Merge d2b22f5 into 9bd31ff

Author: jimklimov

NEWS.adoc

@@ -12,7 +12,6 @@ For a complete and more detailed list of changes, please refer to the
 ChangeLog file (generated for release archives), or to the Git version
 control history for "live" codebase.
 
-
 PLANNED: Release notes for NUT 2.8.5 - what's new since 2.8.4
 -------------------------------------------------------------
 
@@ -287,6 +286,9 @@ https://github.com/networkupstools/nut/milestone/12
      comes into play and breaks things. [issue #661]
    * Fixed `LISTEN *` handling for `upsd.exe` in NUT for Windows builds.
      [PR #3237]
+   * Extended processing of `MAXCONN` setting to allow larger values than the
+     operating system allows, by only waiting for that amount of Unix sockets
+     or Windows `HANDLE`'s at a time, and moving on to another chunk. [#3302]
 
  - `upsdrvctl` tool updates:
    * Make use of `setproctag()` and `getproctag()` to report parent/child

conf/upsd.conf.sample

@@ -121,6 +121,11 @@
 # LISTEN address and each client count as one connection.  If the server
 # runs out of connections, it will no longer accept new incoming client
 # connections.  Only set this if you know exactly what you're doing.
+# Note that on some platforms there may be a smaller amount of file descriptors
+# or handles that can be polled in one operation, the server would then poll
+# several smaller groups until it handles all the connections it tracks.
+# With a large amount of connections this may however impact the delays between
+# processing loops, and time before an incoming message is seen and processed.
 
 # =======================================================================
 # CERTFILE <certificate file>

docs/man/upsd.conf.txt

@@ -150,6 +150,12 @@ This defaults to maximum number allowed on your system.  Each UPS, each
 `LISTEN` address and each client count as one connection.  If the server
 runs out of connections, it will no longer accept new incoming client
 connections.  Only set this if you know exactly what you're doing.
++
+Note that on some platforms there may be a smaller amount of file descriptors
+or handles that can be polled in one operation, the server would then poll
+several smaller groups until it handles all the connections it tracks.
+With a large amount of connections this may however impact the delays between
+processing loops, and time before an incoming message is seen and processed.
 
 *CERTFILE 'certificate file'*::
 

server/upsd.c

@@ -5,7 +5,7 @@
 	2008		Arjen de Korte <adkorte-guest@alioth.debian.org>
 	2011 - 2012	Arnaud Quette <arnaud.quette.free.fr>
 	2019 		Eaton (author: Arnaud Quette <ArnaudQuette@eaton.com>)
-	2020 - 2025	Jim Klimov <jimklimov+nut@gmail.com>
+	2020 - 2026	Jim Klimov <jimklimov+nut@gmail.com>
 
    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
@@ -94,8 +94,12 @@ int allow_no_device = 0;
  */
 int allow_not_all_listeners = 0;
 
-/* preloaded to {OPEN_MAX} in main, can be overridden via upsd.conf */
+/* preloaded to POSIX sysconf(_SC_OPEN_MAX) or WIN32 MAX_WAIT_OBJECTS in main
+ * and elsewhere, the run-time value can be overridden via upsd.conf `MAXCONN`
+ * option (may cause partial waits chunk by chunk, if sysmaxconn is smaller).
+ */
 nfds_t	maxconn = 0;
+static nfds_t	sysmaxconn = 0;
 
 /* preloaded to STATEPATH in main, can be overridden via upsd.conf */
 char	*statepath = NULL;
@@ -112,7 +116,7 @@ nut_ctype_t	*firstclient = NULL;
 /* default is to listen on all local interfaces */
 static stype_t	*firstaddr = NULL;
 
-static int 	opt_af = AF_UNSPEC;
+static int	opt_af = AF_UNSPEC;
 
 typedef enum {
 	DRIVER = 1,
@@ -1224,30 +1228,37 @@ static void poll_reload(void)
 	size_t	maxalloc;
 
 #ifndef WIN32
+	/* Not likely this would change, but refresh just in case */
 	ret = sysconf(_SC_OPEN_MAX);
 #else	/* WIN32 */
 	ret = (long)MAXIMUM_WAIT_OBJECTS;
 #endif	/* WIN32 */
 
-	if ((intmax_t)ret < (intmax_t)maxconn) {
+	if (ret < 1) {
+		/* TOTHINK: Not fail, but use a conservative fallback number? */
 		fatalx(EXIT_FAILURE,
+			"System reported an absurd value %ld as maximum number of connections.\n"
+			"The server won't start until this problem is resolved.\n", ret);
+	}
+
+	if ((intmax_t)ret < (intmax_t)maxconn) {
+		upslogx(LOG_WARNING,
 			"Your system limits the maximum number of connections to %ld\n"
-			"but you requested %" PRIdMAX ". The server won't start until this\n"
-			"problem is resolved.\n", ret, (intmax_t)maxconn);
+			"but you requested %" PRIdMAX ". The server may handle connections\n"
+			"in smaller groups, maybe affecting efficiency and response time.\n",
+			ret, (intmax_t)maxconn);
 	}
 
