diff options
Diffstat (limited to 'src')
| -rw-r--r-- | src/Makefile.am | 2 | ||||
| -rw-r--r-- | src/dns.c | 57 | ||||
| -rw-r--r-- | src/dnsperf.1.in | 268 | ||||
| -rw-r--r-- | src/dnsperf.c | 131 | ||||
| -rw-r--r-- | src/net.c | 500 | ||||
| -rw-r--r-- | src/net.h | 128 | ||||
| -rw-r--r-- | src/net_dot.c | 468 | ||||
| -rw-r--r-- | src/net_tcp.c | 348 | ||||
| -rw-r--r-- | src/net_udp.c | 123 | ||||
| -rw-r--r-- | src/opt.c | 4 | ||||
| -rw-r--r-- | src/opt.h | 3 | ||||
| -rw-r--r-- | src/os.c | 13 | ||||
| -rw-r--r-- | src/os.h | 4 | ||||
| -rw-r--r-- | src/resperf.1.in | 671 | ||||
| -rw-r--r-- | src/resperf.c | 169 | ||||
| -rw-r--r-- | src/test/Makefile.am | 6 | ||||
| -rw-r--r-- | src/test/datafile4 | 3 | ||||
| -rwxr-xr-x | src/test/test2.sh | 6 | ||||
| -rwxr-xr-x | src/test/test4.sh | 9 |
19 files changed, 1917 insertions, 996 deletions
diff --git a/src/Makefile.am b/src/Makefile.am index 35d10a3..d08b272 100644 --- a/src/Makefile.am +++ b/src/Makefile.am @@ -30,7 +30,7 @@ bin_PROGRAMS = dnsperf resperf dist_bin_SCRIPTS = resperf-report _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 \ list.h result.h buffer.h qtype.h edns.h tsig.h @@ -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) { - size_t label_len; + size_t label_len, at; const char* orig_str = str; + bool is_quoted; if (perf_buffer_availablelength(target) < len) { return PERF_R_NOSPACE; } while (len) { - for (label_len = 0; label_len < len; label_len++) { - if (*(str + label_len) == '.') { + is_quoted = false; + 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; } + label_len++; + at++; } if (!label_len) { // 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; } perf_buffer_putuint8(target, label_len); - perf_buffer_putmem(target, str, label_len); - str += label_len; - len -= 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); + str += label_len; + len -= label_len; + } if (len < 2) { // Last label/dot perf_buffer_putuint8(target, 0); diff --git a/src/dnsperf.1.in b/src/dnsperf.1.in index f6def3b..e38aec1 100644 --- a/src/dnsperf.1.in +++ b/src/dnsperf.1.in @@ -20,67 +20,72 @@ dnsperf \- test the performance of a DNS server .SH SYNOPSIS .hy 0 .ad l -\fBdnsperf\fR\ [\fB\-a\ \fIlocal_addr\fB\fR] -[\fB\-b\ \fIbufsize\fB\fR] -[\fB\-c\ \fIclients\fB\fR] -[\fB\-d\ \fIdatafile\fB\fR] +\fBdnsperf\fR\ [\fB\-a\ \fIlocal_addr\fR] +[\fB\-b\ \fIbufsize\fR] +[\fB\-c\ \fIclients\fR] +[\fB\-d\ \fIdatafile\fR] [\fB\-D\fR] [\fB\-e\fR] -[\fB\-E\ \fIcode:secret\fB\fR] -[\fB\-f\ \fIfamily\fB\fR] +[\fB\-E\ \fIcode:secret\fR] +[\fB\-f\ \fIfamily\fR] [\fB\-h\fR] -[\fB\-l\ \fIlimit\fB\fR] -[\fB\-m\ \fImode\fB\fR] -[\fB\-n\ \fIruns_through_file\fB\fR] -[\fB\-p\ \fIport\fB\fR] -[\fB\-q\ \fInum_queries\fB\fR] -[\fB\-Q\ \fImax_qps\fB\fR] -[\fB\-s\ \fIserver_addr\fB\fR] -[\fB\-S\ \fIstats_interval\fB\fR] -[\fB\-t\ \fItimeout\fB\fR] -[\fB\-T\ \fIthreads\fB\fR] +[\fB\-l\ \fIlimit\fR] +[\fB\-m\ \fImode\fR] +[\fB\-n\ \fIruns_through_file\fR] +[\fB\-p\ \fIport\fR] +[\fB\-q\ \fInum_queries\fR] +[\fB\-Q\ \fImax_qps\fR] +[\fB\-s\ \fIserver_addr\fR] +[\fB\-S\ \fIstats_interval\fR] +[\fB\-t\ \fItimeout\fR] +[\fB\-T\ \fIthreads\fR] [\fB\-u\fR] [\fB\-v\fR] -[\fB\-x\ \fIlocal_port\fB\fR] -[\fB\-y\ \fI[alg:]name:secret\fB\fR] +[\fB\-W\fR] +[\fB\-x\ \fIlocal_port\fR] +[\fB\-y\ \fI[alg:]name:secret\fR] .ad .hy .SH DESCRIPTION -\fBdnsperf\fR is a DNS server performance testing tool. It is primarily -intended for measuring the performance of authoritative DNS servers, but it -can also be used for measuring caching server performance in a closed -laboratory environment. For testing caching servers resolving against the -live Internet, the \fBresperf\fR program is preferred. +\fBdnsperf\fR is a DNS server performance testing tool. +It is primarily intended for measuring the performance of authoritative DNS +servers, but it can also be used for measuring caching server performance in +a closed laboratory environment. +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 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 -network, preferably a dedicated Gigabit Ethernet segment. Testing through a -router or firewall is not advisable. +slow down the name server. +The two machines should be connected with a fast network, preferably a +dedicated Gigabit Ethernet segment. +Testing through a router or firewall is not advisable. .SS "Configuring 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 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 -8/9, specify "recursion no;" in the options block). In BIND 8, you should -also specify "fetch-glue no;"; otherwise the server may attempt to retrieve -glue information from the Internet during the test, slowing it down by an -unpredictable factor. +8/9, specify "recursion no;" in the options block). +In BIND 8, you should also specify "fetch-glue no;"; otherwise the server +may attempt to retrieve glue information from the Internet during the test, +slowing it down by an unpredictable factor. .SS "Constructing a query input file" 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 -one line per query, consisting of a domain name and an RR type name -separated by a space. The class of the query is implicitly IN. +queries, on the order of ten thousand to a million. +The input file contains one line per query, consisting of a domain name and +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 zone or TLDs, note that such servers spend most of their time providing -referral responses, not authoritative answers. Therefore, a realistic input -file might consist mostly of queries for type A for names *below*, not at, -the delegations present in the zone. For example, when testing the -performance of a server configured to be authoritative for the top-level -domain "fi.", which contains delegations for domains like "helsinki.fi" and -"turku.fi", the input file could contain lines like +referral responses, not authoritative answers. +Therefore, a realistic input file might consist mostly of queries for type +A for names *below*, not at, the delegations present in the zone. +For example, when testing the performance of a server configured to be +authoritative for the top-level domain "fi.", which contains delegations +for domains like "helsinki.fi" and "turku.fi", the input file could contain +lines like .RS .hy 0 @@ -98,7 +103,8 @@ should be in a random order. .SS "Constructing a dynamic update input file" 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 -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 .hy 0 @@ -108,12 +114,14 @@ example.com .hy .RE -Subsequent lines contain prerequisites, if there are any. Prerequisites can -specify that a name may or may not exist, an rrset may or may not exist, or -an rrset exists and its rdata matches all specified rdata for that name and -type. The keywords "require" and "prohibit" are followed by the appropriate -information. All relative names are considered to be relative to the zone -name. The following lines show the 5 types of prerequisites. +Subsequent lines contain prerequisites, if there are any. +Prerequisites can specify that a name may or may not exist, an rrset may or +may not exist, or an rrset exists and its rdata matches all specified rdata +for that name and type. +The keywords "require" and "prohibit" are followed by the appropriate +information. +All relative names are considered to be relative to the zone name. +The following lines show the 5 types of prerequisites. .RS .hy 0 @@ -128,10 +136,10 @@ prohibit x A .RE 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 -followed by the appropriate information. All relative names are considered -to be relative to the zone name. The following lines show the 4 types of -updates. +to be deleted, or names to be deleted. +The keywords "add" or "delete" are followed by the appropriate information. +All relative names are considered to be relative to the zone name. +The following lines show the 4 types of updates. .RS .hy 0 @@ -155,47 +163,49 @@ send .RE .SS "Running the tests" 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 -mostly self-explanatory. Pay attention to the number of dropped packets -reported - when running the test over a local Ethernet connection, it should -be zero. If one or more packets has been dropped, there may be a problem -with the network connection. In that case, the results should be considered -suspect and the test repeated. +(the \fB\-s\fR option) will normally be specified. +The output of dnsperf is mostly self-explanatory. +Pay attention to the number of dropped packets reported - when running the +test over a local Ethernet connection, it should be zero. +If one or more packets has been dropped, there may be a problem with the +network connection. +In that case, the results should be considered suspect and the test repeated. .SH OPTIONS -\fB-a \fIlocal_addr\fB\fR +\fB-a \fIlocal_addr\fR .br .RS -Specifies the local address from which to send requests. The default