static int
nullfs_readdir(struct vnop_readdir_args * ap)
{
struct vnode *vp, *lvp;
int error;
struct null_mount * null_mp = MOUNTTONULLMOUNT(vnode_mount(ap->a_vp));
NULLFSDEBUG("%s %p\n", __FUNCTION__, ap->a_vp);
/* assumption is that any vp that comes through here had to go through lookup
*/
lck_mtx_lock(&null_mp->nullm_lock);
if (nullfs_isspecialvp(ap->a_vp)) {
error = nullfs_special_readdir(ap);
lck_mtx_unlock(&null_mp->nullm_lock);
return error;
}
lck_mtx_unlock(&null_mp->nullm_lock);
vp = ap->a_vp;
lvp = NULLVPTOLOWERVP(vp);
error = vnode_getwithref(lvp);
if (error == 0) {
error = VNOP_READDIR(lvp, ap->a_uio, ap->a_flags, ap->a_eofflag, ap->a_numdirent, ap->a_context);
vnode_put(lvp);
}
return error;
}
static int
nullfs_readlink(struct vnop_readlink_args * ap)
{
NULLFSDEBUG("%s %p\n", __FUNCTION__, ap->a_vp);
int error;
struct vnode *vp, *lvp;
if (nullfs_checkspecialvp(ap->a_vp)) {
return ENOTSUP; /* the special vnodes aren't links */
}
vp = ap->a_vp;
lvp = NULLVPTOLOWERVP(vp);
error = vnode_getwithref(lvp);
if (error == 0) {
error = VNOP_READLINK(lvp, ap->a_uio, ap->a_context);
vnode_put(lvp);
if (error) {
NULLFSDEBUG("readlink failed: %d\n", error);
}
}
return error;
}
static int
nullfs_getattr(struct vnop_getattr_args * args)
{
int error;
struct null_mount * null_mp = MOUNTTONULLMOUNT(vnode_mount(args->a_vp));
NULLFSDEBUG("%s %p\n", __FUNCTION__, args->a_vp);
lck_mtx_lock(&null_mp->nullm_lock);
if (nullfs_isspecialvp(args->a_vp)) {
error = nullfs_special_getattr(args);
lck_mtx_unlock(&null_mp->nullm_lock);
return error;
}
lck_mtx_unlock(&null_mp->nullm_lock);
/* this will return a different inode for third than read dir will */
struct vnode * lowervp = NULLVPTOLOWERVP(args->a_vp);
error = vnode_getwithref(lowervp);
if (error == 0) {
error = VNOP_GETATTR(lowervp, args->a_vap, args->a_context);
vnode_put(lowervp);
if (error == 0) {
/* fix up fsid so it doesn't say the underlying fs*/
VATTR_RETURN(args->a_vap, va_fsid, vfs_statfs(vnode_mount(args->a_vp))->f_fsid.val[0]);
}
}
return error;
}
static int
null_reclaim(struct vnop_reclaim_args * ap)
{
struct vnode * vp;
struct null_node * xp;
struct vnode * lowervp;
struct null_mount * null_mp = MOUNTTONULLMOUNT(vnode_mount(ap->a_vp));
NULLFSDEBUG("%s %p\n", __FUNCTION__, ap->a_vp);
vp = ap->a_vp;
xp = VTONULL(vp);
lowervp = xp->null_lowervp;
lck_mtx_lock(&null_mp->nullm_lock);
vnode_removefsref(vp);
if (lowervp != NULL) {
/* root and second don't have a lowervp, so nothing to release and nothing
* got hashed */
if (xp->null_flags & NULL_FLAG_HASHED) {
/* only call this if we actually made it into the hash list. reclaim gets
called also to
clean up a vnode that got created when it didn't need to under race
conditions */
null_hashrem(xp);
}
vnode_getwithref(lowervp);
vnode_rele(lowervp);
vnode_put(lowervp);
}
if (vp == null_mp->nullm_rootvp) {
null_mp->nullm_rootvp = NULL;
} else if (vp == null_mp->nullm_secondvp) {
null_mp->nullm_secondvp = NULL;
} else if (vp == null_mp->nullm_thirdcovervp) {
null_mp->nullm_thirdcovervp = NULL;
}
lck_mtx_unlock(&null_mp->nullm_lock);
cache_purge(vp);
vnode_clearfsnode(vp);
FREE(xp, M_TEMP);
return 0;
}
void
vm_swapfile_close(uint64_t path_addr, vnode_t vp)
{
struct nameidata nd;
vfs_context_t context = vfs_context_current();
int error = 0;
vnode_getwithref(vp);
vnode_close(vp, 0, context);
NDINIT(&nd, DELETE, OP_UNLINK, AUDITVNPATH1, UIO_SYSSPACE,
path_addr, context);
error = unlink1(context, &nd, 0);
}
void
vm_swapfile_close(uint64_t path_addr, vnode_t vp)
{
vfs_context_t context = vfs_context_current();
int error;
vnode_getwithref(vp);
vnode_close(vp, 0, context);
error = unlink1(context, NULLVP, CAST_USER_ADDR_T(path_addr),
UIO_SYSSPACE, 0);
#if DEVELOPMENT || DEBUG
if (error)
printf("%s : unlink of %s failed with error %d", __FUNCTION__,
(char *)path_addr, error);
#endif
}
static int
nullfs_listxattr(struct vnop_listxattr_args * args)
{
int error;
struct vnode *vp, *lvp;
NULLFSDEBUG("%s %p\n", __FUNCTION__, args->a_vp);
if (nullfs_checkspecialvp(args->a_vp)) {
return 0; /* nothing extra needed */
}
vp = args->a_vp;
lvp = NULLVPTOLOWERVP(vp);
error = vnode_getwithref(lvp);
if (error == 0) {
error = VNOP_LISTXATTR(lvp, args->a_uio, args->a_size, args->a_options, args->a_context);
vnode_put(lvp);
}
return error;
}
static int
nullfs_mnomap(struct vnop_mnomap_args * args)
{
int error;
struct vnode *vp, *lvp;
NULLFSDEBUG("%s %p\n", __FUNCTION__, args->a_vp);
if (nullfs_checkspecialvp(args->a_vp)) {
return 0; /* nothing extra needed */
}
vp = args->a_vp;
lvp = NULLVPTOLOWERVP(vp);
error = vnode_getwithref(lvp);
if (error == 0) {
error = VNOP_MNOMAP(lvp, args->a_context);
vnode_put(lvp);
}
return error;
}
/*
* forkproc
*
* Description: Create a new process structure, given a parent process
* structure.
*
* Parameters: parent_proc The parent process
*
* Returns: !NULL The new process structure
* NULL Error (insufficient free memory)
*
* Note: When successful, the newly created process structure is
* partially initialized; if a caller needs to deconstruct the
* returned structure, they must call forkproc_free() to do so.
*/
proc_t
forkproc(proc_t parent_proc)
{
proc_t child_proc; /* Our new process */
static int nextpid = 0, pidwrap = 0, nextpidversion = 0;
int error = 0;
struct session *sessp;
uthread_t parent_uthread = (uthread_t)get_bsdthread_info(current_thread());
MALLOC_ZONE(child_proc, proc_t , sizeof *child_proc, M_PROC, M_WAITOK);
if (child_proc == NULL) {
printf("forkproc: M_PROC zone exhausted\n");
goto bad;
}
/* zero it out as we need to insert in hash */
bzero(child_proc, sizeof *child_proc);
MALLOC_ZONE(child_proc->p_stats, struct pstats *,
sizeof *child_proc->p_stats, M_PSTATS, M_WAITOK);
if (child_proc->p_stats == NULL) {
printf("forkproc: M_SUBPROC zone exhausted (p_stats)\n");
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
MALLOC_ZONE(child_proc->p_sigacts, struct sigacts *,
sizeof *child_proc->p_sigacts, M_SIGACTS, M_WAITOK);
if (child_proc->p_sigacts == NULL) {
printf("forkproc: M_SUBPROC zone exhausted (p_sigacts)\n");
FREE_ZONE(child_proc->p_stats, sizeof *child_proc->p_stats, M_PSTATS);
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
/* allocate a callout for use by interval timers */
child_proc->p_rcall = thread_call_allocate((thread_call_func_t)realitexpire, child_proc);
if (child_proc->p_rcall == NULL) {
FREE_ZONE(child_proc->p_sigacts, sizeof *child_proc->p_sigacts, M_SIGACTS);
FREE_ZONE(child_proc->p_stats, sizeof *child_proc->p_stats, M_PSTATS);
FREE_ZONE(child_proc, sizeof *child_proc, M_PROC);
child_proc = NULL;
goto bad;
}
/*
* Find an unused PID.
*/
proc_list_lock();
nextpid++;
retry:
/*
* If the process ID prototype has wrapped around,
* restart somewhat above 0, as the low-numbered procs
* tend to include daemons that don't exit.
*/
if (nextpid >= PID_MAX) {
nextpid = 100;
pidwrap = 1;
}
if (pidwrap != 0) {
/* if the pid stays in hash both for zombie and runniing state */
if (pfind_locked(nextpid) != PROC_NULL) {
nextpid++;
goto retry;
}
if (pgfind_internal(nextpid) != PGRP_NULL) {
nextpid++;
goto retry;
}
if (session_find_internal(nextpid) != SESSION_NULL) {
nextpid++;
goto retry;
}
}
nprocs++;
child_proc->p_pid = nextpid;
child_proc->p_idversion = nextpidversion++;
#if 1
if (child_proc->p_pid != 0) {
if (pfind_locked(child_proc->p_pid) != PROC_NULL)
panic("proc in the list already\n");
}
#endif
/* Insert in the hash */
child_proc->p_listflag |= (P_LIST_INHASH | P_LIST_INCREATE);
LIST_INSERT_HEAD(PIDHASH(child_proc->p_pid), child_proc, p_hash);
proc_list_unlock();
/*
* We've identified the PID we are going to use; initialize the new
* process structure.
*/
child_proc->p_stat = SIDL;
child_proc->p_pgrpid = PGRPID_DEAD;
/*
* The zero'ing of the proc was at the allocation time due to need
* for insertion to hash. Copy the section that is to be copied
* directly from the parent.
*/
bcopy(&parent_proc->p_startcopy, &child_proc->p_startcopy,
(unsigned) ((caddr_t)&child_proc->p_endcopy - (caddr_t)&child_proc->p_startcopy));
/*
* Some flags are inherited from the parent.
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
* The p_stats and p_sigacts substructs are set in vm_fork.
*/
child_proc->p_flag = (parent_proc->p_flag & (P_LP64 | P_TRANSLATED | P_AFFINITY));
if (parent_proc->p_flag & P_PROFIL)
startprofclock(child_proc);
/*
* Note that if the current thread has an assumed identity, this
* credential will be granted to the new process.
