cgroup: add documentation on extended attributes usage
[cascardo/linux.git] / fs / nfsd / nfs4state.c
index 94effd5..cc894ed 100644 (file)
 #include <linux/namei.h>
 #include <linux/swap.h>
 #include <linux/pagemap.h>
+#include <linux/ratelimit.h>
 #include <linux/sunrpc/svcauth_gss.h>
 #include <linux/sunrpc/clnt.h>
 #include "xdr4.h"
 #include "vfs.h"
 #include "current_stateid.h"
+#include "fault_inject.h"
+
+#include "netns.h"
 
 #define NFSDDBG_FACILITY                NFSDDBG_PROC
 
 /* Globals */
 time_t nfsd4_lease = 90;     /* default lease time */
 time_t nfsd4_grace = 90;
-static time_t boot_time;
 
 #define all_ones {{~0,~0},~0}
 static const stateid_t one_stateid = {
@@ -862,6 +865,11 @@ static __be32 nfsd4_new_conn(struct svc_rqst *rqstp, struct nfsd4_session *ses,
        if (ret)
                /* oops; xprt is already down: */
                nfsd4_conn_lost(&conn->cn_xpt_user);
+       if (ses->se_client->cl_cb_state == NFSD4_CB_DOWN &&
+               dir & NFS4_CDFC4_BACK) {
+               /* callback channel may be back up */
+               nfsd4_probe_callback(ses->se_client);
+       }
        return nfs_ok;
 }
 
@@ -1047,12 +1055,12 @@ renew_client(struct nfs4_client *clp)
 
 /* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
 static int
-STALE_CLIENTID(clientid_t *clid)
+STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
 {
-       if (clid->cl_boot == boot_time)
+       if (clid->cl_boot == nn->boot_time)
                return 0;
        dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
-               clid->cl_boot, clid->cl_id, boot_time);
+               clid->cl_boot, clid->cl_id, nn->boot_time);
        return 1;
 }
 
@@ -1215,7 +1223,7 @@ static bool groups_equal(struct group_info *g1, struct group_info *g2)
        return true;
 }
 
-static int
+static bool
 same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
 {
        if ((cr1->cr_flavor != cr2->cr_flavor)
@@ -1227,14 +1235,15 @@ same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
                return true;
        if (!cr1->cr_principal || !cr2->cr_principal)
                return false;
-       return 0 == strcmp(cr1->cr_principal, cr1->cr_principal);
+       return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
 }
 
 static void gen_clid(struct nfs4_client *clp)
 {
        static u32 current_clientid = 1;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
-       clp->cl_clientid.cl_boot = boot_time;
+       clp->cl_clientid.cl_boot = nn->boot_time;
        clp->cl_clientid.cl_id = current_clientid++; 
 }
 
@@ -2217,8 +2226,9 @@ nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
        nfs4_verifier confirm = setclientid_confirm->sc_confirm; 
        clientid_t * clid = &setclientid_confirm->sc_clientid;
        __be32 status;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
-       if (STALE_CLIENTID(clid))
+       if (STALE_CLIENTID(clid, nn))
                return nfserr_stale_clientid;
        nfs4_lock_state();
 
