X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/378393581903b274cb7a4d18e0d978071a6b592d..eee3565979933af707c711411001ba11fe406a3c:/bsd/sys/vnode_internal.h diff --git a/bsd/sys/vnode_internal.h b/bsd/sys/vnode_internal.h index df02742df..22750b6b9 100644 --- a/bsd/sys/vnode_internal.h +++ b/bsd/sys/vnode_internal.h @@ -1,23 +1,29 @@ /* - * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2012 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. + * 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. 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. * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * 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 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. * - * @APPLE_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */ /* @@ -54,12 +60,16 @@ * * @(#)vnode.h 8.17 (Berkeley) 5/20/95 */ - +/* + * 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. + */ + #ifndef _SYS_VNODE_INTERNAL_H_ #define _SYS_VNODE_INTERNAL_H_ -#define INTERIM_FSNODE_LOCK 1 - #include #include #include @@ -75,18 +85,36 @@ #include #include #include +#include struct lockf; +struct label; LIST_HEAD(buflists, buf); - -struct unsafe_fsnode { - lck_mtx_t fsnodelock; - int32_t fsnode_count; - void * fsnodeowner; +#if CONFIG_TRIGGERS +/* + * VFS Internal (private) trigger vnode resolver info. + */ +struct vnode_resolve { + lck_mtx_t vr_lock; /* protects vnode_resolve_t fields */ + trigger_vnode_resolve_callback_t vr_resolve_func; + trigger_vnode_unresolve_callback_t vr_unresolve_func; + trigger_vnode_rearm_callback_t vr_rearm_func; + trigger_vnode_reclaim_callback_t vr_reclaim_func; + void * vr_data; /* private data for resolver */ + uint32_t vr_flags; + uint32_t vr_lastseq; }; +typedef struct vnode_resolve *vnode_resolve_t; + +/* private vr_flags */ +#define VNT_RESOLVED (1UL << 31) +#define VNT_VFS_UNMOUNTED (1UL << 30) +#define VNT_EXTERNAL (1UL << 29) + +#endif /* CONFIG_TRIGGERS */ /* * Reading or writing any of these items requires holding the appropriate lock. @@ -100,19 +128,21 @@ struct vnode { lck_mtx_t v_lock; /* vnode mutex */ TAILQ_ENTRY(vnode) v_freelist; /* vnode freelist */ TAILQ_ENTRY(vnode) v_mntvnodes; /* vnodes for mount point */ + TAILQ_HEAD(, namecache) v_ncchildren; /* name cache entries that regard us as their parent */ LIST_HEAD(, namecache) v_nclinks; /* name cache entries that name this vnode */ - LIST_HEAD(, namecache) v_ncchildren; /* name cache entries that regard us as there parent */ vnode_t v_defer_reclaimlist; /* in case we have to defer the reclaim to avoid recursion */ - u_long v_flag; /* vnode flags (see below) */ - u_short v_lflag; /* vnode local and named ref flags */ - u_char v_iterblkflags; /* buf iterator flags */ - u_char v_references; /* number of times io_count has been granted */ + uint32_t v_listflag; /* flags protected by the vnode_list_lock (see below) */ + uint32_t v_flag; /* vnode flags (see below) */ + uint16_t v_lflag; /* vnode local and named ref flags */ + uint8_t v_iterblkflags; /* buf iterator flags */ + uint8_t v_references; /* number of times io_count has been granted */ int32_t v_kusecount; /* count of in-kernel refs */ int32_t v_usecount; /* reference count of users */ int32_t v_iocount; /* iocounters */ void * v_owner; /* act that owns the vnode */ - enum vtype v_type; /* vnode type */ - u_long v_id; /* identity of vnode contents */ + uint16_t v_type; /* vnode type */ + uint16_t v_tag; /* type of