*: introduce and use FAST_FUNC: regparm on i386, otherwise no-on

   text    data     bss     dec     hex filename
 808035     611    6868  815514   c719a busybox_old
 804472     611    6868  811951   c63af busybox_unstripped

diff --git a/libbb/appletlib.c b/libbb/appletlib.c
index 06a83e5..0017a4b 100644
--- a/libbb/appletlib.c
+++ b/libbb/appletlib.c
@@ -101,7 +101,7 @@
 	full_write(STDERR_FILENO, str, strlen(str));
 }
 
-void bb_show_usage(void)
+void FAST_FUNC bb_show_usage(void)
 {
 	if (ENABLE_SHOW_USAGE) {
 #ifdef SINGLE_APPLET_STR
@@ -153,7 +153,7 @@
 	return strcmp(name, APPLET_NAME(i));
 }
 #endif
-int find_applet_by_name(const char *name)
+int FAST_FUNC find_applet_by_name(const char *name)
 {
 #if NUM_APPLETS > 8
 	/* Do a binary search to find the applet entry given the name. */
@@ -704,7 +704,7 @@
 	xfunc_die();
 }
 
-void run_applet_no_and_exit(int applet_no, char **argv)
+void FAST_FUNC run_applet_no_and_exit(int applet_no, char **argv)
 {
 	int argc = 1;
 
@@ -722,7 +722,7 @@
 	exit(applet_main[applet_no](argc, argv));
 }
 
-void run_applet_and_exit(const char *name, char **argv)
+void FAST_FUNC run_applet_and_exit(const char *name, char **argv)
 {
 	int applet = find_applet_by_name(name);
 	if (applet >= 0)
diff --git a/libbb/ask_confirmation.c b/libbb/ask_confirmation.c
index 646ec4b..d08bc51 100644
--- a/libbb/ask_confirmation.c
+++ b/libbb/ask_confirmation.c
@@ -13,7 +13,7 @@
 
 #include "libbb.h"
 
-int bb_ask_confirmation(void)
+int FAST_FUNC bb_ask_confirmation(void)
 {
 	int retval = 0;
 	int first = 1;
diff --git a/libbb/bb_askpass.c b/libbb/bb_askpass.c
index 3ad0e97..ca18ade 100644
--- a/libbb/bb_askpass.c
+++ b/libbb/bb_askpass.c
@@ -17,7 +17,7 @@
 {
 }
 
-char *bb_askpass(int timeout, const char *prompt)
+char* FAST_FUNC bb_askpass(int timeout, const char *prompt)
 {
 	/* Was static char[BIGNUM] */
 	enum { sizeof_passwd = 128 };
diff --git a/libbb/bb_basename.c b/libbb/bb_basename.c
index e6832f8..bab4166 100644
--- a/libbb/bb_basename.c
+++ b/libbb/bb_basename.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-const char *bb_basename(const char *name)
+const char* FAST_FUNC bb_basename(const char *name)
 {
 	const char *cp = strrchr(name, '/');
 	if (cp)
diff --git a/libbb/bb_do_delay.c b/libbb/bb_do_delay.c
index aa26ade..3d52cc5 100644
--- a/libbb/bb_do_delay.c
+++ b/libbb/bb_do_delay.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_do_delay(int seconds)
+void FAST_FUNC bb_do_delay(int seconds)
 {
 	time_t start, now;
 
diff --git a/libbb/bb_pwd.c b/libbb/bb_pwd.c
index 2bdb662..b03dc83 100644
--- a/libbb/bb_pwd.c
+++ b/libbb/bb_pwd.c
@@ -45,7 +45,7 @@
  * bb_getXXXid(NULL, -1, id) - return user/group name or exit
  */
 /* gets a username given a uid */
-char* bb_getpwuid(char *name, int bufsize, long uid)
+char* FAST_FUNC bb_getpwuid(char *name, int bufsize, long uid)
 {
 	struct passwd *myuser = getpwuid(uid);
 
@@ -54,7 +54,7 @@
 			uid, 'u');
 }
 /* gets a groupname given a gid */
-char* bb_getgrgid(char *group, int bufsize, long gid)
+char* FAST_FUNC bb_getgrgid(char *group, int bufsize, long gid)
 {
 	struct group *mygroup = getgrgid(gid);
 
@@ -64,7 +64,7 @@
 }
 
 /* returns a gid given a group name */
-long xgroup2gid(const char *name)
+long FAST_FUNC xgroup2gid(const char *name)
 {
 	struct group *mygroup;
 
@@ -76,7 +76,7 @@
 }
 
 /* returns a uid given a username */
-long xuname2uid(const char *name)
+long FAST_FUNC xuname2uid(const char *name)
 {
 	struct passwd *myuser;
 
@@ -87,8 +87,8 @@
 	return myuser->pw_uid;
 }
 
-unsigned long get_ug_id(const char *s,
-		long (*xname2id)(const char *))
+unsigned long FAST_FUNC get_ug_id(const char *s,
+		long FAST_FUNC (*xname2id)(const char *))
 {
 	unsigned long r;
 
diff --git a/libbb/bb_qsort.c b/libbb/bb_qsort.c
index e8673ab..9773afa 100644
--- a/libbb/bb_qsort.c
+++ b/libbb/bb_qsort.c
@@ -9,12 +9,12 @@
 
 #include "libbb.h"
 
-int bb_pstrcmp(const void *a, const void *b)
+int /* not FAST_FUNC! */ bb_pstrcmp(const void *a, const void *b)
 {
 	return strcmp(*(char**)a, *(char**)b);
 }
 
-void qsort_string_vector(char **sv, unsigned count)
+void FAST_FUNC qsort_string_vector(char **sv, unsigned count)
 {
 	qsort(sv, count, sizeof(char*), bb_pstrcmp);
 }
diff --git a/libbb/bb_strtonum.c b/libbb/bb_strtonum.c
index 50ef0ba..525c830 100644
--- a/libbb/bb_strtonum.c
+++ b/libbb/bb_strtonum.c
@@ -48,7 +48,7 @@
 }
 
 
-unsigned long long bb_strtoull(const char *arg, char **endp, int base)
+unsigned long long FAST_FUNC bb_strtoull(const char *arg, char **endp, int base)
 {
 	unsigned long long v;
 	char *endptr;
@@ -63,7 +63,7 @@
 	return handle_errors(v, endp, endptr);
 }
 
-long long bb_strtoll(const char *arg, char **endp, int base)
+long long FAST_FUNC bb_strtoll(const char *arg, char **endp, int base)
 {
 	unsigned long long v;
 	char *endptr;
@@ -75,7 +75,7 @@
 }
 
 #if ULONG_MAX != ULLONG_MAX
-unsigned long bb_strtoul(const char *arg, char **endp, int base)
+unsigned long FAST_FUNC bb_strtoul(const char *arg, char **endp, int base)
 {
 	unsigned long v;
 	char *endptr;
@@ -86,7 +86,7 @@
 	return handle_errors(v, endp, endptr);
 }
 
-long bb_strtol(const char *arg, char **endp, int base)
+long FAST_FUNC bb_strtol(const char *arg, char **endp, int base)
 {
 	long v;
 	char *endptr;
@@ -99,7 +99,7 @@
 #endif
 
 #if UINT_MAX != ULONG_MAX
-unsigned bb_strtou(const char *arg, char **endp, int base)
+unsigned FAST_FUNC bb_strtou(const char *arg, char **endp, int base)
 {
 	unsigned long v;
 	char *endptr;
@@ -111,7 +111,7 @@
 	return handle_errors(v, endp, endptr);
 }
 
-int bb_strtoi(const char *arg, char **endp, int base)
+int FAST_FUNC bb_strtoi(const char *arg, char **endp, int base)
 {
 	long v;
 	char *endptr;
@@ -131,7 +131,7 @@
 
 #include <math.h>  /* just for HUGE_VAL */
 #define NOT_DIGIT(a) (((unsigned char)(a-'0')) > 9)
-double bb_strtod(const char *arg, char **endp)
+double FAST_FUNC bb_strtod(const char *arg, char **endp)
 {
 	double v;
 	char *endptr;
diff --git a/libbb/change_identity.c b/libbb/change_identity.c
index da840bf..619db09 100644
--- a/libbb/change_identity.c
+++ b/libbb/change_identity.c
@@ -31,7 +31,7 @@
 #include "libbb.h"
 
 /* Become the user and group(s) specified by PW.  */
-void change_identity(const struct passwd *pw)
+void FAST_FUNC change_identity(const struct passwd *pw)
 {
 	if (initgroups(pw->pw_name, pw->pw_gid) == -1)
 		bb_perror_msg_and_die("can't set groups");
diff --git a/libbb/chomp.c b/libbb/chomp.c
index 8ffaff5..ed4bf6b 100644
--- a/libbb/chomp.c
+++ b/libbb/chomp.c
@@ -10,7 +10,7 @@
 
 #include "libbb.h"
 
-void chomp(char *s)
+void FAST_FUNC chomp(char *s)
 {
 	char *lc = last_char_is(s, '\n');
 
diff --git a/libbb/compare_string_array.c b/libbb/compare_string_array.c
index ec7f94e..43c59e8 100644
--- a/libbb/compare_string_array.c
+++ b/libbb/compare_string_array.c
@@ -7,7 +7,7 @@
 
 /* returns the array index of the string */
 /* (index of first match is returned, or -1) */
-int index_in_str_array(const char *const string_array[], const char *key)
+int FAST_FUNC index_in_str_array(const char *const string_array[], const char *key)
 {
 	int i;
 
@@ -19,7 +19,7 @@
 	return -1;
 }
 
-int index_in_strings(const char *strings, const char *key)
+int FAST_FUNC index_in_strings(const char *strings, const char *key)
 {
 	int idx = 0;
 
@@ -36,7 +36,7 @@
 /* returns the array index of the string, even if it matches only a beginning */
 /* (index of first match is returned, or -1) */
 #ifdef UNUSED
-int index_in_substr_array(const char *const string_array[], const char *key)
+int FAST_FUNC index_in_substr_array(const char *const string_array[], const char *key)
 {
 	int i;
 	int len = strlen(key);
@@ -51,7 +51,7 @@
 }
 #endif
 
-int index_in_substrings(const char *strings, const char *key)
+int FAST_FUNC index_in_substrings(const char *strings, const char *key)
 {
 	int len = strlen(key);
 
@@ -68,7 +68,7 @@
 	return -1;
 }
 
-const char *nth_string(const char *strings, int n)
+const char* FAST_FUNC nth_string(const char *strings, int n)
 {
 	while (n) {
 		n--;
diff --git a/libbb/concat_path_file.c b/libbb/concat_path_file.c
index dd6909f..fb53354 100644
--- a/libbb/concat_path_file.c
+++ b/libbb/concat_path_file.c
@@ -16,7 +16,7 @@
 
 #include "libbb.h"
 
-char *concat_path_file(const char *path, const char *filename)
+char* FAST_FUNC concat_path_file(const char *path, const char *filename)
 {
 	char *lc;
 
diff --git a/libbb/concat_subpath_file.c b/libbb/concat_subpath_file.c
index 1c00588..313fa63 100644
--- a/libbb/concat_subpath_file.c
+++ b/libbb/concat_subpath_file.c
@@ -15,7 +15,7 @@
 
 #include "libbb.h"
 
-char *concat_subpath_file(const char *path, const char *f)
+char* FAST_FUNC concat_subpath_file(const char *path, const char *f)
 {
 	if (f && DOT_OR_DOTDOT(f))
 		return NULL;
diff --git a/libbb/copy_file.c b/libbb/copy_file.c
index 3b83e12..d804ecc 100644
--- a/libbb/copy_file.c
+++ b/libbb/copy_file.c
@@ -71,7 +71,7 @@
  *  0 copy is made or user answered "no" in interactive mode
  *    (failures to preserve mode/owner/times are not reported in exit code)
  */
-int copy_file(const char *source, const char *dest, int flags)
+int FAST_FUNC copy_file(const char *source, const char *dest, int flags)
 {
 	/* This is a recursive function, try to minimize stack usage */
 	/* NB: each struct stat is ~100 bytes */
diff --git a/libbb/copyfd.c b/libbb/copyfd.c
index 08bc6f8..c5f8b5b 100644
--- a/libbb/copyfd.c
+++ b/libbb/copyfd.c
@@ -85,7 +85,7 @@
 
 
 #if 0
-void complain_copyfd_and_die(off_t sz)
+void FAST_FUNC complain_copyfd_and_die(off_t sz)
 {
 	if (sz != -1)
 		bb_error_msg_and_die("short read");
@@ -94,7 +94,7 @@
 }
 #endif
 
-off_t bb_copyfd_size(int fd1, int fd2, off_t size)
+off_t FAST_FUNC bb_copyfd_size(int fd1, int fd2, off_t size)
 {
 	if (size) {
 		return bb_full_fd_action(fd1, fd2, size);
@@ -102,7 +102,7 @@
 	return 0;
 }
 
-void bb_copyfd_exact_size(int fd1, int fd2, off_t size)
+void FAST_FUNC bb_copyfd_exact_size(int fd1, int fd2, off_t size)
 {
 	off_t sz = bb_copyfd_size(fd1, fd2, size);
 	if (sz == size)
@@ -113,7 +113,7 @@
 	xfunc_die();
 }
 
-off_t bb_copyfd_eof(int fd1, int fd2)
+off_t FAST_FUNC bb_copyfd_eof(int fd1, int fd2)
 {
 	return bb_full_fd_action(fd1, fd2, 0);
 }
diff --git a/libbb/correct_password.c b/libbb/correct_password.c
index f47642f..255b048 100644
--- a/libbb/correct_password.c
+++ b/libbb/correct_password.c
@@ -36,7 +36,7 @@
  *
  * NULL pw means "just fake it for login with bad username" */
 
-int correct_password(const struct passwd *pw)
+int FAST_FUNC correct_password(const struct passwd *pw)
 {
 	char *unencrypted, *encrypted;
 	const char *correct;
diff --git a/libbb/crc32.c b/libbb/crc32.c
index acbc458..42079b9 100644
--- a/libbb/crc32.c
+++ b/libbb/crc32.c
@@ -16,7 +16,7 @@
 
 #include "libbb.h"
 
-uint32_t *crc32_filltable(uint32_t *crc_table, int endian)
+uint32_t* FAST_FUNC crc32_filltable(uint32_t *crc_table, int endian)
 {
 	uint32_t polynomial = endian ? 0x04c11db7 : 0xedb88320;
 	uint32_t c;
diff --git a/libbb/create_icmp6_socket.c b/libbb/create_icmp6_socket.c
index a22ac5d..2065517 100644
--- a/libbb/create_icmp6_socket.c
+++ b/libbb/create_icmp6_socket.c
@@ -9,7 +9,7 @@
 #include "libbb.h"
 
 #if ENABLE_FEATURE_IPV6
-int create_icmp6_socket(void)
+int FAST_FUNC create_icmp6_socket(void)
 {
 	int sock;
 #if 0
diff --git a/libbb/create_icmp_socket.c b/libbb/create_icmp_socket.c
index 64beba8..1fa016a 100644
--- a/libbb/create_icmp_socket.c
+++ b/libbb/create_icmp_socket.c
@@ -8,7 +8,7 @@
 
