* subsequent patch.
*/
-#include <linux/config.h>
#include <linux/slab.h>
+#include <linux/smp_lock.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_idmap.h>
+#include <linux/kthread.h>
+#include <linux/module.h>
#include <linux/workqueue.h>
#include <linux/bitops.h>
-#define OPENOWNER_POOL_SIZE 8
-
-static spinlock_t state_spinlock = SPIN_LOCK_UNLOCKED;
+#include "nfs4_fs.h"
+#include "callback.h"
+#include "delegation.h"
+#include "internal.h"
-nfs4_stateid zero_stateid;
+#define OPENOWNER_POOL_SIZE 8
-#if 0
-nfs4_stateid one_stateid =
- { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
-#endif
+const nfs4_stateid zero_stateid;
static LIST_HEAD(nfs4_clientid_list);
-static void nfs4_recover_state(void *);
-extern void nfs4_renew_state(void *);
-
-void
-init_nfsv4_state(struct nfs_server *server)
-{
- server->nfs4_state = NULL;
- INIT_LIST_HEAD(&server->nfs4_siblings);
-}
-
-void
-destroy_nfsv4_state(struct nfs_server *server)
+static int nfs4_init_client(struct nfs_client *clp, struct rpc_cred *cred)
{
- if (server->mnt_path) {
- kfree(server->mnt_path);
- server->mnt_path = NULL;
- }
- if (server->nfs4_state) {
- nfs4_put_client(server->nfs4_state);
- server->nfs4_state = NULL;
- }
-}
-
-/*
- * nfs4_get_client(): returns an empty client structure
- * nfs4_put_client(): drops reference to client structure
- *
- * Since these are allocated/deallocated very rarely, we don't
- * bother putting them in a slab cache...
- */
-static struct nfs4_client *
-nfs4_alloc_client(struct in_addr *addr)
-{
- struct nfs4_client *clp;
-
- if ((clp = kmalloc(sizeof(*clp), GFP_KERNEL))) {
- memset(clp, 0, sizeof(*clp));
- memcpy(&clp->cl_addr, addr, sizeof(clp->cl_addr));
- init_rwsem(&clp->cl_sem);
- INIT_LIST_HEAD(&clp->cl_state_owners);
- INIT_LIST_HEAD(&clp->cl_unused);
- spin_lock_init(&clp->cl_lock);
- atomic_set(&clp->cl_count, 1);
- INIT_WORK(&clp->cl_recoverd, nfs4_recover_state, clp);
- INIT_WORK(&clp->cl_renewd, nfs4_renew_state, clp);
- INIT_LIST_HEAD(&clp->cl_superblocks);
- init_waitqueue_head(&clp->cl_waitq);
- rpc_init_wait_queue(&clp->cl_rpcwaitq, "NFS4 client");
- clp->cl_state = 1 << NFS4CLNT_NEW;
- }
- return clp;
-}
-
-static void
-nfs4_free_client(struct nfs4_client *clp)
-{
- struct nfs4_state_owner *sp;
-
- while (!list_empty(&clp->cl_unused)) {
- sp = list_entry(clp->cl_unused.next,
- struct nfs4_state_owner,
- so_list);
- list_del(&sp->so_list);
- kfree(sp);
- }
- BUG_ON(!list_empty(&clp->cl_state_owners));
- if (clp->cl_cred)
- put_rpccred(clp->cl_cred);
- nfs_idmap_delete(clp);
- if (clp->cl_rpcclient)
- rpc_shutdown_client(clp->cl_rpcclient);
- kfree(clp);
-}
-
-struct nfs4_client *
-nfs4_get_client(struct in_addr *addr)
-{
- struct nfs4_client *new, *clp = NULL;
-
- new = nfs4_alloc_client(addr);
- spin_lock(&state_spinlock);
- list_for_each_entry(clp, &nfs4_clientid_list, cl_servers) {
- if (memcmp(&clp->cl_addr, addr, sizeof(clp->cl_addr)) == 0)
- goto found;
- }
- if (new)
- list_add(&new->cl_servers, &nfs4_clientid_list);
- spin_unlock(&state_spinlock);
- return new;
-found:
- atomic_inc(&clp->cl_count);
- spin_unlock(&state_spinlock);
- if (new)
- nfs4_free_client(new);
- return clp;
-}
-
-void
-nfs4_put_client(struct nfs4_client *clp)
-{
- if (!atomic_dec_and_lock(&clp->cl_count, &state_spinlock))
- return;
- list_del(&clp->cl_servers);
- spin_unlock(&state_spinlock);
- BUG_ON(!list_empty(&clp->cl_superblocks));
- wake_up_all(&clp->cl_waitq);
- rpc_wake_up(&clp->cl_rpcwaitq);
