xnu-2422.110.17.tar.gz

[apple/xnu.git] / bsd / kern / sysv_sem.c

diff --git a/bsd/kern/sysv_sem.c b/bsd/kern/sysv_sem.c
index 8c4c3d47c81e31f7188b664cf77211d9ae4e61f6..f98530962b733037533b7628b432132c04c3ef32 100644 (file)
--- a/bsd/kern/sysv_sem.c
+++ b/bsd/kern/sysv_sem.c
@@ -1,16 +1,19 @@
 /*
 /*

- * Copyright (c) 2000-2003 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2007 Apple Inc. All rights reserved.

  *
  *

- * @APPLE_LICENSE_HEADER_START@
- * 
- * Copyright (c) 1999-2003 Apple Computer, Inc.  All Rights Reserved.
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@

  * 
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * 
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in

- * compliance with the License. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.

  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * 
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER


@@ -20,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 

- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@

  */
 /*
  * Implementation of SVID semaphores
  */
 /*
  * Implementation of SVID semaphores


@@ -32,52 +35,51 @@
 /*
  * John Bellardo modified the implementation for Darwin. 12/2000
  */
 /*
  * John Bellardo modified the implementation for Darwin. 12/2000
  */

+/*
+ * NOTICE: This file was modified by McAfee Research in 2004 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ * Copyright (c) 2005-2006 SPARTA, Inc.
+ */

 
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/kernel.h>
 
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/kernel.h>

-#include <sys/proc.h>
-#include <sys/sem.h>
+#include <sys/proc_internal.h>
+#include <sys/kauth.h>
+#include <sys/sem_internal.h>

 #include <sys/malloc.h>
 #include <mach/mach_types.h>
 
 #include <sys/filedesc.h>
 #include <sys/malloc.h>
 #include <mach/mach_types.h>
 
 #include <sys/filedesc.h>

-#include <sys/file.h>
-#include <sys/kern_audit.h>
+#include <sys/file_internal.h>

 #include <sys/sysctl.h>
 #include <sys/sysctl.h>

+#include <sys/ipcs.h>
+#include <sys/sysent.h>
+#include <sys/sysproto.h>
+#if CONFIG_MACF
+#include <security/mac_framework.h>
+#endif

 
 

-/*#include <sys/sysproto.h>*/
-/*#include <sys/sysent.h>*/
+#include <security/audit/audit.h>

 
 

-/* Uncomment this line to see the debugging output */
-/* #define SEM_DEBUG */
+#if SYSV_SEM

 
 

-/* Macros to deal with the semaphore subsystem lock.  The lock currently uses
- * the semlock_holder static variable as a mutex.  NULL means no lock, any
- * value other than NULL means locked.  semlock_holder is used because it was
- * present in the code before the Darwin port, and for no other reason.
- * When the time comes to relax the funnel requirements of the kernel only
- * these macros should need to be changed.  A spin lock would work well.
- */
-/* Aquire the lock */
-#define SUBSYSTEM_LOCK_AQUIRE(p) { sysv_sem_aquiring_threads++; \
-    while (semlock_holder != NULL) \
-        (void) tsleep((caddr_t)&semlock_holder, (PZERO - 4), "sysvsem", 0); \
-    semlock_holder = p; \
-    sysv_sem_aquiring_threads--; }

 
 

-/* Release the lock */
-#define SUBSYSTEM_LOCK_RELEASE { semlock_holder = NULL; wakeup((caddr_t)&semlock_holder); }
+/* Uncomment this line to see the debugging output */
+/* #define SEM_DEBUG */

 
 

-/* Release the lock and return a value */
-#define UNLOCK_AND_RETURN(ret) { SUBSYSTEM_LOCK_RELEASE; return(ret); }
+/* Uncomment this line to see MAC debugging output. */
+/* #define     MAC_DEBUG */
+#if CONFIG_MACF_DEBUG
+#define        MPRINTF(a)      printf(a)
+#else
+#define        MPRINTF(a)
+#endif

 
 

-#define M_SYSVSEM      M_SUBPROC
+#define M_SYSVSEM      M_TEMP

 
 

-#if 0
-static void seminit __P((void *));
-SYSINIT(sysv_sem, SI_SUB_SYSV_SEM, SI_ORDER_FIRST, seminit, NULL)
-#endif 0

 
 /* Hard system limits to avoid resource starvation / DOS attacks.
  * These are not needed if we can make the semaphore pages swappable.
 
 /* Hard system limits to avoid resource starvation / DOS attacks.
  * These are not needed if we can make the semaphore pages swappable.


@@ -112,141 +114,162 @@ struct seminfo seminfo = {
        SEMAEM  /* adjust on exit max value */
 };
 
        SEMAEM  /* adjust on exit max value */
 };
 

-/* A counter so the module unload code knows when there are no more processes using
- * the sysv_sem code */
-static long sysv_sem_sleeping_threads = 0;
-static long sysv_sem_aquiring_threads = 0;
-
-struct semctl_args;
-int semctl __P((struct proc *p, struct semctl_args *uap, int *));
-struct semget_args;
-int semget __P((struct proc *p, struct semget_args *uap, int *));
-struct semop_args;
-int semop __P((struct proc *p, struct semop_args *uap, int *));
-struct semconfig_args;
-int semconfig __P((struct proc *p, struct semconfig_args *uap, int *));
-
-
-static struct sem_undo *semu_alloc __P((struct proc *p));
-static int semundo_adjust __P((struct proc *p, struct sem_undo **supptr, 
-               int semid, int semnum, int adjval));
-static void semundo_clear __P((int semid, int semnum));

 
 

-typedef int     sy_call_t __P((struct proc *, void *, int *));
+static int semu_alloc(struct proc *p);
+static int semundo_adjust(struct proc *p, int *supidx, 
+               int semid, int semnum, int adjval);
+static void semundo_clear(int semid, int semnum);

 
 /* XXX casting to (sy_call_t *) is bogus, as usual. */
 static sy_call_t *semcalls[] = {
        (sy_call_t *)semctl, (sy_call_t *)semget,
 
 /* XXX casting to (sy_call_t *) is bogus, as usual. */
 static sy_call_t *semcalls[] = {
        (sy_call_t *)semctl, (sy_call_t *)semget,

-       (sy_call_t *)semop, (sy_call_t *)semconfig
+       (sy_call_t *)semop

 };
 
 };
 

-static int     semtot = 0;                     /* # of used semaphores */
-struct semid_ds *sema = NULL;                  /* semaphore id pool */
-struct sem *sem =  NULL;                       /* semaphore pool */
-static struct sem_undo *semu_list = NULL;   /* list of active undo structures */
-struct sem_undo *semu = NULL;                  /* semaphore undo pool */
+static int             semtot = 0;             /* # of used semaphores */
+struct semid_kernel    *sema = NULL;           /* semaphore id pool */
+struct sem             *sem_pool =  NULL;      /* semaphore pool */
+static int             semu_list_idx = -1;     /* active undo structures */
+struct sem_undo                *semu = NULL;           /* semaphore undo pool */

 
 

-static struct proc *semlock_holder = NULL;

 
 

-/* seminit no longer needed.  The data structures are grown dynamically */
-void
-seminit()
+void sysv_sem_lock_init(void);
+static lck_grp_t       *sysv_sem_subsys_lck_grp;
+static lck_grp_attr_t  *sysv_sem_subsys_lck_grp_attr;
+static lck_attr_t      *sysv_sem_subsys_lck_attr;
+static lck_mtx_t        sysv_sem_subsys_mutex;
+
+#define SYSV_SEM_SUBSYS_LOCK() lck_mtx_lock(&sysv_sem_subsys_mutex)
+#define SYSV_SEM_SUBSYS_UNLOCK() lck_mtx_unlock(&sysv_sem_subsys_mutex)
+
+
+__private_extern__ void
+sysv_sem_lock_init( void )

 {
 {

+
+    sysv_sem_subsys_lck_grp_attr = lck_grp_attr_alloc_init();
+
+    sysv_sem_subsys_lck_grp = lck_grp_alloc_init("sysv_sem_subsys_lock", sysv_sem_subsys_lck_grp_attr);
+
+    sysv_sem_subsys_lck_attr = lck_attr_alloc_init();
+    lck_mtx_init(&sysv_sem_subsys_mutex, sysv_sem_subsys_lck_grp, sysv_sem_subsys_lck_attr);
+}
+
+static __inline__ user_time_t
+sysv_semtime(void)
+{
+       struct timeval  tv;
+       microtime(&tv);
+       return (tv.tv_sec);

 }
 
 /*
 }
 
 /*

- * Entry point for all SEM calls
+ * XXX conversion of internal user_time_t to external tume_t loses
+ * XXX precision; not an issue for us now, since we are only ever
+ * XXX setting 32 bits worth of time into it.

  *
  *

- * In Darwin this is no longer the entry point.  It will be removed after
- *  the code has been tested better.
+ * pad field contents are not moved correspondingly; contents will be lost
+ *
+ * NOTE: Source and target may *NOT* overlap! (target is smaller)

  */
  */

-struct semsys_args {
-       u_int   which;
-       int     a2;
-       int     a3;
-       int     a4;
-       int     a5;
-};
-int
-semsys(p, uap, retval)
-       struct proc *p;
-       /* XXX actually varargs. */
-       struct semsys_args *uap;
-       register_t *retval;
+static void
+semid_ds_kernelto32(struct user_semid_ds *in, struct user32_semid_ds *out)

 {
 {

+       out->sem_perm = in->sem_perm;
+       out->sem_base = CAST_DOWN_EXPLICIT(__int32_t,in->sem_base);
+       out->sem_nsems = in->sem_nsems;
+       out->sem_otime = in->sem_otime;         /* XXX loses precision */
+       out->sem_ctime = in->sem_ctime;         /* XXX loses precision */
+}

 
 

-       /* The individual calls handling the locking now */
-       /*while (semlock_holder != NULL && semlock_holder != p)
-               (void) tsleep((caddr_t)&semlock_holder, (PZERO - 4), "semsys", 0);
-        */
-
-       if (uap->which >= sizeof(semcalls)/sizeof(semcalls[0]))
-               return (EINVAL);
-       return ((*semcalls[uap->which])(p, &uap->a2, retval));
+static void
+semid_ds_kernelto64(struct user_semid_ds *in, struct user64_semid_ds *out)
+{
+       out->sem_perm = in->sem_perm;
+       out->sem_base = CAST_DOWN_EXPLICIT(__int32_t,in->sem_base);
+       out->sem_nsems = in->sem_nsems;
+       out->sem_otime = in->sem_otime;         /* XXX loses precision */
+       out->sem_ctime = in->sem_ctime;         /* XXX loses precision */

 }
 
 /*
 }
 
 /*

- * Lock or unlock the entire semaphore facility.
+ * pad field contents are not moved correspondingly; contents will be lost

  *
  *

- * This will probably eventually evolve into a general purpose semaphore
- * facility status enquiry mechanism (I don't like the "read /dev/kmem"
- * approach currently taken by ipcs and the amount of info that we want
- * to be able to extract for ipcs is probably beyond what the capability
- * of the getkerninfo facility.
+ * NOTE: Source and target may are permitted to overlap! (source is smaller);
+ * this works because we copy fields in order from the end of the struct to
+ * the beginning.

  *
  *

- * At the time that the current version of semconfig was written, ipcs is
- * the only user of the semconfig facility.  It uses it to ensure that the
- * semaphore facility data structures remain static while it fishes around
- * in /dev/kmem.
+ * XXX use CAST_USER_ADDR_T() for lack of a CAST_USER_TIME_T(); net effect
+ * XXX is the same.

