From b09734b1f4abd86e046777f0f268215b4ef1b523 Mon Sep 17 00:00:00 2001 From: Tommi Virtanen Date: Wed, 2 Feb 2011 11:39:32 -0800 Subject: libceph: Fix base64-decoding when input ends in newline. It used to return -EINVAL because it thought the end was not aligned to 4 bytes. Clean up superfluous src < end test in if, the while itself guarantees that. Signed-off-by: Tommi Virtanen Signed-off-by: Sage Weil --- net/ceph/armor.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'net') diff --git a/net/ceph/armor.c b/net/ceph/armor.c index eb2a666b0be..1fc1ee11dfa 100644 --- a/net/ceph/armor.c +++ b/net/ceph/armor.c @@ -78,8 +78,10 @@ int ceph_unarmor(char *dst, const char *src, const char *end) while (src < end) { int a, b, c, d; - if (src < end && src[0] == '\n') + if (src[0] == '\n') { src++; + continue; + } if (src + 4 > end) return -EINVAL; a = decode_bits(src[0]); -- cgit v1.2.3-18-g5258 From 6f6c7006755b667f9f6c1f3b6f08cd65f75cc471 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Mon, 17 Jan 2011 20:34:08 -0800 Subject: libceph: fix osd request queuing on osdmap updates If we send a request to osd A, and the request's pg remaps to osd B and then back to A in quick succession, we need to resend the request to A. The old code was only calling kick_requests after processing all incremental maps in a message, so it was very possible to not resend a request that needed to be resent. This would make the osd eventually time out (at least with the current default of osd timeouts enabled). The correct approach is to scan requests on every map incremental. This patch refactors the kick code in a few ways: - all requests are either on req_lru (in flight), req_unsent (ready to send), or req_notarget (currently map to no up osd) - mapping always done by map_request (previous map_osds) - if the mapping changes, we requeue. requests are resent only after all map incrementals are processed. - some osd reset code is moved out of kick_requests into a separate function - the "kick this osd" functionality is moved to kick_osd_requests, as it is unrelated to scanning for request->pg->osd mapping changes Signed-off-by: Sage Weil --- net/ceph/osd_client.c | 255 ++++++++++++++++++++++++-------------------------- 1 file changed, 122 insertions(+), 133 deletions(-) (limited to 'net') diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index 3e20a122ffa..b85ed5a5503 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -22,10 +22,9 @@ #define OSD_OPREPLY_FRONT_LEN 512 static const struct ceph_connection_operations osd_con_ops; -static int __kick_requests(struct ceph_osd_client *osdc, - struct ceph_osd *kickosd); -static void kick_requests(struct ceph_osd_client *osdc, struct ceph_osd *osd); +static void send_queued(struct ceph_osd_client *osdc); +static int __reset_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd); static int op_needs_trail(int op) { @@ -529,6 +528,35 @@ __lookup_request_ge(struct ceph_osd_client *osdc, return NULL; } +/* + * Resubmit requests pending on the given osd. + */ +static void __kick_osd_requests(struct ceph_osd_client *osdc, + struct ceph_osd *osd) +{ + struct ceph_osd_request *req; + int err; + + dout("__kick_osd_requests osd%d\n", osd->o_osd); + err = __reset_osd(osdc, osd); + if (err == -EAGAIN) + return; + + list_for_each_entry(req, &osd->o_requests, r_osd_item) { + list_move(&req->r_req_lru_item, &osdc->req_unsent); + dout("requeued %p tid %llu osd%d\n", req, req->r_tid, + osd->o_osd); + req->r_flags |= CEPH_OSD_FLAG_RETRY; + } +} + +static void kick_osd_requests(struct ceph_osd_client *osdc, + struct ceph_osd *kickosd) +{ + mutex_lock(&osdc->request_mutex); + __kick_osd_requests(osdc, kickosd); + mutex_unlock(&osdc->request_mutex); +} /* * If the osd connection drops, we need to resubmit all requests. @@ -543,7 +571,8 @@ static void osd_reset(struct ceph_connection *con) dout("osd_reset osd%d\n", osd->o_osd); osdc = osd->o_osdc; down_read(&osdc->map_sem); - kick_requests(osdc, osd); + kick_osd_requests(osdc, osd); + send_queued(osdc); up_read(&osdc->map_sem); } @@ -781,20 +810,20 @@ static void __cancel_request(struct ceph_osd_request *req) ceph_con_revoke(&req->r_osd->o_con, req->r_request); req->r_sent = 0; } - list_del_init(&req->r_req_lru_item); } /* * Pick an osd (the first 'up' osd in the pg), allocate the osd struct * (as needed), and set the request r_osd appropriately. If there is - * no up osd, set r_osd to NULL. + * no up osd, set r_osd to NULL. Move the request to the appropiate list + * (unsent, homeless) or leave on in-flight lru. * * Return 0 if unchanged, 1 if changed, or negative on error. * * Caller should hold map_sem for read and request_mutex. */ -static int __map_osds(struct ceph_osd_client *osdc, - struct ceph_osd_request *req) +static int __map_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) { struct ceph_osd_request_head *reqhead = req->r_request->front.iov_base; struct ceph_pg pgid; @@ -802,11 +831,13 @@ static int __map_osds(struct ceph_osd_client *osdc, int o = -1, num = 0; int err; - dout("map_osds %p tid %lld\n", req, req->r_tid); + dout("map_request %p tid %lld\n", req, req->r_tid); err = ceph_calc_object_layout(&reqhead->layout, req->r_oid, &req->r_file_layout, osdc->osdmap); - if (err) + if (err) { + list_move(&req->r_req_lru_item, &osdc->req_notarget); return err; + } pgid = reqhead->layout.ol_pgid; req->r_pgid = pgid; @@ -823,7 +854,7 @@ static int __map_osds(struct ceph_osd_client *osdc, (req->r_osd == NULL && o == -1)) return 0; /* no change */ - dout("map_osds tid %llu pgid %d.%x osd%d (was osd%d)\n", + dout("map_request tid %llu pgid %d.%x osd%d (was osd%d)\n", req->r_tid, le32_to_cpu(pgid.pool), le16_to_cpu(pgid.ps), o, req->r_osd ? req->r_osd->o_osd : -1); @@ -841,10 +872,12 @@ static int __map_osds(struct ceph_osd_client *osdc, if (!req->r_osd && o >= 0) { err = -ENOMEM; req->r_osd = create_osd(osdc); - if (!req->r_osd) + if (!req->r_osd) { + list_move(&req->r_req_lru_item, &osdc->req_notarget); goto out; + } - dout("map_osds osd %p is osd%d\n", req->r_osd, o); + dout("map_request osd %p is osd%d\n", req->r_osd, o); req->r_osd->o_osd = o; req->r_osd->o_con.peer_name.num = cpu_to_le64(o); __insert_osd(osdc, req->r_osd); @@ -855,6 +888,9 @@ static int __map_osds(struct ceph_osd_client *osdc, if (req->r_osd) { __remove_osd_from_lru(req->r_osd); list_add(&req->r_osd_item, &req->r_osd->o_requests); + list_move(&req->r_req_lru_item, &osdc->req_unsent); + } else { + list_move(&req->r_req_lru_item, &osdc->req_notarget); } err = 1; /* osd or pg changed */ @@ -869,16 +905,6 @@ static int __send_request(struct ceph_osd_client *osdc, struct ceph_osd_request *req) { struct ceph_osd_request_head *reqhead; - int err; - - err = __map_osds(osdc, req); - if (err < 0) - return err; - if (req->r_osd == NULL) { - dout("send_request %p no up osds in pg\n", req); - ceph_monc_request_next_osdmap(&osdc->client->monc); - return 0; - } dout("send_request %p tid %llu to osd%d flags %d\n", req, req->r_tid, req->r_osd->o_osd, req->r_flags); @@ -897,6 +923,21 @@ static int __send_request(struct ceph_osd_client *osdc, return 0; } +/* + * Send any requests in the queue (req_unsent). + */ +static void send_queued(struct ceph_osd_client *osdc) +{ + struct ceph_osd_request *req, *tmp; + + dout("send_queued\n"); + mutex_lock(&osdc->request_mutex); + list_for_each_entry_safe(req, tmp, &osdc->req_unsent, r_req_lru_item) { + __send_request(osdc, req); + } + mutex_unlock(&osdc->request_mutex); +} + /* * Timeout callback, called every N seconds when 1 or more osd * requests has been active for more than N seconds. When this @@ -916,7 +957,6 @@ static void handle_timeout(struct work_struct *work) unsigned long keepalive = osdc->client->options->osd_keepalive_timeout * HZ; unsigned long last_stamp = 0; - struct rb_node *p; struct list_head slow_osds; dout("timeout\n"); @@ -925,21 +965,6 @@ static void handle_timeout(struct work_struct *work) ceph_monc_request_next_osdmap(&osdc->client->monc); mutex_lock(&osdc->request_mutex); - for (p = rb_first(&osdc->requests); p; p = rb_next(p)) { - req = rb_entry(p, struct ceph_osd_request, r_node); - - if (req->r_resend) { - int err; - - dout("osdc resending prev failed %lld\n", req->r_tid); - err = __send_request(osdc, req); - if (err) - dout("osdc failed again on %lld\n", req->r_tid); - else - req->r_resend = false; - continue; - } - } /* * reset osds that appear to be _really_ unresponsive. this @@ -963,7 +988,7 @@ static void handle_timeout(struct work_struct *work) BUG_ON(!osd); pr_warning(" tid %llu timed out on osd%d, will reset osd\n", req->r_tid, osd->o_osd); - __kick_requests(osdc, osd); + __kick_osd_requests(osdc, osd); } /* @@ -991,7 +1016,7 @@ static void handle_timeout(struct work_struct *work) __schedule_osd_timeout(osdc); mutex_unlock(&osdc->request_mutex); - + send_queued(osdc); up_read(&osdc->map_sem); } @@ -1109,108 +1134,61 @@ bad: ceph_msg_dump(msg); } - -static int __kick_requests(struct ceph_osd_client *osdc, - struct ceph_osd *kickosd) +static void reset_changed_osds(struct ceph_osd_client *osdc) { - struct ceph_osd_request *req; struct rb_node *p, *n; - int needmap = 0; - int err; - - dout("kick_requests osd%d\n", kickosd ? kickosd->o_osd : -1); - if (kickosd) { - err = __reset_osd(osdc, kickosd); - if (err == -EAGAIN) - return 1; - } else { - for (p = rb_first(&osdc->osds); p; p = n) { - struct ceph_osd *osd = - rb_entry(p, struct ceph_osd, o_node); - - n = rb_next(p); - if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) || - memcmp(&osd->o_con.peer_addr, - ceph_osd_addr(osdc->osdmap, - osd->o_osd), - sizeof(struct ceph_entity_addr)) != 0) - __reset_osd(osdc, osd); - } - } - - for (p = rb_first(&osdc->requests); p; p = rb_next(p)) { - req = rb_entry(p, struct ceph_osd_request, r_node); - - if (req->r_resend) { - dout(" r_resend set on tid %llu\n", req->r_tid); - __cancel_request(req); - goto kick; - } - if (req->r_osd && kickosd == req->r_osd) { - __cancel_request(req); - goto kick; - } - err = __map_osds(osdc, req); - if (err == 0) - continue; /* no change */ - if (err < 0) { - /* - * FIXME: really, we should set the request - * error and fail if this isn't a 'nofail' - * request, but that's a fair bit more - * complicated to do. So retry! - */ - dout(" setting r_resend on %llu\n", req->r_tid); - req->r_resend = true; - continue; - } - if (req->r_osd == NULL) { - dout("tid %llu maps to no valid osd\n", req->r_tid); - needmap++; /* request a newer map */ - continue; - } + for (p = rb_first(&osdc->osds); p; p = n) { + struct ceph_osd *osd = rb_entry(p, struct ceph_osd, o_node); -kick: - dout("kicking %p tid %llu osd%d\n", req, req->r_tid, - req->r_osd ? req->r_osd->o_osd : -1); - req->r_flags |= CEPH_OSD_FLAG_RETRY; - err = __send_request(osdc, req); - if (err) { - dout(" setting r_resend on %llu\n", req->r_tid); - req->r_resend = true; - } + n = rb_next(p); + if (!ceph_osd_is_up(osdc->osdmap, osd->o_osd) || + memcmp(&osd->o_con.peer_addr, + ceph_osd_addr(osdc->osdmap, + osd->o_osd), + sizeof(struct ceph_entity_addr)) != 0) + __reset_osd(osdc, osd); } - - return needmap; } /* - * Resubmit osd requests whose osd or osd address has changed. Request - * a new osd map if osds are down, or we are otherwise unable to determine - * how to direct a request. - * - * Close connections to down osds. - * - * If @who is specified, resubmit requests for that specific osd. + * Requeue requests whose mapping to an OSD has changed. If requests map to + * no osd, request a new map. * * Caller should hold map_sem for read and request_mutex. */ -static void kick_requests(struct ceph_osd_client *osdc, - struct ceph_osd *kickosd) +static void kick_requests(struct ceph_osd_client *osdc) { - int needmap; + struct ceph_osd_request *req; + struct rb_node *p; + int needmap = 0; + int err; + dout("kick_requests\n"); mutex_lock(&osdc->request_mutex); - needmap = __kick_requests(osdc, kickosd); + for (p = rb_first(&osdc->requests); p; p = rb_next(p)) { + req = rb_entry(p, struct ceph_osd_request, r_node); + err = __map_request(osdc, req); + if (err < 