+	sysmaxconn = (nfds_t)ret;
+
 	if (1 > maxconn) {
 		fatalx(EXIT_FAILURE,
 			"You requested %" PRIdMAX " as maximum number of connections.\n"
 			"The server won't start until this problem is resolved.\n", (intmax_t)maxconn);
 	}
 
-#ifndef WIN32
 	/* How many items can we stuff into the array? */
 	maxalloc = SIZE_MAX / sizeof(void *);
-#else	/* WIN32 */
-	maxalloc = MAXIMUM_WAIT_OBJECTS;
-#endif	/* WIN32 */
 	if ((uintmax_t)maxalloc < (uintmax_t)maxconn) {
 		fatalx(EXIT_FAILURE,
 			"You requested %" PRIdMAX " as maximum number of connections, but we can only allocate %" PRIuSIZE ".\n"
@@ -1497,14 +1508,14 @@ static void mainloop(void)
 	nfds_t	i;
 #else	/* WIN32 */
 	DWORD	ret;
-	pipe_conn_t * conn;
+	pipe_conn_t	*conn;
 #endif	/* WIN32 */
 
 	size_t	nfds_wanted = 0,	/* Connections we looked at (some may be invalid) */
 		nfds_considered = 0;	/* Connections we wanted to poll (but might be over maxconn limit) */
 	nfds_t	nfds = 0;
 	upstype_t	*ups;
-	nut_ctype_t		*client, *cnext;
+	nut_ctype_t	*client, *cnext;
 	stype_t		*server;
 	time_t	now;
 
@@ -1644,18 +1655,53 @@ static void mainloop(void)
 	upsdebugx(2, "%s: polling %" PRIdMAX " filedescriptors; some stats: "
 		"considered %" PRIdMAX " connections, "
 		"wanted to actually poll %" PRIdMAX
-		" and was constrained by maxconn=%" PRIdMAX,
+		" and was constrained by maxconn=%" PRIdMAX
+		" and chunked by sysmaxconn=%" PRIdMAX,
 		__func__, (intmax_t)nfds, (intmax_t)nfds_considered,
-		(intmax_t)nfds_wanted, (intmax_t)maxconn);
+		(intmax_t)nfds_wanted, (intmax_t)maxconn, (intmax_t)sysmaxconn);
 
 	if (nfds_wanted != nfds || nfds_wanted >= maxconn) {
 		upslogx(LOG_ERR, "upsd polling %" PRIdMAX " filedescriptors,"
 			" but wanted to poll %" PRIdMAX
-			" and was constrained by maxconn=%" PRIdMAX,
+			" and was constrained by maxconn=%" PRIdMAX
+			" (see upsd.conf MAXCONN setting to adjust)",
 			(intmax_t)nfds, (intmax_t)nfds_wanted, (intmax_t)maxconn);
 	}
 
-	ret = poll(fds, nfds, 2000);
+	if (nfds <= sysmaxconn) {
+		ret = poll(fds, nfds, 2000);
+	} else {
+		/* Chunk it all; try to fit into same 2 sec as above.
+		 * Note that nfds at the moment may be smaller than
+		 * maxconn (allocated array size).
+		 */
+		size_t	last_chunk = nfds % sysmaxconn, chunk,
+			chunks = nfds / sysmaxconn + (last_chunk ? 1 : 0);
+		int	poll_TO = 2000 / chunks, tmpret;
+
+		if (poll_TO < 10)
+			poll_TO = 10;
+
+		ret = 0;
+		for (chunk = 0; chunk < chunks; chunk++) {
+			upsdebugx(5,
+				"%s: chunked filedescriptor polling #%" PRIuSIZE
+				" of %" PRIuSIZE " chunks, with %d hits so far",
+				__func__, chunk, chunks, ret);
+			tmpret = poll(&fds[chunk * sysmaxconn],
+				(last_chunk && chunk == chunks - 1 ? last_chunk : sysmaxconn),
+				poll_TO);
+			if (tmpret < 0) {
+				upsdebug_with_errno(2,
+					"%s: failed during chunked polling, handled %" PRIuSIZE
+					" of %" PRIuSIZE " chunks so far, with %d hits",
+					__func__, chunk, chunks, ret);
+				ret = tmpret;
+				break;
+			}
+			ret += tmpret;
+		}
+	}
 
 	if (ret == 0) {
 		upsdebugx(2, "%s: no data available", __func__);
@@ -1942,19 +1988,55 @@ static void mainloop(void)
 	upsdebugx(2, "%s: wait for %" PRIdMAX " filedescriptors; some stats: "
 		"considered %" PRIdMAX " connections, "
 		"wanted to actually poll %" PRIdMAX
-		" and was constrained by maxconn=%" PRIdMAX,
+		" and was constrained by maxconn=%" PRIdMAX
+		" and chunked by sysmaxconn=%" PRIdMAX,
 		__func__, (intmax_t)nfds, (intmax_t)nfds_considered,
-		(intmax_t)nfds_wanted, (intmax_t)maxconn);
+		(intmax_t)nfds_wanted, (intmax_t)maxconn, (intmax_t)sysmaxconn);
 
 	if (nfds_wanted != nfds || nfds_wanted >= maxconn) {
 		upslogx(LOG_ERR, "upsd polling %" PRIuMAX " filedescriptors,"
 			" but wanted to poll %" PRIuMAX
-			" and was constrained by maxconn=%" PRIuMAX,
+			" and was constrained by maxconn=%" PRIuMAX
+			" (see upsd.conf MAXCONN setting to adjust)",
 			(uintmax_t)nfds, (uintmax_t)nfds_wanted, (uintmax_t)maxconn);
 	}
 