*/
child_proc->p_ucred = kauth_cred_get_with_ref();
#ifdef CONFIG_EMBEDDED
lck_mtx_init(&child_proc->p_mlock, proc_lck_grp, proc_lck_attr);
lck_mtx_init(&child_proc->p_fdmlock, proc_lck_grp, proc_lck_attr);
#if CONFIG_DTRACE
lck_mtx_init(&child_proc->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
#endif
lck_spin_init(&child_proc->p_slock, proc_lck_grp, proc_lck_attr);
#else /* !CONFIG_EMBEDDED */
lck_mtx_init(&child_proc->p_mlock, proc_mlock_grp, proc_lck_attr);
lck_mtx_init(&child_proc->p_fdmlock, proc_fdmlock_grp, proc_lck_attr);
#if CONFIG_DTRACE
lck_mtx_init(&child_proc->p_dtrace_sprlock, proc_lck_grp, proc_lck_attr);
#endif
lck_spin_init(&child_proc->p_slock, proc_slock_grp, proc_lck_attr);
#endif /* !CONFIG_EMBEDDED */
klist_init(&child_proc->p_klist);
if (child_proc->p_textvp != NULLVP) {
/* bump references to the text vnode */
/* Need to hold iocount across the ref call */
if (vnode_getwithref(child_proc->p_textvp) == 0) {
error = vnode_ref(child_proc->p_textvp);
vnode_put(child_proc->p_textvp);
if (error != 0)
child_proc->p_textvp = NULLVP;
}
}
/*
* Copy the parents per process open file table to the child; if
* there is a per-thread current working directory, set the childs
* per-process current working directory to that instead of the
* parents.
*
* XXX may fail to copy descriptors to child
*/
child_proc->p_fd = fdcopy(parent_proc, parent_uthread->uu_cdir);
#if SYSV_SHM
if (parent_proc->vm_shm) {
/* XXX may fail to attach shm to child */
(void)shmfork(parent_proc, child_proc);
}
#endif
/*
* inherit the limit structure to child
*/
proc_limitfork(parent_proc, child_proc);
if (child_proc->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
uint64_t rlim_cur = child_proc->p_limit->pl_rlimit[RLIMIT_CPU].rlim_cur;
child_proc->p_rlim_cpu.tv_sec = (rlim_cur > __INT_MAX__) ? __INT_MAX__ : rlim_cur;
}
/* Intialize new process stats, including start time */
/* <rdar://6640543> non-zeroed portion contains garbage AFAICT */
bzero(&child_proc->p_stats->pstat_startzero,
(unsigned) ((caddr_t)&child_proc->p_stats->pstat_endzero -
(caddr_t)&child_proc->p_stats->pstat_startzero));
bzero(&child_proc->p_stats->user_p_prof, sizeof(struct user_uprof));
microtime(&child_proc->p_start);
child_proc->p_stats->p_start = child_proc->p_start; /* for compat */
if (parent_proc->p_sigacts != NULL)
(void)memcpy(child_proc->p_sigacts,
parent_proc->p_sigacts, sizeof *child_proc->p_sigacts);
else
(void)memset(child_proc->p_sigacts, 0, sizeof *child_proc->p_sigacts);
sessp = proc_session(parent_proc);
if (sessp->s_ttyvp != NULL && parent_proc->p_flag & P_CONTROLT)
OSBitOrAtomic(P_CONTROLT, &child_proc->p_flag);
session_rele(sessp);
/*
* block all signals to reach the process.
* no transition race should be occuring with the child yet,
* but indicate that the process is in (the creation) transition.
*/
proc_signalstart(child_proc, 0);
proc_transstart(child_proc, 0);
child_proc->p_pcaction = (parent_proc->p_pcaction) & P_PCMAX;
TAILQ_INIT(&child_proc->p_uthlist);
TAILQ_INIT(&child_proc->p_aio_activeq);
TAILQ_INIT(&child_proc->p_aio_doneq);
/* Inherit the parent flags for code sign */
child_proc->p_csflags = parent_proc->p_csflags;
/*
* All processes have work queue locks; cleaned up by
* reap_child_locked()
*/
workqueue_init_lock(child_proc);
/*
* Copy work queue information
*
* Note: This should probably only happen in the case where we are
* creating a child that is a copy of the parent; since this
* routine is called in the non-duplication case of vfork()
* or posix_spawn(), then this information should likely not
* be duplicated.
*
* <rdar://6640553> Work queue pointers that no longer point to code
*/
child_proc->p_wqthread = parent_proc->p_wqthread;
child_proc->p_threadstart = parent_proc->p_threadstart;
child_proc->p_pthsize = parent_proc->p_pthsize;
child_proc->p_targconc = parent_proc->p_targconc;
if ((parent_proc->p_lflag & P_LREGISTER) != 0) {
child_proc->p_lflag |= P_LREGISTER;
}
child_proc->p_dispatchqueue_offset = parent_proc->p_dispatchqueue_offset;
#if PSYNCH
pth_proc_hashinit(child_proc);
#endif /* PSYNCH */
#if CONFIG_LCTX
child_proc->p_lctx = NULL;
/* Add new process to login context (if any). */
if (parent_proc->p_lctx != NULL) {
/*
* <rdar://6640564> This should probably be delayed in the
* vfork() or posix_spawn() cases.
*/
LCTX_LOCK(parent_proc->p_lctx);
enterlctx(child_proc, parent_proc->p_lctx, 0);
}
#endif
bad:
return(child_proc);
}
/*
* Search a pathname.
* This is a very central and rather complicated routine.
*
* The pathname is pointed to by ni_ptr and is of length ni_pathlen.
* The starting directory is taken from ni_startdir. The pathname is
* descended until done, or a symbolic link is encountered. The variable
* ni_more is clear if the path is completed; it is set to one if a
* symbolic link needing interpretation is encountered.
*
* The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
* whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it, the parent directory is returned
* locked. If flag has WANTPARENT or'ed into it, the parent directory is
* returned unlocked. Otherwise the parent directory is not returned. If
* the target of the pathname exists and LOCKLEAF is or'ed into the flag
* the target is returned locked, otherwise it is returned unlocked.
* When creating or renaming and LOCKPARENT is specified, the target may not
* be ".". When deleting and LOCKPARENT is specified, the target may be ".".
*
* Overall outline of lookup:
*
* dirloop:
* identify next component of name at ndp->ni_ptr
* handle degenerate case where name is null string
* if .. and crossing mount points and on mounted filesys, find parent
* call VNOP_LOOKUP routine for next component name
* directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
* component vnode returned in ni_vp (if it exists), locked.
* if result vnode is mounted on and crossing mount points,
* find mounted on vnode
* if more components of name, do next level at dirloop
* return the answer in ni_vp, locked if LOCKLEAF set
* if LOCKPARENT set, return locked parent in ni_dvp
* if WANTPARENT set, return unlocked parent in ni_dvp
*
* Returns: 0 Success
* ENOENT No such file or directory
* EBADF Bad file descriptor
* ENOTDIR Not a directory
* EROFS Read-only file system [CREATE]
* EISDIR Is a directory [CREATE]
* cache_lookup_path:ERECYCLE (vnode was recycled from underneath us, redrive lookup again)
* vnode_authorize:EROFS
* vnode_authorize:EACCES
* vnode_authorize:EPERM
* vnode_authorize:???
* VNOP_LOOKUP:ENOENT No such file or directory
* VNOP_LOOKUP:EJUSTRETURN Restart system call (INTERNAL)
* VNOP_LOOKUP:???
* VFS_ROOT:ENOTSUP
* VFS_ROOT:ENOENT
* VFS_ROOT:???
*/
int
lookup(struct nameidata *ndp)
{
char *cp; /* pointer into pathname argument */
vnode_t tdp; /* saved dp */
vnode_t dp; /* the directory we are searching */
int docache = 1; /* == 0 do not cache last component */
int wantparent; /* 1 => wantparent or lockparent flag */
int rdonly; /* lookup read-only flag bit */
int dp_authorized = 0;
int error = 0;
struct componentname *cnp = &ndp->ni_cnd;
vfs_context_t ctx = cnp->cn_context;
int vbusyflags = 0;
int nc_generation = 0;
vnode_t last_dp = NULLVP;
int keep_going;
int atroot;
/*
* Setup: break out flag bits into variables.
*/
if (cnp->cn_flags & (NOCACHE | DOWHITEOUT)) {
if ((cnp->cn_flags & NOCACHE) || (cnp->cn_nameiop == DELETE))
docache = 0;
}
wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
rdonly = cnp->cn_flags & RDONLY;
cnp->cn_flags &= ~ISSYMLINK;
cnp->cn_consume = 0;
dp = ndp->ni_startdir;
ndp->ni_startdir = NULLVP;
if ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0)
vbusyflags = LK_NOWAIT;
cp = cnp->cn_nameptr;
if (*cp == '\0') {
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_vp = dp;
error = lookup_handle_emptyname(ndp, cnp, wantparent);
if (error) {
goto bad;
}
return 0;
}
dirloop:
atroot = 0;
ndp->ni_vp = NULLVP;
if ( (error = cache_lookup_path(ndp, cnp, dp, ctx, &dp_authorized, last_dp)) ) {
dp = NULLVP;
goto bad;
}
if ((cnp->cn_flags & ISLASTCN)) {
if (docache)
cnp->cn_flags |= MAKEENTRY;
} else
cnp->cn_flags |= MAKEENTRY;
dp = ndp->ni_dvp;
if (ndp->ni_vp != NULLVP) {
/*
* cache_lookup_path returned a non-NULL ni_vp then,
* we're guaranteed that the dp is a VDIR, it's
* been authorized, and vp is not ".."
*
* make sure we don't try to enter the name back into
* the cache if this vp is purged before we get to that
* check since we won't have serialized behind whatever
* activity is occurring in the FS that caused the purge
*/
if (dp != NULLVP)
nc_generation = dp->v_nc_generation - 1;
goto returned_from_lookup_path;
}
/*
* Handle "..": two special cases.
* 1. If at root directory (e.g. after chroot)
* or at absolute root directory
* then ignore it so can't get out.
* 2. If this vnode is the root of a mounted
* filesystem, then replace it with the
* vnode which was mounted on so we take the
* .. in the other file system.
*/
if ( (cnp->cn_flags & ISDOTDOT) ) {
for (;;) {
if (dp == ndp->ni_rootdir || dp == rootvnode) {
ndp->ni_dvp = dp;
ndp->ni_vp = dp;
/*
* we're pinned at the root
* we've already got one reference on 'dp'
* courtesy of cache_lookup_path... take
* another one for the ".."