@@ -2577,8 +2587,9 @@ nfsd4_process_open1(struct nfsd4_compound_state *cstate,
        unsigned int strhashval;
        struct nfs4_openowner *oo = NULL;
        __be32 status;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
-       if (STALE_CLIENTID(&open->op_clientid))
+       if (STALE_CLIENTID(&open->op_clientid, nn))
                return nfserr_stale_clientid;
        /*
         * In case we need it later, after we've already created the
@@ -2876,7 +2887,8 @@ static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
  * Attempt to hand out a delegation.
  */
 static void
-nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
+nfs4_open_delegation(struct net *net, struct svc_fh *fh,
+                    struct nfsd4_open *open, struct nfs4_ol_stateid *stp)
 {
        struct nfs4_delegation *dp;
        struct nfs4_openowner *oo = container_of(stp->st_stateowner, struct nfs4_openowner, oo_owner);
@@ -2897,7 +2909,7 @@ nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open, struct nfs4_ol_
                case NFS4_OPEN_CLAIM_NULL:
                        /* Let's not give out any delegations till everyone's
                         * had the chance to reclaim theirs.... */
-                       if (locks_in_grace())
+                       if (locks_in_grace(net))
                                goto out;
                        if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
                                goto out;
@@ -3005,16 +3017,14 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
                        goto out;
        } else {
                status = nfs4_get_vfs_file(rqstp, fp, current_fh, open);
+               if (status)
+                       goto out;
+               status = nfsd4_truncate(rqstp, current_fh, open);
                if (status)
                        goto out;
                stp = open->op_stp;
                open->op_stp = NULL;
                init_open_stateid(stp, fp, open);
-               status = nfsd4_truncate(rqstp, current_fh, open);
-               if (status) {
-                       release_open_stateid(stp);
-                       goto out;
-               }
        }
        update_stateid(&stp->st_stid.sc_stateid);
        memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
@@ -3033,7 +3043,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
        * Attempt to hand out a delegation. No error return, because the
        * OPEN succeeds even if we fail.
        */
-       nfs4_open_delegation(current_fh, open, stp);
+       nfs4_open_delegation(SVC_NET(rqstp), current_fh, open, stp);
 nodeleg:
        status = nfs_ok;
 
@@ -3087,12 +3097,13 @@ nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 {
        struct nfs4_client *clp;
        __be32 status;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
        nfs4_lock_state();
        dprintk("process_renew(%08x/%08x): starting\n", 
                        clid->cl_boot, clid->cl_id);
        status = nfserr_stale_clientid;
-       if (STALE_CLIENTID(clid))
+       if (STALE_CLIENTID(clid, nn))
                goto out;
        clp = find_confirmed_client(clid);
        status = nfserr_expired;
@@ -3111,22 +3122,19 @@ out:
        return status;
 }
 
-static struct lock_manager nfsd4_manager = {
-};
-
-static bool grace_ended;
-
 static void
-nfsd4_end_grace(void)
+nfsd4_end_grace(struct net *net)
 {
+       struct nfsd_net *nn = net_generic(net, nfsd_net_id);
+
        /* do nothing if grace period already ended */
-       if (grace_ended)
+       if (nn->grace_ended)
                return;
 
        dprintk("NFSD: end of grace period\n");
-       grace_ended = true;
-       nfsd4_record_grace_done(&init_net, boot_time);
-       locks_end_grace(&nfsd4_manager);
+       nn->grace_ended = true;
+       nfsd4_record_grace_done(net, nn->boot_time);
+       locks_end_grace(&nn->nfsd4_manager);
        /*
         * Now that every NFSv4 client has had the chance to recover and
         * to see the (possibly new, possibly shorter) lease time, we
@@ -3149,7 +3157,7 @@ nfs4_laundromat(void)
        nfs4_lock_state();
 
        dprintk("NFSD: laundromat service - starting\n");
-       nfsd4_end_grace();
+       nfsd4_end_grace(&init_net);
        INIT_LIST_HEAD(&reaplist);
        spin_lock(&client_lock);
        list_for_each_safe(pos, next, &client_lru) {
@@ -3231,9 +3239,9 @@ static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *s
 }
 
 static int
-STALE_STATEID(stateid_t *stateid)
+STALE_STATEID(stateid_t *stateid, struct nfsd_net *nn)
 {
-       if (stateid->si_opaque.so_clid.cl_boot == boot_time)
+       if (stateid->si_opaque.so_clid.cl_boot == nn->boot_time)
                return 0;
        dprintk("NFSD: stale stateid " STATEID_FMT "!\n",
                STATEID_VAL(stateid));
@@ -3273,11 +3281,11 @@ out:
 }
 
