/*
* Callback supporting lookup in a GFS XATTR directory.
*/
static int
xattr_lookup_cb(vnode_t *vp, const char *nm, vnode_t **vpp, ino64_t *inop,
cred_t *cr, int flags, int *deflags, pathname_t *rpnp)
{
vnode_t *pvp;
struct pathname pn;
int error;
*vpp = NULL;
*inop = 0;
error = xattr_dir_realdir(vp, &pvp, LOOKUP_XATTR, cr, NULL);
/*
* Return ENOENT for EACCES requests during lookup. Once an
* attribute create is attempted EACCES will be returned.
*/
if (error) {
if (error == EACCES)
return (ENOENT);
return (error);
}
error = pn_get((char *)nm, UIO_SYSSPACE, &pn);
if (error == 0) {
error = VOP_LOOKUP(pvp, (char *)nm, vpp, &pn, flags, rootvp,
cr, NULL, deflags, rpnp);
pn_free(&pn);
}
return (error);
}
/*
* readdir_xattr_casecmp: given a system attribute name, see if there
* is a real xattr with the same normalized name.
*/
static int
readdir_xattr_casecmp(vnode_t *dvp, char *nm, cred_t *cr, caller_context_t *ct,
int *eflags)
{
int error;
vnode_t *vp;
struct pathname pn;
*eflags = 0;
error = pn_get(nm, UIO_SYSSPACE, &pn);
if (error == 0) {
error = VOP_LOOKUP(dvp, nm, &vp, &pn,
FIGNORECASE, rootvp, cr, ct, NULL, NULL);
if (error == 0) {
*eflags = ED_CASE_CONFLICT;
VN_RELE(vp);
} else if (error == ENOENT) {
error = 0;
}
pn_free(&pn);
}
return (error);
}
static int
is_nonempty_dir(char *name, char *pathleft, struct sdev_node *dir)
{
struct match_arg marg;
struct pathname pn;
struct vnode *gvp;
struct sdev_node *gdir = dir->sdev_origin;
if (VOP_LOOKUP(SDEVTOV(gdir), name, &gvp, NULL, 0, NULL, kcred,
NULL, NULL, NULL) != 0)
return (0);
if (gvp->v_type != VDIR) {
VN_RELE(gvp);
return (0);
}
if (pn_get(pathleft, UIO_SYSSPACE, &pn) != 0) {
VN_RELE(gvp);
return (0);
}
marg.expr = kmem_alloc(MAXNAMELEN, KM_SLEEP);
(void) pn_getcomponent(&pn, marg.expr);
marg.match = 0;
walk_dir(gvp, &marg, match_name);
VN_RELE(gvp);
kmem_free(marg.expr, MAXNAMELEN);
pn_free(&pn);
return (marg.match);
}
static int zfsfuse_lookup(fuse_req_t req, fuse_ino_t parent, const char *name)
{
if(strlen(name) >= MAXNAMELEN)
return ENAMETOOLONG;
vfs_t *vfs = (vfs_t *) fuse_req_userdata(req);
zfsvfs_t *zfsvfs = vfs->vfs_data;
ZFS_ENTER(zfsvfs);
znode_t *znode;
int error = zfs_zget(zfsvfs, parent, &znode, B_TRUE);
if(error) {
ZFS_EXIT(zfsvfs);
/* If the inode we are trying to get was recently deleted
dnode_hold_impl will return EEXIST instead of ENOENT */
return error == EEXIST ? ENOENT : error;
}
ASSERT(znode != NULL);
vnode_t *dvp = ZTOV(znode);
ASSERT(dvp != NULL);
vnode_t *vp = NULL;
cred_t cred;
zfsfuse_getcred(req, &cred);
error = VOP_LOOKUP(dvp, (char *) name, &vp, NULL, 0, NULL, &cred, NULL, NULL, NULL);
if(error)
goto out;
struct fuse_entry_param e = { 0 };
e.attr_timeout = 0.0;
e.entry_timeout = 0.0;
if(vp == NULL)
goto out;
e.ino = VTOZ(vp)->z_id;
if(e.ino == 3)
e.ino = 1;
e.generation = VTOZ(vp)->z_phys->zp_gen;
error = zfsfuse_stat(vp, &e.attr, &cred);
out:
if(vp != NULL)
VN_RELE(vp);
VN_RELE(dvp);
ZFS_EXIT(zfsvfs);
if(!error)
fuse_reply_entry(req, &e);
return error;
}
STATIC struct dentry *
linvfs_get_parent(
struct dentry *child)
{
int error;
vnode_t *vp, *cvp;
struct dentry *parent;
struct inode *ip = NULL;
struct dentry dotdot;
dotdot.d_name.name = "..";
dotdot.d_name.len = 2;
dotdot.d_inode = 0;
cvp = NULL;
vp = LINVFS_GET_VP(child->d_inode);
VOP_LOOKUP(vp, &dotdot, &cvp, 0, NULL, NULL, error);
if (!error) {
ASSERT(cvp);
ip = LINVFS_GET_IP(cvp);
if (!ip) {
VN_RELE(cvp);
return ERR_PTR(-EACCES);
}
}
if (error)
return ERR_PTR(-error);
parent = d_alloc_anon(ip);
if (!parent) {
VN_RELE(cvp);
parent = ERR_PTR(-ENOMEM);
}
return parent;
}
struct dentry *
vnlayer_get_parent(struct dentry *child)
{
VNODE_T *parentvp;
struct dentry *rdentp;
struct lookup_ctx ctx = {0};
CALL_DATA_T cd;
int err;
if (!MDKI_INOISMVFS(child->d_inode))
return ERR_PTR(-ESTALE);
mdki_linux_init_call_data(&cd);
err = VOP_LOOKUP(ITOV(child->d_inode), "..", &parentvp, NULL,
VNODE_LF_LOOKUP, NULL, &cd, &ctx);
mdki_linux_destroy_call_data(&cd);
if (err == 0) {
ASSERT(ctx.dentrypp == NULL);
ASSERT(parentvp != NULL);
if (!MDKI_INOISMVFS(VTOI(parentvp))) {
rdentp = ERR_PTR(-ESTALE);
} else {
rdentp = vnlayer_find_dentry(parentvp);
}
/* always drop vnode's refcount */
VN_RELE(parentvp);
} else {
rdentp = ERR_PTR(mdki_errno_unix_to_linux(err));
}
return rdentp;
}
STATIC struct dentry *
linvfs_lookup(
struct inode *dir,
struct dentry *dentry)
{
struct inode *ip = NULL;
vnode_t *vp, *cvp = NULL;
int error;
if (dentry->d_name.len >= MAXNAMELEN)
return ERR_PTR(-ENAMETOOLONG);
vp = LINVFS_GET_VP(dir);
VOP_LOOKUP(vp, dentry, &cvp, 0, NULL, NULL, error);
if (!error) {
ASSERT(cvp);
ip = LINVFS_GET_IP(cvp);
if (!ip) {
VN_RELE(cvp);
return ERR_PTR(-EACCES);
}
}
if (error && (error != ENOENT))
return ERR_PTR(-error);
d_add(dentry, ip); /* Negative entry goes in if ip is NULL */
return NULL;
}
/* ARGSUSED */
static int
zfsctl_shares_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, pathname_t *pnp,
int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
int *direntflags, pathname_t *realpnp)
{
zfsvfs_t *zfsvfs = dvp->v_vfsp->vfs_data;
znode_t *dzp;
int error;
ZFS_ENTER(zfsvfs);
if (gfs_lookup_dot(vpp, dvp, zfsvfs->z_ctldir, nm) == 0) {
ZFS_EXIT(zfsvfs);
return (0);
}
if (zfsvfs->z_shares_dir == 0) {
ZFS_EXIT(zfsvfs);
return (ENOTSUP);
}
if ((error = zfs_zget(zfsvfs, zfsvfs->z_shares_dir, &dzp)) == 0)
error = VOP_LOOKUP(ZTOV(dzp), nm, vpp, pnp,
flags, rdir, cr, ct, direntflags, realpnp);
VN_RELE(ZTOV(dzp));
ZFS_EXIT(zfsvfs);
return (error);
}
int
nfs4_readdir_getvp(vnode_t *dvp, char *d_name, vnode_t **vpp,
struct exportinfo **exi, struct svc_req *req,
struct compound_state *cs, int expseudo)
{
int error;
int ismntpt;
fid_t fid;
vnode_t *vp, *pre_tvp;
nfsstat4 status;
struct exportinfo *newexi, *saveexi;
cred_t *scr;
*vpp = vp = NULL;
if (error = VOP_LOOKUP(dvp, d_name, &vp, NULL, 0, NULL, cs->cr,
NULL, NULL, NULL))
return (error);
/*
* If the directory is a referral point, don't return the
* attrs, instead set rdattr_error to MOVED.
