static int
p9fs_mount(struct mount *mp)
{
struct p9fsmount *p9mp;
struct p9fs_session *p9s;
int error;
error = EINVAL;
if (vfs_filteropt(mp->mnt_optnew, p9_opts))
goto out;
if (mp->mnt_flag & MNT_UPDATE)
return (p9fs_mount_parse_opts(mp));
/* Allocate and initialize the private mount structure. */
p9mp = malloc(sizeof (struct p9fsmount), M_P9MNT, M_WAITOK | M_ZERO);
mp->mnt_data = p9mp;
p9mp->p9_mountp = mp;
p9fs_init_session(&p9mp->p9_session);
p9s = &p9mp->p9_session;
p9s->p9s_mount = mp;
error = p9fs_mount_parse_opts(mp);
if (error != 0)
goto out;
error = p9fs_connect(mp);
if (error != 0) {
goto out;
}
/* Negotiate with the remote service. XXX: Add auth call. */
error = p9fs_client_version(p9s);
if (error == 0) {
/* Initialize the root vnode just before attaching. */
struct vnode *vp, *ivp;
struct p9fs_node *np = &p9s->p9s_rootnp;
np->p9n_fid = ROOTFID;
np->p9n_session = p9s;
error = getnewvnode("p9fs", mp, &p9fs_vnops, &vp);
if (error == 0) {
vn_lock(vp, LK_EXCLUSIVE);
error = insmntque(vp, mp);
}
ivp = NULL;
if (error == 0)
error = vfs_hash_insert(vp, ROOTFID, LK_EXCLUSIVE,
curthread, &ivp, NULL, NULL);
if (error == 0 && ivp != NULL)
error = EBUSY;
if (error == 0) {
np->p9n_vnode = vp;
vp->v_data = np;
vp->v_type = VDIR;
vp->v_vflag |= VV_ROOT;
VOP_UNLOCK(vp, 0);
}
}
if (error == 0)
error = p9fs_client_attach(p9s);
if (error == 0)
p9s->p9s_state = P9S_RUNNING;
out:
if (error != 0)
(void) p9fs_unmount(mp, MNT_FORCE);
return (error);
}
int
nfs_nget(struct mount *mntp, nfsfh_t *fhp, int fhsize, struct nfsnode **npp)
{
struct nfsnode *np, *np2;
struct nfsnodehashhead *nhpp;
struct vnode *vp;
int error;
int lkflags;
struct nfsmount *nmp;
/*
* Calculate nfs mount point and figure out whether the rslock should
* be interruptable or not.
*/
nmp = VFSTONFS(mntp);
if (nmp->nm_flag & NFSMNT_INT)
lkflags = LK_PCATCH;
else
lkflags = 0;
lwkt_gettoken(&nfsnhash_token);
retry:
nhpp = NFSNOHASH(fnv_32_buf(fhp->fh_bytes, fhsize, FNV1_32_INIT));
loop:
for (np = nhpp->lh_first; np; np = np->n_hash.le_next) {
if (mntp != NFSTOV(np)->v_mount || np->n_fhsize != fhsize ||
bcmp((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize)) {
continue;
}
vp = NFSTOV(np);
if (vget(vp, LK_EXCLUSIVE))
goto loop;
for (np = nhpp->lh_first; np; np = np->n_hash.le_next) {
if (mntp == NFSTOV(np)->v_mount &&
np->n_fhsize == fhsize &&
bcmp((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize) == 0
) {
break;
}
}
if (np == NULL || NFSTOV(np) != vp) {
vput(vp);
goto loop;
}
*npp = np;
lwkt_reltoken(&nfsnhash_token);
return(0);
}
/*
* Obtain a lock to prevent a race condition if the getnewvnode()
* or MALLOC() below happens to block.
*/
if (lockmgr(&nfsnhash_lock, LK_EXCLUSIVE | LK_SLEEPFAIL))
goto loop;
/*
* Allocate before getnewvnode since doing so afterward
* might cause a bogus v_data pointer to get dereferenced
* elsewhere if objcache should block.
*/
np = objcache_get(nfsnode_objcache, M_WAITOK);
error = getnewvnode(VT_NFS, mntp, &vp, 0, 0);
if (error) {
lockmgr(&nfsnhash_lock, LK_RELEASE);
*npp = NULL;
objcache_put(nfsnode_objcache, np);
lwkt_reltoken(&nfsnhash_token);
return (error);
}
/*
* Initialize most of (np).
*/
bzero(np, sizeof (*np));
if (fhsize > NFS_SMALLFH) {
MALLOC(np->n_fhp, nfsfh_t *, fhsize, M_NFSBIGFH, M_WAITOK);
} else {
/*
* Look up a vnode/nfsnode by file handle.
* Callers must check for mount points!!
* In all cases, a pointer to a
* nfsnode structure is returned.
*/
int
nfs_nget(struct mount *mntp, nfsfh_t *fhp, int fhsize, struct nfsnode **npp, int flags)
{
struct thread *td = curthread; /* XXX */
struct nfsnode *np;
struct vnode *vp;
struct vnode *nvp;
int error;
u_int hash;
struct nfsmount *nmp;
struct nfs_vncmp ncmp;
nmp = VFSTONFS(mntp);
*npp = NULL;
hash = fnv_32_buf(fhp->fh_bytes, fhsize, FNV1_32_INIT);
ncmp.fhsize = fhsize;
ncmp.fh = fhp;
error = vfs_hash_get(mntp, hash, flags,
td, &nvp, nfs_vncmpf, &ncmp);
if (error)
return (error);
if (nvp != NULL) {
*npp = VTONFS(nvp);
return (0);
}
/*
* Allocate before getnewvnode since doing so afterward
* might cause a bogus v_data pointer to get dereferenced
* elsewhere if zalloc should block.
*/
np = uma_zalloc(nfsnode_zone, M_WAITOK | M_ZERO);
error = getnewvnode("nfs", mntp, &nfs_vnodeops, &nvp);
if (error) {
uma_zfree(nfsnode_zone, np);
return (error);
}
vp = nvp;
vp->v_bufobj.bo_ops = &buf_ops_nfs;
vp->v_data = np;
np->n_vnode = vp;
/*
* Initialize the mutex even if the vnode is going to be a loser.
* This simplifies the logic in reclaim, which can then unconditionally
* destroy the mutex (in the case of the loser, or if hash_insert happened
* to return an error no special casing is needed).
*/
mtx_init(&np->n_mtx, "NFSnode lock", NULL, MTX_DEF);
/*
* NFS supports recursive and shared locking.
*/
lockmgr(vp->v_vnlock, LK_EXCLUSIVE | LK_NOWITNESS, NULL);
VN_LOCK_AREC(vp);
VN_LOCK_ASHARE(vp);
if (fhsize > NFS_SMALLFH) {
np->n_fhp = malloc(fhsize, M_NFSBIGFH, M_WAITOK);
} else
np->n_fhp = &np->n_fh;
bcopy((caddr_t)fhp, (caddr_t)np->n_fhp, fhsize);
np->n_fhsize = fhsize;
error = insmntque(vp, mntp);
if (error != 0) {
*npp = NULL;
if (np->n_fhsize > NFS_SMALLFH) {
free((caddr_t)np->n_fhp, M_NFSBIGFH);
}
mtx_destroy(&np->n_mtx);
uma_zfree(nfsnode_zone, np);
return (error);
}
error = vfs_hash_insert(vp, hash, flags,
td, &nvp, nfs_vncmpf, &ncmp);
if (error)
return (error);
if (nvp != NULL) {
*npp = VTONFS(nvp);
/* vfs_hash_insert() vput()'s the losing vnode */
return (0);
}
*npp = np;
return (0);
}
/* ARGSUSED */
int
mfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct vnode *devvp;
struct mfs_args *args = data;
struct ufsmount *ump;
struct fs *fs;
struct mfsnode *mfsp;
struct proc *p;
int flags, error = 0;
if (args == NULL)
return EINVAL;
if (*data_len < sizeof *args)
return EINVAL;
p = l->l_proc;
if (mp->mnt_flag & MNT_GETARGS) {
struct vnode *vp;
ump = VFSTOUFS(mp);
if (ump == NULL)
return EIO;
vp = ump->um_devvp;
if (vp == NULL)
return EIO;
mfsp = VTOMFS(vp);
if (mfsp == NULL)
return EIO;
args->fspec = NULL;
args->base = mfsp->mfs_baseoff;
args->size = mfsp->mfs_size;
*data_len = sizeof *args;
return 0;
}
/*
* XXX turn off async to avoid hangs when writing lots of data.
* the problem is that MFS needs to allocate pages to clean pages,
* so if we wait until the last minute to clean pages then there
* may not be any pages available to do the cleaning.
* ... and since the default partially-synchronous mode turns out
* to not be sufficient under heavy load, make it full synchronous.
*/
mp->mnt_flag &= ~MNT_ASYNC;
mp->mnt_flag |= MNT_SYNCHRONOUS;
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
ump = VFSTOUFS(mp);
fs = ump->um_fs;
if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = ffs_flushfiles(mp, flags, l);
if (error)
return (error);
}
if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR))
fs->fs_ronly = 0;
if (args->fspec == NULL)
return EINVAL;
return (0);
}
error = getnewvnode(VT_MFS, NULL, mfs_vnodeop_p, NULL, &devvp);
if (error)
return (error);
devvp->v_vflag |= VV_MPSAFE;
devvp->v_type = VBLK;
spec_node_init(devvp, makedev(255, mfs_minor));
mfs_minor++;
mfsp = kmem_alloc(sizeof(*mfsp), KM_SLEEP);
devvp->v_data = mfsp;
mfsp->mfs_baseoff = args->base;
mfsp->mfs_size = args->size;
mfsp->mfs_vnode = devvp;
mfsp->mfs_proc = p;
mfsp->mfs_shutdown = 0;
cv_init(&mfsp->mfs_cv, "mfsidl");
mfsp->mfs_refcnt = 1;
bufq_alloc(&mfsp->mfs_buflist, "fcfs", 0);
if ((error = ffs_mountfs(devvp, mp, l)) != 0) {
mfsp->mfs_shutdown = 1;
vrele(devvp);
return (error);
}
ump = VFSTOUFS(mp);
fs = ump->um_fs;
error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
if (error)
return error;
(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
sizeof(fs->fs_fsmnt));
fs->fs_fsmnt[sizeof(fs->fs_fsmnt) - 1] = '\0';
/* XXX: cleanup on error */
return 0;
}
/*
* VFS Operations.
*
* mount system call
*/
int
mfs_mount(struct mount *mp, const char *path, void *data,
struct nameidata *ndp, struct proc *p)
{
struct vnode *devvp;
struct mfs_args args;
struct ufsmount *ump;
struct fs *fs;
struct mfsnode *mfsp;
size_t size;
int flags, error;
error = copyin(data, (caddr_t)&args, sizeof (struct mfs_args));
if (error)
return (error);
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
ump = VFSTOUFS(mp);
fs = ump->um_fs;
if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = ffs_flushfiles(mp, flags, p);
if (error)
return (error);
}
if (fs->fs_ronly && (mp->mnt_flag & MNT_WANTRDWR))
fs->fs_ronly = 0;
#ifdef EXPORTMFS
if (args.fspec == 0)
return (vfs_export(mp, &ump->um_export,
&args.export_info));
#endif
return (0);
}
error = getnewvnode(VT_MFS, NULL, &mfs_vops, &devvp);
if (error)
return (error);
devvp->v_type = VBLK;
if (checkalias(devvp, makedev(255, mfs_minor), (struct mount *)0))
panic("mfs_mount: dup dev");
mfs_minor++;
mfsp = malloc(sizeof *mfsp, M_MFSNODE, M_WAITOK);
mfsp->mfs_dying = 0;
devvp->v_data = mfsp;
mfsp->mfs_baseoff = args.base;
mfsp->mfs_size = args.size;
mfsp->mfs_vnode = devvp;
mfsp->mfs_pid = p->p_pid;
bufq_init(&mfsp->mfs_bufq, BUFQ_FIFO);
if ((error = ffs_mountfs(devvp, mp, p)) != 0) {
mfsp->mfs_dying = 1;
vrele(devvp);
return (error);
}
ump = VFSTOUFS(mp);
fs = ump->um_fs;
(void) copyinstr(path, fs->fs_fsmnt, sizeof(fs->fs_fsmnt) - 1, &size);
bzero(fs->fs_fsmnt + size, sizeof(fs->fs_fsmnt) - size);
bcopy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname, MNAMELEN);
(void) copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1,
&size);
bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
bcopy(&args, &mp->mnt_stat.mount_info.mfs_args, sizeof(args));
return (0);
}
int
fdesc_allocvp(fdntype ftype, unsigned fd_fd, int ix, struct mount *mp,
struct vnode **vpp)
{
struct fdescmount *fmp;
struct fdhashhead *fc;
struct fdescnode *fd, *fd2;
struct vnode *vp, *vp2;
struct thread *td;
int error = 0;
td = curthread;
fc = FD_NHASH(ix);
loop:
mtx_lock(&fdesc_hashmtx);
/*
* If a forced unmount is progressing, we need to drop it. The flags are
* protected by the hashmtx.
*/
fmp = (struct fdescmount *)mp->mnt_data;
if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) {
mtx_unlock(&fdesc_hashmtx);
return (-1);
}
LIST_FOREACH(fd, fc, fd_hash) {
if (fd->fd_ix == ix && fd->fd_vnode->v_mount == mp) {
/* Get reference to vnode in case it's being free'd */
vp = fd->fd_vnode;
VI_LOCK(vp);
mtx_unlock(&fdesc_hashmtx);
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td))
goto loop;
*vpp = vp;
return (0);
}
}
mtx_unlock(&fdesc_hashmtx);
fd = malloc(sizeof(struct fdescnode), M_TEMP, M_WAITOK);
error = getnewvnode("fdescfs", mp, &fdesc_vnodeops, &vp);
if (error) {
free(fd, M_TEMP);
return (error);
}
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
vp->v_data = fd;
fd->fd_vnode = vp;
fd->fd_type = ftype;
fd->fd_fd = fd_fd;
fd->fd_ix = ix;
error = insmntque1(vp, mp, fdesc_insmntque_dtr, NULL);
if (error != 0) {
*vpp = NULLVP;
return (error);
}
/* Make sure that someone didn't beat us when inserting the vnode. */
mtx_lock(&fdesc_hashmtx);
/*
* If a forced unmount is progressing, we need to drop it. The flags are
* protected by the hashmtx.
*/
fmp = (struct fdescmount *)mp->mnt_data;
if (fmp == NULL || fmp->flags & FMNT_UNMOUNTF) {
mtx_unlock(&fdesc_hashmtx);
vgone(vp);
vput(vp);
*vpp = NULLVP;
return (-1);
}
LIST_FOREACH(fd2, fc, fd_hash) {
if (fd2->fd_ix == ix && fd2->fd_vnode->v_mount == mp) {
/* Get reference to vnode in case it's being free'd */
vp2 = fd2->fd_vnode;
VI_LOCK(vp2);
mtx_unlock(&fdesc_hashmtx);
error = vget(vp2, LK_EXCLUSIVE | LK_INTERLOCK, td);
/* Someone beat us, dec use count and wait for reclaim */
vgone(vp);
vput(vp);
/* If we didn't get it, return no vnode. */
if (error)
vp2 = NULLVP;
*vpp = vp2;
return (error);
}
}
/* If we came here, we can insert it safely. */
LIST_INSERT_HEAD(fc, fd, fd_hash);
mtx_unlock(&fdesc_hashmtx);
*vpp = vp;
return (0);
}
/*
* If xtaf_deget() succeeds it returns with the gotten denode locked().
*
* pmp - address of xtafmount structure of the filesystem containing
* the denode of interest. The address of the xtafmount structure
* is used.
* dirclust - which cluster bp contains, if dirclust is 0 (root directory)
* diroffset is relative to the beginning of the root directory,
* otherwise it is cluster relative. This is the cluster this
directory entry came from.
* diroffset - offset past begin of cluster of denode we want
* depp - returns the address of the gotten denode.
*/
int
xtaf_deget(struct xtafmount *pmp, u_long dirclust, u_long diroffset,
struct denode **depp)
{
int error;
uint32_t inode;
struct mount *mntp = pmp->pm_mountp;
struct direntry *ep;
struct denode *ldep;
struct vnode *nvp, *xvp;
struct buf *bp;
#ifdef XTAF_DEBUG
printf("xtaf_deget(pmp %p, dirclust %lx, diroffset %lx, depp %p)\n",
pmp, dirclust, diroffset, depp);
#endif
/*
* See if the denode is in the denode cache. Use the location of
* the directory entry to compute the hash value. For subdir use
* address of the first dir entry. For root dir use cluster XTAFROOT,
* offset XTAFROOT_OFS
*/
inode = (uint32_t)pmp->pm_bpcluster * dirclust + diroffset;
error = vfs_hash_get(mntp, inode, LK_EXCLUSIVE, curthread, &nvp,
de_vncmpf, &inode);
if (error)
return (error);
if (nvp != NULL) {
*depp = VTODE(nvp);
KASSERT((*depp)->de_dirclust == dirclust, ("wrong dirclust"));
KASSERT((*depp)->de_diroffset == diroffset, ("wrong diroffset"));
return (0);
}
/*
* Do the malloc before the getnewvnode since doing so afterward
* might cause a bogus v_data pointer to get dereferenced
* elsewhere if malloc should block.
*/
ldep = malloc(sizeof(struct denode), M_XTAFNODE, M_WAITOK | M_ZERO);
/*
* Directory entry was not in cache, have to create a vnode and
* copy it from the passed disk buffer.
*/
/* getnewvnode() does a VREF() on the vnode */
error = getnewvnode("xtaf", mntp, &xtaf_vnodeops, &nvp);
if (error) {
*depp = NULL;
free(ldep, M_XTAFNODE);
return error;
}
nvp->v_data = ldep;
ldep->de_vnode = nvp;
ldep->de_flag = 0;
ldep->de_dirclust = dirclust;
ldep->de_diroffset = diroffset;
ldep->de_inode = inode;
lockmgr(nvp->v_vnlock, LK_EXCLUSIVE, NULL);
xtaf_fc_purge(ldep, 0); /* init the fat cache for this denode */
error = insmntque(nvp, mntp);
if (error != 0) {
free(ldep, M_XTAFNODE);
*depp = NULL;
return (error);
}
error = vfs_hash_insert(nvp, inode, LK_EXCLUSIVE, curthread, &xvp,
de_vncmpf, &inode);
if (error) {
*depp = NULL;
return (error);
}
if (xvp != NULL) {
*depp = xvp->v_data;
return (0);
}
ldep->de_pmp = pmp;
ldep->de_refcnt = 1;
#ifdef XTAF_DEBUG
printf("xtaf_deget(): dirclust=%lx diroffset=%lx\n", dirclust, diroffset);
#endif
/*
* Copy the directory entry into the denode area of the vnode.
*/
if (dirclust == XTAFROOT && diroffset == XTAFROOT_OFS) {
/*
* Directory entry for the root directory. There isn't one,
* so we manufacture one.
*/
nvp->v_vflag |= VV_ROOT;
FAKEDIR(ldep, pmp->pm_rootdirsize, XTAFROOT);
#ifdef XTAF_DEBUG
printf("xtaf_deget(): FAKEDIR root\n");
#endif
} else {
error = xtaf_readep(pmp, dirclust, diroffset, &bp, &ep);
if (error) {
/*
* The denode does not contain anything useful, so
* it would be wrong to leave it on its hash chain.
* Arrange for vput() to just forget about it.
*/
ldep->de_Length = LEN_DELETED;
vput(nvp);
*depp = NULL;
return (error);
}
DE_INTERNALIZE(ldep, ep);
brelse(bp);
}
/*
* Fill in a few fields of the vnode and finish filling in the
* denode. Then return the address of the found denode.
*/
if (ldep->de_Attributes & ATTR_DIRECTORY) {
/*
* Since XTAF directory entries that describe directories
* have 0 in the filesize field, we take this opportunity
* to find out the length of the directory and plug it into
* the denode structure.
*/
u_long size;
#ifdef XTAF_DEBUG
/* FIXME something goes wrong here, StartCluster shouldn't
* be 0xffffffff ? Filename also gets blanked here and
* SF_IMMUTABLE set
*/
printf("ldep->de_StartCluster = %lx dirclust = %lx\n", ldep->de_StartCluster, dirclust);
#endif
nvp->v_type = VDIR;
if (ldep->de_StartCluster != XTAFROOT) {
error = xtaf_pcbmap(ldep, 0xffff, 0, &size, 0);
if (error == E2BIG) {
ldep->de_FileSize = de_cn2off(pmp, size);
error = 0;
} else
printf("xtaf_deget(): xtaf_pcbmap returned "
"%d\n", error);
}
} else
nvp->v_type = VREG;
ldep->de_modrev = init_va_filerev();
*depp = ldep;
return (0);
}
/*
* Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
* in from disk. If it is in core, wait for the lock bit to clear, then
* return the inode locked. Detection and handling of mount points must be
* done by the calling routine.
*/
int
ext2fs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
{
struct m_ext2fs *fs;
struct inode *ip;
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error;
void *cp;
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
return (0);
/* Allocate a new vnode/inode. */
error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, NULL, &vp);
if (error) {
*vpp = NULL;
return (error);
}
ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
mutex_enter(&ufs_hashlock);
if ((*vpp = ufs_ihashget(dev, ino, 0)) != NULL) {
mutex_exit(&ufs_hashlock);
ungetnewvnode(vp);
pool_put(&ext2fs_inode_pool, ip);
goto retry;
}
vp->v_vflag |= VV_LOCKSWORK;
memset(ip, 0, sizeof(struct inode));
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_ump = ump;
ip->i_e2fs = fs = ump->um_e2fs;
ip->i_dev = dev;
ip->i_number = ino;
ip->i_e2fs_last_lblk = 0;
ip->i_e2fs_last_blk = 0;
genfs_node_init(vp, &ext2fs_genfsops);
/*
* Put it onto its hash chain and lock it so that other requests for
* this inode will block if they arrive while we are sleeping waiting
* for old data structures to be purged or for the contents of the
* disk portion of this inode to be read.
*/
ufs_ihashins(ip);
mutex_exit(&ufs_hashlock);
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, NOCRED, 0, &bp);
if (error) {
/*
* The inode does not contain anything useful, so it would
* be misleading to leave it on its hash chain. With mode
* still zero, it will be unlinked and returned to the free
* list by vput().
*/
vput(vp);
*vpp = NULL;
return (error);
}
cp = (char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs));
ip->i_din.e2fs_din = pool_get(&ext2fs_dinode_pool, PR_WAITOK);
e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
ext2fs_set_inode_guid(ip);
brelse(bp, 0);
/* If the inode was deleted, reset all fields */
if (ip->i_e2fs_dtime != 0) {
ip->i_e2fs_mode = 0;
(void)ext2fs_setsize(ip, 0);
(void)ext2fs_setnblock(ip, 0);
memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
}
/*
* Initialize the vnode from the inode, check for aliases.
*/
error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
if (error) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization now that aliasing has been resolved.
*/
ip->i_devvp = ump->um_devvp;
vref(ip->i_devvp);
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (ip->i_e2fs_gen == 0) {
if (++ext2gennumber < (u_long)time_second)
ext2gennumber = time_second;
ip->i_e2fs_gen = ext2gennumber;
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
uvm_vnp_setsize(vp, ext2fs_size(ip));
*vpp = vp;
return (0);
}
static int
fuse_vnode_alloc(struct mount *mp,
struct thread *td,
uint64_t nodeid,
enum vtype vtyp,
struct vnode **vpp)
{
const int lkflags = LK_EXCLUSIVE | LK_RETRY;
struct fuse_vnode_data *fvdat;
struct vnode *vp2;
int err = 0;
DEBUG("been asked for vno #%ju\n", (uintmax_t)nodeid);
if (vtyp == VNON) {
return EINVAL;
}
*vpp = NULL;
err = vfs_hash_get(mp, fuse_vnode_hash(nodeid), lkflags, td, vpp,
fuse_vnode_cmp, &nodeid);
if (err)
return (err);
if (*vpp) {
MPASS((*vpp)->v_type == vtyp && (*vpp)->v_data != NULL);
DEBUG("vnode taken from hash\n");
return (0);
}
fvdat = malloc(sizeof(*fvdat), M_FUSEVN, M_WAITOK | M_ZERO);
err = getnewvnode("fuse", mp, &fuse_vnops, vpp);
if (err) {
free(fvdat, M_FUSEVN);
return (err);
}
vn_lock(*vpp, lkflags);
err = insmntque(*vpp, mp);
ASSERT_VOP_ELOCKED(*vpp, "fuse_vnode_alloc");
if (err) {
VOP_UNLOCK(*vpp, 0);
free(fvdat, M_FUSEVN);
*vpp = NULL;
return (err);
}
fuse_vnode_init(*vpp, fvdat, nodeid, vtyp);
err = vfs_hash_insert(*vpp, fuse_vnode_hash(nodeid), lkflags,
td, &vp2, fuse_vnode_cmp, &nodeid);
if (err) {
VOP_UNLOCK(*vpp, 0);
fuse_vnode_destroy(*vpp);
*vpp = NULL;
return (err);
}
/*
* XXXIP: Prevent silent vnode reuse. It may happen because several fuse
* filesystems ignore inode numbers
*/
KASSERT(vp2 == NULL,
("vfs hash collision for node #%ju\n", (uintmax_t)nodeid));
ASSERT_VOP_ELOCKED(*vpp, "fuse_vnode_alloc");
return (0);
}
/*
* Make a new or get existing pefs node.
* vp is the alias vnode
* lvp is the lower vnode
* ldvp is the lower directory vnode, used if no key specified
*
* The lvp assumed to be locked and having "spare" reference. This routine
* vrele lvp if pefs node was taken from hash. Otherwise it "transfers" the
* caller's "spare" reference to created pefs vnode.
*/
static int
pefs_node_get(struct mount *mp, struct vnode *lvp, struct vnode **vpp,
pefs_node_init_fn *init_fn, void *context)
{
struct pefs_node *pn;
struct vnode *vp;
int error;
ASSERT_VOP_LOCKED(lvp, "pefs_node_get");
/* Lookup the hash firstly */
*vpp = pefs_nodehash_get(mp, lvp);
if (*vpp != NULL) {
vrele(lvp);
return (0);
}
/*
* We do not serialize vnode creation, instead we will check for
* duplicates later, when adding new vnode to hash.
*
* Note that duplicate can only appear in hash if the lvp is
* locked LK_SHARED.
*/
/*
* Do the MALLOC before the getnewvnode since doing so afterward
* might cause a bogus v_data pointer to get dereferenced
* elsewhere if MALLOC should block.