0) + continue; /* error */ + if (req->r_osd == NULL) { + dout("%p tid %llu maps to no osd\n", req, req->r_tid); + needmap++; /* request a newer map */ + } else if (err > 0) { + dout("%p tid %llu requeued on osd%d\n", req, req->r_tid, + req->r_osd ? req->r_osd->o_osd : -1); + req->r_flags |= CEPH_OSD_FLAG_RETRY; + } + } mutex_unlock(&osdc->request_mutex); if (needmap) { dout("%d requests for down osds, need new map\n", needmap); ceph_monc_request_next_osdmap(&osdc->client->monc); } - } + + /* * Process updated osd map. * @@ -1263,6 +1241,8 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg) ceph_osdmap_destroy(osdc->osdmap); osdc->osdmap = newmap; } + kick_requests(osdc); + reset_changed_osds(osdc); } else { dout("ignoring incremental map %u len %d\n", epoch, maplen); @@ -1300,6 +1280,7 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg) osdc->osdmap = newmap; if (oldmap) ceph_osdmap_destroy(oldmap); + kick_requests(osdc); } p += maplen; nr_maps--; @@ -1308,8 +1289,7 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg) done: downgrade_write(&osdc->map_sem); ceph_monc_got_osdmap(&osdc->client->monc, osdc->osdmap->epoch); - if (newmap) - kick_requests(osdc, NULL); + send_queued(osdc); up_read(&osdc->map_sem); wake_up_all(&osdc->client->auth_wq); return; @@ -1347,15 +1327,22 @@ int ceph_osdc_start_request(struct ceph_osd_client *osdc, * the request still han't been touched yet. */ if (req->r_sent == 0) { - rc = __send_request(osdc, req); - if (rc) { - if (nofail) { - dout("osdc_start_request failed send, " - " marking %lld\n", req->r_tid); - req->r_resend = true; - rc = 0; - } else { - __unregister_request(osdc, req); + rc = __map_request(osdc, req); + if (rc < 0) + return rc; + if (req->r_osd == NULL) { + dout("send_request %p no up osds in pg\n", req); + ceph_monc_request_next_osdmap(&osdc->client->monc); + } else { + rc = __send_request(osdc, req); + if (rc) { + if (nofail) { + dout("osdc_start_request failed send, " + " will retry %lld\n", req->r_tid); + rc = 0; + } else { + __unregister_request(osdc, req); + } } } } @@ -1441,6 +1428,8 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) INIT_LIST_HEAD(&osdc->osd_lru); osdc->requests = RB_ROOT; INIT_LIST_HEAD(&osdc->req_lru); + INIT_LIST_HEAD(&osdc->req_unsent); + INIT_LIST_HEAD(&osdc->req_notarget); osdc->num_requests = 0; INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout); INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout); -- cgit v1.2.3-18-g5258 From a40c4f10e3fb96030358e49abd010c1f08446fa3 Mon Sep 17 00:00:00 2001 From: Yehuda Sadeh Date: Mon, 21 Mar 2011 15:07:16 -0700 Subject: libceph: add lingering request and watch/notify event framework Lingering requests are requests that are sent to the OSD normally but tracked also after we get a successful request. This keeps the OSD connection open and resends the original request if the object moves to another OSD. The OSD can then send notification messages back to us if another client initiates a notify. This framework will be used by RBD so that the client gets notification when a snapshot is created by another node or tool. Signed-off-by: Yehuda Sadeh Signed-off-by: Sage Weil --- net/ceph/ceph_common.c | 1 + net/ceph/osd_client.c | 385 +++++++++++++++++++++++++++++++++++++++++++++++-- 2 files changed, 374 insertions(+), 12 deletions(-) (limited to 'net') diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c index f3e4a13fea0..95f96ab94bb 100644 --- a/net/ceph/ceph_common.c +++ b/net/ceph/ceph_common.c @@ -62,6 +62,7 @@ const char *ceph_msg_type_name(int type) case CEPH_MSG_OSD_MAP: return "osd_map"; case CEPH_MSG_OSD_OP: return "osd_op"; case CEPH_MSG_OSD_OPREPLY: return "osd_opreply"; + case CEPH_MSG_WATCH_NOTIFY: return "watch_notify"; default: return "unknown"; } } diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index b85ed5a5503..02212ed5085 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -25,6 +25,12 @@ static const struct ceph_connection_operations osd_con_ops; static void send_queued(struct ceph_osd_client *osdc); static int __reset_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd); +static void __register_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req); +static void __unregister_linger_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req); +static int __send_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req); static int op_needs_trail(int op) { @@ -33,6 +39,7 @@ static int op_needs_trail(int op) case CEPH_OSD_OP_SETXATTR: case CEPH_OSD_OP_CMPXATTR: case CEPH_OSD_OP_CALL: + case CEPH_OSD_OP_NOTIFY: return 1; default: return 0; @@ -208,6 +215,8 @@ struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc, init_completion(&req->r_completion); init_completion(&req->r_safe_completion); INIT_LIST_HEAD(&req->r_unsafe_item); + INIT_LIST_HEAD(&req->r_linger_item); + INIT_LIST_HEAD(&req->r_linger_osd); req->r_flags = flags; WARN_ON((flags & (CEPH_OSD_FLAG_READ|CEPH_OSD_FLAG_WRITE)) == 0); @@ -314,6 +323,24 @@ static void osd_req_encode_op(struct ceph_osd_request *req, break; case CEPH_OSD_OP_STARTSYNC: break; + case CEPH_OSD_OP_NOTIFY: + { + __le32 prot_ver = cpu_to_le32(src->watch.prot_ver); + __le32 timeout = cpu_to_le32(src->watch.timeout); + + BUG_ON(!req->r_trail); + + ceph_pagelist_append(req->r_trail, + &prot_ver, sizeof(prot_ver)); + ceph_pagelist_append(req->r_trail, + &timeout, sizeof(timeout)); + } + case CEPH_OSD_OP_NOTIFY_ACK: + case CEPH_OSD_OP_WATCH: + dst->watch.cookie = cpu_to_le64(src->watch.cookie); + dst->watch.ver = cpu_to_le64(src->watch.ver); + dst->watch.flag = src->watch.flag; + break; default: pr_err("unrecognized osd opcode %d\n", dst->op); WARN_ON(1); @@ -534,7 +561,7 @@ __lookup_request_ge(struct ceph_osd_client *osdc, static void __kick_osd_requests(struct ceph_osd_client *osdc, struct ceph_osd *osd) { - struct ceph_osd_request *req; + struct ceph_osd_request *req, *nreq; int err; dout("__kick_osd_requests osd%d\n", osd->o_osd); @@ -546,7 +573,17 @@ static void __kick_osd_requests(struct ceph_osd_client *osdc, list_move(&req->r_req_lru_item, &osdc->req_unsent); dout("requeued %p tid %llu osd%d\n", req, req->r_tid, osd->o_osd); - req->r_flags |= CEPH_OSD_FLAG_RETRY; + if (!req->r_linger) + req->r_flags |= CEPH_OSD_FLAG_RETRY; + } + + list_for_each_entry_safe(req, nreq, &osd->o_linger_requests, + r_linger_osd) { + __unregister_linger_request(osdc, req); + __register_request(osdc, req); + list_move(&req->r_req_lru_item, &osdc->req_unsent); + dout("requeued lingering %p tid %llu osd%d\n", req, req->r_tid, + osd->o_osd); } } @@ -590,6 +627,7 @@ static struct ceph_osd *create_osd(struct ceph_osd_client *osdc) atomic_set(&osd->o_ref, 1); osd->o_osdc = osdc; INIT_LIST_HEAD(&osd->o_requests); + INIT_LIST_HEAD(&osd->o_linger_requests); INIT_LIST_HEAD(&osd->o_osd_lru); osd->o_incarnation = 1; @@ -679,7 +717,8 @@ static int __reset_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd) int ret = 0; dout("__reset_osd %p osd%d\n", osd, osd->o_osd); - if (list_empty(&osd->o_requests)) { + if (list_empty(&osd->o_requests) && + list_empty(&osd->o_linger_requests)) { __remove_osd(osdc, osd); } else if (memcmp(&osdc->osdmap->osd_addr[osd->o_osd], &osd->o_con.peer_addr, @@ -752,10 +791,9 @@ static void __cancel_osd_timeout(struct ceph_osd_client *osdc) * Register request, assign tid. If this is the first request, set up * the timeout event. */ -static void register_request(struct ceph_osd_client *osdc, - struct ceph_osd_request *req) +static void __register_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) { - mutex_lock(&osdc->request_mutex); req->r_tid = ++osdc->last_tid; req->r_request->hdr.tid = cpu_to_le64(req->r_tid); INIT_LIST_HEAD(&req->r_req_lru_item); @@ -769,6 +807,13 @@ static void register_request(struct ceph_osd_client *osdc, dout(" first request, scheduling timeout\n"); __schedule_osd_timeout(osdc); } +} + +static void register_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) +{ + mutex_lock(&osdc->request_mutex); + __register_request(osdc, req); mutex_unlock(&osdc->request_mutex); } @@ -787,9 +832,14 @@ static void __unregister_request(struct ceph_osd_client *osdc, ceph_con_revoke(&req->r_osd->o_con, req->r_request); list_del_init(&req->r_osd_item); - if (list_empty(&req->r_osd->o_requests)) + if (list_empty(&req->r_osd->o_requests) && + list_empty(&req->r_osd->o_linger_requests)) { + dout("moving osd to %p lru\n", req->r_osd); __move_osd_to_lru(osdc, req->r_osd); - req->r_osd = NULL; + } + if (list_empty(&req->r_osd_item) && + list_empty(&req->r_linger_item)) + req->r_osd = NULL; } ceph_osdc_put_request(req); @@ -812,6 +862,58 @@ static void __cancel_request(struct ceph_osd_request *req) } } +static void __register_linger_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) +{ + dout("__register_linger_request %p\n", req); + list_add_tail(&req->r_linger_item, &osdc->req_linger); + list_add_tail(&req->r_linger_osd, &req->r_osd->o_linger_requests); +} + +static void __unregister_linger_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) +{ + dout("__unregister_linger_request %p\n", req); + if (req->r_osd) { + list_del_init(&req->r_linger_item); + list_del_init(&req->r_linger_osd); + + if (list_empty(&req->r_osd->o_requests) && + list_empty(&req->r_osd->o_linger_requests)) { + dout("moving osd to %p lru\n", req->r_osd); + __move_osd_to_lru(osdc, req->r_osd); + } + req->r_osd = NULL; + } +} + +void ceph_osdc_unregister_linger_request(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) +{ + mutex_lock(&osdc->request_mutex); + if (req->r_linger) { + __unregister_linger_request(osdc, req); + ceph_osdc_put_request(req); + } + mutex_unlock(&osdc->request_mutex); +} +EXPORT_SYMBOL(ceph_osdc_unregister_linger_request); + +void ceph_osdc_set_request_linger(struct ceph_osd_client *osdc, + struct ceph_osd_request *req) +{ + if (!req->r_linger) { + dout("set_request_linger %p\n", req); + req->r_linger = 1; + /* + * caller is now responsible for calling + * unregister_linger_request + */ + ceph_osdc_get_request(req); + } +} +EXPORT_SYMBOL(ceph_osdc_set_request_linger); + /* * Pick an osd (the first 'up' osd in the pg), allocate the osd struct * (as needed), and set the request r_osd appropriately. If there is @@ -958,7 +1060,6 @@ static void handle_timeout(struct work_struct *work) osdc->client->options->osd_keepalive_timeout * HZ; unsigned long last_stamp = 0; struct list_head slow_osds; - dout("timeout\n"); down_read(&osdc->map_sem); @@ -1060,7 +1161,6 @@ static void handle_reply(struct ceph_osd_client *osdc, struct ceph_msg *msg, numops * sizeof(struct ceph_osd_op)) goto bad; dout("handle_reply %p tid %llu result %d\n", msg, tid, (int)result); - /* lookup */ mutex_lock(&osdc->request_mutex); req = __lookup_request(osdc, tid); @@ -1104,6 +1204,9 @@ static void handle_reply(struct ceph_osd_client *osdc, struct ceph_msg *msg, dout("handle_reply tid %llu flags %d\n", tid, flags); + if (req->r_linger && (flags & CEPH_OSD_FLAG_ONDISK)) + __register_linger_request(osdc, req); + /* either this is a read, or we got the safe response */ if (result < 0 || (flags & CEPH_OSD_FLAG_ONDISK) || @@ -1124,6 +1227,7 @@ static void handle_reply(struct ceph_osd_client *osdc, struct ceph_msg *msg, } done: + dout("req=%p req->r_linger=%d\n", req, req->r_linger); ceph_osdc_put_request(req); return; @@ -1159,7 +1263,7 @@ static void reset_changed_osds(struct ceph_osd_client *osdc) */ static void kick_requests(struct ceph_osd_client *osdc) { - struct ceph_osd_request *req; + struct ceph_osd_request *req, *nreq; struct rb_node *p; int needmap = 0; int err; @@ -1177,8 +1281,30 @@ static void kick_requests(struct ceph_osd_client *osdc) } else if (err > 0) { dout("%p tid %llu requeued on osd%d\n", req, req->r_tid, req->r_osd ? req->r_osd->o_osd : -1); - req->r_flags |= CEPH_OSD_FLAG_RETRY; + if (!req->r_linger) + req->r_flags |= CEPH_OSD_FLAG_RETRY; + } + } + + list_for_each_entry_safe(req, nreq, &osdc->req_linger, + r_linger_item) { + dout("linger req=%p req->r_osd=%p\n", req, req->r_osd); + + err = __map_request(osdc, req); + if (err == 0) + continue; /* no change