NFS: Convert lookups of the open context to RCU

Reduce contention on the inode->i_lock by ensuring that we use RCU
when looking up the NFS open context.

Signed-off-by: Trond Myklebust <[e-mail]>

fs/nfs/delegation.c

···
 	int err;
 
 again:
-	spin_lock(&inode->i_lock);
-	list_for_each_entry(ctx, &nfsi->open_files, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
 		state = ctx->state;
 		if (state == NULL)
 			continue;
···
 			continue;
 		if (!nfs4_stateid_match(&state->stateid, stateid))
 			continue;
-		get_nfs_open_context(ctx);
-		spin_unlock(&inode->i_lock);
+		if (!get_nfs_open_context(ctx))
+			continue;
+		rcu_read_unlock();
 		sp = state->owner;
 		/* Block nfs4_proc_unlck */
 		mutex_lock(&sp->so_delegreturn_mutex);
···
 			return err;
 		goto again;
 	}
-	spin_unlock(&inode->i_lock);
+	rcu_read_unlock();
 	return 0;
 }
 

fs/nfs/inode.c

···
 
 struct nfs_open_context *get_nfs_open_context(struct nfs_open_context *ctx)
 {
-	if (ctx != NULL)
-		refcount_inc(&ctx->lock_context.count);
-	return ctx;
+	if (ctx != NULL && refcount_inc_not_zero(&ctx->lock_context.count))
+		return ctx;
+	return NULL;
 }
 EXPORT_SYMBOL_GPL(get_nfs_open_context);
 
···
 	struct inode *inode = d_inode(ctx->dentry);
 	struct super_block *sb = ctx->dentry->d_sb;
 
-	if (!list_empty(&ctx->list)) {
-		if (!refcount_dec_and_lock(&ctx->lock_context.count, &inode->i_lock))
-			return;
-		list_del(&ctx->list);
-		spin_unlock(&inode->i_lock);
-	} else if (!refcount_dec_and_test(&ctx->lock_context.count))
+	if (!refcount_dec_and_test(&ctx->lock_context.count))
 		return;
+	if (!list_empty(&ctx->list)) {
+		spin_lock(&inode->i_lock);
+		list_del_rcu(&ctx->list);
+		spin_unlock(&inode->i_lock);
+	}
 	if (inode != NULL)
 		NFS_PROTO(inode)->close_context(ctx, is_sync);
 	if (ctx->cred != NULL)
···
 	dput(ctx->dentry);
 	nfs_sb_deactive(sb);
 	kfree(ctx->mdsthreshold);
-	kfree(ctx);
+	kfree_rcu(ctx, rcu_head);
 }
 
 void put_nfs_open_context(struct nfs_open_context *ctx)
···
 	struct nfs_inode *nfsi = NFS_I(inode);
 
 	spin_lock(&inode->i_lock);
-	if (ctx->mode & FMODE_WRITE)
-		list_add(&ctx->list, &nfsi->open_files);
-	else
-		list_add_tail(&ctx->list, &nfsi->open_files);
+	list_add_tail_rcu(&ctx->list, &nfsi->open_files);
 	spin_unlock(&inode->i_lock);
 }
 EXPORT_SYMBOL_GPL(nfs_inode_attach_open_context);
···
 	struct nfs_inode *nfsi = NFS_I(inode);
 	struct nfs_open_context *pos, *ctx = NULL;
 
-	spin_lock(&inode->i_lock);
-	list_for_each_entry(pos, &nfsi->open_files, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(pos, &nfsi->open_files, list) {
 		if (cred != NULL && pos->cred != cred)
 			continue;
 		if ((pos->mode & (FMODE_READ|FMODE_WRITE)) != mode)
 			continue;
 		ctx = get_nfs_open_context(pos);
-		break;
+		if (ctx)
+			break;
 	}
-	spin_unlock(&inode->i_lock);
+	rcu_read_unlock();
 	return ctx;
 }
 
