VServer 1.9.2 (patch-2.6.8.1-vs1.9.2.diff)
[linux-2.6.git] / fs / nfsd / nfs4state.c
index 848c485..0ff4425 100644 (file)
@@ -51,6 +51,9 @@
 #define NFSDDBG_FACILITY                NFSDDBG_PROC
 
 /* Globals */
+static time_t lease_time = 90;     /* default lease time */
+static time_t old_lease_time = 90; /* past incarnation lease time */
+static u32 nfs4_reclaim_init = 0;
 time_t boot_time;
 static time_t grace_end = 0;
 static u32 current_clientid = 1;
@@ -82,7 +85,7 @@ struct nfs4_stateid * find_stateid(stateid_t *stid, int flags);
  *     protects clientid_hashtbl[], clientstr_hashtbl[],
  *     unconfstr_hashtbl[], uncofid_hashtbl[].
  */
-static struct semaphore client_sema;
+static DECLARE_MUTEX(client_sema);
 
 void
 nfs4_lock_state(void)
@@ -131,8 +134,11 @@ static void release_file(struct nfs4_file *fp);
        ((id) & CLIENT_HASH_MASK)
 #define clientstr_hashval(name, namelen) \
        (opaque_hashval((name), (namelen)) & CLIENT_HASH_MASK)
-
-/* conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
+/*
+ * reclaim_str_hashtbl[] holds known client info from previous reset/reboot
+ * used in reboot/reset lease grace period processing
+ *
+ * conf_id_hashtbl[], and conf_str_hashtbl[] hold confirmed
  * setclientid_confirmed info. 
  *
  * unconf_str_hastbl[] and unconf_id_hashtbl[] hold unconfirmed 
@@ -144,6 +150,8 @@ static void release_file(struct nfs4_file *fp);
  * close_lru holds (open) stateowner queue ordered by nfs4_stateowner.so_time
  * for last close replay.
  */
+static struct list_head        reclaim_str_hashtbl[CLIENT_HASH_SIZE];
+static int reclaim_str_hashtbl_size;
 static struct list_head        conf_id_hashtbl[CLIENT_HASH_SIZE];
 static struct list_head        conf_str_hashtbl[CLIENT_HASH_SIZE];
 static struct list_head        unconf_str_hashtbl[CLIENT_HASH_SIZE];
@@ -339,6 +347,95 @@ move_to_confirmed(struct nfs4_client *clp, unsigned int idhashval)
        renew_client(clp);
 }
 
