int
vfs_mountroot_devfs(void)
{
struct vnode *vp;
struct nchandle nch;
struct nlookupdata nd;
struct mount *mp;
struct vfsconf *vfsp;
int error;
struct ucred *cred = proc0.p_ucred;
const char *devfs_path, *init_chroot;
char *dev_malloced = NULL;
if ((init_chroot = kgetenv("init_chroot")) != NULL) {
size_t l;
l = strlen(init_chroot) + sizeof("/dev");
dev_malloced = kmalloc(l, M_MOUNT, M_WAITOK);
ksnprintf(dev_malloced, l, "%s/dev", init_chroot);
devfs_path = dev_malloced;
} else {
devfs_path = "/dev";
}
/*
* Lookup the requested path and extract the nch and vnode.
*/
error = nlookup_init_raw(&nd,
devfs_path, UIO_SYSSPACE, NLC_FOLLOW,
cred, &rootnch);
if (error == 0) {
devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup_init is ok...\n");
if ((error = nlookup(&nd)) == 0) {
devfs_debug(DEVFS_DEBUG_DEBUG, "vfs_mountroot_devfs: nlookup is ok...\n");
if (nd.nl_nch.ncp->nc_vp == NULL) {
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup: simply not found\n");
error = ENOENT;
}
}
}
if (dev_malloced != NULL)
kfree(dev_malloced, M_MOUNT), dev_malloced = NULL;
devfs_path = NULL;
if (error) {
nlookup_done(&nd);
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: nlookup failed, error: %d\n", error);
return (error);
}
/*
* Extract the locked+refd ncp and cleanup the nd structure
*/
nch = nd.nl_nch;
cache_zero(&nd.nl_nch);
nlookup_done(&nd);
/*
* now we have the locked ref'd nch and unreferenced vnode.
*/
vp = nch.ncp->nc_vp;
if ((error = vget(vp, LK_EXCLUSIVE)) != 0) {
cache_put(&nch);
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vget failed\n");
return (error);
}
cache_unlock(&nch);
if ((error = vinvalbuf(vp, V_SAVE, 0, 0)) != 0) {
cache_drop(&nch);
vput(vp);
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vinvalbuf failed\n");
return (error);
}
if (vp->v_type != VDIR) {
cache_drop(&nch);
vput(vp);
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vp is not VDIR\n");
return (ENOTDIR);
}
vfsp = vfsconf_find_by_name("devfs");
vsetflags(vp, VMOUNT);
/*
* Allocate and initialize the filesystem.
*/
mp = kmalloc(sizeof(struct mount), M_MOUNT, M_ZERO|M_WAITOK);
mount_init(mp);
vfs_busy(mp, LK_NOWAIT);
mp->mnt_op = vfsp->vfc_vfsops;
mp->mnt_vfc = vfsp;
vfsp->vfc_refcount++;
mp->mnt_stat.f_type = vfsp->vfc_typenum;
mp->mnt_flag |= vfsp->vfc_flags & MNT_VISFLAGMASK;
strncpy(mp->mnt_stat.f_fstypename, vfsp->vfc_name, MFSNAMELEN);
mp->mnt_stat.f_owner = cred->cr_uid;
vn_unlock(vp);
/*
* Mount the filesystem.
*/
error = VFS_MOUNT(mp, "/dev", NULL, cred);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
/*
* Put the new filesystem on the mount list after root. The mount
* point gets its own mnt_ncmountpt (unless the VFS already set one
* up) which represents the root of the mount. The lookup code
* detects the mount point going forward and checks the root of
* the mount going backwards.
*
* It is not necessary to invalidate or purge the vnode underneath
* because elements under the mount will be given their own glue
* namecache record.
*/
if (!error) {
if (mp->mnt_ncmountpt.ncp == NULL) {
/*
* allocate, then unlock, but leave the ref intact
*/
cache_allocroot(&mp->mnt_ncmountpt, mp, NULL);
cache_unlock(&mp->mnt_ncmountpt);
}
mp->mnt_ncmounton = nch; /* inherits ref */
nch.ncp->nc_flag |= NCF_ISMOUNTPT;
/* XXX get the root of the fs and cache_setvp(mnt_ncmountpt...) */
vclrflags(vp, VMOUNT);
mountlist_insert(mp, MNTINS_LAST);
vn_unlock(vp);
//checkdirs(&mp->mnt_ncmounton, &mp->mnt_ncmountpt);
error = vfs_allocate_syncvnode(mp);
if (error) {
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: vfs_allocate_syncvnode failed\n");
}
vfs_unbusy(mp);
error = VFS_START(mp, 0);
vrele(vp);
} else {
vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_coherency_ops);
vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_journal_ops);
vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_norm_ops);
vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_spec_ops);
vfs_rm_vnodeops(mp, NULL, &mp->mnt_vn_fifo_ops);
vclrflags(vp, VMOUNT);
mp->mnt_vfc->vfc_refcount--;
vfs_unbusy(mp);
kfree(mp, M_MOUNT);
cache_drop(&nch);
vput(vp);
devfs_debug(DEVFS_DEBUG_SHOW, "vfs_mountroot_devfs: mount failed\n");
}
devfs_debug(DEVFS_DEBUG_DEBUG, "rootmount_devfs done with error: %d\n", error);
return (error);
}
/*
* Reload all incore data for a filesystem (used after running fsck on
* the root filesystem and finding things to fix). The filesystem must
* be mounted read-only.
*
* Things to do to update the mount:
* 1) invalidate all cached meta-data.
* 2) re-read superblock from disk.
* 3) re-read summary information from disk.
* 4) invalidate all inactive vnodes.
* 5) invalidate all cached file data.
* 6) re-read inode data for all active vnodes.
*/
int
ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
{
struct vnode *vp, *devvp;
struct inode *ip;
struct buf *bp;
struct m_ext2fs *fs;
struct ext2fs *newfs;
int i, error;
struct ufsmount *ump;
struct vnode_iterator *marker;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
return EINVAL;
ump = VFSTOUFS(mp);
/*
* Step 1: invalidate all cached meta-data.
*/
devvp = ump->um_devvp;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = vinvalbuf(devvp, 0, cred, l, 0, 0);
VOP_UNLOCK(devvp);
if (error)
panic("ext2fs_reload: dirty1");
fs = ump->um_e2fs;
/*
* Step 2: re-read superblock from disk. Copy in new superblock, and compute
* in-memory values.
*/
error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
if (error)
return error;
newfs = (struct ext2fs *)bp->b_data;
e2fs_sbload(newfs, &fs->e2fs);
brelse(bp, 0);
error = ext2fs_sbfill(fs, (mp->mnt_flag & MNT_RDONLY) != 0);
if (error)
return error;
/*
* Step 3: re-read summary information from disk.
*/
for (i = 0; i < fs->e2fs_ngdb; i++) {
error = bread(devvp ,
EXT2_FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
1 /* superblock */ + i),
fs->e2fs_bsize, 0, &bp);
if (error) {
return error;
}
e2fs_cgload((struct ext2_gd *)bp->b_data,
&fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
fs->e2fs_bsize);
brelse(bp, 0);
}
vfs_vnode_iterator_init(mp, &marker);
while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
/*
* Step 4: invalidate all inactive vnodes.
*/
if (vrecycle(vp))
continue;
/*
* Step 5: invalidate all cached file data.
*/
if (vn_lock(vp, LK_EXCLUSIVE)) {
vrele(vp);
continue;
}
if (vinvalbuf(vp, 0, cred, l, 0, 0))
panic("ext2fs_reload: dirty2");
/*
* Step 6: re-read inode data for all active vnodes.
*/
ip = VTOI(vp);
error = bread(devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
(int)fs->e2fs_bsize, 0, &bp);
if (error) {
vput(vp);
break;
}
error = ext2fs_loadvnode_content(fs, ip->i_number, bp, ip);
brelse(bp, 0);
if (error) {
vput(vp);
break;
}
vput(vp);
}
vfs_vnode_iterator_destroy(marker);
return error;
}
/*
* Allocate a new inode.
*/
int
ext2fs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp,
struct componentname *cnp)
{
struct inode *ip, *pdir;
struct vnode *tvp;
int error;
pdir = VTOI(dvp);
#ifdef DIAGNOSTIC
if ((cnp->cn_flags & HASBUF) == 0)
panic("ext2fs_makeinode: no name");
#endif
*vpp = NULL;
if ((mode & IFMT) == 0)
mode |= IFREG;
if ((error = ext2fs_inode_alloc(pdir, mode, cnp->cn_cred, &tvp))
!= 0) {
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
return (error);
}
ip = VTOI(tvp);
ip->i_e2fs_gid = pdir->i_e2fs_gid;
ip->i_e2fs_uid = cnp->cn_cred->cr_uid;
ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
ip->i_e2fs_mode = mode;
tvp->v_type = IFTOVT(mode); /* Rest init'd in getnewvnode(). */
ip->i_e2fs_nlink = 1;
if ((ip->i_e2fs_mode & ISGID) &&
!groupmember(ip->i_e2fs_gid, cnp->cn_cred) &&
suser_ucred(cnp->cn_cred))
ip->i_e2fs_mode &= ~ISGID;
/*
* Make sure inode goes to disk before directory entry.
*/
if ((error = ext2fs_update(ip, NULL, NULL, 1)) != 0)
goto bad;
error = ext2fs_direnter(ip, dvp, cnp);
if (error != 0)
goto bad;
if ((cnp->cn_flags & SAVESTART) == 0)
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
*vpp = tvp;
return (0);
bad:
/*
* Write error occurred trying to update the inode
* or the directory so must deallocate the inode.
*/
pool_put(&namei_pool, cnp->cn_pnbuf);
vput(dvp);
ip->i_e2fs_nlink = 0;
ip->i_flag |= IN_CHANGE;
tvp->v_type = VNON;
vput(tvp);
return (error);
}
static int
ufs_extattr_autostart_locked(struct mount *mp, struct thread *td)
{
struct vnode *rvp, *attr_dvp, *attr_system_dvp, *attr_user_dvp;
struct ufsmount *ump = VFSTOUFS(mp);
int error;
/*
* UFS_EXTATTR applies only to UFS1, as UFS2 uses native extended
* attributes, so don't autostart.
*/
if (ump->um_fstype != UFS1)
return (0);
/*
* Does UFS_EXTATTR_FSROOTSUBDIR exist off the filesystem root?
* If so, automatically start EA's.
*/
error = VFS_ROOT(mp, LK_EXCLUSIVE, &rvp);
if (error) {
printf("ufs_extattr_autostart.VFS_ROOT() returned %d\n",
error);
return (error);
}
error = ufs_extattr_lookup(rvp, UE_GETDIR_LOCKPARENT_DONT,
UFS_EXTATTR_FSROOTSUBDIR, &attr_dvp, td);
if (error) {
/* rvp ref'd but now unlocked */
vrele(rvp);
return (error);
}
if (rvp == attr_dvp) {
/* Should never happen. */
vput(rvp);
vrele(attr_dvp);
return (EINVAL);
}
vrele(rvp);
if (attr_dvp->v_type != VDIR) {
printf("ufs_extattr_autostart: %s != VDIR\n",
UFS_EXTATTR_FSROOTSUBDIR);
goto return_vput_attr_dvp;
}
error = ufs_extattr_start_locked(ump, td);
if (error) {
printf("ufs_extattr_autostart: ufs_extattr_start failed (%d)\n",
error);
goto return_vput_attr_dvp;
}
/*
* Look for two subdirectories: UFS_EXTATTR_SUBDIR_SYSTEM,
* UFS_EXTATTR_SUBDIR_USER. For each, iterate over the sub-directory,
* and start with appropriate type. Failures in either don't
* result in an over-all failure. attr_dvp is left locked to
* be cleaned up on exit.
*/
error = ufs_extattr_lookup(attr_dvp, UE_GETDIR_LOCKPARENT,
UFS_EXTATTR_SUBDIR_SYSTEM, &attr_system_dvp, td);
if (!error) {
error = ufs_extattr_iterate_directory(VFSTOUFS(mp),
attr_system_dvp, EXTATTR_NAMESPACE_SYSTEM, td);
if (error)
printf("ufs_extattr_iterate_directory returned %d\n",
error);
vput(attr_system_dvp);
}
error = ufs_extattr_lookup(attr_dvp, UE_GETDIR_LOCKPARENT,
UFS_EXTATTR_SUBDIR_USER, &attr_user_dvp, td);
if (!error) {
error = ufs_extattr_iterate_directory(VFSTOUFS(mp),
attr_user_dvp, EXTATTR_NAMESPACE_USER, td);
if (error)
printf("ufs_extattr_iterate_directory returned %d\n",
error);
vput(attr_user_dvp);
}
/* Mask startup failures in sub-directories. */
error = 0;
return_vput_attr_dvp:
vput(attr_dvp);
return (error);
}
/*
* XXX Ugly bit of code. But, this is the only safe time that the
* match between BIOS disks and native disks can be done.
