl;
char recvbuf[TCP_RECV_BUF_SIZE], sendbuf[TCP_SEND_BUF_SIZE];
size_t at, sending;
bool is_ready, is_conn_ready, have_more, is_sending, do_reconnect;
perf_sockaddr_t server, local;
size_t bufsize;
uint16_t qid;
uint64_t conn_ts;
perf_socket_event_t conn_event;
};
static void perf__dot_connect(struct perf_net_socket* sock)
{
int ret;
self->is_ready = true;
int fd = socket(self->server.sa.sa.sa_family, SOCK_STREAM, 0);
if (fd == -1) {
char __s[256];
perf_log_fatal("socket: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
ck_pr_store_int(&sock->fd, fd);
if (self->ssl) {
SSL_free(self->ssl);
}
if (!(self->ssl = SSL_new(ssl_ctx))) {
perf_log_fatal("SSL_new(): %s", ERR_error_string(ERR_get_error(), 0));
}
if (!(ret = SSL_set_fd(self->ssl, sock->fd))) {
perf_log_fatal("SSL_set_fd(): %s", ERR_error_string(SSL_get_error(self->ssl, ret), 0));
}
if (self->server.sa.sa.sa_family == AF_INET6) {
int on = 1;
if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) == -1) {
perf_log_warning("setsockopt(IPV6_V6ONLY) failed");
}
}
if (bind(sock->fd, &self->local.sa.sa, self->local.length) == -1) {
char __s[256];
perf_log_fatal("bind: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
if (self->bufsize > 0) {
ret = setsockopt(sock->fd, SOL_SOCKET, SO_RCVBUF,
&self->bufsize, sizeof(self->bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_RCVBUF) failed");
ret = setsockopt(sock->fd, SOL_SOCKET, SO_SNDBUF,
&self->bufsize, sizeof(self->bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_SNDBUF) failed");
}
int flags = fcntl(sock->fd, F_GETFL, 0);
if (flags < 0)
perf_log_fatal("fcntl(F_GETFL)");
ret = fcntl(sock->fd, F_SETFL, flags | O_NONBLOCK);
if (ret < 0)
perf_log_fatal("fcntl(F_SETFL)");
self->conn_ts = perf_get_time();
if (connect(sock->fd, &self->server.sa.sa, self->server.length)) {
if (errno == EINPROGRESS) {
self->is_ready = false;
} else {
char __s[256];
perf_log_fatal("connect() failed: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
}
}
static void perf__dot_reconnect(struct perf_net_socket* sock)
{
close(sock->fd);
self->have_more = false;
self->at = 0;
if (self->sending) {
self->sending = 0;
self->is_sending = false;
}
self->is_conn_ready = false;
perf__dot_connect(sock);
}
static ssize_t perf__dot_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags)
{
ssize_t n;
uint16_t dnslen, dnslen2;
if (!self->have_more) {
PERF_LOCK(&self->lock);
if (!self->is_ready) {
PERF_UNLOCK(&self->lock);
errno = EAGAIN;
return -1;
}
n = SSL_read(self->ssl, self->recvbuf + self->at, TCP_RECV_BUF_SIZE - self->at);
if (!n) {
perf__dot_reconnect(sock);
PERF_UNLOCK(&self->lock);
errno = EAGAIN;
return -1;
}
if (n < 0) {
int err = SSL_get_error(self->ssl, n);
switch (err) {
case SSL_ERROR_WANT_READ:
errno = EAGAIN;
break;
case SSL_ERROR_SYSCALL:
switch (errno) {
case ECONNREFUSED:
case ECONNRESET:
case ENOTCONN:
perf__dot_reconnect(sock);
errno = EAGAIN;
break;
default:
break;
}
break;
default:
errno = EBADF;
break;
}
PERF_UNLOCK(&self->lock);
return -1;
}
PERF_UNLOCK(&self->lock);
self->at += n;
if (self->at < 3) {
errno = EAGAIN;
return -1;
}
}
memcpy(&dnslen, self->recvbuf, 2);
dnslen = ntohs(dnslen);
if (self->at < dnslen + 2) {
errno = EAGAIN;
return -1;
}
memcpy(buf, self->recvbuf + 2, len < dnslen ? len : dnslen);
memmove(self->recvbuf, self->recvbuf + 2 + dnslen, self->at - 2 - dnslen);
self->at -= 2 + dnslen;
if (self->at > 2) {
memcpy(&dnslen2, self->recvbuf, 2);
dnslen2 = ntohs(dnslen2);
if (self->at >= dnslen2 + 2) {
self->have_more = true;
return dnslen;
}
}
self->have_more = false;
return dnslen;
}
static ssize_t perf__dot_sendto(struct perf_net_socket* sock, uint16_t qid, const void* buf, size_t len, int flags, const struct sockaddr* dest_addr, socklen_t addrlen)
{
size_t send = len < TCP_SEND_BUF_SIZE - 2 ? len : (TCP_SEND_BUF_SIZE - 2);
// TODO: We only send what we can send, because we can't continue sending
uint16_t dnslen = htons(send);
ssize_t n;
PERF_LOCK(&self->lock);
if (!self->is_ready) {
PERF_UNLOCK(&self->lock);
errno = EAGAIN;
return -1;
}
memcpy(self->sendbuf, &dnslen, 2);
memcpy(self->sendbuf + 2, buf, send);
self->qid = qid;
n = SSL_write(self->ssl, self->sendbuf, send + 2);
if (n < 1) {
switch (SSL_get_error(self->ssl, n)) {
case SSL_ERROR_SYSCALL:
switch (errno) {
case ECONNREFUSED:
case ECONNRESET:
case ENOTCONN:
case EPIPE:
perf__dot_reconnect(sock);
self->is_sending = true;
self->sending = 0;
PERF_UNLOCK(&self->lock);
errno = EINPROGRESS;
return -1;
default:
break;
}
PERF_UNLOCK(&self->lock);
return -1;
case SSL_ERROR_WANT_READ:
case SSL_ERROR_WANT_WRITE:
self->is_sending = true;
self->sending = 0;
PERF_UNLOCK(&self->lock);
errno = EINPROGRESS;
return -1;
default:
break;
}
perf_log_warning("SSL_write(): %s", ERR_error_string(SSL_get_error(self->ssl, n), 0));
errno = EBADF;
return -1;
}
if (n < send + 2) {
self->sending = n;
self->is_sending = true;
PERF_UNLOCK(&self->lock);
errno = EINPROGRESS;
return -1;
}
PERF_UNLOCK(&self->lock);
return n - 2;
}
static int perf__dot_close(struct perf_net_socket* sock)
{
// TODO
return close(sock->fd);
}
static int perf__dot_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b)
{
return sock_a->fd == sock_b->fd;
}
static int perf__dot_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{
PERF_LOCK(&self->lock);
if (self->do_reconnect) {
perf__dot_reconnect(sock);
self->do_reconnect = false;
}
if (self->is_ready) {
if (self->is_sending) {
uint16_t dnslen;
ssize_t n;
memcpy(&dnslen, self->sendbuf, 2);
dnslen = ntohs(dnslen);
n = SSL_write(self->ssl, self->sendbuf + self->sending, dnslen + 2 - self->sending);
if (n < 1) {
switch (SSL_get_error(self->ssl, n)) {
case SSL_ERROR_SYSCALL:
switch (errno) {
case ECONNREFUSED:
case ECONNRESET:
case ENOTCONN:
case EPIPE:
perf__dot_reconnect(sock);
PERF_UNLOCK(&self->lock);
errno = EINPROGRESS;
return -1;
default:
break;
}
PERF_UNLOCK(&self->lock);
return -1;
case SSL_ERROR_WANT_READ:
case SSL_ERROR_WANT_WRITE:
PERF_UNLOCK(&self->lock);
errno = EINPROGRESS;
return -1;
default:
break;
}
perf_log_warning("SSL_write(): %s", ERR_error_string(SSL_get_error(self->ssl, n), 0));
errno = EBADF;
PERF_UNLOCK(&self->lock);
return -1;
}
PERF_UNLOCK(&self->lock);
self->sending += n;
if (self->sending < dnslen + 2) {
errno = EINPROGRESS;
return -1;
}
self->sending = 0;
self->is_sending = false;
if (sock->sent) {
sock->sent(sock, self->qid);
}
return 1;
}
PERF_UNLOCK(&self->lock);
return 1;
}
if (!self->is_conn_ready) {
switch (perf_os_waituntilanywritable(&sock, 1, pipe_fd, timeout)) {
case PERF_R_TIMEDOUT:
PERF_UNLOCK(&self->lock);
return -1;
case PERF_R_SUCCESS: {
int error = 0;
socklen_t len = (socklen_t)sizeof(error);
getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void*)&error, &len);
if (error != 0) {
if (error == EINPROGRESS
#if EWOULDBLOCK != EAGAIN
|| error == EWOULDBLOCK
#endif
|| error == EAGAIN) {
PERF_UNLOCK(&self->lock);
return 0;
}
PERF_UNLOCK(&self->lock);
return -1;
}
break;
}
default:
PERF_UNLOCK(&self->lock);
return -1;
}
self->is_conn_ready = true;
}
int ret = SSL_connect(self->ssl);
if (!ret) {
perf_log_warning("SSL_connect(): %s", ERR_error_string(SSL_get_error(self->ssl, ret), 0));
PERF_UNLOCK(&self->lock);
return -1;
}
if (ret < 0) {
int err = SSL_get_error(self->ssl, ret);
if (err == SSL_ERROR_WANT_READ || err == SSL_ERROR_WANT_WRITE) {
PERF_UNLOCK(&self->lock);
return 0;
}
self->do_reconnect = true;
PERF_UNLOCK(&self->lock);
perf_log_warning("SSL_connect(): %s", ERR_error_string(err, 0));
return -1;
}
self->is_ready = true;
PERF_UNLOCK(&self->lock);
if (sock->event) {
sock->event(sock, self->conn_event, perf_get_time() - self->conn_ts);
self->conn_event = perf_socket_event_reconnect;
}
if (self->is_sending) {
errno = EINPROGRESS;
return -1;
}
return 1;
}
static bool perf__dot_have_more(struct perf_net_socket* sock)
{
return self->have_more;
}
struct perf_net_socket* perf_net_dot_opensocket(const perf_sockaddr_t* server, const perf_sockaddr_t* local, size_t bufsize)
{
struct perf__dot_socket* tmp = calloc(1, sizeof(struct perf__dot_socket)); // clang scan-build
struct perf_net_socket* sock = (struct perf_net_socket*)tmp;
if (!sock) {
perf_log_fatal("perf_net_dot_opensocket() out of memory");
return 0; // needed for clang scan build
}
sock->recv = perf__dot_recv;
sock->sendto = perf__dot_sendto;
sock->close = perf__dot_close;
sock->sockeq = perf__dot_sockeq;
sock->sockready = perf__dot_sockready;
sock->have_more = perf__dot_have_more;
self->server = *server;
self->local = *local;
self->bufsize = bufsize;
if (self->bufsize > 0) {
self->bufsize *= 1024;
}
self->conn_event = perf_socket_event_connect;
PERF_MUTEX_INIT(&self->lock);
if (!ssl_ctx) {
#ifdef HAVE_TLS_METHOD
if (!(ssl_ctx = SSL_CTX_new(TLS_method()))) {
perf_log_fatal("SSL_CTX_new(): %s", ERR_error_string(ERR_get_error(), 0));
}
if (!SSL_CTX_set_min_proto_version(ssl_ctx, TLS1_2_VERSION)) {
perf_log_fatal("SSL_CTX_set_min_proto_version(TLS1_2_VERSION): %s", ERR_error_string(ERR_get_error(), 0));
}
#else
if (!(ssl_ctx = SSL_CTX_new(SSLv23_client_method()))) {
perf_log_fatal("SSL_CTX_new(): %s", ERR_error_string(ERR_get_error(), 0));
}
#endif
SSL_CTX_set_mode(ssl_ctx, SSL_MODE_ENABLE_PARTIAL_WRITE);
}
perf__dot_connect(sock);
return sock;
}