is the -wildcard address. +Specifies the local address from which to send requests. +The default is the wildcard address. .RE -\fB-b \fIbufsize\fB\fR +\fB-b \fIbufsize\fR .br .RS -Sets the size of the socket's send and receive buffers, in kilobytes. If not -specified, the operating system's default is used. +Sets the size of the socket's send and receive buffers, in kilobytes. +If not specified, the operating system's default is used. .RE -\fB-c \fIclients\fB\fR +\fB-c \fIclients\fR .br .RS -Act as multiple clients. Requests are sent from multiple sockets. The -default is to act as 1 client. +Act as multiple clients. +Requests are sent from multiple sockets. +The default is to act as 1 client. .RE -\fB-d \fIdatafile\fB\fR +\fB-d \fIdatafile\fR .br .RS -Specifies the input data file. If not specified, \fBdnsperf\fR will read -from standard input. +Specifies the input data file. +If not specified, \fBdnsperf\fR will read from standard input. .RE \fB-D\fR .br .RS -Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent. This also enables -EDNS0, which is required for DNSSEC. +Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent. +This also enables EDNS0, which is required for DNSSEC. .RE \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. .RE -\fB-E \fIcode:value\fB\fR +\fB-E \fIcode:value\fR .br .RS 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 -enables EDNS0. +numeric option code and value expressed as a a hex-encoded string. +This also enables EDNS0. .RE -\fB-f \fIfamily\fB\fR +\fB-f \fIfamily\fR .br .RS -Specifies the address family used for sending DNS packets. The possible -values are "inet", "inet6", or "any". If "any" (the default value) is -specified, \fBdnsperf\fR will use whichever address family is appropriate -for the server it is sending packets to. +Specifies the address family used for sending DNS packets. +The possible values are "inet", "inet6", or "any". +If "any" (the default value) is specified, \fBdnsperf\fR will use whichever +address family is appropriate for the server it is sending packets to. .RE \fB-h\fR @@ -227,111 +237,129 @@ for the server it is sending packets to. Print a usage statement and exit. .RE -\fB-l \fIlimit\fB\fR +\fB-l \fIlimit\fR .br .RS -Specifies a time limit for the run, in seconds. This may cause the input to -be read multiple times, or only some of the input to be read. The default -behavior is to read the input once, and have no specific time limit. +Specifies a time limit for the run, in seconds. +This may cause the input to be read multiple times, or only some of the +input to be read. +The default behavior is to read the input once, and have no specific time +limit. .RE -\fB-n \fIruns_through_file\fB\fR +\fB-n \fIruns_through_file\fR .br .RS -Run through the input file at most this many times. If no time limit is set, -the file will be read exactly this number of times; if a time limit is set, -the file may be read fewer times. +Run through the input file at most this many times. +If no time limit is set, the file will be read exactly this number of +times; if a time limit is set, the file may be read fewer times. .RE -\fB-p \fIport\fB\fR +\fB-p \fIport\fR .br .RS -Sets the port on which the DNS packets are sent. If not specified, the -standard DNS port (udp/tcp 53, dot/tls 853) is used. +Sets the port on which the DNS packets are sent. +If not specified, the standard DNS port (udp/tcp 53, DoT 853) is used. .RE -\fB-q \fInum_queries\fB\fR +\fB-q \fInum_queries\fR .br .RS -Sets the maximum number of outstanding requests. When this value is reached, -\fBdnsperf\fR will not send any more requests until either responses are -received or requests time out. The default value is 100. +Sets the maximum number of outstanding requests. +When this value is reached, \fBdnsperf\fR will not send any more requests +until either responses are received or requests time out. +The default value is 100. .RE -\fB-Q \fImax_qps\fB\fR +\fB-Q \fImax_qps\fR .br .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 -\fB-m \fImode\fB\fR +\fB-m \fImode\fR .br .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". .RE -\fB-s \fIserver_addr\fB\fR +\fB-s \fIserver_addr\fR .br .RS Specifies the name or address of the server to which requests will be sent. The default is the loopback address, 127.0.0.1. .RE -\fB-S \fIstats_interval\fB\fR +\fB-S \fIstats_interval\fR .br .RS 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. .RE -\fB-t \fItimeout\fB\fR +\fB-t \fItimeout\fR .br .RS -Specifies the request timeout value, in seconds. \fBdnsperf\fR will no -longer wait for a response to a particular request after this many seconds -have elapsed. The default is 5 seconds. +Specifies the request timeout value, in seconds. +\fBdnsperf\fR will no longer wait for a response to a particular request +after this many seconds have elapsed. +The default is 5 seconds. .RE -\fB-T \fIthreads\fB\fR +\fB-T \fIthreads\fR .br .RS -Run multiple client threads. By default, \fBdnsperf\fR uses one thread for -sending requests and one thread for receiving responses. If this option is -specified, \fBdnsperf\fR will instead use N pairs of send/receive threads. +Run multiple client threads. +By default, \fBdnsperf\fR uses one thread for sending requests and one +thread for receiving responses. +If this option is specified, \fBdnsperf\fR will instead use N pairs of +send/receive threads. .RE \fB-u\fR .br .RS 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. .RE \fB-v\fR .br .RS -Enables verbose mode. The DNS RCODE of each response will be reported to -standard output when the response is received, as will the latency. If a -query times out, it will be reported with the special string "T" instead of -a normal DNS RCODE. 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. +Enables verbose mode. +The DNS RCODE of each response will be reported to standard output when +the response is received, as will the latency. +If a query times out, it will be reported with the special string "T" +instead of a normal DNS RCODE. +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 -\fB-x \fIlocal_port\fB\fR +\fB-W\fR .br .RS -Specifies the local port from which to send requests. The default is the -wildcard port (0). +Log warnings and errors to standard output instead of standard error making +it easier for script, test and automation to capture all output. +.RE + +\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. +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\fB\fR +\fB-y \fI[alg:]name:secret\fR .br .RS Add a TSIG record [RFC2845] to all packets sent, using the specified TSIG diff --git a/src/dnsperf.c b/src/dnsperf.c index f001370..cf17af0 100644 --- a/src/dnsperf.c +++ b/src/dnsperf.c @@ -51,8 +51,8 @@ #define DEFAULT_SERVER_NAME "127.0.0.1" #define DEFAULT_SERVER_PORT 53 -#define DEFAULT_SERVER_TLS_PORT 853 -#define DEFAULT_SERVER_PORTS "udp/tcp 53 or dot/tls 853" +#define DEFAULT_SERVER_DOT_PORT 853 +#define DEFAULT_SERVER_PORTS "udp/tcp 53 or DoT 853" #define DEFAULT_LOCAL_PORT 0 #define DEFAULT_MAX_OUTSTANDING 100 #define DEFAULT_TIMEOUT 5 @@ -111,6 +111,14 @@ typedef struct { uint64_t latency_sum_squares; uint64_t latency_min; 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; typedef perf_list(struct query_info) query_list; @@ -140,9 +148,9 @@ typedef struct { pthread_mutex_t lock; pthread_cond_t cond; - unsigned int nsocks; - int current_sock; - struct perf_net_socket* socks; + unsigned int nsocks; + int current_sock; + struct perf_net_socket** socks; bool done_sending; uint64_t done_send_time; @@ -194,8 +202,13 @@ print_initial_status(const config_t* config) printf("\n"); perf_sockaddr_format(&config->server_addr, buf, sizeof(buf)); - printf("[Status] Sending %s (to %s)\n", - config->updates ? "updates" : "queries", buf); + if (perf_sockaddr_isinet6(&config->server_addr)) { + 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); 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"); + + 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 @@ -339,6 +373,16 @@ sum_stats(const config_t* config, stats_t* total) total->latency_min = stats->latency_min; if (stats->latency_max > total->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", "address family of DNS transport, inet or inet6", "any", &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", "the server to query", DEFAULT_SERVER_NAME, &server_name); 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); 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) @@ -572,6 +616,7 @@ do_send(void* arg) unsigned int length; int n, i, any_inprogress = 0; perf_result_t result; + bool all_fail; tinfo = (threadinfo_t*)arg; config = tinfo->config; @@ -618,15 +663,17 @@ do_send(void* arg) query_move(tinfo, q, prepend_outstanding); q->timestamp = UINT64_MAX; - i = tinfo->nsocks * 2; + i = tinfo->nsocks * 2; + all_fail = true; 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)) { case 0: if (config->verbose) { perf_log_warning("socket %p not ready", q->sock); } - q->sock = 0; + q->sock = 0; + all_fail = false; continue; case -1: if (errno == EINPROGRESS) { @@ -637,12 +684,19 @@ do_send(void* arg) if (config->verbose) { perf_log_warning("socket %p readiness check timed out", q->sock); } + q->sock = 0; + continue; default: break; } + all_fail = false; break; }; + if (all_fail) { + perf_log_fatal("all sockets reported failure, can not continue"); + } + if (!q->sock) { query_move(tinfo, q, prepend_unused); PERF_UNLOCK(&tinfo->lock); @@ -686,7 +740,7 @@ do_send(void* arg) } 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); if (n < 0) { if (errno == EINPROGRESS) { @@ -719,7 +773,7 @@ do_send(void* arg) while (any_inprogress) { any_inprogress = 0; 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; } } @@ -788,7 +842,7 @@ recv_one(threadinfo_t* tinfo, int which_sock, 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(); if (n < 0) { *saved_errnop = errno; @@ -799,7 +853,7 @@ recv_one(threadinfo_t* tinfo, int which_sock, *saved_errnop = EAGAIN; return false; } - recvd->sock = &tinfo->socks[which_sock]; + recvd->sock = tinfo->socks[which_sock]; recvd->qid = ntohs(packet_header[0]); recvd->rcode = ntohs(packet_header[1]) & 0xF; recvd->size = n; @@ -1045,6 +1099,36 @@ per_thread(uint32_t total, uint32_t nthreads, unsigned int offset) 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 threadinfo_init(threadinfo_t* tinfo, const config_t* config, const times_t* times) @@ -1091,11 +1175,18 @@ threadinfo_init(threadinfo_t* tinfo, const config_t* config, socket_offset = 0; for (i = 0; i < offset; i++) 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, &config->local_addr, socket_offset++, 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; PERF_THREAD(&tinfo->receiver, do_recv, tinfo); @@ -1118,7 +1209,7 @@ threadinfo_cleanup(threadinfo_t* tinfo, times_t* times) if (interrupted) cancel_queries(tinfo); 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) times->end_time = tinfo->last_recv; } @@ -1152,8 +1243,8 @@ int main(int argc, char** argv) perf_os_blocksignal(SIGINT, true); switch (config.mode) { case sock_tcp: - case sock_tls: - // block SIGPIPE for TCP/TLS mode, if connection is closed it will generate a signal + 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: @@ -23,25 +23,28 @@ #include "log.h" #include "opt.h" -#include "os.h" -#include "strerror.h" #include <errno.h> #include <fcntl.h> -#include <stdio.h> -#include <stdlib.h> #include <string.h> -#include <unistd.h> -#include <errno.h> #include <poll.h> -#include <openssl/err.h> #include <netdb.h> #include <arpa/inet.h> -#define TCP_RECV_BUF_SIZE (16 * 1024) -#define TCP_SEND_BUF_SIZE (4 * 1024) +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_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) { @@ -49,10 +52,8 @@ int perf_net_parsefamily(const char* family) return AF_UNSPEC; else if (strcmp(family, "inet") == 0) return AF_INET; -#ifdef AF_INET6 else if (strcmp(family, "inet6") == 0) return AF_INET6; -#endif else { fprintf(stderr, "invalid family %s\n", family); perf_opt_usage(); @@ -82,9 +83,9 @@ in_port_t perf_sockaddr_port(const perf_sockaddr_t* sockaddr) { switch (sockaddr->sa.sa.sa_family) { case AF_INET: - return sockaddr->sa.sin.sin_port; + return ntohs(sockaddr->sa.sin.sin_port); case AF_INET6: - return sockaddr->sa.sin6.sin6_port; + return ntohs(sockaddr->sa.sin6.sin6_port); default: break; } @@ -95,10 +96,10 @@ void perf_sockaddr_setport(perf_sockaddr_t* sockaddr, in_port_t port) { switch (sockaddr->sa.sa.sa_family) { case AF_INET: - sockaddr->sa.sin.sin_port = port; + sockaddr->sa.sin.sin_port = htons(port); break; case AF_INET6: - sockaddr->sa.sin6.sin6_port = port; + sockaddr->sa.sin6.sin6_port = htons(port); break; default: break; @@ -188,76 +189,15 @@ void perf_net_parselocal(int family, const char* name, unsigned int port, 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 ret; - int flags; - struct perf_net_socket sock = { .mode = mode, .is_ready = 1 }; - - family = server->sa.sa.sa_family; + int port; + perf_sockaddr_t tmp; - if (local->sa.sa.sa_family != family) { + if (server->sa.sa.sa_family != local->sa.sa.sa_family) { 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; port = perf_sockaddr_port(&tmp); 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); } - if (bind(sock.fd, &tmp.sa.sa, tmp.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)"); - - 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; - } + switch (mode) { + case sock_udp: + return perf_net_udp_opensocket(server, &tmp, bufsize); case sock_tcp: - if (sock->sending) { - uint16_t dnslen; - ssize_t n; - - 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; + return perf_net_tcp_opensocket(server, &tmp, bufsize); + case sock_dot: + return perf_net_dot_opensocket(server, &tmp, bufsize); default: - break; + perf_log_fatal("perf_net_opensocket(): invalid mode"); } - return -1; + return 0; } @@ -25,6 +25,11 @@ #include <openssl/ssl.h> #include <pthread.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 { union { @@ -42,43 +47,120 @@ enum perf_net_mode { sock_pipe, sock_udp, 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 { - enum perf_net_mode mode; - int fd, have_more, is_ready, flags, is_ssl_ready; - char* recvbuf; - size_t at, sending; - char* sendbuf; - struct sockaddr_storage dest_addr; - socklen_t addrlen; - SSL* ssl; - pthread_mutex_t lock; + void* data; /* user data */ + + enum perf_net_mode mode; + perf_net_recv_t recv; + perf_net_sendto_t sendto; + perf_net_close_t close; + perf_net_sockeq_t sockeq; + perf_net_sockready_t sockready; + perf_net_have_more_t have_more; + + /* + * 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; }; -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); -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_format(const perf_sockaddr_t* sockaddr, char* buf, size_t len); +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); +} -ssize_t perf_net_recv(struct perf_net_socket* sock, void* buf, size_t len, int 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); +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); +} -int perf_net_close(struct perf_net_socket* sock); -int perf_net_sockeq(struct perf_net_socket* sock_a, struct perf_net_socket* 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); -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); +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); +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_format(const perf_sockaddr_t* sockaddr, char* buf, size_t len); -enum perf_net_mode perf_net_parsemode(const char* mode); +static inline int perf_sockaddr_isinet6(const perf_sockaddr_t* sockaddr) +{ + return sockaddr->sa.sa.sa_family == AF_INET6; +} + +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_sockready(struct perf_net_socket* sock, int pipe_fd, int64_t timeout); +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); +struct perf_net_socket* perf_net_dot_opensocket(const perf_sockaddr_t*, const perf_sockaddr_t*, size_t); #endif diff --git a/src/net_dot.c b/src/net_dot.c new file mode 100644 index 0000000..90ceb22 --- /dev/null +++ b/src/net_dot.c @@ -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; +} diff --git a/src/net_tcp.c b/src/net_tcp.c new file mode 100644 index 0000000..94d2a6a --- /dev/null +++ b/src/net_tcp.c @@ -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; +} diff --git a/src/net_udp.c b/src/net_udp.c new file mode 100644 index 0000000..03e0ae1 --- /dev/null +++ b/src/net_udp.c @@ -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; +} @@ -214,6 +214,10 @@ void perf_opt_parse(int argc, char** argv) *opt->u.uintp = parse_uint(opt->desc, optarg, 1, 0xFFFFFFFF); break; + case perf_opt_zpint: + *opt->u.uintp = parse_uint(opt->desc, optarg, + 0, 0xFFFFFFFF); + break; case perf_opt_timeval: *opt->u.uint64p = parse_timeval(opt->desc, optarg); break; @@ -23,7 +23,8 @@ typedef enum { perf_opt_string, perf_opt_boolean, - perf_opt_uint, + perf_opt_uint, // can not be zero + perf_opt_zpint, // zero or positive perf_opt_timeval, perf_opt_double, perf_opt_port, @@ -61,11 +61,12 @@ void perf_os_handlesignal(int sig, void (*handler)(int)) perf_result_t 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_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) { struct pollfd fds[nfds + 1]; @@ -73,10 +74,10 @@ perf_os_waituntilanyreadable(struct perf_net_socket* socks, unsigned int nfds, i int to, n; for (i = 0; i < nfds; i++) { - if (socks[i].have_more) + if (perf_net_have_more(socks[i])) return (PERF_R_SUCCESS); - fds[i].fd = socks[i].fd; + fds[i].fd = socks[i]->fd; fds[i].events = POLLIN; } @@ -109,7 +110,7 @@ perf_os_waituntilanyreadable(struct perf_net_socket* socks, unsigned int nfds, i } 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) { struct pollfd fds[nfds + 1]; @@ -117,7 +118,7 @@ perf_os_waituntilanywritable(struct perf_net_socket* socks, unsigned int nfds, i