nfsd: Add reference counting to the lock and open stateids
authorTrond Myklebust <trond.myklebust@primarydata.com>
Wed, 30 Jul 2014 01:34:06 +0000 (21:34 -0400)
committerJ. Bruce Fields <bfields@redhat.com>
Thu, 31 Jul 2014 16:43:53 +0000 (12:43 -0400)
When we remove the client_mutex, we'll need to be able to ensure that
these objects aren't destroyed while we're not holding locks.

Add a ->free() callback to the struct nfs4_stid, so that we can
release a reference to the stid without caring about the contents.

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
fs/nfsd/nfs4callback.c
fs/nfsd/nfs4state.c
fs/nfsd/state.h

index e981338..8574c70 100644 (file)
@@ -905,7 +905,7 @@ static void nfsd4_cb_recall_release(void *calldata)
                spin_lock(&clp->cl_lock);
                list_del(&cb->cb_per_client);
                spin_unlock(&clp->cl_lock);
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(&dp->dl_stid);
        }
 }
 
index b0f83be..60ab22b 100644 (file)
@@ -71,6 +71,7 @@ static u64 current_sessionid = 1;
 
 /* forward declarations */
 static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
+static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
 
 /* Locking: */
 
@@ -463,8 +464,8 @@ static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
                __nfs4_file_put_access(fp, O_RDONLY);
 }
 
-static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct
-kmem_cache *slab)
+static struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl,
+                                        struct kmem_cache *slab)
 {
        struct idr *stateids = &cl->cl_stateids;
        struct nfs4_stid *stid;
@@ -500,7 +501,26 @@ out_free:
 
 static struct nfs4_ol_stateid * nfs4_alloc_stateid(struct nfs4_client *clp)
 {
-       return openlockstateid(nfs4_alloc_stid(clp, stateid_slab));
+       struct nfs4_stid *stid;
+       struct nfs4_ol_stateid *stp;
+
+       stid = nfs4_alloc_stid(clp, stateid_slab);
+       if (!stid)
+               return NULL;
+
+       stp = openlockstateid(stid);
+       stp->st_stid.sc_free = nfs4_free_ol_stateid;
+       return stp;
+}
+
+static void nfs4_free_deleg(struct nfs4_stid *stid)
+{
+       struct nfs4_delegation *dp = delegstateid(stid);
+
+       if (dp->dl_file)
+               put_nfs4_file(dp->dl_file);
+       kmem_cache_free(deleg_slab, stid);
+       atomic_long_dec(&num_delegations);
 }
 
 /*
@@ -594,6 +614,8 @@ alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh)
        dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
        if (dp == NULL)
                goto out_dec;
+
+       dp->dl_stid.sc_free = nfs4_free_deleg;
        /*
         * delegation seqid's are never incremented.  The 4.1 special
         * meaning of seqid 0 isn't meaningful, really, but let's avoid
@@ -611,28 +633,15 @@ out_dec:
        return NULL;
 }
 
-static void remove_stid(struct nfs4_stid *s)
-{
-       struct idr *stateids = &s->sc_client->cl_stateids;
-
-       idr_remove(stateids, s->sc_stateid.si_opaque.so_id);
-}
-
-static void nfs4_free_stid(struct kmem_cache *slab, struct nfs4_stid *s)
-{
-       kmem_cache_free(slab, s);
-}
-
 void
-nfs4_put_delegation(struct nfs4_delegation *dp)
+nfs4_put_stid(struct nfs4_stid *s)
 {
-       if (atomic_dec_and_test(&dp->dl_stid.sc_count)) {
-               if (dp->dl_file)
-                       put_nfs4_file(dp->dl_file);
-               remove_stid(&dp->dl_stid);
-               nfs4_free_stid(deleg_slab, &dp->dl_stid);
-               atomic_long_dec(&num_delegations);
-       }
+       struct nfs4_client *clp = s->sc_client;
+
+       if (!atomic_dec_and_test(&s->sc_count))
+               return;
+       idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
+       s->sc_free(s);
 }
 
 static void nfs4_put_deleg_lease(struct nfs4_file *fp)
@@ -689,7 +698,7 @@ static void destroy_delegation(struct nfs4_delegation *dp)
        spin_lock(&state_lock);
        unhash_delegation_locked(dp);
        spin_unlock(&state_lock);
-       nfs4_put_delegation(dp);
+       nfs4_put_stid(&dp->dl_stid);
 }
 
 static void revoke_delegation(struct nfs4_delegation *dp)
@@ -699,7 +708,7 @@ static void revoke_delegation(struct nfs4_delegation *dp)
        WARN_ON(!list_empty(&dp->dl_recall_lru));
 
        if (clp->cl_minorversion == 0)
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(&dp->dl_stid);
        else {
                dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
                spin_lock(&clp->cl_lock);
@@ -885,17 +894,14 @@ static void unhash_generic_stateid(struct nfs4_ol_stateid *stp)
        list_del(&stp->st_perstateowner);
 }
 
-static void close_generic_stateid(struct nfs4_ol_stateid *stp)
+static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
 {
-       release_all_access(stp);
-}
+       struct nfs4_ol_stateid *stp = openlockstateid(stid);
 