*/
static void
matchbiosdisks(void)
{
struct btinfo_biosgeom *big;
struct bi_biosgeom_entry *be;
device_t dv;
deviter_t di;
int i, ck, error, m, n;
struct vnode *tv;
char mbr[DEV_BSIZE];
int dklist_size;
int numbig;
big = lookup_bootinfo(BTINFO_BIOSGEOM);
numbig = big ? big->num : 0;
/* First, count all native disks. */
for (dv = deviter_first(&di, DEVITER_F_ROOT_FIRST); dv != NULL;
dv = deviter_next(&di)) {
if (is_valid_disk(dv))
x86_ndisks++;
}
deviter_release(&di);
dklist_size = sizeof(struct disklist) + (x86_ndisks - 1) *
sizeof(struct nativedisk_info);
/* XXX M_TEMP is wrong */
x86_alldisks = malloc(dklist_size, M_TEMP, M_NOWAIT | M_ZERO);
if (x86_alldisks == NULL)
return;
x86_alldisks->dl_nnativedisks = x86_ndisks;
x86_alldisks->dl_nbiosdisks = numbig;
for (i = 0; i < numbig; i++) {
x86_alldisks->dl_biosdisks[i].bi_dev = big->disk[i].dev;
x86_alldisks->dl_biosdisks[i].bi_sec = big->disk[i].sec;
x86_alldisks->dl_biosdisks[i].bi_head = big->disk[i].head;
x86_alldisks->dl_biosdisks[i].bi_cyl = big->disk[i].cyl;
x86_alldisks->dl_biosdisks[i].bi_lbasecs = big->disk[i].totsec;
x86_alldisks->dl_biosdisks[i].bi_flags = big->disk[i].flags;
#ifdef GEOM_DEBUG
#ifdef notyet
printf("disk %x: flags %x, interface %x, device %llx\n",
big->disk[i].dev, big->disk[i].flags,
big->disk[i].interface_path, big->disk[i].device_path);
#endif
#endif
}
/* XXX Code duplication from findroot(). */
n = -1;
for (dv = deviter_first(&di, DEVITER_F_ROOT_FIRST); dv != NULL;
dv = deviter_next(&di)) {
if (device_class(dv) != DV_DISK)
continue;
#ifdef GEOM_DEBUG
printf("matchbiosdisks: trying to match (%s) %s\n",
device_xname(dv), device_cfdata(dv)->cf_name);
#endif
if (is_valid_disk(dv)) {
n++;
snprintf(x86_alldisks->dl_nativedisks[n].ni_devname,
sizeof(x86_alldisks->dl_nativedisks[n].ni_devname),
"%s", device_xname(dv));
if ((tv = opendisk(dv)) == NULL)
continue;
error = vn_rdwr(UIO_READ, tv, mbr, DEV_BSIZE, 0,
UIO_SYSSPACE, 0, NOCRED, NULL, NULL);
VOP_CLOSE(tv, FREAD, NOCRED);
vput(tv);
if (error) {
#ifdef GEOM_DEBUG
printf("matchbiosdisks: %s: MBR read failure\n",
device_xname(dv));
#endif
continue;
}
for (ck = i = 0; i < DEV_BSIZE; i++)
ck += mbr[i];
for (m = i = 0; i < numbig; i++) {
be = &big->disk[i];
#ifdef GEOM_DEBUG
printf("match %s with %d "
"dev ck %x bios ck %x\n", device_xname(dv), i,
ck, be->cksum);
#endif
if (be->flags & BI_GEOM_INVALID)
continue;
if (be->cksum == ck &&
memcmp(&mbr[MBR_PART_OFFSET], be->mbrparts,
MBR_PART_COUNT *
sizeof(struct mbr_partition)) == 0) {
#ifdef GEOM_DEBUG
printf("matched BIOS disk %x with %s\n",
be->dev, device_xname(dv));
#endif
x86_alldisks->dl_nativedisks[n].
ni_biosmatches[m++] = i;
}
}
x86_alldisks->dl_nativedisks[n].ni_nmatches = m;
}
}
deviter_release(&di);
}
/*
* nfs statfs call
*/
static int
nfs_statfs(struct mount *mp, struct statfs *sbp)
{
struct vnode *vp;
struct thread *td;
struct nfs_statfs *sfp;
caddr_t bpos, dpos;
struct nfsmount *nmp = VFSTONFS(mp);
int error = 0, v3 = (nmp->nm_flag & NFSMNT_NFSV3), retattr;
struct mbuf *mreq, *mrep, *md, *mb;
struct nfsnode *np;
u_quad_t tquad;
td = curthread;
#ifndef nolint
sfp = NULL;
#endif
error = vfs_busy(mp, MBF_NOWAIT);
if (error)
return (error);
error = nfs_nget(mp, (nfsfh_t *)nmp->nm_fh, nmp->nm_fhsize, &np, LK_EXCLUSIVE);
if (error) {
vfs_unbusy(mp);
return (error);
}
vp = NFSTOV(np);
mtx_lock(&nmp->nm_mtx);
if (v3 && (nmp->nm_state & NFSSTA_GOTFSINFO) == 0) {
mtx_unlock(&nmp->nm_mtx);
(void)nfs_fsinfo(nmp, vp, td->td_ucred, td);
} else
mtx_unlock(&nmp->nm_mtx);
nfsstats.rpccnt[NFSPROC_FSSTAT]++;
mreq = nfsm_reqhead(vp, NFSPROC_FSSTAT, NFSX_FH(v3));
mb = mreq;
bpos = mtod(mb, caddr_t);
nfsm_fhtom(vp, v3);
nfsm_request(vp, NFSPROC_FSSTAT, td, td->td_ucred);
if (v3)
nfsm_postop_attr(vp, retattr);
if (error) {
if (mrep != NULL)
m_freem(mrep);
goto nfsmout;
}
sfp = nfsm_dissect(struct nfs_statfs *, NFSX_STATFS(v3));
mtx_lock(&nmp->nm_mtx);
sbp->f_iosize = nfs_iosize(nmp);
mtx_unlock(&nmp->nm_mtx);
if (v3) {
sbp->f_bsize = NFS_FABLKSIZE;
tquad = fxdr_hyper(&sfp->sf_tbytes);
sbp->f_blocks = tquad / NFS_FABLKSIZE;
tquad = fxdr_hyper(&sfp->sf_fbytes);
sbp->f_bfree = tquad / NFS_FABLKSIZE;
tquad = fxdr_hyper(&sfp->sf_abytes);
sbp->f_bavail = tquad / NFS_FABLKSIZE;
sbp->f_files = (fxdr_unsigned(int32_t,
sfp->sf_tfiles.nfsuquad[1]) & 0x7fffffff);
sbp->f_ffree = (fxdr_unsigned(int32_t,
sfp->sf_ffiles.nfsuquad[1]) & 0x7fffffff);
} else {
sbp->f_bsize = fxdr_unsigned(int32_t, sfp->sf_bsize);
sbp->f_blocks = fxdr_unsigned(int32_t, sfp->sf_blocks);
sbp->f_bfree = fxdr_unsigned(int32_t, sfp->sf_bfree);
sbp->f_bavail = fxdr_unsigned(int32_t, sfp->sf_bavail);
sbp->f_files = 0;
sbp->f_ffree = 0;
}
m_freem(mrep);
nfsmout:
vput(vp);
vfs_unbusy(mp);
return (error);
}
int
nnpfs_fhopen (d_thread_t *proc,
struct nnpfs_fhandle_t *fhp,
int user_flags,
register_t *retval)
{
int error;
struct vnode *vp;
#ifdef HAVE_FREEBSD_THREAD
struct ucred *cred = proc->td_proc->p_ucred;
#else
struct ucred *cred = proc->p_ucred;
#endif
int flags = FFLAGS(user_flags);
int index;
struct file *fp;
int mode;
struct nnpfs_fhandle_t fh;
NNPFSDEB(XDEBVFOPS, ("nnpfs_fhopen: flags = %d\n", user_flags));
error = copyin (fhp, &fh, sizeof(fh));
if (error)
return error;
error = nnpfs_fhlookup (proc, &fh, &vp);
NNPFSDEB(XDEBVFOPS, ("nnpfs_fhlookup returned %d\n", error));
if (error)
return error;
switch (vp->v_type) {
case VDIR :
case VREG :
break;
case VLNK :
error = EMLINK;
goto out;
default :
error = EOPNOTSUPP;
goto out;
}
mode = 0;
if (flags & FWRITE) {
switch (vp->v_type) {
case VREG :
break;
case VDIR :
error = EISDIR;
goto out;
default :
error = EOPNOTSUPP;
goto out;
}
error = vn_writechk (vp);
if (error)
goto out;
mode |= VWRITE;
}
if (flags & FREAD)
mode |= VREAD;
if (mode) {
error = VOP_ACCESS(vp, mode, cred, proc);
if (error)
goto out;
}
error = VOP_OPEN(vp, flags, cred, proc);
if (error)
goto out;
error = falloc(proc, &fp, &index);
if (error)
goto out;
if (flags & FWRITE)
vp->v_writecount++;
#if defined(__FreeBSD_version) && __FreeBSD_version >= 300000
if (vp->v_type == VREG) {
#ifdef HAVE_FREEBSD_THREAD
error = nnpfs_vfs_object_create(vp, proc, proc->td_proc->p_ucred);
#else
error = nnpfs_vfs_object_create(vp, proc, proc->p_ucred);
#endif
if (error)
goto out;
}
#endif
fp->f_flag = flags & FMASK;
fp->f_type = DTYPE_VNODE;
fp->f_ops = &vnops;
fp->f_data = (caddr_t)vp;
nnpfs_vfs_unlock(vp, proc);
*retval = index;
#ifdef FILE_UNUSE
FILE_UNUSE(fp, proc);
#endif
#ifdef __APPLE__
*fdflags(proc, index) &= ~UF_RESERVED;
#endif
return 0;
out:
NNPFSDEB(XDEBVFOPS, ("nnpfs_fhopen: error = %d\n", error));
vput(vp);
return error;
}
/*
* 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);
}
static int
ntfs_mount(struct mount *mp)
{
int err = 0, error;
struct vnode *devvp;
struct nameidata ndp;
struct thread *td;
char *from;
td = curthread;
if (vfs_filteropt(mp->mnt_optnew, ntfs_opts))
return (EINVAL);
/* Force mount as read-only. */
MNT_ILOCK(mp);
mp->mnt_flag |= MNT_RDONLY;
MNT_IUNLOCK(mp);
from = vfs_getopts(mp->mnt_optnew, "from", &error);
if (error)
return (error);
/*
* If updating, check whether changing from read-only to
* read/write.
*/
if (mp->mnt_flag & MNT_UPDATE) {
if (vfs_flagopt(mp->mnt_optnew, "export", NULL, 0)) {
/* Process export requests in vfs_mount.c */
return (0);
} else {
printf("ntfs_mount(): MNT_UPDATE not supported\n");
return (EINVAL);
}
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
NDINIT(&ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, from, td);
err = namei(&ndp);
if (err)
return (err);
NDFREE(&ndp, NDF_ONLY_PNBUF);
devvp = ndp.ni_vp;
if (!vn_isdisk(devvp, &err)) {
vput(devvp);
return (err);
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
err = VOP_ACCESS(devvp, VREAD, td->td_ucred, td);
if (err)
err = priv_check(td, PRIV_VFS_MOUNT_PERM);
if (err) {
vput(devvp);
return (err);
}
/*
* Since this is a new mount, we want the names for the device and
* the mount point copied in. If an error occurs, the mountpoint is
* discarded by the upper level code. Note that vfs_mount() handles
* copying the mountpoint f_mntonname for us, so we don't have to do
* it here unless we want to set it to something other than "path"
* for some rason.
*/
err = ntfs_mountfs(devvp, mp, td);
if (err == 0) {
/* Save "mounted from" info for mount point. */
vfs_mountedfrom(mp, from);
} else
vrele(devvp);
return (err);
}
static int
tmpfs_lookup(struct vop_cachedlookup_args *v)
{
struct vnode *dvp = v->a_dvp;
struct vnode **vpp = v->a_vpp;
struct componentname *cnp = v->a_cnp;
struct tmpfs_dirent *de;
struct tmpfs_node *dnode;
int error;
dnode = VP_TO_TMPFS_DIR(dvp);
*vpp = NULLVP;
/* Check accessibility of requested node as a first step. */
error = VOP_ACCESS(dvp, VEXEC, cnp->cn_cred, cnp->cn_thread);
if (error != 0)
goto out;
/* We cannot be requesting the parent directory of the root node. */
MPASS(IMPLIES(dnode->tn_type == VDIR &&
dnode->tn_dir.tn_parent == dnode,
!(cnp->cn_flags & ISDOTDOT)));
TMPFS_ASSERT_LOCKED(dnode);
if (dnode->tn_dir.tn_parent == NULL) {
error = ENOENT;
goto out;
}
if (cnp->cn_flags & ISDOTDOT) {
error = vn_vget_ino_gen(dvp, tmpfs_vn_get_ino_alloc,
dnode->tn_dir.tn_parent, cnp->cn_lkflags, vpp);
if (error != 0)
goto out;
} else if (cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') {
VREF(dvp);
*vpp = dvp;
error = 0;
} else {
de = tmpfs_dir_lookup(dnode, NULL, cnp);
if (de != NULL && de->td_node == NULL)
cnp->cn_flags |= ISWHITEOUT;
if (de == NULL || de->td_node == NULL) {
/* The entry was not found in the directory.
* This is OK if we are creating or renaming an
* entry and are working on the last component of
* the path name. */
if ((cnp->cn_flags & ISLASTCN) &&
(cnp->cn_nameiop == CREATE || \
cnp->cn_nameiop == RENAME ||
(cnp->cn_nameiop == DELETE &&
cnp->cn_flags & DOWHITEOUT &&
cnp->cn_flags & ISWHITEOUT))) {
error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred,
cnp->cn_thread);
if (error != 0)
goto out;
/* Keep the component name in the buffer for
* future uses. */
cnp->cn_flags |= SAVENAME;
error = EJUSTRETURN;
} else
error = ENOENT;
} else {
struct tmpfs_node *tnode;
/* The entry was found, so get its associated
* tmpfs_node. */
tnode = de->td_node;
/* If we are not at the last path component and
* found a non-directory or non-link entry (which
* may itself be pointing to a directory), raise
* an error. */
if ((tnode->tn_type != VDIR &&
tnode->tn_type != VLNK) &&
!(cnp->cn_flags & ISLASTCN)) {
error = ENOTDIR;
goto out;
}
/* If we are deleting or renaming the entry, keep
* track of its tmpfs_dirent so that it can be
* easily deleted later. */
if ((cnp->cn_flags & ISLASTCN) &&
(cnp->cn_nameiop == DELETE ||
cnp->cn_nameiop == RENAME)) {
error = VOP_ACCESS(dvp, VWRITE, cnp->cn_cred,
cnp->cn_thread);
if (error != 0)
goto out;
/* Allocate a new vnode on the matching entry. */
error = tmpfs_alloc_vp(dvp->v_mount, tnode,
cnp->cn_lkflags, vpp);
if (error != 0)
goto out;
if ((dnode->tn_mode & S_ISTXT) &&
VOP_ACCESS(dvp, VADMIN, cnp->cn_cred, cnp->cn_thread) &&
VOP_ACCESS(*vpp, VADMIN, cnp->cn_cred, cnp->cn_thread)) {
error = EPERM;
vput(*vpp);
*vpp = NULL;
goto out;
}
cnp->cn_flags |= SAVENAME;
} else {
error = tmpfs_alloc_vp(dvp->v_mount, tnode,
cnp->cn_lkflags, vpp);
if (error != 0)
goto out;
}
}
}
/* Store the result of this lookup in the cache. Avoid this if the
* request was for creation, as it does not improve timings on
* emprical tests. */
if ((cnp->cn_flags & MAKEENTRY) != 0)
cache_enter(dvp, *vpp, cnp);
out:
/* If there were no errors, *vpp cannot be null and it must be
* locked. */
MPASS(IFF(error == 0, *vpp != NULLVP && VOP_ISLOCKED(*vpp)));
return error;
}
/*
* Convert a component of a pathname into a pointer to a locked inode.