and no osd was specified */ + if (err < 0) + continue; /* hrm! */ + if (req->r_osd == NULL) { + dout("tid %llu maps to no valid osd\n", req->r_tid); + needmap++; /* request a newer map */ + continue; } + + dout("kicking lingering %p tid %llu osd%d\n", req, req->r_tid, + req->r_osd ? req->r_osd->o_osd : -1); + __unregister_linger_request(osdc, req); + __register_request(osdc, req); } mutex_unlock(&osdc->request_mutex); @@ -1301,6 +1427,223 @@ bad: return; } +/* + * watch/notify callback event infrastructure + * + * These callbacks are used both for watch and notify operations. + */ +static void __release_event(struct kref *kref) +{ + struct ceph_osd_event *event = + container_of(kref, struct ceph_osd_event, kref); + + dout("__release_event %p\n", event); + kfree(event); +} + +static void get_event(struct ceph_osd_event *event) +{ + kref_get(&event->kref); +} + +void ceph_osdc_put_event(struct ceph_osd_event *event) +{ + kref_put(&event->kref, __release_event); +} +EXPORT_SYMBOL(ceph_osdc_put_event); + +static void __insert_event(struct ceph_osd_client *osdc, + struct ceph_osd_event *new) +{ + struct rb_node **p = &osdc->event_tree.rb_node; + struct rb_node *parent = NULL; + struct ceph_osd_event *event = NULL; + + while (*p) { + parent = *p; + event = rb_entry(parent, struct ceph_osd_event, node); + if (new->cookie < event->cookie) + p = &(*p)->rb_left; + else if (new->cookie > event->cookie) + p = &(*p)->rb_right; + else + BUG(); + } + + rb_link_node(&new->node, parent, p); + rb_insert_color(&new->node, &osdc->event_tree); +} + +static struct ceph_osd_event *__find_event(struct ceph_osd_client *osdc, + u64 cookie) +{ + struct rb_node **p = &osdc->event_tree.rb_node; + struct rb_node *parent = NULL; + struct ceph_osd_event *event = NULL; + + while (*p) { + parent = *p; + event = rb_entry(parent, struct ceph_osd_event, node); + if (cookie < event->cookie) + p = &(*p)->rb_left; + else if (cookie > event->cookie) + p = &(*p)->rb_right; + else + return event; + } + return NULL; +} + +static void __remove_event(struct ceph_osd_event *event) +{ + struct ceph_osd_client *osdc = event->osdc; + + if (!RB_EMPTY_NODE(&event->node)) { + dout("__remove_event removed %p\n", event); + rb_erase(&event->node, &osdc->event_tree); + ceph_osdc_put_event(event); + } else { + dout("__remove_event didn't remove %p\n", event); + } +} + +int ceph_osdc_create_event(struct ceph_osd_client *osdc, + void (*event_cb)(u64, u64, u8, void *), + int one_shot, void *data, + struct ceph_osd_event **pevent) +{ + struct ceph_osd_event *event; + + event = kmalloc(sizeof(*event), GFP_NOIO); + if (!event) + return -ENOMEM; + + dout("create_event %p\n", event); + event->cb = event_cb; + event->one_shot = one_shot; + event->data = data; + event->osdc = osdc; + INIT_LIST_HEAD(&event->osd_node); + kref_init(&event->kref); /* one ref for us */ + kref_get(&event->kref); /* one ref for the caller */ + init_completion(&event->completion); + + spin_lock(&osdc->event_lock); + event->cookie = ++osdc->event_count; + __insert_event(osdc, event); + spin_unlock(&osdc->event_lock); + + *pevent = event; + return 0; +} +EXPORT_SYMBOL(ceph_osdc_create_event); + +void ceph_osdc_cancel_event(struct ceph_osd_event *event) +{ + struct ceph_osd_client *osdc = event->osdc; + + dout("cancel_event %p\n", event); + spin_lock(&osdc->event_lock); + __remove_event(event); + spin_unlock(&osdc->event_lock); + ceph_osdc_put_event(event); /* caller's */ +} +EXPORT_SYMBOL(ceph_osdc_cancel_event); + + +static void do_event_work(struct work_struct *work) +{ + struct ceph_osd_event_work *event_work = + container_of(work, struct ceph_osd_event_work, work); + struct ceph_osd_event *event = event_work->event; + u64 ver = event_work->ver; + u64 notify_id = event_work->notify_id; + u8 opcode = event_work->opcode; + + dout("do_event_work completing %p\n", event); + event->cb(ver, notify_id, opcode, event->data); + complete(&event->completion); + dout("do_event_work completed %p\n", event); + ceph_osdc_put_event(event); + kfree(event_work); +} + + +/* + * Process