-static void free_generic_stateid(struct nfs4_ol_stateid *stp)
-{
+       release_all_access(stp);
        if (stp->st_file)
                put_nfs4_file(stp->st_file);
-       remove_stid(&stp->st_stid);
-       nfs4_free_stid(stateid_slab, &stp->st_stid);
+       kmem_cache_free(stateid_slab, stid);
 }
 
 static void __release_lock_stateid(struct nfs4_ol_stateid *stp)
@@ -908,8 +914,7 @@ static void __release_lock_stateid(struct nfs4_ol_stateid *stp)
        file = find_any_file(stp->st_file);
        if (file)
                filp_close(file, (fl_owner_t)lockowner(stp->st_stateowner));
-       close_generic_stateid(stp);
-       free_generic_stateid(stp);
+       nfs4_put_stid(&stp->st_stid);
 }
 
 static void unhash_lockowner(struct nfs4_lockowner *lo)
@@ -966,13 +971,12 @@ static void unhash_open_stateid(struct nfs4_ol_stateid *stp)
 {
        unhash_generic_stateid(stp);
        release_open_stateid_locks(stp);
-       close_generic_stateid(stp);
 }
 
 static void release_open_stateid(struct nfs4_ol_stateid *stp)
 {
        unhash_open_stateid(stp);
-       free_generic_stateid(stp);
+       nfs4_put_stid(&stp->st_stid);
 }
 
 static void unhash_openowner(struct nfs4_openowner *oo)
@@ -993,7 +997,7 @@ static void release_last_closed_stateid(struct nfs4_openowner *oo)
        struct nfs4_ol_stateid *s = oo->oo_last_closed_stid;
 
        if (s) {
-               free_generic_stateid(s);
+               nfs4_put_stid(&s->st_stid);
                oo->oo_last_closed_stid = NULL;
        }
 }
@@ -1467,12 +1471,12 @@ destroy_client(struct nfs4_client *clp)
        while (!list_empty(&reaplist)) {
                dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
                list_del_init(&dp->dl_recall_lru);
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(&dp->dl_stid);
        }
        while (!list_empty(&clp->cl_revoked)) {
                dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
                list_del_init(&dp->dl_recall_lru);
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(&dp->dl_stid);
        }
        while (!list_empty(&clp->cl_openowners)) {
                oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
@@ -3588,7 +3592,7 @@ out_unlock:
        spin_unlock(&state_lock);
 out:
        if (status) {
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(&dp->dl_stid);
                return ERR_PTR(status);
        }
        return dp;
@@ -3818,7 +3822,7 @@ void nfsd4_cleanup_open_state(struct nfsd4_open *open, __be32 status)
        if (open->op_file)
                nfsd4_free_file(open->op_file);
        if (open->op_stp)
-               free_generic_stateid(open->op_stp);
+               nfs4_put_stid(&open->op_stp->st_stid);
 }
 
 __be32
@@ -4266,7 +4270,7 @@ nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
                spin_lock(&cl->cl_lock);
                list_del_init(&dp->dl_recall_lru);
                spin_unlock(&cl->cl_lock);
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(s);
                ret = nfs_ok;
                break;
        default:
@@ -4477,19 +4481,22 @@ static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
        unhash_open_stateid(s);
 
        if (clp->cl_minorversion) {
-               free_generic_stateid(s);
                if (list_empty(&oo->oo_owner.so_stateids))
                        release_openowner(oo);
+               nfs4_put_stid(&s->st_stid);
        } else {
+               /*
+                * In the 4.0 case we need to keep the owners around a
+                * little while to handle CLOSE replay. We still do need
+                * to release any file access that is held by them
+                * before returning however.
+                */
+               release_all_access(s);
                if (s->st_file) {
                        put_nfs4_file(s->st_file);
                        s->st_file = NULL;
                }
                oo->oo_last_closed_stid = s;
-               /*
-                * 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, clp->net);
        }
@@ -5665,7 +5672,7 @@ nfs4_state_shutdown_net(struct net *net)
        list_for_each_safe(pos, next, &reaplist) {
                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
                list_del_init(&dp->dl_recall_lru);
-               nfs4_put_delegation(dp);
+               nfs4_put_stid(&dp->dl_stid);
        }
 
        nfsd4_client_tracking_exit(net);
index 3974773..32c4662 100644 (file)
@@ -85,6 +85,7 @@ struct nfs4_stid {
        unsigned char sc_type;
        stateid_t sc_stateid;
        struct nfs4_client *sc_client;
+       void (*sc_free)(struct nfs4_stid *);
 };
 
 struct nfs4_delegation {
@@ -429,6 +430,7 @@ extern __be32 nfs4_preprocess_stateid_op(struct net *net,
                stateid_t *stateid, int flags, struct file **filp);
 extern void nfs4_lock_state(void);
 extern void nfs4_unlock_state(void);
+void nfs4_put_stid(struct nfs4_stid *s);
 void nfs4_remove_reclaim_record(struct nfs4_client_reclaim *, struct nfsd_net *);
 extern void nfs4_release_reclaim(struct nfsd_net *);
 extern struct nfs4_client_reclaim *nfsd4_find_reclaim_client(const char *recdir,
@@ -446,7 +448,6 @@ extern int nfsd4_create_callback_queue(void);
 extern void nfsd4_destroy_callback_queue(void);
 extern void nfsd4_shutdown_callback(struct nfs4_client *);
 extern void nfsd4_prepare_cb_recall(struct nfs4_delegation *dp);
-extern void nfs4_put_delegation(struct nfs4_delegation *dp);
 extern struct nfs4_client_reclaim *nfs4_client_to_reclaim(const char *name,
                                                        struct nfsd_net *nn);
 extern bool nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn);