Merge pull request #253 from jelu/release/2.13.1

Release/2.13.1
DNS-OARC · Aug 23, 2023 · 9339ee4 · 9339ee4
2 parents 1e97514 + aa5f31c
commit 9339ee4
Show file tree

Hide file tree

Showing 9 changed files with 190 additions and 59 deletions.
diff --git a/CHANGES b/CHANGES
@@ -1,3 +1,35 @@
+2023-08-23 Jerry Lundström
+
+    Release 2.13.1
+
+    This release fixes a few issues with in-progress queries and the TCP
+    transport module.
+
+    When using stateful connections, such as TCP, if the full query couldn't
+    be sent in one go then the query and connection could get stuck as
+    "in-progress".
+    This could easily happen if you limited the in-flight queries to 1 and
+    sent very large DNS messages using the stream binary format.
+    Additional socket ready checks has been added to flush in-progress
+    queries as quickly as possible.
+
+    The TCP module's handling of errors when continuing in-progress queries
+    has been fixed. Previously it would interpret EAGAIN as an unrecoverable
+    error and trigger a reconnect, dropping the quer-y/ies in-progress for
+    that connection.
+
+    Other changes:
+    - Fix input data buffer to allow for maximum binary blob wire format, 2 byte size + max DNS message
+    - Mention PowerTools repository for building on CentOS etc
+
+    56e180c In-progress, bitmaps, clang-format
+    bbc48db In-progress
+    059619d In-progress
+    752575a TCP EAGAIN
+    d51a453 Max input
+    76764d7 64k TCP/DoT payload support
+    8201f50 Doc
+
 2023-06-15 Jerry Lundström
 
     Release 2.13.0

diff --git a/README.md b/README.md
@@ -53,7 +53,7 @@ To install the dependencies under Debian/Ubuntu:
 apt-get install -y libssl-dev libldns-dev libck-dev libnghttp2-dev
 ```
 
-To install the dependencies under CentOS (with EPEL enabled):
+To install the dependencies under CentOS (with EPEL/PowerTools enabled):
 ```
 yum install -y openssl-devel ldns-devel ck-devel libnghttp2-devel
 ```

diff --git a/configure.ac b/configure.ac
@@ -16,7 +16,7 @@
 # limitations under the License.
 
 AC_PREREQ(2.64)
-AC_INIT([dnsperf], [2.13.0], [[email protected]], [dnsperf], [https://github.com/DNS-OARC/dnsperf/issues])
+AC_INIT([dnsperf], [2.13.1], [[email protected]], [dnsperf], [https://github.com/DNS-OARC/dnsperf/issues])
 AM_INIT_AUTOMAKE([-Wall -Werror foreign subdir-objects])
 AC_CONFIG_SRCDIR([src/dnsperf.c])
 AC_CONFIG_HEADER([src/config.h])

diff --git a/debian/changelog b/debian/changelog
@@ -1,3 +1,37 @@
+dnsperf (2.13.1-1~unstable+1) unstable; urgency=low
+
+  * Release 2.13.1
+
+    This release fixes a few issues with in-progress queries and the TCP
+    transport module.
+
+    When using stateful connections, such as TCP, if the full query couldn't
+    be sent in one go then the query and connection could get stuck as
+    "in-progress".
+    This could easily happen if you limited the in-flight queries to 1 and
+    sent very large DNS messages using the stream binary format.
+    Additional socket ready checks has been added to flush in-progress
+    queries as quickly as possible.
+
+    The TCP module's handling of errors when continuing in-progress queries
+    has been fixed. Previously it would interpret EAGAIN as an unrecoverable
+    error and trigger a reconnect, dropping the quer-y/ies in-progress for
+    that connection.
