emulation is now controlled by sysctl. changes:

add e_flags to struct emul.  this stores on/off and native flags.
check for emul enabled in check_exec().  gather all the emuls into a
emulsw so a sysctl can find them.  create sysctl.  move maxhdrsiz calcualation
into init_main so it cleans up sys_execve codepath.  teach sysctl utility
to grok kern.emul hierarchy.
requested and ok deraadt@  some comments from mickey@

9 files changed, 284 insertions, 24 deletions

diff --git a/sbin/sysctl/sysctl.c b/sbin/sysctl/sysctl.c
index 9b7c3095ed9..0cf34004c74 100644
--- a/sbin/sysctl/sysctl.c
+++ b/sbin/sysctl/sysctl.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: sysctl.c,v 1.95 2003/06/11 06:22:15 deraadt Exp $	*/
+/*	$OpenBSD: sysctl.c,v 1.96 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: sysctl.c,v 1.9 1995/09/30 07:12:50 thorpej Exp $	*/
 
 /*
@@ -40,7 +40,7 @@ static const char copyright[] =
 #if 0
 static const char sccsid[] = "@(#)sysctl.c	8.5 (Berkeley) 5/9/95";
 #else
-static char *rcsid = "$OpenBSD: sysctl.c,v 1.95 2003/06/11 06:22:15 deraadt Exp $";
+static char *rcsid = "$OpenBSD: sysctl.c,v 1.96 2003/08/21 18:56:07 tedu Exp $";
 #endif
 #endif /* not lint */
 
@@ -206,6 +206,7 @@ int sysctl_seminfo(char *, char **, int *, int, int *);
 int sysctl_shminfo(char *, char **, int *, int, int *);
 int sysctl_watchdog(char *, char **, int *, int, int *);
 int sysctl_sensors(char *, char **, int *, int, int *);
+int sysctl_emul(char *, char *, int);
 #ifdef CPU_CHIPSET
 int sysctl_chipset(char *, char **, int *, int, int *);
 #endif
@@ -430,6 +431,9 @@ parse(char *string, int flags)
 			if (len < 0)
 				return;
 			break;
+		case KERN_EMUL:
+			sysctl_emul(string, newval, flags);
+			return;
 		}
 		break;
 
@@ -1964,6 +1968,133 @@ sysctl_sensors(char *string, char **bufpp, int mib[], int flags, int *typep)
 	return (3);
 }
 
