xnu-2050.7.9.tar.gz

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

diff --git a/bsd/kern/posix_shm.c b/bsd/kern/posix_shm.c
index 7ad07ace0907f992af8c99f0544aa296875ed3a5..d46eb0b3c1268ed385b2841d55824213b98190eb 100644 (file)
--- a/bsd/kern/posix_shm.c
+++ b/bsd/kern/posix_shm.c
@@ -67,7 +67,7 @@
 #include <sys/stat.h>
 #include <sys/sysproto.h>
 #include <sys/proc_info.h>
 #include <sys/stat.h>
 #include <sys/sysproto.h>
 #include <sys/proc_info.h>

-#include <bsm/audit_kernel.h>
+#include <security/audit/audit.h>

 
 #if CONFIG_MACF
 #include <security/mac_framework.h>
 
 #if CONFIG_MACF
 #include <security/mac_framework.h>


@@ -93,6 +93,12 @@
 #define f_data f_fglob->fg_data
 #define        PSHMNAMLEN      31      /* maximum name segment length we bother with */
 
 #define f_data f_fglob->fg_data
 #define        PSHMNAMLEN      31      /* maximum name segment length we bother with */
 

+struct pshmobj {
+       void *                  pshmo_memobject;
+       memory_object_size_t    pshmo_size;
+       struct pshmobj *        pshmo_next;
+};
+

 struct pshminfo {
        unsigned int    pshm_flags;
        unsigned int    pshm_usecount;
 struct pshminfo {
        unsigned int    pshm_flags;
        unsigned int    pshm_usecount;


@@ -101,7 +107,7 @@ struct pshminfo {
        uid_t           pshm_uid;
        gid_t           pshm_gid;
        char            pshm_name[PSHMNAMLEN + 1];      /* segment name */
        uid_t           pshm_uid;
        gid_t           pshm_gid;
        char            pshm_name[PSHMNAMLEN + 1];      /* segment name */

-       void *          pshm_memobject;
+       struct pshmobj *pshm_memobjects;

 #if DIAGNOSTIC
        unsigned int    pshm_readcount;
        unsigned int    pshm_writecount;
 #if DIAGNOSTIC
        unsigned int    pshm_readcount;
        unsigned int    pshm_writecount;


@@ -111,14 +117,15 @@ struct pshminfo {
 };
 #define PSHMINFO_NULL (struct pshminfo *)0
 
 };
 #define PSHMINFO_NULL (struct pshminfo *)0
 

-#define        PSHM_NONE       1
-#define        PSHM_DEFINED    2
-#define        PSHM_ALLOCATED  4
-#define        PSHM_MAPPED     8
-#define        PSHM_INUSE      0x10
-#define        PSHM_REMOVED    0x20
-#define        PSHM_INCREATE   0x40
-#define        PSHM_INDELETE   0x80
+#define        PSHM_NONE       0x001
+#define        PSHM_DEFINED    0x002
+#define        PSHM_ALLOCATED  0x004
+#define        PSHM_MAPPED     0x008
+#define        PSHM_INUSE      0x010
+#define        PSHM_REMOVED    0x020
+#define        PSHM_INCREATE   0x040
+#define        PSHM_INDELETE   0x080
+#define PSHM_ALLOCATING        0x100

 
 struct pshmcache {
        LIST_ENTRY(pshmcache) pshm_hash;        /* hash chain */
 
 struct pshmcache {
        LIST_ENTRY(pshmcache) pshm_hash;        /* hash chain */


@@ -145,7 +152,7 @@ struct pshmname {
 
 struct pshmnode {
        off_t           mapp_addr;
 
 struct pshmnode {
        off_t           mapp_addr;

-       user_size_t     map_size;
+       user_size_t     map_size;       /* XXX unused ? */

        struct pshminfo *pinfo;
        unsigned int    pshm_usecount;
 #if DIAGNOSTIC
        struct pshminfo *pinfo;
        unsigned int    pshm_usecount;
 #if DIAGNOSTIC


@@ -171,7 +178,7 @@ static int pshm_write (struct fileproc *fp, struct uio *uio,
 static int pshm_ioctl (struct fileproc *fp, u_long com,
                    caddr_t data, vfs_context_t ctx);
 static int pshm_select (struct fileproc *fp, int which, void *wql, vfs_context_t ctx);
 static int pshm_ioctl (struct fileproc *fp, u_long com,
                    caddr_t data, vfs_context_t ctx);
 static int pshm_select (struct fileproc *fp, int which, void *wql, vfs_context_t ctx);

-static int pshm_close(struct pshmnode *pnode);
+static int pshm_close(struct pshminfo *pinfo, int dropref);

 static int pshm_closefile (struct fileglob *fg, vfs_context_t ctx);
 
 static int pshm_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx);
 static int pshm_closefile (struct fileglob *fg, vfs_context_t ctx);
 
 static int pshm_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx);


@@ -183,7 +190,7 @@ static void pshm_cache_delete(struct pshmcache *pcp);
 static void pshm_cache_purge(void);
 #endif /* NOT_USED */
 static int pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,
 static void pshm_cache_purge(void);
 #endif /* NOT_USED */
 static int pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,

-       struct pshmcache **pcache);
+       struct pshmcache **pcache, int addref);

 
 struct         fileops pshmops =
        { pshm_read, pshm_write, pshm_ioctl, pshm_select, pshm_closefile, pshm_kqfilter, 0 };
 
 struct         fileops pshmops =
        { pshm_read, pshm_write, pshm_ioctl, pshm_select, pshm_closefile, pshm_kqfilter, 0 };


@@ -222,7 +229,7 @@ pshm_lock_init( void )
 
 static int
 pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,
 
 static int
 pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,

-       struct pshmcache **pcache)
+       struct pshmcache **pcache, int addref)

 {
        struct pshmcache *pcp, *nnp;
        struct pshmhashhead *pcpp;
 {
        struct pshmcache *pcp, *nnp;
        struct pshmhashhead *pcpp;