+
+/* a helper function for parse_callback */
+static int
+parse_octet(unsigned int *lenp, char **addrp)
+{
+       unsigned int len = *lenp;
+       char *p = *addrp;
+       int n = -1;
+       char c;
+
+       for (;;) {
+               if (!len)
+                       break;
+               len--;
+               c = *p++;
+               if (c == '.')
+                       break;
+               if ((c < '0') || (c > '9')) {
+                       n = -1;
+                       break;
+               }
+               if (n < 0)
+                       n = 0;
+               n = (n * 10) + (c - '0');
+               if (n > 255) {
+                       n = -1;
+                       break;
+               }
+       }
+       *lenp = len;
+       *addrp = p;
+       return n;
+}
+
+/* parse and set the setclientid ipv4 callback address */
+int
+parse_ipv4(unsigned int addr_len, char *addr_val, unsigned int *cbaddrp, unsigned short *cbportp)
+{
+       int temp = 0;
+       u32 cbaddr = 0;
+       u16 cbport = 0;
+       u32 addrlen = addr_len;
+       char *addr = addr_val;
+       int i, shift;
+
+       /* ipaddress */
+       shift = 24;
+       for(i = 4; i > 0  ; i--) {
+               if ((temp = parse_octet(&addrlen, &addr)) < 0) {
+                       return 0;
+               }
+               cbaddr |= (temp << shift);
+               if(shift > 0)
+               shift -= 8;
+       }
+       *cbaddrp = cbaddr;
+
+       /* port */
+       shift = 8;
+       for(i = 2; i > 0  ; i--) {
+               if ((temp = parse_octet(&addrlen, &addr)) < 0) {
+                       return 0;
+               }
+               cbport |= (temp << shift);
+               if(shift > 0)
+                       shift -= 8;
+       }
+       *cbportp = cbport;
+       return 1;
+}
+
+void
+gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se)
+{
+       struct nfs4_callback *cb = &clp->cl_callback;
+
+       if( !(parse_ipv4(se->se_callback_addr_len, se->se_callback_addr_val,
+                        &cb->cb_addr, &cb->cb_port))) {
+               printk(KERN_INFO "NFSD: BAD callback address. client will not receive delegations\n");
+               cb->cb_parsed = 0;
+               return;
+       }
+       cb->cb_netid.len = se->se_callback_netid_len;
+       cb->cb_netid.data = se->se_callback_netid_val;
+        cb->cb_prog = se->se_callback_prog;
+        cb->cb_ident = se->se_callback_ident;
+        cb->cb_parsed = 1;
+}
+
 /*
  * RFC 3010 has a complex implmentation description of processing a 
  * SETCLIENTID request consisting of 5 bullets, labeled as 
@@ -450,6 +547,7 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
                copy_cred(&new->cl_cred,&rqstp->rq_cred);
                gen_clid(new);
                gen_confirm(new);
+               gen_callback(new, setclid);
                add_to_unconfirmed(new, strhashval);
        } else if (cmp_verf(&conf->cl_verifier, &clverifier)) {
                /*
@@ -477,6 +575,7 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
                copy_cred(&new->cl_cred,&rqstp->rq_cred);
                copy_clid(new, conf);
                gen_confirm(new);
+               gen_callback(new, setclid);
                add_to_unconfirmed(new,strhashval);
        } else if (!unconf) {
                /*
@@ -494,6 +593,7 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
                copy_cred(&new->cl_cred,&rqstp->rq_cred);
                gen_clid(new);
                gen_confirm(new);
+               gen_callback(new, setclid);
                add_to_unconfirmed(new, strhashval);
        } else if (!cmp_verf(&conf->cl_confirm, &unconf->cl_confirm)) {
                /*      
@@ -519,6 +619,7 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
                copy_cred(&new->cl_cred,&rqstp->rq_cred);
                gen_clid(new);
                gen_confirm(new);
+               gen_callback(new, setclid);
                add_to_unconfirmed(new, strhashval);
        } else {
                /* No cases hit !!! */
@@ -529,7 +630,6 @@ nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_setclientid *setclid)
        setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
        setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
        memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
-       printk(KERN_INFO "NFSD: this client will not receive delegations\n");
        status = nfs_ok;
 out:
        nfs4_unlock_state();
@@ -1040,10 +1140,9 @@ int status;
 
        if (share_access & NFS4_SHARE_ACCESS_WRITE) {
                status = get_write_access(filp->f_dentry->d_inode);
-               if (!status)
-                       filp->f_mode = FMODE_WRITE;
-               else
+               if (status)
                        return nfserrno(status);
+               filp->f_mode = (filp->f_mode | FMODE_WRITE) & ~FMODE_READ;
        }
        return nfs_ok;
 }
@@ -1053,7 +1152,7 @@ nfs4_file_downgrade(struct file *filp, unsigned int share_access)
 {
        if (share_access & NFS4_SHARE_ACCESS_WRITE) {
                put_write_access(filp->f_dentry->d_inode);
-               filp->f_mode = FMODE_READ;
+               filp->f_mode = (filp->f_mode | FMODE_READ) & ~FMODE_WRITE;
        }
 }
 
@@ -1510,10 +1609,12 @@ nfs4_preprocess_seqid_op(struct svc_fh *current_fh, u32 seqid, stateid_t *statei
 
        status = nfserr_bad_stateid;
 
-       /* for new lock stateowners, check that the lock->v.new.open_stateid
-        * refers to an open stateowner, and that the lockclid
-        * (nfs4_lock->v.new.clientid) is the same as the
-        * open_stateid->st_stateowner->so_client->clientid
+       /* for new lock stateowners:
+        * check that the lock->v.new.open_stateid
+        * refers to an open stateowner
+        *
+        * check that the lockclid (nfs4_lock->v.new.clientid) is the same
+        * as the open_stateid->st_stateowner->so_client->clientid
         */
        if (lockclid) {
                struct nfs4_stateowner *sop = stp->st_stateowner;
@@ -1598,6 +1699,17 @@ check_replay:
        goto out;
 }
 