 #include "libbb.h"
 
-int create_icmp_socket(void)
+int FAST_FUNC create_icmp_socket(void)
 {
 	int sock;
 #if 0
diff --git a/libbb/crypt_make_salt.c b/libbb/crypt_make_salt.c
index ebdf024..393eba5 100644
--- a/libbb/crypt_make_salt.c
+++ b/libbb/crypt_make_salt.c
@@ -24,7 +24,7 @@
 	return ('a' - 38 + i);
 }
 
-int crypt_make_salt(char *p, int cnt, int x)
+int FAST_FUNC crypt_make_salt(char *p, int cnt, int x)
 {
 	x += getpid() + time(NULL);
 	do {
diff --git a/libbb/device_open.c b/libbb/device_open.c
index 6907e98..cf8bcf6 100644
--- a/libbb/device_open.c
+++ b/libbb/device_open.c
@@ -10,7 +10,7 @@
 #include "libbb.h"
 
 /* try to open up the specified device */
-int device_open(const char *device, int mode)
+int FAST_FUNC device_open(const char *device, int mode)
 {
 	int m, f, fd;
 
diff --git a/libbb/die_if_bad_username.c b/libbb/die_if_bad_username.c
index 337ac60..602aadc 100644
--- a/libbb/die_if_bad_username.c
+++ b/libbb/die_if_bad_username.c
@@ -16,7 +16,7 @@
  * at the end of the username.
  */
 
-void die_if_bad_username(const char *name)
+void FAST_FUNC die_if_bad_username(const char *name)
 {
 	goto skip; /* 1st char being dash isn't valid */
 	do {
diff --git a/libbb/dump.c b/libbb/dump.c
index 4d6472e..c455952 100644
--- a/libbb/dump.c
+++ b/libbb/dump.c
@@ -33,7 +33,7 @@
 
 static const char lcc[] ALIGN1 = "diouxX";
 
-int bb_dump_size(FS * fs)
+int FAST_FUNC bb_dump_size(FS *fs)
 {
 	FU *fu;
 	int bcnt, cur_size;
@@ -652,7 +652,7 @@
 	}
 }
 
-int bb_dump_dump(char **argv)
+int FAST_FUNC bb_dump_dump(char **argv)
 {
 	FS *tfs;
 
@@ -674,7 +674,7 @@
 	return exitval;
 }
 
-void bb_dump_add(const char *fmt)
+void FAST_FUNC bb_dump_add(const char *fmt)
 {
 	const char *p;
 	char *p1;
diff --git a/libbb/error_msg.c b/libbb/error_msg.c
index 5f53f03..802fd57 100644
--- a/libbb/error_msg.c
+++ b/libbb/error_msg.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_error_msg(const char *s, ...)
+void FAST_FUNC bb_error_msg(const char *s, ...)
 {
 	va_list p;
 
diff --git a/libbb/error_msg_and_die.c b/libbb/error_msg_and_die.c
index addd818..243433b 100644
--- a/libbb/error_msg_and_die.c
+++ b/libbb/error_msg_and_die.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_error_msg_and_die(const char *s, ...)
+void FAST_FUNC bb_error_msg_and_die(const char *s, ...)
 {
 	va_list p;
 
diff --git a/libbb/execable.c b/libbb/execable.c
index 5c2b450..5c7ac16 100644
--- a/libbb/execable.c
+++ b/libbb/execable.c
@@ -13,7 +13,7 @@
  * return 1 if found;
  * return 0 otherwise;
  */
-int execable_file(const char *name)
+int FAST_FUNC execable_file(const char *name)
 {
 	struct stat s;
 	return (!access(name, X_OK) && !stat(name, &s) && S_ISREG(s.st_mode));
@@ -28,7 +28,7 @@
  * return NULL otherwise; (PATHp is undefined)
  * in all cases (*PATHp) contents will be trashed (s/:/NUL/).
  */
-char *find_execable(const char *filename, char **PATHp)
+char* FAST_FUNC find_execable(const char *filename, char **PATHp)
 {
 	char *p, *n;
 
@@ -54,7 +54,7 @@
  * return 1 if found;
  * return 0 otherwise;
  */
-int exists_execable(const char *filename)
+int FAST_FUNC exists_execable(const char *filename)
 {
 	char *path = xstrdup(getenv("PATH"));
 	char *tmp = path;
@@ -70,7 +70,7 @@
 #if ENABLE_FEATURE_PREFER_APPLETS
 /* just like the real execvp, but try to launch an applet named 'file' first
  */
-int bb_execvp(const char *file, char *const argv[])
+int FAST_FUNC bb_execvp(const char *file, char *const argv[])
 {
 	return execvp(find_applet_by_name(file) >= 0 ? bb_busybox_exec_path : file,
 					argv);
diff --git a/libbb/fclose_nonstdin.c b/libbb/fclose_nonstdin.c
index 768ee94..6f3f373 100644
--- a/libbb/fclose_nonstdin.c
+++ b/libbb/fclose_nonstdin.c
@@ -14,7 +14,7 @@
 
 #include "libbb.h"
 
-int fclose_if_not_stdin(FILE *f)
+int FAST_FUNC fclose_if_not_stdin(FILE *f)
 {
 	/* Some more paranoid applets want ferror() check too */
 	int r = ferror(f); /* NB: does NOT set errno! */
diff --git a/libbb/fflush_stdout_and_exit.c b/libbb/fflush_stdout_and_exit.c
index 9f05500..742fb9f 100644
--- a/libbb/fflush_stdout_and_exit.c
+++ b/libbb/fflush_stdout_and_exit.c
@@ -13,7 +13,7 @@
 
 #include "libbb.h"
 
-void fflush_stdout_and_exit(int retval)
+void FAST_FUNC fflush_stdout_and_exit(int retval)
 {
 	if (fflush(stdout))
 		bb_perror_msg_and_die(bb_msg_standard_output);
diff --git a/libbb/fgets_str.c b/libbb/fgets_str.c
index d6fada1..8026a15 100644
--- a/libbb/fgets_str.c
+++ b/libbb/fgets_str.c
@@ -55,12 +55,12 @@
  * including terminating string.
  * Non-terminated string can be returned if EOF is reached.
  * Return NULL if EOF is reached immediately.  */
-char *xmalloc_fgets_str(FILE *file, const char *terminating_string)
+char* FAST_FUNC xmalloc_fgets_str(FILE *file, const char *terminating_string)
 {
 	return xmalloc_fgets_internal(file, terminating_string, 0);
 }
 
-char *xmalloc_fgetline_str(FILE *file, const char *terminating_string)
+char* FAST_FUNC xmalloc_fgetline_str(FILE *file, const char *terminating_string)
 {
 	return xmalloc_fgets_internal(file, terminating_string, 1);
 }
diff --git a/libbb/find_mount_point.c b/libbb/find_mount_point.c
index cb00b98..4cd6b16 100644
--- a/libbb/find_mount_point.c
+++ b/libbb/find_mount_point.c
@@ -17,7 +17,7 @@
  * Given any other file (or directory), find the mount table entry for its
  * filesystem.
  */
-struct mntent *find_mount_point(const char *name, const char *table)
+struct mntent* FAST_FUNC find_mount_point(const char *name, const char *table)
 {
 	struct stat s;
 	dev_t mountDevice;
diff --git a/libbb/find_pid_by_name.c b/libbb/find_pid_by_name.c
index 8dcdb13..ae2f116 100644
--- a/libbb/find_pid_by_name.c
+++ b/libbb/find_pid_by_name.c
@@ -48,7 +48,7 @@
  *  Returns a list of all matching PIDs
  *  It is the caller's duty to free the returned pidlist.
  */
-pid_t* find_pid_by_name(const char* procName)
+pid_t* FAST_FUNC find_pid_by_name(const char* procName)
 {
 	pid_t* pidList;
 	int i = 0;
@@ -74,7 +74,7 @@
 	return pidList;
 }
 
-pid_t *pidlist_reverse(pid_t *pidList)
+pid_t* FAST_FUNC pidlist_reverse(pid_t *pidList)
 {
 	int i = 0;
 	while (pidList[i])
diff --git a/libbb/find_root_device.c b/libbb/find_root_device.c
index 9779f7e..ca46bf5 100644
--- a/libbb/find_root_device.c
+++ b/libbb/find_root_device.c
@@ -62,7 +62,7 @@
 	return retpath;
 }
 
-char *find_block_device(const char *path)
+char* FAST_FUNC find_block_device(const char *path)
 {
 	struct arena a;
 
diff --git a/libbb/full_write.c b/libbb/full_write.c
index 7503c8b..f353b7d 100644
--- a/libbb/full_write.c
+++ b/libbb/full_write.c
@@ -14,7 +14,7 @@
  * This does multiple writes as necessary.
  * Returns the amount written, or -1 on an error.
  */
-ssize_t full_write(int fd, const void *buf, size_t len)
+ssize_t FAST_FUNC full_write(int fd, const void *buf, size_t len)
 {
 	ssize_t cc;
 	ssize_t total;
diff --git a/libbb/get_console.c b/libbb/get_console.c
index 36fe204..d042afa 100644
--- a/libbb/get_console.c
+++ b/libbb/get_console.c
@@ -38,7 +38,7 @@
  * if someone else used X (which does a chown on /dev/console).
  */
 
-int get_console_fd(void)
+int FAST_FUNC get_console_fd(void)
 {
 	static const char *const console_names[] = {
 		DEV_CONSOLE, CURRENT_VC, CURRENT_TTY
@@ -75,7 +75,7 @@
 	VT_WAITACTIVE = 0x5607  /* wait for vt active */
 };
 
-void console_make_active(int fd, const int vt_num)
+void FAST_FUNC console_make_active(int fd, const int vt_num)
 {
 	xioctl(fd, VT_ACTIVATE, (void *)(ptrdiff_t)vt_num);
 	xioctl(fd, VT_WAITACTIVE, (void *)(ptrdiff_t)vt_num);
diff --git a/libbb/get_last_path_component.c b/libbb/get_last_path_component.c
index 0f60215..7c99116 100644
--- a/libbb/get_last_path_component.c
+++ b/libbb/get_last_path_component.c
@@ -14,7 +14,7 @@
  * "abc/def"  -> "def"
  * "abc/def/" -> ""
  */
-char *bb_get_last_path_component_nostrip(const char *path)
+char* FAST_FUNC bb_get_last_path_component_nostrip(const char *path)
 {
 	char *slash = strrchr(path, '/');
 
@@ -30,7 +30,7 @@
  * "abc/def"  -> "def"
  * "abc/def/" -> "def" !!
  */
-char *bb_get_last_path_component_strip(char *path)
+char* FAST_FUNC bb_get_last_path_component_strip(char *path)
 {
 	char *slash = last_char_is(path, '/');
 
diff --git a/libbb/get_line_from_file.c b/libbb/get_line_from_file.c
index b88872d..66ea5a1 100644
--- a/libbb/get_line_from_file.c
+++ b/libbb/get_line_from_file.c
@@ -16,7 +16,7 @@
  * must be free'ed by the caller.  If end is NULL '\n' isn't considered
  * end of line.  If end isn't NULL, length of the chunk read is stored in it.
  * Return NULL if EOF/error */
-char *bb_get_chunk_from_file(FILE *file, int *end)
+char* FAST_FUNC bb_get_chunk_from_file(FILE *file, int *end)
 {
 	int ch;
 	int idx = 0;
@@ -49,7 +49,7 @@
 }
 
 /* Get line, including trailing \n if any */
-char *xmalloc_fgets(FILE *file)
+char* FAST_FUNC xmalloc_fgets(FILE *file)
 {
 	int i;
 
@@ -57,7 +57,7 @@
 }
 
 /* Get line.  Remove trailing \n */
-char *xmalloc_fgetline(FILE *file)
+char* FAST_FUNC xmalloc_fgetline(FILE *file)
 {
 	int i;
 	char *c = bb_get_chunk_from_file(file, &i);
diff --git a/libbb/getopt32.c b/libbb/getopt32.c
index 86c3348..9dba44d 100644
--- a/libbb/getopt32.c
+++ b/libbb/getopt32.c
@@ -316,7 +316,7 @@
 
 uint32_t option_mask32;
 
-uint32_t
+uint32_t FAST_FUNC
 getopt32(char **argv, const char *applet_opts, ...)
 {
 	int argc;
diff --git a/libbb/getpty.c b/libbb/getpty.c
index d43fb82..bc143c2 100644
--- a/libbb/getpty.c
+++ b/libbb/getpty.c
@@ -10,7 +10,7 @@
 
 #define DEBUG 0
 
-int xgetpty(char *line)
+int FAST_FUNC xgetpty(char *line)
 {
 	int p;
 #if ENABLE_FEATURE_DEVPTS
diff --git a/libbb/herror_msg.c b/libbb/herror_msg.c
index 264690b..7e4f640 100644
--- a/libbb/herror_msg.c
+++ b/libbb/herror_msg.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_herror_msg(const char *s, ...)
+void FAST_FUNC bb_herror_msg(const char *s, ...)
 {
 	va_list p;
 
diff --git a/libbb/herror_msg_and_die.c b/libbb/herror_msg_and_die.c
index 894c80f..230fe64 100644
--- a/libbb/herror_msg_and_die.c
+++ b/libbb/herror_msg_and_die.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_herror_msg_and_die(const char *s, ...)
+void FAST_FUNC bb_herror_msg_and_die(const char *s, ...)
 {
 	va_list p;
 
diff --git a/libbb/human_readable.c b/libbb/human_readable.c
index d60ef61..dad26ed 100644
--- a/libbb/human_readable.c
+++ b/libbb/human_readable.c
@@ -28,7 +28,7 @@
 
 #include "libbb.h"
 
-const char *make_human_readable_str(unsigned long long size,
+const char* FAST_FUNC make_human_readable_str(unsigned long long size,
 	unsigned long block_size, unsigned long display_unit)
 {
 	/* The code will adjust for additional (appended) units */
diff --git a/libbb/inet_common.c b/libbb/inet_common.c
index 9c4f496..3a20b4a 100644
--- a/libbb/inet_common.c
+++ b/libbb/inet_common.c
@@ -11,7 +11,7 @@
 #include "libbb.h"
 #include "inet_common.h"
 
-int INET_resolve(const char *name, struct sockaddr_in *s_in, int hostfirst)
+int FAST_FUNC INET_resolve(const char *name, struct sockaddr_in *s_in, int hostfirst)
 {
 	struct hostent *hp;
 #if ENABLE_FEATURE_ETC_NETWORKS
@@ -81,7 +81,7 @@
  *          & 0x4000: host instead of net,
  *          & 0x0fff: don't resolve
  */
-char *INET_rresolve(struct sockaddr_in *s_in, int numeric, uint32_t netmask)
+char* FAST_FUNC INET_rresolve(struct sockaddr_in *s_in, int numeric, uint32_t netmask)
 {
 	/* addr-to-name cache */
 	struct addr {
@@ -165,7 +165,7 @@
 