+char	**emul_names;
+int	emul_num;
+void	emul_init(void);
+
+int
+sysctl_emul(char *string, char *newval, int flags)
+{
+	int mib[4], enabled, i, old, found = 0;
+	char *head, *target;
+	size_t len;
+
+	emul_init();
+
+	mib[0] = CTL_KERN;
+	mib[1] = KERN_EMUL;
+	mib[3] = KERN_EMUL_ENABLED;
+	head = "kern.emul.";
+
+	if (aflag || strcmp(string, "kern.emul") == 0) {
+		if (strcmp(string, "kern.emul") == 0 && wflag) {
+			warnx("%s: specification is incomplete", string);
+			return (1);
+		}
+		if (nflag)
+			printf("%d\n", emul_num);
+		else
+			printf("%snemuls = %d\n", head, emul_num);
+		for (i = 0; i < emul_num; i++) {
+			mib[2] = i + 1;
+			len = sizeof(int);
+			if (sysctl(mib, 4, &enabled, &len, NULL, 0) == -1) {
+				warn("%s", string);
+				continue;
+			}
+			if (nflag)
+				printf("%d\n", enabled);
+			else
+				printf("%s%s = %d\n", head, emul_names[i],
+				    enabled);
+		}
+		return (0);
+	}
+	/* User specified a third level name */
+	target = strrchr(string, '.');
+	target++;
+	if (strcmp(string, "nemuls") == 0) {
+		if (newval) {
+			warnx("Operation not permitted");
+			return (1);
+		}
+		if (nflag)
+			printf("%d\n", emul_num);
+		else
+			printf("%snemuls = %d\n", head, emul_num);
+		return (0);
+	}
+	for (i = 0; i < emul_num; i++) {
+		if (strcmp(target, emul_names[i]))
+			continue;
+		found = 1;
+		mib[2] = i + 1;
+		len = sizeof(int);
+		if (newval) {
+			enabled = atoi(newval);
+			if (sysctl(mib, 4, &old, &len, &enabled, len) == -1) {
+				warn("%s", string);
+				continue;
+			}
+			if (nflag)
+				printf("%d\n", enabled);
+			else
+				printf("%s%s: %d -> %d\n", head, target, old,
+				    enabled);
+		} else {
+			if (sysctl(mib, 4, &enabled, &len, NULL, 0) == -1) {
+				warn("%s", string);
+				continue;
+			}
+			if (nflag)
+				printf("%d\n", enabled);
+			else
+				printf("%s%s = %d\n", head, target, enabled);
+		}
+	}
+	if (!found)
+		warnx("third level name %s in kern.emul is invalid",
+		    string);
+	return (0);
+
+
+}
+
+void
+emul_init(void)
+{
+	static int done;
+	char string[16];
+	int mib[4], i;
+	size_t len;
+
+	if (done)
+		return;
+	done = 1;
+
+	mib[0] = CTL_KERN;
+	mib[1] = KERN_EMUL;
+	mib[2] = KERN_EMUL_NUM;
+	len = sizeof(int);
+	if (sysctl(mib, 3, &emul_num, &len, NULL, 0) == -1)
+		return;
+
+	emul_names = malloc(emul_num * sizeof(char *));
+	if (emul_names == NULL) {
+		warn("emul_init");
+		return;
+	}
+
+	for (i = 0; i < emul_num; i++) {
+		mib[2] = i + 1;
+		mib[3] = 0;
+		len = sizeof(string);
+		if (sysctl(mib, 4, string, &len, NULL, 0) == -1)
+			break;
+		emul_names[i] = strdup(string);
+	}
+}
+
 /*
  * Scan a list of names searching for a particular name.
  */
diff --git a/sys/compat/linux/linux_exec.c b/sys/compat/linux/linux_exec.c
index 204eca3692a..b0203f6f956 100644
--- a/sys/compat/linux/linux_exec.c
+++ b/sys/compat/linux/linux_exec.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: linux_exec.c,v 1.21 2003/06/21 00:42:58 tedu Exp $	*/
+/*	$OpenBSD: linux_exec.c,v 1.22 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: linux_exec.c,v 1.13 1996/04/05 00:01:10 christos Exp $	*/
 
 /*-
@@ -110,6 +110,7 @@ struct emul emul_linux_aout = {
 	NULL,
 	linux_sigcode,
 	linux_esigcode,
+	0,
 	NULL,
 	linux_e_proc_exec,
 	linux_e_proc_fork,
@@ -134,6 +135,7 @@ struct emul emul_linux_elf = {
 	exec_elf32_fixup,
 	linux_sigcode,
 	linux_esigcode,
+	0,
 	NULL,
 	linux_e_proc_exec,
 	linux_e_proc_fork,
diff --git a/sys/kern/exec_conf.c b/sys/kern/exec_conf.c
index 0af97b9e16b..60d92b590d4 100644
--- a/sys/kern/exec_conf.c
+++ b/sys/kern/exec_conf.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: exec_conf.c,v 1.12 2001/11/14 14:37:22 hugh Exp $	*/
+/*	$OpenBSD: exec_conf.c,v 1.13 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: exec_conf.c,v 1.16 1995/12/09 05:34:47 cgd Exp $	*/
 
