struct vnode *
vn_initialize(
struct inode *inode)
{
struct vnode *vp = LINVFS_GET_VP(inode);
XFS_STATS_INC(vn_active);
XFS_STATS_INC(vn_alloc);
vp->v_flag = VMODIFIED;
spinlock_init(&vp->v_lock, "v_lock");
spin_lock(&vnumber_lock);
if (!++vn_generation) /* v_number shouldn't be zero */
vn_generation++;
vp->v_number = vn_generation;
spin_unlock(&vnumber_lock);
ASSERT(VN_CACHED(vp) == 0);
/* Initialize the first behavior and the behavior chain head. */
vn_bhv_head_init(VN_BHV_HEAD(vp), "vnode");
#ifdef XFS_VNODE_TRACE
vp->v_trace = ktrace_alloc(VNODE_TRACE_SIZE, KM_SLEEP);
#endif /* XFS_VNODE_TRACE */
vn_trace_exit(vp, "vn_initialize", (inst_t *)__return_address);
return vp;
}
/*
* Get a reference on a vnode.
*/
vnode_t *
vn_get(
struct vnode *vp,
vmap_t *vmap)
{
struct inode *inode;
XFS_STATS_INC(vn_get);
inode = LINVFS_GET_IP(vp);
if (inode->i_state & I_FREEING)
return NULL;
inode = VFS_GET_INODE(vmap->v_vfsp, vmap->v_ino, IGET_NOALLOC);
if (!inode) /* Inode not present */
return NULL;
/* We do not want to create new inodes via vn_get,
* returning NULL here is OK.
*/
if (inode->i_state & I_NEW) {
vn_mark_bad(vp);
unlock_new_inode(inode);
iput(inode);
return NULL;
}
vn_trace_exit(vp, "vn_get", (inst_t *)__return_address);
return vp;
}
/*
* Call VOP_INACTIVE on last reference.
*/
void
vn_rele(
struct vnode *vp)
{
int vcnt;
int cache;
XFS_STATS_INC(vn_rele);
VN_LOCK(vp);
vn_trace_entry(vp, "vn_rele", (inst_t *)__return_address);
vcnt = vn_count(vp);
/*
* Since we always get called from put_inode we know
* that i_count won't be decremented after we
* return.
*/
if (!vcnt) {
/*
* As soon as we turn this on, noone can find us in vn_get
* until we turn off VINACT or VRECLM
*/
vp->v_flag |= VINACT;
VN_UNLOCK(vp, 0);
/*
* Do not make the VOP_INACTIVE call if there
* are no behaviors attached to the vnode to call.
*/
if (vp->v_fbhv)
VOP_INACTIVE(vp, NULL, cache);
VN_LOCK(vp);
if (vp->v_flag & VWAIT)
sv_broadcast(vptosync(vp));
vp->v_flag &= ~(VINACT|VWAIT|VRECLM|VMODIFIED);
}
VN_UNLOCK(vp, 0);
vn_trace_exit(vp, "vn_rele", (inst_t *)__return_address);
}
static int
xfs_vn_allocate(xfs_mount_t *mp, xfs_inode_t *ip, struct xfs_vnode **vpp)
{
struct vnode *vp;
struct xfs_vnode *vdata;
int error;
/* Use zone allocator here? */
vdata = kmem_zalloc(sizeof(*vdata), KM_SLEEP);
error = getnewvnode("xfs", XVFSTOMNT(XFS_MTOVFS(mp)),
&xfs_vnops, &vp);
if (error) {
kmem_free(vdata, sizeof(*vdata));
return (error);
}
vp->v_vnlock->lk_flags |= LK_CANRECURSE;
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, curthread);
error = insmntque(vp, XVFSTOMNT(XFS_MTOVFS(mp)));
if (error != 0) {
kmem_free(vdata, sizeof(*vdata));
return (error);
}
vp->v_data = (void *)vdata;
vdata->v_number= 0;
vdata->v_inode = ip;
vdata->v_vfsp = XFS_MTOVFS(mp);
vdata->v_vnode = vp;
vn_bhv_head_init(VN_BHV_HEAD(vdata), "vnode");
#ifdef CONFIG_XFS_VNODE_TRACING
vp->v_trace = ktrace_alloc(VNODE_TRACE_SIZE, KM_SLEEP);
#endif /* CONFIG_XFS_VNODE_TRACING */
vn_trace_exit(vp, "vn_initialize", (inst_t *)__return_address);
if (error == 0)
*vpp = vdata;
return (error);
}
/*
* Finish the removal of a vnode.