*/
if (vn_is_nfs_reparse(vp, cs->cr) && !client_is_downrev(req)) {
VN_RELE(vp);
DTRACE_PROBE2(nfs4serv__func__referral__moved,
vnode_t *, vp, char *, "nfs4_readdir_getvp");
return (NFS4ERR_MOVED);
}
/* ARGSUSED */
void
acl3_getxattrdir(GETXATTRDIR3args *args, GETXATTRDIR3res *resp,
struct exportinfo *exi, struct svc_req *req, cred_t *cr, bool_t ro)
{
int error;
int flags;
vnode_t *vp, *avp;
vp = nfs3_fhtovp(&args->fh, exi);
if (vp == NULL) {
resp->status = NFS3ERR_STALE;
return;
}
flags = LOOKUP_XATTR;
if (args->create)
flags |= CREATE_XATTR_DIR;
else {
ulong_t val = 0;
error = VOP_PATHCONF(vp, _PC_SATTR_EXISTS, &val, cr, NULL);
if (!error && val == 0) {
error = VOP_PATHCONF(vp, _PC_XATTR_EXISTS,
&val, cr, NULL);
if (!error && val == 0) {
VN_RELE(vp);
resp->status = NFS3ERR_NOENT;
return;
}
}
}
error = VOP_LOOKUP(vp, "", &avp, NULL, flags, NULL, cr,
NULL, NULL, NULL);
if (!error && avp == vp) { /* lookup of "" on old FS? */
error = EINVAL;
VN_RELE(avp);
}
if (!error) {
struct vattr va;
va.va_mask = AT_ALL;
error = rfs4_delegated_getattr(avp, &va, 0, cr);
if (!error) {
vattr_to_post_op_attr(&va, &resp->resok.attr);
error = makefh3(&resp->resok.fh, avp, exi);
}
VN_RELE(avp);
}
VN_RELE(vp);
if (error) {
resp->status = puterrno3(error);
return;
}
resp->status = NFS3_OK;
}
static int
zfs_replay_remove(zfsvfs_t *zfsvfs, void *data, boolean_t byteswap)
{
#ifndef TODO_OSV
kprintf("TX_REMOVE\n");
return EOPNOTSUPP;
#else
lr_remove_t *lr = data;
char *name = (char *)(lr + 1); /* name follows lr_remove_t */
znode_t *dzp;
struct componentname cn;
vnode_t *vp;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_doid, &dzp)) != 0)
return (error);
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
cn.cn_nameptr = name;
cn.cn_namelen = strlen(name);
cn.cn_nameiop = DELETE;
cn.cn_flags = ISLASTCN | SAVENAME;
cn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
cn.cn_cred = kcred;
cn.cn_thread = curthread;
vn_lock(ZTOV(dzp), LK_EXCLUSIVE | LK_RETRY);
error = VOP_LOOKUP(ZTOV(dzp), &vp, &cn);
if (error != 0) {
VOP_UNLOCK(ZTOV(dzp), 0);
goto fail;
}
switch ((int)lr->lr_common.lrc_txtype) {
case TX_REMOVE:
error = VOP_REMOVE(ZTOV(dzp), vp, &cn /*,vflg*/);
break;
case TX_RMDIR:
error = VOP_RMDIR(ZTOV(dzp), vp, &cn /*,vflg*/);
break;
default:
error = ENOTSUP;
}
vput(vp);
VOP_UNLOCK(ZTOV(dzp), 0);
fail:
VN_RELE(ZTOV(dzp));
return (error);
#endif
}
int
smb_vop_lookup_xattrdir(vnode_t *fvp, vnode_t **xattrdirvpp, int flags,
cred_t *cr)
{
int error;
error = VOP_LOOKUP(fvp, "", xattrdirvpp, NULL, flags, NULL, cr,
&smb_ct, NULL, NULL);
return (error);
}
/*
* Given a global path (from rootdir), and a vnode that is the current root,
* return the portion of the path that is beneath the current root or NULL on
* failure. The path MUST be a resolved path (no '..' entries or symlinks),
* otherwise this function will fail.
*/
static char *
localpath(char *path, struct vnode *vrootp, cred_t *cr)
{
vnode_t *vp;
vnode_t *cvp;
char component[MAXNAMELEN];
char *ret = NULL;
pathname_t pn;
/*
* We use vn_compare() instead of VN_CMP() in order to detect lofs
* mounts and stacked vnodes.
*/
if (vn_compare(vrootp, rootdir))
return (path);
if (pn_get(path, UIO_SYSSPACE, &pn) != 0)
return (NULL);
vp = rootdir;
VN_HOLD(vp);
if (vn_ismntpt(vp) && traverse(&vp) != 0) {
VN_RELE(vp);
pn_free(&pn);
return (NULL);
}
while (pn_pathleft(&pn)) {
pn_skipslash(&pn);
if (pn_getcomponent(&pn, component) != 0)
break;
if (VOP_LOOKUP(vp, component, &cvp, &pn, 0, rootdir, cr,
NULL, NULL, NULL) != 0)
break;
VN_RELE(vp);
vp = cvp;
if (vn_ismntpt(vp) && traverse(&vp) != 0)
break;
if (vn_compare(vp, vrootp)) {
ret = path + (pn.pn_path - pn.pn_buf);
break;
}
}
VN_RELE(vp);
pn_free(&pn);
return (ret);
}
static int
VMBlockLookup(struct vnode *dvp, // IN: Directory to look in
char *nm, // IN: Name of component to lookup in directory
struct vnode **vpp, // OUT: Pointer to vnode representing found file
struct pathname *pnp,// IN: Full pathname being looked up
int flags, // IN: Lookup flags (see vnode.h)
struct vnode *rdir, // IN: Vnode of root device
struct cred *cr // IN: Credentials of caller
#if OS_VFS_VERSION >= 5
, caller_context_t *ctx // IN: Caller's context
, int *direntflags // IN:
, struct pathname *rpnp // IN:
#endif
)
{
struct vnode *realVp;
VMBlockMountInfo *mip;
int ret;
Debug(VMBLOCK_ENTRY_LOGLEVEL, "VMblockLookup: entry\n");
/* First ensure that we are looking in a directory. */
if (dvp->v_type != VDIR) {
return ENOTDIR;
}
/* Don't invoke lookup for ourself. */
if (nm[0] == '\0' || (nm[0] == '.' && nm[1] == '\0')) {
VN_HOLD(dvp);
*vpp = dvp;
return 0;
}
*vpp = NULL;
/* Make sure nm exists before creating our link to it. */
mip = VPTOMIP(dvp);
ret = VOP_LOOKUP(mip->redirectVnode, nm, &realVp, pnp, flags, rdir, cr
#if OS_VFS_VERSION >= 5
, ctx, direntflags, rpnp
#endif
);
if (ret) {
return ret;
}
ret = VMBlockVnodeGet(vpp, realVp, nm, strlen(nm), dvp, dvp->v_vfsp, FALSE);
if (ret) {
VN_RELE(realVp);
return ret;
}
return 0;
}
int
RUMP_VOP_LOOKUP(struct vnode *dvp,
struct vnode **vpp,
struct componentname *cnp)
{
int error;
rump_schedule();
error = VOP_LOOKUP(dvp, vpp, cnp);
rump_unschedule();
return error;
}
/*
* Create a directory node in a non-global dev instance.
* Always create shadow vnode. Set sdev_origin to the corresponding
* global directory sdev_node if it exists. This facilitates the
* lookup operation.
*/
static int
prof_make_dir(char *name, struct sdev_node **gdirp, struct sdev_node **dirp)
{
struct sdev_node *dir = *dirp;
struct sdev_node *gdir = *gdirp;
struct sdev_node *newdv;
struct vnode *avp, *gnewdir = NULL;
struct vattr vattr;
int error;
/* see if name already exists */
rw_enter(&dir->sdev_contents, RW_READER);
if (newdv = sdev_cache_lookup(dir, name)) {
*dirp = newdv;
*gdirp = newdv->sdev_origin;
rw_exit(&dir->sdev_contents);
SDEV_RELE(dir);
return (0);
}
rw_exit(&dir->sdev_contents);
/* find corresponding dir node in global dev */
if (gdir) {
error = VOP_LOOKUP(SDEVTOV(gdir), name, &gnewdir,
NULL, 0, NULL, kcred, NULL, NULL, NULL);
if (error == 0) {
*gdirp = VTOSDEV(gnewdir);
} else { /* it's ok if there no global dir */
*gdirp = NULL;
}
}
/* get attribute from shadow, also create shadow dir */
prof_getattr(dir, name, gnewdir, &vattr, &avp, NULL);
/* create dev directory vnode */
rw_enter(&dir->sdev_contents, RW_WRITER);
error = prof_mknode(dir, name, &newdv, &vattr, avp, (void *)*gdirp,
kcred);
rw_exit(&dir->sdev_contents);
if (error == 0) {
ASSERT(newdv);
*dirp = newdv;
}
SDEV_RELE(dir);
return (error);
}
static void
prof_getattr(struct sdev_node *dir, char *name, struct vnode *gdv,
struct vattr *vap, struct vnode **avpp, int *no_fs_perm)
{
struct vnode *advp;
/* get attribute from shadow, if present; else get default */
advp = dir->sdev_attrvp;
if (advp && VOP_LOOKUP(advp, name, avpp, NULL, 0, NULL, kcred,
NULL, NULL, NULL) == 0) {
(void) VOP_GETATTR(*avpp, vap, 0, kcred, NULL);
} else if (gdv == NULL || gdv->v_type == VDIR) {
/* always create shadow directory */
*vap = sdev_vattr_dir;
if (advp && VOP_MKDIR(advp, name, &sdev_vattr_dir,
avpp, kcred, NULL, 0, NULL) != 0) {
*avpp = NULLVP;
sdcmn_err10(("prof_getattr: failed to create "
"shadow directory %s/%s\n", dir->sdev_path, name));
}
} else {
/*
* get default permission from devfs
* Before calling devfs_get_defattr, we need to get
* the realvp (the dv_node). If realvp is not a dv_node,
* devfs_get_defattr() will return a system-wide default
* attr for device nodes.