* if we fail to get the new reference, we'll
* drop our original down in 'bad'
*/
if ( (vnode_get(dp)) ) {
error = ENOENT;
goto bad;
}
atroot = 1;
goto returned_from_lookup_path;
}
if ((dp->v_flag & VROOT) == 0 ||
(cnp->cn_flags & NOCROSSMOUNT))
break;
if (dp->v_mount == NULL) { /* forced umount */
error = EBADF;
goto bad;
}
tdp = dp;
dp = tdp->v_mount->mnt_vnodecovered;
vnode_put(tdp);
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
dp_authorized = 0;
}
}
/*
* We now have a segment name to search for, and a directory to search.
*/
unionlookup:
ndp->ni_vp = NULLVP;
if (dp->v_type != VDIR) {
error = ENOTDIR;
goto lookup_error;
}
if ( (cnp->cn_flags & DONOTAUTH) != DONOTAUTH ) {
error = lookup_authorize_search(dp, cnp, dp_authorized, ctx);
if (error) {
goto lookup_error;
}
}
/*
* Now that we've authorized a lookup, can bail out if the filesystem
* will be doing a batched operation. Return an iocount on dvp.
*/
#if NAMEDRSRCFORK
if ((cnp->cn_flags & ISLASTCN) && namei_compound_available(dp, ndp) && !(cnp->cn_flags & CN_WANTSRSRCFORK)) {
#else
if ((cnp->cn_flags & ISLASTCN) && namei_compound_available(dp, ndp)) {
#endif /* NAMEDRSRCFORK */
ndp->ni_flag |= NAMEI_UNFINISHED;
ndp->ni_ncgeneration = dp->v_nc_generation;
return 0;
}
nc_generation = dp->v_nc_generation;
error = VNOP_LOOKUP(dp, &ndp->ni_vp, cnp, ctx);
if ( error ) {
lookup_error:
if ((error == ENOENT) &&
(dp->v_flag & VROOT) && (dp->v_mount != NULL) &&
(dp->v_mount->mnt_flag & MNT_UNION)) {
#if CONFIG_VFS_FUNNEL
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(dp, NULL);
}
#endif /* CONFIG_VFS_FUNNEL */
tdp = dp;
dp = tdp->v_mount->mnt_vnodecovered;
vnode_put(tdp);
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
dp_authorized = 0;
goto unionlookup;
}
if (error != EJUSTRETURN)
goto bad;
if (ndp->ni_vp != NULLVP)
panic("leaf should be empty");
error = lookup_validate_creation_path(ndp);
if (error)
goto bad;
/*
* We return with ni_vp NULL to indicate that the entry
* doesn't currently exist, leaving a pointer to the
* referenced directory vnode in ndp->ni_dvp.
*/
if (cnp->cn_flags & SAVESTART) {
if ( (vnode_get(ndp->ni_dvp)) ) {
error = ENOENT;
goto bad;
}
ndp->ni_startdir = ndp->ni_dvp;
}
if (!wantparent)
vnode_put(ndp->ni_dvp);
if (kdebug_enable)
kdebug_lookup(ndp->ni_dvp, cnp);
return (0);
}
returned_from_lookup_path:
/* We'll always have an iocount on ni_vp when this finishes. */
error = lookup_handle_found_vnode(ndp, cnp, rdonly, vbusyflags, &keep_going, nc_generation, wantparent, atroot, ctx);
if (error != 0) {
goto bad2;
}
if (keep_going) {
dp = ndp->ni_vp;
/* namei() will handle symlinks */
if ((dp->v_type == VLNK) &&
((cnp->cn_flags & FOLLOW) || (ndp->ni_flag & NAMEI_TRAILINGSLASH) || *ndp->ni_next == '/')) {
return 0;
}
/*
* Otherwise, there's more path to process.
* cache_lookup_path is now responsible for dropping io ref on dp
* when it is called again in the dirloop. This ensures we hold
* a ref on dp until we complete the next round of lookup.
*/
last_dp = dp;
goto dirloop;
}
return (0);
bad2:
#if CONFIG_VFS_FUNNEL
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
#endif /* CONFIG_VFS_FUNNEL */
if (ndp->ni_dvp)
vnode_put(ndp->ni_dvp);
vnode_put(ndp->ni_vp);
ndp->ni_vp = NULLVP;
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (error);
bad:
#if CONFIG_VFS_FUNNEL
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
#endif /* CONFIG_VFS_FUNNEL */
if (dp)
vnode_put(dp);
ndp->ni_vp = NULLVP;
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (error);
}
int
lookup_validate_creation_path(struct nameidata *ndp)
{
struct componentname *cnp = &ndp->ni_cnd;
/*
* If creating and at end of pathname, then can consider
* allowing file to be created.
*/
if (cnp->cn_flags & RDONLY) {
return EROFS;
}
if ((cnp->cn_flags & ISLASTCN) && (ndp->ni_flag & NAMEI_TRAILINGSLASH) && !(cnp->cn_flags & WILLBEDIR)) {
return ENOENT;
}
return 0;
}
/*
* Modifies only ni_vp. Always returns with ni_vp still valid (iocount held).
*/
int
lookup_traverse_mountpoints(struct nameidata *ndp, struct componentname *cnp, vnode_t dp,
int vbusyflags, vfs_context_t ctx)
{
mount_t mp;
vnode_t tdp;
int error = 0;
uthread_t uth;
uint32_t depth = 0;
int dont_cache_mp = 0;
vnode_t mounted_on_dp;
int current_mount_generation = 0;
mounted_on_dp = dp;
current_mount_generation = mount_generation;
while ((dp->v_type == VDIR) && dp->v_mountedhere &&
((cnp->cn_flags & NOCROSSMOUNT) == 0)) {
#if CONFIG_TRIGGERS
/*
* For a trigger vnode, call its resolver when crossing its mount (if requested)
*/
if (dp->v_resolve) {
(void) vnode_trigger_resolve(dp, ndp, ctx);
}
#endif
vnode_lock(dp);
if ((dp->v_type == VDIR) && (mp = dp->v_mountedhere)) {
mp->mnt_crossref++;
vnode_unlock(dp);
if (vfs_busy(mp, vbusyflags)) {
mount_dropcrossref(mp, dp, 0);
if (vbusyflags == LK_NOWAIT) {
error = ENOENT;
goto out;
}
continue;
}
/*
* XXX - if this is the last component of the
* pathname, and it's either not a lookup operation
* or the NOTRIGGER flag is set for the operation,
* set a uthread flag to let VFS_ROOT() for autofs
* know it shouldn't trigger a mount.
*/
uth = (struct uthread *)get_bsdthread_info(current_thread());
if ((cnp->cn_flags & ISLASTCN) &&
(cnp->cn_nameiop != LOOKUP ||
(cnp->cn_flags & NOTRIGGER))) {
uth->uu_notrigger = 1;
dont_cache_mp = 1;
}
error = VFS_ROOT(mp, &tdp, ctx);
/* XXX - clear the uthread flag */
uth->uu_notrigger = 0;
mount_dropcrossref(mp, dp, 0);
vfs_unbusy(mp);
if (error) {
goto out;
}
vnode_put(dp);
ndp->ni_vp = dp = tdp;
depth++;
#if CONFIG_TRIGGERS
/*
* Check if root dir is a trigger vnode
*/
if (dp->v_resolve) {
error = vnode_trigger_resolve(dp, ndp, ctx);
if (error) {
goto out;
}
}
#endif
} else {
vnode_unlock(dp);
break;
}
}
if (depth && !dont_cache_mp) {
mp = mounted_on_dp->v_mountedhere;
if (mp) {
mount_lock_spin(mp);
mp->mnt_realrootvp_vid = dp->v_id;
mp->mnt_realrootvp = dp;
mp->mnt_generation = current_mount_generation;
mount_unlock(mp);
}
}
return 0;
out:
return error;
}
/*
* XXX Internally, we use VM_PROT_* somewhat interchangeably, but the correct
* XXX usage is PROT_* from an interface perspective. Thus the values of
* XXX VM_PROT_* and PROT_* need to correspond.
*/
int
mmap(proc_t p, struct mmap_args *uap, user_addr_t *retval)
{
/*
* Map in special device (must be SHARED) or file
*/
struct fileproc *fp;
register struct vnode *vp;
int flags;
int prot, file_prot;
int err=0;
vm_map_t user_map;
kern_return_t result;
mach_vm_offset_t user_addr;
mach_vm_size_t user_size;
vm_object_offset_t pageoff;
vm_object_offset_t file_pos;
int alloc_flags=0;
boolean_t docow;
vm_prot_t maxprot;
void *handle;
vm_pager_t pager;
int mapanon=0;
int fpref=0;
int error =0;
int fd = uap->fd;
user_addr = (mach_vm_offset_t)uap->addr;
user_size = (mach_vm_size_t) uap->len;
AUDIT_ARG(addr, user_addr);
AUDIT_ARG(len, user_size);
AUDIT_ARG(fd, uap->fd);
prot = (uap->prot & VM_PROT_ALL);
#if 3777787
/*
* Since the hardware currently does not support writing without
* read-before-write, or execution-without-read, if the request is
* for write or execute access, we must imply read access as well;
* otherwise programs expecting this to work will fail to operate.
*/
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
#endif /* radar 3777787 */
flags = uap->flags;
vp = NULLVP;
/*
* The vm code does not have prototypes & compiler doesn't do the'
* the right thing when you cast 64bit value and pass it in function
* call. So here it is.
*/
file_pos = (vm_object_offset_t)uap->pos;
/* make sure mapping fits into numeric range etc */
if (file_pos + user_size > (vm_object_offset_t)-PAGE_SIZE_64)
return (EINVAL);
/*
* Align the file position to a page boundary,
* and save its page offset component.
*/
pageoff = (file_pos & PAGE_MASK);
file_pos -= (vm_object_offset_t)pageoff;
/* Adjust size for rounding (on both ends). */
user_size += pageoff; /* low end... */
user_size = mach_vm_round_page(user_size); /* hi end */
/*
* Check for illegal addresses. Watch out for address wrap... Note
* that VM_*_ADDRESS are not constants due to casts (argh).
*/
if (flags & MAP_FIXED) {
/*
* The specified address must have the same remainder
* as the file offset taken modulo PAGE_SIZE, so it
* should be aligned after adjustment by pageoff.
*/
user_addr -= pageoff;
if (user_addr & PAGE_MASK)
return (EINVAL);
}
#ifdef notyet
/* DO not have apis to get this info, need to wait till then*/
/*
* XXX for non-fixed mappings where no hint is provided or
* the hint would fall in the potential heap space,
* place it after the end of the largest possible heap.
*
* There should really be a pmap call to determine a reasonable
* location.