* This is a very central and rather complicated routine.
* If the file system is not maintained in a strict tree hierarchy,
* this can result in a deadlock situation (see comments in code below).
*
* The cnp->cn_nameiop 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 and the target of the pathname
* exists, lookup returns both the target and its parent directory locked.
* When creating or renaming and LOCKPARENT is specified, the target may
* not be ".". When deleting and LOCKPARENT is specified, the target may
* be "."., but the caller must check to ensure it does an vrele and vput
* instead of two vputs.
*
* Overall outline of ufs_lookup:
*
* search for name in directory, to found or notfound
* notfound:
* if creating, return locked directory, leaving info on available slots
* else return error
* found:
* if at end of path and deleting, return information to allow delete
* if at end of path and rewriting (RENAME and LOCKPARENT), lock target
* inode and return info to allow rewrite
* if not at end, add name to cache; if at end and neither creating
* nor deleting, add name to cache
*
* ext2_lookup(struct vnode *a_dvp, struct vnode **a_vpp,
* struct componentname *a_cnp)
*/
int
ext2_lookup(struct vop_old_lookup_args *ap)
{
struct vnode *vdp; /* vnode for directory being searched */
struct inode *dp; /* inode for directory being searched */
struct buf *bp; /* a buffer of directory entries */
struct ext2_dir_entry_2 *ep; /* the current directory entry */
int entryoffsetinblock; /* offset of ep in bp's buffer */
enum {NONE, COMPACT, FOUND} slotstatus;
doff_t slotoffset; /* offset of area with free space */
int slotsize; /* size of area at slotoffset */
int slotfreespace; /* amount of space free in slot */
int slotneeded; /* size of the entry we're seeking */
int numdirpasses; /* strategy for directory search */
doff_t endsearch; /* offset to end directory search */
doff_t prevoff; /* prev entry dp->i_offset */
struct vnode *pdp; /* saved dp during symlink work */
struct vnode *tdp; /* returned by VFS_VGET */
doff_t enduseful; /* pointer past last used dir slot */
u_long bmask; /* block offset mask */
int lockparent; /* 1 => lockparent flag is set */
int wantparent; /* 1 => wantparent or lockparent flag */
int namlen, error;
struct vnode **vpp = ap->a_vpp;
struct componentname *cnp = ap->a_cnp;
struct ucred *cred = cnp->cn_cred;
int flags = cnp->cn_flags;
int nameiop = cnp->cn_nameiop;
int DIRBLKSIZ = VTOI(ap->a_dvp)->i_e2fs->s_blocksize;
bp = NULL;
slotoffset = -1;
*vpp = NULL;
vdp = ap->a_dvp;
dp = VTOI(vdp);
lockparent = flags & CNP_LOCKPARENT;
wantparent = flags & (CNP_LOCKPARENT|CNP_WANTPARENT);
/*
* We now have a segment name to search for, and a directory to search.
*/
/*
* Suppress search for slots unless creating
* file and at end of pathname, in which case
* we watch for a place to put the new file in
* case it doesn't already exist.
*/
slotstatus = FOUND;
slotfreespace = slotsize = slotneeded = 0;
if (nameiop == NAMEI_CREATE || nameiop == NAMEI_RENAME) {
slotstatus = NONE;
slotneeded = EXT2_DIR_REC_LEN(cnp->cn_namelen);
/* was
slotneeded = (sizeof(struct direct) - MAXNAMLEN +
cnp->cn_namelen + 3) &~ 3; */
}
/*
* If there is cached information on a previous search of
* this directory, pick up where we last left off.
* We cache only lookups as these are the most common
* and have the greatest payoff. Caching CREATE has little
* benefit as it usually must search the entire directory
* to determine that the entry does not exist. Caching the
* location of the last DELETE or RENAME has not reduced
* profiling time and hence has been removed in the interest
* of simplicity.
*/
bmask = VFSTOEXT2(vdp->v_mount)->um_mountp->mnt_stat.f_iosize - 1;
if (nameiop != NAMEI_LOOKUP || dp->i_diroff == 0 ||
dp->i_diroff > dp->i_size) {
entryoffsetinblock = 0;
dp->i_offset = 0;
numdirpasses = 1;
} else {
dp->i_offset = dp->i_diroff;
if ((entryoffsetinblock = dp->i_offset & bmask) &&
(error = EXT2_BLKATOFF(vdp, (off_t)dp->i_offset, NULL, &bp)))
return (error);
numdirpasses = 2;
}
prevoff = dp->i_offset;
endsearch = roundup(dp->i_size, DIRBLKSIZ);
enduseful = 0;
searchloop:
while (dp->i_offset < endsearch) {
/*
* If necessary, get the next directory block.
*/
if ((dp->i_offset & bmask) == 0) {
if (bp != NULL)
brelse(bp);
if ((error =
EXT2_BLKATOFF(vdp, (off_t)dp->i_offset, NULL, &bp)) != 0)
return (error);
entryoffsetinblock = 0;
}
/*
* If still looking for a slot, and at a DIRBLKSIZE
* boundary, have to start looking for free space again.
*/
if (slotstatus == NONE &&
(entryoffsetinblock & (DIRBLKSIZ - 1)) == 0) {
slotoffset = -1;
slotfreespace = 0;
}
/*
* Get pointer to next entry.
* Full validation checks are slow, so we only check
* enough to insure forward progress through the
* directory. Complete checks can be run by patching
* "dirchk" to be true.
*/
ep = (struct ext2_dir_entry_2 *)
((char *)bp->b_data + entryoffsetinblock);
if (ep->rec_len == 0 ||
(dirchk && ext2_dirbadentry(vdp, ep, entryoffsetinblock))) {
int i;
ext2_dirbad(dp, dp->i_offset, "mangled entry");
i = DIRBLKSIZ - (entryoffsetinblock & (DIRBLKSIZ - 1));
dp->i_offset += i;
entryoffsetinblock += i;
continue;
}
/*
* If an appropriate sized slot has not yet been found,
* check to see if one is available. Also accumulate space
* in the current block so that we can determine if
* compaction is viable.
*/
if (slotstatus != FOUND) {
int size = ep->rec_len;
if (ep->inode != 0)
size -= EXT2_DIR_REC_LEN(ep->name_len);
if (size > 0) {
if (size >= slotneeded) {
slotstatus = FOUND;
slotoffset = dp->i_offset;
slotsize = ep->rec_len;
} else if (slotstatus == NONE) {
slotfreespace += size;
if (slotoffset == -1)
slotoffset = dp->i_offset;
if (slotfreespace >= slotneeded) {
slotstatus = COMPACT;
slotsize = dp->i_offset +
ep->rec_len - slotoffset;
}
}
}
}
/*
* Check for a name match.
*/
if (ep->inode) {
namlen = ep->name_len;
if (namlen == cnp->cn_namelen &&
!bcmp(cnp->cn_nameptr, ep->name,
(unsigned)namlen)) {
/*
* Save directory entry's inode number and
* reclen in ndp->ni_ufs area, and release
* directory buffer.
*/
dp->i_ino = ep->inode;
dp->i_reclen = ep->rec_len;
goto found;
}
}
prevoff = dp->i_offset;
dp->i_offset += ep->rec_len;
entryoffsetinblock += ep->rec_len;
if (ep->inode)
enduseful = dp->i_offset;
}
/* notfound: */
/*
* If we started in the middle of the directory and failed
* to find our target, we must check the beginning as well.
*/
if (numdirpasses == 2) {
numdirpasses--;
dp->i_offset = 0;
endsearch = dp->i_diroff;
goto searchloop;
}
if (bp != NULL)
brelse(bp);
/*
* If creating, and at end of pathname and current
* directory has not been removed, then can consider
* allowing file to be created.
*/
if ((nameiop == NAMEI_CREATE || nameiop == NAMEI_RENAME) &&
dp->i_nlink != 0) {
/*
* Access for write is interpreted as allowing
* creation of files in the directory.
*/
if ((error = VOP_EACCESS(vdp, VWRITE, cred)) != 0)
return (error);
/*
* Return an indication of where the new directory
* entry should be put. If we didn't find a slot,
* then set dp->i_count to 0 indicating
* that the new slot belongs at the end of the
* directory. If we found a slot, then the new entry
* can be put in the range from dp->i_offset to
* dp->i_offset + dp->i_count.
*/
if (slotstatus == NONE) {
dp->i_offset = roundup(dp->i_size, DIRBLKSIZ);
dp->i_count = 0;
enduseful = dp->i_offset;
} else {
dp->i_offset = slotoffset;
dp->i_count = slotsize;
if (enduseful < slotoffset + slotsize)
enduseful = slotoffset + slotsize;
}
dp->i_endoff = roundup(enduseful, DIRBLKSIZ);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
/*
* We return with the directory locked, so that
* the parameters we set up above will still be
* valid if we actually decide to do a direnter().
* We return ni_vp == NULL to indicate that the entry
* does not currently exist; we leave a pointer to
* the (locked) directory inode in ndp->ni_dvp.
* The pathname buffer is saved so that the name
* can be obtained later.
*
* NB - if the directory is unlocked, then this
* information cannot be used.
*/
if (!lockparent)
vn_unlock(vdp);
return (EJUSTRETURN);
}
return (ENOENT);
found:
/*
* Check that directory length properly reflects presence
* of this entry.
*/
if (entryoffsetinblock + EXT2_DIR_REC_LEN(ep->name_len)
> dp->i_size) {
ext2_dirbad(dp, dp->i_offset, "i_size too small");
dp->i_size = entryoffsetinblock+EXT2_DIR_REC_LEN(ep->name_len);
dp->i_flag |= IN_CHANGE | IN_UPDATE;
}
brelse(bp);
/*
* Found component in pathname.
* If the final component of path name, save information
* in the cache as to where the entry was found.
*/
if (nameiop == NAMEI_LOOKUP)
dp->i_diroff = dp->i_offset &~ (DIRBLKSIZ - 1);
/*
* If deleting, and at end of pathname, return
* parameters which can be used to remove file.
* If the wantparent flag isn't set, we return only
* the directory (in ndp->ni_dvp), otherwise we go
* on and lock the inode, being careful with ".".
*/
if (nameiop == NAMEI_DELETE) {
/*
* Write access to directory required to delete files.
*/
if ((error = VOP_EACCESS(vdp, VWRITE, cred)) != 0)
return (error);
/*
* Return pointer to current entry in dp->i_offset,
* and distance past previous entry (if there
* is a previous entry in this block) in dp->i_count.
* Save directory inode pointer in ndp->ni_dvp for dirremove().
*/
if ((dp->i_offset & (DIRBLKSIZ - 1)) == 0)
dp->i_count = 0;
else
dp->i_count = dp->i_offset - prevoff;
if (dp->i_number == dp->i_ino) {
vref(vdp);
*vpp = vdp;
return (0);
}
if ((error = VFS_VGET(vdp->v_mount, NULL, dp->i_ino, &tdp)) != 0)
return (error);
/*
* If directory is "sticky", then user must own
* the directory, or the file in it, else she
* may not delete it (unless she's root). This
* implements append-only directories.
*/
if ((dp->i_mode & ISVTX) &&
cred->cr_uid != 0 &&
cred->cr_uid != dp->i_uid &&
VTOI(tdp)->i_uid != cred->cr_uid) {
vput(tdp);
return (EPERM);
}
*vpp = tdp;
if (!lockparent)
vn_unlock(vdp);
return (0);
}
/*
* If rewriting (RENAME), return the inode and the
* information required to rewrite the present directory
* Must get inode of directory entry to verify it's a
* regular file, or empty directory.
*/
if (nameiop == NAMEI_RENAME && wantparent) {
if ((error = VOP_EACCESS(vdp, VWRITE, cred)) != 0)
return (error);
/*
* Careful about locking second inode.
* This can only occur if the target is ".".
*/
if (dp->i_number == dp->i_ino)
return (EISDIR);
if ((error = VFS_VGET(vdp->v_mount, NULL, dp->i_ino, &tdp)) != 0)
return (error);
*vpp = tdp;
if (!lockparent)
vn_unlock(vdp);
return (0);
}
/*
* Step through the translation in the name. We do not `vput' the
* directory because we may need it again if a symbolic link
* is relative to the current directory. Instead we save it
* unlocked as "pdp". We must get the target inode before unlocking
* the directory to insure that the inode will not be removed
* before we get it. We prevent deadlock by always fetching
* inodes from the root, moving down the directory tree. Thus
* when following backward pointers ".." we must unlock the
* parent directory before getting the requested directory.
* There is a potential race condition here if both the current
* and parent directories are removed before the VFS_VGET for the
* inode associated with ".." returns. We hope that this occurs
* infrequently since we cannot avoid this race condition without
* implementing a sophisticated deadlock detection algorithm.
* Note also that this simple deadlock detection scheme will not
* work if the file system has any hard links other than ".."
* that point backwards in the directory structure.
*/
pdp = vdp;
if (flags & CNP_ISDOTDOT) {
vn_unlock(pdp); /* race to get the inode */
if ((error = VFS_VGET(vdp->v_mount, NULL, dp->i_ino, &tdp)) != 0) {
vn_lock(pdp, LK_EXCLUSIVE | LK_RETRY);
return (error);
}
if (lockparent && (error = vn_lock(pdp, LK_EXCLUSIVE))) {
vput(tdp);
return (error);
}
*vpp = tdp;
} else if (dp->i_number == dp->i_ino) {
vref(vdp); /* we want ourself, ie "." */
*vpp = vdp;
} else {
if ((error = VFS_VGET(vdp->v_mount, NULL, dp->i_ino, &tdp)) != 0)
return (error);
if (!lockparent)
vn_unlock(pdp);
*vpp = tdp;
}
return (0);
}
static int
tmpfs_rename(struct vop_rename_args *v)
{
struct vnode *fdvp = v->a_fdvp;
struct vnode *fvp = v->a_fvp;
struct componentname *fcnp = v->a_fcnp;
struct vnode *tdvp = v->a_tdvp;
struct vnode *tvp = v->a_tvp;
struct componentname *tcnp = v->a_tcnp;
struct mount *mp = NULL;
char *newname;
int error;
struct tmpfs_dirent *de;
struct tmpfs_mount *tmp;
struct tmpfs_node *fdnode;
struct tmpfs_node *fnode;
struct tmpfs_node *tnode;
struct tmpfs_node *tdnode;
MPASS(VOP_ISLOCKED(tdvp));
MPASS(IMPLIES(tvp != NULL, VOP_ISLOCKED(tvp)));
MPASS(fcnp->cn_flags & HASBUF);
MPASS(tcnp->cn_flags & HASBUF);
/* Disallow cross-device renames.