+
+    Other changes:
+    - Fix input data buffer to allow for maximum binary blob wire format, 2 byte size + max DNS message
+    - Mention PowerTools repository for building on CentOS etc
+
+    56e180c In-progress, bitmaps, clang-format
+    bbc48db In-progress
+    059619d In-progress
+    752575a TCP EAGAIN
+    d51a453 Max input
+    76764d7 64k TCP/DoT payload support
+    8201f50 Doc
+
+ -- Jerry Lundström <[email protected]>  Wed, 23 Aug 2023 14:59:05 +0200
+
 dnsperf (2.13.0-1~unstable+1) unstable; urgency=low
 
   * Release 2.13.0

diff --git a/rpm/dnsperf.spec b/rpm/dnsperf.spec
@@ -1,5 +1,5 @@
 Name:           dnsperf
-Version:        2.13.0
+Version:        2.13.1
 Release:        1%{?dist}
 Summary:        DNS Performance Testing Tool
 Group:          Productivity/Networking/DNS/Utilities
@@ -95,6 +95,32 @@ rm -rf $RPM_BUILD_ROOT
 
 
 %changelog
+* Wed Aug 23 2023 Jerry Lundström <[email protected]> 2.13.1-1
+- Release 2.13.1
+  * This release fixes a few issues with in-progress queries and the TCP
+    transport module.
+  * When using stateful connections, such as TCP, if the full query couldn't
+    be sent in one go then the query and connection could get stuck as
+    "in-progress".
+    This could easily happen if you limited the in-flight queries to 1 and
+    sent very large DNS messages using the stream binary format.
+    Additional socket ready checks has been added to flush in-progress
+    queries as quickly as possible.
+  * The TCP module's handling of errors when continuing in-progress queries
+    has been fixed. Previously it would interpret EAGAIN as an unrecoverable
+    error and trigger a reconnect, dropping the quer-y/ies in-progress for
+    that connection.
+  * Other changes:
+    - Fix input data buffer to allow for maximum binary blob wire format, 2 byte size + max DNS message
+    - Mention PowerTools repository for building on CentOS etc
+  * Commits:
+    56e180c In-progress, bitmaps, clang-format
+    bbc48db In-progress
+    059619d In-progress
+    752575a TCP EAGAIN
+    d51a453 Max input
+    76764d7 64k TCP/DoT payload support
+    8201f50 Doc
 * Thu Jun 15 2023 Jerry Lundström <[email protected]> 2.13.0-1
 - Release 2.13.0
   * This release adds a new option `-O tls-sni=...` to set the Server Name

diff --git a/src/dnsperf.c b/src/dnsperf.c
@@ -66,7 +66,7 @@
 
 #define TIMEOUT_CHECK_TIME 100000
 
-#define MAX_INPUT_DATA (64 * 1024)
+#define MAX_INPUT_DATA (64 * 1024) + 2
 
 #define MAX_SOCKETS 256
 
@@ -853,6 +853,40 @@ wait_for_start(void)
     PERF_UNLOCK(&start_lock);
 }
 
+static inline void
+bit_set(unsigned char* bits, unsigned int bit)
+{
+    unsigned int shift, mask;
+
+    shift = 7 - (bit % 8);
+    mask  = 1 << shift;
+
+    bits[bit / 8] |= mask;
+}
+
+static inline void
+bit_clear(unsigned char* bits, unsigned int bit)
+{
+    unsigned int shift, mask;
+
+    shift = 7 - (bit % 8);
+    mask  = 1 << shift;
+
+    bits[bit / 8] &= ~mask;
+}
+
+static inline bool
+bit_check(unsigned char* bits, unsigned int bit)
+{
+    unsigned int shift;
+
+    shift = 7 - (bit % 8);
+
+    if ((bits[bit / 8] >> shift) & 0x01)
+        return true;
+    return false;
+}
+
 static void*
 do_send(void* arg)
 {
@@ -871,9 +905,10 @@ do_send(void* arg)
     unsigned char   packet_buffer[MAX_EDNS_PACKET];
     unsigned char*  base;
     unsigned int    length;
-    int             n, i, any_inprogress = 0;
+    int             n, i, any_inprogress = 0, sock = 0;
     perf_result_t   result;
     bool            all_fail;
+    unsigned char   socketbits[(MAX_SOCKETS / 8) + 1] = {};
 
     tinfo           = (threadinfo_t*)arg;
     config          = tinfo->config;
@@ -900,6 +935,21 @@ do_send(void* arg)
             now = perf_get_time();
         }
 
+        /* Some sock might still be sending, try flush all of them */
+        if (any_inprogress) {
+            any_inprogress = 0;
+            for (i = 0; i < tinfo->nsocks; i++) {
+                if (!bit_check(socketbits, i)) {
