/*
- * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
*
* @(#)vfs_init.c 8.5 (Berkeley) 5/11/95
*/
+/*
+ * NOTICE: This file was modified by SPARTA, Inc. in 2005 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.
+ */
#include <sys/param.h>
#include <sys/ucred.h>
#include <sys/errno.h>
#include <sys/malloc.h>
+#include <sys/decmpfs.h>
+#if CONFIG_MACF
+#include <security/mac_framework.h>
+#include <sys/kauth.h>
+#endif
+#if QUOTA
+#include <sys/quota.h>
+#endif
/*
* Sigh, such primitive tools are these...
#define DODEBUG(A)
#endif
-extern uid_t console_user;
+__private_extern__ void vntblinit(void);
+
extern struct vnodeopv_desc *vfs_opv_descs[];
/* a list of lists of vnodeops defns */
extern struct vnodeop_desc *vfs_op_descs[];
*/
int vfs_opv_numops;
-typedef (*PFI)(); /* the standard Pointer to a Function returning an Int */
+typedef int (*PFIvp)(void *);
/*
* A miscellaneous routine.
* A generic "default" routine that just returns an error.
*/
int
-vn_default_error()
+vn_default_error(void)
{
return (ENOTSUP);
* that is a(whole)nother story.) This is a feature.
*/
void
-vfs_opv_init()
+vfs_opv_init(void)
{
int i, j, k;
int (***opv_desc_vector_p)(void *);
* Also handle backwards compatibility.
*/
if (*opv_desc_vector_p == NULL) {
- MALLOC(*opv_desc_vector_p, PFI*,
- vfs_opv_numops*sizeof(PFI), M_TEMP, M_WAITOK);
- bzero (*opv_desc_vector_p, vfs_opv_numops*sizeof(PFI));
+ MALLOC(*opv_desc_vector_p, PFIvp*,
+ vfs_opv_numops*sizeof(PFIvp), M_TEMP, M_WAITOK);
+ bzero (*opv_desc_vector_p, vfs_opv_numops*sizeof(PFIvp));
DODEBUG(printf("vector at %x allocated\n",
opv_desc_vector_p));
}
for (j=0; vfs_opv_descs[i]->opv_desc_ops[j].opve_op; j++) {
opve_descp = &(vfs_opv_descs[i]->opv_desc_ops[j]);
+ /* Silently skip known-disabled operations */
+ if (opve_descp->opve_op->vdesc_flags & VDESC_DISABLED) {
+ printf("vfs_fsadd: Ignoring reference in %p to disabled operation %s.\n",
+ vfs_opv_descs[i], opve_descp->opve_op->vdesc_name);
+ continue;
+ }
+
/*
* Sanity check: is this operation listed
* in the list of operations? We check this
* list of supported operations.
*/
if (opve_descp->opve_op->vdesc_offset == 0 &&
- opve_descp->opve_op->vdesc_offset !=
- VOFFSET(vnop_default)) {
+ opve_descp->opve_op !=
+ VDESC(vnop_default)) {
printf("operation %s not listed in %s.\n",
opve_descp->opve_op->vdesc_name,
"vfs_op_descs");
* Initialize known vnode operations vectors.
*/
void
-vfs_op_init()
+vfs_op_init(void)
{
int i;
* and assign each its offset.
*/
for (vfs_opv_numops = 0, i = 0; vfs_op_descs[i]; i++) {
+ /* Silently skip known-disabled operations */
+ if (vfs_op_descs[i]->vdesc_flags & VDESC_DISABLED) {
+ continue;
+ }
vfs_op_descs[i]->vdesc_offset = vfs_opv_numops;
vfs_opv_numops++;
}
lck_grp_attr_t * vnode_lck_grp_attr;
lck_attr_t * vnode_lck_attr;
+#if CONFIG_TRIGGERS
+/* vars for vnode trigger resolver */
+lck_grp_t * trigger_vnode_lck_grp;
+lck_grp_attr_t * trigger_vnode_lck_grp_attr;
+lck_attr_t * trigger_vnode_lck_attr;
+#endif
+
+lck_grp_t * fd_vn_lck_grp;
+lck_grp_attr_t * fd_vn_lck_grp_attr;
+lck_attr_t * fd_vn_lck_attr;
/* vars for vnode list lock */
lck_grp_t * vnode_list_lck_grp;
lck_grp_attr_t * vnode_list_lck_grp_attr;
lck_attr_t * vnode_list_lck_attr;
-lck_mtx_t * vnode_list_mtx_lock;
+lck_spin_t * vnode_list_spin_lock;
lck_mtx_t * spechash_mtx_lock;
-/* Routine to lock and unlock the vnode lists */
-void vnode_list_lock(void);
-void vnode_list_unlock(void);
/* vars for vfsconf lock */
lck_grp_t * fsconf_lck_grp;
lck_attr_t * mnt_list_lck_attr;
lck_mtx_t * mnt_list_mtx_lock;
-extern void journal_init();
+/* vars for sync mutex */
+lck_grp_t * sync_mtx_lck_grp;
+lck_grp_attr_t * sync_mtx_lck_grp_attr;
+lck_attr_t * sync_mtx_lck_attr;
+lck_mtx_t * sync_mtx_lck;
+
+lck_mtx_t *pkg_extensions_lck;
struct mount * dead_mountp;
+
+extern void nspace_handler_init(void);
+
/*
* Initialize the vnode structures and initialize each file system type.