- nfs4_kill_renewd(clp);
- nfs4_free_client(clp);
+ int status = nfs4_proc_setclientid(clp, NFS4_CALLBACK,
+ nfs_callback_tcpport, cred);
+ if (status == 0)
+ status = nfs4_proc_setclientid_confirm(clp, cred);
+ if (status == 0)
+ nfs4_schedule_state_renewal(clp);
+ return status;
}
u32
-nfs4_alloc_lockowner_id(struct nfs4_client *clp)
+nfs4_alloc_lockowner_id(struct nfs_client *clp)
{
return clp->cl_lockowner_id ++;
}
static struct nfs4_state_owner *
-nfs4_client_grab_unused(struct nfs4_client *clp, struct rpc_cred *cred)
+nfs4_client_grab_unused(struct nfs_client *clp, struct rpc_cred *cred)
{
struct nfs4_state_owner *sp = NULL;
atomic_inc(&sp->so_count);
sp->so_cred = cred;
list_move(&sp->so_list, &clp->cl_state_owners);
- sp->so_generation = clp->cl_generation;
clp->cl_nunused--;
}
return sp;
}
+struct rpc_cred *nfs4_get_renew_cred(struct nfs_client *clp)
+{
+ struct nfs4_state_owner *sp;
+ struct rpc_cred *cred = NULL;
+
+ list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
+ if (list_empty(&sp->so_states))
+ continue;
+ cred = get_rpccred(sp->so_cred);
+ break;
+ }
+ return cred;
+}
+
+struct rpc_cred *nfs4_get_setclientid_cred(struct nfs_client *clp)
+{
+ struct nfs4_state_owner *sp;
+
+ if (!list_empty(&clp->cl_state_owners)) {
+ sp = list_entry(clp->cl_state_owners.next,
+ struct nfs4_state_owner, so_list);
+ return get_rpccred(sp->so_cred);
+ }
+ return NULL;
+}
+
static struct nfs4_state_owner *
-nfs4_find_state_owner(struct nfs4_client *clp, struct rpc_cred *cred)
+nfs4_find_state_owner(struct nfs_client *clp, struct rpc_cred *cred)
{
struct nfs4_state_owner *sp, *res = NULL;
{
struct nfs4_state_owner *sp;
- sp = kmalloc(sizeof(*sp),GFP_KERNEL);
+ sp = kzalloc(sizeof(*sp),GFP_KERNEL);
if (!sp)
return NULL;
- init_MUTEX(&sp->so_sema);
- sp->so_seqid = 0; /* arbitrary */
+ spin_lock_init(&sp->so_lock);
INIT_LIST_HEAD(&sp->so_states);
+ INIT_LIST_HEAD(&sp->so_delegations);
+ rpc_init_wait_queue(&sp->so_sequence.wait, "Seqid_waitqueue");
+ sp->so_seqid.sequence = &sp->so_sequence;
+ spin_lock_init(&sp->so_sequence.lock);
+ INIT_LIST_HEAD(&sp->so_sequence.list);
atomic_set(&sp->so_count, 1);
return sp;
}
-static void
-nfs4_unhash_state_owner(struct nfs4_state_owner *sp)
+void
+nfs4_drop_state_owner(struct nfs4_state_owner *sp)
{
- struct nfs4_client *clp = sp->so_client;
+ struct nfs_client *clp = sp->so_client;
spin_lock(&clp->cl_lock);
list_del_init(&sp->so_list);
spin_unlock(&clp->cl_lock);
}
-struct nfs4_state_owner *
-nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
+/*
+ * Note: must be called with clp->cl_sem held in order to prevent races
+ * with reboot recovery!
+ */
+struct nfs4_state_owner *nfs4_get_state_owner(struct nfs_server *server, struct rpc_cred *cred)
{
- struct nfs4_client *clp = server->nfs4_state;
+ struct nfs_client *clp = server->nfs_client;
struct nfs4_state_owner *sp, *new;
get_rpccred(cred);
new->so_client = clp;
new->so_id = nfs4_alloc_lockowner_id(clp);
new->so_cred = cred;
- new->so_generation = clp->cl_generation;
sp = new;
new = NULL;
}
spin_unlock(&clp->cl_lock);
- if (new)
- kfree(new);
- if (sp) {
- if (!test_bit(NFS4CLNT_OK, &clp->cl_state))
- nfs4_wait_clnt_recover(server->client, clp);
- } else
- put_rpccred(cred);
- return sp;
+ kfree(new);
+ if (sp != NULL)
+ return sp;
+ put_rpccred(cred);
+ return NULL;
}
-void
-nfs4_put_state_owner(struct nfs4_state_owner *sp)
+/*
+ * Must be called with clp->cl_sem held in order to avoid races
+ * with state recovery...