@@ -251,6 +258,8 @@ pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,
                /* TOUCH(ncp); */
                *pshmp = pcp->pshminfo;
                *pcache = pcp;
                /* TOUCH(ncp); */
                *pshmp = pcp->pshminfo;
                *pcache = pcp;

+               if (addref)
+                       pcp->pshminfo->pshm_usecount++;

                return (-1);
        }
 
                return (-1);
        }
 


@@ -280,7 +289,7 @@ pshm_cache_add(struct pshminfo *pshmp, struct pshmname *pnp, struct pshmcache *p
 
 
        /*  if the entry has already been added by some one else return */
 
 
        /*  if the entry has already been added by some one else return */

-       if (pshm_cache_search(&dpinfo, pnp, &dpcp) == -1) {
+       if (pshm_cache_search(&dpinfo, pnp, &dpcp, 0) == -1) {

                return(EEXIST);
        }
        pshmnument++;
                return(EEXIST);
        }
        pshmnument++;


@@ -355,34 +364,38 @@ pshm_cache_delete(struct pshmcache *pcp)
 
 
 int
 
 
 int

-shm_open(proc_t p, struct shm_open_args *uap, register_t *retval)
+shm_open(proc_t p, struct shm_open_args *uap, int32_t *retval)

 {
 {

-       struct fileproc *fp;

        size_t  i;
        size_t  i;

-       struct fileproc *nfp;

        int indx, error;
        struct pshmname nd;
        struct pshminfo *pinfo;
        int indx, error;
        struct pshmname nd;
        struct pshminfo *pinfo;

-       char * pnbuf;
+       struct fileproc *fp = NULL;
+       char *pnbuf = NULL;
+       struct pshminfo *new_pinfo = PSHMINFO_NULL;
+       struct pshmnode *new_pnode = PSHMNODE_NULL;
+       struct pshmcache *pcache = PSHMCACHE_NULL;      /* ignored on return */

        char * nameptr;
        char * cp;
        size_t pathlen, plen;
        int fmode ;
        int cmode = uap->mode;
        int incache = 0;
        char * nameptr;
        char * cp;
        size_t pathlen, plen;
        int fmode ;
        int cmode = uap->mode;
        int incache = 0;

-       struct pshmnode * pnode = PSHMNODE_NULL;
-       struct pshmcache * pcache = PSHMCACHE_NULL;
-       struct pshmcache *pcp = NULL;   /* protected by !incache */
-       int pinfo_alloc=0;
+       struct pshmcache *pcp = NULL;

 
        AUDIT_ARG(fflags, uap->oflag);
        AUDIT_ARG(mode, uap->mode);
 
        pinfo = PSHMINFO_NULL;
 
 
        AUDIT_ARG(fflags, uap->oflag);
        AUDIT_ARG(mode, uap->mode);
 
        pinfo = PSHMINFO_NULL;
 

+       /*
+        * Preallocate everything we might need up front to avoid taking
+        * and dropping the lock, opening us up to race conditions.
+        */

        MALLOC_ZONE(pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
        if (pnbuf == NULL) {
        MALLOC_ZONE(pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
        if (pnbuf == NULL) {

-               return(ENOSPC);
+               error = ENOSPC;
+               goto bad;

        }
 
        pathlen = MAXPATHLEN;
        }
 
        pathlen = MAXPATHLEN;


@@ -395,8 +408,6 @@ shm_open(proc_t p, struct shm_open_args *uap, register_t *retval)
                error = ENAMETOOLONG;
                goto bad;
        }
                error = ENAMETOOLONG;
                goto bad;
        }

-
-

 #ifdef PSXSHM_NAME_RESTRICT
        nameptr = pnbuf;
        if (*nameptr == '/') {
 #ifdef PSXSHM_NAME_RESTRICT
        nameptr = pnbuf;
        if (*nameptr == '/') {


@@ -421,178 +432,183 @@ shm_open(proc_t p, struct shm_open_args *uap, register_t *retval)
                nd.pshm_hash += (unsigned char)*cp * i;
        }
 
                nd.pshm_hash += (unsigned char)*cp * i;
        }
 

-       PSHM_SUBSYS_LOCK();
-       error = pshm_cache_search(&pinfo, &nd, &pcache);
-
-       if (error == ENOENT) {
-               PSHM_SUBSYS_UNLOCK();
-               error = EINVAL;
+       /*
+        * attempt to allocate a new fp; if unsuccessful, the fp will be
+        * left unmodified (NULL).
+        */
+       error = falloc(p, &fp, &indx, vfs_context_current());
+       if (error) 

                goto bad;
 
                goto bad;
 

-       }
-       if (!error) {
-               incache = 0;
-       } else
-               incache = 1;
+       cmode &=  ALLPERMS;
+

        fmode = FFLAGS(uap->oflag);
        fmode = FFLAGS(uap->oflag);

-       if ((fmode & (FREAD | FWRITE))==0) {
-               PSHM_SUBSYS_UNLOCK();
+       if ((fmode & (FREAD | FWRITE)) == 0) {

                error = EINVAL;
                goto bad;
        }
 
        /*
                error = EINVAL;
                goto bad;
        }
 
        /*

-        * XXXXXXXXXX TBD XXXXXXXXXX
-        * There is a race that existed with the funnels as well.
-        * Need to be fixed later
+        * We allocate a new entry if we are less than the maximum
+        * allowed and the one at the front of the LRU list is in use.
+        * Otherwise we use the one at the front of the LRU list.