*/
pn = uma_zalloc(pefs_node_zone, M_WAITOK | M_ZERO);
pn->pn_lowervp = lvp;
/* pn->pn_lowervp should be initialized before calling init_fn. */
error = init_fn(mp, pn, context);
MPASS(!(((pn->pn_flags & PN_HASKEY) == 0) ^
(pn->pn_tkey.ptk_key == NULL)));
if (error != 0) {
uma_zfree(pefs_node_zone, pn);
return (error);
}
error = getnewvnode("pefs", mp, &pefs_vnodeops, &vp);
if (error != 0) {
pefs_key_release(pn->pn_tkey.ptk_key);
uma_zfree(pefs_node_zone, pn);
return (error);
}
if (pn->pn_tkey.ptk_key == NULL)
PEFSDEBUG("pefs_node_get: creating node without key: %p\n", pn);
pn->pn_vnode = vp;
vp->v_type = lvp->v_type;
vp->v_data = pn;
vp->v_vnlock = lvp->v_vnlock;
if (vp->v_vnlock == NULL)
panic("pefs_node_get: Passed a NULL vnlock.\n");
error = insmntque1(vp, mp, pefs_insmntque_dtr, pn);
if (error != 0)
return (error);
/*
* Atomically insert our new node into the hash or vget existing
* if someone else has beaten us to it.
*/
*vpp = pefs_nodehash_insert(mp, pn);
if (*vpp != NULL) {
vrele(lvp);
vp->v_vnlock = &vp->v_lock;
pn->pn_lowervp = NULL;
vrele(vp);
MPASS(PEFS_LOWERVP(*vpp) == lvp);
ASSERT_VOP_LOCKED(*vpp, "pefs_node_get: duplicate");
return (0);
}
if (vp->v_type == VDIR)
pn->pn_dircache = pefs_dircache_get();
*vpp = vp;
MPASS(PEFS_LOWERVP(*vpp) == lvp);
ASSERT_VOP_LOCKED(*vpp, "pefs_node_get");
return (0);
}
/*
* Allocate a vnode
*/
int
pfs_vncache_alloc(struct mount *mp, struct vnode **vpp,
struct pfs_node *pn, pid_t pid)
{
struct pfs_vdata *pvd, *pvd2;
struct vnode *vp;
int error;
/*
* See if the vnode is in the cache.
* XXX linear search is not very efficient.
*/
retry:
mtx_lock(&pfs_vncache_mutex);
for (pvd = pfs_vncache; pvd; pvd = pvd->pvd_next) {
if (pvd->pvd_pn == pn && pvd->pvd_pid == pid &&
pvd->pvd_vnode->v_mount == mp) {
vp = pvd->pvd_vnode;
VI_LOCK(vp);
mtx_unlock(&pfs_vncache_mutex);
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, curthread) == 0) {
++pfs_vncache_hits;
*vpp = vp;
/*
* Some callers cache_enter(vp) later, so
* we have to make sure it's not in the
* VFS cache so it doesn't get entered
* twice. A better solution would be to
* make pfs_vncache_alloc() responsible
* for entering the vnode in the VFS
* cache.
*/
cache_purge(vp);
return (0);
}
goto retry;
}
}
mtx_unlock(&pfs_vncache_mutex);
/* nope, get a new one */
pvd = malloc(sizeof *pvd, M_PFSVNCACHE, M_WAITOK);
pvd->pvd_next = pvd->pvd_prev = NULL;
error = getnewvnode("pseudofs", mp, &pfs_vnodeops, vpp);
if (error) {
free(pvd, M_PFSVNCACHE);
return (error);
}
pvd->pvd_pn = pn;
pvd->pvd_pid = pid;
(*vpp)->v_data = pvd;
switch (pn->pn_type) {
case pfstype_root:
(*vpp)->v_vflag = VV_ROOT;
#if 0
printf("root vnode allocated\n");
#endif
/* fall through */
case pfstype_dir:
case pfstype_this:
case pfstype_parent:
case pfstype_procdir:
(*vpp)->v_type = VDIR;
break;
case pfstype_file:
(*vpp)->v_type = VREG;
break;
case pfstype_symlink:
(*vpp)->v_type = VLNK;
break;
case pfstype_none:
KASSERT(0, ("pfs_vncache_alloc called for null node\n"));
default:
panic("%s has unexpected type: %d", pn->pn_name, pn->pn_type);
}
/*
* Propagate flag through to vnode so users know it can change
* if the process changes (i.e. execve)
*/
if ((pn->pn_flags & PFS_PROCDEP) != 0)
(*vpp)->v_vflag |= VV_PROCDEP;
pvd->pvd_vnode = *vpp;
vn_lock(*vpp, LK_EXCLUSIVE | LK_RETRY);
VN_LOCK_AREC(*vpp);
error = insmntque(*vpp, mp);
if (error != 0) {
free(pvd, M_PFSVNCACHE);
*vpp = NULLVP;
return (error);
}
retry2:
mtx_lock(&pfs_vncache_mutex);
/*
* Other thread may race with us, creating the entry we are
* going to insert into the cache. Recheck after
* pfs_vncache_mutex is reacquired.
*/
for (pvd2 = pfs_vncache; pvd2; pvd2 = pvd2->pvd_next) {
if (pvd2->pvd_pn == pn && pvd2->pvd_pid == pid &&
pvd2->pvd_vnode->v_mount == mp) {
vp = pvd2->pvd_vnode;
VI_LOCK(vp);
mtx_unlock(&pfs_vncache_mutex);
if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, curthread) == 0) {
++pfs_vncache_hits;
vgone(*vpp);
vput(*vpp);
*vpp = vp;
cache_purge(vp);
return (0);
}
goto retry2;
}
}
++pfs_vncache_misses;
if (++pfs_vncache_entries > pfs_vncache_maxentries)
pfs_vncache_maxentries = pfs_vncache_entries;
pvd->pvd_prev = NULL;
pvd->pvd_next = pfs_vncache;
if (pvd->pvd_next)
pvd->pvd_next->pvd_prev = pvd;
pfs_vncache = pvd;
mtx_unlock(&pfs_vncache_mutex);
return (0);
}
/*
* Get a p9node. Nodes are represented by (fid, qid) tuples in 9P2000.
* Fids are assigned by the client, while qids are assigned by the server.
*
* The caller is expected to have generated the FID via p9fs_getfid() and
* obtained the QID from the server via p9fs_client_walk() and friends.
*/
int
p9fs_nget(struct p9fs_session *p9s, uint32_t fid, struct p9fs_qid *qid,
int lkflags, struct p9fs_node **npp)
{
int error = 0;
struct p9fs_node *np;
struct vnode *vp, *nvp;
struct vattr vattr = {};
struct thread *td = curthread;
*npp = NULL;
error = vfs_hash_get(p9s->p9s_mount, fid, lkflags, td, &vp, NULL, NULL);
if (error != 0)
return (error);
if (vp != NULL) {
*npp = vp->v_data;
return (0);
}
np = malloc(sizeof (struct p9fs_node), M_P9NODE, M_WAITOK | M_ZERO);
getnewvnode_reserve(1);
error = getnewvnode("p9fs", p9s->p9s_mount, &p9fs_vnops, &nvp);
if (error != 0) {
getnewvnode_drop_reserve();
free(np, M_P9NODE);
return (error);
}
vp = nvp;
vn_lock(vp, LK_EXCLUSIVE);
error = insmntque(nvp, p9s->p9s_mount);
if (error != 0) {
/* vp was vput()'d by insmntque() */
free(np, M_P9NODE);
return (error);
}
error = vfs_hash_insert(nvp, fid, lkflags, td, &nvp, NULL, NULL);
if (error != 0) {
free(np, M_P9NODE);
return (error);
}
if (nvp != NULL) {
free(np, M_P9NODE);
*npp = nvp->v_data;
/* vp was vput()'d by vfs_hash_insert() */
return (0);
}
error = p9fs_client_stat(p9s, fid, &vattr);
if (error != 0) {
free(np, M_P9NODE);
return (error);
}
/* Our vnode is the winner. Set up the new p9node for it. */
vp->v_type = vattr.va_type;
vp->v_data = np;
np->p9n_fid = fid;
np->p9n_session = p9s;
np->p9n_vnode = vp;
bcopy(qid, &np->p9n_qid, sizeof (*qid));
*npp = np;
return (error);
}
/*
* Look up a vnode/nfsnode by file handle.
* Callers must check for mount points!!
* In all cases, a pointer to a
* nfsnode structure is returned.
*/
int
nfs_nget1(struct mount *mntp, nfsfh_t *fhp, int fhsize, struct nfsnode **npp,
int lkflags)
{
struct nfsnode *np;
struct vnode *vp;
struct nfsmount *nmp = VFSTONFS(mntp);
int error;
struct fh_match fhm;
fhm.fhm_fhp = fhp;
fhm.fhm_fhsize = fhsize;
loop:
rw_enter(&nmp->nm_rbtlock, RW_READER);
np = rb_tree_find_node(&nmp->nm_rbtree, &fhm);
if (np != NULL) {
vp = NFSTOV(np);
mutex_enter(vp->v_interlock);
rw_exit(&nmp->nm_rbtlock);
error = vget(vp, LK_EXCLUSIVE | lkflags);
if (error == EBUSY)
return error;
if (error)
goto loop;
*npp = np;
return(0);
}
rw_exit(&nmp->nm_rbtlock);
error = getnewvnode(VT_NFS, mntp, nfsv2_vnodeop_p, NULL, &vp);
if (error) {
*npp = 0;
return (error);
}
np = pool_get(&nfs_node_pool, PR_WAITOK);
memset(np, 0, sizeof *np);
np->n_vnode = vp;
/*
* Insert the nfsnode in the hash queue for its new file handle
*/
if (fhsize > NFS_SMALLFH) {
np->n_fhp = kmem_alloc(fhsize, KM_SLEEP);
} else
np->n_fhp = &np->n_fh;
memcpy(np->n_fhp, fhp, fhsize);
np->n_fhsize = fhsize;
np->n_accstamp = -1;
np->n_vattr = pool_get(&nfs_vattr_pool, PR_WAITOK);
rw_enter(&nmp->nm_rbtlock, RW_WRITER);
if (NULL != rb_tree_find_node(&nmp->nm_rbtree, &fhm)) {
rw_exit(&nmp->nm_rbtlock);
if (fhsize > NFS_SMALLFH) {
kmem_free(np->n_fhp, fhsize);
}
pool_put(&nfs_vattr_pool, np->n_vattr);
pool_put(&nfs_node_pool, np);
ungetnewvnode(vp);
goto loop;
}
vp->v_data = np;
genfs_node_init(vp, &nfs_genfsops);
/*
* Initalize read/write creds to useful values. VOP_OPEN will
* overwrite these.
*/
np->n_rcred = curlwp->l_cred;
kauth_cred_hold(np->n_rcred);
np->n_wcred = curlwp->l_cred;
kauth_cred_hold(np->n_wcred);
VOP_LOCK(vp, LK_EXCLUSIVE);
NFS_INVALIDATE_ATTRCACHE(np);
uvm_vnp_setsize(vp, 0);
(void)rb_tree_insert_node(&nmp->nm_rbtree, np);
rw_exit(&nmp->nm_rbtlock);
*npp = np;
return (0);
}
int
ntfs_vgetex(
struct mount *mp,
ino_t ino,
u_int32_t attrtype,
char *attrname,
u_long lkflags,
u_long flags,
struct vnode **vpp)
{
int error;
struct ntfsmount *ntmp;
struct ntnode *ip;
struct fnode *fp;
struct vnode *vp;
enum vtype f_type = VBAD;
dprintf(("ntfs_vgetex: ino: %llu, attr: 0x%x:%s, lkf: 0x%lx, f:"
" 0x%lx\n", (unsigned long long)ino, attrtype,
attrname ? attrname : "", (u_long)lkflags, (u_long)flags));
ntmp = VFSTONTFS(mp);
*vpp = NULL;
loop:
/* Get ntnode */
error = ntfs_ntlookup(ntmp, ino, &ip);
if (error) {
printf("ntfs_vget: ntfs_ntget failed\n");
return (error);
}
/* It may be not initialized fully, so force load it */
if (!(flags & VG_DONTLOADIN) && !(ip->i_flag & IN_LOADED)) {
error = ntfs_loadntnode(ntmp, ip);
if(error) {
printf("ntfs_vget: CAN'T LOAD ATTRIBUTES FOR INO:"
" %llu\n", (unsigned long long)ip->i_number);
ntfs_ntput(ip);
return (error);
}
}
error = ntfs_fget(ntmp, ip, attrtype, attrname, &fp);
if (error) {
printf("ntfs_vget: ntfs_fget failed\n");
ntfs_ntput(ip);
return (error);
}
if (!(flags & VG_DONTVALIDFN) && !(fp->f_flag & FN_VALID)) {
if ((ip->i_frflag & NTFS_FRFLAG_DIR) &&
(fp->f_attrtype == NTFS_A_DATA && fp->f_attrname == NULL)) {
f_type = VDIR;
} else if (flags & VG_EXT) {
f_type = VNON;
fp->f_size = fp->f_allocated = 0;
} else {
f_type = VREG;
error = ntfs_filesize(ntmp, fp,
&fp->f_size, &fp->f_allocated);
if (error) {
ntfs_ntput(ip);
return (error);
}
}
fp->f_flag |= FN_VALID;
}
/*
* We may be calling vget() now. To avoid potential deadlock, we need
* to release ntnode lock, since due to locking order vnode
* lock has to be acquired first.
* ntfs_fget() bumped ntnode usecount, so ntnode won't be recycled
* prematurely.
* Take v_interlock before releasing ntnode lock to avoid races.