*/
void
-vfsinit()
+vfsinit(void)
{
struct vfstable *vfsp;
int i, maxtypenum;
vnode_list_lck_attr = lck_attr_alloc_init();
/* Allocate vnode list lock */
- vnode_list_mtx_lock = lck_mtx_alloc_init(vnode_list_lck_grp, vnode_list_lck_attr);
+ vnode_list_spin_lock = lck_spin_alloc_init(vnode_list_lck_grp, vnode_list_lck_attr);
/* Allocate spec hash list lock */
spechash_mtx_lock = lck_mtx_alloc_init(vnode_list_lck_grp, vnode_list_lck_attr);
+ /* Allocate the package extensions table lock */
+ pkg_extensions_lck = lck_mtx_alloc_init(vnode_list_lck_grp, vnode_list_lck_attr);
+
/* allocate vnode lock group attribute and group */
vnode_lck_grp_attr= lck_grp_attr_alloc_init();
/* Allocate vnode lock attribute */
vnode_lck_attr = lck_attr_alloc_init();
+#if CONFIG_TRIGGERS
+ trigger_vnode_lck_grp_attr = lck_grp_attr_alloc_init();
+ trigger_vnode_lck_grp = lck_grp_alloc_init("trigger_vnode", trigger_vnode_lck_grp_attr);
+ trigger_vnode_lck_attr = lck_attr_alloc_init();
+#endif
+ /* Allocate per fd vnode data lock attribute and group */
+ fd_vn_lck_grp_attr = lck_grp_attr_alloc_init();
+ fd_vn_lck_grp = lck_grp_alloc_init("fd_vnode_data", fd_vn_lck_grp_attr);
+ fd_vn_lck_attr = lck_attr_alloc_init();
+
/* Allocate fs config lock group attribute and group */
fsconf_lck_grp_attr= lck_grp_attr_alloc_init();
/* Allocate mount lock attribute */
mnt_lck_attr = lck_attr_alloc_init();
- /*
- * Initialize the "console user" for access purposes:
- */
- console_user = (uid_t)0;
-
+ /* Allocate sync lock */
+ sync_mtx_lck_grp_attr = lck_grp_attr_alloc_init();
+ sync_mtx_lck_grp = lck_grp_alloc_init("sync thread", sync_mtx_lck_grp_attr);
+ sync_mtx_lck_attr = lck_attr_alloc_init();
+ sync_mtx_lck = lck_mtx_alloc_init(sync_mtx_lck_grp, sync_mtx_lck_attr);
+
/*
* Initialize the vnode table
*/
* Initialize the vnode name cache
*/
nchinit();
- /*
- * Initialize the journaling locks
- */
- journal_init();
+
+ nspace_handler_init();
+
/*
* Build vnode operation vectors.
*/
* Initialize each file system type in the static list,
* until the first NULL ->vfs_vfsops is encountered.
*/
- numused_vfsslots = maxtypenum = 0;
- for (vfsp = vfsconf, i = 0; i < maxvfsconf; i++, vfsp++) {
+ maxtypenum = VT_NON;
+ for (vfsp = vfsconf, i = 0; i < maxvfsslots; i++, vfsp++) {
+ struct vfsconf vfsc;
if (vfsp->vfc_vfsops == (struct vfsops *)0)
break;
if (i) vfsconf[i-1].vfc_next = vfsp;
if (maxtypenum <= vfsp->vfc_typenum)
maxtypenum = vfsp->vfc_typenum + 1;
- (*vfsp->vfc_vfsops->vfs_init)(vfsp);
- lck_mtx_init(&vfsp->vfc_lock, fsconf_lck_grp, fsconf_lck_attr);
+ bzero(&vfsc, sizeof(struct vfsconf));
+ vfsc.vfc_reserved1 = 0;
+ bcopy(vfsp->vfc_name, vfsc.vfc_name, sizeof(vfsc.vfc_name));
+ vfsc.vfc_typenum = vfsp->vfc_typenum;
+ vfsc.vfc_refcount = vfsp->vfc_refcount;
+ vfsc.vfc_flags = vfsp->vfc_flags;
+ vfsc.vfc_reserved2 = 0;
+ vfsc.vfc_reserved3 = 0;
+
+ if (vfsp->vfc_vfsops->vfs_sysctl) {
+ struct sysctl_oid *oidp = NULL;
+ struct sysctl_oid oid = SYSCTL_STRUCT_INIT(_vfs, vfsp->vfc_typenum, , CTLTYPE_NODE | CTLFLAG_KERN | CTLFLAG_RW | CTLFLAG_LOCKED, NULL, 0, vfs_sysctl_node, "-", "");
+
+ MALLOC(oidp, struct sysctl_oid *, sizeof(struct sysctl_oid), M_TEMP, M_WAITOK);
+ *oidp = oid;
+
+ /* Memory for VFS oid held by vfsentry forever */
+ vfsp->vfc_sysctl = oidp;
+ oidp->oid_name = vfsp->vfc_name;
+ sysctl_register_oid(vfsp->vfc_sysctl);
+ }
+ (*vfsp->vfc_vfsops->vfs_init)(&vfsc);
+
numused_vfsslots++;
+ numregistered_fses++;
}
/* next vfc_typenum to be used */
- maxvfsconf = maxtypenum;
+ maxvfstypenum = maxtypenum;
/*
* Initialize the vnop authorization scope.