···
 		if (ctx->error < 0)
 			invalidate_inode_pages2(inode->i_mapping);
 		filp->private_data = NULL;
-		spin_lock(&inode->i_lock);
-		list_move_tail(&ctx->list, &NFS_I(inode)->open_files);
-		spin_unlock(&inode->i_lock);
 		put_nfs_open_context_sync(ctx);
 	}
 }

fs/nfs/nfs4proc.c

···
 	return ret;
 }
 
-static struct nfs_open_context *nfs4_state_find_open_context(struct nfs4_state *state)
+static struct nfs_open_context *
+nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
 {
 	struct nfs_inode *nfsi = NFS_I(state->inode);
 	struct nfs_open_context *ctx;
 
-	spin_lock(&state->inode->i_lock);
-	list_for_each_entry(ctx, &nfsi->open_files, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
 		if (ctx->state != state)
 			continue;
-		get_nfs_open_context(ctx);
-		spin_unlock(&state->inode->i_lock);
+		if ((ctx->mode & mode) != mode)
+			continue;
+		if (!get_nfs_open_context(ctx))
+			continue;
+		rcu_read_unlock();
 		return ctx;
 	}
-	spin_unlock(&state->inode->i_lock);
+	rcu_read_unlock();
 	return ERR_PTR(-ENOENT);
+}
+
+static struct nfs_open_context *
+nfs4_state_find_open_context(struct nfs4_state *state)
+{
+	struct nfs_open_context *ctx;
+
+	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
+	if (!IS_ERR(ctx))
+		return ctx;
+	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
+	if (!IS_ERR(ctx))
+		return ctx;
+	return nfs4_state_find_open_context_mode(state, FMODE_READ);
 }
 
 static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,

fs/nfs/nfs4state.c

···
 	struct nfs4_state *state;
 	bool found = false;
 
-	spin_lock(&inode->i_lock);
-	list_for_each_entry(ctx, &nfsi->open_files, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
 		state = ctx->state;
 		if (state == NULL)
 			continue;
···
 		    nfs4_state_mark_reclaim_nograce(clp, state))
 			found = true;
 	}
-	spin_unlock(&inode->i_lock);
+	rcu_read_unlock();
 
 	nfs_inode_find_delegation_state_and_recover(inode, stateid);
 	if (found)
···
 	struct nfs_inode *nfsi = NFS_I(inode);
 	struct nfs_open_context *ctx;
 
-	spin_lock(&inode->i_lock);
-	list_for_each_entry(ctx, &nfsi->open_files, list) {
+	rcu_read_lock();
+	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
 		if (ctx->state != state)
 			continue;
 		set_bit(NFS_CONTEXT_BAD, &ctx->flags);
 	}
-	spin_unlock(&inode->i_lock);
+	rcu_read_unlock();
 }
 
 static void nfs4_state_mark_recovery_failed(struct nfs4_state *state, int error)

fs/nfs/pnfs.c

···
 	if (!nfs_have_layout(ino))
 		return false;
 retry:
+	rcu_read_lock();
 	spin_lock(&ino->i_lock);
 	lo = nfsi->layout;
 	if (!lo || !pnfs_layout_is_valid(lo) ||
···
 	pnfs_get_layout_hdr(lo);
 	if (test_bit(NFS_LAYOUT_RETURN_LOCK, &lo->plh_flags)) {
 		spin_unlock(&ino->i_lock);
+		rcu_read_unlock();
 		wait_on_bit(&lo->plh_flags, NFS_LAYOUT_RETURN,
 				TASK_UNINTERRUPTIBLE);
 		pnfs_put_layout_hdr(lo);
···
 		skip_read = true;
 	}
 
-	list_for_each_entry(ctx, &nfsi->open_files, list) {
+	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
 		state = ctx->state;
 		if (state == NULL)
 			continue;
···
 
 out_noroc:
 	spin_unlock(&ino->i_lock);
+	rcu_read_unlock();
 	pnfs_layoutcommit_inode(ino, true);
 	if (roc) {
 		struct pnfs_layoutdriver_type *ld = NFS_SERVER(ino)->pnfs_curr_ld;

include/linux/nfs_fs.h

···
 
 	struct list_head list;
 	struct nfs4_threshold	*mdsthreshold;
+	struct rcu_head	rcu_head;
 };
 
 struct nfs_open_dir_context {