+/*
+ * eventually, this will perform an upcall to the 'state daemon' as well as
+ * set the cl_first_state field.
+ */
+void
+first_state(struct nfs4_client *clp)
+{
+       if (!clp->cl_first_state)
+               clp->cl_first_state = get_seconds();
+}
+
 /*
  * nfs4_unlock_state(); called in encode
  */
@@ -1635,6 +1747,7 @@ nfsd4_open_confirm(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfs
                         stp->st_stateid.si_fileid,
                         stp->st_stateid.si_generation);
        status = nfs_ok;
+       first_state(sop->so_client);
 out:
        return status;
 }
@@ -1850,6 +1963,21 @@ nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
                deny->ld_type = NFS4_WRITE_LT;
 }
 
+static struct nfs4_stateowner *
+find_lockstateowner(struct xdr_netobj *owner, clientid_t *clid)
+{
+       struct nfs4_stateowner *local = NULL;
+       int i;
+
+       for (i = 0; i < LOCK_HASH_SIZE; i++) {
+               list_for_each_entry(local, &lock_ownerid_hashtbl[i], so_idhash) {
+                       if(!cmp_owner_str(local, owner, clid))
+                               continue;
+                       return local;
+               }
+       }
+       return NULL;
+}
 
 static int
 find_lockstateowner_str(unsigned int hashval, struct xdr_netobj *owner, clientid_t *clid, struct nfs4_stateowner **op) {
@@ -1969,7 +2097,7 @@ nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock
 
        if (nfs4_in_grace() && !lock->lk_reclaim)
                return nfserr_grace;
-       if (nfs4_in_no_grace() && lock->lk_reclaim)
+       if (!nfs4_in_grace() && lock->lk_reclaim)
                return nfserr_no_grace;
 
        if (check_lock_length(lock->lk_offset, lock->lk_length))
@@ -1992,7 +2120,11 @@ nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock
                        printk("NFSD: nfsd4_lock: clientid is stale!\n");
                        goto out;
                }
-               /* does the clientid in the lock owner own the open stateid? */
+
+               /* is the new lock seqid presented by the client zero? */
+               status = nfserr_bad_seqid;
+               if (lock->v.new.lock_seqid != 0)
+                       goto out;
 
                /* validate and update open stateid and open seqid */
                status = nfs4_preprocess_seqid_op(current_fh, 
@@ -2011,15 +2143,15 @@ nfsd4_lock(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_lock
                strhashval = lock_ownerstr_hashval(fp->fi_inode, 
                                open_sop->so_client->cl_clientid.cl_id, 
                                lock->v.new.owner);
-
                /* 
                 * If we already have this lock owner, the client is in 
                 * error (or our bookeeping is wrong!) 
                 * for asking for a 'new lock'.
                 */
                status = nfserr_bad_stateid;
-               if (find_lockstateowner_str(strhashval, &lock->v.new.owner,
-                                       &lock->v.new.clientid, &lock_sop))
+               lock_sop = find_lockstateowner(&lock->v.new.owner,
+                                               &lock->v.new.clientid);
+               if (lock_sop)
                        goto out;
                status = nfserr_resource;
                if (!(lock->lk_stateowner = alloc_init_lock_stateowner(strhashval, open_sop->so_client, open_stp, lock)))
@@ -2315,7 +2447,7 @@ nfsd4_release_lockowner(struct svc_rqst *rqstp, struct nfsd4_release_lockowner *
        clientid_t *clid = &rlockowner->rl_clientid;
        struct nfs4_stateowner *local = NULL;
        struct xdr_netobj *owner = &rlockowner->rl_owner;
-       int status, i;
+       int status;
 
        dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
                clid->cl_boot, clid->cl_id);
@@ -2330,34 +2462,136 @@ nfsd4_release_lockowner(struct svc_rqst *rqstp, struct nfsd4_release_lockowner *
 
        nfs4_lock_state();
 
-       /* find the lockowner */
         status = nfs_ok;
-       for (i=0; i < LOCK_HASH_SIZE; i++)
-               list_for_each_entry(local, &lock_ownerstr_hashtbl[i], so_strhash)
-                       if(cmp_owner_str(local, owner, clid)) {
-                               struct nfs4_stateid *stp;
-
-                               /* check for any locks held by any stateid
-                                * associated with the (lock) stateowner */
-                               status = nfserr_locks_held;
-                               list_for_each_entry(stp, &local->so_perfilestate,
-                                                   st_perfilestate) {
-                                       if(stp->st_vfs_set) {
-                                               if (check_for_locks(&stp->st_vfs_file,
-                                                                   local))
-                                                       goto out;
-                                       }
-                               }
-                               /* no locks held by (lock) stateowner */
-                               status = nfs_ok;
-                               release_stateowner(local);
-                               goto out;
+       local = find_lockstateowner(owner, clid);
+       if (local) {
+               struct nfs4_stateid *stp;
+
+               /* check for any locks held by any stateid
+                * associated with the (lock) stateowner */
+               status = nfserr_locks_held;
+               list_for_each_entry(stp, &local->so_perfilestate,
+                               st_perfilestate) {
+                       if(stp->st_vfs_set) {
+                               if (check_for_locks(&stp->st_vfs_file, local))
+                                       goto out;
                        }
+               }
+               /* no locks held by (lock) stateowner */
+               status = nfs_ok;
+               release_stateowner(local);
+       }
 out:
        nfs4_unlock_state();
        return status;
 }
 