int to, n; for (i = 0; i < nfds; i++) { - fds[i].fd = socks[i].fd; + fds[i].fd = socks[i]->fd; fds[i].events = POLLOUT; } @@ -34,11 +34,11 @@ perf_result_t perf_os_waituntilreadable(struct perf_net_socket* sock, int pipe_fd, int64_t timeout); 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); 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); #endif diff --git a/src/resperf.1.in b/src/resperf.1.in index 463a5ac..ea077f2 100644 --- a/src/resperf.1.in +++ b/src/resperf.1.in @@ -20,85 +20,96 @@ resperf \- test the resolution performance of a caching DNS server .SH SYNOPSIS .hy 0 .ad l -\fBresperf\-report\fR\ [\fB\-a\ \fIlocal_addr\fB\fR] -[\fB\-d\ \fIdatafile\fB\fR] -[\fB\-M\ \fImode\fB\fR] -[\fB\-s\ \fIserver_addr\fB\fR] -[\fB\-p\ \fIport\fB\fR] -[\fB\-x\ \fIlocal_port\fB\fR] -[\fB\-t\ \fItimeout\fB\fR] -[\fB\-b\ \fIbufsize\fB\fR] -[\fB\-f\ \fIfamily\fB\fR] +\fBresperf\-report\fR\ [\fB\-a\ \fIlocal_addr\fR] +[\fB\-d\ \fIdatafile\fR] +[\fB\-R\fR] +[\fB\-M\ \fImode\fR] +[\fB\-s\ \fIserver_addr\fR] +[\fB\-p\ \fIport\fR] +[\fB\-x\ \fIlocal_port\fR] +[\fB\-t\ \fItimeout\fR] +[\fB\-b\ \fIbufsize\fR] +[\fB\-f\ \fIfamily\fR] [\fB\-e\fR] [\fB\-D\fR] -[\fB\-y\ \fI[alg:]name:secret\fB\fR] +[\fB\-y\ \fI[alg:]name:secret\fR] [\fB\-h\fR] -[\fB\-i\ \fIinterval\fB\fR] -[\fB\-m\ \fImax_qps\fB\fR] -[\fB\-r\ \fIrampup_time\fB\fR] -[\fB\-c\ \fIconstant_traffic_time\fB\fR] -[\fB\-L\ \fImax_loss\fB\fR] -[\fB\-C\ \fIclients\fB\fR] -[\fB\-q\ \fImax_outstanding\fB\fR] +[\fB\-i\ \fIinterval\fR] +[\fB\-m\ \fImax_qps\fR] +[\fB\-r\ \fIrampup_time\fR] +[\fB\-c\ \fIconstant_traffic_time\fR] +[\fB\-L\ \fImax_loss\fR] +[\fB\-C\ \fIclients\fR] +[\fB\-q\ \fImax_outstanding\fR] +[\fB\-F\ \fIfall_behind\fR] [\fB\-v\fR] +[\fB\-W\fR] .ad .hy .hy 0 .ad l -\fBresperf\fR\ [\fB\-a\ \fIlocal_addr\fB\fR] -[\fB\-d\ \fIdatafile\fB\fR] -[\fB\-M\ \fImode\fB\fR] -[\fB\-s\ \fIserver_addr\fB\fR] -[\fB\-p\ \fIport\fB\fR] -[\fB\-x\ \fIlocal_port\fB\fR] -[\fB\-t\ \fItimeout\fB\fR] -[\fB\-b\ \fIbufsize\fB\fR] -[\fB\-f\ \fIfamily\fB\fR] +\fBresperf\fR\ [\fB\-a\ \fIlocal_addr\fR] +[\fB\-d\ \fIdatafile\fR] +[\fB\-R\fR] +[\fB\-M\ \fImode\fR] +[\fB\-s\ \fIserver_addr\fR] +[\fB\-p\ \fIport\fR] +[\fB\-x\ \fIlocal_port\fR] +[\fB\-t\ \fItimeout\fR] +[\fB\-b\ \fIbufsize\fR] +[\fB\-f\ \fIfamily\fR] [\fB\-e\fR] [\fB\-D\fR] -[\fB\-y\ \fI[alg:]name:secret\fB\fR] +[\fB\-y\ \fI[alg:]name:secret\fR] [\fB\-h\fR] -[\fB\-i\ \fIinterval\fB\fR] -[\fB\-m\ \fImax_qps\fB\fR] -[\fB\-P\ \fIplot_data_file\fB\fR] -[\fB\-r\ \fIrampup_time\fB\fR] -[\fB\-c\ \fIconstant_traffic_time\fB\fR] -[\fB\-L\ \fImax_loss\fB\fR] -[\fB\-C\ \fIclients\fB\fR] -[\fB\-q\ \fImax_outstanding\fB\fR] +[\fB\-i\ \fIinterval\fR] +[\fB\-m\ \fImax_qps\fR] +[\fB\-P\ \fIplot_data_file\fR] +[\fB\-r\ \fIrampup_time\fR] +[\fB\-c\ \fIconstant_traffic_time\fR] +[\fB\-L\ \fImax_loss\fR] +[\fB\-C\ \fIclients\fR] +[\fB\-q\ \fImax_outstanding\fR] +[\fB\-F\ \fIfall_behind\fR] [\fB\-v\fR] +[\fB\-W\fR] .ad .hy .SH DESCRIPTION -\fBresperf\fR is a companion tool to \fBdnsperf\fR. \fBdnsperf\fR was -primarily designed for benchmarking authoritative servers, and it does not -work well with caching servers that are talking to the live Internet. One -reason for this is that dnsperf uses a "self-pacing" approach, which is +\fBresperf\fR is a companion tool to \fBdnsperf\fR. +\fBdnsperf\fR was primarily designed for benchmarking authoritative +servers, and it does not work well with caching servers that are talking +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 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 -works well for authoritative servers that process queries in order and one -at a time; it also works pretty well for a caching server in a closed -laboratory environment talking to a simulated Internet that's all on the -same LAN. Unfortunately, it does not work well with a caching server talking +and then send a new query whenever you get a response back. +This approach works well for authoritative servers that process queries in +order and one at a time; it also works pretty well for a caching server in +a closed laboratory environment talking to a simulated Internet that's all +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 -parallel to achieve its maximum throughput. There have been numerous -attempts to use dnsperf (or its predecessor, queryperf) for benchmarking -live caching servers, usually with poor results. Therefore, a separate tool -designed specifically for caching servers is needed. +parallel to achieve its maximum throughput. +There have been numerous attempts to use dnsperf (or its predecessor, +queryperf) for benchmarking live caching servers, usually with poor results. +Therefore, a separate tool designed specifically for caching servers is +needed. .SS "How resperf works" -Unlike the "self-pacing" approach of dnsperf, resperf works by sending DNS -queries at a controlled, steadily increasing rate. By default, resperf will -send traffic for 60 seconds, linearly increasing the amount of traffic from -zero to 100,000 queries per second. - -During the test, resperf listens for responses from the server and keeps -track of response rates, failure rates, and latencies. It will also continue -listening for responses for an additional 40 seconds after it has stopped -sending traffic, so that there is time for the server to respond to the last -queries sent. This time period was chosen to be longer than the overall -query timeout of both Nominum CacheServe and current versions of BIND. +Unlike the "self-pacing" approach of dnsperf, \fBresperf\fR works by sending +DNS queries at a controlled, steadily increasing rate. +By default, \fBresperf\fR will send traffic for 60 seconds, linearly +increasing the amount of traffic from zero to 100,000 queries per second (or +\fImax_qps\fR). + +During the test, \fBresperf\fR listens for responses from the server and +keeps track of response rates, failure rates, and latencies. +It will also continue listening for responses for an additional 40 seconds +after it has stopped sending traffic, so that there is time for the server +to respond to the last queries sent. +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 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 rate, failure response rate, and average query latency as functions of time. .SS "What you will need" -Benchmarking a live caching server is serious business. A fast caching -server like Nominum CacheServe, resolving a mix of cacheable 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 -send more than 40,000 queries per second to authoritative servers on the -Internet, and receive responses to most of them. Assuming an average request -size of 50 bytes and a response size of 150 bytes, this amounts to some 1216 -Mbps of outgoing and 448 Mbps of incoming traffic. If your Internet -connection can't handle the bandwidth, you will end up measuring the speed -of the connection, not the server, and may saturate the connection causing a -degradation in service for other users. +Benchmarking a live caching server is serious business. +A fast caching server like Nominum CacheServe, resolving a mix of cacheable +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 send more than 40,000 queries per second to +authoritative servers on the Internet, and receive responses to most of them. +Assuming an average request size of 50 bytes and a response size of 150 +bytes, this amounts to some 1216 Mbps of outgoing and 448 Mbps of incoming +traffic. +If your Internet connection can't handle the bandwidth, you will end up +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, 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 -will even lock up or crash. +generate and will end up dropping packets, skewing the test results. +Some will even lock up or crash. -You should run resperf on a machine separate from the server under test, on -the same LAN. Preferably, this should be a Gigabit Ethernet network. The -machine running resperf should be at least as fast as the machine being -tested; otherwise, it may end up being the bottleneck. +You should run \fBresperf\fR on a machine separate from the server under test, +on the same LAN. +Preferably, this should be a Gigabit Ethernet network. +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 -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 -100,000 queries per second, if resperf gets delayed by just 1/100 of a -second, 1000 incoming UDP packets will arrive in the meantime. This is more -than most operating systems will buffer, which means packets will be -dropped. +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 than most operating systems will buffer, which means packets +will be dropped. Because the granularity of the timers provided by operating systems is typically too coarse to accurately schedule packet transmissions at -sub-millisecond intervals, resperf will busy-wait between packet -transmissions, constantly polling for responses in the meantime. Therefore, -it is normal for resperf to consume 100% CPU during the whole test run, even -during periods where query rates are relatively low. +sub-millisecond intervals, \fBresperf\fR will busy-wait between packet +transmissions, constantly polling for responses in the meantime. +Therefore, it is normal for \fBresperf\fR to consume 100% CPU during the +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. 