+ */
+void nfs4_put_state_owner(struct nfs4_state_owner *sp)
{
- struct nfs4_client *clp = sp->so_client;
+ struct nfs_client *clp = sp->so_client;
struct rpc_cred *cred = sp->so_cred;
if (!atomic_dec_and_lock(&sp->so_count, &clp->cl_lock))
{
struct nfs4_state *state;
- state = kmalloc(sizeof(*state), GFP_KERNEL);
+ state = kzalloc(sizeof(*state), GFP_KERNEL);
if (!state)
return NULL;
- state->state = 0;
- state->nreaders = 0;
- state->nwriters = 0;
- state->flags = 0;
- memset(state->stateid.data, 0, sizeof(state->stateid.data));
atomic_set(&state->count, 1);
INIT_LIST_HEAD(&state->lock_states);
- init_MUTEX(&state->lock_sema);
- rwlock_init(&state->state_lock);
+ spin_lock_init(&state->state_lock);
return state;
}
-static struct nfs4_state *
-__nfs4_find_state(struct inode *inode, struct rpc_cred *cred, mode_t mode)
+void
+nfs4_state_set_mode_locked(struct nfs4_state *state, mode_t mode)
{
- struct nfs_inode *nfsi = NFS_I(inode);
- struct nfs4_state *state;
-
- mode &= (FMODE_READ|FMODE_WRITE);
- list_for_each_entry(state, &nfsi->open_states, inode_states) {
- if (state->owner->so_cred != cred)
- continue;
- if ((mode & FMODE_READ) != 0 && state->nreaders == 0)
- continue;
- if ((mode & FMODE_WRITE) != 0 && state->nwriters == 0)
- continue;
- if ((state->state & mode) != mode)
- continue;
- /* Add the state to the head of the inode's list */
- list_move(&state->inode_states, &nfsi->open_states);
- atomic_inc(&state->count);
- if (mode & FMODE_READ)
- state->nreaders++;
+ if (state->state == mode)
+ return;
+ /* NB! List reordering - see the reclaim code for why. */
+ if ((mode & FMODE_WRITE) != (state->state & FMODE_WRITE)) {
if (mode & FMODE_WRITE)
- state->nwriters++;
- return state;
+ list_move(&state->open_states, &state->owner->so_states);
+ else
+ list_move_tail(&state->open_states, &state->owner->so_states);
}
- return NULL;
+ if (mode == 0)
+ list_del_init(&state->inode_states);
+ state->state = mode;
}
static struct nfs4_state *
list_for_each_entry(state, &nfsi->open_states, inode_states) {
/* Is this in the process of being freed? */
- if (state->nreaders == 0 && state->nwriters == 0)
+ if (state->state == 0)
continue;
if (state->owner == owner) {
- /* Add the state to the head of the inode's list */
- list_move(&state->inode_states, &nfsi->open_states);
atomic_inc(&state->count);
return state;
}
return NULL;
}
-struct nfs4_state *
-nfs4_find_state(struct inode *inode, struct rpc_cred *cred, mode_t mode)
-{
- struct nfs4_state *state;
-
- spin_lock(&inode->i_lock);
- state = __nfs4_find_state(inode, cred, mode);
- spin_unlock(&inode->i_lock);
- return state;
-}
-
static void
nfs4_free_open_state(struct nfs4_state *state)
{
if (state)
goto out;
new = nfs4_alloc_open_state();
+ spin_lock(&owner->so_lock);
spin_lock(&inode->i_lock);
state = __nfs4_find_state_byowner(inode, owner);
if (state == NULL && new != NULL) {
state = new;
- /* Caller *must* be holding owner->so_sem */
- list_add(&state->open_states, &owner->so_states);
state->owner = owner;
atomic_inc(&owner->so_count);
list_add(&state->inode_states, &nfsi->open_states);
- state->inode = inode;
+ state->inode = igrab(inode);
spin_unlock(&inode->i_lock);
+ /* Note: The reclaim code dictates that we add stateless
+ * and read-only stateids to the end of the list */
+ list_add_tail(&state->open_states, &owner->so_states);
+ spin_unlock(&owner->so_lock);
} else {
spin_unlock(&inode->i_lock);
+ spin_unlock(&owner->so_lock);
if (new)
nfs4_free_open_state(new);
}
return state;
}
-static void
-__nfs4_put_open_state(struct nfs4_state *state)
+/*
+ * Beware! Caller must be holding exactly one
+ * reference to clp->cl_sem!
+ */
+void nfs4_put_open_state(struct nfs4_state *state)
{
struct inode *inode = state->inode;
struct nfs4_state_owner *owner = state->owner;
- int status = 0;
- if (!atomic_dec_and_lock(&state->count, &inode->i_lock)) {
- up(&owner->so_sema);
+ if (!atomic_dec_and_lock(&state->count, &owner->so_lock))
return;
- }
+ spin_lock(&inode->i_lock);
if (!list_empty(&state->inode_states))
list_del(&state->inode_states);
- spin_unlock(&inode->i_lock);
list_del(&state->open_states);
- if (state->state != 0) {
- do {
- status = nfs4_do_close(inode, state);
- if (!status)
- break;
- up(&owner->so_sema);
- status = nfs4_handle_error(NFS_SERVER(inode), status);
- down(&owner->so_sema);
- } while (!status);
- }
- up(&owner->so_sema);
+ spin_unlock(&inode->i_lock);
+ spin_unlock(&owner->so_lock);
+ iput(inode);
nfs4_free_open_state(state);
nfs4_put_state_owner(owner);
}
-void
-nfs4_put_open_state(struct nfs4_state *state)
-{
- down(&state->owner->so_sema);
- __nfs4_put_open_state(state);
-}
-
-void
-nfs4_close_state(struct nfs4_state *state, mode_t mode)
+/*
+ * Close the current file.