         */
         */

-       PSHM_SUBSYS_UNLOCK();
-       error = falloc(p, &nfp, &indx, vfs_context_current());
-       if (error ) 
+       MALLOC(pcp, struct pshmcache *, sizeof(struct pshmcache), M_SHM, M_WAITOK|M_ZERO);
+       if (pcp == NULL) {
+               error = ENOSPC;
+               goto bad;
+       }
+
+       MALLOC(new_pinfo, struct pshminfo *, sizeof(struct pshminfo), M_SHM, M_WAITOK|M_ZERO);
+       if (new_pinfo == PSHMINFO_NULL) {
+               error = ENOSPC;

                goto bad;
                goto bad;

+       }
+#if CONFIG_MACF
+       mac_posixshm_label_init(new_pinfo);
+#endif
+
+       MALLOC(new_pnode, struct pshmnode *, sizeof(struct pshmnode), M_SHM, M_WAITOK|M_ZERO);
+       if (new_pnode == PSHMNODE_NULL) {
+               error = ENOSPC;
+               goto bad;
+       }
+

        PSHM_SUBSYS_LOCK();
 
        PSHM_SUBSYS_LOCK();
 

-       fp = nfp;
+       /*
+        * If we find the entry in the cache, this will take a reference,
+        * allowing us to unlock it for the permissions check.
+        */
+       error = pshm_cache_search(&pinfo, &nd, &pcache, 1);

 
 

-       cmode &=  ALLPERMS;
+       PSHM_SUBSYS_UNLOCK();

 
 

-       if (fmode & O_CREAT) {
-               if ((fmode & O_EXCL) && incache) {
-                       AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid,
-                                       pinfo->pshm_gid, pinfo->pshm_mode);
+       if (error == ENOENT) {
+               error = EINVAL;
+               goto bad;
+       }

 
 

-                       /* shm obj exists and opened O_EXCL */
-#if notyet
-                       if (pinfo->pshm_flags & PSHM_INDELETE) {
-                       }
-#endif 
-                       error = EEXIST;
-                       PSHM_SUBSYS_UNLOCK();
-                       goto bad1;
-               } 
-               if (!incache) {
-                       PSHM_SUBSYS_UNLOCK();
-                       /*  create a new one */
-                       MALLOC(pinfo, struct pshminfo *, sizeof(struct pshminfo), M_SHM, M_WAITOK|M_ZERO);
-                       if (pinfo == NULL) {
-                               error = ENOSPC;
-                               goto bad1;
-                       }
-                       PSHM_SUBSYS_LOCK();
-                       pinfo_alloc = 1;
+       if (!error) {
+               incache = 0;
+               if (fmode & O_CREAT) {
+                       /*  create a new one (commit the allocation) */
+                       pinfo = new_pinfo;

                        pinfo->pshm_flags = PSHM_DEFINED | PSHM_INCREATE;
                        pinfo->pshm_usecount = 1; /* existence reference */
                        pinfo->pshm_mode = cmode;
                        pinfo->pshm_flags = PSHM_DEFINED | PSHM_INCREATE;
                        pinfo->pshm_usecount = 1; /* existence reference */
                        pinfo->pshm_mode = cmode;

-                       pinfo->pshm_uid = kauth_cred_getuid(kauth_cred_get());
-                       pinfo->pshm_gid = kauth_cred_get()->cr_gid;
-                       bcopy(pnbuf, &pinfo->pshm_name[0], PSHMNAMLEN);
-                       pinfo->pshm_name[PSHMNAMLEN]=0;
+                       pinfo->pshm_uid = kauth_getuid();
+                       pinfo->pshm_gid = kauth_getgid();
+                       bcopy(pnbuf, &pinfo->pshm_name[0], pathlen);
+                       pinfo->pshm_name[pathlen]=0;

 #if CONFIG_MACF
 #if CONFIG_MACF

-                       PSHM_SUBSYS_UNLOCK();
-                       mac_posixshm_label_init(pinfo);
-                       PSHM_SUBSYS_LOCK();

                        error = mac_posixshm_check_create(kauth_cred_get(), nameptr);
                        if (error) {
                        error = mac_posixshm_check_create(kauth_cred_get(), nameptr);
                        if (error) {

-                               PSHM_SUBSYS_UNLOCK();
-                               goto bad2;
+                               goto bad;

                        }
                        mac_posixshm_label_associate(kauth_cred_get(), pinfo, nameptr);
 #endif
                        }
                        mac_posixshm_label_associate(kauth_cred_get(), pinfo, nameptr);
 #endif

-               } else {
+               }
+       } else {
+               incache = 1;
+               if (fmode & O_CREAT) {

                        /*  already exists */
                        /*  already exists */

+                       if ((fmode & O_EXCL)) {
+                               AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid,
+                                               pinfo->pshm_gid,
+                                               pinfo->pshm_mode);
+
+                               /* shm obj exists and opened O_EXCL */
+                               error = EEXIST;
+                               goto bad;
+                       } 
+

                        if( pinfo->pshm_flags & PSHM_INDELETE) {
                        if( pinfo->pshm_flags & PSHM_INDELETE) {

-                               PSHM_SUBSYS_UNLOCK();

                                error = ENOENT;
                                error = ENOENT;

-                               goto bad1;
+                               goto bad;

                        }       
                        AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid,
                                        pinfo->pshm_gid, pinfo->pshm_mode);
 #if CONFIG_MACF        
                        }       
                        AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid,
                                        pinfo->pshm_gid, pinfo->pshm_mode);
 #if CONFIG_MACF        

-                       if ((error = mac_posixshm_check_open(
-                                                       kauth_cred_get(), pinfo))) {
-                               PSHM_SUBSYS_UNLOCK();
-                               goto bad1;
+                       if ((error = mac_posixshm_check_open(kauth_cred_get(), pinfo, fmode))) {
+                               goto bad;

                        }
 #endif
                        if ( (error = pshm_access(pinfo, fmode, kauth_cred_get(), p)) ) {
                        }
 #endif
                        if ( (error = pshm_access(pinfo, fmode, kauth_cred_get(), p)) ) {

-                               PSHM_SUBSYS_UNLOCK();
-                               goto bad1;
+                               goto bad;

                        }
                }
                        }
                }

-       } else {
+       }
+       if (!(fmode & O_CREAT)) {

                if (!incache) {
                if (!incache) {

-                       /* O_CREAT  is not set and the shm obecj does not exist */
-                       PSHM_SUBSYS_UNLOCK();
+                       /* O_CREAT is not set and the object does not exist */