  */
  */

+static void
+semid_ds_32tokernel(struct user32_semid_ds *in, struct user_semid_ds *out)
+{
+       out->sem_ctime = in->sem_ctime;
+       out->sem_otime = in->sem_otime;
+       out->sem_nsems = in->sem_nsems;
+       out->sem_base = (void *)(uintptr_t)in->sem_base;
+       out->sem_perm = in->sem_perm;
+}

 
 

-#ifndef _SYS_SYSPROTO_H_
-struct semconfig_args {
-       semconfig_ctl_t flag;
-};
-#endif
-
-int
-semconfig(p, uap, retval)
-       struct proc *p;
-       struct semconfig_args *uap;
-       register_t *retval;
+static void
+semid_ds_64tokernel(struct user64_semid_ds *in, struct user_semid_ds *out)

 {
 {

-       int eval = 0;
+       out->sem_ctime = in->sem_ctime;
+       out->sem_otime = in->sem_otime;
+       out->sem_nsems = in->sem_nsems;
+       out->sem_base = (void *)(uintptr_t)in->sem_base;
+       out->sem_perm = in->sem_perm;
+}

 
 

-       switch (uap->flag) {
-       case SEM_CONFIG_FREEZE:
-               SUBSYSTEM_LOCK_AQUIRE(p);
-               break;

 
 

-       case SEM_CONFIG_THAW:
-               SUBSYSTEM_LOCK_RELEASE;
-               break;
+/*
+ * semsys
+ *
+ * Entry point for all SEM calls: semctl, semget, semop
+ *
+ * Parameters: p       Process requesting the call
+ *             uap     User argument descriptor (see below)
+ *             retval  Return value of the selected sem call
+ *
+ * Indirect parameters:        uap->which      sem call to invoke (index in array of sem calls)
+ *                     uap->a2         User argument descriptor
+ *                 
+ * Returns:    0       Success
+ *             !0      Not success
+ *
+ * Implicit returns: retval    Return value of the selected sem call
+ *
+ * DEPRECATED:  This interface should not be used to call the other SEM
+ *             functions (semctl, semget, semop). The correct usage is
+ *             to call the other SEM functions directly.
+ *
+ */
+int
+semsys(struct proc *p, struct semsys_args *uap, int32_t *retval)
+{

 
 

-       default:
-               printf("semconfig: unknown flag parameter value (%d) - ignored\n",
-                   uap->flag);
-               eval = EINVAL;
-               break;
-       }
+       /* The individual calls handling the locking now */

 
 

-       *retval = 0;
-       return(eval);
+       if (uap->which >= sizeof(semcalls)/sizeof(semcalls[0]))
+               return (EINVAL);
+       return ((*semcalls[uap->which])(p, &uap->a2, retval));

 }
 
 }
 

-/* Expand the semu array to the given capacity.  If the expansion fails
+/*
+ * Expand the semu array to the given capacity.  If the expansion fails

  * return 0, otherwise return 1.
  *
  * Assumes we already have the subsystem lock.
  */
 static int
  * return 0, otherwise return 1.
  *
  * Assumes we already have the subsystem lock.
  */
 static int

-grow_semu_array(newSize)
-       int newSize;
+grow_semu_array(int newSize)

 {
 {

-       register int i, j;
+       register int i;

        register struct sem_undo *newSemu;
        register struct sem_undo *newSemu;

+

        if (newSize <= seminfo.semmnu)
        if (newSize <= seminfo.semmnu)

-               return 0;
+               return 1;

        if (newSize > limitseminfo.semmnu) /* enforce hard limit */
        {
 #ifdef SEM_DEBUG
        if (newSize > limitseminfo.semmnu) /* enforce hard limit */
        {
 #ifdef SEM_DEBUG


@@ -261,8 +284,8 @@ grow_semu_array(newSize)
 #ifdef SEM_DEBUG
        printf("growing semu[] from %d to %d\n", seminfo.semmnu, newSize);
 #endif
 #ifdef SEM_DEBUG
        printf("growing semu[] from %d to %d\n", seminfo.semmnu, newSize);
 #endif

-       MALLOC(newSemu, struct sem_undo*, sizeof(struct sem_undo)*newSize,
-               M_SYSVSEM, M_WAITOK);
+       MALLOC(newSemu, struct sem_undo *, sizeof (struct sem_undo) * newSize,
+              M_SYSVSEM, M_WAITOK | M_ZERO);

        if (NULL == newSemu)
        {
 #ifdef SEM_DEBUG
        if (NULL == newSemu)
        {
 #ifdef SEM_DEBUG


@@ -271,17 +294,16 @@ grow_semu_array(newSize)
                return 0;
        }
 
                return 0;
        }
 

-               /* Initialize our structure.  */
+               /* copy the old data to the new array */

        for (i = 0; i < seminfo.semmnu; i++)
        {
                newSemu[i] = semu[i];
        for (i = 0; i < seminfo.semmnu; i++)
        {
                newSemu[i] = semu[i];

-               for(j = 0; j < SEMUME; j++)   /* Is this really needed? */
-                       newSemu[i].un_ent[j] = semu[i].un_ent[j];

        }
        }

-               for (i = seminfo.semmnu; i < newSize; i++)
-        {
-                       newSemu[i].un_proc = NULL;
-        }
+       /*
+        * The new elements (from newSemu[i] to newSemu[newSize-1]) have their
+        * "un_proc" set to 0 (i.e. NULL) by the M_ZERO flag to MALLOC() above,
+        * so they're already marked as "not in use".
+        */

 
        /* Clean up the old array */
        if (semu)
 
        /* Clean up the old array */
        if (semu)


@@ -302,10 +324,9 @@ grow_semu_array(newSize)
  * Assumes we already have the subsystem lock.
  */
 static int
  * Assumes we already have the subsystem lock.
  */
 static int

-grow_sema_array(newSize)
-       int newSize;
+grow_sema_array(int newSize)

 {
 {

-       register struct semid_ds *newSema;
+       register struct semid_kernel *newSema;

        register int i;
 
        if (newSize <= seminfo.semmni)
        register int i;
 
        if (newSize <= seminfo.semmni)


@@ -324,8 +345,9 @@ grow_sema_array(newSize)
 #ifdef SEM_DEBUG
        printf("growing sema[] from %d to %d\n", seminfo.semmni, newSize);
 #endif
 #ifdef SEM_DEBUG
        printf("growing sema[] from %d to %d\n", seminfo.semmni, newSize);
 #endif

-       MALLOC(newSema, struct semid_ds*, sizeof(struct semid_ds)*newSize,
-               M_SYSVSEM, M_WAITOK);
+       MALLOC(newSema, struct semid_kernel *,
+              sizeof (struct semid_kernel) * newSize,
+              M_SYSVSEM, M_WAITOK | M_ZERO);

        if (NULL == newSema)
        {
 #ifdef SEM_DEBUG
        if (NULL == newSema)
        {
 #ifdef SEM_DEBUG


@@ -334,7 +356,7 @@ grow_sema_array(newSize)
                return 0;
        }
 
                return 0;
        }
 

-       /* Initialize our new ids, and copy over the old ones */
+       /* copy over the old ids */

        for (i = 0; i < seminfo.semmni; i++)
        {
                newSema[i] = sema[i];
        for (i = 0; i < seminfo.semmni; i++)
        {
                newSema[i] = sema[i];


@@ -344,17 +366,24 @@ grow_sema_array(newSize)
                 * this with the existing code, so we wake up the
                 * process and let it do a lot of work to determine the
                 * semaphore set is really not available yet, and then
                 * this with the existing code, so we wake up the
                 * process and let it do a lot of work to determine the
                 * semaphore set is really not available yet, and then

-                * sleep on the correct, reallocated semid_ds pointer.
+                * sleep on the correct, reallocated semid_kernel pointer.

                 */
                 */

-               if (sema[i].sem_perm.mode & SEM_ALLOC)
+               if (sema[i].u.sem_perm.mode & SEM_ALLOC)

                        wakeup((caddr_t)&sema[i]);
        }
 
                        wakeup((caddr_t)&sema[i]);
        }
 

+#if CONFIG_MACF

        for (i = seminfo.semmni; i < newSize; i++)
        {
        for (i = seminfo.semmni; i < newSize; i++)
        {

-               newSema[i].sem_base = 0;
-               newSema[i].sem_perm.mode = 0;
+               mac_sysvsem_label_init(&newSema[i]);

        }
        }

+#endif
+       
+       /*
+        * The new elements (from newSema[i] to newSema[newSize-1]) have their
+        * "sem_base" and "sem_perm.mode" set to 0 (i.e. NULL) by the M_ZERO
+        * flag to MALLOC() above, so they're already marked as "not in use".
+        */

 
        /* Clean up the old array */
        if (sema)
 
        /* Clean up the old array */
        if (sema)


@@ -369,38 +398,38 @@ grow_sema_array(newSize)
 }
 
 /*
 }
 
 /*

- * Expand the sem array to the given capacity.  If the expansion fails
+ * Expand the sem_pool array to the given capacity.  If the expansion fails

  * we return 0 (fail), otherwise we return 1 (success).
  *
  * Assumes we already hold the subsystem lock.
  */
 static int
  * we return 0 (fail), otherwise we return 1 (success).
  *
  * Assumes we already hold the subsystem lock.
  */
 static int

-grow_sem_array(newSize)
-               int newSize;
+grow_sem_pool(int new_pool_size)

 {
 {

-       register struct sem *newSem = NULL;
-       register int i;
+       struct sem *new_sem_pool = NULL;
+       struct sem *sem_free;
+       int i;

 
 

-       if (newSize < semtot)
+       if (new_pool_size < semtot)

                return 0;
                return 0;

-       if (newSize > limitseminfo.semmns) /* enforce hard limit */
-       {
+       /* enforce hard limit */
+       if (new_pool_size > limitseminfo.semmns) {

 #ifdef SEM_DEBUG
                printf("semaphore hard limit of %d reached, requested %d\n",
 #ifdef SEM_DEBUG
                printf("semaphore hard limit of %d reached, requested %d\n",

-                       limitseminfo.semmns, newSize);
+                       limitseminfo.semmns, new_pool_size);

 #endif
                return 0;
        }
 #endif
                return 0;
        }

-       newSize = (newSize/SEMMNS_INC + 1) * SEMMNS_INC;
-       newSize = newSize > limitseminfo.semmns ? limitseminfo.semmns : newSize;
+
+       new_pool_size = (new_pool_size/SEMMNS_INC + 1) * SEMMNS_INC;
+       new_pool_size = new_pool_size > limitseminfo.semmns ? limitseminfo.semmns : new_pool_size;

 
 #ifdef SEM_DEBUG
 
 #ifdef SEM_DEBUG

-       printf("growing sem array from %d to %d\n", seminfo.semmns, newSize);
+       printf("growing sem_pool array from %d to %d\n", seminfo.semmns, new_pool_size);

 #endif
 #endif

-       MALLOC(newSem, struct sem*, sizeof(struct sem)*newSize,
-               M_SYSVSEM, M_WAITOK);
-       if (NULL == newSem)
-       {
+       MALLOC(new_sem_pool, struct sem *, sizeof (struct sem) * new_pool_size,
+              M_SYSVSEM, M_WAITOK | M_ZERO);
+       if (NULL == new_sem_pool) {

 #ifdef SEM_DEBUG
                printf("allocation failed.  no changes made.\n");
 #endif
 #ifdef SEM_DEBUG
                printf("allocation failed.  no changes made.\n");
 #endif