 static inline __be32
-check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
+check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
 {
        if (ONE_STATEID(stateid) && (flags & RD_STATE))
                return nfs_ok;
-       else if (locks_in_grace()) {
+       else if (locks_in_grace(net)) {
                /* Answer in remaining cases depends on existence of
                 * conflicting state; so we must wait out the grace period. */
                return nfserr_grace;
@@ -3294,9 +3302,9 @@ check_special_stateids(svc_fh *current_fh, stateid_t *stateid, int flags)
  * that are not able to provide mandatory locking.
  */
 static inline int
-grace_disallows_io(struct inode *inode)
+grace_disallows_io(struct net *net, struct inode *inode)
 {
-       return locks_in_grace() && mandatory_lock(inode);
+       return locks_in_grace(net) && mandatory_lock(inode);
 }
 
 /* Returns true iff a is later than b: */
@@ -3333,18 +3341,26 @@ static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_s
        return nfserr_old_stateid;
 }
 
-__be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
+static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
 {
        struct nfs4_stid *s;
        struct nfs4_ol_stateid *ols;
        __be32 status;
 
-       if (STALE_STATEID(stateid))
-               return nfserr_stale_stateid;
-
+       if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
+               return nfserr_bad_stateid;
+       /* Client debugging aid. */
+       if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
+               char addr_str[INET6_ADDRSTRLEN];
+               rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
+                                sizeof(addr_str));
+               pr_warn_ratelimited("NFSD: client %s testing state ID "
+                                       "with incorrect client ID\n", addr_str);
+               return nfserr_bad_stateid;
+       }
        s = find_stateid(cl, stateid);
        if (!s)
-                return nfserr_stale_stateid;
+               return nfserr_bad_stateid;
        status = check_stateid_generation(stateid, &s->sc_stateid, 1);
        if (status)
                return status;
@@ -3360,10 +3376,11 @@ __be32 nfs4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
 static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, struct nfs4_stid **s)
 {
        struct nfs4_client *cl;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
                return nfserr_bad_stateid;
-       if (STALE_STATEID(stateid))
+       if (STALE_STATEID(stateid, nn))
                return nfserr_stale_stateid;
        cl = find_confirmed_client(&stateid->si_opaque.so_clid);
        if (!cl)
@@ -3379,7 +3396,7 @@ static __be32 nfsd4_lookup_stateid(stateid_t *stateid, unsigned char typemask, s
 * Checks for stateid operations
 */
 __be32
-nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
+nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
                           stateid_t *stateid, int flags, struct file **filpp)
 {
        struct nfs4_stid *s;
@@ -3392,11 +3409,11 @@ nfs4_preprocess_stateid_op(struct nfsd4_compound_state *cstate,
        if (filpp)
                *filpp = NULL;
 
-       if (grace_disallows_io(ino))
+       if (grace_disallows_io(net, ino))
                return nfserr_grace;
 
        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
-               return check_special_stateids(current_fh, stateid, flags);
+               return check_special_stateids(net, current_fh, stateid, flags);
 
        status = nfsd4_lookup_stateid(stateid, NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID, &s);
        if (status)
@@ -3463,7 +3480,8 @@ nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
 
        nfs4_lock_state();
        list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
-               stateid->ts_id_status = nfs4_validate_stateid(cl, &stateid->ts_id_stateid);
+               stateid->ts_id_status =
+                       nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
        nfs4_unlock_state();
 
        return nfs_ok;
@@ -3750,12 +3768,19 @@ nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
        nfsd4_close_open_stateid(stp);
        oo->oo_last_closed_stid = stp;
 