                        error = ENOENT;
                        error = ENOENT;

-                       goto bad1;
+                       goto bad;

                }
                if( pinfo->pshm_flags & PSHM_INDELETE) {
                }
                if( pinfo->pshm_flags & PSHM_INDELETE) {

-                       PSHM_SUBSYS_UNLOCK();

                        error = ENOENT;
                        error = ENOENT;

-                       goto bad1;
+                       goto bad;

                }       
 #if CONFIG_MACF        
                }       
 #if CONFIG_MACF        

-               if ((error = mac_posixshm_check_open(
-                                               kauth_cred_get(), pinfo))) {
-                       PSHM_SUBSYS_UNLOCK();
-                       goto bad1;
+               if ((error = mac_posixshm_check_open(kauth_cred_get(), pinfo, fmode))) {
+                       goto bad;

                }
 #endif
 
                }
 #endif
 

-               if ( (error = pshm_access(pinfo, fmode, kauth_cred_get(), p)) ) {
-                       PSHM_SUBSYS_UNLOCK();
-                       goto bad1;
+               if ((error = pshm_access(pinfo, fmode, kauth_cred_get(), p))) {
+                       goto bad;

                }
        }
        if (fmode & O_TRUNC) {
                }
        }
        if (fmode & O_TRUNC) {

-               PSHM_SUBSYS_UNLOCK();

                error = EINVAL;
                error = EINVAL;

-               goto bad2;
+               goto bad;

        }
        }

+
+
+       PSHM_SUBSYS_LOCK();
+

 #if DIAGNOSTIC 
        if (fmode & FWRITE)
                pinfo->pshm_writecount++;
        if (fmode & FREAD)
                pinfo->pshm_readcount++;
 #endif
 #if DIAGNOSTIC 
        if (fmode & FWRITE)
                pinfo->pshm_writecount++;
        if (fmode & FREAD)
                pinfo->pshm_readcount++;
 #endif

-       PSHM_SUBSYS_UNLOCK();
-       MALLOC(pnode, struct pshmnode *, sizeof(struct pshmnode), M_SHM, M_WAITOK|M_ZERO);
-       if (pnode == NULL) {
-               error = ENOSPC;
-               goto bad2;
-       }

        if (!incache) {
        if (!incache) {

+               /* if successful, this will consume the pcp */
+               if ( (error = pshm_cache_add(pinfo, &nd, pcp)) ) {
+                       goto bad_locked;
+               }

                /*
                /*

-                * We allocate a new entry if we are less than the maximum
-                * allowed and the one at the front of the LRU list is in use.
-                * Otherwise we use the one at the front of the LRU list.
+                * add reference for the new entry; otherwise, we obtained
+                * one from the cache hit earlier.

                 */
                 */

-               MALLOC(pcp, struct pshmcache *, sizeof(struct pshmcache), M_SHM, M_WAITOK|M_ZERO);
-               if (pcp == NULL) {
-                       error = ENOSPC;
-                       goto bad2;
-               }
-
+               pinfo->pshm_usecount++;

        }
        }

-       PSHM_SUBSYS_LOCK();
+       pinfo->pshm_flags &= ~PSHM_INCREATE;
+       new_pnode->pinfo = pinfo;

 
 

-       if (!incache) {
-               if ( (error = pshm_cache_add(pinfo, &nd, pcp)) ) {
-                       PSHM_SUBSYS_UNLOCK();
-                       FREE(pcp, M_SHM);
-                       goto bad3;
+       PSHM_SUBSYS_UNLOCK();
+
+       /*
+        * if incache, we did not use the new pcp or new_pinfo and must
+        * free them
+        */
+       if (incache) {
+               FREE(pcp, M_SHM);
+
+               if (new_pinfo != PSHMINFO_NULL) {
+#if CONFIG_MACF
+                       mac_posixshm_label_destroy(new_pinfo);
+#endif
+                       FREE(new_pinfo, M_SHM);

                }
        }
                }
        }

-       pinfo->pshm_flags &= ~PSHM_INCREATE;
-       pinfo->pshm_usecount++; /* extra reference for the new fd */
-       pnode->pinfo = pinfo;

 
 

-       PSHM_SUBSYS_UNLOCK();

        proc_fdlock(p);
        fp->f_flag = fmode & FMASK;
        fp->f_type = DTYPE_PSXSHM;
        fp->f_ops = &pshmops;
        proc_fdlock(p);
        fp->f_flag = fmode & FMASK;
        fp->f_type = DTYPE_PSXSHM;
        fp->f_ops = &pshmops;

-       fp->f_data = (caddr_t)pnode;
+       fp->f_data = (caddr_t)new_pnode;

        *fdflags(p, indx) |= UF_EXCLOSE;
        procfdtbl_releasefd(p, indx, NULL);
        fp_drop(p, indx, fp, 1);
        *fdflags(p, indx) |= UF_EXCLOSE;
        procfdtbl_releasefd(p, indx, NULL);
        fp_drop(p, indx, fp, 1);


@@ -601,34 +617,53 @@ shm_open(proc_t p, struct shm_open_args *uap, register_t *retval)
        *retval = indx;
        FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);
        return (0);
        *retval = indx;
        FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);
        return (0);

-bad3:
-       FREE(pnode, M_SHM);

 
 

-bad2:
-       if (pinfo_alloc) {
+bad_locked:
+       PSHM_SUBSYS_UNLOCK();
+bad:
+       /*
+        * If we obtained the entry from the cache, we need to drop the
+        * reference; holding the reference may have prevented unlinking,
+        * so we need to call pshm_close() to get the full effect.
+        */
+       if (incache) {
+               PSHM_SUBSYS_LOCK();
+               pshm_close(pinfo, 1);
+               PSHM_SUBSYS_UNLOCK();
+       }
+
+       if (pcp != NULL)
+               FREE(pcp, M_SHM);
+
+       if (new_pnode != PSHMNODE_NULL)
+               FREE(new_pnode, M_SHM);
+
+       if (fp != NULL)
+               fp_free(p, indx, fp);
+
+       if (new_pinfo != PSHMINFO_NULL) {