*/
vnode_authorize_init();
+
+ /*
+ * Initialiize the quota system.
+ */
+#if QUOTA
+ dqinit();
+#endif
/*
* create a mount point for dead vnodes
*/
- MALLOC_ZONE(mp, struct mount *, (u_long)sizeof(struct mount),
+ MALLOC_ZONE(mp, struct mount *, sizeof(struct mount),
M_MOUNT, M_WAITOK);
- bzero((char *)mp, (u_long)sizeof(struct mount));
+ bzero((char *)mp, sizeof(struct mount));
/* Initialize the default IO constraints */
mp->mnt_maxreadcnt = mp->mnt_maxwritecnt = MAXPHYS;
mp->mnt_segreadcnt = mp->mnt_segwritecnt = 32;
mp->mnt_maxsegreadsize = mp->mnt_maxreadcnt;
mp->mnt_maxsegwritesize = mp->mnt_maxwritecnt;
mp->mnt_devblocksize = DEV_BSIZE;
+ mp->mnt_alignmentmask = PAGE_MASK;
+ mp->mnt_ioqueue_depth = MNT_DEFAULT_IOQUEUE_DEPTH;
+ mp->mnt_ioscale = 1;
+ mp->mnt_ioflags = 0;
+ mp->mnt_realrootvp = NULLVP;
+ mp->mnt_authcache_ttl = CACHED_LOOKUP_RIGHT_TTL;
TAILQ_INIT(&mp->mnt_vnodelist);
TAILQ_INIT(&mp->mnt_workerqueue);
mp->mnt_flag = MNT_LOCAL;
mp->mnt_lflag = MNT_LDEAD;
mount_lock_init(mp);
+
+#if CONFIG_MACF
+ mac_mount_label_init(mp);
+ mac_mount_label_associate(vfs_context_kernel(), mp);
+#endif
dead_mountp = mp;
+
+#if FS_COMPRESSION
+ decmpfs_init();
+#endif
}
void
-vnode_list_lock()
+vnode_list_lock(void)
{
- lck_mtx_lock(vnode_list_mtx_lock);
+ lck_spin_lock(vnode_list_spin_lock);
}
void
-vnode_list_unlock()
+vnode_list_unlock(void)
{
- lck_mtx_unlock(vnode_list_mtx_lock);
+ lck_spin_unlock(vnode_list_spin_lock);
}
void
-mount_list_lock()
+mount_list_lock(void)
{
lck_mtx_lock(mnt_list_mtx_lock);
}
void
-mount_list_unlock()
+mount_list_unlock(void)
{
lck_mtx_unlock(mnt_list_mtx_lock);
}
mount_lock_init(mount_t mp)
{
lck_mtx_init(&mp->mnt_mlock, mnt_lck_grp, mnt_lck_attr);
+ lck_mtx_init(&mp->mnt_iter_lock, mnt_lck_grp, mnt_lck_attr);
lck_mtx_init(&mp->mnt_renamelock, mnt_lck_grp, mnt_lck_attr);
lck_rw_init(&mp->mnt_rwlock, mnt_lck_grp, mnt_lck_attr);
}
mount_lock_destroy(mount_t mp)
{
lck_mtx_destroy(&mp->mnt_mlock, mnt_lck_grp);
+ lck_mtx_destroy(&mp->mnt_iter_lock, mnt_lck_grp);
lck_mtx_destroy(&mp->mnt_renamelock, mnt_lck_grp);
lck_rw_destroy(&mp->mnt_rwlock, mnt_lck_grp);
}
vfstable_add(struct vfstable *nvfsp)
{
int slot;
- struct vfstable *slotp;
+ struct vfstable *slotp, *allocated = NULL;
+ struct sysctl_oid *oidp = NULL;
+
+ if (nvfsp->vfc_vfsops->vfs_sysctl) {
+ struct sysctl_oid oid = SYSCTL_STRUCT_INIT(_vfs, nvfsp->vfc_typenum, , CTLTYPE_NODE | CTLFLAG_KERN | CTLFLAG_RW | CTLFLAG_LOCKED, NULL, 0, vfs_sysctl_node, "-", "");
+
+ MALLOC(oidp, struct sysctl_oid *, sizeof(struct sysctl_oid), M_TEMP, M_WAITOK);
+ *oidp = oid;
+ }
+
/*
* Find the next empty slot; we recognize an empty slot by a
* NULL-valued ->vfc_vfsops, so if we delete a VFS, we must
* ensure we set the entry back to NULL.