348
src/net_tcp.c Normal file
View file

@ -0,0 +1,348 @@
/*
* Copyright 2019-2021 OARC, Inc.
* Copyright 2017-2018 Akamai Technologies
* Copyright 2006-2016 Nominum, Inc.
* All rights reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "config.h"
#include "net.h"
#include "log.h"
#include "strerror.h"
#include "os.h"
#include "util.h"
#include <errno.h>
#include <fcntl.h>
#include <unistd.h>
#include <string.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <ck_pr.h>
#define self ((struct perf__tcp_socket*)sock)
struct perf__tcp_socket {
struct perf_net_socket base;
char recvbuf[TCP_RECV_BUF_SIZE], sendbuf[TCP_SEND_BUF_SIZE];
size_t at, sending;
bool is_ready, need_reconnect, have_more, is_sending;
int flags;
struct sockaddr_storage dest_addr;
socklen_t addrlen;
perf_sockaddr_t server, local;
size_t bufsize;
int recvfd;
uint16_t qid;
uint64_t conn_ts;
perf_socket_event_t conn_event;
};
static int perf__tcp_connect(struct perf_net_socket* sock)
{
int fd;
self->is_ready = true;
fd = socket(self->server.sa.sa.sa_family, SOCK_STREAM, 0);
if (fd == -1) {
char __s[256];
perf_log_fatal("socket: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
if (self->server.sa.sa.sa_family == AF_INET6) {
int on = 1;
if (setsockopt(fd, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) == -1) {
perf_log_warning("setsockopt(IPV6_V6ONLY) failed");
}
}
if (bind(fd, &self->local.sa.sa, self->local.length) == -1) {
char __s[256];
perf_log_fatal("bind: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
if (self->bufsize) {
int ret = setsockopt(fd, SOL_SOCKET, SO_RCVBUF,
&self->bufsize, sizeof(self->bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_RCVBUF) failed");
ret = setsockopt(fd, SOL_SOCKET, SO_SNDBUF,
&self->bufsize, sizeof(self->bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_SNDBUF) failed");
}
int flags = fcntl(fd, F_GETFL, 0);
if (flags < 0)
perf_log_fatal("fcntl(F_GETFL)");
int ret = fcntl(fd, F_SETFL, flags | O_NONBLOCK);
if (ret < 0)
perf_log_fatal("fcntl(F_SETFL)");
self->conn_ts = perf_get_time();
if (connect(fd, &self->server.sa.sa, self->server.length)) {
if (errno == EINPROGRESS) {
self->is_ready = false;
} else {
char __s[256];
perf_log_fatal("connect() failed: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
}
return fd;
}
static ssize_t perf__tcp_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags)
{
ssize_t n;
uint16_t dnslen, dnslen2;
int fd = ck_pr_load_int(&sock->fd);
if (fd != self->recvfd) {
/* reconnecting happened, reset buffers */
self->have_more = false;
self->at = 0;
self->recvfd = fd;
}
if (!self->have_more) {
n = recv(fd, self->recvbuf + self->at, TCP_RECV_BUF_SIZE - self->at, flags);
if (!n) {
return 0;
} else if (n < 0) {
switch (errno) {
case ECONNREFUSED:
case ECONNRESET:
case ENOTCONN:
errno = EAGAIN;
break;
default:
break;
}
return n;
}
self->at += n;
if (self->at < 3) {
errno = EAGAIN;
return -1;
}
}
memcpy(&dnslen, self->recvbuf, 2);
dnslen = ntohs(dnslen);
if (self->at < dnslen + 2) {
errno = EAGAIN;
return -1;
}
memcpy(buf, self->recvbuf + 2, len < dnslen ? len : dnslen);
memmove(self->recvbuf, self->recvbuf + 2 + dnslen, self->at - 2 - dnslen);
self->at -= 2 + dnslen;
if (self->at > 2) {
memcpy(&dnslen2, self->recvbuf, 2);
dnslen2 = ntohs(dnslen2);
if (self->at >= dnslen2 + 2) {
self->have_more = true;
return dnslen;
}
}
self->have_more = false;
return dnslen;
}
static ssize_t perf__tcp_sendto(struct perf_net_socket* sock, uint16_t qid, const void* buf, size_t len, int flags, const struct sockaddr* dest_addr, socklen_t addrlen)
{
size_t send = len < TCP_SEND_BUF_SIZE - 2 ? len : (TCP_SEND_BUF_SIZE - 2);
// TODO: We only send what we can send, because we can't continue sending
uint16_t dnslen = htons(send);
ssize_t n;
memcpy(self->sendbuf, &dnslen, 2);
memcpy(self->sendbuf + 2, buf, send);
self->qid = qid;
n = sendto(sock->fd, self->sendbuf, send + 2, flags, dest_addr, addrlen);
if (n < 0) {
switch (errno) {
case ECONNREFUSED:
case ECONNRESET:
case ENOTCONN:
case EPIPE:
self->need_reconnect = true;
self->is_sending = true;
self->sending = 0;
errno = EINPROGRESS;
return -1;
default:
break;
}
return -1;
}
if (n < send + 2) {
self->is_sending = true;
self->sending = n;
self->flags = flags;
memcpy(&self->dest_addr, dest_addr, addrlen);
self->addrlen = addrlen;
errno = EINPROGRESS;
return -1;
}
return n - 2;
}
static int perf__tcp_close(struct perf_net_socket* sock)
{
return close(sock->fd);
}
static int perf__tcp_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b)
{
return sock_a->fd == sock_b->fd;
}
static int perf__tcp_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{
if (self->need_reconnect) {
int fd = perf__tcp_connect(sock), oldfd = ck_pr_load_int(&sock->fd);
ck_pr_store_int(&sock->fd, fd);
close(oldfd);
self->need_reconnect = false;
}
if (self->is_ready) {
if (self->is_sending) {
uint16_t dnslen;
ssize_t n;
memcpy(&dnslen, self->sendbuf, 2);
dnslen = ntohs(dnslen);
n = sendto(sock->fd, self->sendbuf + self->sending, dnslen + 2 - self->sending, self->flags, (struct sockaddr*)&self->dest_addr, self->addrlen);
if (n < 1) {
switch (errno) {
case ECONNREFUSED:
case ECONNRESET:
case ENOTCONN:
case EPIPE:
self->need_reconnect = true;
if (self->sending) {
self->sending = 0;
self->is_sending = false;
}
errno = EINPROGRESS;
return -1;
default:
break;
}
return -1;
}
self->sending += n;
if (self->sending < dnslen + 2) {
errno = EINPROGRESS;
return -1;
}
self->sending = 0;
self->is_sending = false;
if (sock->sent) {
sock->sent(sock, self->qid);
}
}
return 1;
}
switch (perf_os_waituntilanywritable(&sock, 1, pipe_fd, timeout)) {
case PERF_R_TIMEDOUT:
return -1;
case PERF_R_SUCCESS: {
int error = 0;
socklen_t len = (socklen_t)sizeof(error);
getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, (void*)&error, &len);
if (error != 0) {
if (error == EINPROGRESS
#if EWOULDBLOCK != EAGAIN
|| error == EWOULDBLOCK
#endif
|| error == EAGAIN) {
return 0;
}
return -1;
}
self->is_ready = true;
if (sock->event) {
sock->event(sock, self->conn_event, perf_get_time() - self->conn_ts);
self->conn_event = perf_socket_event_reconnect;
}
if (self->is_sending) {
errno = EINPROGRESS;
return -1;
}
return 1;
}
default:
break;
}
return -1;
}
static bool perf__tcp_have_more(struct perf_net_socket* sock)
{
return self->have_more;
}
struct perf_net_socket* perf_net_tcp_opensocket(const perf_sockaddr_t* server, const perf_sockaddr_t* local, size_t bufsize)
{
struct perf__tcp_socket* tmp = calloc(1, sizeof(struct perf__tcp_socket)); // clang scan-build
struct perf_net_socket* sock = (struct perf_net_socket*)tmp;
if (!sock) {
perf_log_fatal("perf_net_tcp_opensocket() out of memory");
return 0; // needed for clang scan build
}
sock->recv = perf__tcp_recv;
sock->sendto = perf__tcp_sendto;
sock->close = perf__tcp_close;
sock->sockeq = perf__tcp_sockeq;
sock->sockready = perf__tcp_sockready;
sock->have_more = perf__tcp_have_more;
self->server = *server;
self->local = *local;
self->bufsize = bufsize;
if (self->bufsize > 0) {
self->bufsize *= 1024;
}
self->conn_event = perf_socket_event_connect;
sock->fd = perf__tcp_connect(sock);
self->recvfd = sock->fd;
return sock;
}

123
src/net_udp.c Normal file
View file

@ -0,0 +1,123 @@
/*
* Copyright 2019-2021 OARC, Inc.
* Copyright 2017-2018 Akamai Technologies
* Copyright 2006-2016 Nominum, Inc.
* All rights reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "config.h"
#include "net.h"
#include "log.h"
#include "strerror.h"
#include <errno.h>
#include <fcntl.h>
#include <unistd.h>
#include <sys/socket.h>
#include <netinet/in.h>
#define self ((struct perf__udp_socket*)sock)
struct perf__udp_socket {
struct perf_net_socket base;
};
static ssize_t perf__udp_recv(struct perf_net_socket* sock, void* buf, size_t len, int flags)
{
return recv(sock->fd, buf, len, flags);
}
static ssize_t perf__udp_sendto(struct perf_net_socket* sock, uint16_t qid, const void* buf, size_t len, int flags, const struct sockaddr* dest_addr, socklen_t addrlen)
{
return sendto(sock->fd, buf, len, flags, dest_addr, addrlen);
}
static int perf__udp_close(struct perf_net_socket* sock)
{
return close(sock->fd);
}
static int perf__udp_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* sock_b)
{
return sock_a->fd == sock_b->fd;
}
static int perf__udp_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{
return 1;
}
struct perf_net_socket* perf_net_udp_opensocket(const perf_sockaddr_t* server, const perf_sockaddr_t* local, size_t bufsize)
{
struct perf__udp_socket* tmp = calloc(1, sizeof(struct perf__udp_socket)); // clang scan-build
struct perf_net_socket* sock = (struct perf_net_socket*)tmp;
int ret, flags;
if (!sock) {
perf_log_fatal("perf_net_udp_opensocket() out of memory");
return 0; // needed for clang scan build
}
sock->recv = perf__udp_recv;
sock->sendto = perf__udp_sendto;
sock->close = perf__udp_close;
sock->sockeq = perf__udp_sockeq;
sock->sockready = perf__udp_sockready;
sock->fd = socket(server->sa.sa.sa_family, SOCK_DGRAM, 0);
if (sock->fd == -1) {
char __s[256];
perf_log_fatal("socket: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
if (server->sa.sa.sa_family == AF_INET6) {
int on = 1;
if (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_V6ONLY, &on, sizeof(on)) == -1) {
perf_log_warning("setsockopt(IPV6_V6ONLY) failed");
}
}
if (bind(sock->fd, &local->sa.sa, local->length) == -1) {
char __s[256];
perf_log_fatal("bind: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
if (bufsize > 0) {
bufsize *= 1024;
ret = setsockopt(sock->fd, SOL_SOCKET, SO_RCVBUF,
&bufsize, sizeof(bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_RCVBUF) failed");
ret = setsockopt(sock->fd, SOL_SOCKET, SO_SNDBUF,
&bufsize, sizeof(bufsize));
if (ret < 0)
perf_log_warning("setsockbuf(SO_SNDBUF) failed");
}
flags = fcntl(sock->fd, F_GETFL, 0);
if (flags < 0)
perf_log_fatal("fcntl(F_GETFL)");
ret = fcntl(sock->fd, F_SETFL, flags | O_NONBLOCK);
if (ret < 0)
perf_log_fatal("fcntl(F_SETFL)");
return sock;
}

4
src/opt.c
View file

@ -214,6 +214,10 @@ void perf_opt_parse(int argc, char** argv)
*opt->u.uintp = parse_uint(opt->desc, optarg, *opt->u.uintp = parse_uint(opt->desc, optarg,
1, 0xFFFFFFFF); 1, 0xFFFFFFFF);
break; break;
case perf_opt_zpint:
*opt->u.uintp = parse_uint(opt->desc, optarg,
0, 0xFFFFFFFF);
break;
case perf_opt_timeval: case perf_opt_timeval:
*opt->u.uint64p = parse_timeval(opt->desc, optarg); *opt->u.uint64p = parse_timeval(opt->desc, optarg);
break; break;