*/
void
vn_remove(
struct vnode *vp)
{
vmap_t vmap;
/* Make sure we don't do this to the same vnode twice */
if (!(vp->v_fbhv))
return;
XFS_STATS_INC(vn_remove);
vn_trace_exit(vp, "vn_remove", (inst_t *)__return_address);
/*
* After the following purge the vnode
* will no longer exist.
*/
VMAP(vp, vmap);
vn_purge(vp, &vmap);
}
/*
* Get a reference on a vnode.
*/
vnode_t *
vn_get(
struct vnode *vp,
vmap_t *vmap)
{
struct inode *inode;
XFS_STATS_INC(vn_get);
inode = LINVFS_GET_IP(vp);
if (inode->i_state & I_FREEING)
return NULL;
inode = ilookup(vmap->v_vfsp->vfs_super, vmap->v_ino);
if (!inode) /* Inode not present */
return NULL;
vn_trace_exit(vp, "vn_get", (inst_t *)__return_address);
return vp;
}
/*
* Look up an inode by number in the given file system.
* The inode is looked up in the hash table for the file system
* represented by the mount point parameter mp. Each bucket of
* the hash table is guarded by an individual semaphore.
*
* If the inode is found in the hash table, its corresponding vnode
* is obtained with a call to vn_get(). This call takes care of
* coordination with the reclamation of the inode and vnode. Note
* that the vmap structure is filled in while holding the hash lock.
* This gives us the state of the inode/vnode when we found it and
* is used for coordination in vn_get().
*
* If it is not in core, read it in from the file system's device and
* add the inode into the hash table.
*
* The inode is locked according to the value of the lock_flags parameter.
* This flag parameter indicates how and if the inode's IO lock and inode lock
* should be taken.
*
* mp -- the mount point structure for the current file system. It points
* to the inode hash table.
* tp -- a pointer to the current transaction if there is one. This is
* simply passed through to the xfs_iread() call.
* ino -- the number of the inode desired. This is the unique identifier
* within the file system for the inode being requested.
* lock_flags -- flags indicating how to lock the inode. See the comment
* for xfs_ilock() for a list of valid values.
* bno -- the block number starting the buffer containing the inode,
* if known (as by bulkstat), else 0.
*/
int
xfs_iget(
xfs_mount_t *mp,
xfs_trans_t *tp,
xfs_ino_t ino,
uint flags,
uint lock_flags,
xfs_inode_t **ipp,
xfs_daddr_t bno)
{
xfs_ihash_t *ih;
xfs_inode_t *ip;
xfs_inode_t *iq;
xfs_vnode_t *vp;
ulong version;
int error;
/* REFERENCED */
int newnode;
xfs_chash_t *ch;
xfs_chashlist_t *chl, *chlnew;
vmap_t vmap;
SPLDECL(s);
XFS_STATS_INC(xs_ig_attempts);
ih = XFS_IHASH(mp, ino);
again:
read_lock(&ih->ih_lock);
for (ip = ih->ih_next; ip != NULL; ip = ip->i_next) {
if (ip->i_ino == ino) {
vp = XFS_ITOV(ip);
VMAP(vp, vmap);
/*
* Inode cache hit: if ip is not at the front of
* its hash chain, move it there now.
* Do this with the lock held for update, but
* do statistics after releasing the lock.