 #if CONFIG_MACF
 #if CONFIG_MACF

-               mac_posixshm_label_destroy(pinfo);
+               mac_posixshm_label_destroy(new_pinfo);

 #endif
 #endif

-               FREE(pinfo, M_SHM);
+               FREE(new_pinfo, M_SHM);

        }
        }

-bad1:
-       fp_free(p, indx, fp);
-bad:
-       FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);
+       if (pnbuf != NULL)
+               FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);

        return (error);
 }
 
 
 int
 pshm_truncate(__unused proc_t p, struct fileproc *fp, __unused int fd, 
        return (error);
 }
 
 
 int
 pshm_truncate(__unused proc_t p, struct fileproc *fp, __unused int fd, 

-                               off_t length, __unused register_t *retval)
+                               off_t length, __unused int32_t *retval)

 {
        struct pshminfo * pinfo;
        struct pshmnode * pnode ;
        kern_return_t kret;
 {
        struct pshminfo * pinfo;
        struct pshmnode * pnode ;
        kern_return_t kret;

-       mach_vm_offset_t user_addr;

        mem_entry_name_port_t mem_object;
        mem_entry_name_port_t mem_object;

-       mach_vm_size_t size;
+       mach_vm_size_t total_size, alloc_size;
+       memory_object_size_t mosize;
+       struct pshmobj *pshmobj, *pshmobj_next, **pshmobj_next_p;

 #if CONFIG_MACF
        int error;
 #endif
 #if CONFIG_MACF
        int error;
 #endif


@@ -646,41 +681,78 @@ pshm_truncate(__unused proc_t p, struct fileproc *fp, __unused int fd,
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }

-       if ((pinfo->pshm_flags & (PSHM_DEFINED | PSHM_ALLOCATED)) 
+       if ((pinfo->pshm_flags & (PSHM_DEFINED|PSHM_ALLOCATING|PSHM_ALLOCATED)) 

                        != PSHM_DEFINED) {
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }
 #if CONFIG_MACF
                        != PSHM_DEFINED) {
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }
 #if CONFIG_MACF

-       error = mac_posixshm_check_truncate(kauth_cred_get(), pinfo, size);
+       error = mac_posixshm_check_truncate(kauth_cred_get(), pinfo, length);

        if (error) {
                PSHM_SUBSYS_UNLOCK();
                return(error);
        }
 #endif
        if (error) {
                PSHM_SUBSYS_UNLOCK();
                return(error);
        }
 #endif

-       PSHM_SUBSYS_UNLOCK();
-       size = round_page_64(length);
-       kret = mach_vm_allocate(current_map(), &user_addr, size, VM_FLAGS_ANYWHERE);
-       if (kret != KERN_SUCCESS) 
-               goto out;

 
 

-       kret = mach_make_memory_entry_64 (current_map(), &size,
-                       user_addr, VM_PROT_DEFAULT, &mem_object, 0);
+       pinfo->pshm_flags |= PSHM_ALLOCATING;
+       total_size = round_page_64(length);
+       pshmobj_next_p = &pinfo->pshm_memobjects;

 
 

-       if (kret != KERN_SUCCESS) 
-               goto out;
-       
-       mach_vm_deallocate(current_map(), user_addr, size);
+       for (alloc_size = 0;
+            alloc_size < total_size;
+            alloc_size += mosize) {

 
 

-       PSHM_SUBSYS_LOCK();
-       pinfo->pshm_flags &= ~PSHM_DEFINED;
+               PSHM_SUBSYS_UNLOCK();
+
+               mosize = MIN(total_size - alloc_size, ANON_MAX_SIZE);
+               kret = mach_make_memory_entry_64(
+                       VM_MAP_NULL,
+                       &mosize,
+                       0,
+                       MAP_MEM_NAMED_CREATE | VM_PROT_DEFAULT,
+                       &mem_object,
+                       0);
+
+               if (kret != KERN_SUCCESS) 
+                       goto out;
+
+               MALLOC(pshmobj, struct pshmobj *, sizeof (struct pshmobj),
+                      M_SHM, M_WAITOK);
+               if (pshmobj == NULL) {
+                       kret = KERN_NO_SPACE;
+                       mach_memory_entry_port_release(mem_object);
+                       mem_object = NULL;
+                       goto out;
+               }
+
+               PSHM_SUBSYS_LOCK();
+
+               pshmobj->pshmo_memobject = (void *) mem_object;
+               pshmobj->pshmo_size = mosize;
+               pshmobj->pshmo_next = NULL;
+               
+               *pshmobj_next_p = pshmobj;
+               pshmobj_next_p = &pshmobj->pshmo_next;
+       }
+       

        pinfo->pshm_flags = PSHM_ALLOCATED;
        pinfo->pshm_flags = PSHM_ALLOCATED;

-       pinfo->pshm_memobject = (void *)mem_object;
-       pinfo->pshm_length = size;
+       pinfo->pshm_length = total_size;

        PSHM_SUBSYS_UNLOCK();
        return(0);
 
 out:
        PSHM_SUBSYS_UNLOCK();
        return(0);
 
 out:

+       PSHM_SUBSYS_LOCK();
+       for (pshmobj = pinfo->pshm_memobjects;
+            pshmobj != NULL;
+            pshmobj = pshmobj_next) {
+               pshmobj_next = pshmobj->pshmo_next;
+               mach_memory_entry_port_release(pshmobj->pshmo_memobject);
+               FREE(pshmobj, M_SHM);
+       }
+       pinfo->pshm_memobjects = NULL;
+       pinfo->pshm_flags &= ~PSHM_ALLOCATING;
+       PSHM_SUBSYS_UNLOCK();
+

        switch (kret) {
        case KERN_INVALID_ADDRESS:
        case KERN_NO_SPACE:
        switch (kret) {
        case KERN_INVALID_ADDRESS:
        case KERN_NO_SPACE:


@@ -744,56 +816,34 @@ pshm_stat(struct pshmnode *pnode, void *ub, int isstat64)
 int
 pshm_access(struct pshminfo *pinfo, int mode, kauth_cred_t cred, __unused proc_t p)
 {
 int
 pshm_access(struct pshminfo *pinfo, int mode, kauth_cred_t cred, __unused proc_t p)
 {

-       mode_t mask;
-       int is_member;
+       int mode_req = ((mode & FREAD) ? S_IRUSR : 0) |
+                      ((mode & FWRITE) ? S_IWUSR : 0);

 
        /* Otherwise, user id 0 always gets access. */
        if (!suser(cred, NULL))
                return (0);
 
 
        /* Otherwise, user id 0 always gets access. */
        if (!suser(cred, NULL))
                return (0);
 

-       mask = 0;
-
-       /* Otherwise, check the owner. */
-       if (kauth_cred_getuid(cred) == pinfo->pshm_uid) {
-               if (mode & FREAD)
-                       mask |= S_IRUSR;
-               if (mode & FWRITE)
-                       mask |= S_IWUSR;
-               return ((pinfo->pshm_mode & mask) == mask ? 0 : EACCES);
-       }
-
-       /* Otherwise, check the groups. */
-       if (kauth_cred_ismember_gid(cred, pinfo->pshm_gid, &is_member) == 0 && is_member) {
-               if (mode & FREAD)
-                       mask |= S_IRGRP;
-               if (mode & FWRITE)
-                       mask |= S_IWGRP;
-               return ((pinfo->pshm_mode & mask) == mask ? 0 : EACCES);
-       }
-
-       /* Otherwise, check everyone else. */
-       if (mode & FREAD)
-               mask |= S_IROTH;
-       if (mode & FWRITE)
-               mask |= S_IWOTH;
-       return ((pinfo->pshm_mode & mask) == mask ? 0 : EACCES);
+       return(posix_cred_access(cred, pinfo->pshm_uid, pinfo->pshm_gid, pinfo->pshm_mode, mode_req));

 }
 
 int
 pshm_mmap(__unused proc_t p, struct mmap_args *uap, user_addr_t *retval, struct fileproc *fp, off_t pageoff) 
 {
 }
 
 int
 pshm_mmap(__unused proc_t p, struct mmap_args *uap, user_addr_t *retval, struct fileproc *fp, off_t pageoff) 
 {

-       mach_vm_offset_t        user_addr = (mach_vm_offset_t)uap->addr;
-       mach_vm_size_t          user_size = (mach_vm_size_t)uap->len ;
+       vm_map_offset_t user_addr = (vm_map_offset_t)uap->addr;
+       vm_map_size_t   user_size = (vm_map_size_t)uap->len ;
+       vm_map_offset_t user_start_addr;
+       vm_map_size_t   map_size, mapped_size;

        int prot = uap->prot;
        int flags = uap->flags;
        vm_object_offset_t file_pos = (vm_object_offset_t)uap->pos;
        int prot = uap->prot;
        int flags = uap->flags;
        vm_object_offset_t file_pos = (vm_object_offset_t)uap->pos;

+       vm_object_offset_t map_pos;

        vm_map_t        user_map;
        int             alloc_flags;
        boolean_t       docow;
        kern_return_t   kret;
        struct pshminfo * pinfo;
        struct pshmnode * pnode;
        vm_map_t        user_map;
        int             alloc_flags;
        boolean_t       docow;
        kern_return_t   kret;
        struct pshminfo * pinfo;
        struct pshmnode * pnode;

-       void * mem_object;
+       struct pshmobj * pshmobj;

 #if CONFIG_MACF
        int error;
 #endif
 #if CONFIG_MACF
        int error;
 #endif


@@ -830,7 +880,7 @@ pshm_mmap(__unused proc_t p, struct mmap_args *uap, user_addr_t *retval, struct
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }

-       if ((mem_object =  pinfo->pshm_memobject) == NULL) {
+       if ((pshmobj = pinfo->pshm_memobjects) == NULL) {

                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }
                PSHM_SUBSYS_UNLOCK();
                return(EINVAL);
        }


@@ -848,9 +898,9 @@ pshm_mmap(__unused proc_t p, struct mmap_args *uap, user_addr_t *retval, struct
 
        if ((flags & MAP_FIXED) == 0) {
                alloc_flags = VM_FLAGS_ANYWHERE;
 
        if ((flags & MAP_FIXED) == 0) {
                alloc_flags = VM_FLAGS_ANYWHERE;

-               user_addr = mach_vm_round_page(user_addr); 
+               user_addr = vm_map_round_page(user_addr); 

        } else {
        } else {

-               if (user_addr != mach_vm_trunc_page(user_addr))
+               if (user_addr != vm_map_round_page(user_addr))

                        return (EINVAL);
                /*
                 * We do not get rid of the existing mappings here because
                        return (EINVAL);
                /*
                 * We do not get rid of the existing mappings here because


@@ -862,29 +912,78 @@ pshm_mmap(__unused proc_t p, struct mmap_args *uap, user_addr_t *retval, struct
        }
        docow = FALSE;  
 
        }
        docow = FALSE;  
 