3
src/opt.h
View file

@ -23,7 +23,8 @@
typedef enum { typedef enum {
perf_opt_string, perf_opt_string,
perf_opt_boolean, perf_opt_boolean,
perf_opt_uint, perf_opt_uint, // can not be zero
perf_opt_zpint, // zero or positive
perf_opt_timeval, perf_opt_timeval,
perf_opt_double, perf_opt_double,
perf_opt_port, perf_opt_port,

13
src/os.c
View file

@ -61,11 +61,12 @@ void perf_os_handlesignal(int sig, void (*handler)(int))
perf_result_t perf_result_t
perf_os_waituntilreadable(struct perf_net_socket* sock, int pipe_fd, int64_t timeout) perf_os_waituntilreadable(struct perf_net_socket* sock, int pipe_fd, int64_t timeout)
{ {
return perf_os_waituntilanyreadable(sock, 1, pipe_fd, timeout); struct perf_net_socket* socks[] = { sock };
return perf_os_waituntilanyreadable(socks, 1, pipe_fd, timeout);
} }
perf_result_t perf_result_t
perf_os_waituntilanyreadable(struct perf_net_socket* socks, unsigned int nfds, int pipe_fd, perf_os_waituntilanyreadable(struct perf_net_socket** socks, unsigned int nfds, int pipe_fd,
int64_t timeout) int64_t timeout)
{ {
struct pollfd fds[nfds + 1]; struct pollfd fds[nfds + 1];
@ -73,10 +74,10 @@ perf_os_waituntilanyreadable(struct perf_net_socket* socks, unsigned int nfds, i
int to, n; int to, n;
for (i = 0; i < nfds; i++) { for (i = 0; i < nfds; i++) {
if (socks[i].have_more) if (perf_net_have_more(socks[i]))
return (PERF_R_SUCCESS); return (PERF_R_SUCCESS);
fds[i].fd = socks[i].fd; fds[i].fd = socks[i]->fd;
fds[i].events = POLLIN; fds[i].events = POLLIN;
} }
@ -109,7 +110,7 @@ perf_os_waituntilanyreadable(struct perf_net_socket* socks, unsigned int nfds, i
} }
perf_result_t perf_result_t
perf_os_waituntilanywritable(struct perf_net_socket* socks, unsigned int nfds, int pipe_fd, perf_os_waituntilanywritable(struct perf_net_socket** socks, unsigned int nfds, int pipe_fd,
int64_t timeout) int64_t timeout)
{ {
struct pollfd fds[nfds + 1]; struct pollfd fds[nfds + 1];
@ -117,7 +118,7 @@ perf_os_waituntilanywritable(struct perf_net_socket* socks, unsigned int nfds, i
int to, n; int to, n;
for (i = 0; i < nfds; i++) { for (i = 0; i < nfds; i++) {
fds[i].fd = socks[i].fd; fds[i].fd = socks[i]->fd;
fds[i].events = POLLOUT; fds[i].events = POLLOUT;
} }

4
src/os.h
View file

@ -34,11 +34,11 @@ perf_result_t
perf_os_waituntilreadable(struct perf_net_socket* sock, int pipe_fd, int64_t timeout); perf_os_waituntilreadable(struct perf_net_socket* sock, int pipe_fd, int64_t timeout);
perf_result_t perf_result_t
perf_os_waituntilanyreadable(struct perf_net_socket* socks, unsigned int nfds, int pipe_fd, perf_os_waituntilanyreadable(struct perf_net_socket** socks, unsigned int nfds, int pipe_fd,
int64_t timeout); int64_t timeout);
perf_result_t perf_result_t
perf_os_waituntilanywritable(struct perf_net_socket* socks, unsigned int nfds, int pipe_fd, perf_os_waituntilanywritable(struct perf_net_socket** socks, unsigned int nfds, int pipe_fd,
int64_t timeout); int64_t timeout);
#endif #endif

