1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2002 Cluster File Systems, Inc.
6 * This file is part of Lustre, http://www.lustre.org.
8 * Lustre is free software; you can redistribute it and/or
9 * modify it under the terms of version 2 of the GNU General Public
10 * License as published by the Free Software Foundation.
12 * Lustre is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with Lustre; if not, write to the Free Software
19 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
26 #include <linux/tqueue.h>
27 #include <linux/kp30.h>
28 // #include <linux/obd.h>
29 #include <portals/p30.h>
30 #include <linux/lustre_idl.h>
31 #include <linux/lustre_ha.h>
32 #include <linux/lustre_import.h>
34 /* The following constants determine how much memory is devoted to
35 * buffering in the lustre services.
37 * ?_NEVENTS # event queue entries
39 * ?_NBUFS # request buffers
40 * ?_BUFSIZE # bytes in a single request buffer
41 * total memory = ?_NBUFS * ?_BUFSIZE
43 * ?_MAXREQSIZE # maximum request service will receive
44 * larger messages will get dropped.
45 * request buffers are auto-unlinked when less than ?_MAXREQSIZE
49 #define LDLM_NEVENTS 1024
51 #define LDLM_BUFSIZE (64 * 1024)
52 #define LDLM_MAXREQSIZE 1024
54 #define MDS_NEVENTS 1024
56 #define MDS_BUFSIZE (64 * 1024)
57 #define MDS_MAXREQSIZE 1024
60 #define OST_NEVENTS 1024
62 #define OST_BUFSIZE (64 * 1024)
63 #define OST_MAXREQSIZE (8 * 1024)
65 #define OST_NEVENTS 4096
67 #define OST_BUFSIZE (128 * 1024)
68 #define OST_MAXREQSIZE (8 * 1024)
71 struct ptlrpc_connection {
72 struct list_head c_link;
73 struct lustre_peer c_peer;
74 __u8 c_local_uuid[37]; /* XXX do we need this? */
75 __u8 c_remote_uuid[37];
78 __u32 c_generation; /* changes upon new connection */
79 __u32 c_epoch; /* changes when peer changes */
80 __u32 c_bootcount; /* peer's boot count */
82 spinlock_t c_lock; /* also protects req->rq_list */
91 __u64 c_last_xid; /* protected by c_lock */
92 __u64 c_last_committed;/* protected by c_lock */
93 struct list_head c_delayed_head;/* delayed until post-recovery */
94 struct list_head c_sending_head;/* protected by c_lock */
95 struct list_head c_dying_head; /* protected by c_lock */
96 struct recovd_data c_recovd_data;
98 struct list_head c_imports;
99 struct list_head c_exports;
102 struct ptlrpc_client {
103 __u32 cli_request_portal;
104 __u32 cli_reply_portal;
106 __u32 cli_target_devno;
109 // struct semaphore cli_rpc_sem; /* limits outstanding requests */
114 /* state flags of requests */
115 #define PTL_RPC_FL_INTR (1 << 0)
116 #define PTL_RPC_FL_REPLIED (1 << 1) /* reply was received */
117 #define PTL_RPC_FL_SENT (1 << 2)
118 #define PTL_BULK_FL_SENT (1 << 3)
119 #define PTL_BULK_FL_RCVD (1 << 4)
120 #define PTL_RPC_FL_ERR (1 << 5)
121 #define PTL_RPC_FL_TIMEOUT (1 << 6)
122 #define PTL_RPC_FL_RESEND (1 << 7)
123 #define PTL_RPC_FL_RECOVERY (1 << 8) /* retransmission for recovery */
124 #define PTL_RPC_FL_FINISHED (1 << 9)
125 #define PTL_RPC_FL_RETAIN (1 << 10) /* retain for replay after reply */
126 #define PTL_RPC_FL_REPLAY (1 << 11) /* replay upon recovery */
127 #define PTL_RPC_FL_ALLOCREP (1 << 12) /* reply buffer allocated */