+static inline struct nfs4_client_reclaim *
+alloc_reclaim(int namelen)
+{
+       struct nfs4_client_reclaim *crp = NULL;
+
+       crp = kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
+       if (!crp)
+               return NULL;
+       crp->cr_name.data = kmalloc(namelen, GFP_KERNEL);
+       if (!crp->cr_name.data) {
+               kfree(crp);
+               return NULL;
+       }
+               return crp;
+}
+
+/*
+ * failure => all reset bets are off, nfserr_no_grace...
+ */
+static int
+nfs4_client_to_reclaim(struct nfs4_client *clp)
+{
+       unsigned int strhashval;
+       struct nfs4_client_reclaim *crp = NULL;
+
+       crp = alloc_reclaim(clp->cl_name.len);
+       if (!crp)
+               return 0;
+       strhashval = clientstr_hashval(clp->cl_name.data, clp->cl_name.len);
+       INIT_LIST_HEAD(&crp->cr_strhash);
+       list_add(&crp->cr_strhash, &reclaim_str_hashtbl[strhashval]);
+       memcpy(crp->cr_name.data, clp->cl_name.data, clp->cl_name.len);
+       crp->cr_name.len = clp->cl_name.len;
+       crp->cr_first_state = clp->cl_first_state;
+       crp->cr_expired = 0;
+       return 1;
+}
+
+static void
+nfs4_release_reclaim(void)
+{
+       struct nfs4_client_reclaim *crp = NULL;
+       int i;
+
+       BUG_ON(!nfs4_reclaim_init);
+       for (i = 0; i < CLIENT_HASH_SIZE; i++) {
+               while (!list_empty(&reclaim_str_hashtbl[i])) {
+                       crp = list_entry(reclaim_str_hashtbl[i].next,
+                                       struct nfs4_client_reclaim, cr_strhash);
+                       list_del(&crp->cr_strhash);
+                       kfree(crp->cr_name.data);
+                       kfree(crp);
+                       reclaim_str_hashtbl_size--;
+               }
+       }
+       BUG_ON(reclaim_str_hashtbl_size);
+}
+
+/*
+ * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
+struct nfs4_client_reclaim *
+nfs4_find_reclaim_client(clientid_t *clid)
+{
+       unsigned int idhashval = clientid_hashval(clid->cl_id);
+       unsigned int strhashval;
+       struct nfs4_client *clp, *client = NULL;
+       struct nfs4_client_reclaim *crp = NULL;
+
+
+       /* find clientid in conf_id_hashtbl */
+       list_for_each_entry(clp, &conf_id_hashtbl[idhashval], cl_idhash) {
+               if (cmp_clid(&clp->cl_clientid, clid)) {
+                       client = clp;
+                       break;
+               }
+       }
+       if (!client)
+               return NULL;
+
+       /* find clp->cl_name in reclaim_str_hashtbl */
+       strhashval = clientstr_hashval(client->cl_name.data,
+                                     client->cl_name.len);
+       list_for_each_entry(crp, &reclaim_str_hashtbl[strhashval], cr_strhash) {
+               if(cmp_name(&crp->cr_name, &client->cl_name)) {
+                       return crp;
+               }
+       }
+       return NULL;
+}
+
+/*
+* Called from OPEN. Look for clientid in reclaim list.
+*/
+int
+nfs4_check_open_reclaim(clientid_t *clid)
+{
+       struct nfs4_client_reclaim *crp;
+
+       if ((crp = nfs4_find_reclaim_client(clid)) == NULL)
+               return nfserr_reclaim_bad;
+       if (crp->cr_expired)
+               return nfserr_no_grace;
+       return nfs_ok;
+}
+
+
 /* 
  * Start and stop routines
  */
@@ -2366,10 +2600,16 @@ void
 nfs4_state_init(void)
 {
        int i;
-       time_t start = get_seconds();
+       time_t grace_time;
 