*/
else if (addr < mach_vm_round_page(p->p_vmspace->vm_daddr + MAXDSIZ))
addr = mach_vm_round_page(p->p_vmspace->vm_daddr + MAXDSIZ);
#endif
alloc_flags = 0;
if (flags & MAP_ANON) {
/*
* Mapping blank space is trivial. Use positive fds as the alias
* value for memory tracking.
*/
if (fd != -1) {
/*
* Use "fd" to pass (some) Mach VM allocation flags,
* (see the VM_FLAGS_* definitions).
*/
alloc_flags = fd & (VM_FLAGS_ALIAS_MASK |
VM_FLAGS_PURGABLE);
if (alloc_flags != fd) {
/* reject if there are any extra flags */
return EINVAL;
}
}
handle = NULL;
maxprot = VM_PROT_ALL;
file_pos = 0;
mapanon = 1;
} else {
struct vnode_attr va;
vfs_context_t ctx = vfs_context_current();
/*
* Mapping file, get fp for validation. Obtain vnode and make
* sure it is of appropriate type.
*/
err = fp_lookup(p, fd, &fp, 0);
if (err)
return(err);
fpref = 1;
if(fp->f_fglob->fg_type == DTYPE_PSXSHM) {
uap->addr = (user_addr_t)user_addr;
uap->len = (user_size_t)user_size;
uap->prot = prot;
uap->flags = flags;
uap->pos = file_pos;
error = pshm_mmap(p, uap, retval, fp, (off_t)pageoff);
goto bad;
}
if (fp->f_fglob->fg_type != DTYPE_VNODE) {
error = EINVAL;
goto bad;
}
vp = (struct vnode *)fp->f_fglob->fg_data;
error = vnode_getwithref(vp);
if(error != 0)
goto bad;
if (vp->v_type != VREG && vp->v_type != VCHR) {
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
/*
* POSIX: mmap needs to update access time for mapped files
*/
if ((vnode_vfsvisflags(vp) & MNT_NOATIME) == 0) {
VATTR_INIT(&va);
nanotime(&va.va_access_time);
VATTR_SET_ACTIVE(&va, va_access_time);
vnode_setattr(vp, &va, ctx);
}
/*
* XXX hack to handle use of /dev/zero to map anon memory (ala
* SunOS).
*/
if (vp->v_type == VCHR || vp->v_type == VSTR) {
(void)vnode_put(vp);
error = ENODEV;
goto bad;
} else {
/*
* Ensure that file and memory protections are
* compatible. Note that we only worry about
* writability if mapping is shared; in this case,
* current and max prot are dictated by the open file.
* XXX use the vnode instead? Problem is: what
* credentials do we use for determination? What if
* proc does a setuid?
*/
maxprot = VM_PROT_EXECUTE; /* ??? */
if (fp->f_fglob->fg_flag & FREAD)
maxprot |= VM_PROT_READ;
else if (prot & PROT_READ) {
(void)vnode_put(vp);
error = EACCES;
goto bad;
}
/*
* If we are sharing potential changes (either via
* MAP_SHARED or via the implicit sharing of character
* device mappings), and we are trying to get write
* permission although we opened it without asking
* for it, bail out.
*/
if ((flags & MAP_SHARED) != 0) {
if ((fp->f_fglob->fg_flag & FWRITE) != 0) {
/*
* check for write access
*
* Note that we already made this check when granting FWRITE
* against the file, so it seems redundant here.
*/
error = vnode_authorize(vp, NULL, KAUTH_VNODE_CHECKIMMUTABLE, ctx);
/* if not granted for any reason, but we wanted it, bad */
if ((prot & PROT_WRITE) && (error != 0)) {
vnode_put(vp);
goto bad;
}
/* if writable, remember */
if (error == 0)
maxprot |= VM_PROT_WRITE;
} else if ((prot & PROT_WRITE) != 0) {
(void)vnode_put(vp);
error = EACCES;
goto bad;
}
} else
maxprot |= VM_PROT_WRITE;
handle = (void *)vp;
#if CONFIG_MACF
error = mac_file_check_mmap(vfs_context_ucred(ctx),
fp->f_fglob, prot, flags, &maxprot);
if (error) {
(void)vnode_put(vp);
goto bad;
}
#endif /* MAC */
}
}
if (user_size == 0) {
if (!mapanon)
(void)vnode_put(vp);
error = 0;
goto bad;
}
/*
* We bend a little - round the start and end addresses
* to the nearest page boundary.
*/
user_size = mach_vm_round_page(user_size);
if (file_pos & PAGE_MASK_64) {
if (!mapanon)
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
user_map = current_map();
if ((flags & MAP_FIXED) == 0) {
alloc_flags |= VM_FLAGS_ANYWHERE;
user_addr = mach_vm_round_page(user_addr);
} else {
if (user_addr != mach_vm_trunc_page(user_addr)) {
if (!mapanon)
(void)vnode_put(vp);
error = EINVAL;
goto bad;
}
/*
* mmap(MAP_FIXED) will replace any existing mappings in the
* specified range, if the new mapping is successful.
* If we just deallocate the specified address range here,
* another thread might jump in and allocate memory in that
* range before we get a chance to establish the new mapping,
* and we won't have a chance to restore the old mappings.
* So we use VM_FLAGS_OVERWRITE to let Mach VM know that it
* has to deallocate the existing mappings and establish the
* new ones atomically.
*/
alloc_flags |= VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE;
}
if (flags & MAP_NOCACHE)
alloc_flags |= VM_FLAGS_NO_CACHE;
/*
* Lookup/allocate object.
*/
if (handle == NULL) {
pager = NULL;
#ifdef notyet
/* Hmm .. */
#if defined(VM_PROT_READ_IS_EXEC)
if (prot & VM_PROT_READ)
prot |= VM_PROT_EXECUTE;
if (maxprot & VM_PROT_READ)
maxprot |= VM_PROT_EXECUTE;
#endif
#endif
#if 3777787
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
if (maxprot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
maxprot |= VM_PROT_READ;
#endif /* radar 3777787 */
result = vm_map_enter_mem_object(user_map,
&user_addr, user_size,
0, alloc_flags,
IPC_PORT_NULL, 0, FALSE,
prot, maxprot,
(flags & MAP_SHARED) ?
VM_INHERIT_SHARE :
VM_INHERIT_DEFAULT);
if (result != KERN_SUCCESS)
goto out;
} else {
pager = (vm_pager_t)ubc_getpager(vp);
if (pager == NULL) {
(void)vnode_put(vp);
error = ENOMEM;
goto bad;
}
/*
* Set credentials:
* FIXME: if we're writing the file we need a way to
* ensure that someone doesn't replace our R/W creds
* with ones that only work for read.
*/
ubc_setthreadcred(vp, p, current_thread());
docow = FALSE;
if ((flags & (MAP_ANON|MAP_SHARED)) == 0) {
docow = TRUE;
}
#ifdef notyet
/* Hmm .. */
#if defined(VM_PROT_READ_IS_EXEC)
if (prot & VM_PROT_READ)
prot |= VM_PROT_EXECUTE;
if (maxprot & VM_PROT_READ)
maxprot |= VM_PROT_EXECUTE;
#endif
#endif /* notyet */
#if 3777787
if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
prot |= VM_PROT_READ;
if (maxprot & (VM_PROT_EXECUTE | VM_PROT_WRITE))
maxprot |= VM_PROT_READ;
#endif /* radar 3777787 */
result = vm_map_enter_mem_object(user_map,
&user_addr, user_size,
0, alloc_flags,
(ipc_port_t)pager, file_pos,
docow, prot, maxprot,
(flags & MAP_SHARED) ?
VM_INHERIT_SHARE :
VM_INHERIT_DEFAULT);
if (result != KERN_SUCCESS) {
(void)vnode_put(vp);
goto out;
}
file_prot = prot & (PROT_READ | PROT_WRITE | PROT_EXEC);
if (docow) {
/* private mapping: won't write to the file */
file_prot &= ~PROT_WRITE;
}
(void) ubc_map(vp, file_prot);
}
if (!mapanon)
(void)vnode_put(vp);
out:
switch (result) {
case KERN_SUCCESS:
*retval = user_addr + pageoff;
error = 0;
break;
case KERN_INVALID_ADDRESS:
case KERN_NO_SPACE:
error = ENOMEM;
break;
case KERN_PROTECTION_FAILURE:
error = EACCES;
break;
default:
error = EINVAL;
break;
}
bad:
if (fpref)
fp_drop(p, fd, fp, 0);
KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO, SYS_mmap) | DBG_FUNC_NONE), fd, (uint32_t)(*retval), (uint32_t)user_size, error, 0);
KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO2, SYS_mmap) | DBG_FUNC_NONE), (uint32_t)(*retval >> 32), (uint32_t)(user_size >> 32),
(uint32_t)(file_pos >> 32), (uint32_t)file_pos, 0);
return(error);
}
kern_return_t
map_fd_funneled(
int fd,
vm_object_offset_t offset,
vm_offset_t *va,
boolean_t findspace,
vm_size_t size)
{
kern_return_t result;
struct fileproc *fp;
struct vnode *vp;
void * pager;
vm_offset_t map_addr=0;
vm_size_t map_size;
int err=0;
vm_map_t my_map;
proc_t p = current_proc();
struct vnode_attr vattr;
/*
* Find the inode; verify that it's a regular file.
*/
err = fp_lookup(p, fd, &fp, 0);
if (err)
return(err);
if (fp->f_fglob->fg_type != DTYPE_VNODE){
err = KERN_INVALID_ARGUMENT;
goto bad;
}
if (!(fp->f_fglob->fg_flag & FREAD)) {
err = KERN_PROTECTION_FAILURE;
goto bad;
}
vp = (struct vnode *)fp->f_fglob->fg_data;
err = vnode_getwithref(vp);
if(err != 0)
goto bad;
if (vp->v_type != VREG) {
(void)vnode_put(vp);
err = KERN_INVALID_ARGUMENT;
goto bad;
}
AUDIT_ARG(vnpath, vp, ARG_VNODE1);
/*
* POSIX: mmap needs to update access time for mapped files
*/
if ((vnode_vfsvisflags(vp) & MNT_NOATIME) == 0) {
VATTR_INIT(&vattr);
nanotime(&vattr.va_access_time);
VATTR_SET_ACTIVE(&vattr, va_access_time);
vnode_setattr(vp, &vattr, vfs_context_current());
}
if (offset & PAGE_MASK_64) {
printf("map_fd: file offset not page aligned(%d : %s)\n",p->p_pid, p->p_comm);
(void)vnode_put(vp);
err = KERN_INVALID_ARGUMENT;
goto bad;
}
map_size = round_page(size);
/*
* Allow user to map in a zero length file.