661
src/resperf.1.in
View file

@ -20,85 +20,96 @@ resperf \- test the resolution performance of a caching DNS server
.SH SYNOPSIS .SH SYNOPSIS
.hy 0 .hy 0
.ad l .ad l
\fBresperf\-report\fR\ [\fB\-a\ \fIlocal_addr\fB\fR] \fBresperf\-report\fR\ [\fB\-a\ \fIlocal_addr\fR]
[\fB\-d\ \fIdatafile\fB\fR] [\fB\-d\ \fIdatafile\fR]
[\fB\-M\ \fImode\fB\fR] [\fB\-R\fR]
[\fB\-s\ \fIserver_addr\fB\fR] [\fB\-M\ \fImode\fR]
[\fB\-p\ \fIport\fB\fR] [\fB\-s\ \fIserver_addr\fR]
[\fB\-x\ \fIlocal_port\fB\fR] [\fB\-p\ \fIport\fR]
[\fB\-t\ \fItimeout\fB\fR] [\fB\-x\ \fIlocal_port\fR]
[\fB\-b\ \fIbufsize\fB\fR] [\fB\-t\ \fItimeout\fR]
[\fB\-f\ \fIfamily\fB\fR] [\fB\-b\ \fIbufsize\fR]
[\fB\-f\ \fIfamily\fR]
[\fB\-e\fR] [\fB\-e\fR]
[\fB\-D\fR] [\fB\-D\fR]
[\fB\-y\ \fI[alg:]name:secret\fB\fR] [\fB\-y\ \fI[alg:]name:secret\fR]
[\fB\-h\fR] [\fB\-h\fR]
[\fB\-i\ \fIinterval\fB\fR] [\fB\-i\ \fIinterval\fR]
[\fB\-m\ \fImax_qps\fB\fR] [\fB\-m\ \fImax_qps\fR]
[\fB\-r\ \fIrampup_time\fB\fR] [\fB\-r\ \fIrampup_time\fR]
[\fB\-c\ \fIconstant_traffic_time\fB\fR] [\fB\-c\ \fIconstant_traffic_time\fR]
[\fB\-L\ \fImax_loss\fB\fR] [\fB\-L\ \fImax_loss\fR]
[\fB\-C\ \fIclients\fB\fR] [\fB\-C\ \fIclients\fR]
[\fB\-q\ \fImax_outstanding\fB\fR] [\fB\-q\ \fImax_outstanding\fR]
[\fB\-F\ \fIfall_behind\fR]
[\fB\-v\fR] [\fB\-v\fR]
[\fB\-W\fR]
.ad .ad
.hy .hy
.hy 0 .hy 0
.ad l .ad l
\fBresperf\fR\ [\fB\-a\ \fIlocal_addr\fB\fR] \fBresperf\fR\ [\fB\-a\ \fIlocal_addr\fR]
[\fB\-d\ \fIdatafile\fB\fR] [\fB\-d\ \fIdatafile\fR]
[\fB\-M\ \fImode\fB\fR] [\fB\-R\fR]
[\fB\-s\ \fIserver_addr\fB\fR] [\fB\-M\ \fImode\fR]
[\fB\-p\ \fIport\fB\fR] [\fB\-s\ \fIserver_addr\fR]
[\fB\-x\ \fIlocal_port\fB\fR] [\fB\-p\ \fIport\fR]
[\fB\-t\ \fItimeout\fB\fR] [\fB\-x\ \fIlocal_port\fR]
[\fB\-b\ \fIbufsize\fB\fR] [\fB\-t\ \fItimeout\fR]
[\fB\-f\ \fIfamily\fB\fR] [\fB\-b\ \fIbufsize\fR]
[\fB\-f\ \fIfamily\fR]
[\fB\-e\fR] [\fB\-e\fR]
[\fB\-D\fR] [\fB\-D\fR]
[\fB\-y\ \fI[alg:]name:secret\fB\fR] [\fB\-y\ \fI[alg:]name:secret\fR]
[\fB\-h\fR] [\fB\-h\fR]
[\fB\-i\ \fIinterval\fB\fR] [\fB\-i\ \fIinterval\fR]
[\fB\-m\ \fImax_qps\fB\fR] [\fB\-m\ \fImax_qps\fR]
[\fB\-P\ \fIplot_data_file\fB\fR] [\fB\-P\ \fIplot_data_file\fR]
[\fB\-r\ \fIrampup_time\fB\fR] [\fB\-r\ \fIrampup_time\fR]
[\fB\-c\ \fIconstant_traffic_time\fB\fR] [\fB\-c\ \fIconstant_traffic_time\fR]
[\fB\-L\ \fImax_loss\fB\fR] [\fB\-L\ \fImax_loss\fR]
[\fB\-C\ \fIclients\fB\fR] [\fB\-C\ \fIclients\fR]
[\fB\-q\ \fImax_outstanding\fB\fR] [\fB\-q\ \fImax_outstanding\fR]
[\fB\-F\ \fIfall_behind\fR]
[\fB\-v\fR] [\fB\-v\fR]
[\fB\-W\fR]
.ad .ad
.hy .hy
.SH DESCRIPTION .SH DESCRIPTION
\fBresperf\fR is a companion tool to \fBdnsperf\fR. \fBdnsperf\fR was \fBresperf\fR is a companion tool to \fBdnsperf\fR.
primarily designed for benchmarking authoritative servers, and it does not \fBdnsperf\fR was primarily designed for benchmarking authoritative
work well with caching servers that are talking to the live Internet. One servers, and it does not work well with caching servers that are talking
reason for this is that dnsperf uses a "self-pacing" approach, which is to the live Internet.
One reason for this is that dnsperf uses a "self-pacing" approach, which is
based on the assumption that you can keep the server 100% busy simply by based on the assumption that you can keep the server 100% busy simply by
sending it a small burst of back-to-back queries to fill up network buffers, sending it a small burst of back-to-back queries to fill up network buffers,
and then send a new query whenever you get a response back. This approach and then send a new query whenever you get a response back.
works well for authoritative servers that process queries in order and one This approach works well for authoritative servers that process queries in
at a time; it also works pretty well for a caching server in a closed order and one at a time; it also works pretty well for a caching server in
laboratory environment talking to a simulated Internet that's all on the a closed laboratory environment talking to a simulated Internet that's all
same LAN. Unfortunately, it does not work well with a caching server talking on the same LAN.
Unfortunately, it does not work well with a caching server talking
to the actual Internet, which may need to work on thousands of queries in to the actual Internet, which may need to work on thousands of queries in
parallel to achieve its maximum throughput. There have been numerous parallel to achieve its maximum throughput.
attempts to use dnsperf (or its predecessor, queryperf) for benchmarking There have been numerous attempts to use dnsperf (or its predecessor,
live caching servers, usually with poor results. Therefore, a separate tool queryperf) for benchmarking live caching servers, usually with poor results.
designed specifically for caching servers is needed. Therefore, a separate tool designed specifically for caching servers is
needed.
.SS "How resperf works" .SS "How resperf works"
Unlike the "self-pacing" approach of dnsperf, resperf works by sending DNS Unlike the "self-pacing" approach of dnsperf, \fBresperf\fR works by sending
queries at a controlled, steadily increasing rate. By default, resperf will DNS queries at a controlled, steadily increasing rate.
send traffic for 60 seconds, linearly increasing the amount of traffic from By default, \fBresperf\fR will send traffic for 60 seconds, linearly
zero to 100,000 queries per second. increasing the amount of traffic from zero to 100,000 queries per second (or
\fImax_qps\fR).
During the test, resperf listens for responses from the server and keeps During the test, \fBresperf\fR listens for responses from the server and
track of response rates, failure rates, and latencies. It will also continue keeps track of response rates, failure rates, and latencies.
listening for responses for an additional 40 seconds after it has stopped It will also continue listening for responses for an additional 40 seconds
sending traffic, so that there is time for the server to respond to the last after it has stopped sending traffic, so that there is time for the server
queries sent. This time period was chosen to be longer than the overall to respond to the last queries sent.
query timeout of both Nominum CacheServe and current versions of BIND. This time period was chosen to be longer than the overall query timeout of
both Nominum CacheServe and current versions of BIND.
If the test is successful, the query rate will at some point exceed the If the test is successful, the query rate will at some point exceed the
capacity of the server and queries will be dropped, causing the response capacity of the server and queries will be dropped, causing the response
@ -107,77 +118,87 @@ rate to stop growing or even decrease as the query rate increases.
The result of the test is a set of measurements of the query rate, response The result of the test is a set of measurements of the query rate, response
rate, failure response rate, and average query latency as functions of time. rate, failure response rate, and average query latency as functions of time.
.SS "What you will need" .SS "What you will need"
Benchmarking a live caching server is serious business. A fast caching Benchmarking a live caching server is serious business.
server like Nominum CacheServe, resolving a mix of cacheable and A fast caching server like Nominum CacheServe, resolving a mix of cacheable
non-cacheable queries typical of ISP customer traffic, is capable of and non-cacheable queries typical of ISP customer traffic, is capable of
resolving well over 1,000,000 queries per second. In the process, it will resolving well over 1,000,000 queries per second.
send more than 40,000 queries per second to authoritative servers on the In the process, it will send more than 40,000 queries per second to
Internet, and receive responses to most of them. Assuming an average request authoritative servers on the Internet, and receive responses to most of them.
size of 50 bytes and a response size of 150 bytes, this amounts to some 1216 Assuming an average request size of 50 bytes and a response size of 150
Mbps of outgoing and 448 Mbps of incoming traffic. If your Internet bytes, this amounts to some 1216 Mbps of outgoing and 448 Mbps of incoming
connection can't handle the bandwidth, you will end up measuring the speed traffic.
of the connection, not the server, and may saturate the connection causing a If your Internet connection can't handle the bandwidth, you will end up
degradation in service for other users. measuring the speed of the connection, not the server, and may saturate the
connection causing a degradation in service for other users.
Make sure there is no stateful firewall between the server and the Internet, Make sure there is no stateful firewall between the server and the Internet,
because most of them can't handle the amount of UDP traffic the test will because most of them can't handle the amount of UDP traffic the test will
generate and will end up dropping packets, skewing the test results. Some generate and will end up dropping packets, skewing the test results.
will even lock up or crash. Some will even lock up or crash.
You should run resperf on a machine separate from the server under test, on You should run \fBresperf\fR on a machine separate from the server under test,
the same LAN. Preferably, this should be a Gigabit Ethernet network. The on the same LAN.
machine running resperf should be at least as fast as the machine being Preferably, this should be a Gigabit Ethernet network.
tested; otherwise, it may end up being the bottleneck. The machine running \fBresperf\fR should be at least as fast as the machine
being tested; otherwise, it may end up being the bottleneck.
There should be no other applications running on the machine running There should be no other applications running on the machine running
resperf. Performance testing at the traffic levels involved is essentially a \fBresperf\fR.
Performance testing at the traffic levels involved is essentially a
hard real-time application - consider the fact that at a query rate of hard real-time application - consider the fact that at a query rate of
100,000 queries per second, if resperf gets delayed by just 1/100 of a 100,000 queries per second, if \fBresperf\fR gets delayed by just 1/100 of a
second, 1000 incoming UDP packets will arrive in the meantime. This is more second, 1000 incoming UDP packets will arrive in the meantime.
than most operating systems will buffer, which means packets will be This is more than most operating systems will buffer, which means packets
dropped. will be dropped.
Because the granularity of the timers provided by operating systems is Because the granularity of the timers provided by operating systems is
typically too coarse to accurately schedule packet transmissions at typically too coarse to accurately schedule packet transmissions at
sub-millisecond intervals, resperf will busy-wait between packet sub-millisecond intervals, \fBresperf\fR will busy-wait between packet
transmissions, constantly polling for responses in the meantime. Therefore, transmissions, constantly polling for responses in the meantime.
it is normal for resperf to consume 100% CPU during the whole test run, even Therefore, it is normal for \fBresperf\fR to consume 100% CPU during the
during periods where query rates are relatively low. whole test run, even during periods where query rates are relatively low.
You will also need a set of test queries in the \fBdnsperf\fR file format. You will also need a set of test queries in the \fBdnsperf\fR file format.
See the \fBdnsperf\fR man page for instructions on how to construct this See the \fBdnsperf\fR man page for instructions on how to construct this
query file. To make the test as realistic as possible, the queries should be query file.
derived from recorded production client DNS traffic, without removing To make the test as realistic as possible, the queries should be derived
duplicate queries or other filtering. With the default settings, resperf from recorded production client DNS traffic, without removing duplicate
will use up to 3 million queries in each test run. queries or other filtering.
With the default settings, \fBresperf\fR will use up to 3 million queries
in each test run.
If the caching server to be tested has a configurable limit on the number of If the caching server to be tested has a configurable limit on the number of
simultaneous resolutions, like the \fBmax\-recursive\-clients\fR statement simultaneous resolutions, like the \fBmax\-recursive\-clients\fR statement
in Nominum CacheServe or the \fBrecursive\-clients\fR option in BIND 9, you in Nominum CacheServe or the \fBrecursive\-clients\fR option in BIND 9, you
will probably have to increase it. As a starting point, we recommend a value will probably have to increase it.
of 10000 for Nominum CacheServe and 100000 for BIND 9. Should the limit be As a starting point, we recommend a value of 10000 for Nominum CacheServe
reached, it will show up in the plots as an increase in the number of and 100000 for BIND 9.
failure responses. Should the limit be reached, it will show up in the plots as an increase in
the number of failure responses.
The server being tested should be restarted at the beginning of each test to The server being tested should be restarted at the beginning of each test to
make sure it is starting with an empty cache. If the cache already contains make sure it is starting with an empty cache.
data from a previous test run that used the same set of queries, almost all If the cache already contains data from a previous test run that used the
queries will be answered from the cache, yielding inflated performance same set of queries, almost all queries will be answered from the cache,
numbers. yielding inflated performance numbers.
To use the \fBresperf\-report\fR script, you need to have \fBgnuplot\fR To use the \fBresperf\-report\fR script, you need to have \fBgnuplot\fR
installed. Make sure your installed version of \fBgnuplot\fR supports the installed.
png terminal driver. If your \fBgnuplot\fR doesn't support png but does Make sure your installed version of \fBgnuplot\fR supports the png terminal
support gif, you can change the line saying terminal=png in the driver.
\fBresperf\-report\fR script to terminal=gif. If your \fBgnuplot\fR doesn't support png but does support gif, you can
change the line saying terminal=png in the \fBresperf\-report\fR script
to terminal=gif.
.SS "Running the test" .SS "Running the test"
Resperf is typically invoked via the \fBresperf\-report\fR script, which \fBresperf\fR is typically invoked via the \fBresperf\-report\fR script,
will run \fBresperf\fR with its output redirected to a file and then which will run \fBresperf\fR with its output redirected to a file and then
automatically generate an illustrated report in HTML format. Command line automatically generate an illustrated report in HTML format.
arguments given to resperf-report will be passed on unchanged to resperf. Command line arguments given to \fBresperf\-report\fR will be passed on
unchanged to \fBresperf\fR.
When running resperf-report, you will need to specify at least the server IP When running \fBresperf\-report\fR, you will need to specify at least the
address and the query data file. A typical invocation will look like server IP address and the query data file.
A typical invocation will look like
.RS .RS
.hy 0 .hy 0
@ -189,124 +210,132 @@ resperf\-report \-s 10.0.0.2 \-d queryfile
With default settings, the test run will take at most 100 seconds (60 With default settings, the test run will take at most 100 seconds (60
seconds of ramping up traffic and then 40 seconds of waiting for responses), seconds of ramping up traffic and then 40 seconds of waiting for responses),
but in practice, the 60-second traffic phase will usually be cut short. To but in practice, the 60-second traffic phase will usually be cut short.
be precise, resperf can transition from the traffic-sending phase to the To be precise, \fBresperf\fR can transition from the traffic-sending phase
waiting-for-responses phase in three different ways: to the waiting-for-responses phase in three different ways:
.IP \(bu 2 .IP \(bu 2
Running for the full allotted time and successfully reaching the maximum Running for the full allotted time and successfully reaching the maximum
query rate (by default, 60 seconds and 100,000 qps, respectively). Since query rate (by default, 60 seconds and 100,000 qps, respectively).
this is a very high query rate, this will rarely happen (with today's Since this is a very high query rate, this will rarely happen (with today's
hardware); one of the other two conditions listed below will usually occur hardware); one of the other two conditions listed below will usually occur
first. first.
.IP \(bu 2 .IP \(bu 2
Exceeding 65,536 outstanding queries. This often happens as a result of Exceeding 65,536 outstanding queries.