*/
+findslot:
+ mount_list_lock();
for (slot = 0; slot < maxvfsslots; slot++) {
if (vfsconf[slot].vfc_vfsops == NULL)
break;
}
if (slot == maxvfsslots) {
- /* out of static slots; allocate one instead */
- MALLOC(slotp, struct vfstable *, sizeof(struct vfstable),
- M_TEMP, M_WAITOK);
+ if (allocated == NULL) {
+ mount_list_unlock();
+ /* out of static slots; allocate one instead */
+ MALLOC(allocated, struct vfstable *, sizeof(struct vfstable),
+ M_TEMP, M_WAITOK);
+ goto findslot;
+ } else {
+ slotp = allocated;
+ }
} else {
slotp = &vfsconf[slot];
}
* with the value of 'maxvfslots' in the allocation case.
*/
bcopy(nvfsp, slotp, sizeof(struct vfstable));
- lck_mtx_init(&slotp->vfc_lock, fsconf_lck_grp, fsconf_lck_attr);
if (slot != 0) {
slotp->vfc_next = vfsconf[slot - 1].vfc_next;
vfsconf[slot - 1].vfc_next = slotp;
} else {
slotp->vfc_next = NULL;
}
- numused_vfsslots++;
+
+ if (slotp != allocated) {
+ /* used a statically allocated slot */
+ numused_vfsslots++;
+ }
+ numregistered_fses++;
+
+ if (oidp) {
+ /* Memory freed in vfstable_del after unregistration */
+ slotp->vfc_sysctl = oidp;
+ oidp->oid_name = slotp->vfc_name;
+ sysctl_register_oid(slotp->vfc_sysctl);
+ }
+
+ mount_list_unlock();
+
+ if (allocated && allocated != slotp) {
+ /* did allocation, but ended up using static slot */
+ FREE(allocated, M_TEMP);
+ }
return(slotp);
}
struct vfstable **vcpp;
struct vfstable *vcdelp;
+#if DEBUG
+ lck_mtx_assert(mnt_list_mtx_lock, LCK_MTX_ASSERT_OWNED);
+#endif /* DEBUG */
+
/*
* Traverse the list looking for vtbl; if found, *vcpp
* will contain the address of the pointer to the entry to
*/
for( vcpp = &vfsconf; *vcpp; vcpp = &(*vcpp)->vfc_next) {
if (*vcpp == vtbl)
- break;
- }
+ break;
+ }
if (*vcpp == NULL)
return(ESRCH); /* vtbl not on vfsconf list */
+ if ((*vcpp)->vfc_sysctl) {
+ sysctl_unregister_oid((*vcpp)->vfc_sysctl);
+ (*vcpp)->vfc_sysctl->oid_name = NULL;
+ FREE((*vcpp)->vfc_sysctl, M_TEMP);
+ (*vcpp)->vfc_sysctl = NULL;
+ }
+
/* Unlink entry */
vcdelp = *vcpp;
*vcpp = (*vcpp)->vfc_next;
- lck_mtx_destroy(&vcdelp->vfc_lock, fsconf_lck_grp);
-
/*
* Is this an entry from our static table? We find out by
* seeing if the pointer to the object to be deleted places
if (vcdelp >= vfsconf && vcdelp < (vfsconf + maxvfsslots)) { /* Y */
/* Mark as empty for vfscon_add() */
bzero(vcdelp, sizeof(struct vfstable));
+ numregistered_fses--;
numused_vfsslots--;
} else { /* N */
/*
* vfsconf onto our list, but it may not be persistent
* because of the previous (copying) implementation.
*/
- FREE(vcdelp, M_TEMP);
+ numregistered_fses--;
+ mount_list_unlock();
+ FREE(vcdelp, M_TEMP);
+ mount_list_lock();
}
+#if DEBUG
+ lck_mtx_assert(mnt_list_mtx_lock, LCK_MTX_ASSERT_OWNED);
+#endif /* DEBUG */
+
return(0);
}