]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
NFSv4: Convert LOCK rpc call into an asynchronous RPC call
authorTrond Myklebust <Trond.Myklebust@netapp.com>
Tue, 3 Jan 2006 08:55:17 +0000 (09:55 +0100)
committerTrond Myklebust <Trond.Myklebust@netapp.com>
Fri, 6 Jan 2006 19:58:44 +0000 (14:58 -0500)
 In order to allow users to interrupt/cancel it.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
fs/nfs/nfs4proc.c
fs/nfs/nfs4xdr.c

index 857125705b6fe11ddc3e21c150e6597218cf919c..718fcc7225566a44a18e6a84401b41cd537ee1f5 100644 (file)
@@ -3100,11 +3100,30 @@ static const struct rpc_call_ops nfs4_locku_ops = {
        .rpc_release = nfs4_locku_release_calldata,
 };
 
+static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
+               struct nfs_open_context *ctx,
+               struct nfs4_lock_state *lsp,
+               struct nfs_seqid *seqid)
+{
+       struct nfs4_unlockdata *data;
+       struct rpc_task *task;
+
+       data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
+       if (data == NULL) {
+               nfs_free_seqid(seqid);
+               return ERR_PTR(-ENOMEM);
+       }
+
+       /* Unlock _before_ we do the RPC call */
+       do_vfs_lock(fl->fl_file, fl);
+       task = rpc_run_task(NFS_CLIENT(lsp->ls_state->inode), RPC_TASK_ASYNC, &nfs4_locku_ops, data);
+       if (IS_ERR(task))
+               nfs4_locku_release_calldata(data);
+       return task;
+}
+
 static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
 {
-       struct nfs4_unlockdata *calldata;
-       struct inode *inode = state->inode;
-       struct nfs_server *server = NFS_SERVER(inode);
        struct nfs_seqid *seqid;
        struct nfs4_lock_state *lsp;
        struct rpc_task *task;
@@ -3125,86 +3144,173 @@ static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *
        seqid = nfs_alloc_seqid(&lsp->ls_seqid);
        if (seqid == NULL)
                goto out_unlock;
-       calldata = nfs4_alloc_unlockdata(request, request->fl_file->private_data,
-                       lsp, seqid);
-       if (calldata == NULL)
-               goto out_free_seqid;
-       /* Unlock _before_ we do the RPC call */
-       do_vfs_lock(request->fl_file, request);
-       task = rpc_run_task(server->client, RPC_TASK_ASYNC, &nfs4_locku_ops, calldata);
-       if (!IS_ERR(task)) {
-               status = nfs4_wait_for_completion_rpc_task(task);
-               rpc_release_task(task);
-       } else {
-               status = PTR_ERR(task);
-               nfs4_locku_release_calldata(calldata);
-       }
+       task = nfs4_do_unlck(request, request->fl_file->private_data, lsp, seqid);
+       status = PTR_ERR(task);
+       if (IS_ERR(task))
+               goto out_unlock;
+       status = nfs4_wait_for_completion_rpc_task(task);
+       rpc_release_task(task);
        return status;
-out_free_seqid:
-       nfs_free_seqid(seqid);
 out_unlock:
        do_vfs_lock(request->fl_file, request);
        return status;
 }
 