(successfully) exceeding the capacity of the server being tested, causing This often happens as a result of (successfully) exceeding the capacity of
the excess queries to be dropped. The limit of 65,536 queries comes from the the server being tested, causing the excess queries to be dropped.
number of possible values for the ID field in the DNS packet. Resperf needs The limit of 65,536 queries comes from the number of possible values for
to allocate a unique ID for each outstanding query, and is therefore unable the ID field in the DNS packet.
to send further queries if the set of possible IDs is exhausted. \fBresperf\fR needs to allocate a unique ID for each outstanding query, and is
therefore unable to send further queries if the set of possible IDs is
exhausted.
.IP \(bu 2 .IP \(bu 2
When resperf finds itself unable to send queries fast enough. Resperf will When \fBresperf\fR finds itself unable to send queries fast enough.
notice if it is falling behind in its scheduled query transmissions, and if \fBresperf\fR will notice if it is falling behind in its scheduled query
this backlog reaches 1000 queries, it will print a message like "Fell behind transmissions, and if this backlog reaches 1000 queries, it will print
by 1000 queries" (or whatever the actual number is at the time) and stop a message like "Fell behind by 1000 queries" (or whatever the actual number
sending traffic. is at the time) and stop sending traffic.
.PP .PP
Regardless of which of the above conditions caused the traffic-sending phase Regardless of which of the above conditions caused the traffic-sending phase
of the test to end, you should examine the resulting plots to make sure the of the test to end, you should examine the resulting plots to make sure the
server's response rate is flattening out toward the end of the test. If it server's response rate is flattening out toward the end of the test.
is not, then you are not loading the server enough. If you are getting the If it is not, then you are not loading the server enough.
"Fell behind" message, make sure that the machine running resperf is fast If you are getting the "Fell behind" message, make sure that the machine
enough and has no other applications running. running \fBresperf\fR is fast enough and has no other applications running.
You should also monitor the CPU usage of the server under test. It should You should also monitor the CPU usage of the server under test.
reach close to 100% CPU at the point of maximum traffic; if it does not, you It should reach close to 100% CPU at the point of maximum traffic; if it does
most likely have a bottleneck in some other part of your test setup, for not, you most likely have a bottleneck in some other part of your test setup,
example, your external Internet connection. for example, your external Internet connection.
The report generated by \fBresperf\-report\fR will be stored with a unique The report generated by \fBresperf\-report\fR will be stored with a unique
file name based on the current date and time, e.g., file name based on the current date and time, e.g.,
\fI20060812-1550.html\fR. The PNG images of the plots and other auxiliary \fI20060812-1550.html\fR.
files will be stored in separate files beginning with the same date-time The PNG images of the plots and other auxiliary files will be stored in
string. To view the report, simply open the \fI.html\fR file in a web separate files beginning with the same date-time string.
browser. To view the report, simply open the \fI.html\fR file in a web browser.
If you need to copy the report to a separate machine for viewing, make sure If you need to copy the report to a separate machine for viewing, make sure
to copy the .png files along with the .html file (or simply copy all the to copy the .png files along with the .html file (or simply copy all the
files, e.g., using scp 20060812-1550.* host:directory/). files, e.g., using scp 20060812-1550.* host:directory/).
.SS "Interpreting the report" .SS "Interpreting the report"
The \fI.html\fR file produced by \fBresperf\-report\fR consists of two The \fI.html\fR file produced by \fBresperf\-report\fR consists of two
sections. The first section, "Resperf output", contains output from the sections.
\fBresperf\fR program such as progress messages, a summary of the command The first section, "Resperf output", contains output from the \fBresperf\fR
line arguments, and summary statistics. The second section, "Plots", program such as progress messages, a summary of the command line arguments,
contains two plots generated by \fBgnuplot\fR: "Query/response/failure rate" and summary statistics.
and "Latency". The second section, "Plots", contains two plots generated by \fBgnuplot\fR:
"Query/response/failure rate" and "Latency".
The "Query/response/failure rate" plot contains three graphs. The "Queries The "Query/response/failure rate" plot contains three graphs.
sent per second" graph shows the amount of traffic being sent to the server; The "Queries sent per second" graph shows the amount of traffic being sent to
this should be very close to a straight diagonal line, reflecting the linear the server; this should be very close to a straight diagonal line, reflecting
ramp-up of traffic. the linear ramp-up of traffic.
The "Total responses received per second" graph shows how many of the The "Total responses received per second" graph shows how many of the
queries received a response from the server. All responses are counted, queries received a response from the server.
whether successful (NOERROR or NXDOMAIN) or not (e.g., SERVFAIL). All responses are counted, whether successful (NOERROR or NXDOMAIN) or not
(e.g., SERVFAIL).
The "Failure responses received per second" graph shows how many of the The "Failure responses received per second" graph shows how many of the
queries received a failure response. A response is considered to be a queries received a failure response.
failure if its RCODE is neither NOERROR nor NXDOMAIN. A response is considered to be a failure if its RCODE is neither NOERROR
nor NXDOMAIN.
By visually inspecting the graphs, you can get an idea of how the server By visually inspecting the graphs, you can get an idea of how the server
behaves under increasing load. The "Total responses received per second" behaves under increasing load.
graph will initially closely follow the "Queries sent per second" graph The "Total responses received per second" graph will initially closely
(often rendering it invisible in the plot as the two graphs are plotted on follow the "Queries sent per second" graph (often rendering it invisible in
top of one another), but when the load exceeds the server's capacity, the the plot as the two graphs are plotted on top of one another), but when the
"Total responses received per second" graph may diverge from the "Queries load exceeds the server's capacity, the "Total responses received per second"
sent per second" graph and flatten out, indicating that some of the queries graph may diverge from the "Queries sent per second" graph and flatten out,
are being dropped. indicating that some of the queries are being dropped.
The "Failure responses received per second" graph will normally show a The "Failure responses received per second" graph will normally show a
roughly linear ramp close to the bottom of the plot with some random roughly linear ramp close to the bottom of the plot with some random
fluctuation, since typical query traffic will contain some small percentage fluctuation, since typical query traffic will contain some small percentage
of failing queries randomly interspersed with the successful ones. As the of failing queries randomly interspersed with the successful ones.
total traffic increases, the number of failures will increase As the total traffic increases, the number of failures will increase
proportionally. proportionally.
If the "Failure responses received per second" graph turns sharply upwards, If the "Failure responses received per second" graph turns sharply upwards,
this can be another indication that the load has exceeded the server's this can be another indication that the load has exceeded the server's
capacity. This will happen if the server reacts to overload by sending capacity.
SERVFAIL responses rather than by dropping queries. Since Nominum CacheServe This will happen if the server reacts to overload by sending SERVFAIL
and BIND 9 will both respond with SERVFAIL when they exceed their responses rather than by dropping queries.
\fBmax\-recursive\-clients\fR or \fBrecursive\-clients\fR limit, Since Nominum CacheServe and BIND 9 will both respond with SERVFAIL when
respectively, a sudden increase in the number of failures could mean that they exceed their \fBmax\-recursive\-clients\fR or \fBrecursive\-clients\fR
the limit needs to be increased. limit, respectively, a sudden increase in the number of failures could mean
that the limit needs to be increased.
The "Latency" plot contains a single graph marked "Average latency". This The "Latency" plot contains a single graph marked "Average latency".
shows how the latency varies during the course of the test. Typically, the This shows how the latency varies during the course of the test.
latency graph will exhibit a downwards trend because the cache hit rate Typically, the latency graph will exhibit a downwards trend because the
improves as ever more responses are cached during the test, and the latency cache hit rate improves as ever more responses are cached during the test,
for a cache hit is much smaller than for a cache miss. The latency graph is and the latency for a cache hit is much smaller than for a cache miss.
provided as an aid in determining the point where the server gets The latency graph is provided as an aid in determining the point where the
overloaded, which can be seen as a sharp upwards turn in the graph. The server gets overloaded, which can be seen as a sharp upwards turn in the
latency graph is not intended for making absolute latency measurements or graph.
comparisons between servers; the latencies shown in the graph are not The latency graph is not intended for making absolute latency measurements
or comparisons between servers; the latencies shown in the graph are not
representative of production latencies due to the initially empty cache and representative of production latencies due to the initially empty cache and
the deliberate overloading of the server towards the end of the test. the deliberate overloading of the server towards the end of the test.
Note that all measurements are displayed on the plot at the horizontal Note that all measurements are displayed on the plot at the horizontal
position corresponding to the point in time when the query was sent, not position corresponding to the point in time when the query was sent, not
when the response (if any) was received. This makes it it easy to compare when the response (if any) was received.
the query and response rates; for example, if no queries are dropped, the This makes it it easy to compare the query and response rates; for example,
query and response graphs will be identical. As another example, if the plot if no queries are dropped, the query and response graphs will be identical.
shows 10% failure responses at t=5 seconds, this means that 10% of the As another example, if the plot shows 10% failure responses at t=5 seconds,
queries sent at t=5 seconds eventually failed, not that 10% of the responses this means that 10% of the queries sent at t=5 seconds eventually failed,
received at t=5 seconds were failures. not that 10% of the responses received at t=5 seconds were failures.
.SS "Determining the server's maximum throughput" .SS "Determining the server's maximum throughput"
Often, the goal of running \fBresperf\fR is to determine the server's Often, the goal of running \fBresperf\fR is to determine the server's
maximum throughput, in other words, the number of queries per second it is maximum throughput, in other words, the number of queries per second it is
capable of handling. This is not always an easy task, because as a server is capable of handling.
driven into overload, the service it provides may deteriorate gradually, and This is not always an easy task, because as a server is driven into overload,
this deterioration can manifest itself either as queries being dropped, as the service it provides may deteriorate gradually, and this deterioration
an increase in the number of SERVFAIL responses, or an increase in latency. can manifest itself either as queries being dropped, as an increase in the
number of SERVFAIL responses, or an increase in latency.
The maximum throughput may be defined as the highest level of traffic at The maximum throughput may be defined as the highest level of traffic at
which the server still provides an acceptable level of service, but that which the server still provides an acceptable level of service, but that
means you first need to decide what an acceptable level of service means in means you first need to decide what an acceptable level of service means in
@ -315,22 +344,24 @@ terms of packet drop percentage, SERVFAIL percentage, and latency.
The summary statistics in the "Resperf output" section of the report The summary statistics in the "Resperf output" section of the report
contains a "Maximum throughput" value which by default is determined from contains a "Maximum throughput" value which by default is determined from
the maximum rate at which the server was able to return responses, without the maximum rate at which the server was able to return responses, without
regard to the number of queries being dropped or failing at that point. This regard to the number of queries being dropped or failing at that point.
method of throughput measurement has the advantage of simplicity, but it may This method of throughput measurement has the advantage of simplicity, but
or may not be appropriate for your needs; the reported value should always it may or may not be appropriate for your needs; the reported value should
be validated by a visual inspection of the graphs to ensure that service has always be validated by a visual inspection of the graphs to ensure that
not already deteriorated unacceptably before the maximum response rate is service has not already deteriorated unacceptably before the maximum response
reached. It may also be helpful to look at the "Lost at that point" value in rate is reached.
It may also be helpful to look at the "Lost at that point" value in
the summary statistics; this indicates the percentage of the queries that the summary statistics; this indicates the percentage of the queries that
was being dropped at the point in the test when the maximum throughput was was being dropped at the point in the test when the maximum throughput was
reached. reached.
Alternatively, you can make resperf report the throughput at the point in Alternatively, you can make \fBresperf\fR report the throughput at the point
the test where the percentage of queries dropped exceeds a given limit (or in the test where the percentage of queries dropped exceeds a given limit
the maximum as above if the limit is never exceeded). This can be a more (or the maximum as above if the limit is never exceeded).
realistic indication of how much the server can be loaded while still This can be a more realistic indication of how much the server can be loaded
providing an acceptable level of service. This is done using the \fB\-L\fR while still providing an acceptable level of service.
command line option; for example, specifying \fB\-L 10\fR makes resperf This is done using the \fB\-L\fR command line option; for example, specifying
\fB\-L 10\fR makes \fBresperf\fR
report the highest throughput reached before the server starts dropping more report the highest throughput reached before the server starts dropping more
than 10% of the queries. than 10% of the queries.
@ -338,44 +369,50 @@ There is no corresponding way of automatically constraining results based on
the number of failed queries, because unlike dropped queries, resolution the number of failed queries, because unlike dropped queries, resolution
failures will occur even when the the server is not overloaded, and the failures will occur even when the the server is not overloaded, and the
number of such failures is heavily dependent on the query data and network number of such failures is heavily dependent on the query data and network
conditions. Therefore, the plots should be manually inspected to ensure that conditions.
there is not an abnormal number of failures. Therefore, the plots should be manually inspected to ensure that there is not
an abnormal number of failures.
.SH "GENERATING CONSTANT TRAFFIC" .SH "GENERATING CONSTANT TRAFFIC"
In addition to ramping up traffic linearly, \fBresperf\fR also has the In addition to ramping up traffic linearly, \fBresperf\fR also has the
capability to send a constant stream of traffic. This can be useful when capability to send a constant stream of traffic.