* XXX Why isn't this done by the caller? */
if (fvp->v_mount != tdvp->v_mount ||
(tvp != NULL && fvp->v_mount != tvp->v_mount)) {
error = EXDEV;
goto out;
}
/* If source and target are the same file, there is nothing to do. */
if (fvp == tvp) {
error = 0;
goto out;
}
/* If we need to move the directory between entries, lock the
* source so that we can safely operate on it. */
if (fdvp != tdvp && fdvp != tvp) {
if (vn_lock(fdvp, LK_EXCLUSIVE | LK_NOWAIT) != 0) {
mp = tdvp->v_mount;
error = vfs_busy(mp, 0);
if (error != 0) {
mp = NULL;
goto out;
}
error = tmpfs_rename_relock(fdvp, &fvp, tdvp, &tvp,
fcnp, tcnp);
if (error != 0) {
vfs_unbusy(mp);
return (error);
}
ASSERT_VOP_ELOCKED(fdvp,
"tmpfs_rename: fdvp not locked");
ASSERT_VOP_ELOCKED(tdvp,
"tmpfs_rename: tdvp not locked");
if (tvp != NULL)
ASSERT_VOP_ELOCKED(tvp,
"tmpfs_rename: tvp not locked");
if (fvp == tvp) {
error = 0;
goto out_locked;
}
}
}
tmp = VFS_TO_TMPFS(tdvp->v_mount);
tdnode = VP_TO_TMPFS_DIR(tdvp);
tnode = (tvp == NULL) ? NULL : VP_TO_TMPFS_NODE(tvp);
fdnode = VP_TO_TMPFS_DIR(fdvp);
fnode = VP_TO_TMPFS_NODE(fvp);
de = tmpfs_dir_lookup(fdnode, fnode, fcnp);
/* Entry can disappear before we lock fdvp,
* also avoid manipulating '.' and '..' entries. */
if (de == NULL) {
if ((fcnp->cn_flags & ISDOTDOT) != 0 ||
(fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.'))
error = EINVAL;
else
error = ENOENT;
goto out_locked;
}
MPASS(de->td_node == fnode);
/* If re-naming a directory to another preexisting directory
* ensure that the target directory is empty so that its
* removal causes no side effects.
* Kern_rename gurantees the destination to be a directory
* if the source is one. */
if (tvp != NULL) {
MPASS(tnode != NULL);
if ((tnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND)) ||
(tdnode->tn_flags & (APPEND | IMMUTABLE))) {
error = EPERM;
goto out_locked;
}
if (fnode->tn_type == VDIR && tnode->tn_type == VDIR) {
if (tnode->tn_size > 0) {
error = ENOTEMPTY;
goto out_locked;
}
} else if (fnode->tn_type == VDIR && tnode->tn_type != VDIR) {
error = ENOTDIR;
goto out_locked;
} else if (fnode->tn_type != VDIR && tnode->tn_type == VDIR) {
error = EISDIR;
goto out_locked;
} else {
MPASS(fnode->tn_type != VDIR &&
tnode->tn_type != VDIR);
}
}
if ((fnode->tn_flags & (NOUNLINK | IMMUTABLE | APPEND))
|| (fdnode->tn_flags & (APPEND | IMMUTABLE))) {
error = EPERM;
goto out_locked;
}
/* Ensure that we have enough memory to hold the new name, if it
* has to be changed. */
if (fcnp->cn_namelen != tcnp->cn_namelen ||
bcmp(fcnp->cn_nameptr, tcnp->cn_nameptr, fcnp->cn_namelen) != 0) {
newname = malloc(tcnp->cn_namelen, M_TMPFSNAME, M_WAITOK);
} else
newname = NULL;
/* If the node is being moved to another directory, we have to do
* the move. */
if (fdnode != tdnode) {
/* In case we are moving a directory, we have to adjust its
* parent to point to the new parent. */
if (de->td_node->tn_type == VDIR) {
struct tmpfs_node *n;
/* Ensure the target directory is not a child of the
* directory being moved. Otherwise, we'd end up
* with stale nodes. */
n = tdnode;
/* TMPFS_LOCK garanties that no nodes are freed while
* traversing the list. Nodes can only be marked as
* removed: tn_parent == NULL. */
TMPFS_LOCK(tmp);
TMPFS_NODE_LOCK(n);
while (n != n->tn_dir.tn_parent) {
struct tmpfs_node *parent;
if (n == fnode) {
TMPFS_NODE_UNLOCK(n);
TMPFS_UNLOCK(tmp);
error = EINVAL;
if (newname != NULL)
free(newname, M_TMPFSNAME);
goto out_locked;
}
parent = n->tn_dir.tn_parent;
TMPFS_NODE_UNLOCK(n);
if (parent == NULL) {
n = NULL;
break;
}
TMPFS_NODE_LOCK(parent);
if (parent->tn_dir.tn_parent == NULL) {
TMPFS_NODE_UNLOCK(parent);
n = NULL;
break;
}
n = parent;
}
TMPFS_UNLOCK(tmp);
if (n == NULL) {
error = EINVAL;
if (newname != NULL)
free(newname, M_TMPFSNAME);
goto out_locked;
}
TMPFS_NODE_UNLOCK(n);
/* Adjust the parent pointer. */
TMPFS_VALIDATE_DIR(fnode);
TMPFS_NODE_LOCK(de->td_node);
de->td_node->tn_dir.tn_parent = tdnode;
TMPFS_NODE_UNLOCK(de->td_node);
/* As a result of changing the target of the '..'
* entry, the link count of the source and target
* directories has to be adjusted. */
TMPFS_NODE_LOCK(tdnode);
TMPFS_ASSERT_LOCKED(tdnode);
tdnode->tn_links++;
TMPFS_NODE_UNLOCK(tdnode);
TMPFS_NODE_LOCK(fdnode);
TMPFS_ASSERT_LOCKED(fdnode);
fdnode->tn_links--;
TMPFS_NODE_UNLOCK(fdnode);
}
}
/* Do the move: just remove the entry from the source directory
* and insert it into the target one. */
tmpfs_dir_detach(fdvp, de);
if (fcnp->cn_flags & DOWHITEOUT)
tmpfs_dir_whiteout_add(fdvp, fcnp);
if (tcnp->cn_flags & ISWHITEOUT)
tmpfs_dir_whiteout_remove(tdvp, tcnp);
/* If the name has changed, we need to make it effective by changing
* it in the directory entry. */
if (newname != NULL) {
MPASS(tcnp->cn_namelen <= MAXNAMLEN);
free(de->ud.td_name, M_TMPFSNAME);
de->ud.td_name = newname;
tmpfs_dirent_init(de, tcnp->cn_nameptr, tcnp->cn_namelen);
fnode->tn_status |= TMPFS_NODE_CHANGED;
tdnode->tn_status |= TMPFS_NODE_MODIFIED;
}
/* If we are overwriting an entry, we have to remove the old one
* from the target directory. */
if (tvp != NULL) {
struct tmpfs_dirent *tde;
/* Remove the old entry from the target directory. */
tde = tmpfs_dir_lookup(tdnode, tnode, tcnp);
tmpfs_dir_detach(tdvp, tde);
/* Free the directory entry we just deleted. Note that the
* node referred by it will not be removed until the vnode is
* really reclaimed. */
tmpfs_free_dirent(VFS_TO_TMPFS(tvp->v_mount), tde);
}
tmpfs_dir_attach(tdvp, de);
cache_purge(fvp);
if (tvp != NULL)
cache_purge(tvp);
cache_purge_negative(tdvp);
error = 0;
out_locked:
if (fdvp != tdvp && fdvp != tvp)
VOP_UNLOCK(fdvp, 0);
out:
/* Release target nodes. */
/* XXX: I don't understand when tdvp can be the same as tvp, but
* other code takes care of this... */
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp != NULL)
vput(tvp);
/* Release source nodes. */
vrele(fdvp);
vrele(fvp);
if (mp != NULL)
vfs_unbusy(mp);
return error;
}
static int
puffs_vnop_rename(struct vop_nrename_args *ap)
{
PUFFS_MSG_VARS(vn, rename);
struct nchandle *fnch = ap->a_fnch;
struct nchandle *tnch = ap->a_tnch;
struct vnode *fdvp = ap->a_fdvp;
struct vnode *fvp = fnch->ncp->nc_vp;
struct vnode *tdvp = ap->a_tdvp;
struct vnode *tvp = tnch->ncp->nc_vp;
struct ucred *cred = ap->a_cred;
struct puffs_mount *pmp = MPTOPUFFSMP(fdvp->v_mount);
int error;
if (!EXISTSOP(pmp, RENAME))
return EOPNOTSUPP;
error = vget(tdvp, LK_EXCLUSIVE);
if (error != 0) {
DPRINTF(("puffs_vnop_rename: EAGAIN on tdvp vnode %p %s\n",
tdvp, tnch->ncp->nc_name));
return EAGAIN;
}
if (tvp != NULL) {
error = vget(tvp, LK_EXCLUSIVE);
if (error != 0) {
DPRINTF(("puffs_vnop_rename: EAGAIN on tvp vnode %p %s\n",
tvp, tnch->ncp->nc_name));
vput(tdvp);
return EAGAIN;
}
}
if ((fvp->v_mount != tdvp->v_mount) ||
(tvp && (fvp->v_mount != tvp->v_mount))) {
error = EXDEV;
goto out;
}
if (tvp) {
if (fvp->v_type == VDIR && tvp->v_type != VDIR) {
error = ENOTDIR;
goto out;
} else if (fvp->v_type != VDIR && tvp->v_type == VDIR) {
error = EISDIR;
goto out;
}
}
PUFFS_MSG_ALLOC(vn, rename);
rename_msg->pvnr_cookie_src = VPTOPNC(fvp);
rename_msg->pvnr_cookie_targdir = VPTOPNC(tdvp);
if (tvp)
rename_msg->pvnr_cookie_targ = VPTOPNC(tvp);
else
rename_msg->pvnr_cookie_targ = NULL;
puffs_makecn(&rename_msg->pvnr_cn_src, &rename_msg->pvnr_cn_src_cred,
fnch->ncp, cred);
puffs_makecn(&rename_msg->pvnr_cn_targ, &rename_msg->pvnr_cn_targ_cred,
tnch->ncp, cred);
puffs_msg_setinfo(park_rename, PUFFSOP_VN,
PUFFS_VN_RENAME, VPTOPNC(fdvp));
PUFFS_MSG_ENQUEUEWAIT2(pmp, park_rename, fdvp->v_data, NULL, error);
PUFFS_MSG_RELEASE(rename);
error = checkerr(pmp, error, __func__);
if (error == 0)
puffs_updatenode(VPTOPP(fvp), PUFFS_UPDATECTIME);
out:
if (tvp != NULL)
vn_unlock(tvp);
if (tdvp != tvp)
vn_unlock(tdvp);
if (error == 0)
cache_rename(fnch, tnch);
if (tvp != NULL)
vrele(tvp);
vrele(tdvp);
return error;
}
int
ntfs_lookup(void *v)
{
struct vop_lookup_args *ap = v;
struct vnode *dvp = ap->a_dvp;
struct ntnode *dip = VTONT(dvp);
struct ntfsmount *ntmp = dip->i_mp;
struct componentname *cnp = ap->a_cnp;
struct ucred *cred = cnp->cn_cred;
int error;
int lockparent = cnp->cn_flags & LOCKPARENT;
struct proc *p = cnp->cn_proc;
#if NTFS_DEBUG
int wantparent = cnp->cn_flags & (LOCKPARENT|WANTPARENT);
#endif
DPRINTF("ntfs_lookup: \"%.*s\" (%ld bytes) in %u, lp: %d, wp: %d \n",
(unsigned int)cnp->cn_namelen, cnp->cn_nameptr, cnp->cn_namelen,
dip->i_number, lockparent, wantparent);
error = VOP_ACCESS(dvp, VEXEC, cred, cnp->cn_proc);
if(error)
return (error);
if ((cnp->cn_flags & ISLASTCN) &&
(dvp->v_mount->mnt_flag & MNT_RDONLY) &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
return (EROFS);
/*
* We now have a segment name to search for, and a directory
* to search.
*
* Before tediously performing a linear scan of the directory,
* check the name cache to see if the directory/name pair
* we are looking for is known already.