+ */
+void nfs4_close_state(struct nfs4_state *state, mode_t mode)
{
struct inode *inode = state->inode;
struct nfs4_state_owner *owner = state->owner;
- int newstate;
- int status = 0;
+ int oldstate, newstate = 0;
- down(&owner->so_sema);
+ atomic_inc(&owner->so_count);
/* Protect against nfs4_find_state() */
+ spin_lock(&owner->so_lock);
spin_lock(&inode->i_lock);
- if (mode & FMODE_READ)
- state->nreaders--;
- if (mode & FMODE_WRITE)
- state->nwriters--;
- if (state->nwriters == 0 && state->nreaders == 0)
- list_del_init(&state->inode_states);
- spin_unlock(&inode->i_lock);
- do {
- newstate = 0;
- if (state->state == 0)
- break;
- if (state->nreaders)
- newstate |= FMODE_READ;
- if (state->nwriters)
- newstate |= FMODE_WRITE;
- if (state->state == newstate)
+ switch (mode & (FMODE_READ | FMODE_WRITE)) {
+ case FMODE_READ:
+ state->n_rdonly--;
break;
- if (newstate != 0)
- status = nfs4_do_downgrade(inode, state, newstate);
- else
- status = nfs4_do_close(inode, state);
- if (!status) {
- state->state = newstate;
+ case FMODE_WRITE:
+ state->n_wronly--;
break;
- }
- up(&owner->so_sema);
- status = nfs4_handle_error(NFS_SERVER(inode), status);
- down(&owner->so_sema);
- } while (!status);
- __nfs4_put_open_state(state);
+ case FMODE_READ|FMODE_WRITE:
+ state->n_rdwr--;
+ }
+ oldstate = newstate = state->state;
+ if (state->n_rdwr == 0) {
+ if (state->n_rdonly == 0)
+ newstate &= ~FMODE_READ;
+ if (state->n_wronly == 0)
+ newstate &= ~FMODE_WRITE;
+ }
+ if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
+ nfs4_state_set_mode_locked(state, newstate);
+ oldstate = newstate;
+ }
+ spin_unlock(&inode->i_lock);
+ spin_unlock(&owner->so_lock);
+
+ if (oldstate != newstate && nfs4_do_close(inode, state) == 0)
+ return;
+ nfs4_put_open_state(state);
+ nfs4_put_state_owner(owner);
}
/*
return NULL;
}
-struct nfs4_lock_state *
-nfs4_find_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
-{
- struct nfs4_lock_state *lsp;
- read_lock(&state->state_lock);
- lsp = __nfs4_find_lock_state(state, fl_owner);
- read_unlock(&state->state_lock);
- return lsp;
-}
-
/*
* Return a compatible lock_state. If no initialized lock_state structure
* exists, return an uninitialized one.
*
- * The caller must be holding state->lock_sema
*/
-struct nfs4_lock_state *
-nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
+static struct nfs4_lock_state *nfs4_alloc_lock_state(struct nfs4_state *state, fl_owner_t fl_owner)
{
struct nfs4_lock_state *lsp;
- struct nfs4_client *clp = state->owner->so_client;
+ struct nfs_client *clp = state->owner->so_client;
- lsp = kmalloc(sizeof(*lsp), GFP_KERNEL);
+ lsp = kzalloc(sizeof(*lsp), GFP_KERNEL);
if (lsp == NULL)
return NULL;
- lsp->ls_seqid = 0; /* arbitrary */
- lsp->ls_id = -1;
- memset(lsp->ls_stateid.data, 0, sizeof(lsp->ls_stateid.data));
+ lsp->ls_seqid.sequence = &state->owner->so_sequence;
atomic_set(&lsp->ls_count, 1);
lsp->ls_owner = fl_owner;
- lsp->ls_parent = state;
- INIT_LIST_HEAD(&lsp->ls_locks);
spin_lock(&clp->cl_lock);
lsp->ls_id = nfs4_alloc_lockowner_id(clp);
spin_unlock(&clp->cl_lock);
+ INIT_LIST_HEAD(&lsp->ls_locks);
return lsp;
}
/*
- * Byte-range lock aware utility to initialize the stateid of read/write
- * requests.
+ * Return a compatible lock_state. If no initialized lock_state structure
+ * exists, return an uninitialized one.
+ *
+ * The caller must be holding clp->cl_sem
*/
-void
-nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
-{
- if (test_bit(LK_STATE_IN_USE, &state->flags)) {
- struct nfs4_lock_state *lsp;
-
- lsp = nfs4_find_lock_state(state, fl_owner);
- if (lsp) {
- memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
- nfs4_put_lock_state(lsp);
- return;
+static struct nfs4_lock_state *nfs4_get_lock_state(struct nfs4_state *state, fl_owner_t owner)
+{
+ struct nfs4_lock_state *lsp, *new = NULL;
+
+ for(;;) {
+ spin_lock(&state->state_lock);
+ lsp = __nfs4_find_lock_state(state, owner);
+ if (lsp != NULL)
+ break;
+ if (new != NULL) {
+ new->ls_state = state;
+ list_add(&new->ls_locks, &state->lock_states);
+ set_bit(LK_STATE_IN_USE, &state->flags);
+ lsp = new;
+ new = NULL;
+ break;
}
+ spin_unlock(&state->state_lock);
+ new = nfs4_alloc_lock_state(state, owner);
+ if (new == NULL)
+ return NULL;
}
- memcpy(dst, &state->stateid, sizeof(*dst));
+ spin_unlock(&state->state_lock);
+ kfree(new);
+ return lsp;
}
/*
-* Called with state->lock_sema held.
