Commit e8a7b8b1 authored by Yan, Zheng's avatar Yan, Zheng Committed by Ilya Dryomov
Browse files

ceph: exclude setfilelock requests when calculating oldest tid 



setfilelock requests can block for a long time, which can prevent
client from advancing its oldest tid.

Signed-off-by: default avatarYan, Zheng <zyan@redhat.com>
parent 745a8e3b

fs/ceph/mds_client.c

+23 −7
Original line number Diff line number Diff line
@@ -628,6 +628,9 @@ static void __register_request(struct ceph_mds_client *mdsc, 
	req->r_uid = current_fsuid();

	req->r_gid = current_fsgid();



	if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)

		mdsc->oldest_tid = req->r_tid;



	if (dir) {

		struct ceph_inode_info *ci = ceph_inode(dir);
@@ -643,6 +646,21 @@ static void __unregister_request(struct ceph_mds_client *mdsc, 
				 struct ceph_mds_request *req)

{

	dout("__unregister_request %p tid %lld\n", req, req->r_tid);



	if (req->r_tid == mdsc->oldest_tid) {

		struct rb_node *p = rb_next(&req->r_node);

		mdsc->oldest_tid = 0;

		while (p) {

			struct ceph_mds_request *next_req =

				rb_entry(p, struct ceph_mds_request, r_node);

			if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {

				mdsc->oldest_tid = next_req->r_tid;

				break;

			}

			p = rb_next(p);

		}

	}



	rb_erase(&req->r_node, &mdsc->request_tree);

	RB_CLEAR_NODE(&req->r_node);
@@ -1682,13 +1700,9 @@ static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc) 
			struct ceph_mds_request, r_node);

}



static u64 __get_oldest_tid(struct ceph_mds_client *mdsc)

static inline  u64 __get_oldest_tid(struct ceph_mds_client *mdsc)

{

	struct ceph_mds_request *req = __get_oldest_req(mdsc);



	if (req)

		return req->r_tid;

	return 0;

	return mdsc->oldest_tid;

}



/*
@@ -3378,6 +3392,7 @@ int ceph_mdsc_init(struct ceph_fs_client *fsc) 
	INIT_LIST_HEAD(&mdsc->snap_empty);

	spin_lock_init(&mdsc->snap_empty_lock);

	mdsc->last_tid = 0;

	mdsc->oldest_tid = 0;

	mdsc->request_tree = RB_ROOT;

	INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);

	mdsc->last_renew_caps = jiffies;
@@ -3471,7 +3486,8 @@ static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid) 
			nextreq = rb_entry(n, struct ceph_mds_request, r_node);

		else

			nextreq = NULL;

		if ((req->r_op & CEPH_MDS_OP_WRITE)) {

		if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&

		    (req->r_op & CEPH_MDS_OP_WRITE)) {

			/* write op */

			ceph_mdsc_get_request(req);

			if (nextreq)

fs/ceph/mds_client.h

+2 −0
Original line number Diff line number Diff line
@@ -296,6 +296,8 @@ struct ceph_mds_client { 
	spinlock_t              snap_empty_lock;  /* protect snap_empty */



	u64                    last_tid;      /* most recent mds request */

	u64                    oldest_tid;    /* oldest incomplete mds request,

						 excluding setfilelock requests */

	struct rb_root         request_tree;  /* pending mds requests */

	struct delayed_work    delayed_work;  /* delayed work */

	unsigned long    last_renew_caps;  /* last time we renewed our caps */