-static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *request, int reclaim)
+struct nfs4_lockdata {
+       struct nfs_lock_args arg;
+       struct nfs_lock_res res;
+       struct nfs4_lock_state *lsp;
+       struct nfs_open_context *ctx;
+       struct file_lock fl;
+       int rpc_status;
+       int cancelled;
+};
+
+static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
+               struct nfs_open_context *ctx, struct nfs4_lock_state *lsp)
 {
-       struct inode *inode = state->inode;
+       struct nfs4_lockdata *p;
+       struct inode *inode = lsp->ls_state->inode;
        struct nfs_server *server = NFS_SERVER(inode);
-       struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
-       struct nfs_lock_args arg = {
-               .fh = NFS_FH(inode),
-               .fl = request,
-               .lock_stateid = &lsp->ls_stateid,
-               .open_stateid = &state->stateid,
-               .lock_owner = {
-                       .clientid = server->nfs4_state->cl_clientid,
-                       .id = lsp->ls_id,
-               },
-               .block = (IS_SETLKW(cmd)) ? 1 : 0,
-               .reclaim = reclaim,
-       };
-       struct nfs_lock_res res;
+
+       p = kzalloc(sizeof(*p), GFP_KERNEL);
+       if (p == NULL)
+               return NULL;
+
+       p->arg.fh = NFS_FH(inode);
+       p->arg.fl = &p->fl;
+       p->arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
+       if (p->arg.lock_seqid == NULL)
+               goto out_free;
+       p->arg.lock_stateid = &lsp->ls_stateid;
+       p->arg.lock_owner.clientid = server->nfs4_state->cl_clientid;
+       p->arg.lock_owner.id = lsp->ls_id;
+       p->lsp = lsp;
+       atomic_inc(&lsp->ls_count);
+       p->ctx = get_nfs_open_context(ctx);
+       memcpy(&p->fl, fl, sizeof(p->fl));
+       return p;
+out_free:
+       kfree(p);
+       return NULL;
+}
+
+static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
+{
+       struct nfs4_lockdata *data = calldata;
+       struct nfs4_state *state = data->lsp->ls_state;
+       struct nfs4_state_owner *sp = state->owner;
        struct rpc_message msg = {
-               .rpc_proc       = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
-               .rpc_argp       = &arg,
-               .rpc_resp       = &res,
-               .rpc_cred       = state->owner->so_cred,
+               .rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
+               .rpc_argp = &data->arg,
+               .rpc_resp = &data->res,
+               .rpc_cred = sp->so_cred,
        };
-       int status = -ENOMEM;
-
-       arg.lock_seqid = nfs_alloc_seqid(&lsp->ls_seqid);
-       if (arg.lock_seqid == NULL)
-               return -ENOMEM;
-       if (!(lsp->ls_seqid.flags & NFS_SEQID_CONFIRMED)) {
-               struct nfs4_state_owner *owner = state->owner;
 