129 struct ptlrpc_request {
130 int rq_type; /* one of PTL_RPC_MSG_* */
131 struct list_head rq_list;
132 struct list_head rq_multi;
133 struct obd_device *rq_obd;
137 atomic_t rq_refcount;
140 struct lustre_msg *rq_reqmsg;
143 struct lustre_msg *rq_repmsg;
153 // void * rq_reply_handle;
154 wait_queue_head_t rq_wait_for_rep;
157 ptl_md_t rq_reply_md;
158 ptl_handle_md_t rq_reply_md_h;
159 ptl_handle_me_t rq_reply_me_h;
161 /* outgoing req/rep */
163 ptl_handle_md_t rq_req_md_h;
165 struct lustre_peer rq_peer; /* XXX see service.c can this be factored away? */
166 struct obd_export *rq_export;
167 struct ptlrpc_connection *rq_connection;
168 struct obd_import *rq_import;
169 struct ptlrpc_service *rq_svc;
171 void (*rq_replay_cb)(struct ptlrpc_request *, void *);
172 void *rq_replay_cb_data;
175 struct ptlrpc_bulk_page {
176 struct ptlrpc_bulk_desc *bp_desc;
177 struct list_head bp_link;
180 struct page *bp_page;
183 struct dentry *bp_dentry;
184 int (*bp_cb)(struct ptlrpc_bulk_page *);
187 struct ptlrpc_bulk_desc {
189 struct ptlrpc_connection *bd_connection;
190 struct ptlrpc_client *bd_client;
192 struct lustre_handle bd_conn;
193 void (*bd_cb)(struct ptlrpc_bulk_desc *, void *);
196 wait_queue_head_t bd_waitq;
197 struct list_head bd_page_list;
199 atomic_t bd_refcount;
200 void *bd_desc_private;
201 struct tq_struct bd_queue;
204 ptl_handle_md_t bd_md_h;
205 ptl_handle_me_t bd_me_h;
207 atomic_t bd_source_callback_count;
209 struct iovec bd_iov[16]; /* self-sized pre-allocated iov */
212 struct ptlrpc_thread {
213 struct list_head t_link;
216 wait_queue_head_t t_ctl_waitq;
219 struct ptlrpc_request_buffer_desc {
220 struct list_head rqbd_list;
221 struct ptlrpc_service *rqbd_service;
222 ptl_handle_me_t rqbd_me_h;
223 atomic_t rqbd_refcount;
227 struct ptlrpc_service {
231 /* incoming request buffers */
232 /* FIXME: perhaps a list of EQs, if multiple NIs are used? */
234 __u32 srv_max_req_size; /* biggest request to receive */
235 __u32 srv_buf_size; /* # bytes in a request buffer */
236 struct list_head srv_rqbds; /* all the request buffer descriptors */
237 __u32 srv_nrqbds; /* # request buffers */
238 atomic_t srv_nrqbds_receiving; /* # request buffers posted for input */
240 __u32 srv_req_portal;
241 __u32 srv_rep_portal;
246 ptl_handle_eq_t srv_eq_h;
248 struct lustre_peer srv_self;
250 wait_queue_head_t srv_waitq; /* all threads sleep on this */
253 struct list_head srv_threads;
254 int (*srv_handler)(struct ptlrpc_request *req);
255 char *srv_name; /* only statically allocated strings here; we don't clean them */
258 static inline void ptlrpc_hdl2req(struct ptlrpc_request *req, struct lustre_handle *h)
260 req->rq_reqmsg->addr = h->addr;
261 req->rq_reqmsg->cookie = h->cookie;
264 typedef void (*bulk_callback_t)(struct ptlrpc_bulk_desc *, void *);
266 typedef int (*svc_handler_t)(struct ptlrpc_request *req);
268 /* rpc/connection.c */
269 void ptlrpc_readdress_connection(struct ptlrpc_connection *conn, obd_uuid_t uuid);
270 struct ptlrpc_connection *ptlrpc_get_connection(struct lustre_peer *peer,
272 int ptlrpc_put_connection(struct ptlrpc_connection *c);
273 struct ptlrpc_connection *ptlrpc_connection_addref(struct ptlrpc_connection *);