 	/* https://docs.microsoft.com/en-us/windows/win32/api/synchapi/nf-synchapi-waitformultipleobjects */
-	ret = WaitForMultipleObjects(nfds,fds,FALSE,2000);
+	if (nfds <= sysmaxconn) {
+		ret = WaitForMultipleObjects(nfds, fds, FALSE, 2000);
+	} else {
+		/* Chunk it all; try to fit into same 2 sec as above.
+		 * Note that nfds at the moment may be smaller than
+		 * maxconn (allocated array size).
+		 */
+		size_t	last_chunk = nfds % sysmaxconn, chunk,
+			chunks = nfds / sysmaxconn + (last_chunk ? 1 : 0);
+		DWORD	poll_TO = 2000 / chunks, tmpret;
+
+		if (poll_TO < 10)
+			poll_TO = 10;
+
+		ret = 0;
+		for (chunk = 0; chunk < chunks; chunk++) {
+			upsdebugx(5,
+				"%s: chunked filedescriptor polling #%" PRIuSIZE
+				" of %" PRIuSIZE " chunks, with %" PRIu64 " hits so far",
+				__func__, chunk, chunks, ret);
+			tmpret = WaitForMultipleObjects(
+				(last_chunk && chunk == chunks - 1 ? last_chunk : sysmaxconn),
+				&fds[chunk * sysmaxconn],
+				FALSE, poll_TO);
+			if (tmpret < 0) {
+				upsdebug_with_errno(2,
+					"%s: failed during chunked polling, handled %" PRIuSIZE
+					" of %" PRIuSIZE " chunks so far, with %" PRIu64 " hits",
+					__func__, chunk, chunks, ret);
+				ret = tmpret;
+				break;
+			}
+			ret += tmpret;
+		}
+	}
 
 	upsdebugx(6, "%s: wait for filedescriptors done: %" PRIu64, __func__, ret);
 
@@ -2150,6 +2232,7 @@ void check_perms(const char *fn)
 int main(int argc, char **argv)
 {
 	int	i, cmdret = 0, foreground = -1;
+	long	l;
 #ifndef WIN32
 	int	cmd = 0;
 	pid_t	oldpid = -1;
@@ -2284,14 +2367,14 @@ int main(int argc, char **argv)
 	}
 
 	{ /* scoping */
-		char *s = getenv("NUT_DEBUG_LEVEL");
-		int l;
-		if (s && str_to_int(s, &l, 10)) {
-			if (l > 0 && nut_debug_level_args < 1) {
+		char	*s = getenv("NUT_DEBUG_LEVEL");
+		int	lvl;
+		if (s && str_to_int(s, &lvl, 10)) {
+			if (lvl > 0 && nut_debug_level_args < 1) {
 				upslogx(LOG_INFO, "Defaulting debug verbosity to NUT_DEBUG_LEVEL=%d "
-					"since none was requested by command-line options", l);
-				nut_debug_level = l;
-				nut_debug_level_args = l;
+					"since none was requested by command-line options", lvl);
+				nut_debug_level = lvl;
+				nut_debug_level_args = lvl;
 			}	/* else follow -D settings */
 		}	/* else nothing to bother about */
 	}
@@ -2447,12 +2530,21 @@ int main(int argc, char **argv)
 #ifndef WIN32
 	/* default to system limit (may be overridden in upsd.conf) */
 	/* FIXME: Check for overflows (and int size of nfds_t vs. long) - see get_max_pid_t() for example */
-	maxconn = (nfds_t)sysconf(_SC_OPEN_MAX);
+	l = sysconf(_SC_OPEN_MAX);
 #else	/* WIN32 */
 	/* hard-coded 64 (from ddk/wdm.h or winnt.h) */
-	maxconn = MAXIMUM_WAIT_OBJECTS;
+	l = (long)MAXIMUM_WAIT_OBJECTS;
 #endif	/* WIN32 */
 
+	if (l < 1) {
+		/* TOTHINK: Not fail, but use a conservative fallback number? */
+		fatalx(EXIT_FAILURE,
+			"System reported an absurd value %ld as maximum number of connections.\n"
+			"The server won't start until this problem is resolved.\n", l);
+	}
+
+	maxconn = sysmaxconn = (nfds_t)l;
+
 	/* handle upsd.conf */
 	load_upsdconf(0);	/* 0 = initial */