*/
vp = FTOV(fp);
if (vp) {
mutex_enter(vp->v_interlock);
ntfs_ntput(ip);
if (vget(vp, lkflags) != 0)
goto loop;
*vpp = vp;
return 0;
}
ntfs_ntput(ip);
error = getnewvnode(VT_NTFS, ntmp->ntm_mountp, ntfs_vnodeop_p,
NULL, &vp);
if(error) {
ntfs_frele(fp);
return (error);
}
ntfs_ntget(ip);
error = ntfs_fget(ntmp, ip, attrtype, attrname, &fp);
if (error) {
printf("ntfs_vget: ntfs_fget failed\n");
ntfs_ntput(ip);
return (error);
}
if (FTOV(fp)) {
/*
* Another thread beat us, put back freshly allocated
* vnode and retry.
*/
ntfs_ntput(ip);
ungetnewvnode(vp);
goto loop;
}
dprintf(("ntfs_vget: vnode: %p for ntnode: %llu\n", vp,
(unsigned long long)ino));
fp->f_vp = vp;
vp->v_data = fp;
if (f_type != VBAD)
vp->v_type = f_type;
genfs_node_init(vp, &ntfs_genfsops);
if (ino == NTFS_ROOTINO)
vp->v_vflag |= VV_ROOT;
ntfs_ntput(ip);
if (lkflags & (LK_EXCLUSIVE | LK_SHARED)) {
error = vn_lock(vp, lkflags);
if (error) {
vput(vp);
return (error);
}
}
uvm_vnp_setsize(vp, fp->f_size); /* XXX: mess, cf. ntfs_lookupfile() */
vref(ip->i_devvp);
*vpp = vp;
return (0);
}
/*
* Get the vnode associated with the given inode, allocating the vnode if
* necessary. The vnode will be returned exclusively locked.
*
* *errorp is set to a UNIX error, not a HAMMER2 error.
*
* The caller must lock the inode (shared or exclusive).
*
* Great care must be taken to avoid deadlocks and vnode acquisition/reclaim
* races.
*/
struct vnode *
hammer2_igetv(hammer2_inode_t *ip, int *errorp)
{
hammer2_pfs_t *pmp;
struct vnode *vp;
pmp = ip->pmp;
KKASSERT(pmp != NULL);
*errorp = 0;
for (;;) {
/*
* Attempt to reuse an existing vnode assignment. It is
* possible to race a reclaim so the vget() may fail. The
* inode must be unlocked during the vget() to avoid a
* deadlock against a reclaim.
*/
int wasexclusive;
vp = ip->vp;
if (vp) {
/*
* Inode must be unlocked during the vget() to avoid
* possible deadlocks, but leave the ip ref intact.
*
* vnode is held to prevent destruction during the
* vget(). The vget() can still fail if we lost
* a reclaim race on the vnode.
*/
hammer2_mtx_state_t ostate;
vhold(vp);
ostate = hammer2_inode_lock_temp_release(ip);
if (vget(vp, LK_EXCLUSIVE)) {
vdrop(vp);
hammer2_inode_lock_temp_restore(ip, ostate);
continue;
}
hammer2_inode_lock_temp_restore(ip, ostate);
vdrop(vp);
/* vp still locked and ref from vget */
if (ip->vp != vp) {
kprintf("hammer2: igetv race %p/%p\n",
ip->vp, vp);
vput(vp);
continue;
}
*errorp = 0;
break;
}
/*
* No vnode exists, allocate a new vnode. Beware of
* allocation races. This function will return an
* exclusively locked and referenced vnode.
*/
*errorp = getnewvnode(VT_HAMMER2, pmp->mp, &vp, 0, 0);
if (*errorp) {
kprintf("hammer2: igetv getnewvnode failed %d\n",
*errorp);
vp = NULL;
break;
}
/*
* Lock the inode and check for an allocation race.
*/
wasexclusive = hammer2_inode_lock_upgrade(ip);
if (ip->vp != NULL) {
vp->v_type = VBAD;
vx_put(vp);
hammer2_inode_lock_downgrade(ip, wasexclusive);
continue;
}
switch (ip->meta.type) {
case HAMMER2_OBJTYPE_DIRECTORY:
vp->v_type = VDIR;
break;
case HAMMER2_OBJTYPE_REGFILE:
/*
* Regular file must use buffer cache I/O
* (VKVABIO cpu sync semantics supported)
*/
vp->v_type = VREG;
vsetflags(vp, VKVABIO);
vinitvmio(vp, ip->meta.size,
HAMMER2_LBUFSIZE,
(int)ip->meta.size & HAMMER2_LBUFMASK);
break;
case HAMMER2_OBJTYPE_SOFTLINK:
/*
* XXX for now we are using the generic file_read
* and file_write code so we need a buffer cache
* association.
*
* (VKVABIO cpu sync semantics supported)
*/
vp->v_type = VLNK;
vsetflags(vp, VKVABIO);
vinitvmio(vp, ip->meta.size,
HAMMER2_LBUFSIZE,
(int)ip->meta.size & HAMMER2_LBUFMASK);
break;
case HAMMER2_OBJTYPE_CDEV:
vp->v_type = VCHR;
/* fall through */
case HAMMER2_OBJTYPE_BDEV:
vp->v_ops = &pmp->mp->mnt_vn_spec_ops;
if (ip->meta.type != HAMMER2_OBJTYPE_CDEV)
vp->v_type = VBLK;
addaliasu(vp,
ip->meta.rmajor,
ip->meta.rminor);
break;
case HAMMER2_OBJTYPE_FIFO:
vp->v_type = VFIFO;
vp->v_ops = &pmp->mp->mnt_vn_fifo_ops;
break;
case HAMMER2_OBJTYPE_SOCKET:
vp->v_type = VSOCK;
break;
default:
panic("hammer2: unhandled objtype %d",
ip->meta.type);
break;
}
if (ip == pmp->iroot)
vsetflags(vp, VROOT);
vp->v_data = ip;
ip->vp = vp;
hammer2_inode_ref(ip); /* vp association */
hammer2_inode_lock_downgrade(ip, wasexclusive);
break;
}
/*
* Return non-NULL vp and *errorp == 0, or NULL vp and *errorp != 0.
*/
if (hammer2_debug & 0x0002) {
kprintf("igetv vp %p refs 0x%08x aux 0x%08x\n",
vp, vp->v_refcnt, vp->v_auxrefs);
}
return (vp);
}
/*
* Mount the per-process file descriptors (/dev/fd)
*/
static int
portal_mount(struct mount *mp)
{
struct file *fp;
struct portalmount *fmp;
struct socket *so;
struct vnode *rvp;
struct thread *td;
struct portalnode *pn;
int error, v;
char *p;
td = curthread;
if (vfs_filteropt(mp->mnt_optnew, portal_opts))
return (EINVAL);
error = vfs_scanopt(mp->mnt_optnew, "socket", "%d", &v);
if (error != 1)
return (EINVAL);
error = vfs_getopt(mp->mnt_optnew, "config", (void **)&p, NULL);
if (error)
return (error);
/*
* Capsicum is not incompatible with portalfs, but we don't really
* know what rights are required. In the spirit of "better safe than
* sorry", pretend that all rights are required for now.
*/
if ((error = fget(td, v, CAP_MASK_VALID, &fp)) != 0)
return (error);
if (fp->f_type != DTYPE_SOCKET) {
fdrop(fp, td);
return(ENOTSOCK);
}
so = fp->f_data; /* XXX race against userland */
if (so->so_proto->pr_domain->dom_family != AF_UNIX) {
fdrop(fp, td);
return (ESOCKTNOSUPPORT);
}
pn = malloc(sizeof(struct portalnode),
M_TEMP, M_WAITOK);
fmp = malloc(sizeof(struct portalmount),
M_PORTALFSMNT, M_WAITOK); /* XXX */
error = getnewvnode("portal", mp, &portal_vnodeops, &rvp); /* XXX */
if (error) {
free(fmp, M_PORTALFSMNT);
free(pn, M_TEMP);
fdrop(fp, td);
return (error);
}
error = insmntque(rvp, mp); /* XXX: Too early for mpsafe fs */
if (error != 0) {
free(fmp, M_PORTALFSMNT);
free(pn, M_TEMP);
fdrop(fp, td);
return (error);
}
rvp->v_data = pn;
rvp->v_type = VDIR;
rvp->v_vflag |= VV_ROOT;
VTOPORTAL(rvp)->pt_arg = 0;
VTOPORTAL(rvp)->pt_size = 0;
VTOPORTAL(rvp)->pt_fileid = PORTAL_ROOTFILEID;
fmp->pm_root = rvp;
fhold(fp);
fmp->pm_server = fp;
MNT_ILOCK(mp);
mp->mnt_flag |= MNT_LOCAL;
MNT_IUNLOCK(mp);
mp->mnt_data = fmp;
vfs_getnewfsid(mp);
vfs_mountedfrom(mp, p);
fdrop(fp, td);
return (0);
}
/*
* Make a new or get existing unionfs node.
*
* uppervp and lowervp should be unlocked. Because if new unionfs vnode is
* locked, uppervp or lowervp is locked too. In order to prevent dead lock,
* you should not lock plurality simultaneously.
*/
int
unionfs_nodeget(struct mount *mp, struct vnode *uppervp,
struct vnode *lowervp, struct vnode *dvp,
struct vnode **vpp, struct componentname *cnp,
struct thread *td)
{
struct unionfs_mount *ump;
struct unionfs_node *unp;
struct vnode *vp;
int error;
int lkflags;
enum vtype vt;
char *path;
ump = MOUNTTOUNIONFSMOUNT(mp);
lkflags = (cnp ? cnp->cn_lkflags : 0);
path = (cnp ? cnp->cn_nameptr : NULL);
*vpp = NULLVP;
if (uppervp == NULLVP && lowervp == NULLVP)
panic("unionfs_nodeget: upper and lower is null");
vt = (uppervp != NULLVP ? uppervp->v_type : lowervp->v_type);
/* If it has no ISLASTCN flag, path check is skipped. */
if (cnp && !(cnp->cn_flags & ISLASTCN))
path = NULL;
/* check the cache */
if (path != NULL && dvp != NULLVP && vt == VDIR) {
vp = unionfs_get_cached_vnode(uppervp, lowervp, dvp, path);
if (vp != NULLVP) {
vref(vp);
*vpp = vp;
goto unionfs_nodeget_out;
}
}
if ((uppervp == NULLVP || ump->um_uppervp != uppervp) ||
(lowervp == NULLVP || ump->um_lowervp != lowervp)) {
/* dvp will be NULLVP only in case of root vnode. */
if (dvp == NULLVP)
return (EINVAL);
}
/*
* Do the MALLOC before the getnewvnode since doing so afterward
* might cause a bogus v_data pointer to get dereferenced elsewhere
* if MALLOC should block.
*/
unp = malloc(sizeof(struct unionfs_node),
M_UNIONFSNODE, M_WAITOK | M_ZERO);
error = getnewvnode("unionfs", mp, &unionfs_vnodeops, &vp);
if (error != 0) {
free(unp, M_UNIONFSNODE);
return (error);
}
error = insmntque(vp, mp); /* XXX: Too early for mpsafe fs */
if (error != 0) {
free(unp, M_UNIONFSNODE);
return (error);
}
if (dvp != NULLVP)
vref(dvp);
if (uppervp != NULLVP)
vref(uppervp);
if (lowervp != NULLVP)
vref(lowervp);
if (vt == VDIR)
unp->un_hashtbl = hashinit(NUNIONFSNODECACHE, M_UNIONFSHASH,
&(unp->un_hashmask));
unp->un_vnode = vp;
unp->un_uppervp = uppervp;
unp->un_lowervp = lowervp;
unp->un_dvp = dvp;
if (uppervp != NULLVP)
vp->v_vnlock = uppervp->v_vnlock;
else
vp->v_vnlock = lowervp->v_vnlock;
if (path != NULL) {
unp->un_path = (char *)
malloc(cnp->cn_namelen +1, M_UNIONFSPATH, M_WAITOK|M_ZERO);
bcopy(cnp->cn_nameptr, unp->un_path, cnp->cn_namelen);
unp->un_path[cnp->cn_namelen] = '\0';
}
vp->v_type = vt;
vp->v_data = unp;
if ((uppervp != NULLVP && ump->um_uppervp == uppervp) &&
(lowervp != NULLVP && ump->um_lowervp == lowervp))
vp->v_vflag |= VV_ROOT;
if (path != NULL && dvp != NULLVP && vt == VDIR)
*vpp = unionfs_ins_cached_vnode(unp, dvp, path);
if ((*vpp) != NULLVP) {
if (dvp != NULLVP)
vrele(dvp);
if (uppervp != NULLVP)
vrele(uppervp);
if (lowervp != NULLVP)
vrele(lowervp);
unp->un_uppervp = NULLVP;
unp->un_lowervp = NULLVP;
unp->un_dvp = NULLVP;
vrele(vp);
vp = *vpp;
vref(vp);
} else
*vpp = vp;
unionfs_nodeget_out:
if (lkflags & LK_TYPE_MASK)
vn_lock(vp, lkflags | LK_RETRY, td);
return (0);
}
/*
* layer_node_alloc: make a new layerfs vnode.
*
* => vp is the alias vnode, lowervp is the lower vnode.
* => We will hold a reference to lowervp.