-       kret = vm_map_enter_mem_object(user_map, &user_addr, user_size,
-                                      0, alloc_flags,
-                                      pinfo->pshm_memobject, file_pos, docow,
-                                      prot, VM_PROT_DEFAULT, 
-                                      VM_INHERIT_SHARE);
-       if (kret != KERN_SUCCESS) 
-                       goto out;
-       /* LP64todo - this should be superfluous at this point */
-       kret = mach_vm_inherit(user_map, user_addr, user_size,
-                               VM_INHERIT_SHARE);
+       mapped_size = 0;
+
+       /* reserver the entire space first... */
+       kret = vm_map_enter_mem_object(user_map,
+                                      &user_addr,
+                                      user_size,
+                                      0,
+                                      alloc_flags,
+                                      IPC_PORT_NULL,
+                                      0,
+                                      FALSE,
+                                      VM_PROT_NONE,
+                                      VM_PROT_NONE,
+                                      VM_INHERIT_NONE);
+       user_start_addr = user_addr;

        if (kret != KERN_SUCCESS) {
        if (kret != KERN_SUCCESS) {

-               (void) mach_vm_deallocate(user_map, user_addr, user_size);

                goto out;
        }
                goto out;
        }

+
+       /* ... and overwrite with the real mappings */
+       for (map_pos = 0, pshmobj = pinfo->pshm_memobjects;
+            user_size != 0;
+            map_pos += pshmobj->pshmo_size, pshmobj = pshmobj->pshmo_next) {
+               if (pshmobj == NULL) {
+                       /* nothing there to map !? */
+                       goto out;
+               }
+               if (file_pos >= map_pos + pshmobj->pshmo_size) {
+                       continue;
+               }
+               map_size = pshmobj->pshmo_size - (file_pos - map_pos);
+               if (map_size > user_size) {
+                       map_size = user_size;
+               }
+               kret = vm_map_enter_mem_object(
+                       user_map,
+                       &user_addr,
+                       map_size,
+                       0,
+                       VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE,
+                       pshmobj->pshmo_memobject,
+                       file_pos - map_pos,
+                       docow,
+                       prot,
+                       VM_PROT_DEFAULT, 
+                       VM_INHERIT_SHARE);
+               if (kret != KERN_SUCCESS) 
+                       goto out;
+
+               user_addr += map_size;
+               user_size -= map_size;
+               mapped_size += map_size;
+               file_pos += map_size;
+       }
+

        PSHM_SUBSYS_LOCK();
        PSHM_SUBSYS_LOCK();

-       pnode->mapp_addr = user_addr;
-       pnode->map_size = user_size;
+       pnode->mapp_addr = user_start_addr;
+       pnode->map_size = mapped_size;

        pinfo->pshm_flags |= (PSHM_MAPPED | PSHM_INUSE);
        PSHM_SUBSYS_UNLOCK();
 out:
        pinfo->pshm_flags |= (PSHM_MAPPED | PSHM_INUSE);
        PSHM_SUBSYS_UNLOCK();
 out:

+       if (kret != KERN_SUCCESS) {
+               if (mapped_size != 0) {
+                       (void) mach_vm_deallocate(current_map(),
+                                                 user_start_addr,
+                                                 mapped_size);
+               }
+       }
+

        switch (kret) {
        case KERN_SUCCESS:
        switch (kret) {
        case KERN_SUCCESS:

-               *retval = (user_addr + pageoff);
+               *retval = (user_start_addr + pageoff);

                return (0);
        case KERN_INVALID_ADDRESS:
        case KERN_NO_SPACE:
                return (0);
        case KERN_INVALID_ADDRESS:
        case KERN_NO_SPACE:


@@ -899,7 +998,7 @@ out:
 
 int
 shm_unlink(__unused proc_t p, struct shm_unlink_args *uap, 
 
 int
 shm_unlink(__unused proc_t p, struct shm_unlink_args *uap, 

-                       __unused register_t *retval)
+                       __unused int32_t *retval)

 {
        size_t i;
        int error=0;
 {
        size_t i;
        int error=0;


@@ -911,6 +1010,7 @@ shm_unlink(__unused proc_t p, struct shm_unlink_args *uap,
        size_t pathlen, plen;
        int incache = 0;
        struct pshmcache *pcache = PSHMCACHE_NULL;
        size_t pathlen, plen;
        int incache = 0;
        struct pshmcache *pcache = PSHMCACHE_NULL;

+       struct pshmobj *pshmobj, *pshmobj_next;

 
        pinfo = PSHMINFO_NULL;
 
 
        pinfo = PSHMINFO_NULL;
 


@@ -955,7 +1055,7 @@ shm_unlink(__unused proc_t p, struct shm_unlink_args *uap,
        }
 
        PSHM_SUBSYS_LOCK();
        }
 
        PSHM_SUBSYS_LOCK();

-       error = pshm_cache_search(&pinfo, &nd, &pcache);
+       error = pshm_cache_search(&pinfo, &nd, &pcache, 0);

 
        if (error == ENOENT) {
                PSHM_SUBSYS_UNLOCK();
 
        if (error == ENOENT) {
                PSHM_SUBSYS_UNLOCK();


@@ -976,6 +1076,13 @@ shm_unlink(__unused proc_t p, struct shm_unlink_args *uap,
                goto bad;
        }
 
                goto bad;
        }
 

+       if (pinfo->pshm_flags & PSHM_ALLOCATING) {
+               /* XXX should we wait for flag to clear and then proceed ? */
+               PSHM_SUBSYS_UNLOCK();
+               error = EAGAIN;
+               goto bad;
+       }
+

        if (pinfo->pshm_flags & PSHM_INDELETE) {
                PSHM_SUBSYS_UNLOCK();
                error = 0;
        if (pinfo->pshm_flags & PSHM_INDELETE) {
                PSHM_SUBSYS_UNLOCK();
                error = 0;


@@ -992,29 +1099,42 @@ shm_unlink(__unused proc_t p, struct shm_unlink_args *uap,
        AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid, pinfo->pshm_gid,
                  pinfo->pshm_mode);
 
        AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid, pinfo->pshm_gid,
                  pinfo->pshm_mode);
 

-       /*
-        * JMM - How should permissions be checked?
+       /* 
+        * following file semantics, unlink should be allowed 
+        * for users with write permission only. 

