find: use sysconf(_SC_ARG_MAX) to determine the command-line size limit

The find utility uses a hardcoded value of 32 * 1024 as the limit of
the command-line length when calling 'find -exec ... {} +'. This results
in over 4 times more execve() calls than in coreutils' find.

This patch uses the limit defined in system headers.

Based on the patch by  Bartosz Golaszewski <bartekgola@gmail.com>.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

findutils/find.c

@@ -419,6 +419,7 @@ struct globals {
 	smallint need_print;
 	smallint xdev_on;
 	recurse_flags_t recurse_flags;
+	IF_FEATURE_FIND_EXEC_PLUS(unsigned max_argv_len;)
 } FIX_ALIASING;
 #define G (*(struct globals*)&bb_common_bufsiz1)
 #define INIT_G() do { \
@@ -428,6 +429,7 @@ struct globals {
 	/* we have to zero it out because of NOEXEC */ \
 	memset(&G, 0, sizeof(G)); \
 	IF_FEATURE_FIND_MAXDEPTH(G.minmaxdepth[1] = INT_MAX;) \
+	IF_FEATURE_FIND_EXEC_PLUS(G.max_argv_len = bb_arg_max() - 2048;) \
 	G.need_print = 1; \
 	G.recurse_flags = ACTION_RECURSE; \
 } while (0)
@@ -677,7 +679,7 @@ ACTF(exec)
 		ap->file_len += strlen(fileName) + sizeof(char*) + 1;
 		/* If we have lots of files already, exec the command */
 		rc = 1;
-		if (ap->file_len >= 32*1024)
+		if (ap->file_len >= G.max_argv_len)
 			rc = do_exec(ap, NULL);
 		return rc;
 	}

findutils/xargs.c

@@ -523,12 +523,7 @@ int xargs_main(int argc, char **argv)
 		argc++;
 	}
 
-	/* -s NUM default. fileutils-4.4.2 uses 128k, but I heasitate
+	/*
-	 * to use such a big value - first need to change code to use
-	 * growable buffer instead of fixed one.
-	 */
-	n_max_chars = 32 * 1024;
-	/* Make smaller if system does not allow our default value.
 	 * The Open Group Base Specifications Issue 6:
 	 * "The xargs utility shall limit the command line length such that
 	 * when the command line is invoked, the combined argument
@@ -536,16 +531,15 @@ int xargs_main(int argc, char **argv)
 	 * in the System Interfaces volume of IEEE Std 1003.1-2001)
 	 * shall not exceed {ARG_MAX}-2048 bytes".
 	 */
-	{
+	n_max_chars = bb_arg_max();
-		long arg_max = 0;
+	if (n_max_chars > 32 * 1024)
-#if defined _SC_ARG_MAX
+		n_max_chars = 32 * 1024;
-		arg_max = sysconf(_SC_ARG_MAX) - 2048;
+	/*
-#elif defined ARG_MAX
+	 * POSIX suggests substracting 2048 bytes from sysconf(_SC_ARG_MAX)
-		arg_max = ARG_MAX - 2048;
+	 * so that the process may safely modify its environment.
-#endif
+	 */
-		if (arg_max > 0 && n_max_chars > arg_max)
+	n_max_chars -= 2048;
-			n_max_chars = arg_max;
+
-	}
 	if (opt & OPT_UPTO_SIZE) {
 		n_max_chars = xatou_range(max_chars, 1, INT_MAX);
 	}

include/libbb.h

@@ -731,6 +731,14 @@ extern void *xmalloc_open_read_close(const char *filename, size_t *maxsz_p) FAST
 /* Never returns NULL */
 extern void *xmalloc_xopen_read_close(const char *filename, size_t *maxsz_p) FAST_FUNC RETURNS_MALLOC;
 
+#if defined ARG_MAX
+# define bb_arg_max() ((unsigned)ARG_MAX)
+#elif defined _SC_ARG_MAX
+unsigned bb_arg_max(void) FAST_FUNC;
+#else
+# define bb_arg_max() ((unsigned)(32 * 1024))
+#endif
+
 #define SEAMLESS_COMPRESSION (0 \
  || ENABLE_FEATURE_SEAMLESS_XZ \
  || ENABLE_FEATURE_SEAMLESS_LZMA \

libbb/Kbuild.src

@@ -92,6 +92,7 @@ lib-y += skip_whitespace.o
 lib-y += speed_table.o
 lib-y += str_tolower.o
 lib-y += strrstr.o
+lib-y += sysconf.o
 lib-y += time.o
 lib-y += trim.o
 lib-y += u_signal_names.o

libbb/sysconf.c

@@ -0,0 +1,16 @@
+/* vi: set sw=4 ts=4: */
+/*
+ * Various system configuration helpers.
+ *
+ * Copyright (C) 2014 Bartosz Golaszewski <bartekgola@gmail.com>
+ *
+ * Licensed under GPLv2 or later, see file LICENSE in this source tree.
+ */
+#include "libbb.h"
+
+#if defined _SC_ARG_MAX
+unsigned FAST_FUNC bb_arg_max(void)
+{
+	return sysconf(_SC_ARG_MAX);
+}
+#endif