*/
int
layer_node_alloc(struct mount *mp, struct vnode *lowervp, struct vnode **vpp)
{
struct layer_mount *lmp = MOUNTTOLAYERMOUNT(mp);
struct layer_node_hashhead *hd;
struct layer_node *xp;
struct vnode *vp, *nvp;
int error;
/* Get a new vnode and share its interlock with underlying vnode. */
error = getnewvnode(lmp->layerm_tag, mp, lmp->layerm_vnodeop_p,
lowervp->v_interlock, &vp);
if (error) {
return error;
}
vp->v_type = lowervp->v_type;
mutex_enter(vp->v_interlock);
vp->v_iflag |= VI_LAYER;
mutex_exit(vp->v_interlock);
xp = kmem_alloc(lmp->layerm_size, KM_SLEEP);
if (xp == NULL) {
ungetnewvnode(vp);
return ENOMEM;
}
if (vp->v_type == VBLK || vp->v_type == VCHR) {
spec_node_init(vp, lowervp->v_rdev);
}
/*
* Before inserting the node into the hash, check if other thread
* did not race with us. If so - return that node, destroy ours.
*/
mutex_enter(&lmp->layerm_hashlock);
if ((nvp = layer_node_find(mp, lowervp)) != NULL) {
ungetnewvnode(vp);
kmem_free(xp, lmp->layerm_size);
*vpp = nvp;
return 0;
}
vp->v_data = xp;
vp->v_vflag = (vp->v_vflag & ~VV_MPSAFE) |
(lowervp->v_vflag & VV_MPSAFE);
xp->layer_vnode = vp;
xp->layer_lowervp = lowervp;
xp->layer_flags = 0;
/*
* Insert the new node into the hash.
* Add a reference to the lower node.
*/
vref(lowervp);
hd = LAYER_NHASH(lmp, lowervp);
LIST_INSERT_HEAD(hd, xp, layer_hash);
uvm_vnp_setsize(vp, 0);
mutex_exit(&lmp->layerm_hashlock);
*vpp = vp;
return 0;
}
static int
smbfs_node_alloc(struct mount *mp, struct vnode *dvp, const char *dirnm,
int dirlen, const char *name, int nmlen, char sep,
struct smbfattr *fap, struct vnode **vpp)
{
struct vattr vattr;
struct thread *td = curthread; /* XXX */
struct smbmount *smp = VFSTOSMBFS(mp);
struct smbnode *np, *dnp;
struct vnode *vp, *vp2;
struct smbcmp sc;
char *p, *rpath;
int error, rplen;
sc.n_parent = dvp;
sc.n_nmlen = nmlen;
sc.n_name = name;
if (smp->sm_root != NULL && dvp == NULL) {
SMBERROR("do not allocate root vnode twice!\n");
return EINVAL;
}
if (nmlen == 2 && bcmp(name, "..", 2) == 0) {
if (dvp == NULL)
return EINVAL;
vp = VTOSMB(VTOSMB(dvp)->n_parent)->n_vnode;
error = vget(vp, LK_EXCLUSIVE, td);
if (error == 0)
*vpp = vp;
return error;
} else if (nmlen == 1 && name[0] == '.') {
SMBERROR("do not call me with dot!\n");
return EINVAL;
}
dnp = dvp ? VTOSMB(dvp) : NULL;
if (dnp == NULL && dvp != NULL) {
vprint("smbfs_node_alloc: dead parent vnode", dvp);
return EINVAL;
}
error = vfs_hash_get(mp, smbfs_hash(name, nmlen), LK_EXCLUSIVE, td,
vpp, smbfs_vnode_cmp, &sc);
if (error)
return (error);
if (*vpp) {
np = VTOSMB(*vpp);
/* Force cached attributes to be refreshed if stale. */
(void)VOP_GETATTR(*vpp, &vattr, td->td_ucred);
/*
* If the file type on the server is inconsistent with
* what it was when we created the vnode, kill the
* bogus vnode now and fall through to the code below
* to create a new one with the right type.
*/
if (((*vpp)->v_type == VDIR &&
(np->n_dosattr & SMB_FA_DIR) == 0) ||
((*vpp)->v_type == VREG &&
(np->n_dosattr & SMB_FA_DIR) != 0)) {
vgone(*vpp);
vput(*vpp);
}
else {
SMBVDEBUG("vnode taken from the hashtable\n");
return (0);
}
}
/*
* If we don't have node attributes, then it is an explicit lookup
* for an existing vnode.
*/
if (fap == NULL)
return ENOENT;
error = getnewvnode("smbfs", mp, &smbfs_vnodeops, vpp);
if (error)
return (error);
vp = *vpp;
np = malloc(sizeof *np, M_SMBNODE, M_WAITOK | M_ZERO);
rplen = dirlen;
if (sep != '\0')
rplen++;
rplen += nmlen;
rpath = malloc(rplen + 1, M_SMBNODENAME, M_WAITOK);
p = rpath;
bcopy(dirnm, p, dirlen);
p += dirlen;
if (sep != '\0')
*p++ = sep;
if (name != NULL) {
bcopy(name, p, nmlen);
p += nmlen;
}
*p = '\0';
MPASS(p == rpath + rplen);
lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
/* Vnode initialization */
vp->v_type = fap->fa_attr & SMB_FA_DIR ? VDIR : VREG;
vp->v_data = np;
np->n_vnode = vp;
np->n_mount = VFSTOSMBFS(mp);
np->n_rpath = rpath;
np->n_rplen = rplen;
np->n_nmlen = nmlen;
np->n_name = smbfs_name_alloc(name, nmlen);
np->n_ino = fap->fa_ino;
if (dvp) {
ASSERT_VOP_LOCKED(dvp, "smbfs_node_alloc");
np->n_parent = dvp;
np->n_parentino = VTOSMB(dvp)->n_ino;
if (/*vp->v_type == VDIR &&*/ (dvp->v_vflag & VV_ROOT) == 0) {
vref(dvp);
np->n_flag |= NREFPARENT;
}
} else if (vp->v_type == VREG)
SMBERROR("new vnode '%s' born without parent ?\n", np->n_name);
error = insmntque(vp, mp);
if (error) {
free(np, M_SMBNODE);
return (error);
}
error = vfs_hash_insert(vp, smbfs_hash(name, nmlen), LK_EXCLUSIVE,
td, &vp2, smbfs_vnode_cmp, &sc);
if (error)
return (error);
if (vp2 != NULL)
*vpp = vp2;
return (0);
}
/*
* If deget() succeeds it returns with the gotten denode locked().
*
* pmp - address of msdosfsmount structure of the filesystem containing
* the denode of interest. The pm_dev field and the address of
* the msdosfsmount structure are used.
* dirclust - which cluster bp contains, if dirclust is 0 (root directory)
* diroffset is relative to the beginning of the root directory,
* otherwise it is cluster relative.
* diroffset - offset past begin of cluster of denode we want
* depp - returns the address of the gotten denode.
*/
int
deget(struct msdosfsmount *pmp, uint32_t dirclust, uint32_t diroffset,
struct denode **depp)
{
int error;
extern struct vops msdosfs_vops;
struct direntry *direntptr;
struct denode *ldep;
struct vnode *nvp;
struct buf *bp;
struct proc *p = curproc; /* XXX */
#ifdef MSDOSFS_DEBUG
printf("deget(pmp %08x, dirclust %d, diroffset %x, depp %08x)\n",
pmp, dirclust, diroffset, depp);
#endif
/*
* On FAT32 filesystems, root is a (more or less) normal
* directory
*/
if (FAT32(pmp) && dirclust == MSDOSFSROOT)
dirclust = pmp->pm_rootdirblk;
/*
* See if the denode is in the denode cache. Use the location of
* the directory entry to compute the hash value. For subdir use
* address of "." entry. For root dir (if not FAT32) use cluster
* MSDOSFSROOT, offset MSDOSFSROOT_OFS
*
* NOTE: The check for de_refcnt > 0 below insures the denode being
* examined does not represent an unlinked but still open file.
* These files are not to be accessible even when the directory
* entry that represented the file happens to be reused while the
* deleted file is still open.
*/
retry:
ldep = msdosfs_hashget(pmp->pm_dev, dirclust, diroffset);
if (ldep) {
*depp = ldep;
return (0);
}
/*
* Directory entry was not in cache, have to create a vnode and
* copy it from the passed disk buffer.
*/
/* getnewvnode() does a vref() on the vnode */
error = getnewvnode(VT_MSDOSFS, pmp->pm_mountp, &msdosfs_vops, &nvp);
if (error) {
*depp = 0;
return (error);
}
ldep = malloc(sizeof(*ldep), M_MSDOSFSNODE, M_WAITOK | M_ZERO);
lockinit(&ldep->de_lock, PINOD, "denode", 0, 0);
nvp->v_data = ldep;
ldep->de_vnode = nvp;
ldep->de_flag = 0;
ldep->de_devvp = 0;
ldep->de_lockf = 0;
ldep->de_dev = pmp->pm_dev;
ldep->de_dirclust = dirclust;
ldep->de_diroffset = diroffset;
fc_purge(ldep, 0); /* init the fat cache for this denode */
/*
* Insert the denode into the hash queue and lock the denode so it
* can't be accessed until we've read it in and have done what we
* need to it.
*/
vn_lock(nvp, LK_EXCLUSIVE | LK_RETRY, p);
error = msdosfs_hashins(ldep);
if (error) {
vput (nvp);
if (error == EEXIST)
goto retry;
return (error);
}
ldep->de_pmp = pmp;
ldep->de_devvp = pmp->pm_devvp;
ldep->de_refcnt = 1;
/*
* Copy the directory entry into the denode area of the vnode.
*/
if ((dirclust == MSDOSFSROOT
|| (FAT32(pmp) && dirclust == pmp->pm_rootdirblk))
&& diroffset == MSDOSFSROOT_OFS) {
/*
* Directory entry for the root directory. There isn't one,
* so we manufacture one. We should probably rummage
* through the root directory and find a label entry (if it
* exists), and then use the time and date from that entry
* as the time and date for the root denode.
*/
nvp->v_flag |= VROOT; /* should be further down XXX */
ldep->de_Attributes = ATTR_DIRECTORY;
if (FAT32(pmp))
ldep->de_StartCluster = pmp->pm_rootdirblk;
/* de_FileSize will be filled in further down */
else {
ldep->de_StartCluster = MSDOSFSROOT;
ldep->de_FileSize = pmp->pm_rootdirsize * pmp->pm_BytesPerSec;
}
/*
* fill in time and date so that dos2unixtime() doesn't
* spit up when called from msdosfs_getattr() with root
* denode
*/
ldep->de_CTime = 0x0000; /* 00:00:00 */
ldep->de_CTimeHundredth = 0;
ldep->de_CDate = (0 << DD_YEAR_SHIFT) | (1 << DD_MONTH_SHIFT)
| (1 << DD_DAY_SHIFT);
/* Jan 1, 1980 */
ldep->de_ADate = ldep->de_CDate;
ldep->de_MTime = ldep->de_CTime;
ldep->de_MDate = ldep->de_CDate;
/* leave the other fields as garbage */
} else {
error = readep(pmp, dirclust, diroffset, &bp, &direntptr);
if (error)
return (error);
DE_INTERNALIZE(ldep, direntptr);
brelse(bp);
}
/*
* Fill in a few fields of the vnode and finish filling in the
* denode. Then return the address of the found denode.
*/
if (ldep->de_Attributes & ATTR_DIRECTORY) {
/*
* Since DOS directory entries that describe directories
* have 0 in the filesize field, we take this opportunity
* to find out the length of the directory and plug it into
* the denode structure.
*/
uint32_t size;
nvp->v_type = VDIR;
if (ldep->de_StartCluster != MSDOSFSROOT) {
error = pcbmap(ldep, 0xffff, 0, &size, 0);
if (error == E2BIG) {
ldep->de_FileSize = de_cn2off(pmp, size);
error = 0;
} else if (error) {
printf("deget(): pcbmap returned %d\n", error);
return (error);
}
}
} else
nvp->v_type = VREG;
vref(ldep->de_devvp);
*depp = ldep;
return (0);
}
/*
* internal version with extra arguments to allow accessing resource fork
*/
int
hfs_vget_internal(struct mount *mp, ino_t ino, uint8_t fork,
struct vnode **vpp)
{
struct hfsmount *hmp;
struct hfsnode *hnode;
struct vnode *vp;
hfs_callback_args cbargs;
hfs_cnid_t cnid;
hfs_catalog_keyed_record_t rec;
hfs_catalog_key_t key; /* the search key used to find this file on disk */
dev_t dev;
int error;
#ifdef HFS_DEBUG
printf("vfsop = hfs_vget()\n");
#endif /* HFS_DEBUG */
hnode = NULL;
vp = NULL;
hmp = VFSTOHFS(mp);
dev = hmp->hm_dev;
cnid = (hfs_cnid_t)ino;
if (fork != HFS_RSRCFORK)
fork = HFS_DATAFORK;
retry:
/* Check if this vnode has already been allocated. If so, just return it. */
if ((*vpp = hfs_nhashget(dev, cnid, fork, LK_EXCLUSIVE)) != NULL)
return 0;
/* Allocate a new vnode/inode. */
error = getnewvnode(VT_HFS, mp, hfs_vnodeop_p, NULL, &vp);
if (error) {
goto error;
}
hnode = malloc(sizeof(struct hfsnode), M_TEMP,
M_WAITOK | M_ZERO);
/*
* If someone beat us to it while sleeping in getnewvnode(),
* push back the freshly allocated vnode we don't need, and return.
*/
mutex_enter(&hfs_hashlock);
if (hfs_nhashget(dev, cnid, fork, 0) != NULL) {
mutex_exit(&hfs_hashlock);
ungetnewvnode(vp);
free(hnode, M_TEMP);
goto retry;
}
vp->v_vflag |= VV_LOCKSWORK;
vp->v_data = hnode;
genfs_node_init(vp, &hfs_genfsops);
hnode->h_vnode = vp;
hnode->h_hmp = hmp;
hnode->dummy = 0x1337BABE;
/*
* We need to put this vnode into the hash chain and lock it so that other
* requests for this inode will block if they arrive while we are sleeping
* waiting for old data structures to be purged or for the contents of the
* disk portion of this inode to be read. The hash chain requires the node's
* device and cnid to be known. Since this information was passed in the
* arguments, fill in the appropriate hfsnode fields without reading having
* to read the disk.