*/
if (size == 0) {
(void)vnode_put(vp);
err = KERN_SUCCESS;
goto bad;
}
/*
* Map in the file.
*/
pager = (void *)ubc_getpager(vp);
if (pager == NULL) {
(void)vnode_put(vp);
err = KERN_FAILURE;
goto bad;
}
my_map = current_map();
result = vm_map_64(
my_map,
&map_addr, map_size, (vm_offset_t)0,
VM_FLAGS_ANYWHERE, pager, offset, TRUE,
VM_PROT_DEFAULT, VM_PROT_ALL,
VM_INHERIT_DEFAULT);
if (result != KERN_SUCCESS) {
(void)vnode_put(vp);
err = result;
goto bad;
}
if (!findspace) {
vm_offset_t dst_addr;
vm_map_copy_t tmp;
if (copyin(CAST_USER_ADDR_T(va), &dst_addr, sizeof (dst_addr)) ||
trunc_page_32(dst_addr) != dst_addr) {
(void) vm_map_remove(
my_map,
map_addr, map_addr + map_size,
VM_MAP_NO_FLAGS);
(void)vnode_put(vp);
err = KERN_INVALID_ADDRESS;
goto bad;
}
result = vm_map_copyin(my_map, (vm_map_address_t)map_addr,
(vm_map_size_t)map_size, TRUE, &tmp);
if (result != KERN_SUCCESS) {
(void) vm_map_remove(my_map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size),
VM_MAP_NO_FLAGS);
(void)vnode_put(vp);
err = result;
goto bad;
}
result = vm_map_copy_overwrite(my_map,
(vm_map_address_t)dst_addr, tmp, FALSE);
if (result != KERN_SUCCESS) {
vm_map_copy_discard(tmp);
(void)vnode_put(vp);
err = result;
goto bad;
}
} else {
if (copyout(&map_addr, CAST_USER_ADDR_T(va), sizeof (map_addr))) {
(void) vm_map_remove(my_map, vm_map_trunc_page(map_addr),
vm_map_round_page(map_addr + map_size),
VM_MAP_NO_FLAGS);
(void)vnode_put(vp);
err = KERN_INVALID_ADDRESS;
goto bad;
}
}
ubc_setthreadcred(vp, current_proc(), current_thread());
(void)ubc_map(vp, (PROT_READ | PROT_EXEC));
(void)vnode_put(vp);
err = 0;
bad:
fp_drop(p, fd, fp, 0);
return (err);
}
/*
* shared_region_map_np()
*
* This system call is intended for dyld.
*
* dyld uses this to map a shared cache file into a shared region.
* This is usually done only the first time a shared cache is needed.
* Subsequent processes will just use the populated shared region without
* requiring any further setup.
*/
int
shared_region_map_np(
struct proc *p,
struct shared_region_map_np_args *uap,
__unused int *retvalp)
{
int error;
kern_return_t kr;
int fd;
struct fileproc *fp;
struct vnode *vp, *root_vp;
struct vnode_attr va;
off_t fs;
memory_object_size_t file_size;
user_addr_t user_mappings;
struct shared_file_mapping_np *mappings;
#define SFM_MAX_STACK 8
struct shared_file_mapping_np stack_mappings[SFM_MAX_STACK];
unsigned int mappings_count;
vm_size_t mappings_size;
memory_object_control_t file_control;
struct vm_shared_region *shared_region;
SHARED_REGION_TRACE_DEBUG(
("shared_region: %p [%d(%s)] -> map\n",
current_thread(), p->p_pid, p->p_comm));
shared_region = NULL;
mappings_count = 0;
mappings_size = 0;
mappings = NULL;
fp = NULL;
vp = NULL;
/* get file descriptor for shared region cache file */
fd = uap->fd;
/* get file structure from file descriptor */
error = fp_lookup(p, fd, &fp, 0);
if (error) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map: "
"fd=%d lookup failed (error=%d)\n",
current_thread(), p->p_pid, p->p_comm, fd, error));
goto done;
}
/* make sure we're attempting to map a vnode */
if (fp->f_fglob->fg_type != DTYPE_VNODE) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map: "
"fd=%d not a vnode (type=%d)\n",
current_thread(), p->p_pid, p->p_comm,
fd, fp->f_fglob->fg_type));
error = EINVAL;
goto done;
}
/* we need at least read permission on the file */
if (! (fp->f_fglob->fg_flag & FREAD)) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map: "
"fd=%d not readable\n",
current_thread(), p->p_pid, p->p_comm, fd));
error = EPERM;
goto done;
}
/* get vnode from file structure */
error = vnode_getwithref((vnode_t) fp->f_fglob->fg_data);
if (error) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map: "
"fd=%d getwithref failed (error=%d)\n",
current_thread(), p->p_pid, p->p_comm, fd, error));
goto done;
}
vp = (struct vnode *) fp->f_fglob->fg_data;
/* make sure the vnode is a regular file */
if (vp->v_type != VREG) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"not a file (type=%d)\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, vp->v_type));
error = EINVAL;
goto done;
}
/* make sure vnode is on the process's root volume */
root_vp = p->p_fd->fd_rdir;
if (root_vp == NULL) {
root_vp = rootvnode;
}
if (vp->v_mount != root_vp->v_mount) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"not on process's root volume\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name));
error = EPERM;
goto done;
}
/* make sure vnode is owned by "root" */
VATTR_INIT(&va);
VATTR_WANTED(&va, va_uid);
error = vnode_getattr(vp, &va, vfs_context_current());
if (error) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"vnode_getattr(%p) failed (error=%d)\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, vp, error));
goto done;
}
if (va.va_uid != 0) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"owned by uid=%d instead of 0\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, va.va_uid));
error = EPERM;
goto done;
}
/* get vnode size */
error = vnode_size(vp, &fs, vfs_context_current());
if (error) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"vnode_size(%p) failed (error=%d)\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, vp, error));
goto done;
}
file_size = fs;
/* get the file's memory object handle */
file_control = ubc_getobject(vp, UBC_HOLDOBJECT);
if (file_control == MEMORY_OBJECT_CONTROL_NULL) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"no memory object\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name));
error = EINVAL;
goto done;
}
/* get the list of mappings the caller wants us to establish */
mappings_count = uap->count; /* number of mappings */
mappings_size = (vm_size_t) (mappings_count * sizeof (mappings[0]));
if (mappings_count == 0) {
SHARED_REGION_TRACE_INFO(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"no mappings\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name));
error = 0; /* no mappings: we're done ! */
goto done;
} else if (mappings_count <= SFM_MAX_STACK) {
mappings = &stack_mappings[0];
} else {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"too many mappings (%d)\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, mappings_count));
error = EINVAL;
goto done;
}
user_mappings = uap->mappings; /* the mappings, in user space */
error = copyin(user_mappings,
mappings,
mappings_size);
if (error) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"copyin(0x%llx, %d) failed (error=%d)\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, (uint64_t)user_mappings, mappings_count, error));
goto done;
}
/* get the process's shared region (setup in vm_map_exec()) */
shared_region = vm_shared_region_get(current_task());
if (shared_region == NULL) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"no shared region\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name));
goto done;
}
/* map the file into that shared region's submap */
kr = vm_shared_region_map_file(shared_region,
mappings_count,
mappings,
file_control,
file_size,
(void *) p->p_fd->fd_rdir);
if (kr != KERN_SUCCESS) {
SHARED_REGION_TRACE_ERROR(
("shared_region: %p [%d(%s)] map(%p:'%s'): "
"vm_shared_region_map_file() failed kr=0x%x\n",
current_thread(), p->p_pid, p->p_comm,
vp, vp->v_name, kr));
switch (kr) {
case KERN_INVALID_ADDRESS:
error = EFAULT;
break;
case KERN_PROTECTION_FAILURE:
error = EPERM;
break;
case KERN_NO_SPACE:
error = ENOMEM;
break;
case KERN_FAILURE:
case KERN_INVALID_ARGUMENT:
default:
error = EINVAL;
break;
}
goto done;
}
/*
* The mapping was successful. Let the buffer cache know
* that we've mapped that file with these protections. This
* prevents the vnode from getting recycled while it's mapped.
*/
(void) ubc_map(vp, VM_PROT_READ);
error = 0;
/* update the vnode's access time */
if (! (vnode_vfsvisflags(vp) & MNT_NOATIME)) {
VATTR_INIT(&va);
nanotime(&va.va_access_time);
VATTR_SET_ACTIVE(&va, va_access_time);
vnode_setattr(vp, &va, vfs_context_current());
}
if (p->p_flag & P_NOSHLIB) {
/* signal that this process is now using split libraries */
OSBitAndAtomic(~((uint32_t)P_NOSHLIB), (UInt32 *)&p->p_flag);
}
done:
if (vp != NULL) {
/*
* release the vnode...
* ubc_map() still holds it for us in the non-error case
*/
(void) vnode_put(vp);
vp = NULL;
}
if (fp != NULL) {
/* release the file descriptor */
fp_drop(p, fd, fp, 0);
fp = NULL;
}
if (shared_region != NULL) {
vm_shared_region_deallocate(shared_region);
}
SHARED_REGION_TRACE_DEBUG(
("shared_region: %p [%d(%s)] <- map\n",
current_thread(), p->p_pid, p->p_comm));
return error;
}
/*
* Write out process accounting information, on process exit.
* Data to be written out is specified in Leffler, et al.
* and are enumerated below. (They're also noted in the system
* "acct.h" header file.)
*/
int
acct_process(proc_t p)
{
struct acct an_acct;
struct rusage rup, *r;
struct timeval ut, st, tmp;
int t;
int error;
struct vnode *vp;
kauth_cred_t safecred;
struct session * sessp;
boolean_t fstate;
/* If accounting isn't enabled, don't bother */
vp = acctp;
if (vp == NULLVP)
return (0);
/*
* Get process accounting information.