@@ -408,26 +437,25 @@ grow_sem_array(newSize)
        }
 
        /* We have our new memory, now copy the old contents over */
        }
 
        /* We have our new memory, now copy the old contents over */

-       if (sem)
+       if (sem_pool)

                for(i = 0; i < seminfo.semmns; i++)
                for(i = 0; i < seminfo.semmns; i++)

-                       newSem[i] = sem[i];
+                       new_sem_pool[i] = sem_pool[i];

 
        /* Update our id structures to point to the new semaphores */
 
        /* Update our id structures to point to the new semaphores */

-       for(i = 0; i < seminfo.semmni; i++)
-               if (sema[i].sem_perm.mode & SEM_ALLOC)  /* ID in use */
-               {
-                       if (newSem > sem)
-                               sema[i].sem_base += newSem - sem;
-                       else
-                               sema[i].sem_base -= sem - newSem;
-               }
+       for(i = 0; i < seminfo.semmni; i++) {
+               if (sema[i].u.sem_perm.mode & SEM_ALLOC)  /* ID in use */
+                       sema[i].u.sem_base = new_sem_pool + 
+                               (sema[i].u.sem_base - sem_pool);
+       }
+
+       sem_free = sem_pool;
+       sem_pool = new_sem_pool;

 
        /* clean up the old array */
 
        /* clean up the old array */

-       if (sem)
-               FREE(sem, M_SYSVSEM);
+       if (sem_free != NULL)
+               FREE(sem_free, M_SYSVSEM);

 
 

-       sem = newSem;
-       seminfo.semmns = newSize;
+       seminfo.semmns = new_pool_size;

 #ifdef SEM_DEBUG
        printf("expansion complete\n");
 #endif
 #ifdef SEM_DEBUG
        printf("expansion complete\n");
 #endif


@@ -441,13 +469,12 @@ grow_sem_array(newSize)
  * Assumes we already hold the subsystem lock.
  */
 
  * Assumes we already hold the subsystem lock.
  */
 

-static struct sem_undo *
-semu_alloc(p)
-       struct proc *p;
+static int
+semu_alloc(struct proc *p)

 {
        register int i;
        register struct sem_undo *suptr;
 {
        register int i;
        register struct sem_undo *suptr;

-       register struct sem_undo **supptr;
+       int *supidx;

        int attempt;
 
        /*
        int attempt;
 
        /*


@@ -465,11 +492,12 @@ semu_alloc(p)
                for (i = 0; i < seminfo.semmnu; i++) {
                        suptr = SEMU(i);
                        if (suptr->un_proc == NULL) {
                for (i = 0; i < seminfo.semmnu; i++) {
                        suptr = SEMU(i);
                        if (suptr->un_proc == NULL) {

-                               suptr->un_next = semu_list;
-                               semu_list = suptr;
+                               suptr->un_next_idx = semu_list_idx;
+                               semu_list_idx = i;

                                suptr->un_cnt = 0;
                                suptr->un_cnt = 0;

+                               suptr->un_ent = NULL;

                                suptr->un_proc = p;
                                suptr->un_proc = p;

-                               return(suptr);
+                               return i;

                        }
                }
 
                        }
                }
 


@@ -482,14 +510,15 @@ semu_alloc(p)
                        /* All the structures are in use - try to free some */
                        int did_something = 0;
 
                        /* All the structures are in use - try to free some */
                        int did_something = 0;
 

-                       supptr = &semu_list;
-                       while ((suptr = *supptr) != NULL) {
+                       supidx = &semu_list_idx;
+                       while (*supidx != -1) {
+                               suptr = SEMU(*supidx);

                                if (suptr->un_cnt == 0)  {
                                        suptr->un_proc = NULL;
                                if (suptr->un_cnt == 0)  {
                                        suptr->un_proc = NULL;

-                                       *supptr = suptr->un_next;
+                                       *supidx = suptr->un_next_idx;

                                        did_something = 1;
                                } else
                                        did_something = 1;
                                } else

-                                       supptr = &(suptr->un_next);
+                                       supidx = &(suptr->un_next_idx);

                        }
 
                        /* If we didn't free anything. Try expanding
                        }
 
                        /* If we didn't free anything. Try expanding


@@ -499,7 +528,7 @@ semu_alloc(p)
                         */
                        if (!did_something)
                                if (!grow_semu_array(seminfo.semmnu + 1))
                         */
                        if (!did_something)
                                if (!grow_semu_array(seminfo.semmnu + 1))

-                                       return(NULL);
+                                       return -1;

                } else {
                        /*
                         * The second pass failed even though we freed
                } else {
                        /*
                         * The second pass failed even though we freed


@@ -509,7 +538,7 @@ semu_alloc(p)
                        panic("semu_alloc - second attempt failed");
                }
        }
                        panic("semu_alloc - second attempt failed");
                }
        }

-       return (NULL);
+       return -1;

 }
 
 /*
 }
 
 /*


@@ -517,37 +546,36 @@ semu_alloc(p)
  *
  * Assumes we already hold the subsystem lock.
  */
  *
  * Assumes we already hold the subsystem lock.
  */

-

 static int
 static int

-semundo_adjust(p, supptr, semid, semnum, adjval)
-       register struct proc *p;
-       struct sem_undo **supptr;
-       int semid, semnum;
-       int adjval;
+semundo_adjust(struct proc *p, int *supidx, int semid,
+       int semnum, int adjval)

 {
        register struct sem_undo *suptr;
 {
        register struct sem_undo *suptr;

-       register struct undo *sunptr;
+       int suidx;
+       register struct undo *sueptr, **suepptr, *new_sueptr;

        int i;
 
        int i;
 

-       /* Look for and remember the sem_undo if the caller doesn't provide
-          it */
+       /*
+        * Look for and remember the sem_undo if the caller doesn't provide it
+        */

 
 

-       suptr = *supptr;
-       if (suptr == NULL) {
-               for (suptr = semu_list; suptr != NULL;
-                   suptr = suptr->un_next) {
+       suidx = *supidx;
+       if (suidx == -1) {
+               for (suidx = semu_list_idx; suidx != -1;
+                   suidx = suptr->un_next_idx) {
+                       suptr = SEMU(suidx);

                        if (suptr->un_proc == p) {
                        if (suptr->un_proc == p) {

-                               *supptr = suptr;
+                               *supidx = suidx;

                                break;
                        }
                }
                                break;
                        }
                }

-               if (suptr == NULL) {
+               if (suidx == -1) {

                        if (adjval == 0)
                                return(0);
                        if (adjval == 0)
                                return(0);

-                       suptr = semu_alloc(p);
-                       if (suptr == NULL)
+                       suidx = semu_alloc(p);
+                       if (suidx == -1)

                                return(ENOSPC);
                                return(ENOSPC);

-                       *supptr = suptr;
+                       *supidx = suidx;

                }
        }
 
                }
        }
 


@@ -555,311 +583,388 @@ semundo_adjust(p, supptr, semid, semnum, adjval)
         * Look for the requested entry and adjust it (delete if adjval becomes
         * 0).
         */
         * Look for the requested entry and adjust it (delete if adjval becomes
         * 0).
         */

-       sunptr = &suptr->un_ent[0];
-       for (i = 0; i < suptr->un_cnt; i++, sunptr++) {
-               if (sunptr->un_id != semid || sunptr->un_num != semnum)
+       suptr = SEMU(suidx);
+       new_sueptr = NULL;
+       for (i = 0, suepptr = &suptr->un_ent, sueptr = suptr->un_ent;
+            i < suptr->un_cnt;
+            i++, suepptr = &sueptr->une_next, sueptr = sueptr->une_next) {
+               if (sueptr->une_id != semid || sueptr->une_num != semnum)

                        continue;
                if (adjval == 0)
                        continue;
                if (adjval == 0)

-                       sunptr->un_adjval = 0;
+                       sueptr->une_adjval = 0;

                else
                else

-                       sunptr->un_adjval += adjval;
-               if (sunptr->un_adjval == 0) {
+                       sueptr->une_adjval += adjval;
+               if (sueptr->une_adjval == 0) {

                        suptr->un_cnt--;
                        suptr->un_cnt--;

-                       if (i < suptr->un_cnt)
-                               suptr->un_ent[i] =
-                                   suptr->un_ent[suptr->un_cnt];
+                       *suepptr = sueptr->une_next;
+                       FREE(sueptr, M_SYSVSEM);
+                       sueptr = NULL;

                }
                }

-               return(0);
+               return 0;

        }
 
        /* Didn't find the right entry - create it */
        }
 
        /* Didn't find the right entry - create it */

-       if (adjval == 0)
-               return(0);
-       if (suptr->un_cnt != limitseminfo.semume) {
-               sunptr = &suptr->un_ent[suptr->un_cnt];
-               suptr->un_cnt++;
-               sunptr->un_adjval = adjval;
-               sunptr->un_id = semid; sunptr->un_num = semnum;
-       } else
-               return(EINVAL);
-       return(0);
+       if (adjval == 0) {
+               /* no adjustment: no need for a new entry */
+               return 0;
+       }
+
+       if (suptr->un_cnt == limitseminfo.semume) {
+               /* reached the limit number of semaphore undo entries */
+               return EINVAL;
+       }
+
+       /* allocate a new semaphore undo entry */
+       MALLOC(new_sueptr, struct undo *, sizeof (struct undo),
+              M_SYSVSEM, M_WAITOK);
+       if (new_sueptr == NULL) {
+               return ENOMEM;
+       }
+
+       /* fill in the new semaphore undo entry */
+       new_sueptr->une_next = suptr->un_ent;
+       suptr->un_ent = new_sueptr;
+       suptr->un_cnt++;
+       new_sueptr->une_adjval = adjval;
+       new_sueptr->une_id = semid;
+       new_sueptr->une_num = semnum;
+
+       return 0;

 }
 
 /* Assumes we already hold the subsystem lock.
  */
 static void
 }
 
 /* Assumes we already hold the subsystem lock.
  */
 static void

-semundo_clear(semid, semnum)
-       int semid, semnum;
+semundo_clear(int semid, int semnum)

 {
 {

-       register struct sem_undo *suptr;
+       struct sem_undo *suptr;
+       int suidx;

 
 

-       for (suptr = semu_list; suptr != NULL; suptr = suptr->un_next) {
-               register struct undo *sunptr = &suptr->un_ent[0];
-               register int i = 0;
+       for (suidx = semu_list_idx; suidx != -1; suidx = suptr->un_next_idx) {
+               struct undo *sueptr;
+               struct undo **suepptr;
+               int i = 0;

 
 

+               suptr = SEMU(suidx);
+               sueptr = suptr->un_ent;
+               suepptr = &suptr->un_ent;

                while (i < suptr->un_cnt) {
                while (i < suptr->un_cnt) {

-                       if (sunptr->un_id == semid) {
-                               if (semnum == -1 || sunptr->un_num == semnum) {
+                       if (sueptr->une_id == semid) {
+                               if (semnum == -1 || sueptr->une_num == semnum) {

                                        suptr->un_cnt--;
                                        suptr->un_cnt--;

-                                       if (i < suptr->un_cnt) {
-                                               suptr->un_ent[i] =
-                                                 suptr->un_ent[suptr->un_cnt];
-                                               continue;
-                                       }
+                                       *suepptr = sueptr->une_next;
+                                       FREE(sueptr, M_SYSVSEM);
+                                       sueptr = *suepptr;
+                                       continue;

                                }
                                if (semnum != -1)
                                        break;
                        }
                                }
                                if (semnum != -1)
                                        break;
                        }

-                       i++, sunptr++;
+                       i++;
+                       suepptr = &sueptr->une_next;
+                       sueptr = sueptr->une_next;

                }
        }
 }
 
 /*
                }
        }
 }
 
 /*

- * Note that the user-mode half of this passes a union, not a pointer
+ * Note that the user-mode half of this passes a union coerced to a
+ * user_addr_t.  The union contains either an int or a pointer, and
+ * so we have to coerce it back, variant on whether the calling
+ * process is 64 bit or not.  The coercion works for the 'val' element
+ * because the alignment is the same in user and kernel space.