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 -derived from recorded production client DNS traffic, without removing -duplicate queries or other filtering. With the default settings, resperf -will use up to 3 million queries in each test run. +query file. +To make the test as realistic as possible, the queries should be derived +from recorded production client DNS traffic, without removing duplicate +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 simultaneous resolutions, like the \fBmax\-recursive\-clients\fR statement 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 -of 10000 for Nominum CacheServe and 100000 for BIND 9. Should the limit be -reached, it will show up in the plots as an increase in the number of -failure responses. +will probably have to increase it. +As a starting point, we recommend a value of 10000 for Nominum CacheServe +and 100000 for BIND 9. +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 -make sure it is starting with an empty cache. If the cache already contains -data from a previous test run that used the same set of queries, almost all -queries will be answered from the cache, yielding inflated performance -numbers. +make sure it is starting with an empty cache. +If the cache already contains data from a previous test run that used the +same set of queries, almost all queries will be answered from the cache, +yielding inflated performance numbers. To use the \fBresperf\-report\fR script, you need to have \fBgnuplot\fR -installed. Make sure your installed version of \fBgnuplot\fR supports the -png terminal driver. 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. +installed. +Make sure your installed version of \fBgnuplot\fR supports the png terminal +driver. +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" -Resperf is typically invoked via the \fBresperf\-report\fR script, which -will run \fBresperf\fR with its output redirected to a file and then -automatically generate an illustrated report in HTML format. Command line -arguments given to resperf-report will be passed on unchanged to resperf. +\fBresperf\fR is typically invoked via the \fBresperf\-report\fR script, +which will run \fBresperf\fR with its output redirected to a file and then +automatically generate an illustrated report in HTML format. +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 -address and the query data file. A typical invocation will look like +When running \fBresperf\-report\fR, you will need to specify at least the +server IP address and the query data file. +A typical invocation will look like .RS .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 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 -be precise, resperf can transition from the traffic-sending phase to the -waiting-for-responses phase in three different ways: +but in practice, the 60-second traffic phase will usually be cut short. +To be precise, \fBresperf\fR can transition from the traffic-sending phase +to the waiting-for-responses phase in three different ways: .IP \(bu 2 Running for the full allotted time and successfully reaching the maximum -query rate (by default, 60 seconds and 100,000 qps, respectively). Since -this is a very high query rate, this will rarely happen (with today's +query rate (by default, 60 seconds and 100,000 qps, respectively). +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 first. .IP \(bu 2 -Exceeding 65,536 outstanding queries. This often happens as a result of -(successfully) exceeding the capacity of the server being tested, causing -the excess queries to be dropped. The limit of 65,536 queries comes from the -number of possible values for the ID field in the DNS packet. Resperf 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. +Exceeding 65,536 outstanding queries. +This often happens as a result of (successfully) exceeding the capacity of +the server being tested, causing the excess queries to be dropped. +The limit of 65,536 queries comes from the number of possible values for +the ID field in the DNS packet. +\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 -When resperf finds itself unable to send queries fast enough. Resperf will -notice if it is falling behind in its scheduled query transmissions, and if -this backlog reaches 1000 queries, it will print a message like "Fell behind -by 1000 queries" (or whatever the actual number is at the time) and stop -sending traffic. +When \fBresperf\fR finds itself unable to send queries fast enough. +\fBresperf\fR will notice if it is falling behind in its scheduled query +transmissions, and if this backlog reaches 1000 queries, it will print +a message like "Fell behind by 1000 queries" (or whatever the actual number +is at the time) and stop sending traffic. .PP 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 -server's response rate is flattening out toward the end of the test. If it -is not, then you are not loading the server enough. If you are getting the -"Fell behind" message, make sure that the machine running resperf is fast -enough and has no other applications running. +server's response rate is flattening out toward the end of the test. +If it is not, then you are not loading the server enough. +If you are getting the "Fell behind" message, make sure that the machine +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 -reach close to 100% CPU at the point of maximum traffic; if it does not, you -most likely have a bottleneck in some other part of your test setup, for -example, your external Internet connection. +You should also monitor the CPU usage of the server under test. +It should reach close to 100% CPU at the point of maximum traffic; if it does +not, you most likely have a bottleneck in some other part of your test setup, +for example, your external Internet connection. The report generated by \fBresperf\-report\fR will be stored with a unique file name based on the current date and time, e.g., -\fI20060812-1550.html\fR. The PNG images of the plots and other auxiliary -files will be stored in separate files beginning with the same date-time -string. To view the report, simply open the \fI.html\fR file in a web -browser. +\fI20060812-1550.html\fR. +The PNG images of the plots and other auxiliary files will be stored in +separate files beginning with the same date-time string. +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 to copy the .png files along with the .html file (or simply copy all the files, e.g., using scp 20060812-1550.* host:directory/). .SS "Interpreting the report" The \fI.html\fR file produced by \fBresperf\-report\fR consists of two -sections. The first section, "Resperf output", contains output from the -\fBresperf\fR program such as progress messages, a summary of the command -line arguments, and summary statistics. 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 -sent per second" graph shows the amount of traffic being sent to the server; -this should be very close to a straight diagonal line, reflecting the linear -ramp-up of traffic. +sections. +The first section, "Resperf output", contains output from the \fBresperf\fR +program such as progress messages, a summary of the command line arguments, +and summary statistics. +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 sent per second" graph shows the amount of traffic being sent to +the server; this should be very close to a straight diagonal line, reflecting +the linear ramp-up of traffic. The "Total responses received per second" graph shows how many of the -queries received a response from the server. All responses are counted, -whether successful (NOERROR or NXDOMAIN) or not (e.g., SERVFAIL). +queries received a response from the server. +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 -queries received a failure response. A response is considered to be a -failure if its RCODE is neither NOERROR nor NXDOMAIN. +queries received a failure response. +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 -behaves under increasing load. The "Total responses received per second" -graph will initially closely follow the "Queries sent per second" graph -(often rendering it invisible in the plot as the two graphs are plotted on -top of one another), but when the load exceeds the server's capacity, the -"Total responses received per second" graph may diverge from the "Queries -sent per second" graph and flatten out, indicating that some of the queries -are being dropped. +behaves under increasing load. +The "Total responses received per