 #if ENABLE_FEATURE_IPV6
 
-int INET6_resolve(const char *name, struct sockaddr_in6 *sin6)
+int FAST_FUNC INET6_resolve(const char *name, struct sockaddr_in6 *sin6)
 {
 	struct addrinfo req, *ai;
 	int s;
@@ -189,7 +189,7 @@
 #endif
 
 
-char *INET6_rresolve(struct sockaddr_in6 *sin6, int numeric)
+char* FAST_FUNC INET6_rresolve(struct sockaddr_in6 *sin6, int numeric)
 {
 	char name[128];
 	int s;
diff --git a/libbb/info_msg.c b/libbb/info_msg.c
index 3231bc8..ffef05e 100644
--- a/libbb/info_msg.c
+++ b/libbb/info_msg.c
@@ -10,7 +10,7 @@
 #include "libbb.h"
 #include <syslog.h>
 
-void bb_info_msg(const char *s, ...)
+void FAST_FUNC bb_info_msg(const char *s, ...)
 {
 	va_list p;
 	/* va_copy is used because it is not portable
diff --git a/libbb/inode_hash.c b/libbb/inode_hash.c
index 9cca74b..4469671 100644
--- a/libbb/inode_hash.c
+++ b/libbb/inode_hash.c
@@ -27,7 +27,7 @@
  * Return name if statbuf->st_ino && statbuf->st_dev are recorded in
  * ino_dev_hashtable, else return NULL
  */
-char *is_in_ino_dev_hashtable(const struct stat *statbuf)
+char* FAST_FUNC is_in_ino_dev_hashtable(const struct stat *statbuf)
 {
 	ino_dev_hashtable_bucket_t *bucket;
 
@@ -47,7 +47,7 @@
 }
 
 /* Add statbuf to statbuf hash table */
-void add_to_ino_dev_hashtable(const struct stat *statbuf, const char *name)
+void FAST_FUNC add_to_ino_dev_hashtable(const struct stat *statbuf, const char *name)
 {
 	int i;
 	ino_dev_hashtable_bucket_t *bucket;
@@ -69,7 +69,7 @@
 
 #if ENABLE_FEATURE_CLEAN_UP
 /* Clear statbuf hash table */
-void reset_ino_dev_hashtable(void)
+void FAST_FUNC reset_ino_dev_hashtable(void)
 {
 	int i;
 	ino_dev_hashtable_bucket_t *bucket;
diff --git a/libbb/isdirectory.c b/libbb/isdirectory.c
index 1d2477f..28ed3ec 100644
--- a/libbb/isdirectory.c
+++ b/libbb/isdirectory.c
@@ -15,7 +15,7 @@
  * Return TRUE if fileName is a directory.
  * Nonexistent files return FALSE.
  */
-int is_directory(const char *fileName, const int followLinks, struct stat *statBuf)
+int FAST_FUNC is_directory(const char *fileName, const int followLinks, struct stat *statBuf)
 {
 	int status;
 	struct stat astatBuf;
diff --git a/libbb/kernel_version.c b/libbb/kernel_version.c
index 50b82ae..8b9c4ec 100644
--- a/libbb/kernel_version.c
+++ b/libbb/kernel_version.c
@@ -16,7 +16,7 @@
  *
  *     if (get_linux_version_code() > KERNEL_VERSION(2,2,11)) { <stuff> }
  */
-int get_linux_version_code(void)
+int FAST_FUNC get_linux_version_code(void)
 {
 	struct utsname name;
 	char *s;
diff --git a/libbb/last_char_is.c b/libbb/last_char_is.c
index aaa85dd..b059256 100644
--- a/libbb/last_char_is.c
+++ b/libbb/last_char_is.c
@@ -12,7 +12,7 @@
 /* Find out if the last character of a string matches the one given.
  * Don't underrun the buffer if the string length is 0.
  */
-char* last_char_is(const char *s, int c)
+char* FAST_FUNC last_char_is(const char *s, int c)
 {
 	if (s && *s) {
 		size_t sz = strlen(s) - 1;
diff --git a/libbb/lineedit.c b/libbb/lineedit.c
index fb595c0..42f372f 100644
--- a/libbb/lineedit.c
+++ b/libbb/lineedit.c
@@ -1352,7 +1352,7 @@
  * 0  on ctrl-C (the line entered is still returned in 'command'),
  * >0 length of input string, including terminating '\n'
  */
-int read_line_input(const char *prompt, char *command, int maxsize, line_input_t *st)
+int FAST_FUNC read_line_input(const char *prompt, char *command, int maxsize, line_input_t *st)
 {
 #if ENABLE_FEATURE_TAB_COMPLETION
 	smallint lastWasTab = FALSE;
@@ -1845,7 +1845,7 @@
 	return command_len;
 }
 
-line_input_t *new_line_input_t(int flags)
+line_input_t* FAST_FUNC new_line_input_t(int flags)
 {
 	line_input_t *n = xzalloc(sizeof(*n));
 	n->flags = flags;
@@ -1855,7 +1855,7 @@
 #else
 
 #undef read_line_input
-int read_line_input(const char* prompt, char* command, int maxsize)
+int FAST_FUNC read_line_input(const char* prompt, char* command, int maxsize)
 {
 	fputs(prompt, stdout);
 	fflush(stdout);
diff --git a/libbb/llist.c b/libbb/llist.c
index 4b3971b..094c652 100644
--- a/libbb/llist.c
+++ b/libbb/llist.c
@@ -13,7 +13,7 @@
 #include "libbb.h"
 
 /* Add data to the start of the linked list.  */
-void llist_add_to(llist_t **old_head, void *data)
+void FAST_FUNC llist_add_to(llist_t **old_head, void *data)
 {
 	llist_t *new_head = xmalloc(sizeof(llist_t));
 
@@ -23,7 +23,7 @@
 }
 
 /* Add data to the end of the linked list.  */
-void llist_add_to_end(llist_t **list_head, void *data)
+void FAST_FUNC llist_add_to_end(llist_t **list_head, void *data)
 {
 	llist_t *new_item = xmalloc(sizeof(llist_t));
 
@@ -42,7 +42,7 @@
 }
 
 /* Remove first element from the list and return it */
-void *llist_pop(llist_t **head)
+void* FAST_FUNC llist_pop(llist_t **head)
 {
 	void *data, *next;
 
@@ -58,7 +58,7 @@
 }
 
 /* Unlink arbitrary given element from the list */
-void llist_unlink(llist_t **head, llist_t *elm)
+void FAST_FUNC llist_unlink(llist_t **head, llist_t *elm)
 {
 	llist_t *crt;
 
@@ -80,7 +80,7 @@
 
 /* Recursively free all elements in the linked list.  If freeit != NULL
  * call it on each datum in the list */
-void llist_free(llist_t *elm, void (*freeit) (void *data))
+void FAST_FUNC llist_free(llist_t *elm, void (*freeit) (void *data))
 {
 	while (elm) {
 		void *data = llist_pop(&elm);
@@ -92,7 +92,7 @@
 
 #ifdef UNUSED
 /* Reverse list order. */
-llist_t *llist_rev(llist_t *list)
+llist_t* FAST_FUNC llist_rev(llist_t *list)
 {
 	llist_t *rev = NULL;
 
diff --git a/libbb/login.c b/libbb/login.c
index a5be2c8..346ff13 100644
--- a/libbb/login.c
+++ b/libbb/login.c
@@ -18,7 +18,7 @@
 static const char fmtstr_d[] ALIGN1 = "%A, %d %B %Y";
 static const char fmtstr_t[] ALIGN1 = "%H:%M:%S";
 
-void print_login_issue(const char *issue_file, const char *tty)
+void FAST_FUNC print_login_issue(const char *issue_file, const char *tty)
 {
 	FILE *fd;
 	int c;
@@ -86,7 +86,7 @@
 	fflush(stdout);
 }
 
-void print_login_prompt(void)
+void FAST_FUNC print_login_prompt(void)
 {
 	char *hostname = safe_gethostname();
 
@@ -112,7 +112,7 @@
 	"LD_NOWARN" "\0"
 	"LD_KEEPDIR" "\0";
 
-int sanitize_env_if_suid(void)
+int FAST_FUNC sanitize_env_if_suid(void)
 {
 	const char *p;
 
diff --git a/libbb/loop.c b/libbb/loop.c
index 6934b7a..7d2b420 100644
--- a/libbb/loop.c
+++ b/libbb/loop.c
@@ -44,7 +44,7 @@
 } bb_loop_info;
 #endif
 
-char *query_loop(const char *device)
+char* FAST_FUNC query_loop(const char *device)
 {
 	int fd;
 	bb_loop_info loopinfo;
@@ -61,7 +61,7 @@
 }
 
 
-int del_loop(const char *device)
+int FAST_FUNC del_loop(const char *device)
 {
 	int fd, rc;
 
@@ -79,7 +79,7 @@
    search will re-use an existing loop device already bound to that
    file/offset if it finds one.
  */
-int set_loop(char **device, const char *file, unsigned long long offset)
+int FAST_FUNC set_loop(char **device, const char *file, unsigned long long offset)
 {
 	char dev[LOOP_NAMESIZE];
 	char *try;
diff --git a/libbb/make_directory.c b/libbb/make_directory.c
index 8841c95..5c71aff 100644
--- a/libbb/make_directory.c
+++ b/libbb/make_directory.c
@@ -26,7 +26,7 @@
 
 /* This function is used from NOFORK applets. It must not allocate anything */
 
-int bb_make_directory (char *path, long mode, int flags)
+int FAST_FUNC bb_make_directory(char *path, long mode, int flags)
 {
 	mode_t mask;
 	const char *fail_msg;
diff --git a/libbb/match_fstype.c b/libbb/match_fstype.c
index bd4dbb0..99e2767 100644
--- a/libbb/match_fstype.c
+++ b/libbb/match_fstype.c
@@ -12,7 +12,7 @@
 
 #include "libbb.h"
 
-int match_fstype(const struct mntent *mt, const char *fstype)
+int FAST_FUNC match_fstype(const struct mntent *mt, const char *fstype)
 {
 	int no = 0;
 	int len;
diff --git a/libbb/md5.c b/libbb/md5.c
index 8d4b9fe..4ab06eb 100644
--- a/libbb/md5.c
+++ b/libbb/md5.c
@@ -24,7 +24,7 @@
 /* Initialize structure containing state of computation.
  * (RFC 1321, 3.3: Step 3)
  */
-void md5_begin(md5_ctx_t *ctx)
+void FAST_FUNC md5_begin(md5_ctx_t *ctx)
 {
 	ctx->A = 0x67452301;
 	ctx->B = 0xefcdab89;
@@ -371,7 +371,7 @@
  * This function's internal buffer remembers previous data until it has 64
  * bytes worth to pass on.  Call md5_end() to flush this buffer. */
 
-void md5_hash(const void *buffer, size_t len, md5_ctx_t *ctx)
+void FAST_FUNC md5_hash(const void *buffer, size_t len, md5_ctx_t *ctx)
 {
 	char *buf=(char *)buffer;
 
@@ -410,7 +410,7 @@
  * IMPORTANT: On some systems it is required that RESBUF is correctly
  * aligned for a 32 bits value.
  */
-void *md5_end(void *resbuf, md5_ctx_t *ctx)
+void* FAST_FUNC md5_end(void *resbuf, md5_ctx_t *ctx)
 {
 	char *buf = ctx->buffer;
 	int i;
diff --git a/libbb/mode_string.c b/libbb/mode_string.c
index d17cc4a..b9975f4 100644
--- a/libbb/mode_string.c
+++ b/libbb/mode_string.c
@@ -51,7 +51,7 @@
 /*                                  0123456789abcdef */
 static const char mode_chars[7] ALIGN1 = "rwxSTst";
 
-const char *bb_mode_string(mode_t mode)
+const char* FAST_FUNC bb_mode_string(mode_t mode)
 {
 	static char buf[12];
 	char *p = buf;
@@ -91,7 +91,7 @@
 /*                                  0123456789abcdef */
 static const char mode_chars[7] = "rwxSTst";
 
-const char *bb_mode_string(mode_t mode)
+const char* FAST_FUNC bb_mode_string(mode_t mode)
 {
 	static char buf[12];
 	char *p = buf;
diff --git a/libbb/mtab.c b/libbb/mtab.c
index 18386ef..57654a6 100644
--- a/libbb/mtab.c
+++ b/libbb/mtab.c
@@ -11,7 +11,7 @@
 #include "libbb.h"
 
 #if ENABLE_FEATURE_MTAB_SUPPORT
-void erase_mtab(const char *name)
+void FAST_FUNC erase_mtab(const char *name)
 {
 	struct mntent *entries = NULL;
 	int i, count = 0;
diff --git a/libbb/obscure.c b/libbb/obscure.c
index 1841b27..19b8752 100644
--- a/libbb/obscure.c
+++ b/libbb/obscure.c
@@ -157,7 +157,7 @@
 	return NULL;
 }
 
-int obscure(const char *old, const char *newval, const struct passwd *pw)
+int FAST_FUNC obscure(const char *old, const char *newval, const struct passwd *pw)
 {
 	const char *msg;
 
diff --git a/libbb/parse_mode.c b/libbb/parse_mode.c
index fd54900..40105dd 100644
--- a/libbb/parse_mode.c
+++ b/libbb/parse_mode.c
@@ -15,7 +15,7 @@
 
 #define FILEMODEBITS (S_ISUID | S_ISGID | S_ISVTX | S_IRWXU | S_IRWXG | S_IRWXO)
 
-int bb_parse_mode(const char *s, mode_t *current_mode)
+int FAST_FUNC bb_parse_mode(const char *s, mode_t *current_mode)
 {
 	static const mode_t who_mask[] = {
 		S_ISUID | S_ISGID | S_ISVTX | S_IRWXU | S_IRWXG | S_IRWXO, /* a */
diff --git a/libbb/perror_msg.c b/libbb/perror_msg.c
index af9ff59..6c8e1b5 100644
--- a/libbb/perror_msg.c
+++ b/libbb/perror_msg.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_perror_msg(const char *s, ...)
+void FAST_FUNC bb_perror_msg(const char *s, ...)
 {
 	va_list p;
 
@@ -19,7 +19,7 @@
 	va_end(p);
 }
 
-void bb_simple_perror_msg(const char *s)
+void FAST_FUNC bb_simple_perror_msg(const char *s)
 {
 	bb_perror_msg("%s", s);
 }
diff --git a/libbb/perror_msg_and_die.c b/libbb/perror_msg_and_die.c
index 7b50073..15615fa 100644
--- a/libbb/perror_msg_and_die.c
+++ b/libbb/perror_msg_and_die.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_perror_msg_and_die(const char *s, ...)
+void FAST_FUNC bb_perror_msg_and_die(const char *s, ...)
 {
 	va_list p;
 
@@ -20,7 +20,7 @@
 	xfunc_die();
 }
 
-void bb_simple_perror_msg_and_die(const char *s)
+void FAST_FUNC bb_simple_perror_msg_and_die(const char *s)
 {
 	bb_perror_msg_and_die("%s", s);
 }
diff --git a/libbb/perror_nomsg.c b/libbb/perror_nomsg.c
index 62ce888..a157caa 100644
--- a/libbb/perror_nomsg.c
+++ b/libbb/perror_nomsg.c
@@ -11,11 +11,12 @@
  * modified definition without "attribute (format)"
  * instead of including libbb.h */
 //#include "libbb.h"
-extern void bb_perror_msg(const char *s, ...);
+#include "platform.h"
+extern void bb_perror_msg(const char *s, ...) FAST_FUNC;
 