*/
/* (1) The name of the command that ran */
bcopy(p->p_comm, an_acct.ac_comm, sizeof an_acct.ac_comm);
/* (2) The amount of user and system time that was used */
calcru(p, &ut, &st, NULL);
an_acct.ac_utime = encode_comp_t(ut.tv_sec, ut.tv_usec);
an_acct.ac_stime = encode_comp_t(st.tv_sec, st.tv_usec);
/* (3) The elapsed time the commmand ran (and its starting time) */
an_acct.ac_btime = p->p_start.tv_sec;
microtime(&tmp);
timevalsub(&tmp, &p->p_start);
an_acct.ac_etime = encode_comp_t(tmp.tv_sec, tmp.tv_usec);
/* (4) The average amount of memory used */
proc_lock(p);
rup = p->p_stats->p_ru;
proc_unlock(p);
r = &rup;
tmp = ut;
timevaladd(&tmp, &st);
t = tmp.tv_sec * hz + tmp.tv_usec / tick;
if (t)
an_acct.ac_mem = (r->ru_ixrss + r->ru_idrss + r->ru_isrss) / t;
else
an_acct.ac_mem = 0;
/* (5) The number of disk I/O operations done */
an_acct.ac_io = encode_comp_t(r->ru_inblock + r->ru_oublock, 0);
/* (6) The UID and GID of the process */
safecred = kauth_cred_proc_ref(p);
an_acct.ac_uid = safecred->cr_ruid;
an_acct.ac_gid = safecred->cr_rgid;
/* (7) The terminal from which the process was started */
sessp = proc_session(p);
if ((p->p_flag & P_CONTROLT) && (sessp != SESSION_NULL) && (sessp->s_ttyp != TTY_NULL)) {
fstate = thread_funnel_set(kernel_flock, TRUE);
an_acct.ac_tty = sessp->s_ttyp->t_dev;
(void) thread_funnel_set(kernel_flock, fstate);
}else
an_acct.ac_tty = NODEV;
if (sessp != SESSION_NULL)
session_rele(sessp);
/* (8) The boolean flags that tell how the process terminated, etc. */
an_acct.ac_flag = p->p_acflag;
/*
* Now, just write the accounting information to the file.
*/
if ((error = vnode_getwithref(vp)) == 0) {
error = vn_rdwr(UIO_WRITE, vp, (caddr_t)&an_acct, sizeof (an_acct),
(off_t)0, UIO_SYSSPACE32, IO_APPEND|IO_UNIT, safecred,
(int *)0, p);
vnode_put(vp);
}
kauth_cred_unref(&safecred);
return (error);
}
/*
* We have to carry on the locking protocol on the null layer vnodes
* as we progress through the tree. We also have to enforce read-only
* if this layer is mounted read-only.
*/
static int
null_lookup(struct vnop_lookup_args * ap)
{
struct componentname * cnp = ap->a_cnp;
struct vnode * dvp = ap->a_dvp;
struct vnode *vp, *ldvp, *lvp;
struct mount * mp;
struct null_mount * null_mp;
int error;
NULLFSDEBUG("%s parent: %p component: %.*s\n", __FUNCTION__, ap->a_dvp, cnp->cn_namelen, cnp->cn_nameptr);
mp = vnode_mount(dvp);
/* rename and delete are not allowed. this is a read only file system */
if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME || cnp->cn_nameiop == CREATE) {
return (EROFS);
}
null_mp = MOUNTTONULLMOUNT(mp);
lck_mtx_lock(&null_mp->nullm_lock);
if (nullfs_isspecialvp(dvp)) {
error = null_special_lookup(ap);
lck_mtx_unlock(&null_mp->nullm_lock);
return error;
}
lck_mtx_unlock(&null_mp->nullm_lock);
// . and .. handling
if (cnp->cn_nameptr[0] == '.') {
if (cnp->cn_namelen == 1) {
vp = dvp;
} else if (cnp->cn_namelen == 2 && cnp->cn_nameptr[1] == '.') {
/* mount point crossing is handled in null_special_lookup */
vp = vnode_parent(dvp);
} else {
goto notdot;
}
error = vp ? vnode_get(vp) : ENOENT;
if (error == 0) {
*ap->a_vpp = vp;
}
return error;
}
notdot:
ldvp = NULLVPTOLOWERVP(dvp);
vp = lvp = NULL;
/*
* Hold ldvp. The reference on it, owned by dvp, is lost in
* case of dvp reclamation.
*/
error = vnode_getwithref(ldvp);
if (error) {
return error;
}
error = VNOP_LOOKUP(ldvp, &lvp, cnp, ap->a_context);
vnode_put(ldvp);
if ((error == 0 || error == EJUSTRETURN) && lvp != NULL) {
if (ldvp == lvp) {
vp = dvp;
error = vnode_get(vp);
} else {
error = null_nodeget(mp, lvp, dvp, &vp, cnp, 0);
}
if (error == 0) {
*ap->a_vpp = vp;
}
}
/* if we got lvp, drop the iocount from VNOP_LOOKUP */
if (lvp != NULL) {
vnode_put(lvp);
}
return (error);
}
/*
* Mount null layer
*/
static int
nullfs_mount(struct mount * mp, __unused vnode_t devvp, user_addr_t user_data, vfs_context_t ctx)
{
int error = 0;
struct vnode *lowerrootvp = NULL, *vp = NULL;
struct vfsstatfs * sp = NULL;
struct null_mount * xmp = NULL;
char data[MAXPATHLEN];
size_t count;
struct vfs_attr vfa;
/* set defaults (arbitrary since this file system is readonly) */
uint32_t bsize = BLKDEV_IOSIZE;
size_t iosize = BLKDEV_IOSIZE;
uint64_t blocks = 4711 * 4711;
uint64_t bfree = 0;
uint64_t bavail = 0;
uint64_t bused = 4711;
uint64_t files = 4711;
uint64_t ffree = 0;
kauth_cred_t cred = vfs_context_ucred(ctx);
NULLFSDEBUG("nullfs_mount(mp = %p) %llx\n", (void *)mp, vfs_flags(mp));
if (vfs_flags(mp) & MNT_ROOTFS)
return (EOPNOTSUPP);
/*
* Update is a no-op
*/
if (vfs_isupdate(mp)) {
return ENOTSUP;
}
/* check entitlement */
if (!IOTaskHasEntitlement(current_task(), NULLFS_ENTITLEMENT)) {
return EPERM;
}
/*
* Get argument
*/
error = copyinstr(user_data, data, MAXPATHLEN - 1, &count);
if (error) {
NULLFSDEBUG("nullfs: error copying data form user %d\n", error);
goto error;
}
/* This could happen if the system is configured for 32 bit inodes instead of
* 64 bit */
if (count > MAX_MNT_FROM_LENGTH) {
error = EINVAL;
NULLFSDEBUG("nullfs: path to translocate too large for this system %d vs %d\n", count, MAX_MNT_FROM_LENGTH);
goto error;
}
error = vnode_lookup(data, 0, &lowerrootvp, ctx);
if (error) {
NULLFSDEBUG("lookup %s -> %d\n", data, error);
goto error;
}
/* lowervrootvp has an iocount after vnode_lookup, drop that for a usecount.
Keep this to signal what we want to keep around the thing we are mirroring.
Drop it in unmount.*/
error = vnode_ref(lowerrootvp);
vnode_put(lowerrootvp);
if (error)
{
// If vnode_ref failed, then null it out so it can't be used anymore in cleanup.
lowerrootvp = NULL;
goto error;
}
NULLFSDEBUG("mount %s\n", data);
MALLOC(xmp, struct null_mount *, sizeof(*xmp), M_TEMP, M_WAITOK | M_ZERO);
if (xmp == NULL) {
error = ENOMEM;
goto error;
}
/*
* Save reference to underlying FS
*/
xmp->nullm_lowerrootvp = lowerrootvp;
xmp->nullm_lowerrootvid = vnode_vid(lowerrootvp);
error = null_getnewvnode(mp, NULL, NULL, &vp, NULL, 1);
if (error) {
goto error;
}
/* vp has an iocount on it from vnode_create. drop that for a usecount. This
* is our root vnode so we drop the ref in unmount
*
* Assuming for now that because we created this vnode and we aren't finished mounting we can get a ref*/
vnode_ref(vp);
vnode_put(vp);
error = nullfs_init_lck(&xmp->nullm_lock);
if (error) {
goto error;
}
xmp->nullm_rootvp = vp;
/* read the flags the user set, but then ignore some of them, we will only
allow them if they are set on the lower file system */
uint64_t flags = vfs_flags(mp) & (~(MNT_IGNORE_OWNERSHIP | MNT_LOCAL));
uint64_t lowerflags = vfs_flags(vnode_mount(lowerrootvp)) & (MNT_LOCAL | MNT_QUARANTINE | MNT_IGNORE_OWNERSHIP | MNT_NOEXEC);
if (lowerflags) {
flags |= lowerflags;
}
/* force these flags */
flags |= (MNT_DONTBROWSE | MNT_MULTILABEL | MNT_NOSUID | MNT_RDONLY);
vfs_setflags(mp, flags);
vfs_setfsprivate(mp, xmp);
vfs_getnewfsid(mp);
vfs_setlocklocal(mp);
/* fill in the stat block */
sp = vfs_statfs(mp);
strlcpy(sp->f_mntfromname, data, MAX_MNT_FROM_LENGTH);
sp->f_flags = flags;
xmp->nullm_flags = NULLM_CASEINSENSITIVE; /* default to case insensitive */
error = nullfs_vfs_getlowerattr(vnode_mount(lowerrootvp), &vfa, ctx);
if (error == 0) {
if (VFSATTR_IS_SUPPORTED(&vfa, f_bsize)) {
bsize = vfa.f_bsize;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_iosize)) {
iosize = vfa.f_iosize;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_blocks)) {
blocks = vfa.f_blocks;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_bfree)) {
bfree = vfa.f_bfree;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_bavail)) {
bavail = vfa.f_bavail;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_bused)) {
bused = vfa.f_bused;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_files)) {
files = vfa.f_files;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_ffree)) {
ffree = vfa.f_ffree;
}
if (VFSATTR_IS_SUPPORTED(&vfa, f_capabilities)) {
if ((vfa.f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] & (VOL_CAP_FMT_CASE_SENSITIVE)) &&
(vfa.f_capabilities.valid[VOL_CAPABILITIES_FORMAT] & (VOL_CAP_FMT_CASE_SENSITIVE))) {
xmp->nullm_flags &= ~NULLM_CASEINSENSITIVE;
}
}
} else {
goto error;
}
sp->f_bsize = bsize;
sp->f_iosize = iosize;
sp->f_blocks = blocks;
sp->f_bfree = bfree;
sp->f_bavail = bavail;
sp->f_bused = bused;
sp->f_files = files;
sp->f_ffree = ffree;
/* Associate the mac label information from the mirrored filesystem with the
* mirror */
MAC_PERFORM(mount_label_associate, cred, vnode_mount(lowerrootvp), vfs_mntlabel(mp));
NULLFSDEBUG("nullfs_mount: lower %s, alias at %s\n", sp->f_mntfromname, sp->f_mntonname);
return (0);
error:
if (xmp) {
FREE(xmp, M_TEMP);
}
if (lowerrootvp) {
vnode_getwithref(lowerrootvp);
vnode_rele(lowerrootvp);
vnode_put(lowerrootvp);
}
if (vp) {
/* we made the root vnode but the mount is failed, so clean it up */
vnode_getwithref(vp);
vnode_rele(vp);
/* give vp back */
vnode_recycle(vp);
vnode_put(vp);
}
return error;
}
off_t DldIOShadowFile::expandFile( __in off_t lowTargetSize )
{
assert( preemption_enabled() );
off_t target;
off_t currentLastExpansion;
#if defined( DBG )
currentLastExpansion = DLD_IGNR_FSIZE;
#endif//DBG
//
// Make sure that we are root. Growing a file
// without zero filling the data is a security hole
// root would have access anyway so we'll allow it
//
//
// TO DO - is_suser is not supported for 64 bit kext
//
/*
#if defined(__i386__)
assert( is_suser() );
if( !is_suser() )
return this->lastExpansion;// an unsafe unprotected access to 64b value
#elif defined(__x86_64__)
#endif//defined(__x86_64__)
*/
//
// check that the limit has not been reached
//
if( DLD_IGNR_FSIZE != this->maxSize && this->lastExpansion == this->maxSize )
return this->maxSize;// a safe access to 64b value as the value can't change
this->LockExclusive();
{// start of the lock
off_t valueToAdd;
currentLastExpansion = this->lastExpansion;
//
// try to extend the file on 128 MB
// ( the same value is used as a maximum value for a mapping
// range in DldIOShadow::processFileWriteWQ )
//
valueToAdd = 0x8*0x1000*0x1000;
if( this->offset > this->lastExpansion )
target = this->offset + valueToAdd;
else
target = this->lastExpansion + valueToAdd;
if( target < lowTargetSize )
target = lowTargetSize;
if( DLD_IGNR_FSIZE != this->maxSize && target > this->maxSize )
target = this->maxSize;
}// end of the lock
this->UnLockExclusive();
if( target < lowTargetSize ){
//
// there is no point for extension as the goal won't be reached, fail the request
//
return currentLastExpansion;
}
assert( DLD_IGNR_FSIZE != currentLastExpansion && currentLastExpansion <= target );
//
// extend the file
//
vfs_context_t vfs_context;
int error;
error = vnode_getwithref( this->vnode );
assert( !error );
if( !error ){
struct vnode_attr va;
VATTR_INIT(&va);
VATTR_SET(&va, va_data_size, target);
va.va_vaflags =(IO_NOZEROFILL) & 0xffff;
vfs_context = vfs_context_create( NULL );
assert( vfs_context );
if( vfs_context ){
this->LockExclusive();
{// start of the lock
assert( currentLastExpansion <= this->lastExpansion );
//
// do not truncate the file incidentally!
//
if( target > this->offset ){
error = vnode_setattr( this->vnode, &va, vfs_context );
if( !error ){
this->lastExpansion = target;
currentLastExpansion = target;
}
} else {
//
// the file has been extended by the write operation
//
this->lastExpansion = this->offset;
currentLastExpansion = this->offset;
}// end if( target < this->offset )
}// end of the lock
this->UnLockExclusive();
vfs_context_rele( vfs_context );
DLD_DBG_MAKE_POINTER_INVALID( vfs_context );
}// end if( vfs_context )
vnode_put( this->vnode );
}// end if( !error )
assert( DLD_IGNR_FSIZE != currentLastExpansion );
return currentLastExpansion;
}
/*
* Search a pathname.
* This is a very central and rather complicated routine.
*
* The pathname is pointed to by ni_ptr and is of length ni_pathlen.
* The starting directory is taken from ni_startdir. The pathname is
* descended until done, or a symbolic link is encountered. The variable
* ni_more is clear if the path is completed; it is set to one if a
* symbolic link needing interpretation is encountered.
*
* The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
* whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it, the parent directory is returned
* locked. If flag has WANTPARENT or'ed into it, the parent directory is
* returned unlocked. Otherwise the parent directory is not returned. If
* the target of the pathname exists and LOCKLEAF is or'ed into the flag
* the target is returned locked, otherwise it is returned unlocked.
* When creating or renaming and LOCKPARENT is specified, the target may not
* be ".". When deleting and LOCKPARENT is specified, the target may be ".".
*
* Overall outline of lookup:
*
* dirloop:
* identify next component of name at ndp->ni_ptr
* handle degenerate case where name is null string
* if .. and crossing mount points and on mounted filesys, find parent
* call VNOP_LOOKUP routine for next component name
* directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
* component vnode returned in ni_vp (if it exists), locked.
* if result vnode is mounted on and crossing mount points,
* find mounted on vnode
* if more components of name, do next level at dirloop
* return the answer in ni_vp, locked if LOCKLEAF set
* if LOCKPARENT set, return locked parent in ni_dvp
* if WANTPARENT set, return unlocked parent in ni_dvp
*
* Returns: 0 Success
* ENOENT No such file or directory
* EBADF Bad file descriptor
* ENOTDIR Not a directory
* EROFS Read-only file system [CREATE]
* EISDIR Is a directory [CREATE]
* cache_lookup_path:ERECYCLE (vnode was recycled from underneath us, redrive lookup again)
* vnode_authorize:EROFS
* vnode_authorize:EACCES
* vnode_authorize:EPERM
* vnode_authorize:???
* VNOP_LOOKUP:ENOENT No such file or directory
* VNOP_LOOKUP:EJUSTRETURN Restart system call (INTERNAL)
* VNOP_LOOKUP:???
* VFS_ROOT:ENOTSUP
* VFS_ROOT:ENOENT
* VFS_ROOT:???
*/
int
lookup(struct nameidata *ndp)
{
char *cp; /* pointer into pathname argument */
vnode_t tdp; /* saved dp */
vnode_t dp; /* the directory we are searching */
mount_t mp; /* mount table entry */
int docache = 1; /* == 0 do not cache last component */
int wantparent; /* 1 => wantparent or lockparent flag */
int rdonly; /* lookup read-only flag bit */
int trailing_slash = 0;
int dp_authorized = 0;
int error = 0;
struct componentname *cnp = &ndp->ni_cnd;
vfs_context_t ctx = cnp->cn_context;
int mounted_on_depth = 0;
int dont_cache_mp = 0;
vnode_t mounted_on_dp = NULLVP;
int current_mount_generation = 0;
int vbusyflags = 0;
int nc_generation = 0;
vnode_t last_dp = NULLVP;
/*
* Setup: break out flag bits into variables.
*/
if (cnp->cn_flags & (NOCACHE | DOWHITEOUT)) {
if ((cnp->cn_flags & NOCACHE) || (cnp->cn_nameiop == DELETE))
docache = 0;
}
wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
rdonly = cnp->cn_flags & RDONLY;
cnp->cn_flags &= ~ISSYMLINK;
cnp->cn_consume = 0;
dp = ndp->ni_startdir;
ndp->ni_startdir = NULLVP;
if ((cnp->cn_flags & CN_NBMOUNTLOOK) != 0)
vbusyflags = LK_NOWAIT;
cp = cnp->cn_nameptr;
if (*cp == '\0') {
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
goto emptyname;
}
dirloop:
ndp->ni_vp = NULLVP;
if ( (error = cache_lookup_path(ndp, cnp, dp, ctx, &trailing_slash, &dp_authorized, last_dp)) ) {
dp = NULLVP;
goto bad;
}
if ((cnp->cn_flags & ISLASTCN)) {
if (docache)
cnp->cn_flags |= MAKEENTRY;
} else
cnp->cn_flags |= MAKEENTRY;
dp = ndp->ni_dvp;
if (ndp->ni_vp != NULLVP) {
/*
* cache_lookup_path returned a non-NULL ni_vp then,
* we're guaranteed that the dp is a VDIR, it's
* been authorized, and vp is not ".."
*
* make sure we don't try to enter the name back into
* the cache if this vp is purged before we get to that
* check since we won't have serialized behind whatever
* activity is occurring in the FS that caused the purge
*/
if (dp != NULLVP)
nc_generation = dp->v_nc_generation - 1;
goto returned_from_lookup_path;
}
/*
* Handle "..": two special cases.
* 1. If at root directory (e.g. after chroot)
* or at absolute root directory
* then ignore it so can't get out.
* 2. If this vnode is the root of a mounted
* filesystem, then replace it with the
* vnode which was mounted on so we take the
* .. in the other file system.
*/
if ( (cnp->cn_flags & ISDOTDOT) ) {
for (;;) {
if (dp == ndp->ni_rootdir || dp == rootvnode) {
ndp->ni_dvp = dp;
ndp->ni_vp = dp;
/*
* we're pinned at the root
* we've already got one reference on 'dp'
* courtesy of cache_lookup_path... take
* another one for the ".."
* if we fail to get the new reference, we'll
* drop our original down in 'bad'
*/
if ( (vnode_get(dp)) ) {
error = ENOENT;
goto bad;
}
goto nextname;
}
if ((dp->v_flag & VROOT) == 0 ||
(cnp->cn_flags & NOCROSSMOUNT))
break;
if (dp->v_mount == NULL) { /* forced umount */
error = EBADF;
goto bad;
}
tdp = dp;
dp = tdp->v_mount->mnt_vnodecovered;
vnode_put(tdp);
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
dp_authorized = 0;
}
}
/*
* We now have a segment name to search for, and a directory to search.
*/
unionlookup:
ndp->ni_vp = NULLVP;
if (dp->v_type != VDIR) {
error = ENOTDIR;
goto lookup_error;
}
if ( (cnp->cn_flags & DONOTAUTH) != DONOTAUTH ) {
if (!dp_authorized) {
error = vnode_authorize(dp, NULL, KAUTH_VNODE_SEARCH, ctx);
if (error)
goto lookup_error;
}
#if CONFIG_MACF
error = mac_vnode_check_lookup(ctx, dp, cnp);
if (error)
goto lookup_error;
#endif /* CONFIG_MACF */
}
nc_generation = dp->v_nc_generation;
if ( (error = VNOP_LOOKUP(dp, &ndp->ni_vp, cnp, ctx)) ) {
lookup_error:
if ((error == ENOENT) &&
(dp->v_flag & VROOT) && (dp->v_mount != NULL) &&
(dp->v_mount->mnt_flag & MNT_UNION)) {
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(dp, NULL);
}
tdp = dp;
dp = tdp->v_mount->mnt_vnodecovered;
vnode_put(tdp);
if ( (vnode_getwithref(dp)) ) {
dp = NULLVP;
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
dp_authorized = 0;
goto unionlookup;
}
if (error != EJUSTRETURN)
goto bad;
if (ndp->ni_vp != NULLVP)
panic("leaf should be empty");
/*
* If creating and at end of pathname, then can consider
* allowing file to be created.