*/
struct vnode *rvp;
if (VOP_REALVP(gdv, &rvp, NULL) != 0)
rvp = gdv;
devfs_get_defattr(rvp, vap, no_fs_perm);
*avpp = NULLVP;
}
/* ignore dev_t and vtype from backing store */
if (gdv) {
vap->va_type = gdv->v_type;
vap->va_rdev = gdv->v_rdev;
}
}
STATIC struct dentry *
linvfs_lookup(
struct inode *dir,
struct dentry *dentry,
struct nameidata *nd)
{
struct vnode *vp = LINVFS_GET_VP(dir), *cvp;
int error;
if (dentry->d_name.len >= MAXNAMELEN)
return ERR_PTR(-ENAMETOOLONG);
VOP_LOOKUP(vp, dentry, &cvp, 0, NULL, NULL, error);
if (error) {
if (unlikely(error != ENOENT))
return ERR_PTR(-error);
d_add(dentry, NULL);
return NULL;
}
return d_splice_alias(LINVFS_GET_IP(cvp), dentry);
}
/*
* We have to carry on the locking protocol on the null layer vnodes
* as we progress through the tree. We also have to enforce read-only
* if this layer is mounted read-only.
*/
static int
null_lookup(struct vop_lookup_args *ap)
{
struct componentname *cnp = ap->a_cnp;
struct vnode *dvp = ap->a_dvp;
int flags = cnp->cn_flags;
struct vnode *vp, *ldvp, *lvp;
int error;
if ((flags & ISLASTCN) && (dvp->v_mount->mnt_flag & MNT_RDONLY) &&
(cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME))
return (EROFS);
/*
* Although it is possible to call null_bypass(), we'll do
* a direct call to reduce overhead
*/
ldvp = NULLVPTOLOWERVP(dvp);
vp = lvp = NULL;
error = VOP_LOOKUP(ldvp, &lvp, cnp);
if (error == EJUSTRETURN && (flags & ISLASTCN) &&
(dvp->v_mount->mnt_flag & MNT_RDONLY) &&
(cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME))
error = EROFS;
if ((error == 0 || error == EJUSTRETURN) && lvp != NULL) {
if (ldvp == lvp) {
*ap->a_vpp = dvp;
VREF(dvp);
vrele(lvp);
} else {
error = null_nodeget(dvp->v_mount, lvp, &vp);
if (error)
vput(lvp);
else
*ap->a_vpp = vp;
}
}
return (error);
}
static int
zfs_replay_rename(zfsvfs_t *zfsvfs, lr_rename_t *lr, boolean_t byteswap)
{
char *sname = (char *)(lr + 1); /* sname and tname follow lr_rename_t */
char *tname = sname + strlen(sname) + 1;
znode_t *sdzp, *tdzp;
struct componentname scn, tcn;
vnode_t *svp, *tvp;
kthread_t *td = curthread;
int error;
int vflg = 0;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
if ((error = zfs_zget(zfsvfs, lr->lr_sdoid, &sdzp)) != 0)
return (error);
if ((error = zfs_zget(zfsvfs, lr->lr_tdoid, &tdzp)) != 0) {
VN_RELE(ZTOV(sdzp));
return (error);
}
if (lr->lr_common.lrc_txtype & TX_CI)
vflg |= FIGNORECASE;
svp = tvp = NULL;
scn.cn_nameptr = sname;
scn.cn_namelen = strlen(sname);
scn.cn_nameiop = DELETE;
scn.cn_flags = ISLASTCN | SAVENAME;
scn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
scn.cn_cred = kcred;
scn.cn_thread = td;
vn_lock(ZTOV(sdzp), LK_EXCLUSIVE | LK_RETRY);
error = VOP_LOOKUP(ZTOV(sdzp), &svp, &scn);
VOP_UNLOCK(ZTOV(sdzp), 0);
if (error != 0)
goto fail;
VOP_UNLOCK(svp, 0);
tcn.cn_nameptr = tname;
tcn.cn_namelen = strlen(tname);
tcn.cn_nameiop = RENAME;
tcn.cn_flags = ISLASTCN | SAVENAME;
tcn.cn_lkflags = LK_EXCLUSIVE | LK_RETRY;
tcn.cn_cred = kcred;
tcn.cn_thread = td;
vn_lock(ZTOV(tdzp), LK_EXCLUSIVE | LK_RETRY);
error = VOP_LOOKUP(ZTOV(tdzp), &tvp, &tcn);
if (error == EJUSTRETURN)
tvp = NULL;
else if (error != 0) {
VOP_UNLOCK(ZTOV(tdzp), 0);
goto fail;
}
error = VOP_RENAME(ZTOV(sdzp), svp, &scn, ZTOV(tdzp), tvp, &tcn /*,vflg*/);
return (error);
fail:
if (svp != NULL)
vrele(svp);
if (tvp != NULL)
vrele(tvp);
VN_RELE(ZTOV(tdzp));
VN_RELE(ZTOV(sdzp));
return (error);
}
/*
* This function checks the path to a new export to
* check whether all the pathname components are
* exported. It works by climbing the file tree one
* component at a time via "..", crossing mountpoints
* if necessary until an export entry is found, or the
* system root is reached.
*
* If an unexported mountpoint is found, then
* a new pseudo export is added and the pathname from
* the mountpoint down to the export is added to the
* visible list for the new pseudo export. If an existing
* pseudo export is found, then the pathname is added
* to its visible list.
*
* Note that there's some tests for exportdir.
* The exportinfo entry that's passed as a parameter
* is that of the real export and exportdir is set
* for this case.
*
* Here is an example of a possible setup:
*
* () - a new fs; fs mount point
* EXPORT - a real exported node
* PSEUDO - a pseudo node
* vis - visible list
* f# - security flavor#
* (f#) - security flavor# propagated from its descendents
* "" - covered vnode
*
*
* /
* |
* (a) PSEUDO (f1,f2)
* | vis: b,b,"c","n"
* |
* b
* ---------|------------------
* | |
* (c) EXPORT,f1(f2) (n) PSEUDO (f1,f2)
* | vis: "e","d" | vis: m,m,,p,q,"o"
* | |
* ------------------ -------------------
* | | | | |
* (d) (e) f m EXPORT,f1(f2) p
* EXPORT EXPORT | |
* f1 f2 | |
* | | |
* j (o) EXPORT,f2 q EXPORT f2
*
*/
int
treeclimb_export(struct exportinfo *exip)
{
vnode_t *dvp, *vp;
fid_t fid;
int error;
int exportdir;
struct exportinfo *exi = NULL;
struct exportinfo *new_exi = exip;
struct exp_visible *visp;
struct exp_visible *vis_head = NULL;
struct vattr va;
treenode_t *tree_head = NULL;
ASSERT(RW_WRITE_HELD(&exported_lock));
vp = exip->exi_vp;
VN_HOLD(vp);
exportdir = 1;
for (;;) {
bzero(&fid, sizeof (fid));
fid.fid_len = MAXFIDSZ;
error = vop_fid_pseudo(vp, &fid);
if (error)
break;
if (! exportdir) {
/*
* Check if this exportroot is a VROOT dir. If so,
* then attach the pseudonodes. If not, then
* continue .. traversal until we hit a VROOT
* export (pseudo or real).
*/
exi = checkexport4(&vp->v_vfsp->vfs_fsid, &fid, vp);
if (exi != NULL && vp->v_flag & VROOT) {
/*
* Found an export info
*
* Extend the list of visible
* directories whether it's a pseudo
* or a real export.
*/
more_visible(exi, tree_head);
break; /* and climb no further */
}
}
/*
* If at the root of the filesystem, need
* to traverse across the mountpoint
* and continue the climb on the mounted-on
* filesystem.
*/
if (vp->v_flag & VROOT) {
if (! exportdir) {
/*
* Found the root directory of a filesystem
* that isn't exported. Need to export
* this as a pseudo export so that an NFS v4
* client can do lookups in it.
*/
new_exi = pseudo_exportfs(vp, &fid, vis_head,
NULL);
vis_head = NULL;
}
if (VN_CMP(vp, rootdir)) {
/* at system root */
/*
* If sharing "/", new_exi is shared exportinfo
* (exip). Otherwise, new_exi is exportinfo
* created in pseudo_exportfs() above.
*/
ns_root = tree_prepend_node(tree_head, 0,
new_exi);
break;
}
vp = untraverse(vp);
exportdir = 0;
continue;
}
/*
* Do a getattr to obtain the nodeid (inode num)
* for this vnode.
*/
va.va_mask = AT_NODEID;
error = VOP_GETATTR(vp, &va, 0, CRED(), NULL);
if (error)
break;
/*
* Add this directory fid to visible list
*/
visp = kmem_alloc(sizeof (*visp), KM_SLEEP);
VN_HOLD(vp);
visp->vis_vp = vp;
visp->vis_fid = fid; /* structure copy */
visp->vis_ino = va.va_nodeid;
visp->vis_count = 1;
visp->vis_exported = exportdir;
visp->vis_secinfo = NULL;
visp->vis_seccnt = 0;
visp->vis_next = vis_head;
vis_head = visp;
/*
* Will set treenode's pointer to exportinfo to
* 1. shared exportinfo (exip) - if first visit here
* 2. freshly allocated pseudo export (if any)
* 3. null otherwise
*/
tree_head = tree_prepend_node(tree_head, visp, new_exi);
new_exi = NULL;
/*
* Now, do a ".." to find parent dir of vp.
*/
error = VOP_LOOKUP(vp, "..", &dvp, NULL, 0, NULL, CRED(),
NULL, NULL, NULL);
if (error == ENOTDIR && exportdir) {
dvp = exip->exi_dvp;
ASSERT(dvp != NULL);
VN_HOLD(dvp);
error = 0;
}
if (error)
break;
exportdir = 0;
VN_RELE(vp);
vp = dvp;
}
VN_RELE(vp);
/*
* We can have set error due to error in:
* 1. vop_fid_pseudo()
* 2. VOP_GETATTR()
* 3. VOP_LOOKUP()
* We must free pseudo exportinfos, visibles and treenodes.