 /* suppress gcc "no previous prototype" warning */
-void bb_perror_nomsg(void);
-void bb_perror_nomsg(void)
+void FAST_FUNC bb_perror_nomsg(void);
+void FAST_FUNC bb_perror_nomsg(void)
 {
 	bb_perror_msg(0);
 }
diff --git a/libbb/perror_nomsg_and_die.c b/libbb/perror_nomsg_and_die.c
index dab3df6..d56e05d 100644
--- a/libbb/perror_nomsg_and_die.c
+++ b/libbb/perror_nomsg_and_die.c
@@ -11,11 +11,12 @@
  * modified definition without "attribute (format)"
  * instead of including libbb.h */
 //#include "libbb.h"
-extern void bb_perror_msg_and_die(const char *s, ...);
+#include "platform.h"
+extern void bb_perror_msg_and_die(const char *s, ...) FAST_FUNC;
 
 /* suppress gcc "no previous prototype" warning */
-void bb_perror_nomsg_and_die(void);
-void bb_perror_nomsg_and_die(void)
+void FAST_FUNC bb_perror_nomsg_and_die(void);
+void FAST_FUNC bb_perror_nomsg_and_die(void)
 {
 	bb_perror_msg_and_die(0);
 }
diff --git a/libbb/pidfile.c b/libbb/pidfile.c
index cafa789..7b8fee2 100644
--- a/libbb/pidfile.c
+++ b/libbb/pidfile.c
@@ -13,7 +13,7 @@
 
 smallint wrote_pidfile;
 
-void write_pidfile(const char *path)
+void FAST_FUNC write_pidfile(const char *path)
 {
 	int pid_fd;
 	char *end;
diff --git a/libbb/print_flags.c b/libbb/print_flags.c
index a1dcc01..a8c4f9c 100644
--- a/libbb/print_flags.c
+++ b/libbb/print_flags.c
@@ -9,7 +9,7 @@
 #include <libbb.h>
 
 /* returns a set with the flags not printed */
-int print_flags_separated(const int *masks, const char *labels, int flags, const char *separator)
+int FAST_FUNC print_flags_separated(const int *masks, const char *labels, int flags, const char *separator)
 {
 	const char *need_separator = NULL;
 	while (*labels) {
@@ -19,14 +19,14 @@
 				labels);
 			need_separator = separator;
 			flags &= ~ *masks;
+			masks++;
 		}
-		masks++;
 		labels += strlen(labels) + 1;
 	}
 	return flags;
 }
 
-int print_flags(const masks_labels_t *ml, int flags)
+int FAST_FUNC print_flags(const masks_labels_t *ml, int flags)
 {
 	return print_flags_separated(ml->masks, ml->labels, flags, NULL);
 }
diff --git a/libbb/printable.c b/libbb/printable.c
index 676758a..ae93359 100644
--- a/libbb/printable.c
+++ b/libbb/printable.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void fputc_printable(int ch, FILE *file)
+void FAST_FUNC fputc_printable(int ch, FILE *file)
 {
 	if ((ch & (0x80 + PRINTABLE_META)) == (0x80 + PRINTABLE_META)) {
 		fputs("M-", file);
diff --git a/libbb/process_escape_sequence.c b/libbb/process_escape_sequence.c
index 1cadbd3..4d03bd6 100644
--- a/libbb/process_escape_sequence.c
+++ b/libbb/process_escape_sequence.c
@@ -16,7 +16,7 @@
 #undef _tolower
 #define _tolower(X) ((X)|((char) 0x20))
 
-char bb_process_escape_sequence(const char **ptr)
+char FAST_FUNC bb_process_escape_sequence(const char **ptr)
 {
 	static const char charmap[] ALIGN1 = {
 		'a',  'b',  'f',  'n',  'r',  't',  'v',  '\\', 0,
diff --git a/libbb/procps.c b/libbb/procps.c
index 8946917..7d49d83 100644
--- a/libbb/procps.c
+++ b/libbb/procps.c
@@ -30,7 +30,7 @@
 	cp->cache = NULL;
 	cp->size = 0;
 }
-void clear_username_cache(void)
+void FAST_FUNC clear_username_cache(void)
 {
 	clear_cache(&username);
 	clear_cache(&groupname);
@@ -52,7 +52,7 @@
 }
 #endif
 
-typedef char* ug_func(char *name, int bufsize, long uid);
+typedef char* FAST_FUNC ug_func(char *name, int bufsize, long uid);
 static char* get_cached(cache_t *cp, unsigned id, ug_func* fp)
 {
 	int i;
@@ -66,11 +66,11 @@
 	fp(cp->cache[i].name, sizeof(cp->cache[i].name), id);
 	return cp->cache[i].name;
 }
-const char* get_cached_username(uid_t uid)
+const char* FAST_FUNC get_cached_username(uid_t uid)
 {
 	return get_cached(&username, uid, bb_getpwuid);
 }
-const char* get_cached_groupname(gid_t gid)
+const char* FAST_FUNC get_cached_groupname(gid_t gid)
 {
 	return get_cached(&groupname, gid, bb_getgrgid);
 }
@@ -78,7 +78,7 @@
 
 #define PROCPS_BUFSIZE 1024
 
-static int read_to_buf(const char *filename, void *buf)
+static int FAST_FUNC read_to_buf(const char *filename, void *buf)
 {
 	int fd;
 	/* open_read_close() would do two reads, checking for EOF.
@@ -93,7 +93,7 @@
 	return ret;
 }
 
-static procps_status_t *alloc_procps_scan(void)
+static procps_status_t* FAST_FUNC alloc_procps_scan(void)
 {
 	unsigned n = getpagesize();
 	procps_status_t* sp = xzalloc(sizeof(procps_status_t));
@@ -107,7 +107,7 @@
 	return sp;
 }
 
-void free_procps_scan(procps_status_t* sp)
+void FAST_FUNC free_procps_scan(procps_status_t* sp)
 {
 	closedir(sp->dir);
 	free(sp->argv0);
@@ -163,7 +163,7 @@
 #endif
 
 void BUG_comm_size(void);
-procps_status_t *procps_scan(procps_status_t* sp, int flags)
+procps_status_t* FAST_FUNC procps_scan(procps_status_t* sp, int flags)
 {
 	struct dirent *entry;
 	char buf[PROCPS_BUFSIZE];
@@ -402,7 +402,7 @@
 	return sp;
 }
 
-void read_cmdline(char *buf, int col, unsigned pid, const char *comm)
+void FAST_FUNC read_cmdline(char *buf, int col, unsigned pid, const char *comm)
 {
 	ssize_t sz;
 	char filename[sizeof("/proc//cmdline") + sizeof(int)*3];
diff --git a/libbb/pw_encrypt.c b/libbb/pw_encrypt.c
index 7363186..469e71f 100644
--- a/libbb/pw_encrypt.c
+++ b/libbb/pw_encrypt.c
@@ -50,7 +50,7 @@
 	des_ctx = NULL;
 }
 
-char *pw_encrypt(const char *clear, const char *salt, int cleanup)
+char* FAST_FUNC pw_encrypt(const char *clear, const char *salt, int cleanup)
 {
 	char *encrypted;
 
@@ -70,7 +70,7 @@
 
 #else /* if !ENABLE_USE_BB_CRYPT */
 
-char *pw_encrypt(const char *clear, const char *salt, int cleanup)
+char* FAST_FUNC pw_encrypt(const char *clear, const char *salt, int cleanup)
 {
 #if 0 /* was CONFIG_FEATURE_SHA1_PASSWORDS, but there is no such thing??? */
 	if (strncmp(salt, "$2$", 3) == 0) {
diff --git a/libbb/read.c b/libbb/read.c
index fa9874d..7b80412 100644
--- a/libbb/read.c
+++ b/libbb/read.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-ssize_t safe_read(int fd, void *buf, size_t count)
+ssize_t FAST_FUNC safe_read(int fd, void *buf, size_t count)
 {
 	ssize_t n;
 
@@ -56,7 +56,7 @@
  * which detects EAGAIN and uses poll() to wait on the fd.
  * Thankfully, poll() doesn't care about O_NONBLOCK flag.
  */
-ssize_t nonblock_safe_read(int fd, void *buf, size_t count)
+ssize_t FAST_FUNC nonblock_safe_read(int fd, void *buf, size_t count)
 {
 	struct pollfd pfd[1];
 	ssize_t n;
@@ -78,7 +78,7 @@
  * Returns the amount read, or -1 on an error.
  * A short read is returned on an end of file.
  */
-ssize_t full_read(int fd, void *buf, size_t len)
+ssize_t FAST_FUNC full_read(int fd, void *buf, size_t len)
 {
 	ssize_t cc;
 	ssize_t total;
@@ -107,7 +107,7 @@
 }
 
 // Die with an error message if we can't read the entire buffer.
-void xread(int fd, void *buf, size_t count)
+void FAST_FUNC xread(int fd, void *buf, size_t count)
 {
 	if (count) {
 		ssize_t size = full_read(fd, buf, count);
@@ -117,7 +117,7 @@
 }
 
 // Die with an error message if we can't read one character.
-unsigned char xread_char(int fd)
+unsigned char FAST_FUNC xread_char(int fd)
 {
 	char tmp;
 	xread(fd, &tmp, 1);
@@ -125,7 +125,7 @@
 }
 
 // Read one line a-la fgets. Works only on seekable streams
-char *reads(int fd, char *buffer, size_t size)
+char* FAST_FUNC reads(int fd, char *buffer, size_t size)
 {
 	char *p;
 
@@ -152,7 +152,7 @@
 // Reads one line a-la fgets (but doesn't save terminating '\n').
 // Reads byte-by-byte. Useful when it is important to not read ahead.
 // Bytes are appended to pfx (which must be malloced, or NULL).
-char *xmalloc_reads(int fd, char *buf, size_t *maxsz_p)
+char* FAST_FUNC xmalloc_reads(int fd, char *buf, size_t *maxsz_p)
 {
 	char *p;
 	size_t sz = buf ? strlen(buf) : 0;
@@ -185,7 +185,7 @@
 	return xrealloc(buf, p - buf);
 }
 
-ssize_t read_close(int fd, void *buf, size_t size)
+ssize_t FAST_FUNC read_close(int fd, void *buf, size_t size)
 {
 	/*int e;*/
 	size = full_read(fd, buf, size);
@@ -195,7 +195,7 @@
 	return size;
 }
 
-ssize_t open_read_close(const char *filename, void *buf, size_t size)
+ssize_t FAST_FUNC open_read_close(const char *filename, void *buf, size_t size)
 {
 	int fd = open(filename, O_RDONLY);
 	if (fd < 0)
@@ -205,7 +205,7 @@
 
 // Read (potentially big) files in one go. File size is estimated
 // by stat.
-void *xmalloc_open_read_close(const char *filename, size_t *sizep)
+void* FAST_FUNC xmalloc_open_read_close(const char *filename, size_t *sizep)
 {
 	char *buf;
 	size_t size;
@@ -247,7 +247,7 @@
 
 // Read (potentially big) files in one go. File size is estimated by
 // lseek to end.
-void *xmalloc_open_read_close(const char *filename, size_t *sizep)
+void* FAST_FUNC xmalloc_open_read_close(const char *filename, size_t *sizep)
 {
 	char *buf;
 	size_t size;
@@ -284,7 +284,7 @@
 }
 #endif
 
-void *xmalloc_xopen_read_close(const char *filename, size_t *sizep)
+void* FAST_FUNC xmalloc_xopen_read_close(const char *filename, size_t *sizep)
 {
 	void *buf = xmalloc_open_read_close(filename, sizep);
 	if (!buf)
diff --git a/libbb/recursive_action.c b/libbb/recursive_action.c
index fe9ba2e..9b6951f 100644
--- a/libbb/recursive_action.c
+++ b/libbb/recursive_action.c
@@ -22,7 +22,7 @@
  * is so stinking huge.
  */
 
-static int true_action(const char *fileName ATTRIBUTE_UNUSED,
+static int FAST_FUNC true_action(const char *fileName ATTRIBUTE_UNUSED,
 		struct stat *statbuf ATTRIBUTE_UNUSED,
 		void* userData ATTRIBUTE_UNUSED,
 		int depth ATTRIBUTE_UNUSED)
@@ -53,10 +53,10 @@
  * 1: stat(statbuf). Calls dirAction and optionally recurse on link to dir.
  */
 
-int recursive_action(const char *fileName,
+int FAST_FUNC recursive_action(const char *fileName,
 		unsigned flags,
-		int (*fileAction)(const char *fileName, struct stat *statbuf, void* userData, int depth),
-		int (*dirAction)(const char *fileName, struct stat *statbuf, void* userData, int depth),
+		int FAST_FUNC (*fileAction)(const char *fileName, struct stat *statbuf, void* userData, int depth),
+		int FAST_FUNC (*dirAction)(const char *fileName, struct stat *statbuf, void* userData, int depth),
 		void* userData,
 		unsigned depth)
 {
diff --git a/libbb/remove_file.c b/libbb/remove_file.c
index 21878dc..8b14f07 100644
--- a/libbb/remove_file.c
+++ b/libbb/remove_file.c
@@ -11,7 +11,7 @@
 
 /* Used from NOFORK applets. Must not allocate anything */
 
-int remove_file(const char *path, int flags)
+int FAST_FUNC remove_file(const char *path, int flags)
 {
 	struct stat path_stat;
 
diff --git a/libbb/restricted_shell.c b/libbb/restricted_shell.c
index dc4cfb4..2a5073f 100644
--- a/libbb/restricted_shell.c
+++ b/libbb/restricted_shell.c
@@ -32,7 +32,7 @@
 
 /* Return 1 if SHELL is a restricted shell (one not returned by
    getusershell), else 0, meaning it is a standard shell.  */
-int restricted_shell(const char *shell)
+int FAST_FUNC restricted_shell(const char *shell)
 {
 	char *line;
 
diff --git a/libbb/rtc.c b/libbb/rtc.c
index 78f10c6..1fdeee8 100644
--- a/libbb/rtc.c
+++ b/libbb/rtc.c
@@ -11,7 +11,7 @@
 # define ADJTIME_PATH "/etc/adjtime"
 #endif
 
-int rtc_adjtime_is_utc(void)
+int FAST_FUNC rtc_adjtime_is_utc(void)
 {
 	int utc = 0;
 	FILE *f = fopen(ADJTIME_PATH, "r");
@@ -40,7 +40,7 @@
 	return utc;
 }
 
-int rtc_xopen(const char **default_rtc, int flags)
+int FAST_FUNC rtc_xopen(const char **default_rtc, int flags)
 {
 	int rtc;
 