        if (nfs4_init)
                return;
+       if (!nfs4_reclaim_init) {
+               for (i = 0; i < CLIENT_HASH_SIZE; i++)
+                       INIT_LIST_HEAD(&reclaim_str_hashtbl[i]);
+               reclaim_str_hashtbl_size = 0;
+               nfs4_reclaim_init = 1;
+       }
        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
                INIT_LIST_HEAD(&conf_id_hashtbl[i]);
                INIT_LIST_HEAD(&conf_str_hashtbl[i]);
@@ -2396,27 +2636,36 @@ nfs4_state_init(void)
 
        INIT_LIST_HEAD(&close_lru);
        INIT_LIST_HEAD(&client_lru);
-       init_MUTEX(&client_sema);
-       boot_time = start;
-       grace_end = start + NFSD_LEASE_TIME;
+       boot_time = get_seconds();
+       grace_time = max(old_lease_time, lease_time);
+       if (reclaim_str_hashtbl_size == 0)
+               grace_time = 0;
+       if (grace_time)
+               printk("NFSD: starting %ld-second grace period\n", grace_time);
+       grace_end = boot_time + grace_time;
        INIT_WORK(&laundromat_work,laundromat_main, NULL);
        schedule_delayed_work(&laundromat_work, NFSD_LEASE_TIME*HZ);
        nfs4_init = 1;
-
 }
 
 int
 nfs4_in_grace(void)
 {
-       return time_before(get_seconds(), (unsigned long)grace_end);
+       return get_seconds() < grace_end;
 }
 
-int
-nfs4_in_no_grace(void)
+void
+set_no_grace(void)
 {
-       return (grace_end < get_seconds());
+       printk("NFSD: ERROR in reboot recovery.  State reclaims will fail.\n");
+       grace_end = get_seconds();
 }
 
+time_t
+nfs4_lease_time(void)
+{
+       return lease_time;
+}
 
 static void
 __nfs4_state_shutdown(void)
@@ -2454,6 +2703,61 @@ void
 nfs4_state_shutdown(void)
 {
        nfs4_lock_state();
+       nfs4_release_reclaim();
        __nfs4_state_shutdown();
        nfs4_unlock_state();
 }
+
+/*
+ * Called when leasetime is changed.
+ *
+ * if nfsd is not started, simply set the global lease.
+ *
+ * if nfsd(s) are running, lease change requires nfsv4 state to be reset.
+ * e.g: boot_time is reset, existing nfs4_client structs are
+ * used to fill reclaim_str_hashtbl, then all state (except for the
+ * reclaim_str_hashtbl) is re-initialized.
+ *
+ * if the old lease time is greater than the new lease time, the grace
+ * period needs to be set to the old lease time to allow clients to reclaim
+ * their state. XXX - we may want to set the grace period == lease time
+ * after an initial grace period == old lease time
+ *
+ * if an error occurs in this process, the new lease is set, but the server
+ * will not honor OPEN or LOCK reclaims, and will return nfserr_no_grace
+ * which means OPEN/LOCK/READ/WRITE will fail during grace period.
+ *
+ * clients will attempt to reset all state with SETCLIENTID/CONFIRM, and
+ * OPEN and LOCK reclaims.
+ */
+void
+nfs4_reset_lease(time_t leasetime)
+{
+       struct nfs4_client *clp;
+       int i;
+
+       printk("NFSD: New leasetime %ld\n",leasetime);
+       if (!nfs4_init)
+               return;
+       nfs4_lock_state();
+       old_lease_time = lease_time;
+       lease_time = leasetime;
+
+       nfs4_release_reclaim();
+
+       /* populate reclaim_str_hashtbl with current confirmed nfs4_clientid */
+       for (i = 0; i < CLIENT_HASH_SIZE; i++) {
+               list_for_each_entry(clp, &conf_id_hashtbl[i], cl_idhash) {
+                       if (!nfs4_client_to_reclaim(clp)) {
+                               nfs4_release_reclaim();
+                               goto init_state;
+                       }
+                       reclaim_str_hashtbl_size++;
+               }
+       }
+init_state:
+       __nfs4_state_shutdown();
+       nfs4_state_init();
+       nfs4_unlock_state();
+}
+