* Visibles are referenced from treenode_t::tree_vis and
* exportinfo_t::exi_visible. To avoid double freeing, only
* exi_visible pointer is used, via exi_rele(), for the clean-up.
*/
if (error) {
/* Free unconnected visibles, if there are any. */
if (vis_head)
free_visible(vis_head);
/* Connect unconnected exportinfo, if there is any. */
if (new_exi && new_exi != exip)
tree_head = tree_prepend_node(tree_head, 0, new_exi);
while (tree_head) {
treenode_t *t2 = tree_head;
exportinfo_t *e = tree_head->tree_exi;
/* exip will be freed in exportfs() */
if (e && e != exip) {
exp_kstats_delete(e->exi_kstats);
avl_remove(&exi_id_tree, e);
export_unlink(e);
exi_rele(e);
}
tree_head = tree_head->tree_child_first;
kmem_free(t2, sizeof (*t2));
}
}
return (error);
}
static int
auto_lookup(
vnode_t *dvp,
char *nm,
vnode_t **vpp,
pathname_t *pnp,
int flags,
vnode_t *rdir,
cred_t *cred,
caller_context_t *ct,
int *direntflags,
pathname_t *realpnp)
{
int error = 0;
vnode_t *newvp = NULL;
vfs_t *vfsp;
fninfo_t *dfnip;
fnnode_t *dfnp = NULL;
fnnode_t *fnp = NULL;
char *searchnm;
int operation; /* either AUTOFS_LOOKUP or AUTOFS_MOUNT */
dfnip = vfstofni(dvp->v_vfsp);
AUTOFS_DPRINT((3, "auto_lookup: dvp=%p (%s) name=%s\n",
(void *)dvp, dfnip->fi_map, nm));
if (nm[0] == 0) {
VN_HOLD(dvp);
*vpp = dvp;
return (0);
}
if (error = VOP_ACCESS(dvp, VEXEC, 0, cred, ct))
return (error);
if (nm[0] == '.' && nm[1] == 0) {
VN_HOLD(dvp);
*vpp = dvp;
return (0);
}
if (nm[0] == '.' && nm[1] == '.' && nm[2] == 0) {
fnnode_t *pdfnp;
pdfnp = (vntofn(dvp))->fn_parent;
ASSERT(pdfnp != NULL);
/*
* Since it is legitimate to have the VROOT flag set for the
* subdirectories of the indirect map in autofs filesystem,
* rootfnnodep is checked against fnnode of dvp instead of
* just checking whether VROOT flag is set in dvp
*/
if (pdfnp == pdfnp->fn_globals->fng_rootfnnodep) {
vnode_t *vp;
vfs_rlock_wait(dvp->v_vfsp);
if (dvp->v_vfsp->vfs_flag & VFS_UNMOUNTED) {
vfs_unlock(dvp->v_vfsp);
return (EIO);
}
vp = dvp->v_vfsp->vfs_vnodecovered;
VN_HOLD(vp);
vfs_unlock(dvp->v_vfsp);
error = VOP_LOOKUP(vp, nm, vpp, pnp, flags, rdir, cred,
ct, direntflags, realpnp);
VN_RELE(vp);
return (error);
} else {
*vpp = fntovn(pdfnp);
VN_HOLD(*vpp);
return (0);
}
}
top:
dfnp = vntofn(dvp);
searchnm = nm;
operation = 0;
ASSERT(vn_matchops(dvp, auto_vnodeops));
AUTOFS_DPRINT((3, "auto_lookup: dvp=%p dfnp=%p\n", (void *)dvp,
(void *)dfnp));
/*
* If a lookup or mount of this node is in progress, wait for it
* to finish, and return whatever result it got.
*/
mutex_enter(&dfnp->fn_lock);
if (dfnp->fn_flags & (MF_LOOKUP | MF_INPROG)) {
mutex_exit(&dfnp->fn_lock);
error = auto_wait4mount(dfnp);
if (error == AUTOFS_SHUTDOWN)
error = ENOENT;
if (error == EAGAIN)
goto top;
if (error)
return (error);
} else
mutex_exit(&dfnp->fn_lock);
error = vn_vfsrlock_wait(dvp);
if (error)
return (error);
vfsp = vn_mountedvfs(dvp);
if (vfsp != NULL) {
error = VFS_ROOT(vfsp, &newvp);
vn_vfsunlock(dvp);
if (!error) {
error = VOP_LOOKUP(newvp, nm, vpp, pnp,
flags, rdir, cred, ct, direntflags, realpnp);
VN_RELE(newvp);
}
return (error);
}
vn_vfsunlock(dvp);
rw_enter(&dfnp->fn_rwlock, RW_READER);
error = auto_search(dfnp, nm, &fnp, cred);
if (error) {
if (dfnip->fi_flags & MF_DIRECT) {
/*
* direct map.
*/
if (dfnp->fn_dirents) {
/*
* Mount previously triggered.
* 'nm' not found
*/
error = ENOENT;
} else {
/*
* I need to contact the daemon to trigger
* the mount. 'dfnp' will be the mountpoint.
*/
operation = AUTOFS_MOUNT;
VN_HOLD(fntovn(dfnp));
fnp = dfnp;
error = 0;
}
} else if (dvp == dfnip->fi_rootvp) {
/*
* 'dfnp' is the root of the indirect AUTOFS.
*/
if (rw_tryupgrade(&dfnp->fn_rwlock) == 0) {
/*
* Could not acquire writer lock, release
* reader, and wait until available. We
* need to search for 'nm' again, since we
* had to release the lock before reacquiring
* it.
*/
rw_exit(&dfnp->fn_rwlock);
rw_enter(&dfnp->fn_rwlock, RW_WRITER);
error = auto_search(dfnp, nm, &fnp, cred);
}
ASSERT(RW_WRITE_HELD(&dfnp->fn_rwlock));
if (error) {
/*
* create node being looked-up and request
* mount on it.
*/
error = auto_enter(dfnp, nm, &fnp, kcred);
if (!error)
operation = AUTOFS_LOOKUP;
}
} else if ((dfnp->fn_dirents == NULL) &&
((dvp->v_flag & VROOT) == 0) &&
((fntovn(dfnp->fn_parent))->v_flag & VROOT)) {
/*
* dfnp is the actual 'mountpoint' of indirect map,
* it is the equivalent of a direct mount,
* ie, /home/'user1'
*/
operation = AUTOFS_MOUNT;
VN_HOLD(fntovn(dfnp));
fnp = dfnp;
error = 0;
searchnm = dfnp->fn_name;
}
}
if (error == EAGAIN) {
rw_exit(&dfnp->fn_rwlock);
goto top;
}
if (error) {
rw_exit(&dfnp->fn_rwlock);
return (error);
}
/*
* We now have the actual fnnode we're interested in.
* The 'MF_LOOKUP' indicates another thread is currently
* performing a daemon lookup of this node, therefore we
* wait for its completion.
* The 'MF_INPROG' indicates another thread is currently
* performing a daemon mount of this node, we wait for it
* to be done if we are performing a MOUNT. We don't
* wait for it if we are performing a LOOKUP.
* We can release the reader/writer lock as soon as we acquire
* the mutex, since the state of the lock can only change by
* first acquiring the mutex.
*/
mutex_enter(&fnp->fn_lock);
rw_exit(&dfnp->fn_rwlock);
if ((fnp->fn_flags & MF_LOOKUP) ||
((operation == AUTOFS_MOUNT) && (fnp->fn_flags & MF_INPROG))) {
mutex_exit(&fnp->fn_lock);
error = auto_wait4mount(fnp);
VN_RELE(fntovn(fnp));
if (error == AUTOFS_SHUTDOWN)
error = ENOENT;
if (error && error != EAGAIN)
return (error);
goto top;
}
if (operation == 0) {
/*
* got the fnnode, check for any errors
* on the previous operation on that node.
*/
error = fnp->fn_error;
if ((error == EINTR) || (error == EAGAIN)) {
/*
* previous operation on this node was
* not completed, do a lookup now.
*/
operation = AUTOFS_LOOKUP;
} else {
/*
* previous operation completed. Return
* a pointer to the node only if there was
* no error.
*/
mutex_exit(&fnp->fn_lock);
if (!error)
*vpp = fntovn(fnp);
else
VN_RELE(fntovn(fnp));
return (error);
}
}
/*
* Since I got to this point, it means I'm the one
* responsible for triggering the mount/look-up of this node.