@@ -59,7 +59,7 @@
 	return xopen(*default_rtc, flags);
 }
 
-time_t rtc_read_time(int fd, int utc)
+time_t FAST_FUNC rtc_read_time(int fd, int utc)
 {
 	struct tm tm;
 	char *oldtz = 0;
diff --git a/libbb/run_shell.c b/libbb/run_shell.c
index 239887d..2ccb3a1 100644
--- a/libbb/run_shell.c
+++ b/libbb/run_shell.c
@@ -36,14 +36,12 @@
 #if ENABLE_SELINUX
 static security_context_t current_sid;
 
-void
-renew_current_security_context(void)
+void FAST_FUNC renew_current_security_context(void)
 {
 	freecon(current_sid);  /* Release old context  */
 	getcon(&current_sid);  /* update */
 }
-void
-set_current_security_context(security_context_t sid)
+void FAST_FUNC set_current_security_context(security_context_t sid)
 {
 	freecon(current_sid);  /* Release old context  */
 	current_sid = sid;
@@ -56,7 +54,7 @@
    If ADDITIONAL_ARGS is nonzero, pass it to the shell as more
    arguments.  */
 
-void run_shell(const char *shell, int loginshell, const char *command, const char **additional_args)
+void FAST_FUNC run_shell(const char *shell, int loginshell, const char *command, const char **additional_args)
 {
 	const char **args;
 	int argno = 1;
diff --git a/libbb/safe_gethostname.c b/libbb/safe_gethostname.c
index 3b24edb..1f8b2a8 100644
--- a/libbb/safe_gethostname.c
+++ b/libbb/safe_gethostname.c
@@ -33,7 +33,7 @@
  * This is an illegal first character for a hostname.
  * The returned malloced string must be freed by the caller.
  */
-char *safe_gethostname(void)
+char* FAST_FUNC safe_gethostname(void)
 {
 	struct utsname uts;
 
diff --git a/libbb/safe_poll.c b/libbb/safe_poll.c
index d2b773c..58c7bda 100644
--- a/libbb/safe_poll.c
+++ b/libbb/safe_poll.c
@@ -12,7 +12,7 @@
 /* Wrapper which restarts poll on EINTR or ENOMEM.
  * On other errors does perror("poll") and returns.
  * Warning! May take longer than timeout_ms to return! */
-int safe_poll(struct pollfd *ufds, nfds_t nfds, int timeout)
+int FAST_FUNC safe_poll(struct pollfd *ufds, nfds_t nfds, int timeout)
 {
 	while (1) {
 		int n = poll(ufds, nfds, timeout);
diff --git a/libbb/safe_strncpy.c b/libbb/safe_strncpy.c
index cc42583..649fa10 100644
--- a/libbb/safe_strncpy.c
+++ b/libbb/safe_strncpy.c
@@ -10,7 +10,7 @@
 #include "libbb.h"
 
 /* Like strncpy but make sure the resulting string is always 0 terminated. */
-char *safe_strncpy(char *dst, const char *src, size_t size)
+char* FAST_FUNC safe_strncpy(char *dst, const char *src, size_t size)
 {
 	if (!size) return dst;
 	dst[--size] = '\0';
diff --git a/libbb/safe_write.c b/libbb/safe_write.c
index 5bbb82e..e3561f3 100644
--- a/libbb/safe_write.c
+++ b/libbb/safe_write.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-ssize_t safe_write(int fd, const void *buf, size_t count)
+ssize_t FAST_FUNC safe_write(int fd, const void *buf, size_t count)
 {
 	ssize_t n;
 
diff --git a/libbb/selinux_common.c b/libbb/selinux_common.c
index 7478cc7..5fdbe9d 100644
--- a/libbb/selinux_common.c
+++ b/libbb/selinux_common.c
@@ -7,7 +7,7 @@
 #include "libbb.h"
 #include <selinux/context.h>
 
-context_t set_security_context_component(security_context_t cur_context,
+context_t FAST_FUNC set_security_context_component(security_context_t cur_context,
 					 char *user, char *role, char *type, char *range)
 {
 	context_t con = context_new(cur_context);
@@ -29,7 +29,7 @@
 	return NULL;
 }
 
-void setfscreatecon_or_die(security_context_t scontext)
+void FAST_FUNC setfscreatecon_or_die(security_context_t scontext)
 {
 	if (setfscreatecon(scontext) < 0) {
 		/* Can be NULL. All known printf implementations
@@ -39,7 +39,7 @@
 	}
 }
 
-void selinux_preserve_fcontext(int fdesc)
+void FAST_FUNC selinux_preserve_fcontext(int fdesc)
 {
 	security_context_t context;
 
diff --git a/libbb/setup_environment.c b/libbb/setup_environment.c
index 6e3575c..04e333f 100644
--- a/libbb/setup_environment.c
+++ b/libbb/setup_environment.c
@@ -30,7 +30,7 @@
 
 #include "libbb.h"
 
-void setup_environment(const char *shell, int clear_env, int change_env, const struct passwd *pw)
+void FAST_FUNC setup_environment(const char *shell, int clear_env, int change_env, const struct passwd *pw)
 {
 	if (clear_env) {
 		const char *term;
diff --git a/libbb/sha1.c b/libbb/sha1.c
index 552dcad..cc7edd8 100644
--- a/libbb/sha1.c
+++ b/libbb/sha1.c
@@ -83,7 +83,7 @@
 	ctx->hash[4] += e;
 }
 
-void sha1_begin(sha1_ctx_t *ctx)
+void FAST_FUNC sha1_begin(sha1_ctx_t *ctx)
 {
 	ctx->count[0] = ctx->count[1] = 0;
 	ctx->hash[0] = 0x67452301;
@@ -95,7 +95,7 @@
 
 /* SHA1 hash data in an array of bytes into hash buffer and call the        */
 /* hash_compile function as required.                                       */
-void sha1_hash(const void *data, size_t length, sha1_ctx_t *ctx)
+void FAST_FUNC sha1_hash(const void *data, size_t length, sha1_ctx_t *ctx)
 {
 	uint32_t pos = (uint32_t) (ctx->count[0] & SHA1_MASK);
 	uint32_t freeb = SHA1_BLOCK_SIZE - pos;
@@ -116,7 +116,7 @@
 	memcpy(((unsigned char *) ctx->wbuf) + pos, sp, length);
 }
 
-void *sha1_end(void *resbuf, sha1_ctx_t *ctx)
+void* FAST_FUNC sha1_end(void *resbuf, sha1_ctx_t *ctx)
 {
 	/* SHA1 Final padding and digest calculation  */
 #if BB_BIG_ENDIAN
diff --git a/libbb/signals.c b/libbb/signals.c
index 1342c89..9591146 100644
--- a/libbb/signals.c
+++ b/libbb/signals.c
@@ -12,19 +12,19 @@
 #include "libbb.h"
 
 /* Saves 2 bytes on x86! Oh my... */
-int sigaction_set(int signum, const struct sigaction *act)
+int FAST_FUNC sigaction_set(int signum, const struct sigaction *act)
 {
 	return sigaction(signum, act, NULL);
 }
 
-int sigprocmask_allsigs(int how)
+int FAST_FUNC sigprocmask_allsigs(int how)
 {
 	sigset_t set;
 	sigfillset(&set);
 	return sigprocmask(how, &set, NULL);
 }
 
-void bb_signals(int sigs, void (*f)(int))
+void FAST_FUNC bb_signals(int sigs, void (*f)(int))
 {
 	int sig_no = 0;
 	int bit = 1;
@@ -39,7 +39,7 @@
 	}
 }
 
-void bb_signals_recursive(int sigs, void (*f)(int))
+void FAST_FUNC bb_signals_recursive(int sigs, void (*f)(int))
 {
 	int sig_no = 0;
 	int bit = 1;
@@ -60,7 +60,7 @@
 	}
 }
 
-void sig_block(int sig)
+void FAST_FUNC sig_block(int sig)
 {
 	sigset_t ss;
 	sigemptyset(&ss);
@@ -68,7 +68,7 @@
 	sigprocmask(SIG_BLOCK, &ss, NULL);
 }
 
-void sig_unblock(int sig)
+void FAST_FUNC sig_unblock(int sig)
 {
 	sigset_t ss;
 	sigemptyset(&ss);
@@ -76,7 +76,7 @@
 	sigprocmask(SIG_UNBLOCK, &ss, NULL);
 }
 
-void wait_for_any_sig(void)
+void FAST_FUNC wait_for_any_sig(void)
 {
 	sigset_t ss;
 	sigemptyset(&ss);
@@ -84,7 +84,7 @@
 }
 
 /* Assuming the sig is fatal */
-void kill_myself_with_sig(int sig)
+void FAST_FUNC kill_myself_with_sig(int sig)
 {
 	signal(sig, SIG_DFL);
 	sig_unblock(sig);
@@ -92,7 +92,7 @@
 	_exit(EXIT_FAILURE); /* Should not reach it */
 }
 
-void signal_SA_RESTART_empty_mask(int sig, void (*handler)(int))
+void FAST_FUNC signal_SA_RESTART_empty_mask(int sig, void (*handler)(int))
 {
 	struct sigaction sa;
 	memset(&sa, 0, sizeof(sa));
@@ -102,7 +102,7 @@
 	sigaction_set(sig, &sa);
 }
 
-void signal_no_SA_RESTART_empty_mask(int sig, void (*handler)(int))
+void FAST_FUNC signal_no_SA_RESTART_empty_mask(int sig, void (*handler)(int))
 {
 	struct sigaction sa;
 	memset(&sa, 0, sizeof(sa));
diff --git a/libbb/simplify_path.c b/libbb/simplify_path.c
index 29e371d..367f1f0 100644
--- a/libbb/simplify_path.c
+++ b/libbb/simplify_path.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-char *bb_simplify_path(const char *path)
+char* FAST_FUNC bb_simplify_path(const char *path)
 {
 	char *s, *start, *p;
 
diff --git a/libbb/skip_whitespace.c b/libbb/skip_whitespace.c
index 87b5f23..e85f385 100644
--- a/libbb/skip_whitespace.c
+++ b/libbb/skip_whitespace.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-char *skip_whitespace(const char *s)
+char* FAST_FUNC skip_whitespace(const char *s)
 {
 	/* NB: isspace('\0') returns 0 */
 	while (isspace(*s)) ++s;
@@ -17,7 +17,7 @@
 	return (char *) s;
 }
 
-char *skip_non_whitespace(const char *s)
+char* FAST_FUNC skip_non_whitespace(const char *s)
 {
 	while (*s && !isspace(*s)) ++s;
 
diff --git a/libbb/speed_table.c b/libbb/speed_table.c
index 94a2962..646f914 100644
--- a/libbb/speed_table.c
+++ b/libbb/speed_table.c
@@ -56,7 +56,7 @@
 
 enum { NUM_SPEEDS = ARRAY_SIZE(speeds) };
 
-unsigned int tty_baud_to_value(speed_t speed)
+unsigned FAST_FUNC tty_baud_to_value(speed_t speed)
 {
 	int i = 0;
 
@@ -72,7 +72,7 @@
 	return 0;
 }
 
-speed_t tty_value_to_baud(unsigned int value)
+speed_t FAST_FUNC tty_value_to_baud(unsigned int value)
 {
 	int i = 0;
 
diff --git a/libbb/str_tolower.c b/libbb/str_tolower.c
index 037f717..f402e8e 100644
--- a/libbb/str_tolower.c
+++ b/libbb/str_tolower.c
@@ -4,7 +4,8 @@
  * Licensed under GPLv2 or later, see file LICENSE in this tarball for details.
  */
 #include "libbb.h"
-char* str_tolower(char *str)
+
+char* FAST_FUNC str_tolower(char *str)
 {
 	char *c;
 	for (c = str; *c; ++c)
diff --git a/libbb/strrstr.c b/libbb/strrstr.c
index f61dd51..d5cd44b 100644
--- a/libbb/strrstr.c
+++ b/libbb/strrstr.c
@@ -19,7 +19,7 @@
  * The strrstr() function finds the last occurrence of the substring needle
  * in the string haystack. The terminating nul characters are not compared.
  */
-char* strrstr(const char *haystack, const char *needle)
+char* FAST_FUNC strrstr(const char *haystack, const char *needle)
 {
 	char *r = NULL;
 
diff --git a/libbb/time.c b/libbb/time.c
index 07c0094..7d3ac91 100644
--- a/libbb/time.c
+++ b/libbb/time.c
@@ -20,14 +20,14 @@
 
 /* libc has incredibly messy way of doing this,
  * typically requiring -lrt. We just skip all this mess */
-unsigned long long monotonic_us(void)
+unsigned long long FAST_FUNC monotonic_us(void)
 {
 	struct timespec ts;
 	if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts))
 		bb_error_msg_and_die("clock_gettime(MONOTONIC) failed");
 	return ts.tv_sec * 1000000ULL + ts.tv_nsec/1000;
 }
-unsigned monotonic_sec(void)
+unsigned FAST_FUNC monotonic_sec(void)
 {
 	struct timespec ts;
 	if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts))
@@ -35,14 +35,14 @@
 	return ts.tv_sec;
 }
 #else
-unsigned long long monotonic_us(void)
+unsigned long long FAST_FUNC monotonic_us(void)
 {
 	struct timeval tv;
 	gettimeofday(&tv, NULL);
 	return tv.tv_sec * 1000000ULL + tv.tv_usec;
 }
 
-unsigned monotonic_sec(void)
+unsigned FAST_FUNC monotonic_sec(void)
 {
 	return time(NULL);
 }
diff --git a/libbb/trim.c b/libbb/trim.c
index 94ccaf7..ea20ff3 100644
--- a/libbb/trim.c
+++ b/libbb/trim.c
@@ -10,7 +10,7 @@
 
 #include "libbb.h"
 
-void trim(char *s)
+void FAST_FUNC trim(char *s)
 {
 	size_t len = strlen(s);
 	size_t lws;
diff --git a/libbb/u_signal_names.c b/libbb/u_signal_names.c
index 1dcbf5f..915eea5 100644
--- a/libbb/u_signal_names.c
+++ b/libbb/u_signal_names.c
@@ -121,7 +121,7 @@
 
 // Convert signal name to number.
 