*/
if ((error = cache_lookup(ap->a_dvp, ap->a_vpp, cnp)) >= 0)
return (error);
if(cnp->cn_namelen == 1 && cnp->cn_nameptr[0] == '.') {
DPRINTF("ntfs_lookup: faking . directory in %u\n",
dip->i_number);
vref(dvp);
*ap->a_vpp = dvp;
error = 0;
} else if (cnp->cn_flags & ISDOTDOT) {
struct ntvattr *vap;
DPRINTF("ntfs_lookup: faking .. directory in %u\n",
dip->i_number);
VOP_UNLOCK(dvp, 0, p);
cnp->cn_flags |= PDIRUNLOCK;
error = ntfs_ntvattrget(ntmp, dip, NTFS_A_NAME, NULL, 0, &vap);
if(error)
return (error);
DPRINTF("ntfs_lookup: parentdir: %u\n",
vap->va_a_name->n_pnumber);
error = VFS_VGET(ntmp->ntm_mountp,
vap->va_a_name->n_pnumber,ap->a_vpp);
ntfs_ntvattrrele(vap);
if (error) {
if (vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY, p) == 0)
cnp->cn_flags &= ~PDIRUNLOCK;
return (error);
}
if (lockparent && (cnp->cn_flags & ISLASTCN)) {
error = vn_lock(dvp, LK_EXCLUSIVE, p);
if (error) {
vput( *(ap->a_vpp) );
return (error);
}
cnp->cn_flags &= ~PDIRUNLOCK;
}
} else {
error = ntfs_ntlookupfile(ntmp, dvp, cnp, ap->a_vpp, p);
if (error) {
DPRINTF("ntfs_ntlookupfile: returned %d\n", error);
return (error);
}
DPRINTF("ntfs_lookup: found ino: %u\n",
VTONT(*ap->a_vpp)->i_number);
if(!lockparent || (cnp->cn_flags & ISLASTCN) == 0) {
VOP_UNLOCK(dvp, 0, p);
cnp->cn_flags |= PDIRUNLOCK;
}
}
if (cnp->cn_flags & MAKEENTRY)
cache_enter(dvp, *ap->a_vpp, cnp);
return (error);
}
VMBlockVFSMount(struct mount *mp, // IN: mount(2) parameters
struct thread *td) // IN: caller's thread context
#endif
{
struct VMBlockMount *xmp;
struct nameidata nd, *ndp = &nd;
struct vnode *lowerrootvp, *vp;
char *target;
char *pathname;
int len, error = 0;
VMBLOCKDEBUG("VMBlockVFSMount(mp = %p)\n", (void *)mp);
/*
* TODO: Strip out extraneous export & other misc cruft.
*/
/*
* Disallow the following:
* 1. Mounting over the system root.
* 2. Mount updates/remounts. (Reconsider for rw->ro, ro->rw?)
* 3. Mounting VMBlock on top of a VMBlock.
*/
if ((mp->mnt_flag & MNT_ROOTFS) ||
(mp->mnt_flag & MNT_UPDATE) ||
(mp->mnt_vnodecovered->v_op == &VMBlockVnodeOps)) {
return EOPNOTSUPP;
}
/*
* XXX Should only be unlocked if mnt_flag & MNT_UPDATE.
*/
ASSERT_VOP_UNLOCKED(mp->mnt_vnodecovered, "Covered vnode already locked!");
/*
* Look up path to lower layer (VMBlock source / DnD staging area).
* (E.g., in the command "mount /tmp/VMwareDnD /var/run/vmblock",
* /tmp/VMwareDnD is the staging area.)
*/
error = vfs_getopt(mp->mnt_optnew, "target", (void **)&target, &len);
if (error || target[len - 1] != '\0') {
return EINVAL;
}
pathname = uma_zalloc(VMBlockPathnameZone, M_WAITOK);
if (pathname == NULL) {
return ENOMEM;
}
if (strlcpy(pathname, target, MAXPATHLEN) >= MAXPATHLEN) {
uma_zfree(VMBlockPathnameZone, pathname);
return ENAMETOOLONG;
}
/*
* Find lower node and lock if not already locked.
*/
NDINIT(ndp, LOOKUP, FOLLOW|LOCKLEAF, UIO_SYSSPACE, target, compat_td);
error = namei(ndp);
if (error) {
NDFREE(ndp, 0);
uma_zfree(VMBlockPathnameZone, pathname);
return error;
}
NDFREE(ndp, NDF_ONLY_PNBUF);
/*
* Check multi VMBlock mount to avoid `lock against myself' panic.
*/
lowerrootvp = ndp->ni_vp;
if (lowerrootvp == VPTOVMB(mp->mnt_vnodecovered)->lowerVnode) {
VMBLOCKDEBUG("VMBlockVFSMount: multi vmblock mount?\n");
vput(lowerrootvp);
uma_zfree(VMBlockPathnameZone, pathname);
return EDEADLK;
}
xmp = malloc(sizeof *xmp, M_VMBLOCKFSMNT, M_WAITOK);
/*
* Record pointer (mountVFS) to the staging area's file system. Follow up
* by grabbing a VMBlockNode for our layer's root.
*/
xmp->mountVFS = lowerrootvp->v_mount;
error = VMBlockNodeGet(mp, lowerrootvp, &vp, pathname);
/*
* Make sure the node alias worked
*/
if (error) {
COMPAT_VOP_UNLOCK(vp, 0, compat_td);
vrele(lowerrootvp);
free(xmp, M_VMBLOCKFSMNT); /* XXX */
uma_zfree(VMBlockPathnameZone, pathname);
return error;
}
/*
* Record a reference to the new filesystem's root vnode & mark it as such.
*/
xmp->rootVnode = vp;
xmp->rootVnode->v_vflag |= VV_ROOT;
/*
* Unlock the node (either the lower or the alias)
*/
COMPAT_VOP_UNLOCK(vp, 0, compat_td);
/*
* If the staging area is a local filesystem, reflect that here, too. (We
* could potentially allow NFS staging areas.)
*/
MNT_ILOCK(mp);
mp->mnt_flag |= lowerrootvp->v_mount->mnt_flag & MNT_LOCAL;
#if __FreeBSD_version >= 600000 && __FreeBSD_version < 1000000
mp->mnt_kern_flag |= lowerrootvp->v_mount->mnt_kern_flag & MNTK_MPSAFE;
#endif
MNT_IUNLOCK(mp);
mp->mnt_data = (qaddr_t) xmp;
vfs_getnewfsid(mp);
vfs_mountedfrom(mp, target);
VMBLOCKDEBUG("VMBlockVFSMount: lower %s, alias at %s\n",
mp->mnt_stat.f_mntfromname, mp->mnt_stat.f_mntonname);
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 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);
}
/* ARGSUSED */
int
compat_43_sys_lstat(struct lwp *l, const struct compat_43_sys_lstat_args *uap, register_t *retval)
{
/* {
syscallarg(char *) path;
syscallarg(struct ostat *) ub;
} */
struct vnode *vp, *dvp;
struct stat sb, sb1;
struct stat43 osb;
int error;
struct pathbuf *pb;
struct nameidata nd;
int ndflags;
error = pathbuf_copyin(SCARG(uap, path), &pb);
if (error) {
return error;
}
ndflags = NOFOLLOW | LOCKLEAF | LOCKPARENT | TRYEMULROOT;
again:
NDINIT(&nd, LOOKUP, ndflags, pb);
if ((error = namei(&nd))) {
if (error == EISDIR && (ndflags & LOCKPARENT) != 0) {
/*
* Should only happen on '/'. Retry without LOCKPARENT;
* this is safe since the vnode won't be a VLNK.
*/
ndflags &= ~LOCKPARENT;
goto again;
}
pathbuf_destroy(pb);
return (error);
}
/*
* For symbolic links, always return the attributes of its
* containing directory, except for mode, size, and links.
*/
vp = nd.ni_vp;
dvp = nd.ni_dvp;
pathbuf_destroy(pb);
if (vp->v_type != VLNK) {
if ((ndflags & LOCKPARENT) != 0) {
if (dvp == vp)
vrele(dvp);
else
vput(dvp);
}
error = vn_stat(vp, &sb);
vput(vp);
if (error)
return (error);
} else {
error = vn_stat(dvp, &sb);
vput(dvp);
if (error) {
vput(vp);
return (error);
}
error = vn_stat(vp, &sb1);
vput(vp);
if (error)
return (error);
sb.st_mode &= ~S_IFDIR;
sb.st_mode |= S_IFLNK;
sb.st_nlink = sb1.st_nlink;
sb.st_size = sb1.st_size;
sb.st_blocks = sb1.st_blocks;
}
cvtstat(&sb, &osb);
error = copyout((void *)&osb, (void *)SCARG(uap, ub), sizeof (osb));
return (error);
}
int
nnpfs_fhlookup (d_thread_t *proc,
struct nnpfs_fhandle_t *fhp,
struct vnode **vpp)
{
int error;
struct mount *mp;
#if !(defined(HAVE_GETFH) && defined(HAVE_FHOPEN))
struct ucred *cred = proc->p_ucred;
struct vattr vattr;
fsid_t fsid;
struct nnpfs_fh_args *fh_args = (struct nnpfs_fh_args *)fhp->fhdata;
NNPFSDEB(XDEBVFOPS, ("nnpfs_fhlookup (nnpfs)\n"));
error = nnpfs_suser (proc);
if (error)
return EPERM;
if (fhp->len < sizeof(struct nnpfs_fh_args))
return EINVAL;
fsid = SCARG(fh_args, fsid);
mp = nnpfs_vfs_getvfs (&fsid);
if (mp == NULL)
return ENXIO;
#ifdef __APPLE__
{
uint32_t ino = SCARG(fh_args, fileid);
error = VFS_VGET(mp, &ino, vpp);
}
#else
error = VFS_VGET(mp, SCARG(fh_args, fileid), vpp);
#endif
if (error)
return error;
if (*vpp == NULL)
return ENOENT;
error = VOP_GETATTR(*vpp, &vattr, cred, proc);
if (error) {
vput(*vpp);
return error;
}
if (vattr.va_gen != SCARG(fh_args, gen)) {
vput(*vpp);
return ENOENT;
}
#else /* HAVE_GETFH && HAVE_FHOPEN */
{
fhandle_t *fh = (fhandle_t *) fhp;
NNPFSDEB(XDEBVFOPS, ("nnpfs_fhlookup (native)\n"));
mp = nnpfs_vfs_getvfs (&fh->fh_fsid);
if (mp == NULL)
return ESTALE;
if ((error = VFS_FHTOVP(mp, &fh->fh_fid, vpp)) != 0) {
*vpp = NULL;
return error;
}
}
#endif /* HAVE_GETFH && HAVE_FHOPEN */
#ifdef HAVE_KERNEL_VFS_OBJECT_CREATE
if ((*vpp)->v_type == VREG && (*vpp)->v_object == NULL)
#ifdef HAVE_FREEBSD_THREAD
nnpfs_vfs_object_create (*vpp, proc, proc->td_proc->p_ucred);
#else
nnpfs_vfs_object_create (*vpp, proc, proc->p_ucred);
#endif
#elif __APPLE__
if ((*vpp)->v_type == VREG && (!UBCINFOEXISTS(*vpp))) {
ubc_info_init(*vpp);
}
ubc_hold(*vpp);
#endif
return 0;
}
/*
* Obtain a dquot structure for the specified identifier and quota file
* reading the information from the file if necessary.
*/
static int
dqget(struct vnode *vp, u_long id, struct ufsmount *ump, int type,
struct dquot **dqp)
{
uint8_t buf[sizeof(struct dqblk64)];
off_t base, recsize;
struct dquot *dq, *dq1;
struct dqhash *dqh;
struct vnode *dqvp;
struct iovec aiov;
struct uio auio;
int dqvplocked, error;
#ifdef DEBUG_VFS_LOCKS
if (vp != NULLVP)
ASSERT_VOP_ELOCKED(vp, "dqget");
#endif
if (vp != NULLVP && *dqp != NODQUOT) {
return (0);
}
/* XXX: Disallow negative id values to prevent the
* creation of 100GB+ quota data files.
*/
if ((int)id < 0)
return (EINVAL);
UFS_LOCK(ump);
dqvp = ump->um_quotas[type];
if (dqvp == NULLVP || (ump->um_qflags[type] & QTF_CLOSING)) {
*dqp = NODQUOT;
UFS_UNLOCK(ump);
return (EINVAL);
}
vref(dqvp);
UFS_UNLOCK(ump);
error = 0;
dqvplocked = 0;
/*
* Check the cache first.
*/
dqh = DQHASH(dqvp, id);
DQH_LOCK();
dq = dqhashfind(dqh, id, dqvp);
if (dq != NULL) {
DQH_UNLOCK();
hfound: DQI_LOCK(dq);
DQI_WAIT(dq, PINOD+1, "dqget");
DQI_UNLOCK(dq);
if (dq->dq_ump == NULL) {
dqrele(vp, dq);
dq = NODQUOT;
error = EIO;
}
*dqp = dq;
if (dqvplocked)
vput(dqvp);
else
vrele(dqvp);
return (error);
}
/*
* Quota vnode lock is before DQ_LOCK. Acquire dqvp lock there
* since new dq will appear on the hash chain DQ_LOCKed.
*/
if (vp != dqvp) {
DQH_UNLOCK();
vn_lock(dqvp, LK_SHARED | LK_RETRY);
dqvplocked = 1;
DQH_LOCK();
/*
* Recheck the cache after sleep for quota vnode lock.
*/
dq = dqhashfind(dqh, id, dqvp);
if (dq != NULL) {
DQH_UNLOCK();
goto hfound;
}
}
/*
* Not in cache, allocate a new one or take it from the
* free list.
*/
if (TAILQ_FIRST(&dqfreelist) == NODQUOT &&
numdquot < MAXQUOTAS * desiredvnodes)
desireddquot += DQUOTINC;
if (numdquot < desireddquot) {
numdquot++;
DQH_UNLOCK();
dq1 = malloc(sizeof *dq1, M_DQUOT, M_WAITOK | M_ZERO);
mtx_init(&dq1->dq_lock, "dqlock", NULL, MTX_DEF);
DQH_LOCK();
/*
* Recheck the cache after sleep for memory.
*/
dq = dqhashfind(dqh, id, dqvp);
if (dq != NULL) {
numdquot--;
DQH_UNLOCK();
mtx_destroy(&dq1->dq_lock);
free(dq1, M_DQUOT);
goto hfound;
}
dq = dq1;
} else {
if ((dq = TAILQ_FIRST(&dqfreelist)) == NULL) {
DQH_UNLOCK();
tablefull("dquot");
*dqp = NODQUOT;
if (dqvplocked)
vput(dqvp);
else
vrele(dqvp);
return (EUSERS);
}
if (dq->dq_cnt || (dq->dq_flags & DQ_MOD))
panic("dqget: free dquot isn't %p", dq);
TAILQ_REMOVE(&dqfreelist, dq, dq_freelist);
if (dq->dq_ump != NULL)
LIST_REMOVE(dq, dq_hash);
}
/*
* Dq is put into hash already locked to prevent parallel
* usage while it is being read from file.
*/
dq->dq_flags = DQ_LOCK;
dq->dq_id = id;
dq->dq_type = type;
dq->dq_ump = ump;
LIST_INSERT_HEAD(dqh, dq, dq_hash);
DQREF(dq);
DQH_UNLOCK();
/*
* Read the requested quota record from the quota file, performing
* any necessary conversions.