-*/
-void
-nfs4_increment_lock_seqid(int status, struct nfs4_lock_state *lsp)
+ * Release reference to lock_state, and free it if we see that
+ * it is no longer in use
+ */
+void nfs4_put_lock_state(struct nfs4_lock_state *lsp)
{
- if (status == NFS_OK || seqid_mutating_err(-status))
- lsp->ls_seqid++;
+ struct nfs4_state *state;
+
+ if (lsp == NULL)
+ return;
+ state = lsp->ls_state;
+ if (!atomic_dec_and_lock(&lsp->ls_count, &state->state_lock))
+ return;
+ list_del(&lsp->ls_locks);
+ if (list_empty(&state->lock_states))
+ clear_bit(LK_STATE_IN_USE, &state->flags);
+ spin_unlock(&state->state_lock);
+ kfree(lsp);
}
-/*
-* Check to see if the request lock (type FL_UNLK) effects the fl lock.
-*
-* fl and request must have the same posix owner
-*
-* return:
-* 0 -> fl not effected by request
-* 1 -> fl consumed by request
-*/
+static void nfs4_fl_copy_lock(struct file_lock *dst, struct file_lock *src)
+{
+ struct nfs4_lock_state *lsp = src->fl_u.nfs4_fl.owner;
-static int
-nfs4_check_unlock(struct file_lock *fl, struct file_lock *request)
+ dst->fl_u.nfs4_fl.owner = lsp;
+ atomic_inc(&lsp->ls_count);
+}
+
+static void nfs4_fl_release_lock(struct file_lock *fl)
{
- if (fl->fl_start >= request->fl_start && fl->fl_end <= request->fl_end)
- return 1;
+ nfs4_put_lock_state(fl->fl_u.nfs4_fl.owner);
+}
+
+static struct file_lock_operations nfs4_fl_lock_ops = {
+ .fl_copy_lock = nfs4_fl_copy_lock,
+ .fl_release_private = nfs4_fl_release_lock,
+};
+
+int nfs4_set_lock_state(struct nfs4_state *state, struct file_lock *fl)
+{
+ struct nfs4_lock_state *lsp;
+
+ if (fl->fl_ops != NULL)
+ return 0;
+ lsp = nfs4_get_lock_state(state, fl->fl_owner);
+ if (lsp == NULL)
+ return -ENOMEM;
+ fl->fl_u.nfs4_fl.owner = lsp;
+ fl->fl_ops = &nfs4_fl_lock_ops;
return 0;
}
/*
- * Post an initialized lock_state on the state->lock_states list.
+ * Byte-range lock aware utility to initialize the stateid of read/write
+ * requests.
*/
-void
-nfs4_notify_setlk(struct inode *inode, struct file_lock *request, struct nfs4_lock_state *lsp)
+void nfs4_copy_stateid(nfs4_stateid *dst, struct nfs4_state *state, fl_owner_t fl_owner)
{
- struct nfs4_state *state = lsp->ls_parent;
+ struct nfs4_lock_state *lsp;
- if (!list_empty(&lsp->ls_locks))
+ memcpy(dst, &state->stateid, sizeof(*dst));
+ if (test_bit(LK_STATE_IN_USE, &state->flags) == 0)
return;
- write_lock(&state->state_lock);
- list_add(&lsp->ls_locks, &state->lock_states);
- set_bit(LK_STATE_IN_USE, &state->flags);
- write_unlock(&state->state_lock);
+
+ spin_lock(&state->state_lock);
+ lsp = __nfs4_find_lock_state(state, fl_owner);
+ if (lsp != NULL && (lsp->ls_flags & NFS_LOCK_INITIALIZED) != 0)
+ memcpy(dst, &lsp->ls_stateid, sizeof(*dst));
+ spin_unlock(&state->state_lock);
+ nfs4_put_lock_state(lsp);
}
-/*
- * to decide to 'reap' lock state:
- * 1) search i_flock for file_locks with fl.lock_state = to ls.
- * 2) determine if unlock will consume found lock.
- * if so, reap
- *
- * else, don't reap.