*/
if (ip->i_prevp != &ih->ih_next
&& rwlock_trypromote(&ih->ih_lock)) {
if ((iq = ip->i_next)) {
iq->i_prevp = ip->i_prevp;
}
*ip->i_prevp = iq;
iq = ih->ih_next;
iq->i_prevp = &ip->i_next;
ip->i_next = iq;
ip->i_prevp = &ih->ih_next;
ih->ih_next = ip;
write_unlock(&ih->ih_lock);
} else {
read_unlock(&ih->ih_lock);
}
XFS_STATS_INC(xs_ig_found);
/*
* Get a reference to the vnode/inode.
* vn_get() takes care of coordination with
* the file system inode release and reclaim
* functions. If it returns NULL, the inode
* has been reclaimed so just start the search
* over again. We probably won't find it,
* but we could be racing with another cpu
* looking for the same inode so we have to at
* least look.
*/
if (!(vp = vn_get(vp, &vmap))) {
XFS_STATS_INC(xs_ig_frecycle);
goto again;
}
if (lock_flags != 0) {
ip->i_flags &= ~XFS_IRECLAIM;
xfs_ilock(ip, lock_flags);
}
newnode = (ip->i_d.di_mode == 0);
if (newnode) {
xfs_iocore_inode_reinit(ip);
}
ip->i_flags &= ~XFS_ISTALE;
vn_trace_exit(vp, "xfs_iget.found",
(inst_t *)__return_address);
goto return_ip;
}
}
/*
* Inode cache miss: save the hash chain version stamp and unlock
* the chain, so we don't deadlock in vn_alloc.
*/
XFS_STATS_INC(xs_ig_missed);
version = ih->ih_version;
read_unlock(&ih->ih_lock);
/*
* Read the disk inode attributes into a new inode structure and get
* a new vnode for it. This should also initialize i_ino and i_mount.
*/
error = xfs_iread(mp, tp, ino, &ip, bno);
if (error) {
return error;
}
error = xfs_vn_allocate(mp, ip, &vp);
if (error) {
return error;
}
vn_trace_exit(vp, "xfs_iget.alloc", (inst_t *)__return_address);
xfs_inode_lock_init(ip, vp);
xfs_iocore_inode_init(ip);
if (lock_flags != 0) {
xfs_ilock(ip, lock_flags);
}
/*
* Put ip on its hash chain, unless someone else hashed a duplicate
* after we released the hash lock.
*/
write_lock(&ih->ih_lock);
if (ih->ih_version != version) {
for (iq = ih->ih_next; iq != NULL; iq = iq->i_next) {
if (iq->i_ino == ino) {
write_unlock(&ih->ih_lock);
xfs_idestroy(ip);
XFS_STATS_INC(xs_ig_dup);
goto again;
}
}
}
/*
* These values _must_ be set before releasing ihlock!
*/
ip->i_hash = ih;
if ((iq = ih->ih_next)) {
iq->i_prevp = &ip->i_next;
}
ip->i_next = iq;
ip->i_prevp = &ih->ih_next;
ih->ih_next = ip;
ip->i_udquot = ip->i_gdquot = NULL;
ih->ih_version++;
write_unlock(&ih->ih_lock);
/*
* put ip on its cluster's hash chain
*/
ASSERT(ip->i_chash == NULL && ip->i_cprev == NULL &&
ip->i_cnext == NULL);
chlnew = NULL;
ch = XFS_CHASH(mp, ip->i_blkno);
chlredo:
s = mutex_spinlock(&ch->ch_lock);
for (chl = ch->ch_list; chl != NULL; chl = chl->chl_next) {
if (chl->chl_blkno == ip->i_blkno) {
/* insert this inode into the doubly-linked list
* where chl points */
if ((iq = chl->chl_ip)) {
ip->i_cprev = iq->i_cprev;
iq->i_cprev->i_cnext = ip;
iq->i_cprev = ip;
ip->i_cnext = iq;
} else {
ip->i_cnext = ip;
ip->i_cprev = ip;
}
chl->chl_ip = ip;
ip->i_chash = chl;
break;
}
}
/* no hash list found for this block; add a new hash list */
if (chl == NULL) {
if (chlnew == NULL) {
mutex_spinunlock(&ch->ch_lock, s);
ASSERT(xfs_chashlist_zone != NULL);
chlnew = (xfs_chashlist_t *)
kmem_zone_alloc(xfs_chashlist_zone,
KM_SLEEP);
ASSERT(chlnew != NULL);
goto chlredo;
} else {
ip->i_cnext = ip;
ip->i_cprev = ip;
ip->i_chash = chlnew;
chlnew->chl_ip = ip;
chlnew->chl_blkno = ip->i_blkno;
chlnew->chl_next = ch->ch_list;
ch->ch_list = chlnew;
chlnew = NULL;
}
} else {
if (chlnew != NULL) {
kmem_zone_free(xfs_chashlist_zone, chlnew);
}
}
mutex_spinunlock(&ch->ch_lock, s);
/*
* Link ip to its mount and thread it on the mount's inode list.