underlying data */ + uint32_t v_id; /* identity of vnode contents */ union { struct mount *vu_mountedhere;/* ptr to mounted vfs (VDIR) */ struct socket *vu_socket; /* unix ipc (VSOCK) */ @@ -122,20 +152,33 @@ struct vnode { } v_un; struct buflists v_cleanblkhd; /* clean blocklist head */ struct buflists v_dirtyblkhd; /* dirty blocklist head */ - kauth_cred_t v_cred; - int v_cred_timestamp; - long v_numoutput; /* num of writes in progress */ - long v_writecount; /* reference count of writers */ - char * v_name; /* name component of the vnode */ + struct klist v_knotes; /* knotes attached to this vnode */ + /* + * the following 4 fields are protected + * by the name_cache_lock held in + * excluive mode + */ + kauth_cred_t v_cred; /* last authorized credential */ + kauth_action_t v_authorized_actions; /* current authorized actions for v_cred */ + int v_cred_timestamp; /* determine if entry is stale for MNTK_AUTH_OPAQUE */ + int v_nc_generation; /* changes when nodes are removed from the name cache */ + /* + * back to the vnode lock for protection + */ + int32_t v_numoutput; /* num of writes in progress */ + int32_t v_writecount; /* reference count of writers */ + const char *v_name; /* name component of the vnode */ vnode_t v_parent; /* pointer to parent vnode */ -#ifdef INTERIM_FSNODE_LOCK struct lockf *v_lockf; /* advisory lock list head */ - struct unsafe_fsnode *v_unsafefs; /* pointer to struct used to lock */ -#endif /* vnodes on unsafe filesystems */ int (**v_op)(void *); /* vnode operations vector */ - enum vtagtype v_tag; /* type of underlying data */ mount_t v_mount; /* ptr to vfs we are in */ void * v_data; /* private data for fs */ +#if CONFIG_MACF + struct label *v_label; /* MAC security label */ +#endif +#if CONFIG_TRIGGERS + vnode_resolve_t v_resolve; /* trigger vnode resolve info (VDIR only) */ +#endif /* CONFIG_TRIGGERS */ }; #define v_mountedhere v_un.vu_mountedhere @@ -154,24 +197,34 @@ struct vnode { #define VBI_DIRTY 0x8 #define VBI_NEWBUF 0x10 +/* + * v_listflag + */ +#define VLIST_RAGE 0x01 /* vnode is currently in the rapid age list */ +#define VLIST_DEAD 0x02 /* vnode is currently in the dead list */ +#define VLIST_ASYNC_WORK 0x04 /* vnode is currently on the deferred async work queue */ /* * v_lflags */ #define VL_SUSPENDED 0x0001 /* vnode is suspended */ #define VL_DRAIN 0x0002 /* vnode is being drained */ -#define VL_TERMINATE 0x0004 /* vnode is marked for termination */ -#define VL_TERMWANT 0x0008 /* vnode is marked for termination */ -#define VL_DEAD 0x0010 /* vnode is dead and completed recycle */ -#define VL_MARKTERM 0x0020 /* vnode is dead and completed recycle */ -#define VL_MOUNTDEAD 0x0040 /* v_moutnedhere is dead */ +#define VL_TERMINATE 0x0004 /* vnode is in the process of being recycled */ +#define VL_TERMWANT 0x0008 /* there's a waiter for recycle finish (vnode_getiocount)*/ +#define VL_DEAD 0x0010 /* vnode is dead, cleaned of filesystem-specific info */ +#define VL_MARKTERM 0x0020 /* vnode should be recycled when no longer referenced */ #define VL_NEEDINACTIVE 0x0080 /* delay VNOP_INACTIVE until iocount goes to 0 */ +#define VL_LABEL 0x0100 /* vnode is marked for labeling */ +#define VL_LABELWAIT 0x0200 /* vnode is marked for labeling */ +#define VL_LABELED 0x0400 /* vnode is labeled */ +#define VL_LWARNED 0x0800 +#define VL_HASSTREAMS 0x1000 /* vnode has had at least one associated named stream vnode (may not have one any longer) */ + #define VNAMED_UBC 0x2000 /* ubc