second" graph will initially closely +follow the "Queries sent per second" graph (often rendering it invisible in +the plot as the two graphs are plotted on top of one another), but when the +load exceeds the server's capacity, the "Total responses received per second" +graph may diverge from the "Queries sent per second" graph and flatten out, +indicating that some of the queries are being dropped. The "Failure responses received per second" graph will normally show a roughly linear ramp close to the bottom of the plot with some random fluctuation, since typical query traffic will contain some small percentage -of failing queries randomly interspersed with the successful ones. As the -total traffic increases, the number of failures will increase +of failing queries randomly interspersed with the successful ones. +As the total traffic increases, the number of failures will increase proportionally. If the "Failure responses received per second" graph turns sharply upwards, 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 -SERVFAIL responses rather than by dropping queries. Since Nominum CacheServe -and BIND 9 will both respond with SERVFAIL when they exceed their -\fBmax\-recursive\-clients\fR or \fBrecursive\-clients\fR 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 -shows how the latency varies during the course of the test. Typically, the -latency graph will exhibit a downwards trend because the cache hit rate -improves as ever more responses are cached during the test, and the latency -for a cache hit is much smaller than for a cache miss. The latency graph is -provided as an aid in determining the point where the server gets -overloaded, which can be seen as a sharp upwards turn in the graph. The -latency graph is not intended for making absolute latency measurements or -comparisons between servers; the latencies shown in the graph are not +capacity. +This will happen if the server reacts to overload by sending SERVFAIL +responses rather than by dropping queries. +Since Nominum CacheServe and BIND 9 will both respond with SERVFAIL when +they exceed their \fBmax\-recursive\-clients\fR or \fBrecursive\-clients\fR +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 shows how the latency varies during the course of the test. +Typically, the latency graph will exhibit a downwards trend because the +cache hit rate improves as ever more responses are cached during the test, +and the latency for a cache hit is much smaller than for a cache miss. +The latency graph is provided as an aid in determining the point where the +server gets overloaded, which can be seen as a sharp upwards turn in the +graph. +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 the deliberate overloading of the server towards the end of the test. 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 -when the response (if any) was received. This makes it it easy to compare -the query and response rates; for example, if no queries are dropped, the -query and response graphs will be identical. As another example, if the plot -shows 10% failure responses at t=5 seconds, this means that 10% of the -queries sent at t=5 seconds eventually failed, not that 10% of the responses -received at t=5 seconds were failures. +when the response (if any) was received. +This makes it it easy to compare the query and response rates; for example, +if no queries are dropped, the query and response graphs will be identical. +As another example, if the plot shows 10% failure responses at t=5 seconds, +this means that 10% of the queries sent at t=5 seconds eventually failed, +not that 10% of the responses received at t=5 seconds were failures. .SS "Determining the server's maximum throughput" 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 -capable of handling. This is not always an easy task, because as a server is -driven into overload, the service it provides may deteriorate gradually, and -this deterioration can manifest itself either as queries being dropped, as -an increase in the number of SERVFAIL responses, or an increase in latency. +capable of handling. +This is not always an easy task, because as a server is driven into overload, +the service it provides may deteriorate gradually, and this deterioration +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 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 @@ -315,22 +344,24 @@ terms of packet drop percentage, SERVFAIL percentage, and latency. The summary statistics in the "Resperf output" section of the report contains a "Maximum throughput" value which by default is determined from 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 -method of throughput measurement has the advantage of simplicity, but it may -or may not be appropriate for your needs; the reported value should always -be validated by a visual inspection of the graphs to ensure that service has -not already deteriorated unacceptably before the maximum response rate is -reached. It may also be helpful to look at the "Lost at that point" value in +regard to the number of queries being dropped or failing at that point. +This method of throughput measurement has the advantage of simplicity, but +it may or may not be appropriate for your needs; the reported value should +always be validated by a visual inspection of the graphs to ensure that +service has not already deteriorated unacceptably before the maximum response +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 was being dropped at the point in the test when the maximum throughput was reached. -Alternatively, you can make resperf report the throughput at the point in -the test where the percentage of queries dropped exceeds a given limit (or -the maximum as above if the limit is never exceeded). This can be a more -realistic indication of how much the server can be loaded while still -providing an acceptable level of service. This is done using the \fB\-L\fR -command line option; for example, specifying \fB\-L 10\fR makes resperf +Alternatively, you can make \fBresperf\fR report the throughput at the point +in the test where the percentage of queries dropped exceeds a given limit +(or the maximum as above if the limit is never exceeded). +This can be a more realistic indication of how much the server can be loaded +while still providing an acceptable level of service. +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 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 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 -conditions. Therefore, the plots should be manually inspected to ensure that -there is not an abnormal number of failures. +conditions. +Therefore, the plots should be manually inspected to ensure that there is not +an abnormal number of failures. .SH "GENERATING CONSTANT TRAFFIC" 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 -using \fBresperf\fR for tasks other than performance measurement; for -example, it can be used to "soak test" a server by subjecting it to a -sustained load for an extended period of time. +capability to send a constant stream of traffic. +This can be useful when using \fBresperf\fR for tasks other than performance +measurement; for example, it can be used to "soak test" a server by +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, 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 -\fB\-m 10000 \-c 3600\fR. This will include the usual 30-second gradual -ramp-up of traffic at the 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. +query rate. +For example, to send 10000 queries per second for an hour, use \fB\-m 10000 +\-c 3600\fR. +This will include the usual 30-second gradual ramp-up of traffic at the +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 \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 \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 -duration of zero seconds with \fB\-r 0\fR and \fB\-c 0\fR, respectively. The -latter is the default. +seconds. +Either the ramp-up or the plateau can be suppressed by supplying a duration +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 -large amounts of input data. Also, a long-running test will generate a large -amount of plot data, which is kept in memory for the duration of the test. +large amounts of input data. +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 increasing the interval between measurements from the default of 0.5 seconds using the \fB\-i\fR option in long-running tests. 