-       /* place unused nfs4_stateowners on so_close_lru list to be
-        * released by the laundromat service after the lease period
-        * to enable us to handle CLOSE replay
-        */
-       if (list_empty(&oo->oo_owner.so_stateids))
-               move_to_close_lru(oo);
+       if (list_empty(&oo->oo_owner.so_stateids)) {
+               if (cstate->minorversion) {
+                       release_openowner(oo);
+                       cstate->replay_owner = NULL;
+               } else {
+                       /*
+                        * In the 4.0 case we need to keep the owners around a
+                        * little while to handle CLOSE replay.
+                        */
+                       if (list_empty(&oo->oo_owner.so_stateids))
+                               move_to_close_lru(oo);
+               }
+       }
 out:
        if (!cstate->replay_owner)
                nfs4_unlock_state();
@@ -4027,6 +4052,7 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
        bool new_state = false;
        int lkflg;
        int err;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
        dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
                (long long) lock->lk_offset,
@@ -4044,11 +4070,6 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
        nfs4_lock_state();
 
        if (lock->lk_is_new) {
-               /*
-                * Client indicates that this is a new lockowner.
-                * Use open owner and open stateid to create lock owner and
-                * lock stateid.
-                */
                struct nfs4_ol_stateid *open_stp = NULL;
 
                if (nfsd4_has_session(cstate))
@@ -4058,7 +4079,7 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
                                sizeof(clientid_t));
 
                status = nfserr_stale_clientid;
-               if (STALE_CLIENTID(&lock->lk_new_clientid))
+               if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
                        goto out;
 
                /* validate and update open stateid and open seqid */
@@ -4075,17 +4096,13 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
                        goto out;
                status = lookup_or_create_lock_state(cstate, open_stp, lock,
                                                        &lock_stp, &new_state);
-               if (status)
-                       goto out;
-       } else {
-               /* lock (lock owner + lock stateid) already exists */
+       } else
                status = nfs4_preprocess_seqid_op(cstate,
                                       lock->lk_old_lock_seqid,
                                       &lock->lk_old_lock_stateid,
                                       NFS4_LOCK_STID, &lock_stp);
-               if (status)
-                       goto out;
-       }
+       if (status)
+               goto out;
        lock_sop = lockowner(lock_stp->st_stateowner);
 
        lkflg = setlkflg(lock->lk_type);
@@ -4094,10 +4111,10 @@ nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
                goto out;
 
        status = nfserr_grace;
-       if (locks_in_grace() && !lock->lk_reclaim)
+       if (locks_in_grace(SVC_NET(rqstp)) && !lock->lk_reclaim)
                goto out;
        status = nfserr_no_grace;
-       if (!locks_in_grace() && lock->lk_reclaim)
+       if (!locks_in_grace(SVC_NET(rqstp)) && lock->lk_reclaim)
                goto out;
 
        locks_init_lock(&file_lock);
@@ -4196,8 +4213,9 @@ nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
        struct file_lock file_lock;
        struct nfs4_lockowner *lo;
        __be32 status;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
-       if (locks_in_grace())
+       if (locks_in_grace(SVC_NET(rqstp)))
                return nfserr_grace;
 
        if (check_lock_length(lockt->lt_offset, lockt->lt_length))
@@ -4206,7 +4224,7 @@ nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
        nfs4_lock_state();
 
        status = nfserr_stale_clientid;
-       if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid))
+       if (!nfsd4_has_session(cstate) && STALE_CLIENTID(&lockt->lt_clientid, nn))
                goto out;
 
        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
@@ -4355,6 +4373,7 @@ nfsd4_release_lockowner(struct svc_rqst *rqstp,
        struct list_head matches;
        unsigned int hashval = ownerstr_hashval(clid->cl_id, owner);
        __be32 status;
+       struct nfsd_net *nn = net_generic(&init_net, nfsd_net_id);
 
        dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
                clid->cl_boot, clid->cl_id);
@@ -4362,7 +4381,7 @@ nfsd4_release_lockowner(struct svc_rqst *rqstp,
        /* XXX check for lease expiration */
 
        status = nfserr_stale_clientid;
-       if (STALE_CLIENTID(clid))
+       if (STALE_CLIENTID(clid, nn))
                return status;
 
        nfs4_lock_state();
@@ -4564,7 +4583,7 @@ void nfsd_forget_openowners(u64 num)
        printk(KERN_INFO "NFSD: Forgot %d open owners", count);
 }
 