named reference */ #define VNAMED_MOUNT 0x4000 /* mount point named reference */ #define VNAMED_FSHASH 0x8000 /* FS hash named reference */ - /* * v_flags */ @@ -179,33 +232,82 @@ struct vnode { #define VTEXT 0x000002 /* vnode is a pure text prototype */ #define VSYSTEM 0x000004 /* vnode being used by kernel */ #define VISTTY 0x000008 /* vnode represents a tty */ -#define VWASMAPPED 0x000010 /* vnode was mapped before */ -#define VTERMINATE 0x000020 /* terminating memory object */ -#define VTERMWANT 0x000040 /* wating for memory object death */ +#define VRAGE 0x000010 /* vnode is in rapid age state */ +#define VBDEVVP 0x000020 /* vnode created by bdevvp */ +#define VDEVFLUSH 0x000040 /* device vnode after vflush */ #define VMOUNT 0x000080 /* mount operation in progress */ #define VBWAIT 0x000100 /* waiting for output to complete */ -#define VALIASED 0x000200 /* vnode has an alias */ +#define VSHARED_DYLD 0x000200 /* vnode is a dyld shared cache file */ #define VNOCACHE_DATA 0x000400 /* don't keep data cached once it's been consumed */ #define VSTANDARD 0x000800 /* vnode obtained from common pool */ #define VAGE 0x001000 /* Insert vnode at head of free list */ #define VRAOFF 0x002000 /* read ahead disabled */ #define VNCACHEABLE 0x004000 /* vnode is allowed to be put back in name cache */ -#define VUINACTIVE 0x008000 /* UBC vnode is on inactive list */ +#if NAMEDSTREAMS +#define VISSHADOW 0x008000 /* vnode is a shadow file */ +#endif #define VSWAP 0x010000 /* vnode is being used as swapfile */ #define VTHROTTLED 0x020000 /* writes or pageouts have been throttled */ /* wakeup tasks waiting when count falls below threshold */ #define VNOFLUSH 0x040000 /* don't vflush() if SKIPSYSTEM */ #define VLOCKLOCAL 0x080000 /* this vnode does adv locking in vfs */ #define VISHARDLINK 0x100000 /* hard link needs special processing on lookup and in volfs */ +#define VISUNION 0x200000 /* union special processing */ +#if NAMEDSTREAMS +#define VISNAMEDSTREAM 0x400000 /* vnode is a named stream (eg HFS resource fork) */ +#endif +#define VOPENEVT 0x800000 /* if process is P_CHECKOPENEVT, then or in the O_EVTONLY flag on open */ +#define VNEEDSSNAPSHOT 0x1000000 +#define VNOCS 0x2000000 /* is there no code signature available */ +#define VISDIRTY 0x4000000 /* vnode will need IO if reclaimed */ +#define VFASTDEVCANDIDATE 0x8000000 /* vnode is a candidate to store on a fast device */ +#define VAUTOCANDIDATE 0x10000000 /* vnode was automatically marked as a fast-dev candidate */ +/* + 0x20000000 not used + 0x40000000 not used + 0x80000000 not used. +*/ -#define VCRED_EXPIRED 2 /* number of seconds to keep cached credential valid */ +/* + * This structure describes vnode data which is specific to a file descriptor. + * It is currently used only for file descriptors which are for opened for + * directories. + */ +struct fd_vn_data { + lck_mtx_t fv_lock; /* Used to serialize directory enumeration on fd */ + off_t fv_offset; /* Offset to be used */ + void *fv_dircookie; /* If FS doesn't like offsets in directories */ + caddr_t fv_buf; /* Temporary buffer to store directory entries */ + size_t fv_bufsiz; /* Valid size of fv_buf */ + size_t fv_bufdone; /* How much of fv_buf is processed */ + size_t fv_bufallocsiz; /* Allocation size determined for Buffer*/ + off_t fv_soff; /* Starting FS offset for this buffer */ + off_t fv_eoff; /* Ending FS offset for this buffer */ + int fv_eofflag;/* Does fv_eoff represent EOF ? */ +}; +/* + * FV_DIRBUF_START_SIZ is the initial size of the buffer passed to VNOP_READDIR. + * That may not be enough for some filesytems so the current algorithm works its + * way upto FV_DIRBUF_MAX_SIZ + */ +#define FV_DIRBUF_DIRENTRY_SIZ (sizeof(struct direntry)) +#define FV_DIRBUF_START_SIZ FV_DIRBUF_DIRENTRY_SIZ +#define FV_DIRBUF_MAX_SIZ (4*(sizeof(struct direntry))) + +#define FV_LOCK(fvd) lck_mtx_lock(&(((struct fd_vn_data *)fvd)->fv_lock)) +#define FV_UNLOCK(fvd) lck_mtx_unlock(&(((struct fd_vn_data *)fvd)->fv_lock)) /* * Global vnode data. */ extern struct vnode *rootvnode; /* root (i.e. "/") vnode */ +#ifdef CONFIG_IMGSRC_ACCESS +#define MAX_IMAGEBOOT_NESTING 2 +extern struct vnode *imgsrc_rootvnodes[]; +#endif /* CONFIG_IMGSRC_ACCESS */ + /* * Mods for exensibility. @@ -234,7 +336,7 @@ extern struct vnode *rootvnode; /* root (i.e. "/") vnode */ */ struct vnodeop_desc { int vdesc_offset; /* offset in vector--first for speed */ - char *vdesc_name; /* a readable name for debugging */ + const char *vdesc_name; /* a readable name for debugging */ int vdesc_flags; /* VDESC_* flags */ /* @@ -262,10 +364,6 @@ struct vnodeop_desc { */ extern struct vnodeop_desc *vnodeop_descs[]; -/* - * Interlock for scanning list of vnodes attached to a mountpoint - */ -extern void * mntvnode_slock; /* * This macro is very helpful in defining those offsets in the vdesc struct. @@ -297,74 +395,208 @@ extern void * mntvnode_slock; #define VDESC(OP) (& __CONCAT(OP,_desc)) #define VOFFSET(OP) (VDESC(OP)->vdesc_offset) +struct ostat; - -int build_path(vnode_t first_vp, char *buff, int buflen, int *outlen); -int bdevvp(dev_t dev, struct vnode **vpp); +/* bdevvp moved to vnode.h as private KPI */ void cvtstat(struct stat *st, struct ostat *ost); void vprint(const char *label, struct vnode *vp); -__private_extern__ int is_package_name(char *name, int len); -__private_extern__ int set_package_extensions_table(void *data, int nentries, int maxwidth); -int vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, - int len, off_t offset, enum uio_seg segflg, int ioflg, - struct ucred *cred, int *aresid, struct proc *p); +__private_extern__ int set_package_extensions_table(user_addr_t data, int nentries, int maxwidth); int vn_rdwr_64(enum uio_rw rw, struct vnode *vp, uint64_t base, int64_t len, off_t offset, enum uio_seg segflg, - int ioflg, struct ucred *cred, int *aresid, + int ioflg, kauth_cred_t cred, int64_t *aresid, struct proc *p); +#if CONFIG_MACF +int vn_setlabel (struct vnode *vp, struct label *intlabel, + vfs_context_t context); +#endif void fifo_printinfo(struct vnode *vp); -int vn_lock(struct vnode *vp, int flags, struct proc *p); int vn_open(struct nameidata *ndp, int fmode, int cmode); int vn_open_modflags(struct nameidata *ndp, int *fmode, int cmode); int vn_open_auth(struct nameidata *ndp, int *fmode, struct vnode_attr *); -int vn_close(vnode_t, int flags, struct ucred *cred, struct proc *p); +int vn_close(vnode_t, int flags, vfs_context_t ctx); +errno_t vn_remove(vnode_t dvp, vnode_t *vpp, struct nameidata *ndp, int32_t flags, struct vnode_attr *vap, vfs_context_t ctx); +errno_t vn_rename(struct vnode *fdvp, struct vnode **fvpp, struct componentname *fcnp, struct vnode_attr *fvap, + struct vnode *tdvp, struct vnode **tvpp, struct componentname *tcnp, struct vnode_attr *tvap, + uint32_t flags, vfs_context_t ctx); + +void lock_vnode_and_post(vnode_t, int); + +#define post_event_if_success(_vp, _error, _event) \ + do { \ + if (0 == (_error)) { \ + lock_vnode_and_post((_vp), (_event)); \ + } \ + } while (0) + +/* Authorization subroutines */ +int vn_authorize_open_existing(vnode_t vp, struct componentname *cnp, int fmode, vfs_context_t ctx, void *reserved); +int vn_authorize_create(vnode_t, struct componentname *, struct vnode_attr *, vfs_context_t, void*); +int vn_attribute_prepare(vnode_t dvp, struct vnode_attr *vap, uint32_t *defaulted_fieldsp, vfs_context_t ctx); +void vn_attribute_cleanup(struct vnode_attr *vap, uint32_t defaulted_fields); +int vn_authorize_unlink(vnode_t dvp, vnode_t vp, struct componentname *cnp, vfs_context_t ctx, void *reserved); +int vn_authorize_rename(struct vnode *fdvp, struct vnode *fvp, struct componentname *fcnp, + struct vnode *tdvp, struct vnode *tvp, struct componentname *tcnp, + vfs_context_t ctx, void *reserved); +int vn_authorize_renamex(struct vnode *fdvp, struct vnode *fvp, struct componentname *fcnp, + struct vnode *tdvp, struct vnode *tvp, struct componentname *tcnp, + vfs_context_t ctx, vfs_rename_flags_t flags, void *reserved); +int vn_authorize_rmdir(vnode_t dvp, vnode_t vp, struct componentname *cnp, vfs_context_t ctx, void *reserved); + +typedef int (*vn_create_authorizer_t)(vnode_t, struct componentname *, struct vnode_attr *, vfs_context_t, void*); +int vn_authorize_mkdir(vnode_t, struct componentname *, struct vnode_attr *, vfs_context_t, void*); +int vn_authorize_null(vnode_t, struct componentname *, struct vnode_attr *, vfs_context_t, void*); +/* End of authorization subroutines */ #define VN_CREATE_NOAUTH (1<<0) #define VN_CREATE_NOINHERIT (1<<1) -errno_t vn_create(vnode_t, vnode_t *, struct componentname *, struct vnode_attr *, int flags, vfs_context_t); - +#define VN_CREATE_UNION (1<<2) +#define VN_CREATE_NOLABEL (1<<3) +#define VN_CREATE_DOOPEN (1<<4) /* Open file if a batched operation is available */ +errno_t vn_create(vnode_t, vnode_t *, struct nameidata *, struct vnode_attr *, uint32_t, int, uint32_t*, vfs_context_t); +int vn_mkdir(vnode_t dvp, vnode_t *vpp, struct nameidata *ndp, struct vnode_attr *vap, vfs_context_t ctx); +int vn_rmdir(vnode_t dvp, vnode_t *vpp, struct nameidata *ndp, struct vnode_attr *vap, vfs_context_t ctx); int vn_getxattr(vnode_t, const char *, uio_t, size_t *, int, vfs_context_t); int vn_setxattr(vnode_t, const char *, uio_t, int, vfs_context_t); int vn_removexattr(vnode_t, const char *, int, vfs_context_t); int vn_listxattr(vnode_t, uio_t, size_t *, int, vfs_context_t); +#if NAMEDSTREAMS +errno_t vnode_getnamedstream(vnode_t, vnode_t *, const char *, enum nsoperation, int, vfs_context_t); +errno_t vnode_makenamedstream(vnode_t, vnode_t *, const char *, int, vfs_context_t); +errno_t vnode_removenamedstream(vnode_t, vnode_t, const char *, int, vfs_context_t); +errno_t vnode_flushnamedstream(vnode_t vp, vnode_t svp, vfs_context_t context); +errno_t vnode_relenamedstream(vnode_t vp, vnode_t svp); +errno_t vnode_verifynamedstream (vnode_t vp); +#endif + + +void nchinit(void); +int resize_namecache(uint32_t newsize); +void name_cache_lock_shared(void); void name_cache_lock(void); void name_cache_unlock(void); +void cache_enter_with_gen(vnode_t dvp, vnode_t vp, struct componentname *cnp, int gen); +const char *cache_enter_create(vnode_t dvp, vnode_t vp, struct componentname *cnp); -char * vnode_getname(vnode_t vp); -void vnode_putname(char *name); +extern int nc_disabled; -vnode_t vnode_getparent(vnode_t vp); +#define vnode_lock_convert(v) lck_mtx_convert_spin(&(v)->v_lock) + +void