  */
  */

-#ifndef _SYS_SYSPROTO_H_
-struct semctl_args {
-       int     semid;
-       int     semnum;
-       int     cmd;
-       union   semun arg;
-};
-#endif
-

 int
 int

-semctl(p, uap, retval)
-       struct proc *p;
-       register struct semctl_args *uap;
-       register_t *retval;
+semctl(struct proc *p, struct semctl_args *uap, int32_t *retval)

 {
        int semid = uap->semid;
        int semnum = uap->semnum;
        int cmd = uap->cmd;
 {
        int semid = uap->semid;
        int semnum = uap->semnum;
        int cmd = uap->cmd;

-       union semun arg = uap->arg;
-       union semun real_arg;
-       struct ucred *cred = p->p_ucred;
+       user_semun_t user_arg = (user_semun_t)uap->arg;
+       kauth_cred_t cred = kauth_cred_get();

        int i, rval, eval;
        int i, rval, eval;

-       struct semid_ds sbuf;
-       register struct semid_ds *semaptr;
+       struct user_semid_ds sbuf;
+       struct semid_kernel *semakptr;
+       

 
        AUDIT_ARG(svipc_cmd, cmd);
        AUDIT_ARG(svipc_id, semid);
 
        AUDIT_ARG(svipc_cmd, cmd);
        AUDIT_ARG(svipc_id, semid);

-       SUBSYSTEM_LOCK_AQUIRE(p);
+
+       SYSV_SEM_SUBSYS_LOCK();
+

 #ifdef SEM_DEBUG
 #ifdef SEM_DEBUG

-       printf("call to semctl(%d, %d, %d, 0x%x)\n", semid, semnum, cmd, arg);
+       printf("call to semctl(%d, %d, %d, 0x%qx)\n", semid, semnum, cmd, user_arg);

 #endif
 
        semid = IPCID_TO_IX(semid);
 #endif
 
        semid = IPCID_TO_IX(semid);

-       if (semid < 0 || semid >= seminfo.semmni)
-{
+
+       if (semid < 0 || semid >= seminfo.semmni) {

 #ifdef SEM_DEBUG
                printf("Invalid semid\n");
 #endif
 #ifdef SEM_DEBUG
                printf("Invalid semid\n");
 #endif

-               UNLOCK_AND_RETURN(EINVAL);
-}
+               eval = EINVAL;
+               goto semctlout;
+       }

 
 

-       semaptr = &sema[semid];
-       if ((semaptr->sem_perm.mode & SEM_ALLOC) == 0 ||
-           semaptr->sem_perm.seq != IPCID_TO_SEQ(uap->semid))
-               UNLOCK_AND_RETURN(EINVAL);
+       semakptr = &sema[semid];
+       if ((semakptr->u.sem_perm.mode & SEM_ALLOC) == 0 ||
+           semakptr->u.sem_perm._seq != IPCID_TO_SEQ(uap->semid)) {
+               eval = EINVAL;
+               goto semctlout;
+       }
+#if CONFIG_MACF
+       eval = mac_sysvsem_check_semctl(cred, semakptr, cmd);
+       if (eval)
+               goto semctlout;
+#endif

 
        eval = 0;
        rval = 0;
 
        switch (cmd) {
        case IPC_RMID:
 
        eval = 0;
        rval = 0;
 
        switch (cmd) {
        case IPC_RMID:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_M)))
-                       UNLOCK_AND_RETURN(eval);
-               semaptr->sem_perm.cuid = cred->cr_uid;
-               semaptr->sem_perm.uid = cred->cr_uid;
-               semtot -= semaptr->sem_nsems;
-               for (i = semaptr->sem_base - sem; i < semtot; i++)
-                       sem[i] = sem[i + semaptr->sem_nsems];
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_M))) 
+                       goto semctlout;
+
+               semakptr->u.sem_perm.cuid = kauth_cred_getuid(cred);
+               semakptr->u.sem_perm.uid = kauth_cred_getuid(cred);
+               semtot -= semakptr->u.sem_nsems;
+               for (i = semakptr->u.sem_base - sem_pool; i < semtot; i++)
+                       sem_pool[i] = sem_pool[i + semakptr->u.sem_nsems];

                for (i = 0; i < seminfo.semmni; i++) {
                for (i = 0; i < seminfo.semmni; i++) {

-                       if ((sema[i].sem_perm.mode & SEM_ALLOC) &&
-                           sema[i].sem_base > semaptr->sem_base)
-                               sema[i].sem_base -= semaptr->sem_nsems;
+                       if ((sema[i].u.sem_perm.mode & SEM_ALLOC) &&
+                           sema[i].u.sem_base > semakptr->u.sem_base)
+                               sema[i].u.sem_base -= semakptr->u.sem_nsems;

                }
                }

-               semaptr->sem_perm.mode = 0;
+               semakptr->u.sem_perm.mode = 0;
+#if CONFIG_MACF
+               mac_sysvsem_label_recycle(semakptr);
+#endif

                semundo_clear(semid, -1);
                semundo_clear(semid, -1);

-               wakeup((caddr_t)semaptr);
+               wakeup((caddr_t)semakptr);

                break;
 
        case IPC_SET:
                break;
 
        case IPC_SET:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_M)))
-                       UNLOCK_AND_RETURN(eval);
-               /*if ((eval = copyin(arg, &real_arg, sizeof(real_arg))) != 0)
-                       UNLOCK_AND_RETURN(eval);*/
-               if ((eval = copyin(arg.buf, (caddr_t)&sbuf,
-                   sizeof(sbuf))) != 0)
-                       UNLOCK_AND_RETURN(eval);
-               semaptr->sem_perm.uid = sbuf.sem_perm.uid;
-               semaptr->sem_perm.gid = sbuf.sem_perm.gid;
-               semaptr->sem_perm.mode = (semaptr->sem_perm.mode & ~0777) |
-                   (sbuf.sem_perm.mode & 0777);
-               semaptr->sem_ctime = time_second;
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_M)))
+                               goto semctlout;
+
+               if (IS_64BIT_PROCESS(p)) {
+                       struct user64_semid_ds ds64;
+                       eval = copyin(user_arg.buf, &ds64, sizeof(ds64));
+                       semid_ds_64tokernel(&ds64, &sbuf);
+               } else {
+                       struct user32_semid_ds ds32;
+                       eval = copyin(user_arg.buf, &ds32, sizeof(ds32));
+                       semid_ds_32tokernel(&ds32, &sbuf);
+               }
+               
+               if (eval != 0) {
+                       goto semctlout;
+               }
+
+               semakptr->u.sem_perm.uid = sbuf.sem_perm.uid;
+               semakptr->u.sem_perm.gid = sbuf.sem_perm.gid;
+               semakptr->u.sem_perm.mode = (semakptr->u.sem_perm.mode &
+                   ~0777) | (sbuf.sem_perm.mode & 0777);
+               semakptr->u.sem_ctime = sysv_semtime();

                break;
 
        case IPC_STAT:
                break;
 
        case IPC_STAT:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_R)))
-                       UNLOCK_AND_RETURN(eval);
-               /*if ((eval = copyin(arg, &real_arg, sizeof(real_arg))) != 0)
-                       UNLOCK_AND_RETURN(eval);*/
-               eval = copyout((caddr_t)semaptr, arg.buf,
-                   sizeof(struct semid_ds));
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_R)))
+                               goto semctlout;
+
+               if (IS_64BIT_PROCESS(p)) {
+                       struct user64_semid_ds semid_ds64;
+                       bzero(&semid_ds64, sizeof(semid_ds64));
+                       semid_ds_kernelto64(&semakptr->u, &semid_ds64);
+                       eval = copyout(&semid_ds64, user_arg.buf, sizeof(semid_ds64));
+               } else {
+                       struct user32_semid_ds semid_ds32;
+                       bzero(&semid_ds32, sizeof(semid_ds32));
+                       semid_ds_kernelto32(&semakptr->u, &semid_ds32);
+                       eval = copyout(&semid_ds32, user_arg.buf, sizeof(semid_ds32));
+               }

                break;
 
        case GETNCNT:
                break;
 
        case GETNCNT:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_R)))
-                       UNLOCK_AND_RETURN(eval);
-               if (semnum < 0 || semnum >= semaptr->sem_nsems)
-                       UNLOCK_AND_RETURN(EINVAL);
-               rval = semaptr->sem_base[semnum].semncnt;
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_R)))
+                               goto semctlout;
+               if (semnum < 0 || semnum >= semakptr->u.sem_nsems) {
+                       eval = EINVAL;
+                       goto semctlout;
+               }
+               rval = semakptr->u.sem_base[semnum].semncnt;

                break;
 
        case GETPID:
                break;
 
        case GETPID:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_R)))
-                       UNLOCK_AND_RETURN(eval);
-               if (semnum < 0 || semnum >= semaptr->sem_nsems)
-                       UNLOCK_AND_RETURN(EINVAL);
-               rval = semaptr->sem_base[semnum].sempid;
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_R)))
+                               goto semctlout;
+               if (semnum < 0 || semnum >= semakptr->u.sem_nsems) {
+                       eval = EINVAL;
+                       goto semctlout;
+               }
+               rval = semakptr->u.sem_base[semnum].sempid;

                break;
 
        case GETVAL:
                break;
 
        case GETVAL:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_R)))
-                       UNLOCK_AND_RETURN(eval);
-               if (semnum < 0 || semnum >= semaptr->sem_nsems)
-                       UNLOCK_AND_RETURN(EINVAL);
-               rval = semaptr->sem_base[semnum].semval;
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_R)))
+                               goto semctlout;
+               if (semnum < 0 || semnum >= semakptr->u.sem_nsems) {
+                       eval = EINVAL;
+                       goto semctlout;
+               }
+               rval = semakptr->u.sem_base[semnum].semval;

                break;
 
        case GETALL:
                break;
 
        case GETALL:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_R)))
-                       UNLOCK_AND_RETURN(eval);
-               /*if ((eval = copyin(arg, &real_arg, sizeof(real_arg))) != 0)
-                       UNLOCK_AND_RETURN(eval);*/
-               for (i = 0; i < semaptr->sem_nsems; i++) {
-                       eval = copyout((caddr_t)&semaptr->sem_base[i].semval,
-                           &arg.array[i], sizeof(arg.array[0]));
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_R)))
+                               goto semctlout;
+/* XXXXXXXXXXXXXXXX TBD XXXXXXXXXXXXXXXX */
+               for (i = 0; i < semakptr->u.sem_nsems; i++) {
+                       /* XXX could be done in one go... */
+                       eval = copyout((caddr_t)&semakptr->u.sem_base[i].semval,
+                           user_arg.array + (i * sizeof(unsigned short)),
+                           sizeof(unsigned short));