+                    continue;
+                }
+                if (!perf_net_sockready(tinfo->socks[i], threadpipe[0], TIMEOUT_CHECK_TIME)) {
+                    any_inprogress = 1;
+                } else {
+                    bit_clear(socketbits, i);
+                }
+            }
+        }
+
         /* Rate limiting */
         if (tinfo->max_qps > 0) {
             /* the 1 second time slice where q_sent is calculated over */
@@ -910,36 +960,40 @@ do_send(void* arg)
             }
             /* limit QPS over the 1 second slice */
             if (q_sent >= tinfo->max_qps) {
-                wait_us = q_slice - now;
-                if (config->qps_threshold_wait && wait_us > config->qps_threshold_wait) {
-                    wait_us -= config->qps_threshold_wait;
-                    struct timespec ts = { 0, 0 };
-                    if (wait_us >= MILLION) {
-                        ts.tv_sec  = wait_us / MILLION;
-                        ts.tv_nsec = (wait_us % MILLION) * 1000;
-                    } else {
-                        ts.tv_sec  = 0;
-                        ts.tv_nsec = wait_us * 1000;
+                if (!any_inprogress) { // only if nothing is in-progress
+                    wait_us = q_slice - now;
+                    if (config->qps_threshold_wait && wait_us > config->qps_threshold_wait) {
+                        wait_us -= config->qps_threshold_wait;
+                        struct timespec ts = { 0, 0 };
+                        if (wait_us >= MILLION) {
+                            ts.tv_sec  = wait_us / MILLION;
+                            ts.tv_nsec = (wait_us % MILLION) * 1000;
+                        } else {
+                            ts.tv_sec  = 0;
+                            ts.tv_nsec = wait_us * 1000;
+                        }
+                        nanosleep(&ts, NULL);
                     }
-                    nanosleep(&ts, NULL);
                 }
                 now = perf_get_time();
                 continue;
             }
             /* handle stepping to the next window to send a query on */
             if (req_time > now) {
-                wait_us = req_time - now;
-                if (config->qps_threshold_wait && wait_us > config->qps_threshold_wait) {
-                    wait_us -= config->qps_threshold_wait;
-                    struct timespec ts = { 0, 0 };
-                    if (wait_us >= MILLION) {
-                        ts.tv_sec  = wait_us / MILLION;
-                        ts.tv_nsec = (wait_us % MILLION) * 1000;
-                    } else {
-                        ts.tv_sec  = 0;
-                        ts.tv_nsec = wait_us * 1000;
+                if (!any_inprogress) { // only if nothing is in-progress
+                    wait_us = req_time - now;
+                    if (config->qps_threshold_wait && wait_us > config->qps_threshold_wait) {
+                        wait_us -= config->qps_threshold_wait;
+                        struct timespec ts = { 0, 0 };
+                        if (wait_us >= MILLION) {
+                            ts.tv_sec  = wait_us / MILLION;
+                            ts.tv_nsec = (wait_us % MILLION) * 1000;
+                        } else {
+                            ts.tv_sec  = 0;
+                            ts.tv_nsec = wait_us * 1000;
+                        }
+                        nanosleep(&ts, NULL);
                     }
-                    nanosleep(&ts, NULL);
                 }
                 now = perf_get_time();
                 continue;
@@ -951,7 +1005,9 @@ do_send(void* arg)
 
         /* Limit in-flight queries */
         if (num_outstanding(stats) >= tinfo->max_outstanding) {
-            PERF_TIMEDWAIT(&tinfo->cond, &tinfo->lock, &times->stop_time_ns, NULL);
+            if (!any_inprogress) { // only if nothing is in-progress
+                PERF_TIMEDWAIT(&tinfo->cond, &tinfo->lock, &times->stop_time_ns, NULL);