*/
if (ump->um_qflags[type] & QTF_64BIT) {
recsize = sizeof(struct dqblk64);
base = sizeof(struct dqhdr64);
} else {
recsize = sizeof(struct dqblk32);
base = 0;
}
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
aiov.iov_base = buf;
aiov.iov_len = recsize;
auio.uio_resid = recsize;
auio.uio_offset = base + id * recsize;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = UIO_READ;
auio.uio_td = (struct thread *)0;
error = VOP_READ(dqvp, &auio, 0, ump->um_cred[type]);
if (auio.uio_resid == recsize && error == 0) {
bzero(&dq->dq_dqb, sizeof(dq->dq_dqb));
} else {
if (ump->um_qflags[type] & QTF_64BIT)
dqb64_dq((struct dqblk64 *)buf, dq);
else
dqb32_dq((struct dqblk32 *)buf, dq);
}
if (dqvplocked)
vput(dqvp);
else
vrele(dqvp);
/*
* I/O error in reading quota file, release
* quota structure and reflect problem to caller.
*/
if (error) {
DQH_LOCK();
dq->dq_ump = NULL;
LIST_REMOVE(dq, dq_hash);
DQH_UNLOCK();
DQI_LOCK(dq);
if (dq->dq_flags & DQ_WANT)
wakeup(dq);
dq->dq_flags = 0;
DQI_UNLOCK(dq);
dqrele(vp, dq);
*dqp = NODQUOT;
return (error);
}
DQI_LOCK(dq);
/*
* Check for no limit to enforce.
* Initialize time values if necessary.
*/
if (dq->dq_isoftlimit == 0 && dq->dq_bsoftlimit == 0 &&
dq->dq_ihardlimit == 0 && dq->dq_bhardlimit == 0)
dq->dq_flags |= DQ_FAKE;
if (dq->dq_id != 0) {
if (dq->dq_btime == 0) {
dq->dq_btime = time_second + ump->um_btime[type];
if (dq->dq_bsoftlimit &&
dq->dq_curblocks >= dq->dq_bsoftlimit)
dq->dq_flags |= DQ_MOD;
}
if (dq->dq_itime == 0) {
dq->dq_itime = time_second + ump->um_itime[type];
if (dq->dq_isoftlimit &&
dq->dq_curinodes >= dq->dq_isoftlimit)
dq->dq_flags |= DQ_MOD;
}
}
DQI_WAKEUP(dq);
DQI_UNLOCK(dq);
*dqp = dq;
return (0);
}
/*
* Given a locked directory vnode, iterate over the names in the directory
* and use ufs_extattr_lookup() to retrieve locked vnodes of potential
* attribute files. Then invoke ufs_extattr_enable_with_open() on each
* to attempt to start the attribute. Leaves the directory locked on
* exit.
*/
static int
ufs_extattr_iterate_directory(struct ufsmount *ump, struct vnode *dvp,
int attrnamespace, struct thread *td)
{
struct vop_readdir_args vargs;
struct dirent *dp, *edp;
struct vnode *attr_vp;
struct uio auio;
struct iovec aiov;
char *dirbuf;
int error, eofflag = 0;
if (dvp->v_type != VDIR)
return (ENOTDIR);
dirbuf = malloc(DIRBLKSIZ, M_TEMP, M_WAITOK);
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_rw = UIO_READ;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_td = td;
auio.uio_offset = 0;
vargs.a_gen.a_desc = NULL;
vargs.a_vp = dvp;
vargs.a_uio = &auio;
vargs.a_cred = td->td_ucred;
vargs.a_eofflag = &eofflag;
vargs.a_ncookies = NULL;
vargs.a_cookies = NULL;
while (!eofflag) {
auio.uio_resid = DIRBLKSIZ;
aiov.iov_base = dirbuf;
aiov.iov_len = DIRBLKSIZ;
error = ufs_readdir(&vargs);
if (error) {
printf("ufs_extattr_iterate_directory: ufs_readdir "
"%d\n", error);
return (error);
}
/*
* XXXRW: While in UFS, we always get DIRBLKSIZ returns from
* the directory code on success, on other file systems this
* may not be the case. For portability, we should check the
* read length on return from ufs_readdir().
*/
edp = (struct dirent *)&dirbuf[DIRBLKSIZ];
for (dp = (struct dirent *)dirbuf; dp < edp; ) {
#if (BYTE_ORDER == LITTLE_ENDIAN)
dp->d_type = dp->d_namlen;
dp->d_namlen = 0;
#else
dp->d_type = 0;
#endif
if (dp->d_reclen == 0)
break;
error = ufs_extattr_lookup(dvp, UE_GETDIR_LOCKPARENT,
dp->d_name, &attr_vp, td);
if (error) {
printf("ufs_extattr_iterate_directory: lookup "
"%s %d\n", dp->d_name, error);
} else if (attr_vp == dvp) {
vrele(attr_vp);
} else if (attr_vp->v_type != VREG) {
vput(attr_vp);
} else {
error = ufs_extattr_enable_with_open(ump,
attr_vp, attrnamespace, dp->d_name, td);
vrele(attr_vp);
if (error) {
printf("ufs_extattr_iterate_directory: "
"enable %s %d\n", dp->d_name,
error);
} else if (bootverbose) {
printf("UFS autostarted EA %s\n",
dp->d_name);
}
}
dp = (struct dirent *) ((char *)dp + dp->d_reclen);
if (dp >= edp)
break;
}
}
free(dirbuf, M_TEMP);
return (0);
}
/*
* Update the disk quota in the quota file.
*/
static int
dqsync(struct vnode *vp, struct dquot *dq)
{
uint8_t buf[sizeof(struct dqblk64)];
off_t base, recsize;
struct vnode *dqvp;
struct iovec aiov;
struct uio auio;
int error;
struct mount *mp;
struct ufsmount *ump;
#ifdef DEBUG_VFS_LOCKS
if (vp != NULL)
ASSERT_VOP_ELOCKED(vp, "dqsync");
#endif
mp = NULL;
error = 0;
if (dq == NODQUOT)
panic("dqsync: dquot");
if ((ump = dq->dq_ump) == NULL)
return (0);
UFS_LOCK(ump);
if ((dqvp = ump->um_quotas[dq->dq_type]) == NULLVP)
panic("dqsync: file");
vref(dqvp);
UFS_UNLOCK(ump);
DQI_LOCK(dq);
if ((dq->dq_flags & DQ_MOD) == 0) {
DQI_UNLOCK(dq);
vrele(dqvp);
return (0);
}
DQI_UNLOCK(dq);
(void) vn_start_secondary_write(dqvp, &mp, V_WAIT);
if (vp != dqvp)
vn_lock(dqvp, LK_EXCLUSIVE | LK_RETRY);
DQI_LOCK(dq);
DQI_WAIT(dq, PINOD+2, "dqsync");
if ((dq->dq_flags & DQ_MOD) == 0)
goto out;
dq->dq_flags |= DQ_LOCK;
DQI_UNLOCK(dq);
/*
* Write the quota record to the quota file, performing any
* necessary conversions. See dqget() for additional details.
*/
if (ump->um_qflags[dq->dq_type] & QTF_64BIT) {
dq_dqb64(dq, (struct dqblk64 *)buf);
recsize = sizeof(struct dqblk64);
base = sizeof(struct dqhdr64);
} else {
dq_dqb32(dq, (struct dqblk32 *)buf);
recsize = sizeof(struct dqblk32);
base = 0;
}
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
aiov.iov_base = buf;
aiov.iov_len = recsize;
auio.uio_resid = recsize;
auio.uio_offset = base + dq->dq_id * recsize;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = UIO_WRITE;
auio.uio_td = (struct thread *)0;
error = VOP_WRITE(dqvp, &auio, 0, dq->dq_ump->um_cred[dq->dq_type]);
if (auio.uio_resid && error == 0)
error = EIO;
DQI_LOCK(dq);
DQI_WAKEUP(dq);
dq->dq_flags &= ~DQ_MOD;
out:
DQI_UNLOCK(dq);
if (vp != dqvp)
vput(dqvp);
else
vrele(dqvp);
vn_finished_secondary_write(mp);
return (error);
}
static int
smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred)
{
struct dirent de;
struct componentname cn;
struct smb_cred scred;
struct smbfs_fctx *ctx;
struct vnode *newvp;
struct smbnode *np = VTOSMB(vp);
int error/*, *eofflag = ap->a_eofflag*/;
long offset, limit;
np = VTOSMB(vp);
SMBVDEBUG("dirname='%s'\n", np->n_name);
smb_makescred(&scred, uio->uio_td, cred);
offset = uio->uio_offset / DE_SIZE; /* offset in the directory */
limit = uio->uio_resid / DE_SIZE;
if (uio->uio_resid < DE_SIZE || uio->uio_offset < 0)
return EINVAL;
while (limit && offset < 2) {
limit--;
bzero((caddr_t)&de, DE_SIZE);
de.d_reclen = DE_SIZE;
de.d_fileno = (offset == 0) ? np->n_ino :
(np->n_parent ? VTOSMB(np->n_parent)->n_ino : 2);
if (de.d_fileno == 0)
de.d_fileno = 0x7ffffffd + offset;
de.d_namlen = offset + 1;
de.d_name[0] = '.';
de.d_name[1] = '.';
de.d_name[offset + 1] = '\0';
de.d_type = DT_DIR;
error = uiomove(&de, DE_SIZE, uio);
if (error)
return error;
offset++;
uio->uio_offset += DE_SIZE;
}
if (limit == 0)
return 0;
if (offset != np->n_dirofs || np->n_dirseq == NULL) {
SMBVDEBUG("Reopening search %ld:%ld\n", offset, np->n_dirofs);
if (np->n_dirseq) {
smbfs_findclose(np->n_dirseq, &scred);
np->n_dirseq = NULL;
}
np->n_dirofs = 2;
error = smbfs_findopen(np, "*", 1,
SMB_FA_SYSTEM | SMB_FA_HIDDEN | SMB_FA_DIR,
&scred, &ctx);
if (error) {
SMBVDEBUG("can not open search, error = %d", error);
return error;
}
np->n_dirseq = ctx;
} else
ctx = np->n_dirseq;
while (np->n_dirofs < offset) {
error = smbfs_findnext(ctx, offset - np->n_dirofs++, &scred);
if (error) {
smbfs_findclose(np->n_dirseq, &scred);
np->n_dirseq = NULL;
return error == ENOENT ? 0 : error;
}
}
error = 0;
for (; limit; limit--, offset++) {
error = smbfs_findnext(ctx, limit, &scred);
if (error)
break;
np->n_dirofs++;
bzero((caddr_t)&de, DE_SIZE);
de.d_reclen = DE_SIZE;
de.d_fileno = ctx->f_attr.fa_ino;
de.d_type = (ctx->f_attr.fa_attr & SMB_FA_DIR) ? DT_DIR : DT_REG;
de.d_namlen = ctx->f_nmlen;
bcopy(ctx->f_name, de.d_name, de.d_namlen);
de.d_name[de.d_namlen] = '\0';
if (smbfs_fastlookup) {
error = smbfs_nget(vp->v_mount, vp, ctx->f_name,
ctx->f_nmlen, &ctx->f_attr, &newvp);
if (!error) {
cn.cn_nameptr = de.d_name;
cn.cn_namelen = de.d_namlen;
cache_enter(vp, newvp, &cn);
vput(newvp);
}
}
error = uiomove(&de, DE_SIZE, uio);
if (error)
break;
}
if (error == ENOENT)
error = 0;
uio->uio_offset = offset * DE_SIZE;
return error;
}
/*
* VFS Operations.
*
* mount system call
*/
static int
ext2_mount(struct mount *mp)
{
struct vfsoptlist *opts;
struct vnode *devvp;
struct thread *td;
struct ext2mount *ump = 0;
struct m_ext2fs *fs;
struct nameidata nd, *ndp = &nd;
accmode_t accmode;
char *path, *fspec;
int error, flags, len;
td = curthread;
opts = mp->mnt_optnew;
if (vfs_filteropt(opts, ext2_opts))
return (EINVAL);
vfs_getopt(opts, "fspath", (void **)&path, NULL);
/* Double-check the length of path.. */
if (strlen(path) >= MAXMNTLEN - 1)
return (ENAMETOOLONG);
fspec = NULL;
error = vfs_getopt(opts, "from", (void **)&fspec, &len);
if (!error && fspec[len - 1] != '\0')
return (EINVAL);
/*
* 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 = VFSTOEXT2(mp);
fs = ump->um_e2fs;
error = 0;
if (fs->e2fs_ronly == 0 &&
vfs_flagopt(opts, "ro", NULL, 0)) {
error = VFS_SYNC(mp, MNT_WAIT);
if (error)
return (error);
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = ext2_flushfiles(mp, flags, td);
if ( error == 0 && fs->e2fs_wasvalid && ext2_cgupdate(ump, MNT_WAIT) == 0) {
fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
ext2_sbupdate(ump, MNT_WAIT);
}
fs->e2fs_ronly = 1;
vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY);
DROP_GIANT();
g_topology_lock();
g_access(ump->um_cp, 0, -1, 0);
g_topology_unlock();
PICKUP_GIANT();
}
if (!error && (mp->mnt_flag & MNT_RELOAD))
error = ext2_reload(mp, td);
if (error)
return (error);
devvp = ump->um_devvp;
if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) {
if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
return (EPERM);
/*
* If upgrade to read-write by non-root, then verify
* that user has necessary permissions on the device.
*/
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_ACCESS(devvp, VREAD | VWRITE,
td->td_ucred, td);
if (error)
error = priv_check(td, PRIV_VFS_MOUNT_PERM);
if (error) {
VOP_UNLOCK(devvp, 0);
return (error);
}
VOP_UNLOCK(devvp, 0);
DROP_GIANT();
g_topology_lock();
error = g_access(ump->um_cp, 0, 1, 0);
g_topology_unlock();
PICKUP_GIANT();
if (error)
return (error);
if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 ||
(fs->e2fs->e2fs_state & E2FS_ERRORS)) {
if (mp->mnt_flag & MNT_FORCE) {
printf(
"WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
} else {
printf(
"WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
fs->e2fs_fsmnt);
return (EPERM);
}
}
fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;
(void)ext2_cgupdate(ump, MNT_WAIT);
fs->e2fs_ronly = 0;
MNT_ILOCK(mp);
mp->mnt_flag &= ~MNT_RDONLY;
MNT_IUNLOCK(mp);
}
if (vfs_flagopt(opts, "export", NULL, 0)) {
/* Process export requests in vfs_mount.c. */
return (error);
}
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible disk device.