- *
- */
-void
-nfs4_notify_unlck(struct inode *inode, struct file_lock *request, struct nfs4_lock_state *lsp)
+struct nfs_seqid *nfs_alloc_seqid(struct nfs_seqid_counter *counter)
{
- struct nfs4_state *state = lsp->ls_parent;
- struct file_lock *fl;
+ struct rpc_sequence *sequence = counter->sequence;
+ struct nfs_seqid *new;
- for (fl = inode->i_flock; fl != NULL; fl = fl->fl_next) {
- if (!(fl->fl_flags & FL_POSIX))
- continue;
- if (fl->fl_owner != lsp->ls_owner)
- continue;
- /* Exit if we find at least one lock which is not consumed */
- if (nfs4_check_unlock(fl,request) == 0)
- return;
+ new = kmalloc(sizeof(*new), GFP_KERNEL);
+ if (new != NULL) {
+ new->sequence = counter;
+ spin_lock(&sequence->lock);
+ list_add_tail(&new->list, &sequence->list);
+ spin_unlock(&sequence->lock);
}
+ return new;
+}
- write_lock(&state->state_lock);
- list_del_init(&lsp->ls_locks);
- if (list_empty(&state->lock_states))
- clear_bit(LK_STATE_IN_USE, &state->flags);
- write_unlock(&state->state_lock);
+void nfs_free_seqid(struct nfs_seqid *seqid)
+{
+ struct rpc_sequence *sequence = seqid->sequence->sequence;
+
+ spin_lock(&sequence->lock);
+ list_del(&seqid->list);
+ spin_unlock(&sequence->lock);
+ rpc_wake_up(&sequence->wait);
+ kfree(seqid);
}
/*
- * Release reference to lock_state, and free it if we see that
- * it is no longer in use
+ * Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
+ * failed with a seqid incrementing error -
+ * see comments nfs_fs.h:seqid_mutating_error()
*/
-void
-nfs4_put_lock_state(struct nfs4_lock_state *lsp)
+static inline void nfs_increment_seqid(int status, struct nfs_seqid *seqid)
{
- if (!atomic_dec_and_test(&lsp->ls_count))
- return;
- if (!list_empty(&lsp->ls_locks))
- return;
- kfree(lsp);
+ switch (status) {
+ case 0:
+ break;
+ case -NFS4ERR_BAD_SEQID:
+ case -NFS4ERR_STALE_CLIENTID:
+ case -NFS4ERR_STALE_STATEID:
+ case -NFS4ERR_BAD_STATEID:
+ case -NFS4ERR_BADXDR:
+ case -NFS4ERR_RESOURCE:
+ case -NFS4ERR_NOFILEHANDLE:
+ /* Non-seqid mutating errors */
+ return;
+ };
+ /*
+ * Note: no locking needed as we are guaranteed to be first
+ * on the sequence list
+ */
+ seqid->sequence->counter++;
}
-/*
-* Called with sp->so_sema held.
-*
-* Increment the seqid if the OPEN/OPEN_DOWNGRADE/CLOSE succeeded, or
-* failed with a seqid incrementing error -
-* see comments nfs_fs.h:seqid_mutating_error()
-*/
-void
-nfs4_increment_seqid(int status, struct nfs4_state_owner *sp)
+void nfs_increment_open_seqid(int status, struct nfs_seqid *seqid)
{
- if (status == NFS_OK || seqid_mutating_err(-status))
- sp->so_seqid++;
- /* If the server returns BAD_SEQID, unhash state_owner here */
- if (status == -NFS4ERR_BAD_SEQID)
- nfs4_unhash_state_owner(sp);
+ if (status == -NFS4ERR_BAD_SEQID) {
+ struct nfs4_state_owner *sp = container_of(seqid->sequence,
+ struct nfs4_state_owner, so_seqid);
+ nfs4_drop_state_owner(sp);
+ }
+ return nfs_increment_seqid(status, seqid);
}
-static int reclaimer(void *);
-struct reclaimer_args {
- struct nfs4_client *clp;
- struct completion complete;
-};
-
/*
- * State recovery routine
+ * Increment the seqid if the LOCK/LOCKU succeeded, or
+ * failed with a seqid incrementing error -
+ * see comments nfs_fs.h:seqid_mutating_error()
*/
-void
-nfs4_recover_state(void *data)
+void nfs_increment_lock_seqid(int status, struct nfs_seqid *seqid)
{
- struct nfs4_client *clp = (struct nfs4_client *)data;
- struct reclaimer_args args = {
- .clp = clp,
- };
- might_sleep();
+ return nfs_increment_seqid(status, seqid);
+}
- init_completion(&args.complete);
+int nfs_wait_on_sequence(struct nfs_seqid *seqid, struct rpc_task *task)
+{
+ struct rpc_sequence *sequence = seqid->sequence->sequence;
+ int status = 0;
- down_read(&clp->cl_sem);
- if (test_and_set_bit(NFS4CLNT_SETUP_STATE, &clp->cl_state))
- goto out_failed;
- if (kernel_thread(reclaimer, &args, CLONE_KERNEL) < 0)
- goto out_failed_clear;
- wait_for_completion(&args.complete);
- return;
-out_failed_clear:
+ if (sequence->list.next == &seqid->list)
+ goto out;
+ spin_lock(&sequence->lock);