osd watch notifications + */ +void handle_watch_notify(struct ceph_osd_client *osdc, struct ceph_msg *msg) +{ + void *p, *end; + u8 proto_ver; + u64 cookie, ver, notify_id; + u8 opcode; + struct ceph_osd_event *event; + struct ceph_osd_event_work *event_work; + + p = msg->front.iov_base; + end = p + msg->front.iov_len; + + ceph_decode_8_safe(&p, end, proto_ver, bad); + ceph_decode_8_safe(&p, end, opcode, bad); + ceph_decode_64_safe(&p, end, cookie, bad); + ceph_decode_64_safe(&p, end, ver, bad); + ceph_decode_64_safe(&p, end, notify_id, bad); + + spin_lock(&osdc->event_lock); + event = __find_event(osdc, cookie); + if (event) { + get_event(event); + if (event->one_shot) + __remove_event(event); + } + spin_unlock(&osdc->event_lock); + dout("handle_watch_notify cookie %lld ver %lld event %p\n", + cookie, ver, event); + if (event) { + event_work = kmalloc(sizeof(*event_work), GFP_NOIO); + INIT_WORK(&event_work->work, do_event_work); + if (!event_work) { + dout("ERROR: could not allocate event_work\n"); + goto done_err; + } + event_work->event = event; + event_work->ver = ver; + event_work->notify_id = notify_id; + event_work->opcode = opcode; + if (!queue_work(osdc->notify_wq, &event_work->work)) { + dout("WARNING: failed to queue notify event work\n"); + goto done_err; + } + } + + return; + +done_err: + complete(&event->completion); + ceph_osdc_put_event(event); + return; + +bad: + pr_err("osdc handle_watch_notify corrupt msg\n"); + return; +} + +int ceph_osdc_wait_event(struct ceph_osd_event *event, unsigned long timeout) +{ + int err; + + dout("wait_event %p\n", event); + err = wait_for_completion_interruptible_timeout(&event->completion, + timeout * HZ); + ceph_osdc_put_event(event); + if (err > 0) + err = 0; + dout("wait_event %p returns %d\n", event, err); + return err; +} +EXPORT_SYMBOL(ceph_osdc_wait_event); + /* * Register request, send initial attempt. */ @@ -1430,9 +1773,13 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) INIT_LIST_HEAD(&osdc->req_lru); INIT_LIST_HEAD(&osdc->req_unsent); INIT_LIST_HEAD(&osdc->req_notarget); + INIT_LIST_HEAD(&osdc->req_linger); osdc->num_requests = 0; INIT_DELAYED_WORK(&osdc->timeout_work, handle_timeout); INIT_DELAYED_WORK(&osdc->osds_timeout_work, handle_osds_timeout); + spin_lock_init(&osdc->event_lock); + osdc->event_tree = RB_ROOT; + osdc->event_count = 0; schedule_delayed_work(&osdc->osds_timeout_work, round_jiffies_relative(osdc->client->options->osd_idle_ttl * HZ)); @@ -1452,6 +1799,13 @@ int ceph_osdc_init(struct ceph_osd_client *osdc, struct ceph_client *client) "osd_op_reply"); if (err < 0) goto out_msgpool; + + osdc->notify_wq = create_singlethread_workqueue("ceph-watch-notify"); + if (IS_ERR(osdc->notify_wq)) { + err = PTR_ERR(osdc->notify_wq); + osdc->notify_wq = NULL; + goto out_msgpool; + } return 0; out_msgpool: @@ -1465,6 +1819,8 @@ EXPORT_SYMBOL(ceph_osdc_init); void ceph_osdc_stop(struct ceph_osd_client *osdc) { + flush_workqueue(osdc->notify_wq); + destroy_workqueue(osdc->notify_wq); cancel_delayed_work_sync(&osdc->timeout_work); cancel_delayed_work_sync(&osdc->osds_timeout_work); if (osdc->osdmap) { @@ -1472,6 +1828,7 @@ void ceph_osdc_stop(struct ceph_osd_client *osdc) osdc->osdmap = NULL; } remove_old_osds(osdc, 1); + WARN_ON(!RB_EMPTY_ROOT(&osdc->osds)); mempool_destroy(osdc->req_mempool); ceph_msgpool_destroy(&osdc->msgpool_op); ceph_msgpool_destroy(&osdc->msgpool_op_reply); @@ -1580,6 +1937,9 @@ static void dispatch(struct ceph_connection *con, struct ceph_msg *msg) case CEPH_MSG_OSD_OPREPLY: handle_reply(osdc, msg, con); break; + case CEPH_MSG_WATCH_NOTIFY: + handle_watch_notify(osdc, msg); + break; default: pr_err("received unknown message type %d %s\n", type, @@ -1673,6 +2033,7 @@ static struct ceph_msg *alloc_msg(struct ceph_connection *con, switch (type) { case CEPH_MSG_OSD_MAP: + case CEPH_MSG_WATCH_NOTIFY: return ceph_msg_new(type, front, GFP_NOFS); case CEPH_MSG_OSD_OPREPLY: return get_reply(con, hdr, skip); -- cgit v1.2.3-18-g5258