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 -itself and the server under test can sustain. Otherwise, the test is likely -to be cut short as a result of either running out of query IDs (because of -large numbers of dropped queries) or of resperf falling behind its -transmission schedule. +itself and the server under test can sustain. +Otherwise, the test is likely to be cut short as a result of either running +out of query IDs (because of large numbers of dropped queries) or of +\fBresperf\fR falling behind its transmission schedule. .SH 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 @@ -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 not itself take a \fB\-P\fR option. -\fB-d \fIdatafile\fB\fR +\fB-d \fIdatafile\fR +.br +.RS +Specifies the input data file. +If not specified, \fBresperf\fR will read from standard input. +.RE + +\fB-R\fR .br .RS -Specifies the input data file. If not specified, \fBresperf\fR will read -from standard input. +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 -\fB-M \fImode\fB\fR +\fB-M \fImode\fR .br .RS -Specifies the transport mode to use, "udp", "tcp" or "tls". Default is "udp". +Specifies the transport mode to use, "udp", "tcp" or "dot". +Default is "udp". .RE -\fB-s \fIserver_addr\fB\fR +\fB-s \fIserver_addr\fR .br .RS Specifies the name or address of the server to which requests will be sent. The default is the loopback address, 127.0.0.1. .RE -\fB-p \fIport\fB\fR +\fB-p \fIport\fR .br .RS -Sets the port on which the DNS packets are sent. If not specified, the -standard DNS port (udp/tcp 53, tls 853) is used. +Sets the port on which the DNS packets are sent. +If not specified, the standard DNS port (udp/tcp 53, DoT 853) is used. .RE -\fB-a \fIlocal_addr\fB\fR +\fB-a \fIlocal_addr\fR .br .RS -Specifies the local address from which to send requests. The default is the -wildcard address. +Specifies the local address from which to send requests. +The default is the wildcard address. .RE -\fB-x \fIlocal_port\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). +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. +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-t \fItimeout\fB\fR +\fB-t \fItimeout\fR .br .RS -Specifies the request timeout value, in seconds. \fBresperf\fR will no -longer wait for a response to a particular request after this many seconds -have elapsed. The default is 45 seconds. +Specifies the request timeout value, in seconds. +\fBresperf\fR will no longer wait for a response to a particular request +after this many seconds have elapsed. +The default is 45 seconds. \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 @@ -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. 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 -corresponding default in \fBdnsperf\fR, because caching servers can take -many orders of magnitude longer to answer a query than authoritative servers +timeout of current caching servers. +Note that this is longer than the corresponding default in \fBdnsperf\fR, +because caching servers can take many orders of magnitude longer to answer +a query than authoritative servers do. 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. .RE -\fB-b \fIbufsize\fB\fR +\fB-b \fIbufsize\fR .br .RS -Sets the size of the socket's send and receive buffers, in kilobytes. If not -specified, the operating system's default is used. +Sets the size of the socket's send and receive buffers, in kilobytes. +If not specified, the operating system's default is used. .RE -\fB-f \fIfamily\fB\fR +\fB-f \fIfamily\fR .br .RS -Specifies the address family used for sending DNS packets. The possible -values are "inet", "inet6", or "any". If "any" (the default value) is -specified, \fBresperf\fR will use whichever address family is appropriate -for the server it is sending packets to. +Specifies the address family used for sending DNS packets. +The possible values are "inet", "inet6", or "any". +If "any" (the default value) is specified, \fBresperf\fR will use whichever +address family is appropriate for the server it is sending packets to. .RE \fB-e\fR @@ -478,11 +526,11 @@ Enables EDNS0 [RFC2671], by adding an OPT record to all packets sent. \fB-D\fR .br .RS -Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent. This also enables -EDNS0, which is required for DNSSEC. +Sets the DO (DNSSEC OK) bit [RFC3225] in all packets sent. +This also enables EDNS0, which is required for DNSSEC. .RE -\fB-y \fI[alg:]name:secret\fB\fR +\fB-y \fI[alg:]name:secret\fR .br .RS 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. .RE -\fB-i \fIinterval\fB\fR +\fB-i \fIinterval\fR .br .RS -Specifies the time interval between data points in the plot file. The -default is 0.5 seconds. +Specifies the time interval between data points in the plot file. +The default is 0.5 seconds. .RE -\fB-m \fImax_qps\fB\fR +\fB-m \fImax_qps\fR .br .RS -Specifies the target maximum query rate (in queries per second). This should -be higher than the expected maximum throughput of the server being tested. +Specifies the target maximum query rate (in queries per second). +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 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 -\fB-P \fIplot_data_file\fB\fR +\fB-P \fIplot_data_file\fR .br .RS -Specifies the name of the plot data file. The default is -\fIresperf.gnuplot\fR. +Specifies the name of the plot data file. +The default is \fIresperf.gnuplot\fR. .RE -\fB-r \fIrampup_time\fB\fR +\fB-r \fIrampup_time\fR .br .RS -Specifies the length of time over which traffic will be ramped up. The -default is 60 seconds. +Specifies the length of time over which traffic will be ramped up. +The default is 60 seconds. .RE -\fB-c \fIconstant_traffic_time\fB\fR +\fB-c \fIconstant_traffic_time\fR .br .RS 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 -sending of traffic at a constant rate will be done. +rate following the initial ramp-up. +The default is 0 seconds, meaning no sending of traffic at a constant rate +will be done. .RE -\fB-L \fImax_loss\fB\fR +\fB-L \fImax_loss\fR .br .RS Specifies the maximum acceptable query loss percentage for purposes of -determining the maximum throughput value. The default is 100%, meaning that -\fBresperf\fR will measure the maximum throughput without regard to query +determining the maximum throughput value. +The default is 100%, meaning that \fBresperf\fR will measure the maximum +throughput without regard to query loss. .RE -\fB-C \fIclients\fB\fR +\fB-C \fIclients\fR +.br +.RS +Act as multiple clients. +Requests are sent from multiple sockets. +The default is to act as 1 client. +.RE + +\fB-q \fImax_outstanding\fR .br .RS -Act as multiple clients. Requests are sent from multiple sockets. The -default is to act as 1 client. +Sets the maximum number of outstanding requests. +\fBresperf\fR will stop ramping up traffic when this many queries are +outstanding. +The default is 64k, and the limit is 64k per client. .RE -\fB-q \fImax_outstanding\fB\fR +\fB-F \fIfall_behind\fR .br .RS -Sets the maximum number of outstanding requests. \fBresperf\fR will stop -ramping up traffic when this many queries are outstanding. The default is -64k, and the limit is 64k per client. +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 \fB-v\fR @@ -564,22 +627,31 @@ ramping up traffic when this many queries are outstanding. The default is .RS Enables verbose mode to report about network readiness and congestion. .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" 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 -\fBresperf\-report\fR script, there is no need for the user to deal with -this file directly, but its format and contents are documented here for -completeness and in case you wish to run \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 -recognized as a comment by its leading hash sign (#). - -Subsequent lines contain the actual plot data. For purposes of generating -the plot data file, the test run is divided into time intervals of 0.5 -seconds (or some other length of time specified with the \fB\-i\fR command -line option). Each line corresponds to one such interval, and contains the -following values as floating-point numbers: +data to be plotted using \fBgnuplot\fR. +When running \fBresperf\fR via the \fBresperf\-report\fR script, there is +no need for the user to deal with this file directly, but its format and +contents are documented here for completeness and in case you wish to run +\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 recognized as a comment by its leading hash sign (#). + +Subsequent lines contain the actual plot data. +For purposes of generating the plot data file, the test run is divided into +time intervals of 0.5 seconds (or some other length of time specified with +the \fB\-i\fR command line option). +Each line corresponds to one such interval, and contains the following values +as floating-point numbers: \fBTime\fR .br @@ -621,6 +693,23 @@ length of the interval The average time between sending the query and receiving a response, for queries sent in this time interval .