using \fBresperf\fR for tasks other than performance measurement; for This can be useful when using \fBresperf\fR for tasks other than performance
example, it can be used to "soak test" a server by subjecting it to a measurement; for example, it can be used to "soak test" a server by
sustained load for an extended period of time. subjecting it to a sustained load for an extended period of time.
To generate a constant traffic load, use the \fB\-c\fR command line option, To generate a constant traffic load, use the \fB\-c\fR command line option,
together with the \fB\-m\fR option which specifies the desired constant together with the \fB\-m\fR option which specifies the desired constant
query rate. For example, to send 10000 queries per second for an hour, use query rate.
\fB\-m 10000 \-c 3600\fR. This will include the usual 30-second gradual For example, to send 10000 queries per second for an hour, use \fB\-m 10000
ramp-up of traffic at the beginning, which may be useful to avoid initially \-c 3600\fR.
overwhelming a server that is starting with an empty cache. To start the This will include the usual 30-second gradual ramp-up of traffic at the
onslaught of traffic instantly, use \fB\-m 10000 \-c 3600 \-r 0\fR. beginning, which may be useful to avoid initially overwhelming a server that
is starting with an empty cache.
To start the onslaught of traffic instantly, use \fB\-m 10000 \-c 3600
\-r 0\fR.
To be precise, \fBresperf\fR will do a linear ramp-up of traffic from 0 to To be precise, \fBresperf\fR will do a linear ramp-up of traffic from 0 to
\fB\-m\fR queries per second over a period of \fB\-r\fR seconds, followed by \fB\-m\fR queries per second over a period of \fB\-r\fR seconds, followed by
a plateau of steady traffic at \fB\-m\fR queries per second lasting for a plateau of steady traffic at \fB\-m\fR queries per second lasting for
\fB\-c\fR seconds, followed by waiting for responses for an extra 40 \fB\-c\fR seconds, followed by waiting for responses for an extra 40
seconds. Either the ramp-up or the plateau can be suppressed by supplying a seconds.
duration of zero seconds with \fB\-r 0\fR and \fB\-c 0\fR, respectively. The Either the ramp-up or the plateau can be suppressed by supplying a duration
latter is the default. of zero seconds with \fB\-r 0\fR and \fB\-c 0\fR, respectively.
The latter is the default.
Sending traffic at high rates for hours on end will of course require very Sending traffic at high rates for hours on end will of course require very
large amounts of input data. Also, a long-running test will generate a large large amounts of input data.
amount of plot data, which is kept in memory for the duration of the test. Also, a long-running test will generate a large amount of plot data, which is
kept in memory for the duration of the test.
To reduce the memory usage and the size of the plot file, consider To reduce the memory usage and the size of the plot file, consider
increasing the interval between measurements from the default of 0.5 seconds increasing the interval between measurements from the default of 0.5 seconds
using the \fB\-i\fR option in long-running tests. using the \fB\-i\fR option in long-running tests.
When using \fBresperf\fR for long-running tests, it is important that the When using \fBresperf\fR for long-running tests, it is important that the
traffic rate specified using the \fB\-m\fR is one that both \fBresperf\fR traffic rate specified using the \fB\-m\fR is one that both \fBresperf\fR
itself and the server under test can sustain. Otherwise, the test is likely itself and the server under test can sustain.
to be cut short as a result of either running out of query IDs (because of Otherwise, the test is likely to be cut short as a result of either running
large numbers of dropped queries) or of resperf falling behind its out of query IDs (because of large numbers of dropped queries) or of
transmission schedule. \fBresperf\fR falling behind its transmission schedule.
.SH OPTIONS .SH OPTIONS
Because the \fBresperf\-report\fR script passes its command line options Because the \fBresperf\-report\fR script passes its command line options
directly to the \fBresperf\fR programs, they both accept the same set of directly to the \fBresperf\fR programs, they both accept the same set of
@ -383,57 +420,67 @@ options, with one exception: \fBresperf\-report\fR automatically adds an
appropriate \fB\-P\fR to the \fBresperf\fR command line, and therefore does appropriate \fB\-P\fR to the \fBresperf\fR command line, and therefore does
not itself take a \fB\-P\fR option. not itself take a \fB\-P\fR option.
\fB-d \fIdatafile\fB\fR \fB-d \fIdatafile\fR
.br .br
.RS .RS
Specifies the input data file. If not specified, \fBresperf\fR will read Specifies the input data file.
from standard input. If not specified, \fBresperf\fR will read from standard input.
.RE .RE
\fB-M \fImode\fB\fR \fB-R\fR
.br .br
.RS .RS
Specifies the transport mode to use, "udp", "tcp" or "tls". Default is "udp". Reopen the datafile if it runs out of data before the testing is completed.
This allows for long running tests on very small and simple query datafile.
.RE .RE
\fB-s \fIserver_addr\fB\fR \fB-M \fImode\fR
.br
.RS
Specifies the transport mode to use, "udp", "tcp" or "dot".
Default is "udp".
.RE
\fB-s \fIserver_addr\fR
.br .br
.RS .RS
Specifies the name or address of the server to which requests will be sent. Specifies the name or address of the server to which requests will be sent.
The default is the loopback address, 127.0.0.1. The default is the loopback address, 127.0.0.1.
.RE .RE
\fB-p \fIport\fB\fR \fB-p \fIport\fR
.br .br
.RS .RS
Sets the port on which the DNS packets are sent. If not specified, the Sets the port on which the DNS packets are sent.
standard DNS port (udp/tcp 53, tls 853) is used. If not specified, the standard DNS port (udp/tcp 53, DoT 853) is used.
.RE .RE
\fB-a \fIlocal_addr\fB\fR \fB-a \fIlocal_addr\fR
.br .br
.RS .RS
Specifies the local address from which to send requests. The default is the Specifies the local address from which to send requests.
wildcard address. The default is the wildcard address.
.RE .RE
\fB-x \fIlocal_port\fB\fR \fB-x \fIlocal_port\fR
.br .br
.RS .RS
Specifies the local port from which to send requests. The default is the Specifies the local port from which to send requests.
wildcard port (0). The default is the wildcard port (0).
If acting as multiple clients and the wildcard port is used, each client If acting as multiple clients and the wildcard port is used, each client
will use a different random port. If a port is specified, the clients will will use a different random port.
use a range of ports starting with the specified one. If a port is specified, the clients will use a range of ports starting
with the specified one.
.RE .RE
\fB-t \fItimeout\fB\fR \fB-t \fItimeout\fR
.br .br
.RS .RS
Specifies the request timeout value, in seconds. \fBresperf\fR will no Specifies the request timeout value, in seconds.
longer wait for a response to a particular request after this many seconds \fBresperf\fR will no longer wait for a response to a particular request
have elapsed. The default is 45 seconds. after this many seconds have elapsed.
The default is 45 seconds.
\fBresperf\fR times out unanswered requests in order to reclaim query IDs so \fBresperf\fR times out unanswered requests in order to reclaim query IDs so
that the query ID space will not be exhausted in a long-running test, such that the query ID space will not be exhausted in a long-running test, such
@ -442,9 +489,10 @@ The timeouts and the ability to tune them are of little use in the more
typical use case of a performance test lasting only a minute or two. typical use case of a performance test lasting only a minute or two.
The default timeout of 45 seconds was chosen to be longer than the query The default timeout of 45 seconds was chosen to be longer than the query
timeout of current caching servers. Note that this is longer than the timeout of current caching servers.
corresponding default in \fBdnsperf\fR, because caching servers can take Note that this is longer than the corresponding default in \fBdnsperf\fR,
many orders of magnitude longer to answer a query than authoritative servers because caching servers can take many orders of magnitude longer to answer
a query than authoritative servers
do. do.
If a short timeout is used, there is a possibility that \fBresperf\fR will If a short timeout is used, there is a possibility that \fBresperf\fR will
@ -453,20 +501,20 @@ case, a message like Warning: Received a response with an unexpected id: 141
will be printed. will be printed.
.RE .RE
\fB-b \fIbufsize\fB\fR \fB-b \fIbufsize\fR
.br .br
.RS .RS
Sets the size of the socket's send and receive buffers, in kilobytes. If not Sets the size of the socket's send and receive buffers, in kilobytes.
specified, the operating system's default is used. If not specified, the operating system's default is used.
.RE .RE
\fB-f \fIfamily\fB\fR \fB-f \fIfamily\fR
.br .br
.RS .RS
Specifies the address family used for sending DNS packets. The possible Specifies the address family used for sending DNS packets.
values are "inet", "inet6", or "any". If "any" (the default value) is The possible values are "inet", "inet6", or "any".
specified, \fBresperf\fR will use whichever address family is appropriate If "any" (the default value) is specified, \fBresperf\fR will use whichever
for the server it is sending packets to. address family is appropriate for the server it is sending packets to.
.RE .RE
\fB-e\fR \fB-e\fR
@ -478,11 +526,11 @@ Enables EDNS0 [RFC2671], by adding an OPT record to all packets sent.
\fB-D\fR \fB-D\fR
.br .br
.RS .RS
Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent. This also enables Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent.
EDNS0, which is required for DNSSEC. This also enables EDNS0, which is required for DNSSEC.
.RE .RE
\fB-y \fI[alg:]name:secret\fB\fR \fB-y \fI[alg:]name:secret\fR
.br .br
.RS .RS
Add a TSIG record [RFC2845] to all packets sent, using the specified TSIG Add a TSIG record [RFC2845] to all packets sent, using the specified TSIG
@ -496,67 +544,82 @@ the secret is expressed as a base-64 encoded string.
Print a usage statement and exit. Print a usage statement and exit.
.RE .RE
\fB-i \fIinterval\fB\fR \fB-i \fIinterval\fR
.br .br
.RS .RS
Specifies the time interval between data points in the plot file. The Specifies the time interval between data points in the plot file.
default is 0.5 seconds. The default is 0.5 seconds.
.RE .RE
\fB-m \fImax_qps\fB\fR \fB-m \fImax_qps\fR
.br .br
.RS .RS
Specifies the target maximum query rate (in queries per second). This should Specifies the target maximum query rate (in queries per second).
be higher than the expected maximum throughput of the server being tested. This should be higher than the expected maximum throughput of the server
being tested.
Traffic will be ramped up at a linearly increasing rate until this value is Traffic will be ramped up at a linearly increasing rate until this value is
reached, or until one of the other conditions described in the section reached, or until one of the other conditions described in the section
"Running the test" occurs. The default is 100000 queries per second. "Running the test" occurs.
The default is 100000 queries per second.
.RE .RE
\fB-P \fIplot_data_file\fB\fR \fB-P \fIplot_data_file\fR
.br .br
.RS .RS
Specifies the name of the plot data file. The default is Specifies the name of the plot data file.
\fIresperf.gnuplot\fR. The default is \fIresperf.gnuplot\fR.
.RE .RE
\fB-r \fIrampup_time\fB\fR \fB-r \fIrampup_time\fR
.br .br
.RS .RS
Specifies the length of time over which traffic will be ramped up. The Specifies the length of time over which traffic will be ramped up.
default is 60 seconds. The default is 60 seconds.
.RE .RE
\fB-c \fIconstant_traffic_time\fB\fR \fB-c \fIconstant_traffic_time\fR
.br .br
.RS .RS
Specifies the length of time for which traffic will be sent at a constant Specifies the length of time for which traffic will be sent at a constant
rate following the initial ramp-up. The default is 0 seconds, meaning no rate following the initial ramp-up.
sending of traffic at a constant rate will be done. The default is 0 seconds, meaning no sending of traffic at a constant rate
will be done.
.RE .RE
\fB-L \fImax_loss\fB\fR \fB-L \fImax_loss\fR
.br .br
.RS .RS
Specifies the maximum acceptable query loss percentage for purposes of Specifies the maximum acceptable query loss percentage for purposes of
determining the maximum throughput value. The default is 100%, meaning that determining the maximum throughput value.
\fBresperf\fR will measure the maximum throughput without regard to query The default is 100%, meaning that \fBresperf\fR will measure the maximum
throughput without regard to query
loss. loss.
.RE .RE
\fB-C \fIclients\fB\fR \fB-C \fIclients\fR
.br .br
.RS .RS
Act as multiple clients. Requests are sent from multiple sockets. The Act as multiple clients.
default is to act as 1 client. Requests are sent from multiple sockets.
The default is to act as 1 client.
.RE .RE
\fB-q \fImax_outstanding\fB\fR \fB-q \fImax_outstanding\fR
.br .br
.RS .RS
Sets the maximum number of outstanding requests. \fBresperf\fR will stop Sets the maximum number of outstanding requests.
ramping up traffic when this many queries are outstanding. The default is \fBresperf\fR will stop ramping up traffic when this many queries are
64k, and the limit is 64k per client. outstanding.
The default is 64k, and the limit is 64k per client.
.RE
\fB-F \fIfall_behind\fR
.br
.RS
Sets the maximum number of queries that can fall behind being sent.
\fBresperf\fR will stop when this many queries should have been sent and it
can be relative easy to hit if \fImax_qps\fR is set too high.
The default is 1000 and setting it to zero (0) disables the check.
.RE .RE
\fB-v\fR \fB-v\fR
@ -564,22 +627,31 @@ ramping up traffic when this many queries are outstanding. The default is
.RS .RS
Enables verbose mode to report about network readiness and congestion. Enables verbose mode to report about network readiness and congestion.
.RE .RE
\fB-W\fR
.br
.RS
Log warnings and errors to standard output instead of standard error making
it easier for script, test and automation to capture all output.
.RE
.SH "THE PLOT DATA FILE" .SH "THE PLOT DATA FILE"
The plot data file is written by the \fBresperf\fR program and contains the The plot data file is written by the \fBresperf\fR program and contains the
data to be plotted using \fBgnuplot\fR. When running \fBresperf\fR via the data to be plotted using \fBgnuplot\fR.
\fBresperf\-report\fR script, there is no need for the user to deal with When running \fBresperf\fR via the \fBresperf\-report\fR script, there is
this file directly, but its format and contents are documented here for no need for the user to deal with this file directly, but its format and
completeness and in case you wish to run \fBresperf\fR directly and use its contents are documented here for completeness and in case you wish to run
output for purposes other than viewing it with \fBgnuplot\fR. \fBresperf\fR directly and use its output for purposes other than viewing
it with \fBgnuplot\fR.
The first line of the file is a comment identifying the fields. It may be The first line of the file is a comment identifying the fields.
recognized as a comment by its leading hash sign (#). It may be recognized as a comment by its leading hash sign (#).
Subsequent lines contain the actual plot data. For purposes of generating Subsequent lines contain the actual plot data.
the plot data file, the test run is divided into time intervals of 0.5 For purposes of generating the plot data file, the test run is divided into
seconds (or some other length of time specified with the \fB\-i\fR command time intervals of 0.5 seconds (or some other length of time specified with
line option). Each line corresponds to one such interval, and contains the the \fB\-i\fR command line option).
following values as floating-point numbers: Each line corresponds to one such interval, and contains the following values
as floating-point numbers:
\fBTime\fR \fBTime\fR
.br .br
@ -621,6 +693,23 @@ length of the interval
The average time between sending the query and receiving a response, for The average time between sending the query and receiving a response, for
queries sent in this time interval queries sent in this time interval
.RE .RE
\fBConnections\fR
.br
.RS
The number of connections done, including re-connections, during this time
interval.
This is only relevant to connection oriented protocols, such as TCP and DoT.
.RE
\fBAverage connection latency\fR
.br
.RS
The average time between starting to connect and having the connection ready
for sending queries to, for this time interval.
This is only relevant to connection oriented protocols, such as TCP and DoT.
.RE
.SH "SEE ALSO" .SH "SEE ALSO"
\fBdnsperf\fR(1) \fBdnsperf\fR(1)
.SH AUTHOR .SH AUTHOR