 /*
@@ -121,3 +121,73 @@ struct execsw execsw[] = {
 };
 int nexecs = (sizeof execsw / sizeof(*execsw));
 int exec_maxhdrsz;
+
+extern struct emul emul_native, emul_elf32, emul_elf64, emul_aout,
+	emul_bsdos, emul_aout_freebsd, emul_elf_freebsd, emul_hpux,
+	emul_ibcs2, emul_linux_elf, emul_linux_aout, emul_elf64_netbsd,
+	emul_osf1, emul_sunos, emul_svr4, emul_ultrix;
+struct emul *emulsw[] = {
+	&emul_native,
+#ifdef _KERN_DO_ELF
+	&emul_elf32,
+#endif
+#ifdef _KERN_DO_ELF64
+	&emul_elf64,
+#endif
+#if defined (_KERN_DO_AOUT) && defined (COMPAT_AOUT)
+	&emul_aout,
+#endif
+#ifdef COMPAT_BSDOS
+	&emul_bsdos,
+#endif
+#ifdef COMPAT_FREEBSD
+	&emul_aout_freebsd,
+	&emul_elf_freebsd,
+#endif
+#ifdef COMPAT_HPUX
+	&emul_hpux,
+#endif
+#ifdef COMPAT_IBCS2
+	&emul_ibcs2,
+#endif
+#ifdef COMPAT_LINUX
+	&emul_linux_elf,
+	&emul_linux_aout,
+#endif
+#if defined (COMPAT_NETBSD) && defined (_KERN_DO_ELF64)
+	&emul_elf64_netbsd,
+#endif
+#ifdef COMPAT_OSF1
+	&emul_osf1,
+#endif
+#ifdef COMPAT_SUNOS
+	&emul_sunos,
+#endif
+#ifdef COMPAT_SVR4
+	&emul_svr4,
+#endif
+#ifdef COMPAT_ULTRIX
+	&emul_ultrix,
+#endif
+};
+int	nemuls = sizeof(emulsw) / sizeof(*emulsw);
+
+void	init_exec(void);
+
+void
+init_exec(void)
+{
+	int i;
+
+	/*
+	 * figure out the maximum size of an exec header.
+	 * XXX should be able to keep LKM code from modifying exec switch
+	 * when we're still using it, but...
+	 */
+	if (exec_maxhdrsz == 0) {
+		for (i = 0; i < nexecs; i++)
+			if (execsw[i].es_check != NULL
+			    && execsw[i].es_hdrsz > exec_maxhdrsz)
+				exec_maxhdrsz = execsw[i].es_hdrsz;
+	}
+}
diff --git a/sys/kern/exec_elf.c b/sys/kern/exec_elf.c
index 2ff39545f8d..96083e6bfe9 100644
--- a/sys/kern/exec_elf.c
+++ b/sys/kern/exec_elf.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: exec_elf.c,v 1.47 2003/05/17 14:02:06 grange Exp $	*/
+/*	$OpenBSD: exec_elf.c,v 1.48 2003/08/21 18:56:07 tedu Exp $	*/
 
 /*
  * Copyright (c) 1996 Per Fogelstrom
@@ -137,6 +137,7 @@ struct emul ELFNAMEEND(emul) = {
 	ELFNAME2(exec,fixup),
 	sigcode,
 	esigcode,
+	EMUL_ENABLED | EMUL_NATIVE,
 };
 
 /*
diff --git a/sys/kern/init_main.c b/sys/kern/init_main.c
index ada370a248e..e9393e2a2bd 100644
--- a/sys/kern/init_main.c
+++ b/sys/kern/init_main.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: init_main.c,v 1.105 2003/06/02 23:28:05 millert Exp $	*/
+/*	$OpenBSD: init_main.c,v 1.106 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: init_main.c,v 1.84.4.1 1996/06/02 09:08:06 mrg Exp $	*/
 
 /*
@@ -135,6 +135,7 @@ void	start_cleaner(void *);
 void	start_update(void *);
 void	start_reaper(void *);
 void    start_crypto(void *);
+void	init_exec(void);
 
 extern char sigcode[], esigcode[];
 #ifdef SYSCALL_DEBUG
@@ -159,8 +160,10 @@ struct emul emul_native = {
 	NULL,
 	sigcode,
 	esigcode,
+	EMUL_ENABLED | EMUL_NATIVE,
 };
 
+
 /*
  * System startup; initialize the world, create process 0, mount root
  * filesystem, and fork to create init and pagedaemon.  Most of the
@@ -373,6 +376,9 @@ main(framep)
 		guard[i] = arc4random();
 #endif
 
+	/* init exec and emul */
+	init_exec();
+
 	/* Start the scheduler */
 	scheduler_start();
 
diff --git a/sys/kern/kern_exec.c b/sys/kern/kern_exec.c
index af4347fbfe7..8be38c028df 100644
--- a/sys/kern/kern_exec.c
+++ b/sys/kern/kern_exec.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_exec.c,v 1.80 2003/06/21 00:42:58 tedu Exp $	*/
+/*	$OpenBSD: kern_exec.c,v 1.81 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: kern_exec.c,v 1.75 1996/02/09 18:59:28 christos Exp $	*/
 