*/
switch (operation) {
case AUTOFS_LOOKUP:
AUTOFS_BLOCK_OTHERS(fnp, MF_LOOKUP);
fnp->fn_error = 0;
mutex_exit(&fnp->fn_lock);
error = auto_lookup_aux(fnp, searchnm, cred);
if (!error) {
/*
* Return this vnode
*/
*vpp = fntovn(fnp);
} else {
/*
* release our reference to this vnode
* and return error
*/
VN_RELE(fntovn(fnp));
}
break;
case AUTOFS_MOUNT:
AUTOFS_BLOCK_OTHERS(fnp, MF_INPROG);
fnp->fn_error = 0;
mutex_exit(&fnp->fn_lock);
/*
* auto_new_mount_thread fires up a new thread which
* calls automountd finishing up the work
*/
auto_new_mount_thread(fnp, searchnm, cred);
/*
* At this point, we are simply another thread
* waiting for the mount to complete
*/
error = auto_wait4mount(fnp);
if (error == AUTOFS_SHUTDOWN)
error = ENOENT;
/*
* now release our reference to this vnode
*/
VN_RELE(fntovn(fnp));
if (!error)
goto top;
break;
default:
auto_log(dfnp->fn_globals->fng_verbose,
dfnp->fn_globals->fng_zoneid, CE_WARN,
"auto_lookup: unknown operation %d",
operation);
}
AUTOFS_DPRINT((5, "auto_lookup: name=%s *vpp=%p return=%d\n",
nm, (void *)*vpp, error));
return (error);
}
static int
unionfs_lookup(void *v)
{
struct vop_lookup_args *ap = v;
int iswhiteout;
int lockflag;
int error , uerror, lerror;
u_long nameiop;
u_long cnflags, cnflagsbk;
struct unionfs_node *dunp;
struct vnode *dvp, *udvp, *ldvp, *vp, *uvp, *lvp, *dtmpvp;
struct vattr va;
struct componentname *cnp;
iswhiteout = 0;
lockflag = 0;
error = uerror = lerror = ENOENT;
cnp = ap->a_cnp;
nameiop = cnp->cn_nameiop;
cnflags = cnp->cn_flags;
dvp = ap->a_dvp;
dunp = VTOUNIONFS(dvp);
udvp = dunp->un_uppervp;
ldvp = dunp->un_lowervp;
vp = uvp = lvp = NULLVP;
*(ap->a_vpp) = NULLVP;
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: enter: nameiop=%ld, flags=%lx, path=%s\n", nameiop, cnflags, cnp->cn_nameptr);
if (dvp->v_type != VDIR)
return (ENOTDIR);
/*
* If read-only and op is not LOOKUP, will return EROFS.
*/
if ((cnflags & ISLASTCN) &&
(dvp->v_mount->mnt_flag & MNT_RDONLY) &&
LOOKUP != nameiop)
return (EROFS);
/*
* lookup dotdot
*/
if (cnflags & ISDOTDOT) {
if (LOOKUP != nameiop && udvp == NULLVP)
return (EROFS);
if (udvp != NULLVP) {
dtmpvp = udvp;
if (ldvp != NULLVP)
VOP_UNLOCK(ldvp);
}
else
dtmpvp = ldvp;
error = VOP_LOOKUP(dtmpvp, &vp, cnp);
if (dtmpvp == udvp && ldvp != NULLVP) {
VOP_UNLOCK(udvp);
vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
}
if (error == 0) {
/*
* Exchange lock and reference from vp to
* dunp->un_dvp. vp is upper/lower vnode, but it
* will need to return the unionfs vnode.
*/
if (nameiop == DELETE || nameiop == RENAME)
VOP_UNLOCK(vp);
vrele(vp);
VOP_UNLOCK(dvp);
*(ap->a_vpp) = dunp->un_dvp;
vref(dunp->un_dvp);
vn_lock(dunp->un_dvp, LK_EXCLUSIVE | LK_RETRY);
vn_lock(dvp, LK_EXCLUSIVE | LK_RETRY);
} else if (error == ENOENT && nameiop != CREATE)
cache_enter(dvp, NULLVP, cnp->cn_nameptr,
cnp->cn_namelen, cnp->cn_flags);
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", error);
return (error);
}
/*
* lookup upper layer
*/
if (udvp != NULLVP) {
uerror = VOP_LOOKUP(udvp, &uvp, cnp);
if (uerror == 0) {
if (udvp == uvp) { /* is dot */
vrele(uvp);
*(ap->a_vpp) = dvp;
vref(dvp);
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", uerror);
return (uerror);
}
}
/* check whiteout */
if (uerror == ENOENT || uerror == EJUSTRETURN)
if (cnp->cn_flags & ISWHITEOUT)
iswhiteout = 1; /* don't lookup lower */
if (iswhiteout == 0 && ldvp != NULLVP)
if (VOP_GETATTR(udvp, &va, cnp->cn_cred) == 0 &&
(va.va_flags & OPAQUE))
iswhiteout = 1; /* don't lookup lower */
#if 0
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: debug: whiteout=%d, path=%s\n", iswhiteout, cnp->cn_nameptr);
#endif
}
/*
* lookup lower layer
*/
if (ldvp != NULLVP && !(cnflags & DOWHITEOUT) && iswhiteout == 0) {
/* always op is LOOKUP */
cnp->cn_nameiop = LOOKUP;
cnflagsbk = cnp->cn_flags;
cnp->cn_flags = cnflags;
lerror = VOP_LOOKUP(ldvp, &lvp, cnp);
cnp->cn_nameiop = nameiop;
if (udvp != NULLVP && (uerror == 0 || uerror == EJUSTRETURN))
cnp->cn_flags = cnflagsbk;
if (lerror == 0) {
if (ldvp == lvp) { /* is dot */
if (uvp != NULLVP)
vrele(uvp); /* no need? */
vrele(lvp);
*(ap->a_vpp) = dvp;
vref(dvp);
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", lerror);
if (uvp != NULL)
VOP_UNLOCK(uvp);
return (lerror);
}
}
}
/*
* check lookup result
*/
if (uvp == NULLVP && lvp == NULLVP) {
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n",
(udvp != NULLVP ? uerror : lerror));
return (udvp != NULLVP ? uerror : lerror);
}
/*
* check vnode type
*/
if (uvp != NULLVP && lvp != NULLVP && uvp->v_type != lvp->v_type) {
vput(lvp);
lvp = NULLVP;
}
/*
* check shadow dir
*/
if (uerror != 0 && uerror != EJUSTRETURN && udvp != NULLVP &&
lerror == 0 && lvp != NULLVP && lvp->v_type == VDIR &&
!(dvp->v_mount->mnt_flag & MNT_RDONLY) &&
(1 < cnp->cn_namelen || '.' != *(cnp->cn_nameptr))) {
/* get unionfs vnode in order to create a new shadow dir. */
error = unionfs_nodeget(dvp->v_mount, NULLVP, lvp, dvp, &vp,
cnp);
if (error != 0)
goto unionfs_lookup_out;
error = unionfs_mkshadowdir(MOUNTTOUNIONFSMOUNT(dvp->v_mount),
udvp, VTOUNIONFS(vp), cnp);
if (error != 0) {
UNIONFSDEBUG("unionfs_lookup: Unable to create shadow dir.");
vput(vp);
goto unionfs_lookup_out;
}
}
/*
* get unionfs vnode.
*/
else {
if (uvp != NULLVP)
error = uerror;
else
error = lerror;
if (error != 0)
goto unionfs_lookup_out;
error = unionfs_nodeget(dvp->v_mount, uvp, lvp, dvp, &vp, cnp);
if (error != 0) {
UNIONFSDEBUG("unionfs_lookup: Unable to create unionfs vnode.");
goto unionfs_lookup_out;
}
}
*(ap->a_vpp) = vp;
cache_enter(dvp, vp, cnp->cn_nameptr, cnp->cn_namelen,
cnp->cn_flags);
/* XXXAD lock status on error */
unionfs_lookup_out:
if (uvp != NULLVP)
vrele(uvp);
if (lvp != NULLVP)
vrele(lvp);
if (error == ENOENT && nameiop != CREATE)
cache_enter(dvp, NULLVP, cnp->cn_nameptr, cnp->cn_namelen,
cnp->cn_flags);
UNIONFS_INTERNAL_DEBUG("unionfs_lookup: leave (%d)\n", error);
return (error);
}
static int
copen(int startfd, char *fname, int filemode, int createmode)
{
struct pathname pn;
vnode_t *vp, *sdvp;
file_t *fp, *startfp;
enum vtype type;
int error;
int fd, dupfd;
vnode_t *startvp;
proc_t *p = curproc;
if (startfd == AT_FDCWD) {
/*
* Regular open()
*/
startvp = NULL;
} else {
/*
* We're here via openat()
*/
char startchar;
if (copyin(fname, &startchar, sizeof (char)))
return (set_errno(EFAULT));
/*
* if startchar is / then startfd is ignored
*/
if (startchar == '/')
startvp = NULL;
else {
if ((startfp = getf(startfd)) == NULL)
return (set_errno(EBADF));
startvp = startfp->f_vnode;
VN_HOLD(startvp);
releasef(startfd);
}
}
if (filemode & FXATTR) {
/*
* Make sure we have a valid request.
* We must either have a real fd or AT_FDCWD
*/
if (startfd != AT_FDCWD && startvp == NULL) {
error = EINVAL;
goto out;
}
if (error = pn_get(fname, UIO_USERSPACE, &pn)) {
goto out;
}
if (startfd == AT_FDCWD) {
mutex_enter(&p->p_lock);
startvp = PTOU(p)->u_cdir;
VN_HOLD(startvp);
mutex_exit(&p->p_lock);
}
/*
* Verify permission to put attributes on file
*/
if ((VOP_ACCESS(startvp, VREAD, 0, CRED()) != 0) &&
(VOP_ACCESS(startvp, VWRITE, 0, CRED()) != 0) &&
(VOP_ACCESS(startvp, VEXEC, 0, CRED()) != 0)) {
error = EACCES;
pn_free(&pn);
goto out;
}
if ((startvp->v_vfsp->vfs_flag & VFS_XATTR) != 0) {
error = VOP_LOOKUP(startvp, "", &sdvp, &pn,
LOOKUP_XATTR|CREATE_XATTR_DIR, rootvp, CRED());
} else {
error = EINVAL;
}
pn_free(&pn);
if (error != 0)
goto out;
VN_RELE(startvp);
startvp = sdvp;
}
if ((filemode & (FREAD|FWRITE)) != 0) {
if ((filemode & (FNONBLOCK|FNDELAY)) == (FNONBLOCK|FNDELAY))
filemode &= ~FNDELAY;
error = falloc((vnode_t *)NULL, filemode, &fp, &fd);
if (error == 0) {
#ifdef C2_AUDIT
if (audit_active)
audit_setfsat_path(1);
#endif /* C2_AUDIT */
/*
* Last arg is a don't-care term if
* !(filemode & FCREAT).
