libceph: async MON client generic requests
authorIlya Dryomov <idryomov@gmail.com>
Thu, 28 Apr 2016 14:07:27 +0000 (16:07 +0200)
committerIlya Dryomov <idryomov@gmail.com>
Wed, 25 May 2016 23:15:29 +0000 (01:15 +0200)
For map check, we are going to need to send CEPH_MSG_MON_GET_VERSION
messages asynchronously and get a callback on completion.  Refactor MON
client to allow firing off generic requests asynchronously and add an
async variant of ceph_monc_get_version().  ceph_monc_do_statfs() is
switched over and remains sync.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
drivers/block/rbd.c
include/linux/ceph/mon_client.h
net/ceph/mon_client.c

index d0834c477f9684b407fd053359eaba11295c6b9f..8eae6f56194dd02540445130db1598ae9eaf593e 100644 (file)
@@ -4896,8 +4896,8 @@ static int rbd_add_get_pool_id(struct rbd_client *rbdc, const char *pool_name)
 again:
        ret = ceph_pg_poolid_by_name(rbdc->client->osdc.osdmap, pool_name);
        if (ret == -ENOENT && tries++ < 1) {
-               ret = ceph_monc_do_get_version(&rbdc->client->monc, "osdmap",
-                                              &newest_epoch);
+               ret = ceph_monc_get_version(&rbdc->client->monc, "osdmap",
+                                           &newest_epoch);
                if (ret < 0)
                        return ret;
 
index c14e9d861cda31bc54001906cd1fcf7df7913fdb..19800d9b45f3e51544e44754ba44b054d2a2935f 100644 (file)
@@ -39,20 +39,31 @@ struct ceph_mon_request {
        ceph_monc_request_func_t do_request;
 };
 
+typedef void (*ceph_monc_callback_t)(struct ceph_mon_generic_request *);
+
 /*
  * ceph_mon_generic_request is being used for the statfs and
  * mon_get_version requests which are being done a bit differently
  * because we need to get data back to the caller
  */
 struct ceph_mon_generic_request {
+       struct ceph_mon_client *monc;
        struct kref kref;
        u64 tid;
        struct rb_node node;
        int result;
-       void *buf;
+
        struct completion completion;
+       ceph_monc_callback_t complete_cb;
+       u64 private_data;          /* r_tid/linger_id */
+
        struct ceph_msg *request;  /* original request */
        struct ceph_msg *reply;    /* and reply */
+
+       union {
+               struct ceph_statfs *st;
+               u64 newest;
+       } u;
 };
 
 struct ceph_mon_client {
@@ -124,8 +135,10 @@ extern int ceph_monc_wait_osdmap(struct ceph_mon_client *monc, u32 epoch,
 extern int ceph_monc_do_statfs(struct ceph_mon_client *monc,
                               struct ceph_statfs *buf);
 
-extern int ceph_monc_do_get_version(struct ceph_mon_client *monc,
-                                   const char *what, u64 *newest);
+int ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
+                         u64 *newest);
+int ceph_monc_get_version_async(struct ceph_mon_client *monc, const char *what,
+                               ceph_monc_callback_t cb, u64 private_data);
 
 extern int ceph_monc_open_session(struct ceph_mon_client *monc);
 
index 98bfbe1f680784f518df54ad829d19497b2e0e8b..4e49b2296920437d8c13c631e959249ab39d9a46 100644 (file)
@@ -493,6 +493,10 @@ static void release_generic_request(struct kref *kref)
        struct ceph_mon_generic_request *req =
                container_of(kref, struct ceph_mon_generic_request, kref);
 
+       dout("%s greq %p request %p reply %p\n", __func__, req, req->request,
+            req->reply);
+       WARN_ON(!RB_EMPTY_NODE(&req->node));
+
        if (req->reply)
                ceph_msg_put(req->reply);
        if (req->request)
@@ -503,7 +507,8 @@ static void release_generic_request(struct kref *kref)
 
 static void put_generic_request(struct ceph_mon_generic_request *req)
 {
-       kref_put(&req->kref, release_generic_request);
+       if (req)
+               kref_put(&req->kref, release_generic_request);
 }
 
 static void get_generic_request(struct ceph_mon_generic_request *req)
@@ -511,6 +516,103 @@ static void get_generic_request(struct ceph_mon_generic_request *req)
        kref_get(&req->kref);
 }
 