+ if (sequence->list.next != &seqid->list) {
+ rpc_sleep_on(&sequence->wait, task, NULL, NULL);
+ status = -EAGAIN;
+ }
+ spin_unlock(&sequence->lock);
+out:
+ return status;
+}
+
+static int reclaimer(void *);
+
+static inline void nfs4_clear_recover_bit(struct nfs_client *clp)
+{
smp_mb__before_clear_bit();
- clear_bit(NFS4CLNT_SETUP_STATE, &clp->cl_state);
+ clear_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state);
smp_mb__after_clear_bit();
- wake_up_all(&clp->cl_waitq);
+ wake_up_bit(&clp->cl_state, NFS4CLNT_STATE_RECOVER);
rpc_wake_up(&clp->cl_rpcwaitq);
-out_failed:
- up_read(&clp->cl_sem);
+}
+
+/*
+ * State recovery routine
+ */
+static void nfs4_recover_state(struct nfs_client *clp)
+{
+ struct task_struct *task;
+
+ __module_get(THIS_MODULE);
+ atomic_inc(&clp->cl_count);
+ task = kthread_run(reclaimer, clp, "%u.%u.%u.%u-reclaim",
+ NIPQUAD(clp->cl_addr.sin_addr));
+ if (!IS_ERR(task))
+ return;
+ nfs4_clear_recover_bit(clp);
+ nfs_put_client(clp);
+ module_put(THIS_MODULE);
}
/*
* Schedule a state recovery attempt
*/
-void
-nfs4_schedule_state_recovery(struct nfs4_client *clp)
+void nfs4_schedule_state_recovery(struct nfs_client *clp)
{
if (!clp)
return;
- smp_mb__before_clear_bit();
- clear_bit(NFS4CLNT_OK, &clp->cl_state);
- smp_mb__after_clear_bit();
- schedule_work(&clp->cl_recoverd);
+ if (test_and_set_bit(NFS4CLNT_STATE_RECOVER, &clp->cl_state) == 0)
+ nfs4_recover_state(clp);
}
-static int
-nfs4_reclaim_open_state(struct nfs4_state_owner *sp)
+static int nfs4_reclaim_locks(struct nfs4_state_recovery_ops *ops, struct nfs4_state *state)
{
- struct nfs4_state *state;
+ struct inode *inode = state->inode;
+ struct file_lock *fl;
int status = 0;
- list_for_each_entry(state, &sp->so_states, open_states) {
- status = nfs4_open_reclaim(sp, state);
+ for (fl = inode->i_flock; fl != 0; fl = fl->fl_next) {
+ if (!(fl->fl_flags & (FL_POSIX|FL_FLOCK)))
+ continue;
+ if (((struct nfs_open_context *)fl->fl_file->private_data)->state != state)
+ continue;
+ status = ops->recover_lock(state, fl);
if (status >= 0)
continue;
switch (status) {
case -NFS4ERR_EXPIRED:
case -NFS4ERR_NO_GRACE:
case -NFS4ERR_RECLAIM_BAD:
+ case -NFS4ERR_RECLAIM_CONFLICT:
+ /* kill_proc(fl->fl_pid, SIGLOST, 1); */
+ break;
+ case -NFS4ERR_STALE_CLIENTID:
+ goto out_err;
+ }
+ }
+ return 0;
+out_err:
+ return status;
+}
+
+static int nfs4_reclaim_open_state(struct nfs4_state_recovery_ops *ops, struct nfs4_state_owner *sp)
+{
+ struct nfs4_state *state;
+ struct nfs4_lock_state *lock;
+ int status = 0;
+
+ /* Note: we rely on the sp->so_states list being ordered
+ * so that we always reclaim open(O_RDWR) and/or open(O_WRITE)
+ * states first.
+ * This is needed to ensure that the server won't give us any
+ * read delegations that we have to return if, say, we are
+ * recovering after a network partition or a reboot from a
+ * server that doesn't support a grace period.
+ */
+ list_for_each_entry(state, &sp->so_states, open_states) {
+ if (state->state == 0)
+ continue;
+ status = ops->recover_open(sp, state);
+ if (status >= 0) {
+ status = nfs4_reclaim_locks(ops, state);
+ if (status < 0)
+ goto out_err;
+ list_for_each_entry(lock, &state->lock_states, ls_locks) {
+ if (!(lock->ls_flags & NFS_LOCK_INITIALIZED))
+ printk("%s: Lock reclaim failed!\n",
+ __FUNCTION__);
+ }
+ continue;
+ }
+ switch (status) {
+ default:
+ printk(KERN_ERR "%s: unhandled error %d. Zeroing state\n",
+ __FUNCTION__, status);
+ case -ENOENT:
+ case -NFS4ERR_RECLAIM_BAD:
case -NFS4ERR_RECLAIM_CONFLICT:
/*
* Open state on this file cannot be recovered
/* Mark the file as being 'closed' */
state->state = 0;
break;
+ case -NFS4ERR_EXPIRED:
+ case -NFS4ERR_NO_GRACE:
case -NFS4ERR_STALE_CLIENTID:
goto out_err;
}
return status;
}
-static int
-reclaimer(void *ptr)
+static void nfs4_state_mark_reclaim(struct nfs_client *clp)
{
- struct reclaimer_args *args = (struct reclaimer_args *)ptr;
- struct nfs4_client *clp = args->clp;
struct nfs4_state_owner *sp;
- int generation;
- int status;
+ struct nfs4_state *state;