-int get_signum(const char *name)
+int FAST_FUNC get_signum(const char *name)
 {
 	unsigned i;
 
@@ -155,7 +155,7 @@
 
 // Convert signal number to name
 
-const char *get_signame(int number)
+const char* FAST_FUNC get_signame(int number)
 {
 	if ((unsigned)number < ARRAY_SIZE(signals)) {
 		if (signals[number][0]) /* if it's not an empty str */
@@ -168,7 +168,7 @@
 
 // Print the whole signal list
 
-void print_signames(void)
+void FAST_FUNC print_signames(void)
 {
 	unsigned signo;
 
diff --git a/libbb/udp_io.c b/libbb/udp_io.c
index c99e516..b31f284 100644
--- a/libbb/udp_io.c
+++ b/libbb/udp_io.c
@@ -13,7 +13,7 @@
  * This asks kernel to let us know dst addr/port of incoming packets
  * We don't check for errors here. Not supported == won't be used
  */
-void
+void FAST_FUNC
 socket_want_pktinfo(int fd)
 {
 #ifdef IP_PKTINFO
@@ -25,7 +25,7 @@
 }
 
 
-ssize_t
+ssize_t FAST_FUNC
 send_to_from(int fd, void *buf, size_t len, int flags,
 		const struct sockaddr *to,
 		const struct sockaddr *from,
@@ -100,7 +100,7 @@
  * _Only_ IP/IPv6 address part of 'to' is _maybe_ modified.
  * Typical usage is to preinit 'to' with "default" value
  * before calling recv_from_to(). */
-ssize_t
+ssize_t FAST_FUNC
 recv_from_to(int fd, void *buf, size_t len, int flags,
 		struct sockaddr *from, struct sockaddr *to,
 		socklen_t sa_size)
diff --git a/libbb/update_passwd.c b/libbb/update_passwd.c
index d10e863..88bc28c 100644
--- a/libbb/update_passwd.c
+++ b/libbb/update_passwd.c
@@ -36,7 +36,7 @@
 #define check_selinux_update_passwd(username) ((void)0)
 #endif
 
-int update_passwd(const char *filename, const char *username,
+int FAST_FUNC update_passwd(const char *filename, const char *username,
 			const char *new_pw)
 {
 	struct stat sb;
diff --git a/libbb/uuencode.c b/libbb/uuencode.c
index 0aedf33..67d98d5 100644
--- a/libbb/uuencode.c
+++ b/libbb/uuencode.c
@@ -39,7 +39,7 @@
  * buffer of at least 1+BASE64_LENGTH(length) bytes.
  * where BASE64_LENGTH(len) = (4 * ((LENGTH + 2) / 3))
  */
-void bb_uuencode(char *p, const void *src, int length, const char *tbl)
+void FAST_FUNC bb_uuencode(char *p, const void *src, int length, const char *tbl)
 {
 	const unsigned char *s = src;
 
diff --git a/libbb/vdprintf.c b/libbb/vdprintf.c
index 726d563..09fffbc 100644
--- a/libbb/vdprintf.c
+++ b/libbb/vdprintf.c
@@ -10,7 +10,7 @@
 #include "libbb.h"
 
 #if defined(__GLIBC__) && __GLIBC__ < 2
-int vdprintf(int d, const char *format, va_list ap)
+int FAST_FUNC vdprintf(int d, const char *format, va_list ap)
 {
 	char buf[BUF_SIZE];
 	int len;
diff --git a/libbb/verror_msg.c b/libbb/verror_msg.c
index 5c6df48..58846d5 100644
--- a/libbb/verror_msg.c
+++ b/libbb/verror_msg.c
@@ -13,7 +13,7 @@
 smallint logmode = LOGMODE_STDIO;
 const char *msg_eol = "\n";
 
-void bb_verror_msg(const char *s, va_list p, const char* strerr)
+void FAST_FUNC bb_verror_msg(const char *s, va_list p, const char* strerr)
 {
 	char *msg;
 	int applet_len, strerr_len, msgeol_len, used;
@@ -70,7 +70,7 @@
 /* Code size is approximately the same, but currently it's the only user
  * of writev in entire bbox. __libc_writev in uclibc is ~50 bytes. */
 
-void bb_verror_msg(const char *s, va_list p, const char* strerr)
+void FAST_FUNC bb_verror_msg(const char *s, va_list p, const char* strerr)
 {
 	int strerr_len, msgeol_len;
 	struct iovec iov[3];
diff --git a/libbb/vfork_daemon_rexec.c b/libbb/vfork_daemon_rexec.c
index 9624efb..37d4c27 100644
--- a/libbb/vfork_daemon_rexec.c
+++ b/libbb/vfork_daemon_rexec.c
@@ -20,7 +20,7 @@
 
 /* This does a fork/exec in one call, using vfork().  Returns PID of new child,
  * -1 for failure.  Runs argv[0], searching path if that has no / in it. */
-pid_t spawn(char **argv)
+pid_t FAST_FUNC spawn(char **argv)
 {
 	/* Compiler should not optimize stores here */
 	volatile int failed;
@@ -58,7 +58,7 @@
 }
 
 /* Die with an error message if we can't spawn a child process. */
-pid_t xspawn(char **argv)
+pid_t FAST_FUNC xspawn(char **argv)
 {
 	pid_t pid = spawn(argv);
 	if (pid < 0)
@@ -66,7 +66,7 @@
 	return pid;
 }
 
-int safe_waitpid(int pid, int *wstat, int options)
+int FAST_FUNC safe_waitpid(int pid, int *wstat, int options)
 {
 	int r;
 
@@ -76,13 +76,13 @@
 	return r;
 }
 
-int wait_any_nohang(int *wstat)
+int FAST_FUNC wait_any_nohang(int *wstat)
 {
 	return safe_waitpid(-1, wstat, WNOHANG);
 }
 
 // Wait for the specified child PID to exit, returning child's error return.
-int wait4pid(int pid)
+int FAST_FUNC wait4pid(int pid)
 {
 	int status;
 
@@ -101,7 +101,7 @@
 }
 
 #if ENABLE_FEATURE_PREFER_APPLETS
-void save_nofork_data(struct nofork_save_area *save)
+void FAST_FUNC save_nofork_data(struct nofork_save_area *save)
 {
 	memcpy(&save->die_jmp, &die_jmp, sizeof(die_jmp));
 	save->applet_name = applet_name;
@@ -111,7 +111,7 @@
 	save->saved = 1;
 }
 
-void restore_nofork_data(struct nofork_save_area *save)
+void FAST_FUNC restore_nofork_data(struct nofork_save_area *save)
 {
 	memcpy(&die_jmp, &save->die_jmp, sizeof(die_jmp));
 	applet_name = save->applet_name;
@@ -120,7 +120,7 @@
 	die_sleep = save->die_sleep;
 }
 
-int run_nofork_applet_prime(struct nofork_save_area *old, int applet_no, char **argv)
+int FAST_FUNC run_nofork_applet_prime(struct nofork_save_area *old, int applet_no, char **argv)
 {
 	int rc, argc;
 
@@ -166,7 +166,7 @@
 	return rc & 0xff; /* don't confuse people with "exitcodes" >255 */
 }
 
-int run_nofork_applet(int applet_no, char **argv)
+int FAST_FUNC run_nofork_applet(int applet_no, char **argv)
 {
 	struct nofork_save_area old;
 
@@ -176,7 +176,7 @@
 }
 #endif /* FEATURE_PREFER_APPLETS */
 
-int spawn_and_wait(char **argv)
+int FAST_FUNC spawn_and_wait(char **argv)
 {
 	int rc;
 #if ENABLE_FEATURE_PREFER_APPLETS
@@ -210,7 +210,7 @@
 }
 
 #if !BB_MMU
-void re_exec(char **argv)
+void FAST_FUNC re_exec(char **argv)
 {
 	/* high-order bit of first char in argv[0] is a hidden
 	 * "we have (already) re-execed, don't do it again" flag */
@@ -219,7 +219,7 @@
 	bb_perror_msg_and_die("exec %s", bb_busybox_exec_path);
 }
 
-void forkexit_or_rexec(char **argv)
+void FAST_FUNC forkexit_or_rexec(char **argv)
 {
 	pid_t pid;
 	/* Maybe we are already re-execed and come here again? */
@@ -237,7 +237,7 @@
 #else
 /* Dance around (void)...*/
 #undef forkexit_or_rexec
-void forkexit_or_rexec(void)
+void FAST_FUNC forkexit_or_rexec(void)
 {
 	pid_t pid;
 	pid = fork();
@@ -252,7 +252,7 @@
 
 /* Due to a #define in libbb.h on MMU systems we actually have 1 argument -
  * char **argv "vanishes" */
-void bb_daemonize_or_rexec(int flags, char **argv)
+void FAST_FUNC bb_daemonize_or_rexec(int flags, char **argv)
 {
 	int fd;
 
@@ -286,7 +286,7 @@
 	}
 }
 
-void bb_sanitize_stdio(void)
+void FAST_FUNC bb_sanitize_stdio(void)
 {
 	bb_daemonize_or_rexec(DAEMON_ONLY_SANITIZE, NULL);
 }
diff --git a/libbb/warn_ignoring_args.c b/libbb/warn_ignoring_args.c
index be78a44..65dea32 100644
--- a/libbb/warn_ignoring_args.c
+++ b/libbb/warn_ignoring_args.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-void bb_warn_ignoring_args(int n)
+void FAST_FUNC bb_warn_ignoring_args(int n)
 {
 	if (n) {
 		bb_error_msg("ignoring all arguments");
diff --git a/libbb/wfopen.c b/libbb/wfopen.c
index 9248874..ac365c2 100644
--- a/libbb/wfopen.c
+++ b/libbb/wfopen.c
@@ -9,7 +9,7 @@
 
 #include "libbb.h"
 
-FILE *fopen_or_warn(const char *path, const char *mode)
+FILE* FAST_FUNC fopen_or_warn(const char *path, const char *mode)
 {
 	FILE *fp = fopen(path, mode);
 	if (!fp) {
diff --git a/libbb/wfopen_input.c b/libbb/wfopen_input.c
index a7c1c32..46ff7a6 100644
--- a/libbb/wfopen_input.c
+++ b/libbb/wfopen_input.c
@@ -14,7 +14,7 @@
 
 #include "libbb.h"
 
-FILE *fopen_or_warn_stdin(const char *filename)
+FILE* FAST_FUNC fopen_or_warn_stdin(const char *filename)
 {
 	FILE *fp = stdin;
 
@@ -26,7 +26,7 @@
 	return fp;
 }
 
-FILE *xfopen_stdin(const char *filename)
+FILE* FAST_FUNC xfopen_stdin(const char *filename)
 {
 	FILE *fp = fopen_or_warn_stdin(filename);
 	if (fp)
@@ -34,7 +34,7 @@
 	xfunc_die();	/* We already output an error message. */
 }
 
-int open_or_warn_stdin(const char *filename)
+int FAST_FUNC open_or_warn_stdin(const char *filename)
 {
 	int fd = STDIN_FILENO;
 
diff --git a/libbb/write.c b/libbb/write.c
index b628b49..e8a9eff 100644
--- a/libbb/write.c
+++ b/libbb/write.c
@@ -11,7 +11,7 @@
 
 /* Open file and write string str to it, close file.
  * Die on any open or write-error.  */
-void xopen_xwrite_close(const char* file, const char* str)
+void FAST_FUNC xopen_xwrite_close(const char* file, const char* str)
 {
 	int fd = xopen(file, O_WRONLY);
 
diff --git a/libbb/xatonum.c b/libbb/xatonum.c
index a410ae9..3cdf634 100644
--- a/libbb/xatonum.c
+++ b/libbb/xatonum.c
@@ -59,12 +59,12 @@
 
 /* A few special cases */
 
-int xatoi_u(const char *numstr)
+int FAST_FUNC xatoi_u(const char *numstr)
 {
 	return xatou_range(numstr, 0, INT_MAX);
 }
 
-uint16_t xatou16(const char *numstr)
+uint16_t FAST_FUNC xatou16(const char *numstr)
 {
 	return xatou_range(numstr, 0, 0xffff);
 }
diff --git a/libbb/xatonum_template.c b/libbb/xatonum_template.c
index 9f9dc11..2360ae8 100644
--- a/libbb/xatonum_template.c
+++ b/libbb/xatonum_template.c
@@ -12,7 +12,7 @@
 #define XSTR_STRTOU strtoul
 */
 
-unsigned type xstrtou(_range_sfx)(const char *numstr, int base,
+unsigned type FAST_FUNC xstrtou(_range_sfx)(const char *numstr, int base,
 		unsigned type lower,
 		unsigned type upper,
 		const struct suffix_mult *suffixes)
@@ -71,25 +71,25 @@
 	bb_error_msg_and_die("invalid number '%s'", numstr);
 }
 
-unsigned type xstrtou(_range)(const char *numstr, int base,
+unsigned type FAST_FUNC xstrtou(_range)(const char *numstr, int base,
 		unsigned type lower,
 		unsigned type upper)
 {
 	return xstrtou(_range_sfx)(numstr, base, lower, upper, NULL);
 }
 
-unsigned type xstrtou(_sfx)(const char *numstr, int base,
+unsigned type FAST_FUNC xstrtou(_sfx)(const char *numstr, int base,
 		const struct suffix_mult *suffixes)
 {
 	return xstrtou(_range_sfx)(numstr, base, 0, XSTR_UTYPE_MAX, suffixes);
 }
 
-unsigned type xstrtou()(const char *numstr, int base)
+unsigned type FAST_FUNC xstrtou()(const char *numstr, int base)
 {
 	return xstrtou(_range_sfx)(numstr, base, 0, XSTR_UTYPE_MAX, NULL);
 }
 
-unsigned type xatou(_range_sfx)(const char *numstr,
+unsigned type FAST_FUNC xatou(_range_sfx)(const char *numstr,
 		unsigned type lower,
 		unsigned type upper,
 		const struct suffix_mult *suffixes)
@@ -97,27 +97,27 @@
 	return xstrtou(_range_sfx)(numstr, 10, lower, upper, suffixes);
 }
 
-unsigned type xatou(_range)(const char *numstr,
+unsigned type FAST_FUNC xatou(_range)(const char *numstr,
 		unsigned type lower,
 		unsigned type upper)
 {
 	return xstrtou(_range_sfx)(numstr, 10, lower, upper, NULL);
 }
 
-unsigned type xatou(_sfx)(const char *numstr,
+unsigned type FAST_FUNC xatou(_sfx)(const char *numstr,
 		const struct suffix_mult *suffixes)
 {
 	return xstrtou(_range_sfx)(numstr, 10, 0, XSTR_UTYPE_MAX, suffixes);
 }
 
-unsigned type xatou()(const char *numstr)
+unsigned type FAST_FUNC xatou()(const char *numstr)
 {
 	return xatou(_sfx)(numstr, NULL);
 }
 
 /* Signed ones */
 
-type xstrto(_range_sfx)(const char *numstr, int base,
+type FAST_FUNC xstrto(_range_sfx)(const char *numstr, int base,
 		type lower,
 		type upper,
 		const struct suffix_mult *suffixes)
@@ -148,12 +148,12 @@
 	return r;
 }
 
-type xstrto(_range)(const char *numstr, int base, type lower, type upper)
+type FAST_FUNC xstrto(_range)(const char *numstr, int base, type lower, type upper)
 {
 	return xstrto(_range_sfx)(numstr, base, lower, upper, NULL);
 }
 
-type xato(_range_sfx)(const char *numstr,
+type FAST_FUNC xato(_range_sfx)(const char *numstr,
 		type lower,
 		type upper,
 		const struct suffix_mult *suffixes)
@@ -161,17 +161,17 @@
 	return xstrto(_range_sfx)(numstr, 10, lower, upper, suffixes);
 }
 
-type xato(_range)(const char *numstr, type lower, type upper)
+type FAST_FUNC xato(_range)(const char *numstr, type lower, type upper)
 {
 	return xstrto(_range_sfx)(numstr, 10, lower, upper, NULL);
 }
 
-type xato(_sfx)(const char *numstr, const struct suffix_mult *suffixes)
+type FAST_FUNC xato(_sfx)(const char *numstr, const struct suffix_mult *suffixes)
 {
 	return xstrto(_range_sfx)(numstr, 10, XSTR_TYPE_MIN, XSTR_TYPE_MAX, suffixes);
 }
 