+static struct ceph_mon_generic_request *
+alloc_generic_request(struct ceph_mon_client *monc, gfp_t gfp)
+{
+       struct ceph_mon_generic_request *req;
+
+       req = kzalloc(sizeof(*req), gfp);
+       if (!req)
+               return NULL;
+
+       req->monc = monc;
+       kref_init(&req->kref);
+       RB_CLEAR_NODE(&req->node);
+       init_completion(&req->completion);
+
+       dout("%s greq %p\n", __func__, req);
+       return req;
+}
+
+static void register_generic_request(struct ceph_mon_generic_request *req)
+{
+       struct ceph_mon_client *monc = req->monc;
+
+       WARN_ON(req->tid);
+
+       get_generic_request(req);
+       req->tid = ++monc->last_tid;
+       insert_generic_request(&monc->generic_request_tree, req);
+}
+
+static void send_generic_request(struct ceph_mon_client *monc,
+                                struct ceph_mon_generic_request *req)
+{
+       WARN_ON(!req->tid);
+
+       dout("%s greq %p tid %llu\n", __func__, req, req->tid);
+       req->request->hdr.tid = cpu_to_le64(req->tid);
+       ceph_con_send(&monc->con, ceph_msg_get(req->request));
+}
+
+static void __finish_generic_request(struct ceph_mon_generic_request *req)
+{
+       struct ceph_mon_client *monc = req->monc;
+
+       dout("%s greq %p tid %llu\n", __func__, req, req->tid);
+       erase_generic_request(&monc->generic_request_tree, req);
+
+       ceph_msg_revoke(req->request);
+       ceph_msg_revoke_incoming(req->reply);
+}
+
+static void finish_generic_request(struct ceph_mon_generic_request *req)
+{
+       __finish_generic_request(req);
+       put_generic_request(req);
+}
+
+static void complete_generic_request(struct ceph_mon_generic_request *req)
+{
+       if (req->complete_cb)
+               req->complete_cb(req);
+       else
+               complete_all(&req->completion);
+       put_generic_request(req);
+}
+
+void cancel_generic_request(struct ceph_mon_generic_request *req)
+{
+       struct ceph_mon_client *monc = req->monc;
+       struct ceph_mon_generic_request *lookup_req;
+
+       dout("%s greq %p tid %llu\n", __func__, req, req->tid);
+
+       mutex_lock(&monc->mutex);
+       lookup_req = lookup_generic_request(&monc->generic_request_tree,
+                                           req->tid);
+       if (lookup_req) {
+               WARN_ON(lookup_req != req);
+               finish_generic_request(req);
+       }
+
+       mutex_unlock(&monc->mutex);
+}
+
+static int wait_generic_request(struct ceph_mon_generic_request *req)
+{
+       int ret;
+
+       dout("%s greq %p tid %llu\n", __func__, req, req->tid);
+       ret = wait_for_completion_interruptible(&req->completion);
+       if (ret)
+               cancel_generic_request(req);
+       else
+               ret = req->result; /* completed */
+
+       return ret;
+}
+
 static struct ceph_msg *get_generic_reply(struct ceph_connection *con,
                                         struct ceph_msg_header *hdr,
                                         int *skip)
@@ -540,40 +642,6 @@ static struct ceph_msg *get_generic_reply(struct ceph_connection *con,
        return m;
 }
 
-static int __do_generic_request(struct ceph_mon_client *monc, u64 tid,
-                               struct ceph_mon_generic_request *req)
-{
-       int err;
-
-       /* register request */
-       req->tid = tid != 0 ? tid : ++monc->last_tid;
-       req->request->hdr.tid = cpu_to_le64(req->tid);
-       insert_generic_request(&monc->generic_request_tree, req);
-       ceph_con_send(&monc->con, ceph_msg_get(req->request));
-       mutex_unlock(&monc->mutex);
-
-       err = wait_for_completion_interruptible(&req->completion);
-
-       mutex_lock(&monc->mutex);
-       erase_generic_request(&monc->generic_request_tree, req);
-
-       if (!err)
-               err = req->result;
-       return err;
-}
-
-static int do_generic_request(struct ceph_mon_client *monc,
-                             struct ceph_mon_generic_request *req)
-{
-       int err;
-
-       mutex_lock(&monc->mutex);
-       err = __do_generic_request(monc, 0, req);
-       mutex_unlock(&monc->mutex);
-
-       return err;
-}
-
 /*
  * statfs
  */
@@ -584,22 +652,24 @@ static void handle_statfs_reply(struct ceph_mon_client *monc,
        struct ceph_mon_statfs_reply *reply = msg->front.iov_base;
        u64 tid = le64_to_cpu(msg->hdr.tid);
 