                        if (eval != 0)
                                break;
                }
                break;
 
        case GETZCNT:
                        if (eval != 0)
                                break;
                }
                break;
 
        case GETZCNT:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_R)))
-                       UNLOCK_AND_RETURN(eval);
-               if (semnum < 0 || semnum >= semaptr->sem_nsems)
-                       UNLOCK_AND_RETURN(EINVAL);
-               rval = semaptr->sem_base[semnum].semzcnt;
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_R)))
+                               goto semctlout;
+               if (semnum < 0 || semnum >= semakptr->u.sem_nsems) {
+                       eval = EINVAL;
+                       goto semctlout;
+               }
+               rval = semakptr->u.sem_base[semnum].semzcnt;

                break;
 
        case SETVAL:
                break;
 
        case SETVAL:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_W)))
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_W)))

                 {
 #ifdef SEM_DEBUG
                        printf("Invalid credentials for write\n");
 #endif
                 {
 #ifdef SEM_DEBUG
                        printf("Invalid credentials for write\n");
 #endif

-                       UNLOCK_AND_RETURN(eval);
+                               goto semctlout;

                }
                }

-               if (semnum < 0 || semnum >= semaptr->sem_nsems)
+               if (semnum < 0 || semnum >= semakptr->u.sem_nsems)

                {
 #ifdef SEM_DEBUG
                        printf("Invalid number out of range for set\n");
 #endif
                {
 #ifdef SEM_DEBUG
                        printf("Invalid number out of range for set\n");
 #endif

-                       UNLOCK_AND_RETURN(EINVAL);
+                       eval = EINVAL;
+                       goto semctlout;

                }
                }

-               /*if ((eval = copyin(arg, &real_arg, sizeof(real_arg))) != 0)
+               
+               /*
+                * Cast down a pointer instead of using 'val' member directly
+                * to avoid introducing endieness and a pad field into the
+                * header file.  Ugly, but it works.
+                */
+               u_int newsemval = CAST_DOWN_EXPLICIT(u_int, user_arg.buf);
+               
+               /*
+                * The check is being performed as unsigned values to match 
+                * eventual destination
+                */ 
+               if (newsemval > (u_int)seminfo.semvmx)

                {
 #ifdef SEM_DEBUG
                {
 #ifdef SEM_DEBUG

-                       printf("Error during value copyin\n");
+                       printf("Out of range sem value for set\n");

 #endif
 #endif

-                       UNLOCK_AND_RETURN(eval);
-               }*/
-               semaptr->sem_base[semnum].semval = arg.val;
+                       eval = ERANGE;
+                       goto semctlout;
+               }
+               semakptr->u.sem_base[semnum].semval = newsemval;
+               semakptr->u.sem_base[semnum].sempid = p->p_pid;
+               /* XXX scottl Should there be a MAC call here? */

                semundo_clear(semid, semnum);
                semundo_clear(semid, semnum);

-               wakeup((caddr_t)semaptr);
+               wakeup((caddr_t)semakptr);

                break;
 
        case SETALL:
                break;
 
        case SETALL:

-               if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_W)))
-                       UNLOCK_AND_RETURN(eval);
-               /*if ((eval = copyin(arg, &real_arg, sizeof(real_arg))) != 0)
-                       UNLOCK_AND_RETURN(eval);*/
-               for (i = 0; i < semaptr->sem_nsems; i++) {
-                       eval = copyin(&arg.array[i],
-                           (caddr_t)&semaptr->sem_base[i].semval,
-                           sizeof(arg.array[0]));
+               if ((eval = ipcperm(cred, &semakptr->u.sem_perm, IPC_W)))
+                               goto semctlout;
+/*** XXXXXXXXXXXX TBD ********/
+               for (i = 0; i < semakptr->u.sem_nsems; i++) {
+                       /* XXX could be done in one go... */
+                       eval = copyin(user_arg.array + (i * sizeof(unsigned short)),
+                           (caddr_t)&semakptr->u.sem_base[i].semval,
+                           sizeof(unsigned short));

                        if (eval != 0)
                                break;
                        if (eval != 0)
                                break;

+                       semakptr->u.sem_base[i].sempid = p->p_pid;

                }
                }

+               /* XXX scottl Should there be a MAC call here? */

                semundo_clear(semid, -1);
                semundo_clear(semid, -1);

-               wakeup((caddr_t)semaptr);
+               wakeup((caddr_t)semakptr);

                break;
 
        default:
                break;
 
        default:

-               UNLOCK_AND_RETURN(EINVAL);
+                       eval = EINVAL;
+                       goto semctlout;

        }
 
        if (eval == 0)
                *retval = rval;
        }
 
        if (eval == 0)
                *retval = rval;

-       UNLOCK_AND_RETURN(eval);
+semctlout:
+       SYSV_SEM_SUBSYS_UNLOCK();
+       return(eval);

 }
 
 }
 

-#ifndef _SYS_SYSPROTO_H_
-struct semget_args {
-       key_t   key;
-       int     nsems;
-       int     semflg;
-};
-#endif
-

 int
 int

-semget(p, uap, retval)
-       struct proc *p;
-       register struct semget_args *uap;
-       register_t *retval;
+semget(__unused struct proc *p, struct semget_args *uap, int32_t *retval)

 {
        int semid, eval;
        int key = uap->key;
        int nsems = uap->nsems;
        int semflg = uap->semflg;
 {
        int semid, eval;
        int key = uap->key;
        int nsems = uap->nsems;
        int semflg = uap->semflg;

-       struct ucred *cred = p->p_ucred;
+       kauth_cred_t cred = kauth_cred_get();

 
 

-       SUBSYSTEM_LOCK_AQUIRE(p);

 #ifdef SEM_DEBUG
        if (key != IPC_PRIVATE)
                printf("semget(0x%x, %d, 0%o)\n", key, nsems, semflg);
        else
                printf("semget(IPC_PRIVATE, %d, 0%o)\n", nsems, semflg);
 #endif
 #ifdef SEM_DEBUG
        if (key != IPC_PRIVATE)
                printf("semget(0x%x, %d, 0%o)\n", key, nsems, semflg);
        else
                printf("semget(IPC_PRIVATE, %d, 0%o)\n", nsems, semflg);
 #endif

+
+
+       SYSV_SEM_SUBSYS_LOCK();
+

     
        if (key != IPC_PRIVATE) {
                for (semid = 0; semid < seminfo.semmni; semid++) {
     
        if (key != IPC_PRIVATE) {
                for (semid = 0; semid < seminfo.semmni; semid++) {

-                       if ((sema[semid].sem_perm.mode & SEM_ALLOC) &&
-                           sema[semid].sem_perm.key == key)
+                       if ((sema[semid].u.sem_perm.mode & SEM_ALLOC) &&
+                           sema[semid].u.sem_perm._key == key)

                                break;
                }
                if (semid < seminfo.semmni) {
 #ifdef SEM_DEBUG
                        printf("found public key\n");
 #endif
                                break;
                }
                if (semid < seminfo.semmni) {
 #ifdef SEM_DEBUG
                        printf("found public key\n");
 #endif

-                       if ((eval = ipcperm(cred, &sema[semid].sem_perm,
+                       if ((eval = ipcperm(cred, &sema[semid].u.sem_perm,

                            semflg & 0700)))
                            semflg & 0700)))

-                               UNLOCK_AND_RETURN(eval);
-                       if (nsems > 0 && sema[semid].sem_nsems < nsems) {
+                               goto semgetout;
+                       if (nsems < 0 || sema[semid].u.sem_nsems < nsems) {

 #ifdef SEM_DEBUG
                                printf("too small\n");
 #endif
 #ifdef SEM_DEBUG
                                printf("too small\n");
 #endif

-                               UNLOCK_AND_RETURN(EINVAL);
+                               eval = EINVAL;
+                               goto semgetout;

                        }
                        if ((semflg & IPC_CREAT) && (semflg & IPC_EXCL)) {
 #ifdef SEM_DEBUG
                                printf("not exclusive\n");
 #endif
                        }
                        if ((semflg & IPC_CREAT) && (semflg & IPC_EXCL)) {
 #ifdef SEM_DEBUG
                                printf("not exclusive\n");
 #endif

-                               UNLOCK_AND_RETURN(EEXIST);
+                               eval = EEXIST;
+                               goto semgetout;

                        }
                        }

+#if CONFIG_MACF
+                       eval = mac_sysvsem_check_semget(cred, &sema[semid]);
+                       if (eval) 
+                               goto semgetout;
+#endif

                        goto found;
                }
        }
                        goto found;
                }
        }


@@ -873,23 +978,24 @@ semget(p, uap, retval)
                        printf("nsems out of range (0<%d<=%d)\n", nsems,
                            seminfo.semmsl);
 #endif
                        printf("nsems out of range (0<%d<=%d)\n", nsems,
                            seminfo.semmsl);
 #endif

-                       UNLOCK_AND_RETURN(EINVAL);
+                       eval = EINVAL;
+                       goto semgetout;

                }
                if (nsems > seminfo.semmns - semtot) {
 #ifdef SEM_DEBUG
                        printf("not enough semaphores left (need %d, got %d)\n",
                            nsems, seminfo.semmns - semtot);
 #endif
                }
                if (nsems > seminfo.semmns - semtot) {
 #ifdef SEM_DEBUG
                        printf("not enough semaphores left (need %d, got %d)\n",
                            nsems, seminfo.semmns - semtot);
 #endif

-                       if (!grow_sem_array(semtot + nsems))
-                       {
+                       if (!grow_sem_pool(semtot + nsems)) {

 #ifdef SEM_DEBUG
                                printf("failed to grow the sem array\n");
 #endif
 #ifdef SEM_DEBUG
                                printf("failed to grow the sem array\n");
 #endif

-                               UNLOCK_AND_RETURN(ENOSPC);
+                               eval = ENOSPC;
+                               goto semgetout;

                        }
                }
                for (semid = 0; semid < seminfo.semmni; semid++) {
                        }
                }
                for (semid = 0; semid < seminfo.semmni; semid++) {

-                       if ((sema[semid].sem_perm.mode & SEM_ALLOC) == 0)
+                       if ((sema[semid].u.sem_perm.mode & SEM_ALLOC) == 0)

                                break;
                }
                if (semid == seminfo.semmni) {
                                break;
                }
                if (semid == seminfo.semmni) {


@@ -901,112 +1007,137 @@ semget(p, uap, retval)
 #ifdef SEM_DEBUG
                                printf("failed to grow sema array\n");
 #endif
 #ifdef SEM_DEBUG
                                printf("failed to grow sema array\n");
 #endif

-                               UNLOCK_AND_RETURN(ENOSPC);
+                               eval = ENOSPC;
+                               goto semgetout;

                        }
                }
 #ifdef SEM_DEBUG
                printf("semid %d is available\n", semid);
 #endif
                        }
                }
 #ifdef SEM_DEBUG
                printf("semid %d is available\n", semid);
 #endif

-               sema[semid].sem_perm.key = key;
-               sema[semid].sem_perm.cuid = cred->cr_uid;
-               sema[semid].sem_perm.uid = cred->cr_uid;
-               sema[semid].sem_perm.cgid = cred->cr_gid;
-               sema[semid].sem_perm.gid = cred->cr_gid;
-               sema[semid].sem_perm.mode = (semflg & 0777) | SEM_ALLOC;
-               sema[semid].sem_perm.seq =
-                   (sema[semid].sem_perm.seq + 1) & 0x7fff;
-               sema[semid].sem_nsems = nsems;
-               sema[semid].sem_otime = 0;
-               sema[semid].sem_ctime = time_second;
-               sema[semid].sem_base = &sem[semtot];
+               sema[semid].u.sem_perm._key = key;
+               sema[semid].u.sem_perm.cuid = kauth_cred_getuid(cred);
+               sema[semid].u.sem_perm.uid = kauth_cred_getuid(cred);
+               sema[semid].u.sem_perm.cgid = kauth_cred_getgid(cred);
+               sema[semid].u.sem_perm.gid = kauth_cred_getgid(cred);
+               sema[semid].u.sem_perm.mode = (semflg & 0777) | SEM_ALLOC;
+               sema[semid].u.sem_perm._seq =
+                   (sema[semid].u.sem_perm._seq + 1) & 0x7fff;
+               sema[semid].u.sem_nsems = nsems;
+               sema[semid].u.sem_otime = 0;
+               sema[semid].u.sem_ctime = sysv_semtime();
+               sema[semid].u.sem_base = &sem_pool[semtot];

                semtot += nsems;
                semtot += nsems;