+ struct nfs4_lock_state *lock;
- daemonize("%u.%u.%u.%u-reclaim", NIPQUAD(clp->cl_addr));
- allow_signal(SIGKILL);
+ /* Reset all sequence ids to zero */
+ list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
+ sp->so_seqid.counter = 0;
+ sp->so_seqid.flags = 0;
+ spin_lock(&sp->so_lock);
+ list_for_each_entry(state, &sp->so_states, open_states) {
+ list_for_each_entry(lock, &state->lock_states, ls_locks) {
+ lock->ls_seqid.counter = 0;
+ lock->ls_seqid.flags = 0;
+ lock->ls_flags &= ~NFS_LOCK_INITIALIZED;
+ }
+ }
+ spin_unlock(&sp->so_lock);
+ }
+}
+
+static int reclaimer(void *ptr)
+{
+ struct nfs_client *clp = ptr;
+ struct nfs4_state_owner *sp;
+ struct nfs4_state_recovery_ops *ops;
+ struct rpc_cred *cred;
+ int status = 0;
- complete(&args->complete);
+ allow_signal(SIGKILL);
+ /* Ensure exclusive access to NFSv4 state */
+ lock_kernel();
+ down_write(&clp->cl_sem);
/* Are there any NFS mounts out there? */
if (list_empty(&clp->cl_superblocks))
goto out;
- if (!test_bit(NFS4CLNT_NEW, &clp->cl_state)) {
- status = nfs4_proc_renew(clp);
- if (status == 0) {
- set_bit(NFS4CLNT_OK, &clp->cl_state);
- goto out;
+restart_loop:
+ ops = &nfs4_network_partition_recovery_ops;
+ /* Are there any open files on this volume? */
+ cred = nfs4_get_renew_cred(clp);
+ if (cred != NULL) {
+ /* Yes there are: try to renew the old lease */
+ status = nfs4_proc_renew(clp, cred);
+ switch (status) {
+ case 0:
+ case -NFS4ERR_CB_PATH_DOWN:
+ put_rpccred(cred);
+ goto out;
+ case -NFS4ERR_STALE_CLIENTID:
+ case -NFS4ERR_LEASE_MOVED:
+ ops = &nfs4_reboot_recovery_ops;
}
+ } else {
+ /* "reboot" to ensure we clear all state on the server */
+ clp->cl_boot_time = CURRENT_TIME;
+ cred = nfs4_get_setclientid_cred(clp);
+ }
+ /* We're going to have to re-establish a clientid */
+ nfs4_state_mark_reclaim(clp);
+ status = -ENOENT;
+ if (cred != NULL) {
+ status = nfs4_init_client(clp, cred);
+ put_rpccred(cred);
}
- status = nfs4_proc_setclientid(clp, 0, 0);
- if (status)
- goto out_error;
- status = nfs4_proc_setclientid_confirm(clp);
if (status)
goto out_error;
- generation = ++(clp->cl_generation);
- clear_bit(NFS4CLNT_NEW, &clp->cl_state);
- set_bit(NFS4CLNT_OK, &clp->cl_state);
- up_read(&clp->cl_sem);
- nfs4_schedule_state_renewal(clp);
-restart_loop:
- spin_lock(&clp->cl_lock);
+ /* Mark all delegations for reclaim */
+ nfs_delegation_mark_reclaim(clp);
+ /* Note: list is protected by exclusive lock on cl->cl_sem */
list_for_each_entry(sp, &clp->cl_state_owners, so_list) {
- if (sp->so_generation - generation >= 0)
- continue;
- atomic_inc(&sp->so_count);
- spin_unlock(&clp->cl_lock);
- down(&sp->so_sema);
- if (sp->so_generation - generation < 0) {
- smp_rmb();
- sp->so_generation = clp->cl_generation;
- status = nfs4_reclaim_open_state(sp);
- }
- up(&sp->so_sema);
- nfs4_put_state_owner(sp);
+ status = nfs4_reclaim_open_state(ops, sp);
if (status < 0) {
+ if (status == -NFS4ERR_NO_GRACE) {
+ ops = &nfs4_network_partition_recovery_ops;
+ status = nfs4_reclaim_open_state(ops, sp);
+ }
if (status == -NFS4ERR_STALE_CLIENTID)
- nfs4_schedule_state_recovery(clp);
- goto out;
+ goto restart_loop;
+ if (status == -NFS4ERR_EXPIRED)
+ goto restart_loop;
}
- goto restart_loop;
}
- spin_unlock(&clp->cl_lock);
+ nfs_delegation_reap_unclaimed(clp);
out:
- smp_mb__before_clear_bit();
- clear_bit(NFS4CLNT_SETUP_STATE, &clp->cl_state);
- smp_mb__after_clear_bit();
- wake_up_all(&clp->cl_waitq);
- rpc_wake_up(&clp->cl_rpcwaitq);
+ up_write(&clp->cl_sem);
+ unlock_kernel();
+ if (status == -NFS4ERR_CB_PATH_DOWN)
+ nfs_handle_cb_pathdown(clp);
+ nfs4_clear_recover_bit(clp);
+ nfs_put_client(clp);
+ module_put_and_exit(0);
return 0;
out_error:
- printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %u.%u.%u.%u\n",
- NIPQUAD(clp->cl_addr.s_addr));
- up_read(&clp->cl_sem);
+ printk(KERN_WARNING "Error: state recovery failed on NFSv4 server %u.%u.%u.%u with error %d\n",
+ NIPQUAD(clp->cl_addr.sin_addr), -status);
+ set_bit(NFS4CLNT_LEASE_EXPIRED, &clp->cl_state);
goto out;
}
git://git.onelab.eu / linux-2.6.git / blobdiff