+       dout("%s msg %p tid %llu\n", __func__, msg, tid);
+
        if (msg->front.iov_len != sizeof(*reply))
                goto bad;
-       dout("handle_statfs_reply %p tid %llu\n", msg, tid);
 
        mutex_lock(&monc->mutex);
        req = lookup_generic_request(&monc->generic_request_tree, tid);
-       if (req) {
-               *(struct ceph_statfs *)req->buf = reply->st;
-               req->result = 0;
-               get_generic_request(req);
+       if (!req) {
+               mutex_unlock(&monc->mutex);
+               return;
        }
+
+       req->result = 0;
+       *req->u.st = reply->st; /* struct */
+       __finish_generic_request(req);
        mutex_unlock(&monc->mutex);
-       if (req) {
-               complete_all(&req->completion);
-               put_generic_request(req);
-       }
+
+       complete_generic_request(req);
        return;
 
 bad:
@@ -614,39 +684,38 @@ int ceph_monc_do_statfs(struct ceph_mon_client *monc, struct ceph_statfs *buf)
 {
        struct ceph_mon_generic_request *req;
        struct ceph_mon_statfs *h;
-       int err;
+       int ret = -ENOMEM;
 
-       req = kzalloc(sizeof(*req), GFP_NOFS);
+       req = alloc_generic_request(monc, GFP_NOFS);
        if (!req)
-               return -ENOMEM;
-
-       kref_init(&req->kref);
-       RB_CLEAR_NODE(&req->node);
-       req->buf = buf;
-       init_completion(&req->completion);
+               goto out;
 
-       err = -ENOMEM;
        req->request = ceph_msg_new(CEPH_MSG_STATFS, sizeof(*h), GFP_NOFS,
                                    true);
        if (!req->request)
                goto out;
-       req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 1024, GFP_NOFS,
-                                 true);
+
+       req->reply = ceph_msg_new(CEPH_MSG_STATFS_REPLY, 64, GFP_NOFS, true);
        if (!req->reply)
                goto out;
 
+       req->u.st = buf;
+
+       mutex_lock(&monc->mutex);
+       register_generic_request(req);
        /* fill out request */
        h = req->request->front.iov_base;
        h->monhdr.have_version = 0;
        h->monhdr.session_mon = cpu_to_le16(-1);
        h->monhdr.session_mon_tid = 0;
        h->fsid = monc->monmap->fsid;
+       send_generic_request(monc, req);
+       mutex_unlock(&monc->mutex);
 
-       err = do_generic_request(monc, req);
-
+       ret = wait_generic_request(req);
 out:
        put_generic_request(req);
-       return err;
+       return ret;
 }
 EXPORT_SYMBOL(ceph_monc_do_statfs);
 
@@ -659,7 +728,7 @@ static void handle_get_version_reply(struct ceph_mon_client *monc,
        void *end = p + msg->front_alloc_len;
        u64 handle;
 
-       dout("%s %p tid %llu\n", __func__, msg, tid);
+       dout("%s msg %p tid %llu\n", __func__, msg, tid);
 
        ceph_decode_need(&p, end, 2*sizeof(u64), bad);
        handle = ceph_decode_64(&p);
@@ -668,77 +737,110 @@ static void handle_get_version_reply(struct ceph_mon_client *monc,
 
        mutex_lock(&monc->mutex);
        req = lookup_generic_request(&monc->generic_request_tree, handle);
-       if (req) {
-               *(u64 *)req->buf = ceph_decode_64(&p);
-               req->result = 0;
-               get_generic_request(req);
+       if (!req) {
+               mutex_unlock(&monc->mutex);
+               return;
        }
+
+       req->result = 0;
+       req->u.newest = ceph_decode_64(&p);
+       __finish_generic_request(req);
        mutex_unlock(&monc->mutex);
-       if (req) {
-               complete_all(&req->completion);
-               put_generic_request(req);
-       }
 
+       complete_generic_request(req);
        return;
+
 bad:
        pr_err("corrupt mon_get_version reply, tid %llu\n", tid);
        ceph_msg_dump(msg);
 }
 