*/
hnode->h_dev = dev;
hnode->h_rec.u.cnid = cnid;
hnode->h_fork = fork;
hfs_nhashinsert(hnode);
mutex_exit(&hfs_hashlock);
/*
* Read catalog record from disk.
*/
hfslib_init_cbargs(&cbargs);
if (hfslib_find_catalog_record_with_cnid(&hmp->hm_vol, cnid,
&rec, &key, &cbargs) != 0) {
vput(vp);
error = EBADF;
goto error;
}
memcpy(&hnode->h_rec, &rec, sizeof(hnode->h_rec));
hnode->h_parent = key.parent_cnid;
/* XXX Eventually need to add an "ignore permissions" mount option */
/*
* Now convert some of the catalog record's fields into values that make
* sense on this system.
*/
/* DATE AND TIME */
/*
* Initialize the vnode from the hfsnode, check for aliases.
* Note that the underlying vnode may change.
*/
hfs_vinit(mp, hfs_specop_p, hfs_fifoop_p, &vp);
hnode->h_devvp = hmp->hm_devvp;
vref(hnode->h_devvp); /* Increment the ref count to the volume's device. */
/* Make sure UVM has allocated enough memory. (?) */
if (hnode->h_rec.u.rec_type == HFS_REC_FILE) {
if (hnode->h_fork == HFS_DATAFORK)
uvm_vnp_setsize(vp,
hnode->h_rec.file.data_fork.logical_size);
else
uvm_vnp_setsize(vp,
hnode->h_rec.file.rsrc_fork.logical_size);
}
else
uvm_vnp_setsize(vp, 0); /* no directly reading directories */
*vpp = vp;
return 0;
error:
*vpp = NULL;
return error;
}
/*
* Look up a vnode/nfsnode by file handle and store the pointer in *npp.
* Callers must check for mount points!!
* An error number is returned.
*/
int
nfs_nget(struct mount *mnt, nfsfh_t *fh, int fhsize, struct nfsnode **npp)
{
struct nfsmount *nmp;
struct nfsnode *np, find, *np2;
struct vnode *vp, *nvp;
struct proc *p = curproc; /* XXX */
int error;
nmp = VFSTONFS(mnt);
loop:
rw_enter_write(&nfs_hashlock);
find.n_fhp = fh;
find.n_fhsize = fhsize;
np = RB_FIND(nfs_nodetree, &nmp->nm_ntree, &find);
if (np != NULL) {
rw_exit_write(&nfs_hashlock);
vp = NFSTOV(np);
error = vget(vp, LK_EXCLUSIVE, p);
if (error)
goto loop;
*npp = np;
return (0);
}
/*
* getnewvnode() could recycle a vnode, potentially formerly
* owned by NFS. This will cause a VOP_RECLAIM() to happen,
* which will cause recursive locking, so we unlock before
* calling getnewvnode() lock again afterwards, but must check
* to see if this nfsnode has been added while we did not hold
* the lock.
*/
rw_exit_write(&nfs_hashlock);
error = getnewvnode(VT_NFS, mnt, &nfs_vops, &nvp);
/* note that we don't have this vnode set up completely yet */
rw_enter_write(&nfs_hashlock);
if (error) {
*npp = NULL;
rw_exit_write(&nfs_hashlock);
return (error);
}
nvp->v_flag |= VLARVAL;
np = RB_FIND(nfs_nodetree, &nmp->nm_ntree, &find);
if (np != NULL) {
vgone(nvp);
rw_exit_write(&nfs_hashlock);
goto loop;
}
vp = nvp;
np = pool_get(&nfs_node_pool, PR_WAITOK | PR_ZERO);
vp->v_data = np;
/* we now have an nfsnode on this vnode */
vp->v_flag &= ~VLARVAL;
np->n_vnode = vp;
rw_init(&np->n_commitlock, "nfs_commitlk");
/*
* Are we getting the root? If so, make sure the vnode flags
* are correct
*/
if ((fhsize == nmp->nm_fhsize) && !bcmp(fh, nmp->nm_fh, fhsize)) {
if (vp->v_type == VNON)
vp->v_type = VDIR;
vp->v_flag |= VROOT;
}
np->n_fhp = &np->n_fh;
bcopy(fh, np->n_fhp, fhsize);
np->n_fhsize = fhsize;
np2 = RB_INSERT(nfs_nodetree, &nmp->nm_ntree, np);
KASSERT(np2 == NULL);
np->n_accstamp = -1;
rw_exit(&nfs_hashlock);
*npp = np;
return (0);
}
/*
* Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
* in from disk. If it is in core, wait for the lock bit to clear, then
* return the inode locked. Detection and handling of mount points must be
* done by the calling routine.
*/
static int
ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
{
struct m_ext2fs *fs;
struct inode *ip;
struct ext2mount *ump;
struct buf *bp;
struct vnode *vp;
struct cdev *dev;
struct thread *td;
int i, error;
int used_blocks;
td = curthread;
error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
if (error || *vpp != NULL)
return (error);
ump = VFSTOEXT2(mp);
dev = ump->um_dev;
ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
*vpp = NULL;
free(ip, M_EXT2NODE);
return (error);
}
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_e2fs = fs = ump->um_e2fs;
ip->i_ump = ump;
ip->i_number = ino;
lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
error = insmntque(vp, mp);
if (error != 0) {
free(ip, M_EXT2NODE);
*vpp = NULL;
return (error);
}
error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
if (error || *vpp != NULL)
return (error);
/* Read in the disk contents for the inode, copy into the inode. */
if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
/*
* The inode does not contain anything useful, so it would
* be misleading to leave it on its hash chain. With mode
* still zero, it will be unlinked and returned to the free
* list by vput().
*/
brelse(bp);
vput(vp);
*vpp = NULL;
return (error);
}
/* convert ext2 inode to dinode */
ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
ino_to_fsbo(fs, ino)), ip);
ip->i_block_group = ino_to_cg(fs, ino);
ip->i_next_alloc_block = 0;
ip->i_next_alloc_goal = 0;
/*
* Now we want to make sure that block pointers for unused
* blocks are zeroed out - ext2_balloc depends on this
* although for regular files and directories only
*
* If IN_E4EXTENTS is enabled, unused blocks are not zeroed
* out because we could corrupt the extent tree.
*/
if (!(ip->i_flag & IN_E4EXTENTS) &&
(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
ip->i_db[i] = 0;
}
#ifdef EXT2FS_DEBUG
ext2_print_inode(ip);
#endif
bqrelse(bp);
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Finish inode initialization.
*/
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (ip->i_gen == 0) {
ip->i_gen = random() + 1;
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
*vpp = vp;
return (0);
}
/*
* If deget() succeeds it returns with the gotten denode locked().
*
* pmp - address of msdosfsmount structure of the filesystem containing
* the denode of interest. The address of
* the msdosfsmount structure are used.
* dirclust - which cluster bp contains, if dirclust is 0 (root directory)
* diroffset is relative to the beginning of the root directory,
* otherwise it is cluster relative.
* diroffset - offset past begin of cluster of denode we want
* depp - returns the address of the gotten denode.
*/
int
deget(struct msdosfsmount *pmp, u_long dirclust, u_long diroffset,
struct denode **depp)
{
int error;
uint64_t inode;
struct mount *mntp = pmp->pm_mountp;
struct direntry *direntptr;
struct denode *ldep;
struct vnode *nvp, *xvp;
struct buf *bp;
#ifdef MSDOSFS_DEBUG
printf("deget(pmp %p, dirclust %lu, diroffset %lx, depp %p)\n",
pmp, dirclust, diroffset, depp);
#endif
/*
* On FAT32 filesystems, root is a (more or less) normal
* directory
*/
if (FAT32(pmp) && dirclust == MSDOSFSROOT)
dirclust = pmp->pm_rootdirblk;
/*
* See if the denode is in the denode cache. Use the location of
* the directory entry to compute the hash value. For subdir use
* address of "." entry. For root dir (if not FAT32) use cluster
* MSDOSFSROOT, offset MSDOSFSROOT_OFS
*
* NOTE: The check for de_refcnt > 0 below insures the denode being
* examined does not represent an unlinked but still open file.
* These files are not to be accessible even when the directory
* entry that represented the file happens to be reused while the
* deleted file is still open.
*/
inode = (uint64_t)pmp->pm_bpcluster * dirclust + diroffset;
error = vfs_hash_get(mntp, inode, LK_EXCLUSIVE, curthread, &nvp,
de_vncmpf, &inode);
if (error)
return (error);
if (nvp != NULL) {
*depp = VTODE(nvp);
KASSERT((*depp)->de_dirclust == dirclust, ("wrong dirclust"));
KASSERT((*depp)->de_diroffset == diroffset, ("wrong diroffset"));
return (0);
}
ldep = malloc(sizeof(struct denode), M_MSDOSFSNODE, M_WAITOK | M_ZERO);
/*
* Directory entry was not in cache, have to create a vnode and
* copy it from the passed disk buffer.
*/
/* getnewvnode() does a VREF() on the vnode */
error = getnewvnode("msdosfs", mntp, &msdosfs_vnodeops, &nvp);
if (error) {
*depp = NULL;
free(ldep, M_MSDOSFSNODE);
return error;
}
nvp->v_data = ldep;
ldep->de_vnode = nvp;
ldep->de_flag = 0;
ldep->de_dirclust = dirclust;
ldep->de_diroffset = diroffset;
ldep->de_inode = inode;
lockmgr(nvp->v_vnlock, LK_EXCLUSIVE, NULL);
fc_purge(ldep, 0); /* init the FAT cache for this denode */
error = insmntque(nvp, mntp);
if (error != 0) {
free(ldep, M_MSDOSFSNODE);
*depp = NULL;
return (error);
}
error = vfs_hash_insert(nvp, inode, LK_EXCLUSIVE, curthread, &xvp,
de_vncmpf, &inode);
if (error) {
*depp = NULL;
return (error);
}
if (xvp != NULL) {
*depp = xvp->v_data;
return (0);
}
ldep->de_pmp = pmp;
ldep->de_refcnt = 1;
/*
* Copy the directory entry into the denode area of the vnode.
*/
if ((dirclust == MSDOSFSROOT
|| (FAT32(pmp) && dirclust == pmp->pm_rootdirblk))
&& diroffset == MSDOSFSROOT_OFS) {
/*
* Directory entry for the root directory. There isn't one,
* so we manufacture one. We should probably rummage
* through the root directory and find a label entry (if it
* exists), and then use the time and date from that entry
* as the time and date for the root denode.
*/
nvp->v_vflag |= VV_ROOT; /* should be further down XXX */
ldep->de_Attributes = ATTR_DIRECTORY;
ldep->de_LowerCase = 0;
if (FAT32(pmp))
ldep->de_StartCluster = pmp->pm_rootdirblk;
/* de_FileSize will be filled in further down */
else {
ldep->de_StartCluster = MSDOSFSROOT;
ldep->de_FileSize = pmp->pm_rootdirsize * DEV_BSIZE;
}
/*
* fill in time and date so that fattime2timespec() doesn't
* spit up when called from msdosfs_getattr() with root
* denode
*/
ldep->de_CHun = 0;
ldep->de_CTime = 0x0000; /* 00:00:00 */
ldep->de_CDate = (0 << DD_YEAR_SHIFT) | (1 << DD_MONTH_SHIFT)
| (1 << DD_DAY_SHIFT);
/* Jan 1, 1980 */
ldep->de_ADate = ldep->de_CDate;
ldep->de_MTime = ldep->de_CTime;
ldep->de_MDate = ldep->de_CDate;
/* leave the other fields as garbage */
} else {
error = readep(pmp, dirclust, diroffset, &bp, &direntptr);
if (error) {
/*
* The denode does not contain anything useful, so
* it would be wrong to leave it on its hash chain.
* Arrange for vput() to just forget about it.
*/
ldep->de_Name[0] = SLOT_DELETED;
vput(nvp);
*depp = NULL;
return (error);
}
(void)DE_INTERNALIZE(ldep, direntptr);
brelse(bp);
}
/*
* Fill in a few fields of the vnode and finish filling in the
* denode. Then return the address of the found denode.
*/
if (ldep->de_Attributes & ATTR_DIRECTORY) {
/*
* Since DOS directory entries that describe directories
* have 0 in the filesize field, we take this opportunity
* to find out the length of the directory and plug it into
* the denode structure.
*/
u_long size;
/*
* XXX it sometimes happens that the "." entry has cluster
* number 0 when it shouldn't. Use the actual cluster number
* instead of what is written in directory entry.
*/
if (diroffset == 0 && ldep->de_StartCluster != dirclust) {
#ifdef MSDOSFS_DEBUG
printf("deget(): \".\" entry at clust %lu != %lu\n",
dirclust, ldep->de_StartCluster);
#endif
ldep->de_StartCluster = dirclust;
}
nvp->v_type = VDIR;
if (ldep->de_StartCluster != MSDOSFSROOT) {
error = pcbmap(ldep, 0xffff, 0, &size, 0);
if (error == E2BIG) {
ldep->de_FileSize = de_cn2off(pmp, size);
error = 0;
} else {
#ifdef MSDOSFS_DEBUG
printf("deget(): pcbmap returned %d\n", error);
#endif
}
}
} else
nvp->v_type = VREG;
ldep->de_modrev = init_va_filerev();
*depp = ldep;
return (0);
}
int
ntfs_vgetex(struct mount *mp, ino_t ino, u_int32_t attrtype, char *attrname,
u_long lkflags, u_long flags, struct thread *td,
struct vnode **vpp)
{
int error;
struct ntfsmount *ntmp;
struct ntnode *ip;
struct fnode *fp;
struct vnode *vp;
enum vtype f_type;
dprintf(("ntfs_vgetex: ino: %ju, attr: 0x%x:%s, lkf: 0x%lx, f: 0x%lx\n",
(uintmax_t) ino, attrtype, attrname?attrname:"", lkflags, flags));
ntmp = VFSTONTFS(mp);
*vpp = NULL;
/* Get ntnode */
error = ntfs_ntlookup(ntmp, ino, &ip);
if (error) {
kprintf("ntfs_vget: ntfs_ntget failed\n");
return (error);
}
/* It may be not initialized fully, so force load it */
if (!(flags & VG_DONTLOADIN) && !(ip->i_flag & IN_LOADED)) {
error = ntfs_loadntnode(ntmp, ip);
if(error) {
kprintf("ntfs_vget: CAN'T LOAD ATTRIBUTES FOR INO: %"PRId64"\n",
ip->i_number);
ntfs_ntput(ip);
return (error);
}
}
error = ntfs_fget(ntmp, ip, attrtype, attrname, &fp);
if (error) {
kprintf("ntfs_vget: ntfs_fget failed\n");
ntfs_ntput(ip);
return (error);
}
f_type = VINT;
if (!(flags & VG_DONTVALIDFN) && !(fp->f_flag & FN_VALID)) {
if ((ip->i_frflag & NTFS_FRFLAG_DIR) &&
(fp->f_attrtype == NTFS_A_DATA && fp->f_attrname == NULL)) {
f_type = VDIR;
} else if (flags & VG_EXT) {
f_type = VINT;
fp->f_size = fp->f_allocated = 0;
} else {
f_type = VREG;
error = ntfs_filesize(ntmp, fp,
&fp->f_size, &fp->f_allocated);
if (error) {
ntfs_ntput(ip);
return (error);
}
}
fp->f_flag |= FN_VALID;
}
if (FTOV(fp)) {
VGET(FTOV(fp), lkflags);
*vpp = FTOV(fp);
ntfs_ntput(ip);
return (0);
}
error = getnewvnode(VT_NTFS, ntmp->ntm_mountp, &vp, VLKTIMEOUT, 0);
if(error) {
ntfs_frele(fp);
ntfs_ntput(ip);
return (error);
}
dprintf(("ntfs_vget: vnode: %p for ntnode: %ju\n", vp, (uintmax_t)ino));
fp->f_vp = vp;
vp->v_data = fp;
vp->v_type = f_type;
if (ino == NTFS_ROOTINO)
vsetflags(vp, VROOT);
/*
* Normal files use the buffer cache
*/
if (f_type == VREG)
vinitvmio(vp, fp->f_size, PAGE_SIZE, -1);
ntfs_ntput(ip);
KKASSERT(lkflags & LK_TYPE_MASK);
/* XXX leave vnode locked exclusively from getnewvnode */
*vpp = vp;
return (0);
}
/*
* Allocate a vnode
*/
int
pfs_vncache_alloc(struct mount *mp, struct vnode **vpp,
struct pfs_node *pn, pid_t pid)
{
struct pfs_vdata *pvd;
int error;
/*
* See if the vnode is in the cache.
* XXX linear search is not very efficient.
*/
mtx_lock(&pfs_vncache_mutex);
for (pvd = pfs_vncache; pvd; pvd = pvd->pvd_next) {
if (pvd->pvd_pn == pn && pvd->pvd_pid == pid) {
if (vget(pvd->pvd_vnode, 0, curthread) == 0) {
++pfs_vncache_hits;
*vpp = pvd->pvd_vnode;
mtx_unlock(&pfs_vncache_mutex);
/* XXX see comment at top of pfs_lookup() */
cache_purge(*vpp);
vn_lock(*vpp, LK_RETRY | LK_EXCLUSIVE,
curthread);
return (0);
}
/* XXX if this can happen, we're in trouble */
break;
}
}
mtx_unlock(&pfs_vncache_mutex);
++pfs_vncache_misses;
/* nope, get a new one */
MALLOC(pvd, struct pfs_vdata *, sizeof *pvd, M_PFSVNCACHE, M_WAITOK);
if (++pfs_vncache_entries > pfs_vncache_maxentries)
pfs_vncache_maxentries = pfs_vncache_entries;
error = getnewvnode("pseudofs", mp, pfs_vnodeop_p, vpp);
if (error) {
FREE(pvd, M_PFSVNCACHE);
return (error);
}
pvd->pvd_pn = pn;
pvd->pvd_pid = pid;
(*vpp)->v_data = pvd;
switch (pn->pn_type) {
case pfstype_root:
(*vpp)->v_vflag = VV_ROOT;
#if 0
printf("root vnode allocated\n");
#endif
/* fall through */
case pfstype_dir:
case pfstype_this:
case pfstype_parent:
case pfstype_procdir:
(*vpp)->v_type = VDIR;
break;
case pfstype_file:
(*vpp)->v_type = VREG;
break;
case pfstype_symlink:
(*vpp)->v_type = VLNK;
break;
case pfstype_none:
KASSERT(0, ("pfs_vncache_alloc called for null node\n"));
default:
panic("%s has unexpected type: %d", pn->pn_name, pn->pn_type);
}
/*
* Propagate flag through to vnode so users know it can change
* if the process changes (i.e. execve)
*/
if ((pn->pn_flags & PFS_PROCDEP) != 0)
(*vpp)->v_vflag |= VV_PROCDEP;
pvd->pvd_vnode = *vpp;
mtx_lock(&pfs_vncache_mutex);
pvd->pvd_prev = NULL;
pvd->pvd_next = pfs_vncache;
if (pvd->pvd_next)
pvd->pvd_next->pvd_prev = pvd;
pfs_vncache = pvd;
mtx_unlock(&pfs_vncache_mutex);
(*vpp)->v_vnlock->lk_flags |= LK_CANRECURSE;
vn_lock(*vpp, LK_RETRY | LK_EXCLUSIVE, curthread);
return (0);
}
/*
* lookup an anode, check mount's hash table if not found, create
* return locked and referenced al la vget(vp, 1);
*/
int
adosfs_vget(struct mount *mp, ino_t an, struct vnode **vpp)
{
struct adosfsmount *amp;
struct vnode *vp;
struct anode *ap;
struct buf *bp;
char *nam, *tmp;
int namlen, error;
error = 0;
amp = VFSTOADOSFS(mp);
bp = NULL;
/*
* check hash table. we are done if found
*/
if ((*vpp = adosfs_ahashget(mp, an)) != NULL)
return (0);
error = getnewvnode(VT_ADOSFS, mp, adosfs_vnodeop_p, NULL, &vp);
if (error)
return (error);
/*
* setup, insert in hash, and lock before io.
*/
vp->v_data = ap = pool_get(&adosfs_node_pool, PR_WAITOK);
memset(ap, 0, sizeof(struct anode));
ap->vp = vp;
ap->amp = amp;
ap->block = an;
ap->nwords = amp->nwords;
genfs_node_init(vp, &adosfs_genfsops);
adosfs_ainshash(amp, ap);
if ((error = bread(amp->devvp, an * amp->bsize / DEV_BSIZE,
amp->bsize, NOCRED, 0, &bp)) != 0) {
vput(vp);
return (error);
}
/*
* get type and fill rest in based on that.
*/
switch (ap->type = adosfs_getblktype(amp, bp)) {
case AROOT:
vp->v_type = VDIR;
vp->v_vflag |= VV_ROOT;
ap->mtimev.days = adoswordn(bp, ap->nwords - 10);
ap->mtimev.mins = adoswordn(bp, ap->nwords - 9);
ap->mtimev.ticks = adoswordn(bp, ap->nwords - 8);
ap->created.days = adoswordn(bp, ap->nwords - 7);
ap->created.mins = adoswordn(bp, ap->nwords - 6);
ap->created.ticks = adoswordn(bp, ap->nwords - 5);
break;
case ALDIR:
case ADIR:
vp->v_type = VDIR;
break;
case ALFILE:
case AFILE:
vp->v_type = VREG;
ap->fsize = adoswordn(bp, ap->nwords - 47);
break;
case ASLINK: /* XXX soft link */
vp->v_type = VLNK;
/*
* convert from BCPL string and
* from: "part:dir/file" to: "/part/dir/file"
*/
nam = (char *)bp->b_data + (6 * sizeof(long));
namlen = strlen(nam);
tmp = nam;
while (*tmp && *tmp != ':')
tmp++;
if (*tmp == 0) {
ap->slinkto = malloc(namlen + 1, M_ANODE, M_WAITOK);
memcpy(ap->slinkto, nam, namlen);
} else if (*nam == ':') {
ap->slinkto = malloc(namlen + 1, M_ANODE, M_WAITOK);
memcpy(ap->slinkto, nam, namlen);
ap->slinkto[0] = '/';
} else {
ap->slinkto = malloc(namlen + 2, M_ANODE, M_WAITOK);
ap->slinkto[0] = '/';
memcpy(&ap->slinkto[1], nam, namlen);
ap->slinkto[tmp - nam + 1] = '/';
namlen++;
}
ap->slinkto[namlen] = 0;
ap->fsize = namlen;
break;
default:
brelse(bp, 0);
vput(vp);
return (EINVAL);
}
/*
* Get appropriate data from this block; hard link needs
* to get other data from the "real" block.
*/
/*
* copy in name (from original block)
*/
nam = (char *)bp->b_data + (ap->nwords - 20) * sizeof(u_int32_t);
namlen = *(u_char *)nam++;
if (namlen > 30) {
#ifdef DIAGNOSTIC
printf("adosfs: aget: name length too long blk %llu\n",
(unsigned long long)an);
#endif
brelse(bp, 0);
vput(vp);
return (EINVAL);
}
memcpy(ap->name, nam, namlen);
ap->name[namlen] = 0;
/*
* if dir alloc hash table and copy it in
*/
if (vp->v_type == VDIR) {
int i;
ap->tab = malloc(ANODETABSZ(ap) * 2, M_ANODE, M_WAITOK);
ap->ntabent = ANODETABENT(ap);
ap->tabi = (int *)&ap->tab[ap->ntabent];
memset(ap->tabi, 0, ANODETABSZ(ap));
for (i = 0; i < ap->ntabent; i++)
ap->tab[i] = adoswordn(bp, i + 6);
}
/*
* misc.
*/
ap->pblock = adoswordn(bp, ap->nwords - 3);
ap->hashf = adoswordn(bp, ap->nwords - 4);
ap->linknext = adoswordn(bp, ap->nwords - 10);
ap->linkto = adoswordn(bp, ap->nwords - 11);
/*
* setup last indirect block cache.
*/
ap->lastlindblk = 0;
if (ap->type == AFILE) {
ap->lastindblk = ap->block;
if (adoswordn(bp, ap->nwords - 10))
ap->linkto = ap->block;
} else if (ap->type == ALFILE) {
ap->lastindblk = ap->linkto;
brelse(bp, 0);
bp = NULL;
error = bread(amp->devvp, ap->linkto * amp->bsize / DEV_BSIZE,
amp->bsize, NOCRED, 0, &bp);
if (error) {
vput(vp);
return (error);
}
ap->fsize = adoswordn(bp, ap->nwords - 47);
/*
* Should ap->block be set to the real file header block?
*/
ap->block = ap->linkto;
}
if (ap->type == AROOT) {
ap->adprot = 15;
ap->uid = amp->uid;
ap->gid = amp->gid;
} else {
ap->adprot = adoswordn(bp, ap->nwords - 48) ^ 15;
/*
* ADOS directories do not have a `x' protection bit as
* it is known in VFS; this functionality is fulfilled
* by the ADOS `r' bit.
*
* To retain the ADOS behaviour, fake execute permissions
* in that case.
*/
if ((ap->type == ADIR || ap->type == ALDIR) &&
(ap->adprot & 0x00000008) == 0)
ap->adprot &= ~0x00000002;
/*
* Get uid/gid from extensions in file header
* (really need to know if this is a muFS partition)
*/
ap->uid = (adoswordn(bp, ap->nwords - 49) >> 16) & 0xffff;
ap->gid = adoswordn(bp, ap->nwords - 49) & 0xffff;
if (ap->uid || ap->gid) {
if (ap->uid == 0xffff)
ap->uid = 0;
if (ap->gid == 0xffff)
ap->gid = 0;
ap->adprot |= 0x40000000; /* Kludge */
}
else {
/*
* uid & gid extension don't exist,
* so use the mount-point uid/gid
*/
ap->uid = amp->uid;
ap->gid = amp->gid;
}
}
ap->mtime.days = adoswordn(bp, ap->nwords - 23);
ap->mtime.mins = adoswordn(bp, ap->nwords - 22);
ap->mtime.ticks = adoswordn(bp, ap->nwords - 21);
*vpp = vp;
brelse(bp, 0);
uvm_vnp_setsize(vp, ap->fsize);
return (0);
}
/*
* Look up a FFS dinode number to find its incore vnode, otherwise read it
* in from disk. If it is in core, wait for the lock bit to clear, then
* return the inode locked. Detection and handling of mount points must be
* done by the calling routine.