*/
XFS_MOUNT_ILOCK(mp);
if ((iq = mp->m_inodes)) {
ASSERT(iq->i_mprev->i_mnext == iq);
ip->i_mprev = iq->i_mprev;
iq->i_mprev->i_mnext = ip;
iq->i_mprev = ip;
ip->i_mnext = iq;
} else {
ip->i_mnext = ip;
ip->i_mprev = ip;
}
mp->m_inodes = ip;
XFS_MOUNT_IUNLOCK(mp);
newnode = 1;
return_ip:
ASSERT(ip->i_df.if_ext_max ==
XFS_IFORK_DSIZE(ip) / sizeof(xfs_bmbt_rec_t));
ASSERT(((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != 0) ==
((ip->i_iocore.io_flags & XFS_IOCORE_RT) != 0));
*ipp = ip;
/*
* If we have a real type for an on-disk inode, we can set ops(&unlock)
* now. If it's a new inode being created, xfs_ialloc will handle it.
*/
XVFS_INIT_VNODE(XFS_MTOVFS(mp), vp, XFS_ITOBHV(ip), 1);
return 0;
}
STATIC int
linvfs_fill_super(
struct super_block *sb,
void *data,
int silent)
{
vnode_t *rootvp;
struct vfs *vfsp = vfs_allocate();
struct xfs_mount_args *args = xfs_args_allocate(sb);
struct kstatfs statvfs;
int error;
vfsp->vfs_super = sb;
LINVFS_SET_VFS(sb, vfsp);
if (sb->s_flags & MS_RDONLY)
vfsp->vfs_flag |= VFS_RDONLY;
bhv_insert_all_vfsops(vfsp);
VFS_PARSEARGS(vfsp, (char *)data, args, 0, error);
if (error) {
bhv_remove_all_vfsops(vfsp, 1);
goto fail_vfsop;
}
sb_min_blocksize(sb, BBSIZE);
sb->s_export_op = &linvfs_export_ops;
sb->s_qcop = &linvfs_qops;
sb->s_op = &linvfs_sops;
VFS_MOUNT(vfsp, args, NULL, error);
if (error) {
bhv_remove_all_vfsops(vfsp, 1);
goto fail_vfsop;
}
VFS_STATVFS(vfsp, &statvfs, NULL, error);
if (error)
goto fail_unmount;
sb->s_dirt = 1;
sb->s_magic = statvfs.f_type;
sb->s_blocksize = statvfs.f_bsize;
sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1;
sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
set_posix_acl_flag(sb);
VFS_ROOT(vfsp, &rootvp, error);
if (error)
goto fail_unmount;
sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp));
if (!sb->s_root)
goto fail_vnrele;
if (is_bad_inode(sb->s_root->d_inode))
goto fail_vnrele;
if (linvfs_start_syncd(vfsp))
goto fail_vnrele;
vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address);
kmem_free(args, sizeof(*args));
return 0;
fail_vnrele:
if (sb->s_root) {
dput(sb->s_root);
sb->s_root = NULL;
} else {
VN_RELE(rootvp);
}
fail_unmount:
VFS_UNMOUNT(vfsp, 0, NULL, error);
fail_vfsop:
vfs_deallocate(vfsp);
kmem_free(args, sizeof(*args));
return -error;
}