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" \fBdnsperf\fR(1) .SH AUTHOR diff --git a/src/resperf.c b/src/resperf.c index 97c998d..ed0ccc5 100644 --- a/src/resperf.c +++ b/src/resperf.c @@ -52,12 +52,13 @@ #define DEFAULT_SERVER_NAME "127.0.0.1" #define DEFAULT_SERVER_PORT 53 -#define DEFAULT_SERVER_TLS_PORT 853 -#define DEFAULT_SERVER_PORTS "udp/tcp 53 or dot/tls 853" +#define DEFAULT_SERVER_DOT_PORT 853 +#define DEFAULT_SERVER_PORTS "udp/tcp 53 or DoT 853" #define DEFAULT_LOCAL_PORT 0 #define DEFAULT_SOCKET_BUFFER 32 #define DEFAULT_TIMEOUT 45 #define DEFAULT_MAX_OUTSTANDING (64 * 1024) +#define DEFAULT_MAX_FALL_BEHIND 1000 #define MAX_INPUT_DATA (64 * 1024) @@ -72,6 +73,8 @@ typedef perf_list(struct query_info) query_list; typedef struct query_info { uint64_t sent_timestamp; + bool is_inprogress; + /* * This link links the query into the list of outstanding * queries or the list of available query IDs. @@ -88,11 +91,11 @@ static query_list instanding_list; static query_info* queries; -static perf_sockaddr_t server_addr; -static perf_sockaddr_t local_addr; -static unsigned int nsocks; -static struct perf_net_socket* socks; -static enum perf_net_mode mode; +static perf_sockaddr_t server_addr; +static perf_sockaddr_t local_addr; +static unsigned int nsocks; +static struct perf_net_socket** socks; +static enum perf_net_mode mode; static int dummypipe[2]; @@ -143,6 +146,7 @@ static uint64_t num_queries_outstanding; static uint64_t num_responses_received; static uint64_t num_queries_timed_out; static uint64_t rcodecounts[16]; +static uint64_t num_reconnections; static uint64_t time_now; static uint64_t time_of_program_start; @@ -164,6 +168,9 @@ typedef struct { int responses; int failures; double latency_sum; + + int connections; + double conn_latency_sum; } ramp_bucket; /* Pointer to array of n_buckets ramp_bucket structures */ @@ -192,7 +199,8 @@ static uint64_t sustain_phase_began, wait_phase_began; static perf_tsigkey_t* tsigkey; -static bool verbose; +static bool verbose; +static unsigned int max_fall_behind; const char* progname = "resperf"; @@ -205,6 +213,9 @@ stringify(double value, int precision) 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 setup(int argc, char** argv) { @@ -232,11 +243,12 @@ setup(int argc, char** argv) nsocks = 1; mode = sock_udp; verbose = false; + max_fall_behind = DEFAULT_MAX_FALL_BEHIND; perf_opt_add('f', perf_opt_string, "family", "address family of DNS transport, inet or inet6", "any", &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", "the server to query", DEFAULT_SERVER_NAME, &server_name); perf_opt_add('p', perf_opt_port, "port", @@ -280,7 +292,7 @@ setup(int argc, char** argv) "the maximum acceptable query loss, in percent", stringify(max_loss_percent, 0), &max_loss_percent); 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", "the maximum number of queries outstanding", stringify(DEFAULT_MAX_OUTSTANDING, 0), &max_outstanding); @@ -289,6 +301,10 @@ setup(int argc, char** argv) NULL, &verbose); bool log_stdout = false; 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); @@ -300,7 +316,7 @@ setup(int argc, char** argv) mode = perf_net_parsemode(_mode); 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) @@ -330,6 +346,9 @@ setup(int argc, char** argv) local_port, &local_addr); input = perf_datafile_open(filename); + if (reopen_datafile) { + perf_datafile_setmaxruns(input, -1); + } if (dnssec) edns = true; @@ -340,8 +359,15 @@ setup(int argc, char** argv) if (!(socks = calloc(nsocks, sizeof(*socks)))) { 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); + 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 @@ -351,7 +377,7 @@ cleanup(void) perf_datafile_close(&input); for (i = 0; i < nsocks; i++) - (void)perf_net_close(&socks[i]); + (void)perf_net_close(socks[i]); close(dummypipe[0]); close(dummypipe[1]); } @@ -374,6 +400,30 @@ find_bucket(uint64_t when) 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 out statistics based on the results of the test @@ -409,6 +459,7 @@ print_statistics(void) (rcodecounts[i] * 100.0) / num_responses_received); } printf("\n"); + printf(" Reconnection(s): %" PRIu64 "\n", num_reconnections); printf(" Run time (s): %u.%06u\n", (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; sock = (q - queries) % nsocks; - if (socks[sock].sending) { - if (perf_net_sockready(&socks[sock], dummypipe[0], TIMEOUT_CHECK_TIME) == -1) { + while (q->is_inprogress) { + if (perf_net_sockready(socks[sock], dummypipe[0], TIMEOUT_CHECK_TIME) == -1) { if (errno == EINPROGRESS) { if (verbose) { 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 { if (verbose) { 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); } + q->is_inprogress = false; perf_list_unlink(instanding_list, q); perf_list_prepend(outstanding_list, q); q->list = &outstanding_list; @@ -499,14 +551,20 @@ do_one_line(perf_buffer_t* lines, perf_buffer_t* msg) 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: if (verbose) { perf_log_warning("failed to send packet: socket %d not ready", sock); } return (PERF_R_FAILURE); 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); default: break; @@ -530,13 +588,14 @@ do_one_line(perf_buffer_t* lines, perf_buffer_t* msg) base = perf_buffer_base(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) < 1) { if (errno == EINPROGRESS) { if (verbose) { perf_log_warning("network congested, packet sending in progress"); } + q->is_inprogress = true; } else { if (verbose) { char __s[256]; @@ -591,8 +650,17 @@ try_process_response(unsigned int sockindex) ramp_bucket* b; 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; - 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 (errno == EAGAIN || errno == EINTR) { return; @@ -611,11 +679,12 @@ try_process_response(unsigned int sockindex) qid = ntohs(packet_header[0]); rcode = ntohs(packet_header[1]) & 0xF; - q = &queries[qid * nsocks + sockindex]; - if (q->list != &outstanding_list) { + size_t idx = qid * nsocks + sockindex; + if (idx >= max_outstanding || queries[idx].list != &outstanding_list) { perf_log_warning("received a response with an unexpected id: %u", qid); return; } + q = &queries[idx]; perf_list_unlink(outstanding_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 i; @@ -701,7 +755,15 @@ int main(int argc, char** argv) if (pipe(dummypipe) < 0) 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)); @@ -725,7 +787,8 @@ int main(int argc, char** argv) printf("[Status] Sending\n"); - current_sock = 0; + int try_responses = (max_qps / max_outstanding) + 1; + current_sock = 0; for (;;) { int should_send; uint64_t time_since_start = time_now - time_of_program_start; @@ -745,7 +808,7 @@ int main(int argc, char** argv) } if (phase != PHASE_WAIT) { 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, " "ending test at %.0f qps\n", should_send, (max_qps * time_since_start) / ramp_time); @@ -762,8 +825,11 @@ int main(int argc, char** argv) } } } - try_process_response(current_sock++); - current_sock = current_sock % nsocks; + for (i = try_responses; i--;) { + try_process_response(current_sock++); + if (current_sock >= nsocks) + current_sock = 0; + } retire_old_queries(); time_now = perf_get_time(); } @@ -780,8 +846,8 @@ end_loop: } /* Print column headers */ - fprintf(plotf, "# time target_qps actual_qps " - "responses_per_sec failures_per_sec avg_latency\n"); + fprintf(plotf, "# time target_qps actual_qps responses_per_sec failures_per_sec avg_latency" + " connections conn_avg_latency\n"); /* Don't print unused 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 latency = buckets[i].responses ? buckets[i].latency_sum / buckets[i].responses : 0; 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, target_qps, - buckets[i].queries / interval, - buckets[i].responses / interval, - buckets[i].failures / interval, - latency); + (double)buckets[i].queries / interval, + (double)buckets[i].responses / interval, + (double)buckets[i].failures / interval, + latency, + (double)buckets[i].connections / interval, + conn_latency); } fclose(plotf); diff --git a/src/test/Makefile.am b/src/test/Makefile.am index 9b3e7c2..377edea 100644 --- a/src/test/Makefile.am +++ b/src/test/Makefile.am @@ -1,9 +1,9 @@ MAINTAINERCLEANFILES = $(srcdir)/Makefile.in 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) \ - datafile datafile2 updatefile + datafile datafile2 updatefile datafile3 datafile4 diff --git a/src/test/datafile4 b/src/test/datafile4 new file mode 100644 index 0000000..3193ed1 --- /dev/null +++ b/src/test/datafile4 @@ -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 diff --git a/src/test/test2.sh b/src/test/test2.sh index a4cfce0..56932f9 100755 --- a/src/test/test2.sh +++ b/src/test/test2.sh @@ -16,7 +16,7 @@ grep -q "Queries sent: *4" test2.out ../dnsperf -s $ip -d "$srcdir/datafile" -n 1 -m tcp >test2.out cat 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 grep -q "Queries sent: *2" 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 # 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 @@ -78,7 +78,7 @@ done # for ip sleep 2 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 pkill -KILL -u `id -u` dnsperf || true diff --git a/src/test/test4.sh b/src/test/test4.sh new file mode 100755 index 0000000..5b334af --- /dev/null +++ b/src/test/test4.sh @@ -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 |