155
src/resperf.c
View file

@ -52,12 +52,13 @@
#define DEFAULT_SERVER_NAME "127.0.0.1" #define DEFAULT_SERVER_NAME "127.0.0.1"
#define DEFAULT_SERVER_PORT 53 #define DEFAULT_SERVER_PORT 53
#define DEFAULT_SERVER_TLS_PORT 853 #define DEFAULT_SERVER_DOT_PORT 853
#define DEFAULT_SERVER_PORTS "udp/tcp 53 or dot/tls 853" #define DEFAULT_SERVER_PORTS "udp/tcp 53 or DoT 853"
#define DEFAULT_LOCAL_PORT 0 #define DEFAULT_LOCAL_PORT 0
#define DEFAULT_SOCKET_BUFFER 32 #define DEFAULT_SOCKET_BUFFER 32
#define DEFAULT_TIMEOUT 45 #define DEFAULT_TIMEOUT 45
#define DEFAULT_MAX_OUTSTANDING (64 * 1024) #define DEFAULT_MAX_OUTSTANDING (64 * 1024)
#define DEFAULT_MAX_FALL_BEHIND 1000
#define MAX_INPUT_DATA (64 * 1024) #define MAX_INPUT_DATA (64 * 1024)
@ -72,6 +73,8 @@ typedef perf_list(struct query_info) query_list;
typedef struct query_info { typedef struct query_info {
uint64_t sent_timestamp; uint64_t sent_timestamp;
bool is_inprogress;
/* /*
* This link links the query into the list of outstanding * This link links the query into the list of outstanding
* queries or the list of available query IDs. * queries or the list of available query IDs.
@ -91,7 +94,7 @@ static query_info* queries;
static perf_sockaddr_t server_addr; static perf_sockaddr_t server_addr;
static perf_sockaddr_t local_addr; static perf_sockaddr_t local_addr;
static unsigned int nsocks; static unsigned int nsocks;
static struct perf_net_socket* socks; static struct perf_net_socket** socks;
static enum perf_net_mode mode; static enum perf_net_mode mode;
static int dummypipe[2]; static int dummypipe[2];
@ -143,6 +146,7 @@ static uint64_t num_queries_outstanding;
static uint64_t num_responses_received; static uint64_t num_responses_received;
static uint64_t num_queries_timed_out; static uint64_t num_queries_timed_out;
static uint64_t rcodecounts[16]; static uint64_t rcodecounts[16];
static uint64_t num_reconnections;
static uint64_t time_now; static uint64_t time_now;
static uint64_t time_of_program_start; static uint64_t time_of_program_start;
@ -164,6 +168,9 @@ typedef struct {
int responses; int responses;
int failures; int failures;
double latency_sum; double latency_sum;
int connections;
double conn_latency_sum;
} ramp_bucket; } ramp_bucket;
/* Pointer to array of n_buckets ramp_bucket structures */ /* Pointer to array of n_buckets ramp_bucket structures */
@ -193,6 +200,7 @@ static uint64_t sustain_phase_began, wait_phase_began;
static perf_tsigkey_t* tsigkey; static perf_tsigkey_t* tsigkey;
static bool verbose; static bool verbose;
static unsigned int max_fall_behind;
const char* progname = "resperf"; const char* progname = "resperf";
@ -205,6 +213,9 @@ stringify(double value, int precision)
return buf; return buf;
} }
static void perf__net_event(struct perf_net_socket* sock, perf_socket_event_t event, uint64_t elapsed_time);
static void perf__net_sent(struct perf_net_socket* sock, uint16_t qid);
static void static void
setup(int argc, char** argv) setup(int argc, char** argv)
{ {
@ -232,11 +243,12 @@ setup(int argc, char** argv)
nsocks = 1; nsocks = 1;
mode = sock_udp; mode = sock_udp;
verbose = false; verbose = false;
max_fall_behind = DEFAULT_MAX_FALL_BEHIND;
perf_opt_add('f', perf_opt_string, "family", perf_opt_add('f', perf_opt_string, "family",
"address family of DNS transport, inet or inet6", "any", "address family of DNS transport, inet or inet6", "any",
&family); &family);
perf_opt_add('M', perf_opt_string, "mode", "set transport mode: udp, tcp or dot/tls", "udp", &_mode); perf_opt_add('M', perf_opt_string, "mode", "set transport mode: udp, tcp or dot", "udp", &_mode);
perf_opt_add('s', perf_opt_string, "server_addr", perf_opt_add('s', perf_opt_string, "server_addr",
"the server to query", DEFAULT_SERVER_NAME, &server_name); "the server to query", DEFAULT_SERVER_NAME, &server_name);
perf_opt_add('p', perf_opt_port, "port", perf_opt_add('p', perf_opt_port, "port",
@ -280,7 +292,7 @@ setup(int argc, char** argv)
"the maximum acceptable query loss, in percent", "the maximum acceptable query loss, in percent",
stringify(max_loss_percent, 0), &max_loss_percent); stringify(max_loss_percent, 0), &max_loss_percent);
perf_opt_add('C', perf_opt_uint, "clients", perf_opt_add('C', perf_opt_uint, "clients",
"the number of clients to act as", NULL, &nsocks); "the number of clients to act as", stringify(1, 0), &nsocks);
perf_opt_add('q', perf_opt_uint, "num_outstanding", perf_opt_add('q', perf_opt_uint, "num_outstanding",
"the maximum number of queries outstanding", "the maximum number of queries outstanding",
stringify(DEFAULT_MAX_OUTSTANDING, 0), &max_outstanding); stringify(DEFAULT_MAX_OUTSTANDING, 0), &max_outstanding);
@ -289,6 +301,10 @@ setup(int argc, char** argv)
NULL, &verbose); NULL, &verbose);
bool log_stdout = false; bool log_stdout = false;
perf_opt_add('W', perf_opt_boolean, NULL, "log warnings and errors to stdout instead of stderr", NULL, &log_stdout); perf_opt_add('W', perf_opt_boolean, NULL, "log warnings and errors to stdout instead of stderr", NULL, &log_stdout);
bool reopen_datafile = false;
perf_opt_add('R', perf_opt_boolean, NULL, "reopen datafile on end, allow for infinit use of it", NULL, &reopen_datafile);
perf_opt_add('F', perf_opt_zpint, "fall_behind", "the maximum number of queries that is allowed to fall behind, zero to disable",
stringify(DEFAULT_MAX_FALL_BEHIND, 0), &max_fall_behind);
perf_opt_parse(argc, argv); perf_opt_parse(argc, argv);
@ -300,7 +316,7 @@ setup(int argc, char** argv)
mode = perf_net_parsemode(_mode); mode = perf_net_parsemode(_mode);
if (!server_port) { if (!server_port) {
server_port = mode == sock_tls ? DEFAULT_SERVER_TLS_PORT : DEFAULT_SERVER_PORT; server_port = mode == sock_dot ? DEFAULT_SERVER_DOT_PORT : DEFAULT_SERVER_PORT;
} }
if (max_outstanding > nsocks * DEFAULT_MAX_OUTSTANDING) if (max_outstanding > nsocks * DEFAULT_MAX_OUTSTANDING)
@ -330,6 +346,9 @@ setup(int argc, char** argv)
local_port, &local_addr); local_port, &local_addr);
input = perf_datafile_open(filename); input = perf_datafile_open(filename);
if (reopen_datafile) {
perf_datafile_setmaxruns(input, -1);
}
if (dnssec) if (dnssec)
edns = true; edns = true;
@ -340,8 +359,15 @@ setup(int argc, char** argv)
if (!(socks = calloc(nsocks, sizeof(*socks)))) { if (!(socks = calloc(nsocks, sizeof(*socks)))) {
perf_log_fatal("out of memory"); perf_log_fatal("out of memory");
} }
for (i = 0; i < nsocks; i++) for (i = 0; i < nsocks; i++) {
socks[i] = perf_net_opensocket(mode, &server_addr, &local_addr, i, bufsize); socks[i] = perf_net_opensocket(mode, &server_addr, &local_addr, i, bufsize);
if (!socks[i]) {
perf_log_fatal("perf_net_opensocket(): no socket returned, out of memory?");
}
socks[i]->data = (void*)(intptr_t)i;
socks[i]->sent = perf__net_sent;
socks[i]->event = perf__net_event;
}
} }
static void static void
@ -351,7 +377,7 @@ cleanup(void)
perf_datafile_close(&input); perf_datafile_close(&input);
for (i = 0; i < nsocks; i++) for (i = 0; i < nsocks; i++)
(void)perf_net_close(&socks[i]); (void)perf_net_close(socks[i]);
close(dummypipe[0]); close(dummypipe[0]);
close(dummypipe[1]); close(dummypipe[1]);
} }
@ -374,6 +400,30 @@ find_bucket(uint64_t when)
return &buckets[i]; return &buckets[i];
} }
static void perf__net_event(struct perf_net_socket* sock, perf_socket_event_t event, uint64_t elapsed_time)
{
ramp_bucket* b = find_bucket(time_now);
switch (event) {
case perf_socket_event_reconnect:
num_reconnections++;
case perf_socket_event_connect:
b->connections++;
b->conn_latency_sum += elapsed_time / (double)MILLION;
}
}
static void perf__net_sent(struct perf_net_socket* sock, uint16_t qid)
{
ramp_bucket* b = find_bucket(time_now);
b->queries++;
size_t idx = (size_t)qid * nsocks + (intptr_t)sock->data;
assert(idx < max_outstanding);
queries[idx].sent_timestamp = time_now;
}
/* /*
* print_statistics: * print_statistics:
* Print out statistics based on the results of the test * Print out statistics based on the results of the test
@ -409,6 +459,7 @@ print_statistics(void)
(rcodecounts[i] * 100.0) / num_responses_received); (rcodecounts[i] * 100.0) / num_responses_received);
} }
printf("\n"); printf("\n");
printf(" Reconnection(s): %" PRIu64 "\n", num_reconnections);
printf(" Run time (s): %u.%06u\n", printf(" Run time (s): %u.%06u\n",
(unsigned int)(run_time / MILLION), (unsigned int)(run_time / MILLION),
(unsigned int)(run_time % MILLION)); (unsigned int)(run_time % MILLION));
@ -470,8 +521,8 @@ do_one_line(perf_buffer_t* lines, perf_buffer_t* msg)
qid = (q - queries) / nsocks; qid = (q - queries) / nsocks;
sock = (q - queries) % nsocks; sock = (q - queries) % nsocks;
if (socks[sock].sending) { while (q->is_inprogress) {
if (perf_net_sockready(&socks[sock], dummypipe[0], TIMEOUT_CHECK_TIME) == -1) { if (perf_net_sockready(socks[sock], dummypipe[0], TIMEOUT_CHECK_TIME) == -1) {
if (errno == EINPROGRESS) { if (errno == EINPROGRESS) {
if (verbose) { if (verbose) {
perf_log_warning("network congested, packet sending in progress"); perf_log_warning("network congested, packet sending in progress");
@ -479,12 +530,13 @@ do_one_line(perf_buffer_t* lines, perf_buffer_t* msg)
} else { } else {
if (verbose) { if (verbose) {
char __s[256]; char __s[256];
perf_log_warning("failed to send packet: %s", perf_strerror_r(errno, __s, sizeof(__s))); perf_log_warning("failed to check socket readiness: %s", perf_strerror_r(errno, __s, sizeof(__s)));
} }
} }
return (PERF_R_FAILURE); return (PERF_R_FAILURE);
} }
q->is_inprogress = false;
perf_list_unlink(instanding_list, q); perf_list_unlink(instanding_list, q);
perf_list_prepend(outstanding_list, q); perf_list_prepend(outstanding_list, q);
q->list = &outstanding_list; q->list = &outstanding_list;
@ -499,14 +551,20 @@ do_one_line(perf_buffer_t* lines, perf_buffer_t* msg)
sock = (q - queries) % nsocks; sock = (q - queries) % nsocks;
} }
switch (perf_net_sockready(&socks[sock], dummypipe[0], TIMEOUT_CHECK_TIME)) { switch (perf_net_sockready(socks[sock], dummypipe[0], TIMEOUT_CHECK_TIME)) {
case 0: case 0:
if (verbose) { if (verbose) {
perf_log_warning("failed to send packet: socket %d not ready", sock); perf_log_warning("failed to send packet: socket %d not ready", sock);
} }
return (PERF_R_FAILURE); return (PERF_R_FAILURE);
case -1: case -1:
perf_log_warning("failed to send packet: socket %d readiness check timed out", sock); if (errno == EINPROGRESS) {
if (verbose) {
perf_log_warning("network congested, packet sending in progress");
}
} else {
perf_log_warning("failed to send packet: socket %d not ready", sock);
}
return (PERF_R_FAILURE); return (PERF_R_FAILURE);
default: default:
break; break;
@ -530,13 +588,14 @@ do_one_line(perf_buffer_t* lines, perf_buffer_t* msg)
base = perf_buffer_base(msg); base = perf_buffer_base(msg);
length = perf_buffer_usedlength(msg); length = perf_buffer_usedlength(msg);
if (perf_net_sendto(&socks[sock], base, length, 0, if (perf_net_sendto(socks[sock], qid, base, length, 0,
&server_addr.sa.sa, server_addr.length) &server_addr.sa.sa, server_addr.length)
< 1) { < 1) {
if (errno == EINPROGRESS) { if (errno == EINPROGRESS) {
if (verbose) { if (verbose) {
perf_log_warning("network congested, packet sending in progress"); perf_log_warning("network congested, packet sending in progress");
} }
q->is_inprogress = true;
} else { } else {
if (verbose) { if (verbose) {
char __s[256]; char __s[256];
@ -591,8 +650,17 @@ try_process_response(unsigned int sockindex)
ramp_bucket* b; ramp_bucket* b;
int n; int n;
if (perf_net_sockready(socks[sockindex], dummypipe[0], TIMEOUT_CHECK_TIME) == -1) {
if (errno != EINPROGRESS) {
if (verbose) {
char __s[256];
perf_log_warning("failed to check socket readiness: %s", perf_strerror_r(errno, __s, sizeof(__s)));
}
}
}
packet_header = (uint16_t*)packet_buffer; packet_header = (uint16_t*)packet_buffer;
n = perf_net_recv(&socks[sockindex], packet_buffer, sizeof(packet_buffer), 0); n = perf_net_recv(socks[sockindex], packet_buffer, sizeof(packet_buffer), 0);
if (n < 0) { if (n < 0) {
if (errno == EAGAIN || errno == EINTR) { if (errno == EAGAIN || errno == EINTR) {
return; return;
@ -611,11 +679,12 @@ try_process_response(unsigned int sockindex)
qid = ntohs(packet_header[0]); qid = ntohs(packet_header[0]);
rcode = ntohs(packet_header[1]) & 0xF; rcode = ntohs(packet_header[1]) & 0xF;
q = &queries[qid * nsocks + sockindex]; size_t idx = qid * nsocks + sockindex;
if (q->list != &outstanding_list) { if (idx >= max_outstanding || queries[idx].list != &outstanding_list) {
perf_log_warning("received a response with an unexpected id: %u", qid); perf_log_warning("received a response with an unexpected id: %u", qid);
return; return;
} }
q = &queries[idx];
perf_list_unlink(outstanding_list, q); perf_list_unlink(outstanding_list, q);
perf_list_append(instanding_list, q); perf_list_append(instanding_list, q);
@ -661,21 +730,6 @@ num_scheduled(uint64_t time_since_start)
} }
} }
static void
handle_sigpipe(int sig)
{
(void)sig;
switch (mode) {
case sock_tcp:
case sock_tls:
// if connection is closed it will generate a signal
perf_log_fatal("SIGPIPE received, connection(s) likely closed, can't continue");
break;
default:
break;
}
}
int main(int argc, char** argv) int main(int argc, char** argv)
{ {
int i; int i;
@ -701,7 +755,15 @@ int main(int argc, char** argv)
if (pipe(dummypipe) < 0) if (pipe(dummypipe) < 0)
perf_log_fatal("creating pipe"); perf_log_fatal("creating pipe");
perf_os_handlesignal(SIGPIPE, handle_sigpipe); switch (mode) {
case sock_tcp:
case sock_dot:
// block SIGPIPE for TCP/DOT mode, if connection is closed it will generate a signal
perf_os_blocksignal(SIGPIPE, true);
break;
default:
break;
}
perf_buffer_init(&lines, input_data, sizeof(input_data)); perf_buffer_init(&lines, input_data, sizeof(input_data));
@ -725,6 +787,7 @@ int main(int argc, char** argv)
printf("[Status] Sending\n"); printf("[Status] Sending\n");
int try_responses = (max_qps / max_outstanding) + 1;
current_sock = 0; current_sock = 0;
for (;;) { for (;;) {
int should_send; int should_send;
@ -745,7 +808,7 @@ int main(int argc, char** argv)
} }
if (phase != PHASE_WAIT) { if (phase != PHASE_WAIT) {
should_send = num_scheduled(time_since_start) - num_queries_sent; should_send = num_scheduled(time_since_start) - num_queries_sent;
if (should_send >= 1000) { if (max_fall_behind && should_send >= max_fall_behind) {
printf("[Status] Fell behind by %d queries, " printf("[Status] Fell behind by %d queries, "
"ending test at %.0f qps\n", "ending test at %.0f qps\n",
should_send, (max_qps * time_since_start) / ramp_time); should_send, (max_qps * time_since_start) / ramp_time);
@ -762,8 +825,11 @@ int main(int argc, char** argv)
} }
} }
} }
for (i = try_responses; i--;) {
try_process_response(current_sock++); try_process_response(current_sock++);
current_sock = current_sock % nsocks; if (current_sock >= nsocks)
current_sock = 0;
}
retire_old_queries(); retire_old_queries();
time_now = perf_get_time(); time_now = perf_get_time();
} }
@ -780,8 +846,8 @@ end_loop:
} }
/* Print column headers */ /* Print column headers */
fprintf(plotf, "# time target_qps actual_qps " fprintf(plotf, "# time target_qps actual_qps responses_per_sec failures_per_sec avg_latency"
"responses_per_sec failures_per_sec avg_latency\n"); " connections conn_avg_latency\n");
/* Don't print unused buckets */ /* Don't print unused buckets */
last_bucket_used = find_bucket(wait_phase_began) - buckets; last_bucket_used = find_bucket(wait_phase_began) - buckets;
@ -796,13 +862,18 @@ end_loop:
double target_qps = t <= ramp_dtime ? (t / ramp_dtime) * max_qps : max_qps; double target_qps = t <= ramp_dtime ? (t / ramp_dtime) * max_qps : max_qps;
double latency = buckets[i].responses ? buckets[i].latency_sum / buckets[i].responses : 0; double latency = buckets[i].responses ? buckets[i].latency_sum / buckets[i].responses : 0;
double interval = bucket_interval / (double)MILLION; double interval = bucket_interval / (double)MILLION;
fprintf(plotf, "%7.3f %8.2f %8.2f %8.2f %8.2f %8.6f\n",
double conn_latency = buckets[i].connections ? buckets[i].conn_latency_sum / buckets[i].connections : 0;
fprintf(plotf, "%7.3f %8.2f %8.2f %8.2f %8.2f %8.6f %8.2f %8.6f\n",
t, t,
target_qps, target_qps,
buckets[i].queries / interval, (double)buckets[i].queries / interval,
buckets[i].responses / interval, (double)buckets[i].responses / interval,
buckets[i].failures / interval, (double)buckets[i].failures / interval,
latency); latency,
(double)buckets[i].connections / interval,
conn_latency);
} }
fclose(plotf); fclose(plotf);