+            }
             PERF_UNLOCK(&tinfo->lock);
             now = perf_get_time();
             continue;
@@ -964,7 +1020,8 @@ do_send(void* arg)
         i        = tinfo->nsocks * 2;
         all_fail = true;
         while (i--) {
-            q->sock = tinfo->socks[tinfo->current_sock++ % tinfo->nsocks];
+            sock    = tinfo->current_sock++ % tinfo->nsocks;
+            q->sock = tinfo->socks[sock];
             switch (perf_net_sockready(q->sock, threadpipe[0], TIMEOUT_CHECK_TIME)) {
             case 0:
                 if (config->verbose && !config->suppress.sockready) {
@@ -1060,6 +1117,7 @@ do_send(void* arg)
                     perf_log_warning("network congested, packet sending in progress");
                 }
                 any_inprogress = 1;
+                bit_set(socketbits, sock);
             } else {
                 if (config->verbose && !config->suppress.sendfailed) {
                     char __s[256];
@@ -1185,29 +1243,6 @@ recv_one(threadinfo_t* tinfo, int which_sock,
     return true;
 }
 
-static inline void
-bit_set(unsigned char* bits, unsigned int bit)
-{
-    unsigned int shift, mask;
-
-    shift = 7 - (bit % 8);
-    mask  = 1 << shift;
-
-    bits[bit / 8] |= mask;
-}
-
-static inline bool
-bit_check(unsigned char* bits, unsigned int bit)
-{
-    unsigned int shift;
-
-    shift = 7 - (bit % 8);
-
-    if ((bits[bit / 8] >> shift) & 0x01)
-        return true;
-    return false;
-}
-
 static void*
 do_recv(void* arg)
 {
@@ -1217,7 +1252,7 @@ do_recv(void* arg)
     received_query_t recvd[RECV_BATCH_SIZE] = { { 0, 0, 0, 0, 0, 0, false, false, 0 } };
     unsigned int     nrecvd;
     int              saved_errno;
-    unsigned char    socketbits[MAX_SOCKETS / 8];
+    unsigned char    socketbits[(MAX_SOCKETS / 8) + 1];
     uint64_t         now, latency;
     query_info*      q;
     unsigned int     current_socket, last_socket;
@@ -1640,8 +1675,7 @@ int main(int argc, char** argv)
     perf_os_handlesignal(SIGINT, handle_sigint);
     perf_os_blocksignal(SIGINT, false);
     sock.fd = mainpipe[0];
-    result  = perf_os_waituntilreadable(&sock, intrpipe[0],
-        times.stop_time - times.start_time);
+    result  = perf_os_waituntilreadable(&sock, intrpipe[0], times.stop_time - times.start_time);
     if (result == PERF_R_CANCELED)
         interrupted = true;
 

diff --git a/src/net.h b/src/net.h
@@ -28,8 +28,8 @@
 #include <assert.h>
 #include <stdbool.h>
 
-#define TCP_RECV_BUF_SIZE (16 * 1024)
-#define TCP_SEND_BUF_SIZE (4 * 1024)
+#define TCP_RECV_BUF_SIZE (65535 + 2)
+#define TCP_SEND_BUF_SIZE (65535 + 2)
 
 struct perf_sockaddr {
     union {

diff --git a/src/net_tcp.c b/src/net_tcp.c
@@ -297,6 +297,9 @@ static int perf__tcp_sockready(struct perf_net_socket* sock, int pipe_fd, int64_
             dnslen = ntohs(dnslen);
             n      = sendto(sock->fd, self->sendbuf + self->sending, dnslen + 2 - self->sending, 0, 0, 0);
             if (n < 0) {
+                if (errno == EAGAIN) {
+                    return 0;
+                }
                 int fd = perf__tcp_connect(sock), oldfd = ck_pr_load_int(&sock->fd);
                 ck_pr_store_int(&sock->fd, fd);
                 close(oldfd);
@@ -364,7 +367,9 @@ static int perf__tcp_sockready(struct perf_net_socket* sock, int pipe_fd, int64_
             dnslen = ntohs(dnslen);
             n      = sendto(sock->fd, self->sendbuf + self->sending, dnslen + 2 - self->sending, 0, 0, 0);
             if (n < 0) {
-                self->need_reconnect = true;
+                if (errno != EAGAIN) {
+                    self->need_reconnect = true;
+                }
                 return 0;
             }
             self->sending += n;

diff --git a/src/resperf.c b/src/resperf.c
@@ -61,7 +61,7 @@
 #define DEFAULT_MAX_OUTSTANDING (64 * 1024)
 #define DEFAULT_MAX_FALL_BEHIND 1000
 
-#define MAX_INPUT_DATA (64 * 1024)
+#define MAX_INPUT_DATA (64 * 1024) + 2
 
 #define TIMEOUT_CHECK_TIME 5000000