274 void ptlrpc_init_connection(void);
275 void ptlrpc_cleanup_connection(void);
278 int ptlrpc_check_bulk_sent(struct ptlrpc_bulk_desc *bulk);
279 int ptlrpc_check_bulk_received(struct ptlrpc_bulk_desc *bulk);
280 int ptlrpc_send_bulk(struct ptlrpc_bulk_desc *);
281 int ptlrpc_register_bulk(struct ptlrpc_bulk_desc *);
282 int ptlrpc_abort_bulk(struct ptlrpc_bulk_desc *bulk);
283 int ptlrpc_reply(struct ptlrpc_service *svc, struct ptlrpc_request *req);
284 int ptlrpc_error(struct ptlrpc_service *svc, struct ptlrpc_request *req);
285 void ptlrpc_resend_req(struct ptlrpc_request *request);
286 int ptl_send_rpc(struct ptlrpc_request *request);
287 void ptlrpc_link_svc_me(struct ptlrpc_request_buffer_desc *rqbd);
290 void ptlrpc_init_client(int req_portal, int rep_portal, char *name,
291 struct ptlrpc_client *);
292 void ptlrpc_cleanup_client(struct obd_import *imp);
293 __u8 *ptlrpc_req_to_uuid(struct ptlrpc_request *req);
294 struct ptlrpc_connection *ptlrpc_uuid_to_connection(obd_uuid_t uuid);
296 int ptlrpc_queue_wait(struct ptlrpc_request *req);
297 void ptlrpc_continue_req(struct ptlrpc_request *req);
298 int ptlrpc_replay_req(struct ptlrpc_request *req);
299 void ptlrpc_restart_req(struct ptlrpc_request *req);
301 struct ptlrpc_request *ptlrpc_prep_req(struct obd_import *imp, int opcode,
302 int count, int *lengths, char **bufs);
303 void ptlrpc_free_req(struct ptlrpc_request *request);
304 void ptlrpc_req_finished(struct ptlrpc_request *request);
305 struct ptlrpc_bulk_desc *ptlrpc_prep_bulk(struct ptlrpc_connection *);
306 void ptlrpc_free_bulk(struct ptlrpc_bulk_desc *bulk);
307 struct ptlrpc_bulk_page *ptlrpc_prep_bulk_page(struct ptlrpc_bulk_desc *desc);
308 void ptlrpc_free_bulk_page(struct ptlrpc_bulk_page *page);
309 int ptlrpc_check_status(struct ptlrpc_request *req, int err);
312 struct ptlrpc_service *
313 ptlrpc_init_svc(__u32 nevents, __u32 nbufs, __u32 bufsize, __u32 max_req_size,
314 int req_portal, int rep_portal,
315 obd_uuid_t uuid, svc_handler_t, char *name);
316 void ptlrpc_stop_all_threads(struct ptlrpc_service *svc);
317 int ptlrpc_start_thread(struct obd_device *dev, struct ptlrpc_service *svc,
319 int ptlrpc_unregister_service(struct ptlrpc_service *service);
321 struct ptlrpc_svc_data {
323 struct ptlrpc_service *svc;
324 struct ptlrpc_thread *thread;
325 struct obd_device *dev;
328 /* rpc/pack_generic.c */
329 int lustre_pack_msg(int count, int *lens, char **bufs, int *len,
330 struct lustre_msg **msg);
331 int lustre_msg_size(int count, int *lengths);
332 int lustre_unpack_msg(struct lustre_msg *m, int len);
333 void *lustre_msg_buf(struct lustre_msg *m, int n);
335 static inline void ptlrpc_bulk_decref(struct ptlrpc_bulk_desc *desc)
337 if (atomic_dec_and_test(&desc->bd_refcount)) {
338 CDEBUG(D_PAGE, "Released last ref on %p, freeing\n", desc);
339 ptlrpc_free_bulk(desc);
341 CDEBUG(D_PAGE, "%p -> %d\n", desc,
342 atomic_read(&desc->bd_refcount));
346 static inline void ptlrpc_bulk_addref(struct ptlrpc_bulk_desc *desc)
348 atomic_inc(&desc->bd_refcount);
349 CDEBUG(D_PAGE, "Set refcount of %p to %d\n", desc,
350 atomic_read(&desc->bd_refcount));