-               bzero(sema[semid].sem_base,
-                   sizeof(sema[semid].sem_base[0])*nsems);
+               bzero(sema[semid].u.sem_base,
+                   sizeof(sema[semid].u.sem_base[0])*nsems);
+#if CONFIG_MACF
+               mac_sysvsem_label_associate(cred, &sema[semid]);
+#endif

 #ifdef SEM_DEBUG
 #ifdef SEM_DEBUG

-               printf("sembase = 0x%x, next = 0x%x\n", sema[semid].sem_base,
-                   &sem[semtot]);
+               printf("sembase = 0x%x, next = 0x%x\n", sema[semid].u.sem_base,
+                   &sem_pool[semtot]);

 #endif
        } else {
 #ifdef SEM_DEBUG
                printf("didn't find it and wasn't asked to create it\n");
 #endif
 #endif
        } else {
 #ifdef SEM_DEBUG
                printf("didn't find it and wasn't asked to create it\n");
 #endif

-               UNLOCK_AND_RETURN(ENOENT);
+               eval = ENOENT;
+               goto semgetout;

        }
 
 found:
        }
 
 found:

-       *retval = IXSEQ_TO_IPCID(semid, sema[semid].sem_perm);
+       *retval = IXSEQ_TO_IPCID(semid, sema[semid].u.sem_perm);

        AUDIT_ARG(svipc_id, *retval);
 #ifdef SEM_DEBUG
        printf("semget is done, returning %d\n", *retval);
 #endif
        AUDIT_ARG(svipc_id, *retval);
 #ifdef SEM_DEBUG
        printf("semget is done, returning %d\n", *retval);
 #endif

-       SUBSYSTEM_LOCK_RELEASE;
-       return(0);
-}
+       eval = 0;

 
 

-#ifndef _SYS_SYSPROTO_H_
-struct semop_args {
-       int     semid;
-       struct  sembuf *sops;
-       int     nsops;
-};
-#endif
+semgetout:
+       SYSV_SEM_SUBSYS_UNLOCK();
+       return(eval);
+}

 
 int
 
 int

-semop(p, uap, retval)
-       struct proc *p;
-       register struct semop_args *uap;
-       register_t *retval;
+semop(struct proc *p, struct semop_args *uap, int32_t *retval)

 {
        int semid = uap->semid;
        int nsops = uap->nsops;
 {
        int semid = uap->semid;
        int nsops = uap->nsops;

-       struct sembuf sops[MAX_SOPS];
-       register struct semid_ds *semaptr;
-       register struct sembuf *sopptr;
-       register struct sem *semptr;
-       struct sem_undo *suptr = NULL;
-       struct ucred *cred = p->p_ucred;
+       struct sembuf sops[seminfo.semopm];
+       register struct semid_kernel *semakptr;
+       register struct sembuf *sopptr = NULL;  /* protected by 'semptr' */
+       register struct sem *semptr = NULL;     /* protected by 'if' */
+       int supidx = -1;

        int i, j, eval;
        int do_wakeup, do_undos;
 
        AUDIT_ARG(svipc_id, uap->semid);
        int i, j, eval;
        int do_wakeup, do_undos;
 
        AUDIT_ARG(svipc_id, uap->semid);

-       SUBSYSTEM_LOCK_AQUIRE(p);
+
+       SYSV_SEM_SUBSYS_LOCK();
+

 #ifdef SEM_DEBUG
        printf("call to semop(%d, 0x%x, %d)\n", semid, sops, nsops);
 #endif
 
        semid = IPCID_TO_IX(semid);     /* Convert back to zero origin */
 
 #ifdef SEM_DEBUG
        printf("call to semop(%d, 0x%x, %d)\n", semid, sops, nsops);
 #endif
 
        semid = IPCID_TO_IX(semid);     /* Convert back to zero origin */
 

-       if (semid < 0 || semid >= seminfo.semmni)
-               UNLOCK_AND_RETURN(EINVAL);
+       if (semid < 0 || semid >= seminfo.semmni) {
+               eval = EINVAL;
+               goto semopout;
+       }

 
 

-       semaptr = &sema[semid];
-       if ((semaptr->sem_perm.mode & SEM_ALLOC) == 0)
-               UNLOCK_AND_RETURN(EINVAL);
-       if (semaptr->sem_perm.seq != IPCID_TO_SEQ(uap->semid))
-               UNLOCK_AND_RETURN(EINVAL);
+       semakptr = &sema[semid];
+       if ((semakptr->u.sem_perm.mode & SEM_ALLOC) == 0) {
+               eval = EINVAL;
+               goto semopout;
+       }
+       if (semakptr->u.sem_perm._seq != IPCID_TO_SEQ(uap->semid)) {
+               eval = EINVAL;
+               goto semopout;
+       }

 
 

-       if ((eval = ipcperm(cred, &semaptr->sem_perm, IPC_W))) {
+       if ((eval = ipcperm(kauth_cred_get(), &semakptr->u.sem_perm, IPC_W))) {

 #ifdef SEM_DEBUG
                printf("eval = %d from ipaccess\n", eval);
 #endif
 #ifdef SEM_DEBUG
                printf("eval = %d from ipaccess\n", eval);
 #endif

-               UNLOCK_AND_RETURN(eval);
+               goto semopout;

        }
 
        }
 

-       if (nsops > MAX_SOPS) {
+       if (nsops < 0 || nsops > seminfo.semopm) {

 #ifdef SEM_DEBUG
 #ifdef SEM_DEBUG

-               printf("too many sops (max=%d, nsops=%d)\n", MAX_SOPS, nsops);
+               printf("too many sops (max=%d, nsops=%d)\n",
+                   seminfo.semopm, nsops);

 #endif
 #endif

-               UNLOCK_AND_RETURN(E2BIG);
+               eval = E2BIG;
+               goto semopout;

        }
        }

-
-       if ((eval = copyin(uap->sops, &sops, nsops * sizeof(sops[0]))) != 0) {
+       
+       /*  OK for LP64, since sizeof(struct sembuf) is currently invariant */
+       if ((eval = copyin(uap->sops, &sops, nsops * sizeof(struct sembuf))) != 0) {

 #ifdef SEM_DEBUG
                printf("eval = %d from copyin(%08x, %08x, %ld)\n", eval,
 #ifdef SEM_DEBUG
                printf("eval = %d from copyin(%08x, %08x, %ld)\n", eval,

-                   uap->sops, &sops, nsops * sizeof(sops[0]));
+                   uap->sops, &sops, nsops * sizeof(struct sembuf));

 #endif
 #endif

-               UNLOCK_AND_RETURN(eval);
+               goto semopout;

        }
 
        }
 

+#if CONFIG_MACF
+       /*
+        * Initial pass thru sops to see what permissions are needed.
+        */
+       j = 0;          /* permission needed */
+       for (i = 0; i < nsops; i++)
+               j |= (sops[i].sem_op == 0) ? SEM_R : SEM_A;
+
+       /*
+        * The MAC hook checks whether the thread has read (and possibly
+        * write) permissions to the semaphore array based on the
+        * sopptr->sem_op value.
+        */
+       eval = mac_sysvsem_check_semop(kauth_cred_get(), semakptr, j);
+       if (eval)
+               goto semopout;
+#endif
+

        /*
         * Loop trying to satisfy the vector of requests.
         * If we reach a point where we must wait, any requests already
        /*
         * Loop trying to satisfy the vector of requests.
         * If we reach a point where we must wait, any requests already


@@ -1024,14 +1155,16 @@ semop(p, uap, retval)
                for (i = 0; i < nsops; i++) {
                        sopptr = &sops[i];
 
                for (i = 0; i < nsops; i++) {
                        sopptr = &sops[i];
 

-                       if (sopptr->sem_num >= semaptr->sem_nsems)
-                               UNLOCK_AND_RETURN(EFBIG);
+                       if (sopptr->sem_num >= semakptr->u.sem_nsems) {
+                               eval = EFBIG;
+                               goto semopout;
+                       }

 
 

-                       semptr = &semaptr->sem_base[sopptr->sem_num];
+                       semptr = &semakptr->u.sem_base[sopptr->sem_num];

 
 #ifdef SEM_DEBUG
 
 #ifdef SEM_DEBUG

-                       printf("semop:  semaptr=%x, sem_base=%x, semptr=%x, sem[%d]=%d : op=%d, flag=%s\n",
-                           semaptr, semaptr->sem_base, semptr,
+                       printf("semop:  semakptr=%x, sem_base=%x, semptr=%x, sem[%d]=%d : op=%d, flag=%s\n",
+                           semakptr, semakptr->u.sem_base, semptr,

                            sopptr->sem_num, semptr->semval, sopptr->sem_op,
                            (sopptr->sem_flg & IPC_NOWAIT) ? "nowait" : "wait");
 #endif
                            sopptr->sem_num, semptr->semval, sopptr->sem_op,
                            (sopptr->sem_flg & IPC_NOWAIT) ? "nowait" : "wait");
 #endif


@@ -1079,15 +1212,17 @@ semop(p, uap, retval)
                printf("semop:  rollback 0 through %d\n", i-1);
 #endif
                for (j = 0; j < i; j++)
                printf("semop:  rollback 0 through %d\n", i-1);
 #endif
                for (j = 0; j < i; j++)

-                       semaptr->sem_base[sops[j].sem_num].semval -=
+                       semakptr->u.sem_base[sops[j].sem_num].semval -=

                            sops[j].sem_op;
 
                /*
                 * If the request that we couldn't satisfy has the
                 * NOWAIT flag set then return with EAGAIN.
                 */
                            sops[j].sem_op;
 
                /*
                 * If the request that we couldn't satisfy has the
                 * NOWAIT flag set then return with EAGAIN.
                 */

-               if (sopptr->sem_flg & IPC_NOWAIT)
-                       UNLOCK_AND_RETURN(EAGAIN);
+               if (sopptr->sem_flg & IPC_NOWAIT) {
+                       eval = EAGAIN;
+                       goto semopout;
+               }

 
                if (sopptr->sem_op == 0)
                        semptr->semzcnt++;
 
                if (sopptr->sem_op == 0)
                        semptr->semzcnt++;


@@ -1102,43 +1237,51 @@ semop(p, uap, retval)
                 * waiting for. We will get the lock back after we
                 * wake up.
                 */
                 * waiting for. We will get the lock back after we
                 * wake up.
                 */

-               SUBSYSTEM_LOCK_RELEASE;
-                sysv_sem_sleeping_threads++;
-               eval = tsleep((caddr_t)semaptr, (PZERO - 4) | PCATCH,
+               eval = msleep((caddr_t)semakptr, &sysv_sem_subsys_mutex , (PZERO - 4) | PCATCH,

                    "semwait", 0);
                    "semwait", 0);

-                sysv_sem_sleeping_threads--;

                 
 #ifdef SEM_DEBUG
                printf("semop:  good morning (eval=%d)!\n", eval);
 #endif
                 
 #ifdef SEM_DEBUG
                printf("semop:  good morning (eval=%d)!\n", eval);
 #endif

-               /* There is no need to get the lock if we are just
-                * going to return without performing more semaphore
-                * operations.
-                */
-               if (eval != 0)
-                       return(EINTR);
-
-               SUBSYSTEM_LOCK_AQUIRE(p);       /* Get it back */
-               suptr = NULL;   /* sem_undo may have been reallocated */
-               semaptr = &sema[semid];    /* sema may have been reallocated */
+               if (eval != 0) {
+                       eval = EINTR;
+               }

 
 

+               /*
+                * IMPORTANT: while we were asleep, the semaphore array might
+                * have been reallocated somewhere else (see grow_sema_array()).
+                * When we wake up, we have to re-lookup the semaphore 
+                * structures and re-validate them.
+                */

 
 

-#ifdef SEM_DEBUG
-               printf("semop:  good morning!\n");
-#endif
+               semptr = NULL;

 
                /*
                 * Make sure that the semaphore still exists
 
                /*
                 * Make sure that the semaphore still exists

+                *
+                * XXX POSIX: Third test this 'if' and 'EINTR' precedence may
+                * fail testing; if so, we will need to revert this code.