-type xato()(const char *numstr)
+type FAST_FUNC xato()(const char *numstr)
 {
 	return xstrto(_range_sfx)(numstr, 10, XSTR_TYPE_MIN, XSTR_TYPE_MAX, NULL);
 }
diff --git a/libbb/xconnect.c b/libbb/xconnect.c
index 950aee8..d48c503 100644
--- a/libbb/xconnect.c
+++ b/libbb/xconnect.c
@@ -9,16 +9,16 @@
 #include <netinet/in.h>
 #include "libbb.h"
 
-void setsockopt_reuseaddr(int fd)
+void FAST_FUNC setsockopt_reuseaddr(int fd)
 {
 	setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, &const_int_1, sizeof(const_int_1));
 }
-int setsockopt_broadcast(int fd)
+int FAST_FUNC setsockopt_broadcast(int fd)
 {
 	return setsockopt(fd, SOL_SOCKET, SO_BROADCAST, &const_int_1, sizeof(const_int_1));
 }
 
-void xconnect(int s, const struct sockaddr *s_addr, socklen_t addrlen)
+void FAST_FUNC xconnect(int s, const struct sockaddr *s_addr, socklen_t addrlen)
 {
 	if (connect(s, s_addr, addrlen) < 0) {
 		if (ENABLE_FEATURE_CLEAN_UP)
@@ -35,7 +35,7 @@
  * If "port" is a number use it as the port.
  * If "port" is a name it is looked up in /etc/services, if it isnt found return
  * default_port */
-unsigned bb_lookup_port(const char *port, const char *protocol, unsigned default_port)
+unsigned FAST_FUNC bb_lookup_port(const char *port, const char *protocol, unsigned default_port)
 {
 	unsigned port_nr = default_port;
 	if (port) {
@@ -60,7 +60,7 @@
 /* "Old" networking API - only IPv4 */
 
 /*
-void bb_lookup_host(struct sockaddr_in *s_in, const char *host)
+void FAST_FUNC bb_lookup_host(struct sockaddr_in *s_in, const char *host)
 {
 	struct hostent *he;
 
@@ -71,7 +71,7 @@
 }
 
 
-int xconnect_tcp_v4(struct sockaddr_in *s_addr)
+int FAST_FUNC xconnect_tcp_v4(struct sockaddr_in *s_addr)
 {
 	int s = xsocket(AF_INET, SOCK_STREAM, 0);
 	xconnect(s, (struct sockaddr*) s_addr, sizeof(*s_addr));
@@ -82,7 +82,7 @@
 /* "New" networking API */
 
 
-int get_nport(const struct sockaddr *sa)
+int FAST_FUNC get_nport(const struct sockaddr *sa)
 {
 #if ENABLE_FEATURE_IPV6
 	if (sa->sa_family == AF_INET6) {
@@ -96,7 +96,7 @@
 	return -1;
 }
 
-void set_nport(len_and_sockaddr *lsa, unsigned port)
+void FAST_FUNC set_nport(len_and_sockaddr *lsa, unsigned port)
 {
 #if ENABLE_FEATURE_IPV6
 	if (lsa->u.sa.sa_family == AF_INET6) {
@@ -205,34 +205,34 @@
 #endif
 
 #if ENABLE_FEATURE_IPV6
-len_and_sockaddr* host_and_af2sockaddr(const char *host, int port, sa_family_t af)
+len_and_sockaddr* FAST_FUNC host_and_af2sockaddr(const char *host, int port, sa_family_t af)
 {
 	return str2sockaddr(host, port, af, 0);
 }
 
-len_and_sockaddr* xhost_and_af2sockaddr(const char *host, int port, sa_family_t af)
+len_and_sockaddr* FAST_FUNC xhost_and_af2sockaddr(const char *host, int port, sa_family_t af)
 {
 	return str2sockaddr(host, port, af, DIE_ON_ERROR);
 }
 #endif
 
-len_and_sockaddr* host2sockaddr(const char *host, int port)
+len_and_sockaddr* FAST_FUNC host2sockaddr(const char *host, int port)
 {
 	return str2sockaddr(host, port, AF_UNSPEC, 0);
 }
 
-len_and_sockaddr* xhost2sockaddr(const char *host, int port)
+len_and_sockaddr* FAST_FUNC xhost2sockaddr(const char *host, int port)
 {
 	return str2sockaddr(host, port, AF_UNSPEC, DIE_ON_ERROR);
 }
 
-len_and_sockaddr* xdotted2sockaddr(const char *host, int port)
+len_and_sockaddr* FAST_FUNC xdotted2sockaddr(const char *host, int port)
 {
 	return str2sockaddr(host, port, AF_UNSPEC, AI_NUMERICHOST | DIE_ON_ERROR);
 }
 
 #undef xsocket_type
-int xsocket_type(len_and_sockaddr **lsap, USE_FEATURE_IPV6(int family,) int sock_type)
+int FAST_FUNC xsocket_type(len_and_sockaddr **lsap, USE_FEATURE_IPV6(int family,) int sock_type)
 {
 	SKIP_FEATURE_IPV6(enum { family = AF_INET };)
 	len_and_sockaddr *lsa;
@@ -264,7 +264,7 @@
 	return fd;
 }
 
-int xsocket_stream(len_and_sockaddr **lsap)
+int FAST_FUNC xsocket_stream(len_and_sockaddr **lsap)
 {
 	return xsocket_type(lsap, USE_FEATURE_IPV6(AF_UNSPEC,) SOCK_STREAM);
 }
@@ -288,18 +288,18 @@
 	return fd;
 }
 
-int create_and_bind_stream_or_die(const char *bindaddr, int port)
+int FAST_FUNC create_and_bind_stream_or_die(const char *bindaddr, int port)
 {
 	return create_and_bind_or_die(bindaddr, port, SOCK_STREAM);
 }
 
-int create_and_bind_dgram_or_die(const char *bindaddr, int port)
+int FAST_FUNC create_and_bind_dgram_or_die(const char *bindaddr, int port)
 {
 	return create_and_bind_or_die(bindaddr, port, SOCK_DGRAM);
 }
 
 
-int create_and_connect_stream_or_die(const char *peer, int port)
+int FAST_FUNC create_and_connect_stream_or_die(const char *peer, int port)
 {
 	int fd;
 	len_and_sockaddr *lsa;
@@ -312,7 +312,7 @@
 	return fd;
 }
 
-int xconnect_stream(const len_and_sockaddr *lsa)
+int FAST_FUNC xconnect_stream(const len_and_sockaddr *lsa)
 {
 	int fd = xsocket(lsa->u.sa.sa_family, SOCK_STREAM, 0);
 	xconnect(fd, &lsa->u.sa, lsa->len);
@@ -322,7 +322,7 @@
 /* We hijack this constant to mean something else */
 /* It doesn't hurt because we will add this bit anyway */
 #define IGNORE_PORT NI_NUMERICSERV
-static char* sockaddr2str(const struct sockaddr *sa, int flags)
+static char* FAST_FUNC sockaddr2str(const struct sockaddr *sa, int flags)
 {
 	char host[128];
 	char serv[16];
@@ -361,26 +361,26 @@
 	/*return xstrdup(host);*/
 }
 
-char* xmalloc_sockaddr2host(const struct sockaddr *sa)
+char* FAST_FUNC xmalloc_sockaddr2host(const struct sockaddr *sa)
 {
 	return sockaddr2str(sa, 0);
 }
 
-char* xmalloc_sockaddr2host_noport(const struct sockaddr *sa)
+char* FAST_FUNC xmalloc_sockaddr2host_noport(const struct sockaddr *sa)
 {
 	return sockaddr2str(sa, IGNORE_PORT);
 }
 
-char* xmalloc_sockaddr2hostonly_noport(const struct sockaddr *sa)
+char* FAST_FUNC xmalloc_sockaddr2hostonly_noport(const struct sockaddr *sa)
 {
 	return sockaddr2str(sa, NI_NAMEREQD | IGNORE_PORT);
 }
-char* xmalloc_sockaddr2dotted(const struct sockaddr *sa)
+char* FAST_FUNC xmalloc_sockaddr2dotted(const struct sockaddr *sa)
 {
 	return sockaddr2str(sa, NI_NUMERICHOST);
 }
 
-char* xmalloc_sockaddr2dotted_noport(const struct sockaddr *sa)
+char* FAST_FUNC xmalloc_sockaddr2dotted_noport(const struct sockaddr *sa)
 {
 	return sockaddr2str(sa, NI_NUMERICHOST | IGNORE_PORT);
 }
diff --git a/libbb/xfunc_die.c b/libbb/xfunc_die.c
index 357494d..ba9fe93 100644
--- a/libbb/xfunc_die.c
+++ b/libbb/xfunc_die.c
@@ -17,7 +17,7 @@
 jmp_buf die_jmp;
 #endif
 
-void xfunc_die(void)
+void FAST_FUNC xfunc_die(void)
 {
 	if (die_sleep) {
 		if ((ENABLE_FEATURE_PREFER_APPLETS || ENABLE_HUSH)
diff --git a/libbb/xfuncs.c b/libbb/xfuncs.c
index fe3c647..8ef305b 100644
--- a/libbb/xfuncs.c
+++ b/libbb/xfuncs.c
@@ -25,17 +25,17 @@
 #include "libbb.h"
 
 /* Turn on nonblocking I/O on a fd */
-int ndelay_on(int fd)
+int FAST_FUNC ndelay_on(int fd)
 {
 	return fcntl(fd, F_SETFL, fcntl(fd,F_GETFL) | O_NONBLOCK);
 }
 
-int ndelay_off(int fd)
+int FAST_FUNC ndelay_off(int fd)
 {
 	return fcntl(fd, F_SETFL, fcntl(fd,F_GETFL) & ~O_NONBLOCK);
 }
 
-int close_on_exec_on(int fd)
+int FAST_FUNC close_on_exec_on(int fd)
 {
 	return fcntl(fd, F_SETFD, FD_CLOEXEC);
 }
@@ -43,7 +43,7 @@
 /* Convert unsigned long long value into compact 4-char
  * representation. Examples: "1234", "1.2k", " 27M", "123T"
  * String is not terminated (buf[4] is untouched) */
-void smart_ulltoa4(unsigned long long ul, char buf[5], const char *scale)
+void FAST_FUNC smart_ulltoa4(unsigned long long ul, char buf[5], const char *scale)
 {
 	const char *fmt;
 	char c;
@@ -91,7 +91,7 @@
 
 /* Convert unsigned long long value into compact 5-char representation.
  * String is not terminated (buf[5] is untouched) */
-void smart_ulltoa5(unsigned long long ul, char buf[6], const char *scale)
+void FAST_FUNC smart_ulltoa5(unsigned long long ul, char buf[6], const char *scale)
 {
 	const char *fmt;
 	char c;
@@ -149,7 +149,7 @@
 // A truncated result contains the first few digits of the result ala strncpy.
 // Returns a pointer past last generated digit, does _not_ store NUL.
 void BUG_sizeof_unsigned_not_4(void);
-char *utoa_to_buf(unsigned n, char *buf, unsigned buflen)
+char* FAST_FUNC utoa_to_buf(unsigned n, char *buf, unsigned buflen)
 {
 	unsigned i, out, res;
 	if (sizeof(unsigned) != 4)
@@ -170,7 +170,7 @@
 }
 
 /* Convert signed integer to ascii, like utoa_to_buf() */
-char *itoa_to_buf(int n, char *buf, unsigned buflen)
+char* FAST_FUNC itoa_to_buf(int n, char *buf, unsigned buflen)
 {
 	if (buflen && n < 0) {
 		n = -n;
@@ -190,7 +190,7 @@
 static char local_buf[sizeof(int) * 3];
 
 // Convert unsigned integer to ascii using a static buffer (returned).
-char *utoa(unsigned n)
+char* FAST_FUNC utoa(unsigned n)
 {
 	*(utoa_to_buf(n, local_buf, sizeof(local_buf))) = '\0';
 
@@ -198,7 +198,7 @@
 }
 
 /* Convert signed integer to ascii using a static buffer (returned). */
-char *itoa(int n)
+char* FAST_FUNC itoa(int n)
 {
 	*(itoa_to_buf(n, local_buf, sizeof(local_buf))) = '\0';
 
@@ -206,7 +206,7 @@
 }
 
 /* Emit a string of hex representation of bytes */
-char *bin2hex(char *p, const char *cp, int count)
+char* FAST_FUNC bin2hex(char *p, const char *cp, int count)
 {
 	while (count) {
 		unsigned char c = *cp++;
@@ -220,7 +220,7 @@
 
 /* Return how long the file at fd is, if there's any way to determine it. */
 #ifdef UNUSED
-off_t fdlength(int fd)
+off_t FAST_FUNC fdlength(int fd)
 {
 	off_t bottom = 0, top = 0, pos;
 	long size;
@@ -262,7 +262,7 @@
 
 /* It is perfectly ok to pass in a NULL for either width or for
  * height, in which case that value will not be set.  */
-int get_terminal_width_height(int fd, unsigned *width, unsigned *height)
+int FAST_FUNC get_terminal_width_height(int fd, unsigned *width, unsigned *height)
 {
 	struct winsize win = { 0, 0, 0, 0 };
 	int ret = ioctl(fd, TIOCGWINSZ, &win);
diff --git a/libbb/xfuncs_printf.c b/libbb/xfuncs_printf.c
index 105939b..108e140 100644
--- a/libbb/xfuncs_printf.c
+++ b/libbb/xfuncs_printf.c
@@ -33,7 +33,7 @@
  * included after these prototypes in libbb.h, all is well.
  */
 // Warn if we can't allocate size bytes of memory.
-void *malloc_or_warn(size_t size)
+void* FAST_FUNC malloc_or_warn(size_t size)
 {
 	void *ptr = malloc(size);
 	if (ptr == NULL && size != 0)
@@ -42,7 +42,7 @@
 }
 
 // Die if we can't allocate size bytes of memory.
-void *xmalloc(size_t size)
+void* FAST_FUNC xmalloc(size_t size)
 {
 	void *ptr = malloc(size);
 	if (ptr == NULL && size != 0)
@@ -53,7 +53,7 @@
 // Die if we can't resize previously allocated memory.  (This returns a pointer
 // to the new memory, which may or may not be the same as the old memory.
 // It'll copy the contents to a new chunk and free the old one if necessary.)
-void *xrealloc(void *ptr, size_t size)
+void* FAST_FUNC xrealloc(void *ptr, size_t size)
 {
 	ptr = realloc(ptr, size);
 	if (ptr == NULL && size != 0)
@@ -63,7 +63,7 @@
 #endif /* DMALLOC */
 
 // Die if we can't allocate and zero size bytes of memory.
-void *xzalloc(size_t size)
+void* FAST_FUNC xzalloc(size_t size)
 {
 	void *ptr = xmalloc(size);
 	memset(ptr, 0, size);
@@ -71,7 +71,7 @@
 }
 
 // Die if we can't copy a string to freshly allocated memory.
-char * xstrdup(const char *s)
+char* FAST_FUNC xstrdup(const char *s)
 {
 	char *t;
 