*/
if (fspec == NULL)
return (EINVAL);
NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
if ((error = namei(ndp)) != 0)
return (error);
NDFREE(ndp, NDF_ONLY_PNBUF);
devvp = ndp->ni_vp;
if (!vn_isdisk(devvp, &error)) {
vput(devvp);
return (error);
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*
* XXXRW: VOP_ACCESS() enough?
*/
accmode = VREAD;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
accmode |= VWRITE;
error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
if (error)
error = priv_check(td, PRIV_VFS_MOUNT_PERM);
if (error) {
vput(devvp);
return (error);
}
if ((mp->mnt_flag & MNT_UPDATE) == 0) {
error = ext2_mountfs(devvp, mp);
} else {
if (devvp != ump->um_devvp) {
vput(devvp);
return (EINVAL); /* needs translation */
} else
vput(devvp);
}
if (error) {
vrele(devvp);
return (error);
}
ump = VFSTOEXT2(mp);
fs = ump->um_e2fs;
/*
* Note that this strncpy() is ok because of a check at the start
* of ext2_mount().
*/
strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN);
fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0';
vfs_mountedfrom(mp, fspec);
return (0);
}
/* When successful this function returns data in the following "out"-arguments:
* o res_vnode: the vnode of the parent directory of "name"
* o name: the `basename' (the element of the pathname)
* o namelen: the length of the basename
*
* For example: dir_namev("/s5fs/bin/ls", &namelen, &name, NULL,
* &res_vnode) would put 2 in namelen, "ls" in name, and a pointer to the
* vnode corresponding to "/s5fs/bin" in res_vnode.
*
* The "base" argument defines where we start resolving the path from:
* A base value of NULL means to use the process's current working directory,
* curproc->p_cwd. If pathname[0] == '/', ignore base and start with
* vfs_root_vn. dir_namev() should call lookup() to take care of resolving each
* piece of the pathname.
*
* Note: A successful call to this causes vnode refcount on *res_vnode to
* be incremented.
*/
int
dir_namev(const char *pathname, size_t *namelen, const char **name,
vnode_t *base, vnode_t **res_vnode)
{
/* VFS {{{ */
char *saveptr;
int err;
char *dirtok;
int dirlen;
char *nextok;
int nextlen;
vnode_t *curdir;
vnode_t *nextdir;
KASSERT(NULL != pathname);
KASSERT(NULL != namelen);
KASSERT(NULL != name);
KASSERT(NULL != res_vnode);
if ('\0' == pathname[0]) {
return -EINVAL;
}
/* Choose base directory */
if (pathname[0] == '/') {
curdir = vfs_root_vn;
} else if (base == NULL) {
curdir = curproc->p_cwd;
} else {
curdir = base;
}
KASSERT(NULL != curdir);
vref(curdir);
/* Follow the path */
saveptr = (char *) pathname;
dirtok = namev_tokenize(&saveptr, &dirlen);
nextok = namev_tokenize(&saveptr, &nextlen);
while (nextlen != 0) {
err = lookup(curdir, dirtok, dirlen, &nextdir);
vput(curdir);
if (0 != err) {
return err;
}
curdir = nextdir;
dirtok = nextok;
dirlen = nextlen;
nextok = namev_tokenize(&saveptr, &nextlen);
}
*res_vnode = curdir;
*name = dirtok;
*namelen = dirlen;
/* VFS }}} */
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;
/*
* If this malloc() is performed after the getnewvnode()
* it might block, leaving a vnode with a NULL v_data to be
* found by ext2_sync() if a sync happens to fire right then,
* which will cause a panic because ext2_sync() blindly
* dereferences vp->v_data (as well it should).
*/
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(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;
}
/*
ext2_print_inode(ip);
*/
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() / 2 + 1;
if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
ip->i_flag |= IN_MODIFIED;
}
*vpp = vp;
return (0);
}
/*
* Helper function for findroot():
* Return non-zero if disk device matches bootinfo.
*/
static int
match_bootdisk(struct device *dv, struct btinfo_bootdisk *bid)
{
struct vnode *tmpvn;
int error;
struct disklabel label;
int found = 0;
int bmajor;
/*
* A disklabel is required here. The boot loader doesn't refuse
* to boot from a disk without a label, but this is normally not
* wanted.
*/
if (bid->labelsector == -1)
return (0);
/*
* Lookup major number for disk block device.
*/
bmajor = devsw_name2blk(dv->dv_xname, NULL, 0);
if (bmajor == -1)
return (0); /* XXX panic ??? */
/*
* Fake a temporary vnode for the disk, open it, and read
* the disklabel for comparison.
*/
if (bdevvp(MAKEDISKDEV(bmajor, dv->dv_unit, RAW_PART), &tmpvn))
panic("match_bootdisk: can't alloc vnode");
error = VOP_OPEN(tmpvn, FREAD, NOCRED);
if (error) {
#ifndef DEBUG
/*
* Ignore errors caused by missing device, partition,
* or medium.
*/
if (error != ENXIO && error != ENODEV)
#endif
printf("match_bootdisk: can't open dev %s (%d)\n",
dv->dv_xname, error);
vput(tmpvn);
return (0);
}
error = VOP_IOCTL(tmpvn, DIOCGDINFO, &label, FREAD, NOCRED);
if (error) {
/*
* XXX Can't happen -- open() would have errored out
* or faked one up.
*/
printf("match_bootdisk: can't get label for dev %s (%d)\n",
dv->dv_xname, error);
goto closeout;
}
/* Compare with our data. */
if (label.d_type == bid->label.type &&
label.d_checksum == bid->label.checksum &&
strncmp(label.d_packname, bid->label.packname, 16) == 0)
found = 1;
closeout:
VOP_CLOSE(tmpvn, FREAD, NOCRED);
vput(tmpvn);
return (found);
}
/*
* Common code for vnode open operations. Check permissions, and call
* the VOP_NOPEN or VOP_NCREATE routine.
*
* The caller is responsible for setting up nd with nlookup_init() and
* for cleaning it up with nlookup_done(), whether we return an error
* or not.
*
* On success nd->nl_open_vp will hold a referenced and, if requested,
* locked vnode. A locked vnode is requested via NLC_LOCKVP. If fp
* is non-NULL the vnode will be installed in the file pointer.
*
* NOTE: If the caller wishes the namecache entry to be operated with
* a shared lock it must use NLC_SHAREDLOCK. If NLC_LOCKVP is set
* then the vnode lock will also be shared.
*
* NOTE: The vnode is referenced just once on return whether or not it
* is also installed in the file pointer.
*/
int
vn_open(struct nlookupdata *nd, struct file *fp, int fmode, int cmode)
{
struct vnode *vp;
struct ucred *cred = nd->nl_cred;
struct vattr vat;
struct vattr *vap = &vat;
int error;
u_int flags;
uint64_t osize;
struct mount *mp;
/*
* Certain combinations are illegal
*/
if ((fmode & (FWRITE | O_TRUNC)) == O_TRUNC)
return(EACCES);
/*
* Lookup the path and create or obtain the vnode. After a
* successful lookup a locked nd->nl_nch will be returned.
*
* The result of this section should be a locked vnode.
*
* XXX with only a little work we should be able to avoid locking
* the vnode if FWRITE, O_CREAT, and O_TRUNC are *not* set.
*/
nd->nl_flags |= NLC_OPEN;
if (fmode & O_APPEND)
nd->nl_flags |= NLC_APPEND;
if (fmode & O_TRUNC)
nd->nl_flags |= NLC_TRUNCATE;
if (fmode & FREAD)
nd->nl_flags |= NLC_READ;
if (fmode & FWRITE)
nd->nl_flags |= NLC_WRITE;
if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0)
nd->nl_flags |= NLC_FOLLOW;
if (fmode & O_CREAT) {
/*
* CONDITIONAL CREATE FILE CASE
*
* Setting NLC_CREATE causes a negative hit to store
* the negative hit ncp and not return an error. Then
* nc_error or nc_vp may be checked to see if the ncp
* represents a negative hit. NLC_CREATE also requires
* write permission on the governing directory or EPERM
* is returned.
*/
nd->nl_flags |= NLC_CREATE;
nd->nl_flags |= NLC_REFDVP;
bwillinode(1);
error = nlookup(nd);
} else {
/*
* NORMAL OPEN FILE CASE
*/
error = nlookup(nd);
}
if (error)
return (error);
/*
* split case to allow us to re-resolve and retry the ncp in case
* we get ESTALE.
*/
again:
if (fmode & O_CREAT) {
if (nd->nl_nch.ncp->nc_vp == NULL) {
if ((error = ncp_writechk(&nd->nl_nch)) != 0)
return (error);
VATTR_NULL(vap);
vap->va_type = VREG;
vap->va_mode = cmode;
if (fmode & O_EXCL)
vap->va_vaflags |= VA_EXCLUSIVE;
error = VOP_NCREATE(&nd->nl_nch, nd->nl_dvp, &vp,
nd->nl_cred, vap);
if (error)
return (error);
fmode &= ~O_TRUNC;
/* locked vnode is returned */
} else {
if (fmode & O_EXCL) {
error = EEXIST;
} else {
error = cache_vget(&nd->nl_nch, cred,
LK_EXCLUSIVE, &vp);
}
if (error)
return (error);
fmode &= ~O_CREAT;
}
} else {
if (nd->nl_flags & NLC_SHAREDLOCK) {
error = cache_vget(&nd->nl_nch, cred, LK_SHARED, &vp);
} else {
error = cache_vget(&nd->nl_nch, cred,
LK_EXCLUSIVE, &vp);
}
if (error)
return (error);
}
/*
* We have a locked vnode and ncp now. Note that the ncp will
* be cleaned up by the caller if nd->nl_nch is left intact.
*/
if (vp->v_type == VLNK) {
error = EMLINK;
goto bad;
}
if (vp->v_type == VSOCK) {
error = EOPNOTSUPP;
goto bad;
}
if (vp->v_type != VDIR && (fmode & O_DIRECTORY)) {
error = ENOTDIR;
goto bad;
}
if ((fmode & O_CREAT) == 0) {
if (fmode & (FWRITE | O_TRUNC)) {
if (vp->v_type == VDIR) {
error = EISDIR;
goto bad;
}
error = vn_writechk(vp, &nd->nl_nch);
if (error) {
/*
* Special stale handling, re-resolve the
* vnode.
*/
if (error == ESTALE) {
vput(vp);
vp = NULL;
if (nd->nl_flags & NLC_SHAREDLOCK) {
cache_unlock(&nd->nl_nch);
cache_lock(&nd->nl_nch);
}
cache_setunresolved(&nd->nl_nch);
error = cache_resolve(&nd->nl_nch,
cred);
if (error == 0)
goto again;
}
goto bad;
}
}
}
if (fmode & O_TRUNC) {
vn_unlock(vp); /* XXX */
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */
osize = vp->v_filesize;
VATTR_NULL(vap);
vap->va_size = 0;
error = VOP_SETATTR(vp, vap, cred);
if (error)
goto bad;
error = VOP_GETATTR(vp, vap);
if (error)
goto bad;
mp = vq_vptomp(vp);
VFS_ACCOUNT(mp, vap->va_uid, vap->va_gid, -osize);
}
/*
* Set or clear VNSWAPCACHE on the vp based on nd->nl_nch.ncp->nc_flag.
* These particular bits a tracked all the way from the root.
*
* NOTE: Might not work properly on NFS servers due to the
* disconnected namecache.
*/
flags = nd->nl_nch.ncp->nc_flag;
if ((flags & (NCF_UF_CACHE | NCF_UF_PCACHE)) &&
(flags & (NCF_SF_NOCACHE | NCF_SF_PNOCACHE)) == 0) {
vsetflags(vp, VSWAPCACHE);
} else {
vclrflags(vp, VSWAPCACHE);
}
/*
* Setup the fp so VOP_OPEN can override it. No descriptor has been
* associated with the fp yet so we own it clean.
*
* f_nchandle inherits nl_nch. This used to be necessary only for
* directories but now we do it unconditionally so f*() ops
* such as fchmod() can access the actual namespace that was
* used to open the file.
*/
if (fp) {
if (nd->nl_flags & NLC_APPENDONLY)
fmode |= FAPPENDONLY;
fp->f_nchandle = nd->nl_nch;
cache_zero(&nd->nl_nch);
cache_unlock(&fp->f_nchandle);
}
/*
* Get rid of nl_nch. vn_open does not return it (it returns the
* vnode or the file pointer). Note: we can't leave nl_nch locked
* through the VOP_OPEN anyway since the VOP_OPEN may block, e.g.
* on /dev/ttyd0
*/
if (nd->nl_nch.ncp)
cache_put(&nd->nl_nch);
error = VOP_OPEN(vp, fmode, cred, fp);
if (error) {
/*
* setting f_ops to &badfileops will prevent the descriptor
* code from trying to close and release the vnode, since
* the open failed we do not want to call close.
*/
if (fp) {
fp->f_data = NULL;
fp->f_ops = &badfileops;
}
goto bad;
}
#if 0
/*
* Assert that VREG files have been setup for vmio.
*/
KASSERT(vp->v_type != VREG || vp->v_object != NULL,
("vn_open: regular file was not VMIO enabled!"));
#endif
/*
* Return the vnode. XXX needs some cleaning up. The vnode is
* only returned in the fp == NULL case.
*/
if (fp == NULL) {
nd->nl_open_vp = vp;
nd->nl_vp_fmode = fmode;
if ((nd->nl_flags & NLC_LOCKVP) == 0)
vn_unlock(vp);
} else {
vput(vp);
}
return (0);
bad:
if (vp)
vput(vp);
return (error);
}
/*
* Rename system call.
* rename("foo", "bar");
* is essentially
* unlink("bar");
* link("foo", "bar");
* unlink("foo");
* but ``atomically''. Can't do full commit without saving state in the
* inode on disk which isn't feasible at this time. Best we can do is
* always guarantee the target exists.