*/
error = vn_openat(fname, UIO_USERSPACE, filemode,
(int)(createmode & MODEMASK), &vp, CRCREAT,
u.u_cmask, startvp);
if (startvp != NULL)
VN_RELE(startvp);
if (error == 0) {
#ifdef C2_AUDIT
if (audit_active)
audit_copen(fd, fp, vp);
#endif /* C2_AUDIT */
if ((vp->v_flag & VDUP) == 0) {
fp->f_vnode = vp;
mutex_exit(&fp->f_tlock);
/*
* We must now fill in the slot
* falloc reserved.
*/
setf(fd, fp);
return (fd);
} else {
/*
* Special handling for /dev/fd.
* Give up the file pointer
* and dup the indicated file descriptor
* (in v_rdev). This is ugly, but I've
* seen worse.
*/
unfalloc(fp);
dupfd = getminor(vp->v_rdev);
type = vp->v_type;
mutex_enter(&vp->v_lock);
vp->v_flag &= ~VDUP;
mutex_exit(&vp->v_lock);
VN_RELE(vp);
if (type != VCHR)
return (set_errno(EINVAL));
if ((fp = getf(dupfd)) == NULL) {
setf(fd, NULL);
return (set_errno(EBADF));
}
mutex_enter(&fp->f_tlock);
fp->f_count++;
mutex_exit(&fp->f_tlock);
setf(fd, fp);
releasef(dupfd);
}
return (fd);
} else {
setf(fd, NULL);
unfalloc(fp);
return (set_errno(error));
}
}
} else {
error = EINVAL;
}
out:
if (startvp != NULL)
VN_RELE(startvp);
return (set_errno(error));
}
/*
* Get the "real" XATTR directory associtated with the GFS XATTR directory.
* Note: This does NOT take any additional hold on the returned real_vp,
* because when this lookup succeeds we save the result in xattr_realvp
* and keep that hold until the GFS XATTR directory goes inactive.
*/
static int
xattr_dir_realdir(vnode_t *gfs_dvp, vnode_t **ret_vpp, int flags,
cred_t *cr, caller_context_t *ct)
{
struct pathname pn;
char *nm = "";
xattr_dir_t *xattr_dir;
vnode_t *realvp;
int error;
*ret_vpp = NULL;
/*
* Usually, we've already found the underlying XATTR directory
* during some previous lookup and stored it in xattr_realvp.
*/
mutex_enter(&gfs_dvp->v_lock);
xattr_dir = gfs_dvp->v_data;
realvp = xattr_dir->xattr_realvp;
mutex_exit(&gfs_dvp->v_lock);
if (realvp != NULL) {
*ret_vpp = realvp;
return (0);
}
/*
* Lookup the XATTR dir in the underlying FS, relative to our
* "parent", which is the real object for which this GFS XATTR
* directory was created. Set the LOOKUP_HAVE_SYSATTR_DIR flag
* so that we don't get into an infinite loop with fop_lookup
* calling back to xattr_dir_lookup.
*/
error = pn_get(nm, UIO_SYSSPACE, &pn);
if (error != 0)
return (error);
error = VOP_LOOKUP(gfs_file_parent(gfs_dvp), nm, &realvp, &pn,
flags | LOOKUP_HAVE_SYSATTR_DIR, rootvp, cr, ct, NULL, NULL);
pn_free(&pn);
if (error != 0)
return (error);
/*
* Have the real XATTR directory. Save it -- but first
* check whether we lost a race doing the lookup.
*/
mutex_enter(&gfs_dvp->v_lock);
xattr_dir = gfs_dvp->v_data;
if (xattr_dir->xattr_realvp == NULL) {
/*
* Note that the hold taken by the VOP_LOOKUP above is
* retained from here until xattr_dir_inactive.
*/
xattr_dir->xattr_realvp = realvp;
} else {
/* We lost the race. */
VN_RELE(realvp);
realvp = xattr_dir->xattr_realvp;
}
mutex_exit(&gfs_dvp->v_lock);
*ret_vpp = realvp;
return (0);
}
int
xattr_dir_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp)
{
int error;
vnode_t *pvp, *dvp;
xattr_fid_t *xfidp;
struct pathname pn;
char *nm;
uint16_t orig_len;
*vpp = NULL;
if (fidp->fid_len < XATTR_FIDSZ)
return (EINVAL);
xfidp = (xattr_fid_t *)fidp;
orig_len = fidp->fid_len;
fidp->fid_len = xfidp->parent_len;
error = VFS_VGET(vfsp, &pvp, fidp);
fidp->fid_len = orig_len;
if (error)
return (error);
/*
* Start by getting the GFS sysattr directory. We might need
* to recreate it during the VOP_LOOKUP.
*/
nm = "";
error = pn_get(nm, UIO_SYSSPACE, &pn);
if (error) {
VN_RELE(pvp);
return (EINVAL);
}
error = VOP_LOOKUP(pvp, nm, &dvp, &pn, LOOKUP_XATTR|CREATE_XATTR_DIR,
rootvp, CRED(), NULL, NULL, NULL);
pn_free(&pn);
VN_RELE(pvp);
if (error)
return (error);
if (xfidp->dir_offset == 0) {
/*
* If we were looking for the directory, we're done.
*/
*vpp = dvp;
return (0);
}
if (xfidp->dir_offset > XATTRDIR_NENTS) {
VN_RELE(dvp);
return (EINVAL);
}
nm = xattr_dirents[xfidp->dir_offset - 1].gfse_name;
error = pn_get(nm, UIO_SYSSPACE, &pn);
if (error) {
VN_RELE(dvp);
return (EINVAL);
}
error = VOP_LOOKUP(dvp, nm, vpp, &pn, 0, rootvp, CRED(), NULL,
NULL, NULL);
pn_free(&pn);
VN_RELE(dvp);
return (error);
}
/*
* Get the XATTR dir for some file or directory.
* See vnode.c: fop_lookup()
*
* Note this only gets the GFS XATTR directory. We'll get the
* real XATTR directory later, in xattr_dir_realdir.
*/
int
xattr_dir_lookup(vnode_t *dvp, vnode_t **vpp, int flags, cred_t *cr)
{
int error = 0;
*vpp = NULL;
if (dvp->v_type != VDIR && dvp->v_type != VREG)
return (EINVAL);
mutex_enter(&dvp->v_lock);
/*
* If we're already in sysattr space, don't allow creation
* of another level of sysattrs.
*/
if (dvp->v_flag & V_SYSATTR) {
mutex_exit(&dvp->v_lock);
return (EINVAL);
}
if (dvp->v_xattrdir != NULL) {
*vpp = dvp->v_xattrdir;
VN_HOLD(*vpp);
} else {
ulong_t val;
int xattrs_allowed = dvp->v_vfsp->vfs_flag & VFS_XATTR;
int sysattrs_allowed = 1;
/*
* We have to drop the lock on dvp. gfs_dir_create will
* grab it for a VN_HOLD.
*/
mutex_exit(&dvp->v_lock);
/*
* If dvp allows xattr creation, but not sysattr
* creation, return the real xattr dir vp. We can't
* use the vfs feature mask here because _PC_SATTR_ENABLED
* has vnode-level granularity (e.g. .zfs).
*/
error = VOP_PATHCONF(dvp, _PC_SATTR_ENABLED, &val, cr, NULL);
if (error != 0 || val == 0)
sysattrs_allowed = 0;
if (!xattrs_allowed && !sysattrs_allowed)
return (EINVAL);
if (!sysattrs_allowed) {
struct pathname pn;
char *nm = "";
error = pn_get(nm, UIO_SYSSPACE, &pn);
if (error)
return (error);
error = VOP_LOOKUP(dvp, nm, vpp, &pn,
flags|LOOKUP_HAVE_SYSATTR_DIR, rootvp, cr, NULL,
NULL, NULL);
pn_free(&pn);
return (error);
}
/*
* Note that we act as if we were given CREATE_XATTR_DIR,
* but only for creation of the GFS directory.
*/
*vpp = gfs_dir_create(
sizeof (xattr_dir_t), dvp, xattr_dir_ops, xattr_dirents,
xattrdir_do_ino, MAXNAMELEN, NULL, xattr_lookup_cb);
mutex_enter(&dvp->v_lock);
if (dvp->v_xattrdir != NULL) {
/*
* We lost the race to create the xattr dir.
* Destroy this one, use the winner. We can't
* just call VN_RELE(*vpp), because the vnode
* is only partially initialized.
*/
gfs_dir_t *dp = (*vpp)->v_data;
ASSERT((*vpp)->v_count == 1);
vn_free(*vpp);
mutex_destroy(&dp->gfsd_lock);
kmem_free(dp->gfsd_static,
dp->gfsd_nstatic * sizeof (gfs_dirent_t));
kmem_free(dp, dp->gfsd_file.gfs_size);
/*
* There is an implied VN_HOLD(dvp) here. We should
* be doing a VN_RELE(dvp) to clean up the reference
* from *vpp, and then a VN_HOLD(dvp) for the new
* reference. Instead, we just leave the count alone.