-int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegation *))
+int nfsd_process_n_delegations(u64 num, struct list_head *list)
 {
        int i, count = 0;
        struct nfs4_file *fp, *fnext;
@@ -4573,7 +4592,7 @@ int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegatio
        for (i = 0; i < FILE_HASH_SIZE; i++) {
                list_for_each_entry_safe(fp, fnext, &file_hashtbl[i], fi_hash) {
                        list_for_each_entry_safe(dp, dnext, &fp->fi_delegations, dl_perfile) {
-                               deleg_func(dp);
+                               list_move(&dp->dl_recall_lru, list);
                                if (++count == num)
                                        return count;
                        }
@@ -4586,9 +4605,16 @@ int nfsd_process_n_delegations(u64 num, void (*deleg_func)(struct nfs4_delegatio
 void nfsd_forget_delegations(u64 num)
 {
        unsigned int count;
+       LIST_HEAD(victims);
+       struct nfs4_delegation *dp, *dnext;
+
+       spin_lock(&recall_lock);
+       count = nfsd_process_n_delegations(num, &victims);
+       spin_unlock(&recall_lock);
 
        nfs4_lock_state();
-       count = nfsd_process_n_delegations(num, unhash_delegation);
+       list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru)
+               unhash_delegation(dp);
        nfs4_unlock_state();
 
        printk(KERN_INFO "NFSD: Forgot %d delegations", count);
@@ -4597,12 +4623,16 @@ void nfsd_forget_delegations(u64 num)
 void nfsd_recall_delegations(u64 num)
 {
        unsigned int count;
+       LIST_HEAD(victims);
+       struct nfs4_delegation *dp, *dnext;
 
-       nfs4_lock_state();
        spin_lock(&recall_lock);
-       count = nfsd_process_n_delegations(num, nfsd_break_one_deleg);
+       count = nfsd_process_n_delegations(num, &victims);
+       list_for_each_entry_safe(dp, dnext, &victims, dl_recall_lru) {
+               list_del(&dp->dl_recall_lru);
+               nfsd_break_one_deleg(dp);
+       }
        spin_unlock(&recall_lock);
-       nfs4_unlock_state();
 
        printk(KERN_INFO "NFSD: Recalled %d delegations", count);
 }
@@ -4665,6 +4695,8 @@ set_max_delegations(void)
 int
 nfs4_state_start(void)
 {
+       struct net *net = &init_net;
+       struct nfsd_net *nn = net_generic(net, nfsd_net_id);
        int ret;
 
        /*
@@ -4674,11 +4706,11 @@ nfs4_state_start(void)
         * to that instead and then do most of the rest of this on a per-net
         * basis.
         */
-       get_net(&init_net);
-       nfsd4_client_tracking_init(&init_net);
-       boot_time = get_seconds();
-       locks_start_grace(&nfsd4_manager);
-       grace_ended = false;
+       get_net(net);
+       nfsd4_client_tracking_init(net);
+       nn->boot_time = get_seconds();
+       locks_start_grace(net, &nn->nfsd4_manager);
+       nn->grace_ended = false;
        printk(KERN_INFO "NFSD: starting %ld-second grace period\n",
               nfsd4_grace);
        ret = set_callback_cred();
@@ -4700,8 +4732,8 @@ nfs4_state_start(void)
 out_free_laundry:
        destroy_workqueue(laundry_wq);
 out_recovery:
-       nfsd4_client_tracking_exit(&init_net);
-       put_net(&init_net);
+       nfsd4_client_tracking_exit(net);
+       put_net(net);
        return ret;
 }
 
@@ -4742,9 +4774,12 @@ __nfs4_state_shutdown(void)
 void
 nfs4_state_shutdown(void)
 {
+       struct net *net = &init_net;
+       struct nfsd_net *nn = net_generic(net, nfsd_net_id);
+
        cancel_delayed_work_sync(&laundromat_work);
        destroy_workqueue(laundry_wq);
-       locks_end_grace(&nfsd4_manager);
+       locks_end_grace(&nn->nfsd4_manager);
        nfs4_lock_state();
        __nfs4_state_shutdown();
        nfs4_unlock_state();