*
* Basic algorithm is:
*
* 1) Bump link count on source while we're linking it to the
* target. This also ensure the inode won't be deleted out
* from underneath us while we work (it may be truncated by
* a concurrent `trunc' or `open' for creation).
* 2) Link source to destination. If destination already exists,
* delete it first.
* 3) Unlink source reference to inode if still around. If a
* directory was moved and the parent of the destination
* is different from the source, patch the ".." entry in the
* directory.
*/
int
ext2fs_rename(void *v)
{
struct vop_rename_args *ap = v;
struct vnode *tvp = ap->a_tvp;
struct vnode *tdvp = ap->a_tdvp;
struct vnode *fvp = ap->a_fvp;
struct vnode *fdvp = ap->a_fdvp;
struct componentname *tcnp = ap->a_tcnp;
struct componentname *fcnp = ap->a_fcnp;
struct inode *ip, *xp, *dp;
struct proc *p = fcnp->cn_proc;
struct ext2fs_dirtemplate dirbuf;
/* struct timespec ts; */
int doingdirectory = 0, oldparent = 0, newparent = 0;
int error = 0;
u_char namlen;
#ifdef DIAGNOSTIC
if ((tcnp->cn_flags & HASBUF) == 0 ||
(fcnp->cn_flags & HASBUF) == 0)
panic("ext2fs_rename: no name");
#endif
/*
* Check for cross-device rename.
*/
if ((fvp->v_mount != tdvp->v_mount) ||
(tvp && (fvp->v_mount != tvp->v_mount))) {
error = EXDEV;
abortit:
VOP_ABORTOP(tdvp, tcnp); /* XXX, why not in NFS? */
if (tdvp == tvp)
vrele(tdvp);
else
vput(tdvp);
if (tvp)
vput(tvp);
VOP_ABORTOP(fdvp, fcnp); /* XXX, why not in NFS? */
vrele(fdvp);
vrele(fvp);
return (error);
}
/*
* Check if just deleting a link name.
*/
if (tvp && ((VTOI(tvp)->i_e2fs_flags & (EXT2_IMMUTABLE | EXT2_APPEND)) ||
(VTOI(tdvp)->i_e2fs_flags & EXT2_APPEND))) {
error = EPERM;
goto abortit;
}
if (fvp == tvp) {
if (fvp->v_type == VDIR) {
error = EINVAL;
goto abortit;
}
/* Release destination completely. */
VOP_ABORTOP(tdvp, tcnp);
vput(tdvp);
vput(tvp);
/* Delete source. */
vrele(fdvp);
vrele(fvp);
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
if ((fcnp->cn_flags & SAVESTART) == 0)
panic("ext2fs_rename: lost from startdir");
fcnp->cn_nameiop = DELETE;
(void) vfs_relookup(fdvp, &fvp, fcnp);
return (VOP_REMOVE(fdvp, fvp, fcnp));
}
if ((error = vn_lock(fvp, LK_EXCLUSIVE, p)) != 0)
goto abortit;
dp = VTOI(fdvp);
ip = VTOI(fvp);
if ((nlink_t)ip->i_e2fs_nlink >= LINK_MAX) {
VOP_UNLOCK(fvp, 0);
error = EMLINK;
goto abortit;
}
if ((ip->i_e2fs_flags & (EXT2_IMMUTABLE | EXT2_APPEND)) ||
(dp->i_e2fs_flags & EXT2_APPEND)) {
VOP_UNLOCK(fvp, 0);
error = EPERM;
goto abortit;
}
if ((ip->i_e2fs_mode & IFMT) == IFDIR) {
error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
if (!error && tvp)
error = VOP_ACCESS(tvp, VWRITE, tcnp->cn_cred);
if (error) {
VOP_UNLOCK(fvp, 0);
error = EACCES;
goto abortit;
}
/*
* Avoid ".", "..", and aliases of "." for obvious reasons.
*/
if ((fcnp->cn_namelen == 1 && fcnp->cn_nameptr[0] == '.') ||
dp == ip ||
(fcnp->cn_flags&ISDOTDOT) ||
(tcnp->cn_flags & ISDOTDOT) ||
(ip->i_flag & IN_RENAME)) {
VOP_UNLOCK(fvp, 0);
error = EINVAL;
goto abortit;
}
ip->i_flag |= IN_RENAME;
oldparent = dp->i_number;
doingdirectory++;
}
vrele(fdvp);
/*
* When the target exists, both the directory
* and target vnodes are returned locked.
*/
dp = VTOI(tdvp);
xp = NULL;
if (tvp)
xp = VTOI(tvp);
/*
* 1) Bump link count while we're moving stuff
* around. If we crash somewhere before
* completing our work, the link count
* may be wrong, but correctable.
*/
ip->i_e2fs_nlink++;
ip->i_flag |= IN_CHANGE;
if ((error = ext2fs_update(ip, NULL, NULL, 1)) != 0) {
VOP_UNLOCK(fvp, 0);
goto bad;
}
/*
* If ".." must be changed (ie the directory gets a new
* parent) then the source directory must not be in the
* directory hierarchy above the target, as this would
* orphan everything below the source directory. Also
* the user must have write permission in the source so
* as to be able to change "..". We must repeat the call
* to namei, as the parent directory is unlocked by the
* call to checkpath().
*/
error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred);
VOP_UNLOCK(fvp, 0);
if (oldparent != dp->i_number)
newparent = dp->i_number;
if (doingdirectory && newparent) {
if (error) /* write access check above */
goto bad;
if (xp != NULL)
vput(tvp);
error = ext2fs_checkpath(ip, dp, tcnp->cn_cred);
if (error != 0)
goto out;
if ((tcnp->cn_flags & SAVESTART) == 0)
panic("ext2fs_rename: lost to startdir");
if ((error = vfs_relookup(tdvp, &tvp, tcnp)) != 0)
goto out;
dp = VTOI(tdvp);
xp = NULL;
if (tvp)
xp = VTOI(tvp);
}
/*
* 2) If target doesn't exist, link the target
* to the source and unlink the source.
* Otherwise, rewrite the target directory
* entry to reference the source inode and
* expunge the original entry's existence.
*/
if (xp == NULL) {
if (dp->i_dev != ip->i_dev)
panic("rename: EXDEV");
/*
* Account for ".." in new directory.
* When source and destination have the same
* parent we don't fool with the link count.
*/
if (doingdirectory && newparent) {
if ((nlink_t)dp->i_e2fs_nlink >= LINK_MAX) {
error = EMLINK;
goto bad;
}
dp->i_e2fs_nlink++;
dp->i_flag |= IN_CHANGE;
if ((error = ext2fs_update(dp, NULL, NULL, 1)) != 0)
goto bad;
}
error = ext2fs_direnter(ip, tdvp, tcnp);
if (error != 0) {
if (doingdirectory && newparent) {
dp->i_e2fs_nlink--;
dp->i_flag |= IN_CHANGE;
(void)ext2fs_update(dp, NULL, NULL, 1);
}
goto bad;
}
vput(tdvp);
} else {
if (xp->i_dev != dp->i_dev || xp->i_dev != ip->i_dev)
panic("rename: EXDEV");
/*
* Short circuit rename(foo, foo).
*/
if (xp->i_number == ip->i_number)
panic("rename: same file");
/*
* If the parent directory is "sticky", then the user must
* own the parent directory, or the destination of the rename,
* otherwise the destination may not be changed (except by
* root). This implements append-only directories.
*/
if ((dp->i_e2fs_mode & S_ISTXT) && tcnp->cn_cred->cr_uid != 0 &&
tcnp->cn_cred->cr_uid != dp->i_e2fs_uid &&
xp->i_e2fs_uid != tcnp->cn_cred->cr_uid) {
error = EPERM;
goto bad;
}
/*
* Target must be empty if a directory and have no links
* to it. Also, ensure source and target are compatible
* (both directories, or both not directories).
*/
if ((xp->i_e2fs_mode & IFMT) == IFDIR) {
if (!ext2fs_dirempty(xp, dp->i_number, tcnp->cn_cred) ||
xp->i_e2fs_nlink > 2) {
error = ENOTEMPTY;
goto bad;
}
if (!doingdirectory) {
error = ENOTDIR;
goto bad;
}
cache_purge(tdvp);
} else if (doingdirectory) {
error = EISDIR;
goto bad;
}
error = ext2fs_dirrewrite(dp, ip, tcnp);
if (error != 0)
goto bad;
/*
* If the target directory is in the same
* directory as the source directory,
* decrement the link count on the parent
* of the target directory.
*/
if (doingdirectory && !newparent) {
dp->i_e2fs_nlink--;
dp->i_flag |= IN_CHANGE;
}
vput(tdvp);
/*
* Adjust the link count of the target to
* reflect the dirrewrite above. If this is
* a directory it is empty and there are
* no links to it, so we can squash the inode and
* any space associated with it. We disallowed
* renaming over top of a directory with links to
* it above, as the remaining link would point to
* a directory without "." or ".." entries.
*/
xp->i_e2fs_nlink--;
if (doingdirectory) {
if (--xp->i_e2fs_nlink != 0)
panic("rename: linked directory");
error = ext2fs_truncate(xp, (off_t)0, IO_SYNC,
tcnp->cn_cred);
}
xp->i_flag |= IN_CHANGE;
vput(tvp);
xp = NULL;
}
/*
* 3) Unlink the source.
*/
fcnp->cn_flags &= ~MODMASK;
fcnp->cn_flags |= LOCKPARENT | LOCKLEAF;
if ((fcnp->cn_flags & SAVESTART) == 0)
panic("ext2fs_rename: lost from startdir");
(void) vfs_relookup(fdvp, &fvp, fcnp);
if (fvp != NULL) {
xp = VTOI(fvp);
dp = VTOI(fdvp);
} else {
/*
* From name has disappeared.
*/
if (doingdirectory)
panic("ext2fs_rename: lost dir entry");
vrele(ap->a_fvp);
return (0);
}
/*
* Ensure that the directory entry still exists and has not
* changed while the new name has been entered. If the source is
* a file then the entry may have been unlinked or renamed. In
* either case there is no further work to be done. If the source
* is a directory then it cannot have been rmdir'ed; its link
* count of three would cause a rmdir to fail with ENOTEMPTY.
* The IRENAME flag ensures that it cannot be moved by another
* rename.
*/
if (xp != ip) {
if (doingdirectory)
panic("ext2fs_rename: lost dir entry");
} else {
/*
* If the source is a directory with a
* new parent, the link count of the old
* parent directory must be decremented
* and ".." set to point to the new parent.
*/
if (doingdirectory && newparent) {
dp->i_e2fs_nlink--;
dp->i_flag |= IN_CHANGE;
error = vn_rdwr(UIO_READ, fvp, (caddr_t)&dirbuf,
sizeof (struct ext2fs_dirtemplate), (off_t)0,
UIO_SYSSPACE, IO_NODELOCKED,
tcnp->cn_cred, NULL, curproc);
if (error == 0) {
namlen = dirbuf.dotdot_namlen;
if (namlen != 2 ||
dirbuf.dotdot_name[0] != '.' ||
dirbuf.dotdot_name[1] != '.') {
ufs_dirbad(xp, (doff_t)12,
"ext2fs_rename: mangled dir");
} else {
dirbuf.dotdot_ino = h2fs32(newparent);
(void) vn_rdwr(UIO_WRITE, fvp,
(caddr_t)&dirbuf,
sizeof (struct dirtemplate),
(off_t)0, UIO_SYSSPACE,
IO_NODELOCKED|IO_SYNC,
tcnp->cn_cred, NULL, curproc);
cache_purge(fdvp);
}
}
}
error = ext2fs_dirremove(fdvp, fcnp);
if (!error) {
xp->i_e2fs_nlink--;
xp->i_flag |= IN_CHANGE;
}
xp->i_flag &= ~IN_RENAME;
}
if (dp)
vput(fdvp);
if (xp)
vput(fvp);
vrele(ap->a_fvp);
return (error);
bad:
if (xp)
vput(ITOV(xp));
vput(ITOV(dp));
out:
if (doingdirectory)
ip->i_flag &= ~IN_RENAME;
if (vn_lock(fvp, LK_EXCLUSIVE, p) == 0) {
ip->i_e2fs_nlink--;
ip->i_flag |= IN_CHANGE;
vput(fvp);
} else
vrele(fvp);
return (error);
}
/*
* s5fs_rmdir:
* s5fs_rmdir removes the directory called name from dir. the directory
* to be removed must be empty (except for . and .. of course).
* param *parent: the pointer to the parent dir of the name specified
* param *name: name string
* param namelen: the length of the name string
* return: 0 on success; negative numbers on a variety of errors
*/
static int
s5fs_rmdir(vnode_t *parent, const char *name, size_t namelen)
{
dbg(DBG_S5FS, "{\n");
KASSERT(parent != NULL);
KASSERT(name != NULL);
KASSERT(namelen <= NAME_LEN - 1);
KASSERT((uint32_t)parent->vn_len == VNODE_TO_S5INODE(parent)->s5_size);
kmutex_lock(&parent->vn_mutex);
int inode_number = 0;
if ((inode_number = s5_find_dirent(parent, name, namelen)) < 0)
{
kmutex_unlock(&parent->vn_mutex);
/* Need vput? */
return inode_number;
}
/* May block here */
vnode_t* vn = vget(parent->vn_fs, inode_number);
KASSERT(vn != NULL);
if (!S_ISDIR(vn->vn_mode))
{
/* May block here */
vput(vn);
kmutex_unlock(&parent->vn_mutex);
return -ENOTDIR;
}
/* Check empty */
if (VNODE_TO_S5INODE(vn)->s5_size > sizeof(dirent_t)*2)
{
vput(vn);
kmutex_unlock(&parent->vn_mutex);
return -ENOTEMPTY;
}
int ret;
if ((ret = s5_remove_dirent(parent, name, namelen)) < 0)
{
/* May block here */
vput(vn);
kmutex_unlock(&parent->vn_mutex);
return ret;
}
/* Decrease the linkcount because .. is removed */
s5_inode_t* parent_inode = VNODE_TO_S5INODE(parent);
parent_inode->s5_linkcount--;
s5_dirty_inode(VNODE_TO_S5FS(parent), parent_inode);
/* May block here */
vput(vn);
kmutex_unlock(&parent->vn_mutex);
dbg(DBG_S5FS, "}\n");
return ret;
}