*/
*vpp = dvp->v_xattrdir;
VN_HOLD(*vpp);
} else {
(*vpp)->v_flag |= (V_XATTRDIR|V_SYSATTR);
dvp->v_xattrdir = *vpp;
}
}
mutex_exit(&dvp->v_lock);
return (error);
}
/*
* Search a pathname.
* This is a very central and rather complicated routine.
*
* The pathname is pointed to by ni_cnd.cn_nameptr and is of length
* ni_pathlen. The starting directory is taken from ni_startdir. The
* pathname is descended until done, or a symbolic link is encountered.
* If the path is completed the flag ISLASTCN is set in ni_cnd.cn_flags.
* If a symbolic link need interpretation is encountered, the flag ISSYMLINK
* is set in ni_cnd.cn_flags.
*
* The flag argument is LOOKUP, CREATE, RENAME, or DELETE depending on
* whether the name is to be looked up, created, renamed, or deleted.
* When CREATE, RENAME, or DELETE is specified, information usable in
* creating, renaming, or deleting a directory entry may be calculated.
* If flag has LOCKPARENT or'ed into it, the parent directory is returned
* locked. If flag has WANTPARENT or'ed into it, the parent directory is
* returned unlocked. Otherwise the parent directory is not returned. If
* the target of the pathname exists and LOCKLEAF is or'ed into the flag
* the target is returned locked, otherwise it is returned unlocked.
* When creating or renaming and LOCKPARENT is specified, the target may not
* be ".". When deleting and LOCKPARENT is specified, the target may be ".".
*
* Overall outline of lookup:
*
* dirloop:
* identify next component of name at ndp->ni_ptr
* handle degenerate case where name is null string
* if .. and crossing mount points and on mounted filesys, find parent
* call VOP_LOOKUP routine for next component name
* directory vnode returned in ni_dvp, unlocked unless LOCKPARENT set
* component vnode returned in ni_vp (if it exists), locked.
* if result vnode is mounted on and crossing mount points,
* find mounted on vnode
* if more components of name, do next level at dirloop
* return the answer in ni_vp, locked if LOCKLEAF set
* if LOCKPARENT set, return locked parent in ni_dvp
* if WANTPARENT set, return unlocked parent in ni_dvp
*/
int
lookup(struct nameidata *ndp)
{
char *cp; /* pointer into pathname argument */
struct vnode *dp = 0; /* the directory we are searching */
struct vnode *tdp; /* saved dp */
struct mount *mp; /* mount table entry */
int docache; /* == 0 do not cache last component */
int wantparent; /* 1 => wantparent or lockparent flag */
int rdonly; /* lookup read-only flag bit */
int error = 0;
int dpunlocked = 0; /* dp has already been unlocked */
int slashes;
struct componentname *cnp = &ndp->ni_cnd;
struct proc *p = cnp->cn_proc;
/*
* Setup: break out flag bits into variables.
*/
wantparent = cnp->cn_flags & (LOCKPARENT | WANTPARENT);
docache = (cnp->cn_flags & NOCACHE) ^ NOCACHE;
if (cnp->cn_nameiop == DELETE ||
(wantparent && cnp->cn_nameiop != CREATE))
docache = 0;
rdonly = cnp->cn_flags & RDONLY;
ndp->ni_dvp = NULL;
cnp->cn_flags &= ~ISSYMLINK;
dp = ndp->ni_startdir;
ndp->ni_startdir = NULLVP;
vn_lock(dp, LK_EXCLUSIVE | LK_RETRY, p);
/*
* If we have a leading string of slashes, remove them, and just make
* sure the current node is a directory.
*/
cp = cnp->cn_nameptr;
if (*cp == '/') {
do {
cp++;
} while (*cp == '/');
ndp->ni_pathlen -= cp - cnp->cn_nameptr;
cnp->cn_nameptr = cp;
if (dp->v_type != VDIR) {
error = ENOTDIR;
goto bad;
}
/*
* If we've exhausted the path name, then just return the
* current node. If the caller requested the parent node (i.e.
* it's a CREATE, DELETE, or RENAME), and we don't have one
* (because this is the root directory), then we must fail.
*/
if (cnp->cn_nameptr[0] == '\0') {
if (ndp->ni_dvp == NULL && wantparent) {
error = EISDIR;
goto bad;
}
ndp->ni_vp = dp;
cnp->cn_flags |= ISLASTCN;
goto terminal;
}
}
dirloop:
/*
* Search a new directory.
*
* The cn_hash value is for use by vfs_cache.
* The last component of the filename is left accessible via
* cnp->cn_nameptr for callers that need the name. Callers needing
* the name set the SAVENAME flag. When done, they assume
* responsibility for freeing the pathname buffer.
*/
cp = NULL;
cnp->cn_consume = 0;
cnp->cn_hash = hash32_stre(cnp->cn_nameptr, '/', &cp, HASHINIT);
cnp->cn_namelen = cp - cnp->cn_nameptr;
if (cnp->cn_namelen > NAME_MAX) {
error = ENAMETOOLONG;
goto bad;
}
#ifdef NAMEI_DIAGNOSTIC
{ char c = *cp;
*cp = '\0';
printf("{%s}: ", cnp->cn_nameptr);
*cp = c; }
#endif
ndp->ni_pathlen -= cnp->cn_namelen;
ndp->ni_next = cp;
/*
* If this component is followed by a slash, then move the pointer to
* the next component forward, and remember that this component must be
* a directory.
*/
if (*cp == '/') {
do {
cp++;
} while (*cp == '/');
slashes = cp - ndp->ni_next;
ndp->ni_pathlen -= slashes;
ndp->ni_next = cp;
cnp->cn_flags |= REQUIREDIR;
} else {
slashes = 0;
cnp->cn_flags &= ~REQUIREDIR;
}
/*
* We do special processing on the last component, whether or not it's
* a directory. Cache all intervening lookups, but not the final one.
*/
if (*cp == '\0') {
if (docache)
cnp->cn_flags |= MAKEENTRY;
else
cnp->cn_flags &= ~MAKEENTRY;
cnp->cn_flags |= ISLASTCN;
} else {
cnp->cn_flags |= MAKEENTRY;
cnp->cn_flags &= ~ISLASTCN;
}
if (cnp->cn_namelen == 2 &&
cnp->cn_nameptr[1] == '.' && cnp->cn_nameptr[0] == '.')
cnp->cn_flags |= ISDOTDOT;
else
cnp->cn_flags &= ~ISDOTDOT;
/*
* Handle "..": two special cases.
* 1. If at root directory (e.g. after chroot)
* or at absolute root directory
* then ignore it so can't get out.
* 2. If this vnode is the root of a mounted
* filesystem, then replace it with the
* vnode which was mounted on so we take the
* .. in the other file system.
*/
if (cnp->cn_flags & ISDOTDOT) {
for (;;) {
if (dp == ndp->ni_rootdir || dp == rootvnode) {
ndp->ni_dvp = dp;
ndp->ni_vp = dp;
vref(dp);
goto nextname;
}
if ((dp->v_flag & VROOT) == 0 ||
(cnp->cn_flags & NOCROSSMOUNT))
break;
tdp = dp;
dp = dp->v_mount->mnt_vnodecovered;
vput(tdp);
vref(dp);
vn_lock(dp, LK_EXCLUSIVE | LK_RETRY, p);
}
}
/*
* We now have a segment name to search for, and a directory to search.
*/
ndp->ni_dvp = dp;
ndp->ni_vp = NULL;
cnp->cn_flags &= ~PDIRUNLOCK;
if ((error = VOP_LOOKUP(dp, &ndp->ni_vp, cnp)) != 0) {
#ifdef DIAGNOSTIC
if (ndp->ni_vp != NULL)
panic("leaf should be empty");
#endif
#ifdef NAMEI_DIAGNOSTIC
printf("not found\n");
#endif
if (error != EJUSTRETURN)
goto bad;
/*
* If this was not the last component, or there were trailing
* slashes, then the name must exist.
*/
if (cnp->cn_flags & REQUIREDIR) {
error = ENOENT;
goto bad;
}
/*
* If creating and at end of pathname, then can consider
* allowing file to be created.
*/
if (rdonly || (ndp->ni_dvp->v_mount->mnt_flag & MNT_RDONLY)) {
error = EROFS;
goto bad;
}
/*
* We return with ni_vp NULL to indicate that the entry
* doesn't currently exist, leaving a pointer to the
* (possibly locked) directory inode in ndp->ni_dvp.
*/
if (cnp->cn_flags & SAVESTART) {
ndp->ni_startdir = ndp->ni_dvp;
vref(ndp->ni_startdir);
}
return (0);
}
#ifdef NAMEI_DIAGNOSTIC
printf("found\n");
#endif
/*
* Take into account any additional components consumed by the
* underlying filesystem. This will include any trailing slashes after
* the last component consumed.
*/
if (cnp->cn_consume > 0) {
if (cnp->cn_consume >= slashes) {
cnp->cn_flags &= ~REQUIREDIR;
}
ndp->ni_pathlen -= cnp->cn_consume - slashes;
ndp->ni_next += cnp->cn_consume - slashes;
cnp->cn_consume = 0;
if (ndp->ni_next[0] == '\0')
cnp->cn_flags |= ISLASTCN;
}
dp = ndp->ni_vp;
/*
* Check to see if the vnode has been mounted on;
* if so find the root of the mounted file system.
*/
while (dp->v_type == VDIR && (mp = dp->v_mountedhere) &&
(cnp->cn_flags & NOCROSSMOUNT) == 0) {
if (vfs_busy(mp, VB_READ|VB_WAIT))
continue;
VOP_UNLOCK(dp, 0, p);
error = VFS_ROOT(mp, &tdp);
vfs_unbusy(mp);
if (error) {
dpunlocked = 1;
goto bad2;
}
vrele(dp);
ndp->ni_vp = dp = tdp;
}
/*
* Check for symbolic link. Back up over any slashes that we skipped,
* as we will need them again.