@@ -88,7 +88,7 @@
 
 // Die if we can't allocate n+1 bytes (space for the null terminator) and copy
 // the (possibly truncated to length n) string into it.
-char *xstrndup(const char *s, int n)
+char* FAST_FUNC xstrndup(const char *s, int n)
 {
 	int m;
 	char *t;
@@ -112,9 +112,9 @@
 	return memcpy(t, s, n);
 }
 
-// Die if we can't open a file and return a FILE * to it.
+// Die if we can't open a file and return a FILE* to it.
 // Notice we haven't got xfread(), This is for use with fscanf() and friends.
-FILE *xfopen(const char *path, const char *mode)
+FILE* FAST_FUNC xfopen(const char *path, const char *mode)
 {
 	FILE *fp = fopen(path, mode);
 	if (fp == NULL)
@@ -123,7 +123,7 @@
 }
 
 // Die if we can't open a file and return a fd.
-int xopen3(const char *pathname, int flags, int mode)
+int FAST_FUNC xopen3(const char *pathname, int flags, int mode)
 {
 	int ret;
 
@@ -135,13 +135,13 @@
 }
 
 // Die if we can't open an existing file and return a fd.
-int xopen(const char *pathname, int flags)
+int FAST_FUNC xopen(const char *pathname, int flags)
 {
 	return xopen3(pathname, flags, 0666);
 }
 
 // Warn if we can't open a file and return a fd.
-int open3_or_warn(const char *pathname, int flags, int mode)
+int FAST_FUNC open3_or_warn(const char *pathname, int flags, int mode)
 {
 	int ret;
 
@@ -153,24 +153,24 @@
 }
 
 // Warn if we can't open a file and return a fd.
-int open_or_warn(const char *pathname, int flags)
+int FAST_FUNC open_or_warn(const char *pathname, int flags)
 {
 	return open3_or_warn(pathname, flags, 0666);
 }
 
-void xunlink(const char *pathname)
+void FAST_FUNC xunlink(const char *pathname)
 {
 	if (unlink(pathname))
 		bb_perror_msg_and_die("can't remove file '%s'", pathname);
 }
 
-void xrename(const char *oldpath, const char *newpath)
+void FAST_FUNC xrename(const char *oldpath, const char *newpath)
 {
 	if (rename(oldpath, newpath))
 		bb_perror_msg_and_die("can't move '%s' to '%s'", oldpath, newpath);
 }
 
-int rename_or_warn(const char *oldpath, const char *newpath)
+int FAST_FUNC rename_or_warn(const char *oldpath, const char *newpath)
 {
 	int n = rename(oldpath, newpath);
 	if (n)
@@ -178,20 +178,20 @@
 	return n;
 }
 
-void xpipe(int filedes[2])
+void FAST_FUNC xpipe(int filedes[2])
 {
 	if (pipe(filedes))
 		bb_perror_msg_and_die("can't create pipe");
 }
 
-void xdup2(int from, int to)
+void FAST_FUNC xdup2(int from, int to)
 {
 	if (dup2(from, to) != to)
 		bb_perror_msg_and_die("can't duplicate file descriptor");
 }
 
 // "Renumber" opened fd
-void xmove_fd(int from, int to)
+void FAST_FUNC xmove_fd(int from, int to)
 {
 	if (from == to)
 		return;
@@ -200,7 +200,7 @@
 }
 
 // Die with an error message if we can't write the entire buffer.
-void xwrite(int fd, const void *buf, size_t count)
+void FAST_FUNC xwrite(int fd, const void *buf, size_t count)
 {
 	if (count) {
 		ssize_t size = full_write(fd, buf, count);
@@ -210,7 +210,7 @@
 }
 
 // Die with an error message if we can't lseek to the right spot.
-off_t xlseek(int fd, off_t offset, int whence)
+off_t FAST_FUNC xlseek(int fd, off_t offset, int whence)
 {
 	off_t off = lseek(fd, offset, whence);
 	if (off == (off_t)-1) {
@@ -221,8 +221,8 @@
 	return off;
 }
 
-// Die with supplied filename if this FILE * has ferror set.
-void die_if_ferror(FILE *fp, const char *fn)
+// Die with supplied filename if this FILE* has ferror set.
+void FAST_FUNC die_if_ferror(FILE *fp, const char *fn)
 {
 	if (ferror(fp)) {
 		/* ferror doesn't set useful errno */
@@ -231,13 +231,13 @@
 }
 
 // Die with an error message if stdout has ferror set.
-void die_if_ferror_stdout(void)
+void FAST_FUNC die_if_ferror_stdout(void)
 {
 	die_if_ferror(stdout, bb_msg_standard_output);
 }
 
 // Die with an error message if we have trouble flushing stdout.
-void xfflush_stdout(void)
+void FAST_FUNC xfflush_stdout(void)
 {
 	if (fflush(stdout)) {
 		bb_perror_msg_and_die(bb_msg_standard_output);
@@ -245,7 +245,7 @@
 }
 
 
-int bb_putchar(int ch)
+int FAST_FUNC bb_putchar(int ch)
 {
 	/* time.c needs putc(ch, stdout), not putchar(ch).
 	 * it does "stdout = stderr;", but then glibc's putchar()
@@ -253,9 +253,9 @@
 	return putc(ch, stdout);
 }
 
-/* Die with an error message if we can't copy an entire FILE * to stdout,
+/* Die with an error message if we can't copy an entire FILE* to stdout,
  * then close that file. */
-void xprint_and_close_file(FILE *file)
+void FAST_FUNC xprint_and_close_file(FILE *file)
 {
 	fflush(stdout);
 	// copyfd outputs error messages for us.
@@ -267,7 +267,7 @@
 
 // Die with an error message if we can't malloc() enough space and do an
 // sprintf() into that space.
-char *xasprintf(const char *format, ...)
+char* FAST_FUNC xasprintf(const char *format, ...)
 {
 	va_list p;
 	int r;
@@ -295,7 +295,7 @@
 }
 
 #if 0 /* If we will ever meet a libc which hasn't [f]dprintf... */
-int fdprintf(int fd, const char *format, ...)
+int FAST_FUNC fdprintf(int fd, const char *format, ...)
 {
 	va_list p;
 	int r;
@@ -327,7 +327,7 @@
 }
 #endif
 
-void xsetenv(const char *key, const char *value)
+void FAST_FUNC xsetenv(const char *key, const char *value)
 {
 	if (setenv(key, value, 1))
 		bb_error_msg_and_die(bb_msg_memory_exhausted);
@@ -336,32 +336,32 @@
 // Die with an error message if we can't set gid.  (Because resource limits may
 // limit this user to a given number of processes, and if that fills up the
 // setgid() will fail and we'll _still_be_root_, which is bad.)
-void xsetgid(gid_t gid)
+void FAST_FUNC xsetgid(gid_t gid)
 {
 	if (setgid(gid)) bb_perror_msg_and_die("setgid");
 }
 
 // Die with an error message if we can't set uid.  (See xsetgid() for why.)
-void xsetuid(uid_t uid)
+void FAST_FUNC xsetuid(uid_t uid)
 {
 	if (setuid(uid)) bb_perror_msg_and_die("setuid");
 }
 
 // Die if we can't chdir to a new path.
-void xchdir(const char *path)
+void FAST_FUNC xchdir(const char *path)
 {
 	if (chdir(path))
 		bb_perror_msg_and_die("chdir(%s)", path);
 }
 
-void xchroot(const char *path)
+void FAST_FUNC xchroot(const char *path)
 {
 	if (chroot(path))
 		bb_perror_msg_and_die("can't change root directory to %s", path);
 }
 
 // Print a warning message if opendir() fails, but don't die.
-DIR *warn_opendir(const char *path)
+DIR* FAST_FUNC warn_opendir(const char *path)
 {
 	DIR *dp;
 
@@ -372,7 +372,7 @@
 }
 
 // Die with an error message if opendir() fails.
-DIR *xopendir(const char *path)
+DIR* FAST_FUNC xopendir(const char *path)
 {
 	DIR *dp;
 
@@ -383,7 +383,7 @@
 }
 
 // Die with an error message if we can't open a new socket.
-int xsocket(int domain, int type, int protocol)
+int FAST_FUNC xsocket(int domain, int type, int protocol)
 {
 	int r = socket(domain, type, protocol);
 
@@ -404,20 +404,20 @@
 }
 
 // Die with an error message if we can't bind a socket to an address.
-void xbind(int sockfd, struct sockaddr *my_addr, socklen_t addrlen)
+void FAST_FUNC xbind(int sockfd, struct sockaddr *my_addr, socklen_t addrlen)
 {
 	if (bind(sockfd, my_addr, addrlen)) bb_perror_msg_and_die("bind");
 }
 
 // Die with an error message if we can't listen for connections on a socket.
-void xlisten(int s, int backlog)
+void FAST_FUNC xlisten(int s, int backlog)
 {
 	if (listen(s, backlog)) bb_perror_msg_and_die("listen");
 }
 
 /* Die with an error message if sendto failed.
  * Return bytes sent otherwise  */
-ssize_t xsendto(int s, const  void *buf, size_t len, const struct sockaddr *to,
+ssize_t FAST_FUNC xsendto(int s, const  void *buf, size_t len, const struct sockaddr *to,
 				socklen_t tolen)
 {
 	ssize_t ret = sendto(s, buf, len, 0, to, tolen);
@@ -430,14 +430,14 @@
 }
 
 // xstat() - a stat() which dies on failure with meaningful error message
-void xstat(const char *name, struct stat *stat_buf)
+void FAST_FUNC xstat(const char *name, struct stat *stat_buf)
 {
 	if (stat(name, stat_buf))
 		bb_perror_msg_and_die("can't stat '%s'", name);
 }
 
 // selinux_or_die() - die if SELinux is disabled.
-void selinux_or_die(void)
+void FAST_FUNC selinux_or_die(void)
 {
 #if ENABLE_SELINUX
 	int rc = is_selinux_enabled();
@@ -451,7 +451,7 @@
 #endif
 }
 
-int ioctl_or_perror_and_die(int fd, unsigned request, void *argp, const char *fmt,...)
+int FAST_FUNC ioctl_or_perror_and_die(int fd, unsigned request, void *argp, const char *fmt,...)
 {
 	int ret;
 	va_list p;
@@ -467,7 +467,7 @@
 	return ret;
 }
 
-int ioctl_or_perror(int fd, unsigned request, void *argp, const char *fmt,...)
+int FAST_FUNC ioctl_or_perror(int fd, unsigned request, void *argp, const char *fmt,...)
 {
 	va_list p;
 	int ret = ioctl(fd, request, argp);
@@ -481,7 +481,7 @@
 }
 
 #if ENABLE_IOCTL_HEX2STR_ERROR
-int bb_ioctl_or_warn(int fd, unsigned request, void *argp, const char *ioctl_name)
+int FAST_FUNC bb_ioctl_or_warn(int fd, unsigned request, void *argp, const char *ioctl_name)
 {
 	int ret;
 
@@ -490,7 +490,7 @@
 		bb_simple_perror_msg(ioctl_name);
 	return ret;
 }
-int bb_xioctl(int fd, unsigned request, void *argp, const char *ioctl_name)
+int FAST_FUNC bb_xioctl(int fd, unsigned request, void *argp, const char *ioctl_name)
 {
 	int ret;
 
@@ -500,7 +500,7 @@
 	return ret;
 }
 #else
-int bb_ioctl_or_warn(int fd, unsigned request, void *argp)
+int FAST_FUNC bb_ioctl_or_warn(int fd, unsigned request, void *argp)
 {
 	int ret;
 
@@ -509,7 +509,7 @@
 		bb_perror_msg("ioctl %#x failed", request);
 	return ret;
 }
-int bb_xioctl(int fd, unsigned request, void *argp)
+int FAST_FUNC bb_xioctl(int fd, unsigned request, void *argp)
 {
 	int ret;
 
diff --git a/libbb/xgetcwd.c b/libbb/xgetcwd.c
index c194e23..eefe1d6 100644
--- a/libbb/xgetcwd.c
+++ b/libbb/xgetcwd.c
@@ -14,7 +14,7 @@
    If argument is not NULL (previous usage allocate memory), call free()
 */
 
-char *
+char* FAST_FUNC
 xrealloc_getcwd_or_warn(char *cwd)
 {
 #define PATH_INCR 64
diff --git a/libbb/xgethostbyname.c b/libbb/xgethostbyname.c
index 3bb522d..f1839f7 100644
--- a/libbb/xgethostbyname.c
+++ b/libbb/xgethostbyname.c
@@ -10,7 +10,7 @@
 //#include <netdb.h>
 #include "libbb.h"
 
-struct hostent *xgethostbyname(const char *name)
+struct hostent* FAST_FUNC xgethostbyname(const char *name)
 {
 	struct hostent *retval = gethostbyname(name);
 	if (!retval)
diff --git a/libbb/xreadlink.c b/libbb/xreadlink.c
index 0b961b6..2cfc575 100644
--- a/libbb/xreadlink.c
+++ b/libbb/xreadlink.c
@@ -10,7 +10,7 @@
  * NOTE: This function returns a malloced char* that you will have to free
  * yourself.
  */
-char *xmalloc_readlink(const char *path)
+char* FAST_FUNC xmalloc_readlink(const char *path)
 {
 	enum { GROWBY = 80 }; /* how large we will grow strings by */
 
@@ -42,7 +42,7 @@
  * those at the tail.
  * A malloced char* is returned, which must be freed by the caller.
  */
-char *xmalloc_follow_symlinks(const char *path)
+char* FAST_FUNC xmalloc_follow_symlinks(const char *path)
 {
 	char *buf;
 	char *lpc;
@@ -84,7 +84,7 @@
 	}
 }
 
-char *xmalloc_readlink_or_warn(const char *path)
+char* FAST_FUNC xmalloc_readlink_or_warn(const char *path)
 {
 	char *buf = xmalloc_readlink(path);
 	if (!buf) {
@@ -96,7 +96,7 @@
 
 /* UNUSED */
 #if 0
-char *xmalloc_realpath(const char *path)
+char* FAST_FUNC xmalloc_realpath(const char *path)
 {
 #if defined(__GLIBC__) && !defined(__UCLIBC__)
 	/* glibc provides a non-standard extension */
diff --git a/libbb/xregcomp.c b/libbb/xregcomp.c
index 157132c..abfa35f 100644
--- a/libbb/xregcomp.c
+++ b/libbb/xregcomp.c
@@ -11,7 +11,7 @@
 #include "libbb.h"
 #include "xregex.h"
 
-char* regcomp_or_errmsg(regex_t *preg, const char *regex, int cflags)
+char* FAST_FUNC regcomp_or_errmsg(regex_t *preg, const char *regex, int cflags)
 {
 	int ret = regcomp(preg, regex, cflags);
 	if (ret) {
@@ -23,7 +23,7 @@
 	return NULL;
 }
 
-void xregcomp(regex_t *preg, const char *regex, int cflags)
+void FAST_FUNC xregcomp(regex_t *preg, const char *regex, int cflags)
 {
 	char *errmsg = regcomp_or_errmsg(preg, regex, cflags);
 	if (errmsg) {