vnode_lock_spin(vnode_t); -int vn_pathconf(vnode_t, int, register_t *, vfs_context_t); void vnode_list_lock(void); void vnode_list_unlock(void); -int vnode_ref_ext(vnode_t, int); -void vnode_rele_ext(vnode_t, int, int); + +#define VNODE_REF_FORCE 0x1 +int vnode_ref_ext(vnode_t, int, int); + void vnode_rele_internal(vnode_t, int, int, int); -int vnode_getwithref(vnode_t); +#ifdef BSD_KERNEL_PRIVATE +int vnode_getalways(vnode_t); +int vget_internal(vnode_t, int, int); +errno_t vnode_getiocount(vnode_t, unsigned int, int); +#endif /* BSD_KERNEL_PRIVATE */ +int vnode_get_locked(vnode_t); int vnode_put_locked(vnode_t); int vnode_issock(vnode_t); +int vnode_isaliased(vnode_t); void unlock_fsnode(vnode_t, int *); int lock_fsnode(vnode_t, int *); errno_t vnode_resume(vnode_t); +errno_t vnode_suspend(vnode_t); + + +errno_t vnode_mtime(vnode_t, struct timespec *, vfs_context_t); +errno_t vnode_flags(vnode_t, uint32_t *, vfs_context_t); errno_t vnode_size(vnode_t, off_t *, vfs_context_t); errno_t vnode_setsize(vnode_t, off_t, int ioflag, vfs_context_t); int vnode_setattr_fallback(vnode_t vp, struct vnode_attr *vap, vfs_context_t ctx); +int vnode_isspec(vnode_t vp); + + +#ifdef BSD_KERNEL_PRIVATE + +typedef uint32_t compound_vnop_id_t; +#define COMPOUND_VNOP_OPEN 0x01 +#define COMPOUND_VNOP_MKDIR 0x02 +#define COMPOUND_VNOP_RENAME 0x04 +#define COMPOUND_VNOP_REMOVE 0x08 +#define COMPOUND_VNOP_RMDIR 0x10 + +int vnode_compound_rename_available(vnode_t vp); +int vnode_compound_rmdir_available(vnode_t vp); +int vnode_compound_mkdir_available(vnode_t vp); +int vnode_compound_remove_available(vnode_t vp); +int vnode_compound_open_available(vnode_t vp); +int vnode_compound_op_available(vnode_t, compound_vnop_id_t); +#endif /* BSD_KERNEL_PRIVATE */ + +void vn_setunionwait(vnode_t); +void vn_checkunionwait(vnode_t); +void vn_clearunionwait(vnode_t, int); void SPECHASH_LOCK(void); void SPECHASH_UNLOCK(void); -int check_cdevmounted(dev_t, enum vtype, int *); - void vnode_authorize_init(void); +void vfsinit(void); +void vnode_lock(vnode_t); +void vnode_unlock(vnode_t); + +void vn_print_state(vnode_t /* vp */, const char * /* fmt */, ...) + __printflike(2,3); + +#if DEVELOPMENT || DEBUG +#define VNASSERT(exp, vp, msg) \ +do { \ + if (__improbable(!(exp))) { \ + vn_print_state(vp, "VNASSERT failed %s:%d\n", __FILE__, \ + __LINE__); \ + panic msg; \ + } \ +} while (0) +#else +#define VNASSERT(exp, vp, msg) +#endif /* DEVELOPMENT || DEBUG */ + +/* + * XXX exported symbols; should be static + */ +void vfs_op_init(void); +void vfs_opv_init(void); + +#ifdef BSD_KERNEL_PRIVATE +int vfs_sysctl_node SYSCTL_HANDLER_ARGS; +void vnode_setneedinactive(vnode_t); +int vnode_hasnamedstreams(vnode_t); /* Does this vnode have associated named streams? */ + +void nspace_proc_exit(struct proc *p); + +errno_t +vnode_readdir64(struct vnode *vp, struct uio *uio, int flags, int *eofflag, + int *numdirent, vfs_context_t ctxp); + +void vnode_setswapmount(vnode_t); +int64_t vnode_getswappin_avail(vnode_t); + +int vnode_get_snapdir(vnode_t , vnode_t *, vfs_context_t); + +#if CONFIG_TRIGGERS +/* VFS Internal Vnode Trigger Interfaces (Private) */ +int vnode_trigger_resolve(vnode_t, struct nameidata *, vfs_context_t); +void vnode_trigger_rearm(vnode_t, vfs_context_t); +void vfs_nested_trigger_unmounts(mount_t, int, vfs_context_t); +#endif /* CONFIG_TRIGGERS */ + +#endif /* BSD_KERNEL_PRIVATE */ + #endif /* !_SYS_VNODE_INTERNAL_H_ */