*/
if ((dp->v_type == VLNK) && (cnp->cn_flags & (FOLLOW|REQUIREDIR))) {
ndp->ni_pathlen += slashes;
ndp->ni_next -= slashes;
cnp->cn_flags |= ISSYMLINK;
return (0);
}
/*
* Check for directory, if the component was followed by a series of
* slashes.
*/
if ((dp->v_type != VDIR) && (cnp->cn_flags & REQUIREDIR)) {
error = ENOTDIR;
goto bad2;
}
nextname:
/*
* Not a symbolic link. If this was not the last component, then
* continue at the next component, else return.
*/
if (!(cnp->cn_flags & ISLASTCN)) {
cnp->cn_nameptr = ndp->ni_next;
vrele(ndp->ni_dvp);
goto dirloop;
}
terminal:
/*
* Check for read-only file systems.
*/
if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) {
/*
* Disallow directory write attempts on read-only
* file systems.
*/
if (rdonly || (dp->v_mount->mnt_flag & MNT_RDONLY) ||
(wantparent &&
(ndp->ni_dvp->v_mount->mnt_flag & MNT_RDONLY))) {
error = EROFS;
goto bad2;
}
}
if (ndp->ni_dvp != NULL) {
if (cnp->cn_flags & SAVESTART) {
ndp->ni_startdir = ndp->ni_dvp;
vref(ndp->ni_startdir);
}
if (!wantparent)
vrele(ndp->ni_dvp);
}
if ((cnp->cn_flags & LOCKLEAF) == 0)
VOP_UNLOCK(dp, 0, p);
return (0);
bad2:
if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN) &&
((cnp->cn_flags & PDIRUNLOCK) == 0))
VOP_UNLOCK(ndp->ni_dvp, 0, p);
vrele(ndp->ni_dvp);
bad:
if (dpunlocked)
vrele(dp);
else
vput(dp);
ndp->ni_vp = NULL;
return (error);
}
/*
* Reacquire a path name component.
*/
int
relookup(struct vnode *dvp, struct vnode **vpp, struct componentname *cnp)
{
struct proc *p = cnp->cn_proc;
struct vnode *dp = 0; /* the directory we are searching */
int wantparent; /* 1 => wantparent or lockparent flag */
int rdonly; /* lookup read-only flag bit */
int error = 0;
#ifdef NAMEI_DIAGNOSTIC
u_int32_t newhash; /* DEBUG: check name hash */
char *cp; /* DEBUG: check name ptr/len */
#endif
/*
* Setup: break out flag bits into variables.
*/
wantparent = cnp->cn_flags & (LOCKPARENT|WANTPARENT);
rdonly = cnp->cn_flags & RDONLY;
cnp->cn_flags &= ~ISSYMLINK;
dp = dvp;
vn_lock(dp, LK_EXCLUSIVE | LK_RETRY, p);
/* dirloop: */
/*
* Search a new directory.
*
* The cn_hash value is for use by vfs_cache.
* The last component of the filename is left accessible via
* cnp->cn_nameptr for callers that need the name. Callers needing
* the name set the SAVENAME flag. When done, they assume
* responsibility for freeing the pathname buffer.
*/
#ifdef NAMEI_DIAGNOSTIC
cp = NULL;
newhash = hash32_stre(cnp->cn_nameptr, '/', &cp, HASHINIT);
if (newhash != cnp->cn_hash)
panic("relookup: bad hash");
if (cnp->cn_namelen != cp - cnp->cn_nameptr)
panic ("relookup: bad len");
if (*cp != 0)
panic("relookup: not last component");
printf("{%s}: ", cnp->cn_nameptr);
#endif
/*
* Check for degenerate name (e.g. / or "")
* which is a way of talking about a directory,
* e.g. like "/." or ".".
*/
if (cnp->cn_nameptr[0] == '\0')
panic("relookup: null name");
if (cnp->cn_flags & ISDOTDOT)
panic ("relookup: lookup on dot-dot");
/*
* We now have a segment name to search for, and a directory to search.
*/
if ((error = VOP_LOOKUP(dp, vpp, cnp)) != 0) {
#ifdef DIAGNOSTIC
if (*vpp != NULL)
panic("leaf should be empty");
#endif
if (error != EJUSTRETURN)
goto bad;
/*
* If creating and at end of pathname, then can consider
* allowing file to be created.
*/
if (rdonly || (dvp->v_mount->mnt_flag & MNT_RDONLY)) {
error = EROFS;
goto bad;
}
/* ASSERT(dvp == ndp->ni_startdir) */
if (cnp->cn_flags & SAVESTART)
vref(dvp);
/*
* We return with ni_vp NULL to indicate that the entry
* doesn't currently exist, leaving a pointer to the
* (possibly locked) directory inode in ndp->ni_dvp.
*/
return (0);
}
dp = *vpp;
#ifdef DIAGNOSTIC
/*
* Check for symbolic link
*/
if (dp->v_type == VLNK && (cnp->cn_flags & FOLLOW))
panic ("relookup: symlink found.");
#endif
/*
* Check for read-only file systems.
*/
if (cnp->cn_nameiop == DELETE || cnp->cn_nameiop == RENAME) {
/*
* Disallow directory write attempts on read-only
* file systems.
*/
if (rdonly || (dp->v_mount->mnt_flag & MNT_RDONLY) ||
(wantparent &&
(dvp->v_mount->mnt_flag & MNT_RDONLY))) {
error = EROFS;
goto bad2;
}
}
/* ASSERT(dvp == ndp->ni_startdir) */
if (cnp->cn_flags & SAVESTART)
vref(dvp);
if (!wantparent)
vrele(dvp);
if ((cnp->cn_flags & LOCKLEAF) == 0)
VOP_UNLOCK(dp, 0, p);
return (0);
bad2:
if ((cnp->cn_flags & LOCKPARENT) && (cnp->cn_flags & ISLASTCN))
VOP_UNLOCK(dvp, 0, p);
vrele(dvp);
bad:
vput(dp);
*vpp = NULL;
return (error);
}
/* Find parent vnode of *lvpp, return in *uvpp */
int
vfs_getcwd_scandir(struct vnode **lvpp, struct vnode **uvpp, char **bpp,
char *bufp, struct proc *p)
{
int eofflag, tries, dirbuflen, len, reclen, error = 0;
off_t off;
struct uio uio;
struct iovec iov;
char *dirbuf = NULL;
ino_t fileno;
struct vattr va;
struct vnode *uvp = NULL;
struct vnode *lvp = *lvpp;
struct componentname cn;
tries = 0;
/*
* If we want the filename, get some info we need while the
* current directory is still locked.
*/
if (bufp != NULL) {
error = VOP_GETATTR(lvp, &va, p->p_ucred, p);
if (error) {
vput(lvp);
*lvpp = NULL;
*uvpp = NULL;
return (error);
}
}
cn.cn_nameiop = LOOKUP;
cn.cn_flags = ISLASTCN | ISDOTDOT | RDONLY;
cn.cn_proc = p;
cn.cn_cred = p->p_ucred;
cn.cn_pnbuf = NULL;
cn.cn_nameptr = "..";
cn.cn_namelen = 2;
cn.cn_consume = 0;
/* Get parent vnode using lookup of '..' */
error = VOP_LOOKUP(lvp, uvpp, &cn);
if (error) {
vput(lvp);
*lvpp = NULL;
*uvpp = NULL;
return (error);
}
uvp = *uvpp;
/* If we don't care about the pathname, we're done */
if (bufp == NULL) {
vrele(lvp);
*lvpp = NULL;
return (0);
}
fileno = va.va_fileid;
dirbuflen = DIRBLKSIZ;
if (dirbuflen < va.va_blocksize)
dirbuflen = va.va_blocksize;
dirbuf = malloc(dirbuflen, M_TEMP, M_WAITOK);
off = 0;
do {
char *cpos;
struct dirent *dp;
iov.iov_base = dirbuf;
iov.iov_len = dirbuflen;
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = off;
uio.uio_resid = dirbuflen;
uio.uio_segflg = UIO_SYSSPACE;
uio.uio_rw = UIO_READ;
uio.uio_procp = p;
eofflag = 0;
/* Call VOP_READDIR of parent */
error = VOP_READDIR(uvp, &uio, p->p_ucred, &eofflag);
off = uio.uio_offset;
/* Try again if NFS tosses its cookies */
if (error == EINVAL && tries < 3) {
tries++;
off = 0;
continue;
} else if (error) {
goto out; /* Old userland getcwd() behaviour */
}
cpos = dirbuf;
tries = 0;
/* Scan directory page looking for matching vnode */
for (len = (dirbuflen - uio.uio_resid); len > 0;
len -= reclen) {
dp = (struct dirent *)cpos;
reclen = dp->d_reclen;
/* Check for malformed directory */
if (reclen < DIRENT_RECSIZE(1)) {
error = EINVAL;
goto out;
}
if (dp->d_fileno == fileno) {
char *bp = *bpp;
bp -= dp->d_namlen;
if (bp <= bufp) {
error = ERANGE;
goto out;
}
memmove(bp, dp->d_name, dp->d_namlen);
error = 0;
*bpp = bp;
goto out;
}
cpos += reclen;
}
} while (!eofflag);
error = ENOENT;
out:
vrele(lvp);
*lvpp = NULL;
free(dirbuf, M_TEMP);
return (error);
}