                 */
                 */

-               if ((semaptr->sem_perm.mode & SEM_ALLOC) == 0 ||
-                   semaptr->sem_perm.seq != IPCID_TO_SEQ(uap->semid)) {
+               semakptr = &sema[semid];   /* sema may have been reallocated */
+               if ((semakptr->u.sem_perm.mode & SEM_ALLOC) == 0 ||
+                   semakptr->u.sem_perm._seq != IPCID_TO_SEQ(uap->semid) ||
+                   sopptr->sem_num >= semakptr->u.sem_nsems) {

                        /* The man page says to return EIDRM. */
                        /* Unfortunately, BSD doesn't define that code! */
                        /* The man page says to return EIDRM. */
                        /* Unfortunately, BSD doesn't define that code! */

+                       if (eval == EINTR) {
+                               /*
+                                * EINTR takes precedence over the fact that
+                                * the semaphore disappeared while we were
+                                * sleeping...
+                                */
+                       } else {

 #ifdef EIDRM
 #ifdef EIDRM

-                       UNLOCK_AND_RETURN(EIDRM);
+                               eval = EIDRM;

 #else
 #else

-                       UNLOCK_AND_RETURN(EINVAL);
+                               eval = EINVAL;          /* Ancient past */

 #endif
 #endif

+                       }
+                       goto semopout;

                }
 
                /*
                }
 
                /*


@@ -1147,11 +1290,15 @@ semop(p, uap, retval)
                 * because the sem[] may have been reallocated while
                 * we were sleeping, updating our sem_base pointer.
                 */
                 * because the sem[] may have been reallocated while
                 * we were sleeping, updating our sem_base pointer.
                 */

-               semptr = &semaptr->sem_base[sopptr->sem_num];
+               semptr = &semakptr->u.sem_base[sopptr->sem_num];

                if (sopptr->sem_op == 0)
                        semptr->semzcnt--;
                else
                        semptr->semncnt--;
                if (sopptr->sem_op == 0)
                        semptr->semzcnt--;
                else
                        semptr->semncnt--;

+
+               if (eval != 0) { /* EINTR */
+                       goto semopout;
+               }

        }
 
 done:
        }
 
 done:


@@ -1171,7 +1318,7 @@ done:
                        adjval = sops[i].sem_op;
                        if (adjval == 0)
                                continue;
                        adjval = sops[i].sem_op;
                        if (adjval == 0)
                                continue;

-                       eval = semundo_adjust(p, &suptr, semid,
+                       eval = semundo_adjust(p, &supidx, semid,

                            sops[i].sem_num, -adjval);
                        if (eval == 0)
                                continue;
                            sops[i].sem_num, -adjval);
                        if (eval == 0)
                                continue;


@@ -1191,57 +1338,51 @@ done:
                                adjval = sops[j].sem_op;
                                if (adjval == 0)
                                        continue;
                                adjval = sops[j].sem_op;
                                if (adjval == 0)
                                        continue;

-                               if (semundo_adjust(p, &suptr, semid,
+                               if (semundo_adjust(p, &supidx, semid,

                                    sops[j].sem_num, adjval) != 0)
                                        panic("semop - can't undo undos");
                        }
 
                        for (j = 0; j < nsops; j++)
                                    sops[j].sem_num, adjval) != 0)
                                        panic("semop - can't undo undos");
                        }
 
                        for (j = 0; j < nsops; j++)

-                               semaptr->sem_base[sops[j].sem_num].semval -=
+                               semakptr->u.sem_base[sops[j].sem_num].semval -=

                                    sops[j].sem_op;
 
 #ifdef SEM_DEBUG
                        printf("eval = %d from semundo_adjust\n", eval);
 #endif
                                    sops[j].sem_op;
 
 #ifdef SEM_DEBUG
                        printf("eval = %d from semundo_adjust\n", eval);
 #endif

-                       UNLOCK_AND_RETURN(eval);
+                       goto semopout;

                } /* loop through the sops */
        } /* if (do_undos) */
 
        /* We're definitely done - set the sempid's */
        for (i = 0; i < nsops; i++) {
                sopptr = &sops[i];
                } /* loop through the sops */
        } /* if (do_undos) */
 
        /* We're definitely done - set the sempid's */
        for (i = 0; i < nsops; i++) {
                sopptr = &sops[i];

-               semptr = &semaptr->sem_base[sopptr->sem_num];
+               semptr = &semakptr->u.sem_base[sopptr->sem_num];

                semptr->sempid = p->p_pid;
        }
                semptr->sempid = p->p_pid;
        }

+       semakptr->u.sem_otime = sysv_semtime();

 
 

-       /* Do a wakeup if any semaphore was up'd.
-        *  we will release our lock on the semaphore subsystem before
-        *  we wakeup other processes to prevent a little thrashing.
-        *  Note that this is fine because we are done using the
-        *  semaphore structures at this point in time.  We only use
-        *  a local variable pointer value, and the retval
-        *  parameter.
-        *  Note 2: Future use of sem_wakeup may reqiure the lock.
-        */
-       SUBSYSTEM_LOCK_RELEASE;

        if (do_wakeup) {
 #ifdef SEM_DEBUG
                printf("semop:  doing wakeup\n");
 #ifdef SEM_WAKEUP
        if (do_wakeup) {
 #ifdef SEM_DEBUG
                printf("semop:  doing wakeup\n");
 #ifdef SEM_WAKEUP

-               sem_wakeup((caddr_t)semaptr);
+               sem_wakeup((caddr_t)semakptr);

 #else
 #else

-               wakeup((caddr_t)semaptr);
+               wakeup((caddr_t)semakptr);

 #endif
                printf("semop:  back from wakeup\n");
 #else
 #endif
                printf("semop:  back from wakeup\n");
 #else

-               wakeup((caddr_t)semaptr);
+               wakeup((caddr_t)semakptr);

 #endif
        }
 #ifdef SEM_DEBUG
        printf("semop:  done\n");
 #endif
        *retval = 0;
 #endif
        }
 #ifdef SEM_DEBUG
        printf("semop:  done\n");
 #endif
        *retval = 0;

-       return(0);
+       eval = 0;
+semopout:
+       SYSV_SEM_SUBSYS_UNLOCK();
+       return(eval);

 }
 
 /*
 }
 
 /*


@@ -1249,21 +1390,22 @@ done:
  * semaphores.
  */
 void
  * semaphores.
  */
 void

-semexit(p)
-       struct proc *p;
+semexit(struct proc *p)

 {
 {

-       register struct sem_undo *suptr;
-       register struct sem_undo **supptr;
+       register struct sem_undo *suptr = NULL;
+       int suidx;
+       int *supidx;

        int did_something;
 
        /* If we have not allocated our semaphores yet there can't be
         * anything to undo, but we need the lock to prevent
         * dynamic memory race conditions.
         */
        int did_something;
 
        /* If we have not allocated our semaphores yet there can't be
         * anything to undo, but we need the lock to prevent
         * dynamic memory race conditions.
         */

-       SUBSYSTEM_LOCK_AQUIRE(p);
-       if (!sem)
+       SYSV_SEM_SUBSYS_LOCK();
+
+       if (!sem_pool)

        {
        {

-               SUBSYSTEM_LOCK_RELEASE;
+               SYSV_SEM_SUBSYS_UNLOCK();

                return;
        }
        did_something = 0;
                return;
        }
        did_something = 0;


@@ -1273,13 +1415,14 @@ semexit(p)
         * associated with this process.
         */
 
         * associated with this process.
         */
 

-       for (supptr = &semu_list; (suptr = *supptr) != NULL;
-           supptr = &suptr->un_next) {
+       for (supidx = &semu_list_idx; (suidx = *supidx) != -1;
+           supidx = &suptr->un_next_idx) {
+               suptr = SEMU(suidx);

                if (suptr->un_proc == p)
                        break;
        }
 
                if (suptr->un_proc == p)
                        break;
        }
 

-       if (suptr == NULL)
+       if (suidx == -1)

                goto unlock;
 
 #ifdef SEM_DEBUG
                goto unlock;
 
 #ifdef SEM_DEBUG


@@ -1291,38 +1434,43 @@ semexit(p)
         * If there are any active undo elements then process them.
         */
        if (suptr->un_cnt > 0) {
         * If there are any active undo elements then process them.
         */
        if (suptr->un_cnt > 0) {

-               int ix;
+               while (suptr->un_ent != NULL) {
+                       struct undo *sueptr;
+                       int semid;
+                       int semnum;
+                       int adjval;
+                       struct semid_kernel *semakptr;

 
 

-               for (ix = 0; ix < suptr->un_cnt; ix++) {
-                       int semid = suptr->un_ent[ix].un_id;
-                       int semnum = suptr->un_ent[ix].un_num;
-                       int adjval = suptr->un_ent[ix].un_adjval;
-                       struct semid_ds *semaptr;
+                       sueptr = suptr->un_ent;
+                       semid = sueptr->une_id;
+                       semnum = sueptr->une_num;
+                       adjval = sueptr->une_adjval;

 
 

-                       semaptr = &sema[semid];
-                       if ((semaptr->sem_perm.mode & SEM_ALLOC) == 0)
+                       semakptr = &sema[semid];
+                       if ((semakptr->u.sem_perm.mode & SEM_ALLOC) == 0)

                                panic("semexit - semid not allocated");
                                panic("semexit - semid not allocated");

-                       if (semnum >= semaptr->sem_nsems)
+                       if (semnum >= semakptr->u.sem_nsems)

                                panic("semexit - semnum out of range");
 
 #ifdef SEM_DEBUG
                        printf("semexit:  %08x id=%d num=%d(adj=%d) ; sem=%d\n",
                                panic("semexit - semnum out of range");
 
 #ifdef SEM_DEBUG
                        printf("semexit:  %08x id=%d num=%d(adj=%d) ; sem=%d\n",

-                           suptr->un_proc, suptr->un_ent[ix].un_id,
-                           suptr->un_ent[ix].un_num,
-                           suptr->un_ent[ix].un_adjval,
-                           semaptr->sem_base[semnum].semval);
+                              suptr->un_proc,
+                              semid,
+                              semnum,
+                              adjval,
+                              semakptr->u.sem_base[semnum].semval);