         */
         */

+       if ( (error = pshm_access(pinfo, FWRITE, kauth_cred_get(), p)) ) {
+               PSHM_SUBSYS_UNLOCK();
+               goto bad;
+       }

 
        pinfo->pshm_flags |= PSHM_INDELETE;
        pshm_cache_delete(pcache);
        pinfo->pshm_flags |= PSHM_REMOVED;
        /* release the existence reference */
        if (!--pinfo->pshm_usecount) {
 
        pinfo->pshm_flags |= PSHM_INDELETE;
        pshm_cache_delete(pcache);
        pinfo->pshm_flags |= PSHM_REMOVED;
        /* release the existence reference */
        if (!--pinfo->pshm_usecount) {

+#if CONFIG_MACF
+               mac_posixshm_label_destroy(pinfo);
+#endif

                PSHM_SUBSYS_UNLOCK();
                /*
                 * If this is the last reference going away on the object,
                 * then we need to destroy the backing object.  The name
                 * has an implied but uncounted reference on the object,
                PSHM_SUBSYS_UNLOCK();
                /*
                 * If this is the last reference going away on the object,
                 * then we need to destroy the backing object.  The name
                 * has an implied but uncounted reference on the object,

-                * once it's created, since it's used as a rendesvous, and
+                * once it's created, since it's used as a rendezvous, and

                 * therefore may be subsequently reopened.
                 */
                 * therefore may be subsequently reopened.
                 */

-               if (pinfo->pshm_memobject != NULL)
-                       mach_memory_entry_port_release(pinfo->pshm_memobject);
-               PSHM_SUBSYS_LOCK();
+               for (pshmobj = pinfo->pshm_memobjects;
+                    pshmobj != NULL;
+                    pshmobj = pshmobj_next) {
+                       mach_memory_entry_port_release(pshmobj->pshmo_memobject);
+                       pshmobj_next = pshmobj->pshmo_next;
+                       FREE(pshmobj, M_SHM);
+               }

                FREE(pinfo,M_SHM);
                FREE(pinfo,M_SHM);

+       } else {
+               PSHM_SUBSYS_UNLOCK();

        }
        }

-       PSHM_SUBSYS_UNLOCK();

        FREE(pcache, M_SHM);
        error = 0;
 bad:
        FREE(pcache, M_SHM);
        error = 0;
 bad:


@@ -1024,15 +1144,16 @@ bad:
 
 /* already called locked */
 static int
 
 /* already called locked */
 static int

-pshm_close(struct pshmnode *pnode)
+pshm_close(struct pshminfo *pinfo, int dropref)

 {
 {

-       int error=0;
-       struct pshminfo *pinfo;
+       int error = 0;
+       struct pshmobj *pshmobj, *pshmobj_next;

 
 

-       if ((pinfo = pnode->pinfo) == PSHMINFO_NULL)
-               return(EINVAL);
-
-       if ((pinfo->pshm_flags & PSHM_ALLOCATED) != PSHM_ALLOCATED) {
+       /*
+        * If we are dropping the reference we took on the cache object, don't
+        * enforce the allocation requirement.
+        */
+       if ( !dropref && ((pinfo->pshm_flags & PSHM_ALLOCATED) != PSHM_ALLOCATED)) {

                return(EINVAL);
        }
 #if DIAGNOSTIC
                return(EINVAL);
        }
 #if DIAGNOSTIC


@@ -1043,20 +1164,24 @@ pshm_close(struct pshmnode *pnode)
        pinfo->pshm_usecount--; /* release this fd's reference */
 
        if ((pinfo->pshm_flags & PSHM_REMOVED) && !pinfo->pshm_usecount) {
        pinfo->pshm_usecount--; /* release this fd's reference */
 
        if ((pinfo->pshm_flags & PSHM_REMOVED) && !pinfo->pshm_usecount) {

+#if CONFIG_MACF
+               mac_posixshm_label_destroy(pinfo);
+#endif

                PSHM_SUBSYS_UNLOCK();
                /*
                 * If this is the last reference going away on the object,
                 * then we need to destroy the backing object.
                 */
                PSHM_SUBSYS_UNLOCK();
                /*
                 * If this is the last reference going away on the object,
                 * then we need to destroy the backing object.
                 */

-               if (pinfo->pshm_memobject != NULL)
-                       mach_memory_entry_port_release(pinfo->pshm_memobject);
+               for (pshmobj = pinfo->pshm_memobjects;
+                    pshmobj != NULL;
+                    pshmobj = pshmobj_next) {
+                       mach_memory_entry_port_release(pshmobj->pshmo_memobject);
+                       pshmobj_next = pshmobj->pshmo_next;
+                       FREE(pshmobj, M_SHM);
+               }

                PSHM_SUBSYS_LOCK();
                PSHM_SUBSYS_LOCK();

-#if CONFIG_MACF
-               mac_posixshm_label_destroy(pinfo);
-#endif

                FREE(pinfo,M_SHM);
        }
                FREE(pinfo,M_SHM);
        }

-       FREE(pnode, M_SHM);

        return (error);
 }
 
        return (error);
 }
 


@@ -1064,11 +1189,20 @@ pshm_close(struct pshmnode *pnode)
 static int
 pshm_closefile(struct fileglob *fg, __unused vfs_context_t ctx)
 {
 static int
 pshm_closefile(struct fileglob *fg, __unused vfs_context_t ctx)
 {

-       int error;
+       int error = EINVAL;
+       struct pshmnode *pnode;

 
        PSHM_SUBSYS_LOCK();
 
        PSHM_SUBSYS_LOCK();

-       error =  pshm_close(((struct pshmnode *)fg->fg_data));
+
+       if ((pnode = (struct pshmnode *)fg->fg_data) != NULL) {
+               if (pnode->pinfo != PSHMINFO_NULL) {
+                       error =  pshm_close(pnode->pinfo, 0);
+               }
+               FREE(pnode, M_SHM);
+       }
+

        PSHM_SUBSYS_UNLOCK();
        PSHM_SUBSYS_UNLOCK();

+

        return(error);
 }
 
        return(error);
 }