*/
if (rdonly) {
error = EROFS;
goto bad;
}
if ((cnp->cn_flags & ISLASTCN) && trailing_slash && !(cnp->cn_flags & WILLBEDIR)) {
error = ENOENT;
goto bad;
}
/*
* We return with ni_vp NULL to indicate that the entry
* doesn't currently exist, leaving a pointer to the
* referenced directory vnode in ndp->ni_dvp.
*/
if (cnp->cn_flags & SAVESTART) {
if ( (vnode_get(ndp->ni_dvp)) ) {
error = ENOENT;
goto bad;
}
ndp->ni_startdir = ndp->ni_dvp;
}
if (!wantparent)
vnode_put(ndp->ni_dvp);
if (kdebug_enable)
kdebug_lookup(ndp->ni_dvp, cnp);
return (0);
}
returned_from_lookup_path:
dp = ndp->ni_vp;
/*
* Take into account any additional components consumed by
* the underlying filesystem.
*/
if (cnp->cn_consume > 0) {
cnp->cn_nameptr += cnp->cn_consume;
ndp->ni_next += cnp->cn_consume;
ndp->ni_pathlen -= cnp->cn_consume;
cnp->cn_consume = 0;
} else {
if (dp->v_name == NULL || dp->v_parent == NULLVP) {
int isdot_or_dotdot;
int update_flags = 0;
isdot_or_dotdot = (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') || (cnp->cn_flags & ISDOTDOT);
if (isdot_or_dotdot == 0) {
if (dp->v_name == NULL)
update_flags |= VNODE_UPDATE_NAME;
if (ndp->ni_dvp != NULLVP && dp->v_parent == NULLVP)
update_flags |= VNODE_UPDATE_PARENT;
if (update_flags)
vnode_update_identity(dp, ndp->ni_dvp, cnp->cn_nameptr, cnp->cn_namelen, cnp->cn_hash, update_flags);
}
}
if ( (cnp->cn_flags & MAKEENTRY) && (dp->v_flag & VNCACHEABLE) && LIST_FIRST(&dp->v_nclinks) == NULL) {
/*
* missing from name cache, but should
* be in it... this can happen if volfs
* causes the vnode to be created or the
* name cache entry got recycled but the
* vnode didn't...
* check to make sure that ni_dvp is valid
* cache_lookup_path may return a NULL
* do a quick check to see if the generation of the
* directory matches our snapshot... this will get
* rechecked behind the name cache lock, but if it
* already fails to match, no need to go any further
*/
if (ndp->ni_dvp != NULLVP && (nc_generation == ndp->ni_dvp->v_nc_generation))
cache_enter_with_gen(ndp->ni_dvp, dp, cnp, nc_generation);
}
}
mounted_on_dp = dp;
mounted_on_depth = 0;
dont_cache_mp = 0;
current_mount_generation = mount_generation;
/*
* Check to see if the vnode has been mounted on...
* if so find the root of the mounted file system.
*/
check_mounted_on:
if ((dp->v_type == VDIR) && dp->v_mountedhere &&
((cnp->cn_flags & NOCROSSMOUNT) == 0)) {
vnode_lock(dp);
if ((dp->v_type == VDIR) && (mp = dp->v_mountedhere)) {
struct uthread *uth = (struct uthread *)get_bsdthread_info(current_thread());
mp->mnt_crossref++;
vnode_unlock(dp);
if (vfs_busy(mp, vbusyflags)) {
mount_dropcrossref(mp, dp, 0);
if (vbusyflags == LK_NOWAIT) {
error = ENOENT;
goto bad2;
}
goto check_mounted_on;
}
/*
* XXX - if this is the last component of the
* pathname, and it's either not a lookup operation
* or the NOTRIGGER flag is set for the operation,
* set a uthread flag to let VFS_ROOT() for autofs
* know it shouldn't trigger a mount.
*/
if ((cnp->cn_flags & ISLASTCN) &&
(cnp->cn_nameiop != LOOKUP ||
(cnp->cn_flags & NOTRIGGER))) {
uth->uu_notrigger = 1;
dont_cache_mp = 1;
}
error = VFS_ROOT(mp, &tdp, ctx);
/* XXX - clear the uthread flag */
uth->uu_notrigger = 0;
/*
* mount_dropcrossref does a vnode_put
* on dp if the 3rd arg is non-zero
*/
mount_dropcrossref(mp, dp, 1);
dp = NULL;
vfs_unbusy(mp);
if (error) {
goto bad2;
}
ndp->ni_vp = dp = tdp;
mounted_on_depth++;
goto check_mounted_on;
}
vnode_unlock(dp);
}
#if CONFIG_MACF
if (vfs_flags(vnode_mount(dp)) & MNT_MULTILABEL) {
error = vnode_label(vnode_mount(dp), NULL, dp, NULL,
VNODE_LABEL_NEEDREF, ctx);
if (error)
goto bad2;
}
#endif
if (mounted_on_depth && !dont_cache_mp) {
mp = mounted_on_dp->v_mountedhere;
if (mp) {
mount_lock(mp);
mp->mnt_realrootvp_vid = dp->v_id;
mp->mnt_realrootvp = dp;
mp->mnt_generation = current_mount_generation;
mount_unlock(mp);
}
}
/*
* Check for symbolic link
*/
if ((dp->v_type == VLNK) &&
((cnp->cn_flags & FOLLOW) || trailing_slash || *ndp->ni_next == '/')) {
cnp->cn_flags |= ISSYMLINK;
return (0);
}
/*
* Check for bogus trailing slashes.
*/
if (trailing_slash) {
if (dp->v_type != VDIR) {
error = ENOTDIR;
goto bad2;
}
trailing_slash = 0;
}
nextname:
/*
* Not a symbolic link. If more pathname,
* continue at next component, else return.
*/
if (*ndp->ni_next == '/') {
cnp->cn_nameptr = ndp->ni_next + 1;
ndp->ni_pathlen--;
while (*cnp->cn_nameptr == '/') {
cnp->cn_nameptr++;
ndp->ni_pathlen--;
}
vnode_put(ndp->ni_dvp);
cp = cnp->cn_nameptr;
if (*cp == '\0')
goto emptyname;
/*
* cache_lookup_path is now responsible for dropping io ref on dp
* when it is called again in the dirloop. This ensures we hold
* a ref on dp until we complete the next round of lookup.
*/
last_dp = dp;
goto dirloop;
}
/*
* Disallow directory write attempts on read-only file systems.
*/
if (rdonly &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME)) {
error = EROFS;
goto bad2;
}
if (cnp->cn_flags & SAVESTART) {
/*
* note that we already hold a reference
* on both dp and ni_dvp, but for some reason
* can't get another one... in this case we
* need to do vnode_put on dp in 'bad2'
*/
if ( (vnode_get(ndp->ni_dvp)) ) {
error = ENOENT;
goto bad2;
}
ndp->ni_startdir = ndp->ni_dvp;
}
if (!wantparent && ndp->ni_dvp) {
vnode_put(ndp->ni_dvp);
ndp->ni_dvp = NULLVP;
}
if (cnp->cn_flags & AUDITVNPATH1)
AUDIT_ARG(vnpath, dp, ARG_VNODE1);
else if (cnp->cn_flags & AUDITVNPATH2)
AUDIT_ARG(vnpath, dp, ARG_VNODE2);
#if NAMEDRSRCFORK
/*
* Caller wants the resource fork.
*/
if ((cnp->cn_flags & CN_WANTSRSRCFORK) && (dp != NULLVP)) {
vnode_t svp = NULLVP;
enum nsoperation nsop;
if (dp->v_type != VREG) {
error = ENOENT;
goto bad2;
}
switch (cnp->cn_nameiop) {
case DELETE:
nsop = NS_DELETE;
break;
case CREATE:
nsop = NS_CREATE;
break;
case LOOKUP:
/* Make sure our lookup of "/..namedfork/rsrc" is allowed. */
if (cnp->cn_flags & CN_ALLOWRSRCFORK) {
nsop = NS_OPEN;
} else {
error = EPERM;
goto bad2;
}
break;
default:
error = EPERM;
goto bad2;
}
/* Ask the file system for the resource fork. */
error = vnode_getnamedstream(dp, &svp, XATTR_RESOURCEFORK_NAME, nsop, 0, ctx);
/* During a create, it OK for stream vnode to be missing. */
if (error == ENOATTR || error == ENOENT) {
error = (nsop == NS_CREATE) ? 0 : ENOENT;
}
if (error) {
goto bad2;
}
/* The "parent" of the stream is the file. */
if (wantparent) {
if (ndp->ni_dvp) {
if (ndp->ni_cnd.cn_flags & FSNODELOCKHELD) {
ndp->ni_cnd.cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
vnode_put(ndp->ni_dvp);
}
ndp->ni_dvp = dp;
} else {
vnode_put(dp);
}
ndp->ni_vp = dp = svp; /* on create this may be null */
/* Restore the truncated pathname buffer (for audits). */
if (ndp->ni_pathlen == 1 && ndp->ni_next[0] == '\0') {
ndp->ni_next[0] = '/';
}
cnp->cn_flags &= ~MAKEENTRY;
}
#endif
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (0);
emptyname:
cnp->cn_namelen = 0;
/*
* A degenerate name (e.g. / or "") which is a way of
* talking about a directory, e.g. like "/." or ".".
*/
if (dp->v_type != VDIR) {
error = ENOTDIR;
goto bad;
}
if (cnp->cn_nameiop != LOOKUP) {
error = EISDIR;
goto bad;
}
if (wantparent) {
/*
* note that we already hold a reference
* on dp, but for some reason can't
* get another one... in this case we
* need to do vnode_put on dp in 'bad'
*/
if ( (vnode_get(dp)) ) {
error = ENOENT;
goto bad;
}
ndp->ni_dvp = dp;
}
cnp->cn_flags &= ~ISDOTDOT;
cnp->cn_flags |= ISLASTCN;
ndp->ni_next = cp;
ndp->ni_vp = dp;
if (cnp->cn_flags & AUDITVNPATH1)
AUDIT_ARG(vnpath, dp, ARG_VNODE1);
else if (cnp->cn_flags & AUDITVNPATH2)
AUDIT_ARG(vnpath, dp, ARG_VNODE2);
if (cnp->cn_flags & SAVESTART)
panic("lookup: SAVESTART");
return (0);
bad2:
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
if (ndp->ni_dvp)
vnode_put(ndp->ni_dvp);
if (dp)
vnode_put(dp);
ndp->ni_vp = NULLVP;
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (error);
bad:
if ((cnp->cn_flags & FSNODELOCKHELD)) {
cnp->cn_flags &= ~FSNODELOCKHELD;
unlock_fsnode(ndp->ni_dvp, NULL);
}
if (dp)
vnode_put(dp);
ndp->ni_vp = NULLVP;
if (kdebug_enable)
kdebug_lookup(dp, cnp);
return (error);
}