-               arg.open_seqid = nfs_alloc_seqid(&owner->so_seqid);
-               if (arg.open_seqid == NULL)
+       if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
+               return;
+       dprintk("%s: begin!\n", __FUNCTION__);
+       /* Do we need to do an open_to_lock_owner? */
+       if (!(data->arg.lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)) {
+               data->arg.open_seqid = nfs_alloc_seqid(&sp->so_seqid);
+               if (data->arg.open_seqid == NULL) {
+                       data->rpc_status = -ENOMEM;
+                       task->tk_action = NULL;
                        goto out;
-               arg.new_lock_owner = 1;
-               status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
-               /* increment open seqid on success, and seqid mutating errors */
-               if (arg.new_lock_owner != 0) {
-                       nfs_increment_open_seqid(status, arg.open_seqid);
-                       if (status == 0)
-                               nfs_confirm_seqid(&lsp->ls_seqid, 0);
                }
-               nfs_free_seqid(arg.open_seqid);
-       } else
-               status = rpc_call_sync(server->client, &msg, RPC_TASK_NOINTR);
-       /* increment lock seqid on success, and seqid mutating errors*/
-       nfs_increment_lock_seqid(status, arg.lock_seqid);
-       /* save the returned stateid. */
-       if (status == 0) {
-               memcpy(lsp->ls_stateid.data, res.stateid.data,
-                               sizeof(lsp->ls_stateid.data));
-               lsp->ls_flags |= NFS_LOCK_INITIALIZED;
-       } else if (status == -NFS4ERR_DENIED)
-               status = -EAGAIN;
+               data->arg.open_stateid = &state->stateid;
+               data->arg.new_lock_owner = 1;
+       }
+       rpc_call_setup(task, &msg, 0);
 out:
-       nfs_free_seqid(arg.lock_seqid);
-       return status;
+       dprintk("%s: done!, ret = %d\n", __FUNCTION__, data->rpc_status);
+}
+
+static void nfs4_lock_done(struct rpc_task *task, void *calldata)
+{
+       struct nfs4_lockdata *data = calldata;
+
+       dprintk("%s: begin!\n", __FUNCTION__);
+
+       data->rpc_status = task->tk_status;
+       if (RPC_ASSASSINATED(task))
+               goto out;
+       if (data->arg.new_lock_owner != 0) {
+               nfs_increment_open_seqid(data->rpc_status, data->arg.open_seqid);
+               if (data->rpc_status == 0)
+                       nfs_confirm_seqid(&data->lsp->ls_seqid, 0);
+               else
+                       goto out;
+       }
+       if (data->rpc_status == 0) {
+               memcpy(data->lsp->ls_stateid.data, data->res.stateid.data,
+                                       sizeof(data->lsp->ls_stateid.data));
+               data->lsp->ls_flags |= NFS_LOCK_INITIALIZED;
+       }
+       nfs_increment_lock_seqid(data->rpc_status, data->arg.lock_seqid);
+out:
+       dprintk("%s: done, ret = %d!\n", __FUNCTION__, data->rpc_status);
+}
+
+static void nfs4_lock_release(void *calldata)
+{
+       struct nfs4_lockdata *data = calldata;
+
+       dprintk("%s: begin!\n", __FUNCTION__);
+       if (data->arg.open_seqid != NULL)
+               nfs_free_seqid(data->arg.open_seqid);
+       if (data->cancelled != 0) {
+               struct rpc_task *task;
+               task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
+                               data->arg.lock_seqid);
+               if (!IS_ERR(task))
+                       rpc_release_task(task);
+               dprintk("%s: cancelling lock!\n", __FUNCTION__);
+       } else
+               nfs_free_seqid(data->arg.lock_seqid);
+       nfs4_put_lock_state(data->lsp);
+       put_nfs_open_context(data->ctx);
+       kfree(data);
+       dprintk("%s: done!\n", __FUNCTION__);
+}
+
+static const struct rpc_call_ops nfs4_lock_ops = {
+       .rpc_call_prepare = nfs4_lock_prepare,
+       .rpc_call_done = nfs4_lock_done,
+       .rpc_release = nfs4_lock_release,
+};
+
+static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int reclaim)
+{
+       struct nfs4_lockdata *data;
+       struct rpc_task *task;
+       int ret;
+
+       dprintk("%s: begin!\n", __FUNCTION__);
+       data = nfs4_alloc_lockdata(fl, fl->fl_file->private_data,
+                       fl->fl_u.nfs4_fl.owner);
+       if (data == NULL)
+               return -ENOMEM;
+       if (IS_SETLKW(cmd))
+               data->arg.block = 1;
+       if (reclaim != 0)
+               data->arg.reclaim = 1;
+       task = rpc_run_task(NFS_CLIENT(state->inode), RPC_TASK_ASYNC,
+                       &nfs4_lock_ops, data);
+       if (IS_ERR(task)) {
+               nfs4_lock_release(data);
+               return PTR_ERR(task);
+       }
+       ret = nfs4_wait_for_completion_rpc_task(task);
+       if (ret == 0) {
+               ret = data->rpc_status;
+               if (ret == -NFS4ERR_DENIED)
+                       ret = -EAGAIN;
+       } else
+               data->cancelled = 1;
+       rpc_release_task(task);
+       dprintk("%s: done, ret = %d!\n", __FUNCTION__, ret);
+       return ret;
 }
 
 static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
index a7b5de899c6ddf78fecabac6d1a639b3d6578ab8..5d6bda43dfaa3260066663f16478cc3cced882b2 100644 (file)
@@ -1615,12 +1615,6 @@ static int nfs4_xdr_enc_lock(struct rpc_rqst *req, uint32_t *p, struct nfs_lock_
        };
        int status;
 
-       status = nfs_wait_on_sequence(args->lock_seqid, req->rq_task);
-       if (status != 0)
-               goto out;
-       /* Do we need to do an open_to_lock_owner? */
-       if (args->lock_seqid->sequence->flags & NFS_SEQID_CONFIRMED)
-               args->new_lock_owner = 0;
        xdr_init_encode(&xdr, &req->rq_snd_buf, p);
        encode_compound_hdr(&xdr, &hdr);
        status = encode_putfh(&xdr, args->fh);