 /*-
@@ -175,6 +175,8 @@ check_exec(p, epp)
 		if (execsw[i].es_check == NULL)
 			continue;
 		newerror = (*execsw[i].es_check)(p, epp);
+		if (!newerror && !(epp->ep_emul->e_flags & EMUL_ENABLED))
+			newerror = ENOEXEC;
 		/* make sure the first "interesting" error code is saved. */
 		if (!newerror || error == ENOEXEC)
 			error = newerror;
@@ -235,7 +237,7 @@ sys_execve(p, v, retval)
 		syscallarg(char * *) argp;
 		syscallarg(char * *) envp;
 	} */ *uap = v;
-	int error, i;
+	int error;
 	struct exec_package pack;
 	struct nameidata nid;
 	struct vattr attr;
@@ -259,18 +261,6 @@ sys_execve(p, v, retval)
 	 */
 	p->p_flag |= P_INEXEC;
 
-	/*
-	 * figure out the maximum size of an exec header, if necessary.
-	 * XXX should be able to keep LKM code from modifying exec switch
-	 * when we're still using it, but...
-	 */
-	if (exec_maxhdrsz == 0) {
-		for (i = 0; i < nexecs; i++)
-			if (execsw[i].es_check != NULL
-			    && execsw[i].es_hdrsz > exec_maxhdrsz)
-				exec_maxhdrsz = execsw[i].es_hdrsz;
-	}
-
 	/* init the namei data to point the file user's program name */
 	NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE, SCARG(uap, path), p);
 
diff --git a/sys/kern/kern_sysctl.c b/sys/kern/kern_sysctl.c
index a15d511470b..b75ad814560 100644
--- a/sys/kern/kern_sysctl.c
+++ b/sys/kern/kern_sysctl.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_sysctl.c,v 1.85 2003/08/15 20:32:18 tedu Exp $	*/
+/*	$OpenBSD: kern_sysctl.c,v 1.86 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: kern_sysctl.c,v 1.17 1996/05/20 17:49:05 mrg Exp $	*/
 
 /*-
@@ -92,6 +92,7 @@ int sysctl_diskinit(int, struct proc *);
 int sysctl_proc_args(int *, u_int, void *, size_t *, struct proc *);
 int sysctl_intrcnt(int *, u_int, void *, size_t *);
 int sysctl_sensors(int *, u_int, void *, size_t *, void *, size_t);
+int sysctl_emul(int *, u_int, void *, size_t *, void *, size_t);
 
 /*
  * Lock to avoid too many processes vslocking a large amount of memory
@@ -268,6 +269,7 @@ kern_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
 		case KERN_SHMINFO:
 		case KERN_INTRCNT:
 		case KERN_WATCHDOG:
+		case KERN_EMUL:
 			break;
 		default:
 			return (ENOTDIR);	/* overloaded */
@@ -482,6 +484,9 @@ kern_sysctl(name, namelen, oldp, oldlenp, newp, newlen, p)
 	case KERN_WATCHDOG:
 		return (sysctl_wdog(name + 1, namelen - 1, oldp, oldlenp,
 		    newp, newlen));
+	case KERN_EMUL:
+		return (sysctl_emul(name + 1, namelen - 1, oldp, oldlenp,
+		    newp, newlen));
 	default:
 		return (EOPNOTSUPP);
 	}
@@ -1516,3 +1521,40 @@ sysctl_sensors(int *name, u_int namelen, void *oldp, size_t *oldlenp,
 
 	return (sysctl_rdstruct(oldp, oldlenp, newp, s, sizeof(struct sensor)));
 }
+
+int
+sysctl_emul(int *name, u_int namelen, void *oldp, size_t *oldlenp,
+    void *newp, size_t newlen)
+{
+	int enabled, error;
+	struct emul *e;
+
+	if (name[0] == KERN_EMUL_NUM) {
+		if (namelen != 1)
+			return (ENOTDIR);
+		return (sysctl_rdint(oldp, oldlenp, newp, nemuls));
+	}
+
+	if (namelen != 2)
+		return (ENOTDIR);
+	if (name[0] > nemuls || name[0] < 0)
+		return (EINVAL);
+	e = emulsw[name[0] - 1];
+
+	switch (name[1]) {
+	case KERN_EMUL_NAME:
+		return (sysctl_rdstring(oldp, oldlenp, newp, e->e_name));
+	case KERN_EMUL_ENABLED:
+		if (e->e_flags & EMUL_NATIVE)
+			return (sysctl_rdint(oldp, oldlenp, newp, 1));
+		else {
+			enabled = (e->e_flags & EMUL_ENABLED);
+			error = sysctl_int(oldp, oldlenp, newp, newlen,
+			    &enabled);
+			e->e_flags |= (enabled & EMUL_ENABLED);
+			return (error);
+		}
+	default:
+		return (EINVAL);
+	}
+}
diff --git a/sys/sys/proc.h b/sys/sys/proc.h
index 0f2c883c272..c7a01a41188 100644
--- a/sys/sys/proc.h
+++ b/sys/sys/proc.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: proc.h,v 1.66 2003/08/03 19:25:49 millert Exp $	*/
+/*	$OpenBSD: proc.h,v 1.67 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: proc.h,v 1.44 1996/04/22 01:23:21 christos Exp $	*/
 
 /*-
@@ -95,12 +95,19 @@ struct	emul {
 	int	(*e_fixup)(struct proc *, struct exec_package *);
 	char	*e_sigcode;		/* Start of sigcode */
 	char	*e_esigcode;		/* End of sigcode */
+	int	e_flags;		/* Flags, see below */
 	struct uvm_object *e_sigobject;	/* shared sigcode object */
 					/* Per-process hooks */
 	void	(*e_proc_exec)(struct proc *, struct exec_package *);
 	void	(*e_proc_fork)(struct proc *p, struct proc *parent);
 	void	(*e_proc_exit)(struct proc *);
 };
+/* Flags for e_flags */
+#define	EMUL_ENABLED	0x0001		/* Allow exec to continue */
+#define	EMUL_NATIVE	0x0002		/* Always enabled */
+
+extern struct emul *emulsw[];		/* All emuls in system */
+extern int nemuls;			/* Number of emuls */
 
 /*
  * Description of a process.
diff --git a/sys/sys/sysctl.h b/sys/sys/sysctl.h
index 64474a4aef9..38d9f3706a7 100644
--- a/sys/sys/sysctl.h
+++ b/sys/sys/sysctl.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: sysctl.h,v 1.64 2003/06/02 23:28:22 millert Exp $	*/
+/*	$OpenBSD: sysctl.h,v 1.65 2003/08/21 18:56:07 tedu Exp $	*/
 /*	$NetBSD: sysctl.h,v 1.16 1996/04/09 20:55:36 cgd Exp $	*/
 
 /*
@@ -177,7 +177,8 @@ struct ctlname {
 #define	KERN_SHMINFO		62	/* struct: SysV struct shminfo */
 #define KERN_INTRCNT		63	/* node: interrupt counters */
 #define	KERN_WATCHDOG		64	/* node: watchdog */
-#define	KERN_MAXID		65	/* number of valid kern ids */
+#define	KERN_EMUL		65	/* node: emuls */
+#define	KERN_MAXID		66	/* number of valid kern ids */
 
 #define	CTL_KERN_NAMES { \
 	{ 0, 0 }, \
@@ -245,9 +246,19 @@ struct ctlname {
 	{ "shminfo", CTLTYPE_STRUCT }, \
 	{ "intrcnt", CTLTYPE_NODE }, \
  	{ "watchdog", CTLTYPE_NODE }, \
+ 	{ "emul", CTLTYPE_NODE }, \
 }
 
 /*
+ * KERN_EMUL subtypes.
+ */
+#define	KERN_EMUL_NUM		0
+/* Fourth level sysctl names */
+#define KERN_EMUL_NAME		0
+#define KERN_EMUL_ENABLED	1
+
+
+/*
  * KERN_PROC subtypes
  */
 #define	KERN_PROC_ALL		0	/* everything but kernel threads */