-/*
- * Send MMonGetVersion and wait for the reply.
- *
- * @what: one of "mdsmap", "osdmap" or "monmap"
- */
-int ceph_monc_do_get_version(struct ceph_mon_client *monc, const char *what,
-                            u64 *newest)
+static struct ceph_mon_generic_request *
+__ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
+                       ceph_monc_callback_t cb, u64 private_data)
 {
        struct ceph_mon_generic_request *req;
-       void *p, *end;
-       u64 tid;
-       int err;
 
-       req = kzalloc(sizeof(*req), GFP_NOFS);
+       req = alloc_generic_request(monc, GFP_NOIO);
        if (!req)
-               return -ENOMEM;
-
-       kref_init(&req->kref);
-       RB_CLEAR_NODE(&req->node);
-       req->buf = newest;
-       init_completion(&req->completion);
+               goto err_put_req;
 
        req->request = ceph_msg_new(CEPH_MSG_MON_GET_VERSION,
                                    sizeof(u64) + sizeof(u32) + strlen(what),
-                                   GFP_NOFS, true);
-       if (!req->request) {
-               err = -ENOMEM;
-               goto out;
-       }
+                                   GFP_NOIO, true);
+       if (!req->request)
+               goto err_put_req;
 
-       req->reply = ceph_msg_new(CEPH_MSG_MON_GET_VERSION_REPLY, 1024,
-                                 GFP_NOFS, true);
-       if (!req->reply) {
-               err = -ENOMEM;
-               goto out;
-       }
+       req->reply = ceph_msg_new(CEPH_MSG_MON_GET_VERSION_REPLY, 32, GFP_NOIO,
+                                 true);
+       if (!req->reply)
+               goto err_put_req;
 
-       p = req->request->front.iov_base;
-       end = p + req->request->front_alloc_len;
+       req->complete_cb = cb;
+       req->private_data = private_data;
 
-       /* fill out request */
        mutex_lock(&monc->mutex);
-       tid = ++monc->last_tid;
-       ceph_encode_64(&p, tid); /* handle */
-       ceph_encode_string(&p, end, what, strlen(what));
+       register_generic_request(req);
+       {
+               void *p = req->request->front.iov_base;
+               void *const end = p + req->request->front_alloc_len;
+
+               ceph_encode_64(&p, req->tid); /* handle */
+               ceph_encode_string(&p, end, what, strlen(what));
+               WARN_ON(p != end);
+       }
+       send_generic_request(monc, req);
+       mutex_unlock(&monc->mutex);
 
-       err = __do_generic_request(monc, tid, req);
+       return req;
 
-       mutex_unlock(&monc->mutex);
-out:
+err_put_req:
        put_generic_request(req);
-       return err;
+       return ERR_PTR(-ENOMEM);
+}
+
+/*
+ * Send MMonGetVersion and wait for the reply.
+ *
+ * @what: one of "mdsmap", "osdmap" or "monmap"
+ */
+int ceph_monc_get_version(struct ceph_mon_client *monc, const char *what,
+                         u64 *newest)
+{
+       struct ceph_mon_generic_request *req;
+       int ret;
+
+       req = __ceph_monc_get_version(monc, what, NULL, 0);
+       if (IS_ERR(req))
+               return PTR_ERR(req);
+
+       ret = wait_generic_request(req);
+       if (!ret)
+               *newest = req->u.newest;
+
+       put_generic_request(req);
+       return ret;
+}
+EXPORT_SYMBOL(ceph_monc_get_version);
+
+/*
+ * Send MMonGetVersion,
+ *
+ * @what: one of "mdsmap", "osdmap" or "monmap"
+ */
+int ceph_monc_get_version_async(struct ceph_mon_client *monc, const char *what,
+                               ceph_monc_callback_t cb, u64 private_data)
+{
+       struct ceph_mon_generic_request *req;
+
+       req = __ceph_monc_get_version(monc, what, cb, private_data);
+       if (IS_ERR(req))
+               return PTR_ERR(req);
+
+       put_generic_request(req);
+       return 0;
 }
-EXPORT_SYMBOL(ceph_monc_do_get_version);
+EXPORT_SYMBOL(ceph_monc_get_version_async);
 
 /*
  * Resend pending generic requests.
@@ -923,6 +1025,8 @@ void ceph_monc_stop(struct ceph_mon_client *monc)
 
        ceph_auth_destroy(monc->auth);
 
+       WARN_ON(!RB_EMPTY_ROOT(&monc->generic_request_tree));
+
        ceph_msg_put(monc->m_auth);
        ceph_msg_put(monc->m_auth_reply);
        ceph_msg_put(monc->m_subscribe);