Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 30 Jan 2010 10:45:23 GMT
From:      Mikolaj Golub <to.my.trociny@gmail.com>
To:        freebsd-gnats-submit@FreeBSD.org
Subject:   bin/143368: [patch] awk(1): number of open files is limited to small constant
Message-ID:  <201001301045.o0UAjNsN094092@www.freebsd.org>
Resent-Message-ID: <201001301050.o0UAo1li051418@freefall.freebsd.org>

next in thread | raw e-mail | index | archive | help

>Number:         143368
>Category:       bin
>Synopsis:       [patch] awk(1): number of open files is limited to small constant
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Sat Jan 30 10:50:01 UTC 2010
>Closed-Date:
>Last-Modified:
>Originator:     Mikolaj Golub
>Release:        
>Organization:
>Environment:
FreeBSD zhuzha.ua1 8.0-STABLE FreeBSD 8.0-STABLE #6: Sun Jan 24 21:36:17 EET 2010     root@zhuzha.ua1:/usr/obj/usr/src/sys/GENERIC  i386
>Description:
The maximal number of open files in awk(1) is limited by FOPEN_MAX constant defined for awk in usr.bin/awk/Makefile:

CFLAGS+= -DHAS_ISBLANK -I. -I${AWKSRC} -DFOPEN_MAX=64

This might be low for today needs and today hardware allows much higher values.

NetBSD has fixed their version of awk making the array of files dynamically allocated

http://www.netbsd.org/cgi-bin/query-pr-single.pl?number=37205

It would be nice to do the same in FreeBSD.
>How-To-Repeat:
kopusha:~% awk 'BEGIN {for (i=1; i <= 2000; ++i){print "test" > ("/tmp/file_" i)}}'
awk: /tmp/file_62 makes too many open files
 source line number 1

>Fix:
See the attached patch adopted from NetBSD (PR/37205: Aleksey Cheusov: nawk: a number of open files is limited to very small constant. Make the array of files dynamically allocated.).

Patch attached with submission follows:

diff -ru contrib/one-true-awk.orig/run.c contrib/one-true-awk/run.c
--- contrib/one-true-awk.orig/run.c	2007-06-05 18:33:51.000000000 +0300
+++ contrib/one-true-awk/run.c	2010-01-30 12:27:38.000000000 +0200
@@ -1613,28 +1613,36 @@
 	FILE	*fp;
 	const char	*fname;
 	int	mode;	/* '|', 'a', 'w' => LE/LT, GT */
-} files[FOPEN_MAX] ={
-	{ NULL,  "/dev/stdin",  LT },	/* watch out: don't free this! */
-	{ NULL, "/dev/stdout", GT },
-	{ NULL, "/dev/stderr", GT }
-};
+} *files;
+size_t nfiles;
 
 void stdinit(void)	/* in case stdin, etc., are not constants */
 {
+	nfiles = FOPEN_MAX;
+	files = calloc(nfiles, sizeof(*files));
+	if (files == NULL)
+		FATAL("can't allocate file memory for %zu files", nfiles);
 	files[0].fp = stdin;
+	files[0].fname = "/dev/stdin";
+	files[0].mode = LT;
 	files[1].fp = stdout;
+	files[1].fname = "/dev/stdout";
+	files[1].mode = GT;
 	files[2].fp = stderr;
+	files[2].fname = "/dev/stderr";
+	files[2].mode = GT;
 }
 
 FILE *openfile(int a, const char *us)
 {
 	const char *s = us;
-	int i, m;
+	size_t i;
+	int m;
 	FILE *fp = 0;
 
 	if (*s == '\0')
 		FATAL("null file name in print or getline");
-	for (i=0; i < FOPEN_MAX; i++)
+	for (i = 0; i < nfiles; i++)
 		if (files[i].fname && strcmp(s, files[i].fname) == 0) {
 			if (a == files[i].mode || (a==APPEND && files[i].mode==GT))
 				return files[i].fp;
@@ -1644,11 +1652,19 @@
 	if (a == FFLUSH)	/* didn't find it, so don't create it! */
 		return NULL;
 
-	for (i=0; i < FOPEN_MAX; i++)
-		if (files[i].fp == 0)
+	for (i = 0; i < nfiles; i++)
+		if (files[i].fp == NULL)
 			break;
-	if (i >= FOPEN_MAX)
-		FATAL("%s makes too many open files", s);
+	if (i >= nfiles) {
+		struct files *nf;
+		size_t nnf = nfiles + FOPEN_MAX;
+		nf = realloc(files, nnf * sizeof(*nf));
+		if (nf == NULL)
+			FATAL("cannot grow files for %s and %zu files", s, nnf);
+		(void)memset(&nf[nfiles], 0, FOPEN_MAX * sizeof(*nf));
+		nfiles = nnf;
+		files = nf;
+	}
 	fflush(stdout);	/* force a semblance of order */
 	m = a;
 	if (a == GT) {
@@ -1674,9 +1690,9 @@
 
 const char *filename(FILE *fp)
 {
-	int i;
+	size_t i;
 
-	for (i = 0; i < FOPEN_MAX; i++)
+	for (i = 0; i < nfiles; i++)
 		if (fp == files[i].fp)
 			return files[i].fname;
 	return "???";
@@ -1685,13 +1701,14 @@
 Cell *closefile(Node **a, int n)
 {
 	Cell *x;
-	int i, stat;
+	size_t i;
+	int stat;
 
 	n = n;
 	x = execute(a[0]);
 	getsval(x);
 	stat = -1;
-	for (i = 0; i < FOPEN_MAX; i++) {
+	for (i = 0; i < nfiles; i++) {
 		if (files[i].fname && strcmp(x->sval, files[i].fname) == 0) {
 			if (ferror(files[i].fp))
 				WARNING( "i/o error occurred on %s", files[i].fname );
@@ -1715,9 +1732,10 @@
 
 void closeall(void)
 {
-	int i, stat;
+	size_t i;
+	int stat;
 
-	for (i = 0; i < FOPEN_MAX; i++) {
+	for (i = 0; i < nfiles; i++) {
 		if (files[i].fp) {
 			if (ferror(files[i].fp))
 				WARNING( "i/o error occurred on %s", files[i].fname );
@@ -1733,9 +1751,9 @@
 
 void flush_all(void)
 {
-	int i;
+	size_t i;
 
-	for (i = 0; i < FOPEN_MAX; i++)
+	for (i = 0; i < nfiles; i++)
 		if (files[i].fp)
 			fflush(files[i].fp);
 }


>Release-Note:
>Audit-Trail:
>Unformatted:



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201001301045.o0UAjNsN094092>