6
src/test/Makefile.am
View file

@ -1,9 +1,9 @@
MAINTAINERCLEANFILES = $(srcdir)/Makefile.in MAINTAINERCLEANFILES = $(srcdir)/Makefile.in
CLEANFILES = test*.log test*.trs \ CLEANFILES = test*.log test*.trs \
test2.out test2.out test4.out test4err.out
TESTS = test1.sh test2.sh test3.sh TESTS = test1.sh test2.sh test3.sh test4.sh
EXTRA_DIST = $(TESTS) \ EXTRA_DIST = $(TESTS) \
datafile datafile2 updatefile datafile datafile2 updatefile datafile3 datafile4

3
src/test/datafile4 Normal file
View file

@ -0,0 +1,3 @@
api-read.facebook.com.\002\004\003\002\002\002\002\005\004\004\003\004\006\005\006\006\006\005\006\006\006\007\009\008\006\007\009\007\006\006\008\011\008\009\010\010\010\010\010\006\008\011\012\011\010\012\009\010\010\010\255\219. A
valid\.quote.com A
invalid\0quote.com A

6
src/test/test2.sh
View file

@ -16,7 +16,7 @@ grep -q "Queries sent: *4" test2.out
../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m tcp >test2.out ../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m tcp >test2.out
cat test2.out cat test2.out
grep -q "Queries sent: *2" test2.out grep -q "Queries sent: *2" test2.out
../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m tls >test2.out ../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m dot >test2.out
cat test2.out cat test2.out
grep -q "Queries sent: *2" test2.out grep -q "Queries sent: *2" test2.out
../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m dot >test2.out ../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m dot >test2.out
@ -70,7 +70,7 @@ grep -q "Queries sent: *2" test2.out
# Ignore failure until https://github.com/DNS-OARC/dnsperf/issues/88 is fixed # Ignore failure until https://github.com/DNS-OARC/dnsperf/issues/88 is fixed
# May work on slower systems # May work on slower systems
../resperf -s $ip -m 1 -d "$srcdir/datafile2" -r 2 -c 2 -M tls || true ../resperf -s $ip -m 1 -d "$srcdir/datafile2" -r 2 -c 2 -M dot || true
done # for ip done # for ip
@ -78,7 +78,7 @@ done # for ip
sleep 2 sleep 2
pkill -KILL -u `id -u` dnsperf || true pkill -KILL -u `id -u` dnsperf || true
../dnsperf -s 127.66.66.66 -d "$srcdir/datafile" -vvvv -m tls -n 1 & ../dnsperf -s 127.66.66.66 -d "$srcdir/datafile" -vvvv -m dot -n 1 &
sleep 2 sleep 2
pkill -KILL -u `id -u` dnsperf || true pkill -KILL -u `id -u` dnsperf || true

9
src/test/test4.sh Executable file
View file

@ -0,0 +1,9 @@
#!/bin/sh -xe
test "$TEST_DNSPERF_WITH_NETWORK" = "1" || exit 0
../dnsperf -vvv -d "$srcdir/datafile4" -t 0 -s 127.0.0.1 >test4.out 2>test4err.out
grep 'api-read.facebook.com.\\002\\004\\003\\002\\002\\002\\002\\005\\004\\004\\003\\004\\006\\005\\006\\006\\006\\005\\006\\006\\006\\007\\009\\008\\006\\007\\009\\007\\006\\006\\008\\011\\008\\009\\010\\010\\010\\010\\010\\006\\008\\011\\012\\011\\010\\012\\009\\010\\010\\010\\255\\219. A' test4.out
grep 'T valid\\.quote.com A' test4.out
grep 'Warning: invalid domain name (or out of space): invalid\\0quote.com' test4err.out