*/
int
ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
{
struct fs *fs;
struct inode *ip;
struct ufs1_dinode *dp1;
#ifdef FFS2
struct ufs2_dinode *dp2;
#endif
struct ufsmount *ump;
struct buf *bp;
struct vnode *vp;
dev_t dev;
int error;
if (ino > (ufsino_t)-1)
panic("ffs_vget: alien ino_t %llu", (unsigned long long)ino);
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if ((*vpp = ufs_ihashget(dev, ino)) != NULL)
return (0);
/* Allocate a new vnode/inode. */
if ((error = getnewvnode(VT_UFS, mp, &ffs_vops, &vp)) != 0) {
*vpp = NULL;
return (error);
}
#ifdef VFSLCKDEBUG
vp->v_flag |= VLOCKSWORK;
#endif
ip = pool_get(&ffs_ino_pool, PR_WAITOK|PR_ZERO);
lockinit(&ip->i_lock, PINOD, "inode", 0, 0);
ip->i_ump = ump;
vref(ip->i_devvp);
vp->v_data = ip;
ip->i_vnode = vp;
ip->i_fs = fs = ump->um_fs;
ip->i_dev = dev;
ip->i_number = ino;
ip->i_vtbl = &ffs_vtbl;
/*
* Put it onto its hash chain and lock it so that other requests for
* this inode will block if they arrive while we are sleeping waiting
* for old data structures to be purged or for the contents of the
* disk portion of this inode to be read.
*/
error = ufs_ihashins(ip);
if (error) {
/*
* VOP_INACTIVE will treat this as a stale file
* and recycle it quickly
*/
vrele(vp);
if (error == EEXIST)
goto retry;
return (error);
}
/* Read in the disk contents for the inode, copy into the inode. */
error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
(int)fs->fs_bsize, &bp);
if (error) {
/*
* The inode does not contain anything useful, so it would
* be misleading to leave it on its hash chain. With mode
* still zero, it will be unlinked and returned to the free
* list by vput().
*/
vput(vp);
brelse(bp);
*vpp = NULL;
return (error);
}
#ifdef FFS2
if (ip->i_ump->um_fstype == UM_UFS2) {
ip->i_din2 = pool_get(&ffs_dinode2_pool, PR_WAITOK);
dp2 = (struct ufs2_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
*ip->i_din2 = *dp2;
} else
#endif
{
ip->i_din1 = pool_get(&ffs_dinode1_pool, PR_WAITOK);
dp1 = (struct ufs1_dinode *) bp->b_data + ino_to_fsbo(fs, ino);
*ip->i_din1 = *dp1;
}
brelse(bp);
if (DOINGSOFTDEP(vp))
softdep_load_inodeblock(ip);
else
ip->i_effnlink = DIP(ip, nlink);
/*
* Initialize the vnode from the inode, check for aliases.
* Note that the underlying vnode may have changed.
*/
error = ufs_vinit(mp, &ffs_specvops, FFS_FIFOOPS, &vp);
if (error) {
vput(vp);
*vpp = NULL;
return (error);
}
/*
* Set up a generation number for this inode if it does not
* already have one. This should only happen on old filesystems.
*/
if (DIP(ip, gen) == 0) {
DIP_ASSIGN(ip, gen, arc4random() & INT_MAX);
if (DIP(ip, gen) == 0 || DIP(ip, gen) == -1)
DIP_ASSIGN(ip, gen, 1); /* Shouldn't happen */
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
/*
* Ensure that uid and gid are correct. This is a temporary
* fix until fsck has been changed to do the update.
*/
if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
ip->i_ffs1_uid = ip->i_din1->di_ouid;
ip->i_ffs1_gid = ip->i_din1->di_ogid;
}
*vpp = vp;
return (0);
}
/*
* ONLY USED FOR THE ROOT DIRECTORY. nfscl_nget() does the rest. If this
* function is going to be used to get Regular Files, code must be added
* to fill in the "struct nfsv4node".
* Look up a vnode/nfsnode by file handle.
* Callers must check for mount points!!
* In all cases, a pointer to a
* nfsnode structure is returned.
*/
int
ncl_nget(struct mount *mntp, u_int8_t *fhp, int fhsize, struct nfsnode **npp,
int lkflags)
{
struct thread *td = curthread; /* XXX */
struct nfsnode *np;
struct vnode *vp;
struct vnode *nvp;
int error;
u_int hash;
struct nfsmount *nmp;
struct nfsfh *nfhp;
nmp = VFSTONFS(mntp);
*npp = NULL;
hash = fnv_32_buf(fhp, fhsize, FNV1_32_INIT);
MALLOC(nfhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize,
M_NFSFH, M_WAITOK);
bcopy(fhp, &nfhp->nfh_fh[0], fhsize);
nfhp->nfh_len = fhsize;
error = vfs_hash_get(mntp, hash, lkflags,
td, &nvp, newnfs_vncmpf, nfhp);
FREE(nfhp, M_NFSFH);
if (error)
return (error);
if (nvp != NULL) {
*npp = VTONFS(nvp);
return (0);
}
np = uma_zalloc(newnfsnode_zone, M_WAITOK | M_ZERO);
error = getnewvnode("newnfs", mntp, &newnfs_vnodeops, &nvp);
if (error) {
uma_zfree(newnfsnode_zone, np);
return (error);
}
vp = nvp;
KASSERT(vp->v_bufobj.bo_bsize != 0, ("ncl_nget: bo_bsize == 0"));
vp->v_bufobj.bo_ops = &buf_ops_newnfs;
vp->v_data = np;
np->n_vnode = vp;
/*
* Initialize the mutex even if the vnode is going to be a loser.
* This simplifies the logic in reclaim, which can then unconditionally
* destroy the mutex (in the case of the loser, or if hash_insert
* happened to return an error no special casing is needed).
*/
mtx_init(&np->n_mtx, "NEWNFSnode lock", NULL, MTX_DEF | MTX_DUPOK);
/*
* NFS supports recursive and shared locking.
*/
lockmgr(vp->v_vnlock, LK_EXCLUSIVE | LK_NOWITNESS, NULL);
VN_LOCK_AREC(vp);
VN_LOCK_ASHARE(vp);
/*
* Are we getting the root? If so, make sure the vnode flags
* are correct
*/
if ((fhsize == nmp->nm_fhsize) &&
!bcmp(fhp, nmp->nm_fh, fhsize)) {
if (vp->v_type == VNON)
vp->v_type = VDIR;
vp->v_vflag |= VV_ROOT;
}
MALLOC(np->n_fhp, struct nfsfh *, sizeof (struct nfsfh) + fhsize,
M_NFSFH, M_WAITOK);
bcopy(fhp, np->n_fhp->nfh_fh, fhsize);
np->n_fhp->nfh_len = fhsize;
error = insmntque(vp, mntp);
if (error != 0) {
*npp = NULL;
FREE((caddr_t)np->n_fhp, M_NFSFH);
mtx_destroy(&np->n_mtx);
uma_zfree(newnfsnode_zone, np);
return (error);
}
error = vfs_hash_insert(vp, hash, lkflags,
td, &nvp, newnfs_vncmpf, np->n_fhp);
if (error)
return (error);
if (nvp != NULL) {
*npp = VTONFS(nvp);
/* vfs_hash_insert() vput()'s the losing vnode */
return (0);
}
*npp = np;
return (0);
}
static int
chfs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
{
struct chfs_mount *chmp;
struct chfs_inode *ip;
struct ufsmount *ump;
struct vnode *vp;
dev_t dev;
int error;
struct chfs_vnode_cache* chvc = NULL;
struct chfs_node_ref* nref = NULL;
struct buf *bp;
dbg("vget() | ino: %llu\n", (unsigned long long)ino);
ump = VFSTOUFS(mp);
dev = ump->um_dev;
retry:
if (!vpp) {
vpp = kmem_alloc(sizeof(struct vnode*), KM_SLEEP);
}
/* Get node from inode hash. */
if ((*vpp = chfs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL) {
return 0;
}
/* Allocate a new vnode/inode. */
if ((error = getnewvnode(VT_CHFS,
mp, chfs_vnodeop_p, NULL, &vp)) != 0) {
*vpp = NULL;
return (error);
}
ip = pool_get(&chfs_inode_pool, PR_WAITOK);
mutex_enter(&chfs_hashlock);
if ((*vpp = chfs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL) {
mutex_exit(&chfs_hashlock);
ungetnewvnode(vp);
pool_put(&chfs_inode_pool, ip);
goto retry;
}
vp->v_vflag |= VV_LOCKSWORK;
/* Initialize vnode/inode. */
memset(ip, 0, sizeof(*ip));
vp->v_data = ip;
ip->vp = vp;
ip->ch_type = VTTOCHT(vp->v_type);
ip->ump = ump;
ip->chmp = chmp = ump->um_chfs;
ip->dev = dev;
ip->ino = ino;
vp->v_mount = mp;
genfs_node_init(vp, &chfs_genfsops);
rb_tree_init(&ip->fragtree, &frag_rbtree_ops);
chfs_ihashins(ip);
mutex_exit(&chfs_hashlock);
/* Set root inode. */
if (ino == CHFS_ROOTINO) {
dbg("SETROOT\n");
vp->v_vflag |= VV_ROOT;
vp->v_type = VDIR;
ip->ch_type = CHT_DIR;
ip->mode = IFMT | IEXEC | IWRITE | IREAD;
ip->iflag |= (IN_ACCESS | IN_CHANGE | IN_UPDATE);
chfs_update(vp, NULL, NULL, UPDATE_WAIT);
TAILQ_INIT(&ip->dents);
chfs_set_vnode_size(vp, 512);
}
mutex_enter(&chmp->chm_lock_vnocache);
chvc = chfs_vnode_cache_get(chmp, ino);
mutex_exit(&chmp->chm_lock_vnocache);
if (!chvc) {
dbg("!chvc\n");
/* Initialize the corresponding vnode cache. */
/* XXX, we cant alloc under a lock, refactor this! */
chvc = chfs_vnode_cache_alloc(ino);
mutex_enter(&chmp->chm_lock_vnocache);
if (ino == CHFS_ROOTINO) {
chvc->nlink = 2;
chvc->pvno = CHFS_ROOTINO;
chvc->state = VNO_STATE_CHECKEDABSENT;
}
chfs_vnode_cache_add(chmp, chvc);
mutex_exit(&chmp->chm_lock_vnocache);
ip->chvc = chvc;
TAILQ_INIT(&ip->dents);
} else {
dbg("chvc\n");
ip->chvc = chvc;
/* We had a vnode cache, the node is already on flash, so read it */
if (ino == CHFS_ROOTINO) {
chvc->pvno = CHFS_ROOTINO;
TAILQ_INIT(&chvc->scan_dirents);
} else {
chfs_readvnode(mp, ino, &vp);
}
mutex_enter(&chmp->chm_lock_mountfields);
/* Initialize type specific things. */
switch (ip->ch_type) {
case CHT_DIR:
/* Read every dirent. */
nref = chvc->dirents;
while (nref &&
(struct chfs_vnode_cache *)nref != chvc) {
chfs_readdirent(mp, nref, ip);
nref = nref->nref_next;
}
chfs_set_vnode_size(vp, 512);
break;
case CHT_REG:
/* FALLTHROUGH */
case CHT_SOCK:
/* Collect data. */
dbg("read_inode_internal | ino: %llu\n",
(unsigned long long)ip->ino);
error = chfs_read_inode(chmp, ip);
if (error) {
vput(vp);
*vpp = NULL;
mutex_exit(&chmp->chm_lock_mountfields);
return (error);
}
break;
case CHT_LNK:
/* Collect data. */
dbg("read_inode_internal | ino: %llu\n",
(unsigned long long)ip->ino);
error = chfs_read_inode_internal(chmp, ip);
if (error) {
vput(vp);
*vpp = NULL;
mutex_exit(&chmp->chm_lock_mountfields);
return (error);
}
/* Set link. */
dbg("size: %llu\n", (unsigned long long)ip->size);
bp = getiobuf(vp, true);
bp->b_blkno = 0;
bp->b_bufsize = bp->b_resid =
bp->b_bcount = ip->size;
bp->b_data = kmem_alloc(ip->size, KM_SLEEP);
chfs_read_data(chmp, vp, bp);
if (!ip->target)
ip->target = kmem_alloc(ip->size,
KM_SLEEP);
memcpy(ip->target, bp->b_data, ip->size);
kmem_free(bp->b_data, ip->size);
putiobuf(bp);
break;
case CHT_CHR:
/* FALLTHROUGH */
case CHT_BLK:
/* FALLTHROUGH */
case CHT_FIFO:
/* Collect data. */
dbg("read_inode_internal | ino: %llu\n",
(unsigned long long)ip->ino);
error = chfs_read_inode_internal(chmp, ip);
if (error) {
vput(vp);
*vpp = NULL;
mutex_exit(&chmp->chm_lock_mountfields);
return (error);
}
/* Set device. */
bp = getiobuf(vp, true);
bp->b_blkno = 0;
bp->b_bufsize = bp->b_resid =
bp->b_bcount = sizeof(dev_t);
bp->b_data = kmem_alloc(sizeof(dev_t), KM_SLEEP);
chfs_read_data(chmp, vp, bp);
memcpy(&ip->rdev,
bp->b_data, sizeof(dev_t));
kmem_free(bp->b_data, sizeof(dev_t));
putiobuf(bp);
/* Set specific operations. */
if (ip->ch_type == CHT_FIFO) {
vp->v_op = chfs_fifoop_p;
} else {
vp->v_op = chfs_specop_p;
spec_node_init(vp, ip->rdev);
}
break;
case CHT_BLANK:
/* FALLTHROUGH */
case CHT_BAD:
break;
}
mutex_exit(&chmp->chm_lock_mountfields);
}
/* Finish inode initalization. */
ip->devvp = ump->um_devvp;
vref(ip->devvp);
uvm_vnp_setsize(vp, ip->size);
*vpp = vp;
return 0;
}