 #endif
 
                        if (adjval < 0) {
 #endif
 
                        if (adjval < 0) {

-                               if (semaptr->sem_base[semnum].semval < -adjval)
-                                       semaptr->sem_base[semnum].semval = 0;
+                               if (semakptr->u.sem_base[semnum].semval < -adjval)
+                                       semakptr->u.sem_base[semnum].semval = 0;

                                else
                                else

-                                       semaptr->sem_base[semnum].semval +=
+                                       semakptr->u.sem_base[semnum].semval +=

                                            adjval;
                        } else
                                            adjval;
                        } else

-                               semaptr->sem_base[semnum].semval += adjval;
+                               semakptr->u.sem_base[semnum].semval += adjval;

 
 

-               /* Maybe we should build a list of semaptr's to wake
+               /* Maybe we should build a list of semakptr's to wake

                 * up, finish all access to data structures, release the
                 * subsystem lock, and wake all the processes.  Something
                 * to think about.  It wouldn't buy us anything unless
                 * up, finish all access to data structures, release the
                 * subsystem lock, and wake all the processes.  Something
                 * to think about.  It wouldn't buy us anything unless


@@ -1331,13 +1479,17 @@ semexit(p)
                 * in the BSD code at once.
                 */
 #ifdef SEM_WAKEUP
                 * in the BSD code at once.
                 */
 #ifdef SEM_WAKEUP

-                       sem_wakeup((caddr_t)semaptr);
+                       sem_wakeup((caddr_t)semakptr);

 #else
 #else

-                       wakeup((caddr_t)semaptr);
+                       wakeup((caddr_t)semakptr);

 #endif
 #ifdef SEM_DEBUG
                        printf("semexit:  back from wakeup\n");
 #endif
 #endif
 #ifdef SEM_DEBUG
                        printf("semexit:  back from wakeup\n");
 #endif

+                       suptr->un_cnt--;
+                       suptr->un_ent = sueptr->une_next;
+                       FREE(sueptr, M_SYSVSEM);
+                       sueptr = NULL;

                }
        }
 
                }
        }
 


@@ -1348,7 +1500,7 @@ semexit(p)
        printf("removing vector\n");
 #endif
        suptr->un_proc = NULL;
        printf("removing vector\n");
 #endif
        suptr->un_proc = NULL;

-       *supptr = suptr->un_next;
+       *supidx = suptr->un_next_idx;

 
 unlock:
        /*
 
 unlock:
        /*


@@ -1370,51 +1522,170 @@ unlock:
          * same leaky semaphore problem.
          */
 
          * same leaky semaphore problem.
          */
 

-       SUBSYSTEM_LOCK_RELEASE;
+       SYSV_SEM_SUBSYS_UNLOCK();

 }
 }

+
+

 /* (struct sysctl_oid *oidp, void *arg1, int arg2, \
         struct sysctl_req *req) */
 static int
 /* (struct sysctl_oid *oidp, void *arg1, int arg2, \
         struct sysctl_req *req) */
 static int

-sysctl_seminfo SYSCTL_HANDLER_ARGS
+sysctl_seminfo(__unused struct sysctl_oid *oidp, void *arg1,
+       __unused int arg2, struct sysctl_req *req)

 {
        int error = 0;
 
        error = SYSCTL_OUT(req, arg1, sizeof(int));
 {
        int error = 0;
 
        error = SYSCTL_OUT(req, arg1, sizeof(int));

-       if (error || !req->newptr)
+       if (error || req->newptr == USER_ADDR_NULL)

                return(error);
 
                return(error);
 

-       SUBSYSTEM_LOCK_AQUIRE(current_proc());
+       SYSV_SEM_SUBSYS_LOCK();
+

        /* Set the values only if shared memory is not initialised */
        /* Set the values only if shared memory is not initialised */

-       if ((sem == (struct sem *) 0) && 
-               (sema == (struct semid_ds *) 0) && 
-               (semu == (struct semid_ds *) 0) && 
-               (semu_list == (struct sem_undo *) 0)) {
-                       if (error = SYSCTL_IN(req, arg1, sizeof(int))) {
+       if ((sem_pool == NULL) && 
+               (sema == NULL) && 
+               (semu == NULL) && 
+               (semu_list_idx == -1)) {
+                       if ((error = SYSCTL_IN(req, arg1, sizeof(int)))) {

                                goto out;
                        }
        } else 
                error = EINVAL;
 out:
                                goto out;
                        }
        } else 
                error = EINVAL;
 out:

-       SUBSYSTEM_LOCK_RELEASE;
+       SYSV_SEM_SUBSYS_UNLOCK();

        return(error);
        
 }
 
 /* SYSCTL_NODE(_kern, KERN_SYSV, sysv, CTLFLAG_RW, 0, "SYSV"); */
 extern struct sysctl_oid_list sysctl__kern_sysv_children;
        return(error);
        
 }
 
 /* SYSCTL_NODE(_kern, KERN_SYSV, sysv, CTLFLAG_RW, 0, "SYSV"); */
 extern struct sysctl_oid_list sysctl__kern_sysv_children;

-SYSCTL_PROC(_kern_sysv, KSYSV_SEMMNI, semmni, CTLTYPE_INT | CTLFLAG_RW,
+SYSCTL_PROC(_kern_sysv, OID_AUTO, semmni, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,

     &limitseminfo.semmni, 0, &sysctl_seminfo ,"I","semmni");
 
     &limitseminfo.semmni, 0, &sysctl_seminfo ,"I","semmni");
 

-SYSCTL_PROC(_kern_sysv, KSYSV_SEMMNS, semmns, CTLTYPE_INT | CTLFLAG_RW,
+SYSCTL_PROC(_kern_sysv, OID_AUTO, semmns, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,

     &limitseminfo.semmns, 0, &sysctl_seminfo ,"I","semmns");
 
     &limitseminfo.semmns, 0, &sysctl_seminfo ,"I","semmns");
 

-SYSCTL_PROC(_kern_sysv, KSYSV_SEMMNU, semmnu, CTLTYPE_INT | CTLFLAG_RW,
+SYSCTL_PROC(_kern_sysv, OID_AUTO, semmnu, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,

     &limitseminfo.semmnu, 0, &sysctl_seminfo ,"I","semmnu");
 
     &limitseminfo.semmnu, 0, &sysctl_seminfo ,"I","semmnu");
 

-SYSCTL_PROC(_kern_sysv, KSYSV_SEMMSL, semmsl, CTLTYPE_INT | CTLFLAG_RW,
+SYSCTL_PROC(_kern_sysv, OID_AUTO, semmsl, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,

     &limitseminfo.semmsl, 0, &sysctl_seminfo ,"I","semmsl");
     
     &limitseminfo.semmsl, 0, &sysctl_seminfo ,"I","semmsl");
     

-SYSCTL_PROC(_kern_sysv, KSYSV_SEMUNE, semume, CTLTYPE_INT | CTLFLAG_RW,
+SYSCTL_PROC(_kern_sysv, OID_AUTO, semume, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,

     &limitseminfo.semume, 0, &sysctl_seminfo ,"I","semume");
 
 
     &limitseminfo.semume, 0, &sysctl_seminfo ,"I","semume");
 
 

+static int
+IPCS_sem_sysctl(__unused struct sysctl_oid *oidp, __unused void *arg1,
+       __unused int arg2, struct sysctl_req *req)
+{
+       int error;
+       int cursor;
+       union {
+               struct user32_IPCS_command u32;
+               struct user_IPCS_command u64;
+       } ipcs;
+       struct user32_semid_ds semid_ds32;      /* post conversion, 32 bit version */
+       struct user64_semid_ds semid_ds64;      /* post conversion, 64 bit version */
+       void *semid_dsp;
+       size_t ipcs_sz;
+       size_t semid_ds_sz;
+       struct proc *p = current_proc();
+
+       if (IS_64BIT_PROCESS(p)) {
+               ipcs_sz = sizeof(struct user_IPCS_command);
+               semid_ds_sz = sizeof(struct user64_semid_ds);
+       } else {
+               ipcs_sz = sizeof(struct user32_IPCS_command);
+               semid_ds_sz = sizeof(struct user32_semid_ds);
+       }
+
+       /* Copy in the command structure */
+       if ((error = SYSCTL_IN(req, &ipcs, ipcs_sz)) != 0) {
+               return(error);
+       }
+
+       if (!IS_64BIT_PROCESS(p)) /* convert in place */
+               ipcs.u64.ipcs_data = CAST_USER_ADDR_T(ipcs.u32.ipcs_data);
+
+       /* Let us version this interface... */
+       if (ipcs.u64.ipcs_magic != IPCS_MAGIC) {
+               return(EINVAL);
+       }
+
+       SYSV_SEM_SUBSYS_LOCK();
+       switch(ipcs.u64.ipcs_op) {
+       case IPCS_SEM_CONF:     /* Obtain global configuration data */
+               if (ipcs.u64.ipcs_datalen != sizeof(struct seminfo)) {
+                       error = ERANGE;
+                       break;
+               }
+               if (ipcs.u64.ipcs_cursor != 0) {        /* fwd. compat. */
+                       error = EINVAL;
+                       break;
+               }
+               error = copyout(&seminfo, ipcs.u64.ipcs_data, ipcs.u64.ipcs_datalen);
+               break;
+
+       case IPCS_SEM_ITER:     /* Iterate over existing segments */
+               cursor = ipcs.u64.ipcs_cursor;
+               if (cursor < 0 || cursor >= seminfo.semmni) {
+                       error = ERANGE;
+                       break;
+               }
+               if (ipcs.u64.ipcs_datalen != (int)semid_ds_sz ) {
+                       error = EINVAL;
+                       break;
+               }
+               for( ; cursor < seminfo.semmni; cursor++) {
+                       if (sema[cursor].u.sem_perm.mode & SEM_ALLOC)
+                               break;
+                       continue;
+               }
+               if (cursor == seminfo.semmni) {
+                       error = ENOENT;
+                       break;
+               }
+
+               semid_dsp = &sema[cursor].u;    /* default: 64 bit */
+
+               /*
+                * If necessary, convert the 64 bit kernel segment
+                * descriptor to a 32 bit user one.
+                */
+               if (!IS_64BIT_PROCESS(p)) {
+                       bzero(&semid_ds32, sizeof(semid_ds32));
+                       semid_ds_kernelto32(semid_dsp, &semid_ds32);
+                       semid_dsp = &semid_ds32;
+               } else {
+                       bzero(&semid_ds64, sizeof(semid_ds64));
+                       semid_ds_kernelto64(semid_dsp, &semid_ds64);
+                       semid_dsp = &semid_ds64;
+               }
+
+               error = copyout(semid_dsp, ipcs.u64.ipcs_data, ipcs.u64.ipcs_datalen);
+               if (!error) {
+                       /* update cursor */
+                       ipcs.u64.ipcs_cursor = cursor + 1;
+
+                       if (!IS_64BIT_PROCESS(p))       /* convert in place */
+                               ipcs.u32.ipcs_data = CAST_DOWN_EXPLICIT(user32_addr_t,ipcs.u64.ipcs_data);
+
+                       error = SYSCTL_OUT(req, &ipcs, ipcs_sz);
+               }
+               break;
+
+       default:
+               error = EINVAL;
+               break;
+       }
+       SYSV_SEM_SUBSYS_UNLOCK();
+       return(error);
+}
+
+SYSCTL_DECL(_kern_sysv_ipcs);
+SYSCTL_PROC(_kern_sysv_ipcs, OID_AUTO, sem, CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_LOCKED,
+       0, 0, IPCS_sem_sysctl,
+       "S,IPCS_sem_command",
+       "ipcs sem command interface");
+
+#endif /* SYSV_SEM */