1 /******************************************************************************
2 *******************************************************************************
4 ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
5 ** Copyright (C) 2005 Red Hat, Inc. All rights reserved.
7 ** This copyrighted material is made available to anyone wishing to use,
8 ** modify, copy, or redistribute it subject to the terms and conditions
9 ** of the GNU General Public License v.2.
11 *******************************************************************************
12 ******************************************************************************/
14 #include "dlm_internal.h"
15 #include "lockspace.h"
28 static int rcom_response(struct dlm_ls *ls)
30 return test_bit(LSFL_RCOM_READY, &ls->ls_flags);
33 static int create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len,
34 struct dlm_rcom **rc_ret, struct dlm_mhandle **mh_ret)
37 struct dlm_mhandle *mh;
39 int mb_len = sizeof(struct dlm_rcom) + len;
41 mh = dlm_lowcomms_get_buffer(to_nodeid, mb_len, ls->ls_allocation, &mb);
43 log_print("create_rcom to %d type %d len %d ENOBUFS",
44 to_nodeid, type, len);
47 memset(mb, 0, mb_len);
49 rc = (struct dlm_rcom *) mb;
51 rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
52 rc->rc_header.h_lockspace = ls->ls_global_id;
53 rc->rc_header.h_nodeid = dlm_our_nodeid();
54 rc->rc_header.h_length = mb_len;
55 rc->rc_header.h_cmd = DLM_RCOM;
59 spin_lock(&ls->ls_recover_lock);
60 rc->rc_seq = ls->ls_recover_seq;
61 spin_unlock(&ls->ls_recover_lock);
68 static void send_rcom(struct dlm_ls *ls, struct dlm_mhandle *mh,
72 dlm_lowcomms_commit_buffer(mh);
75 /* When replying to a status request, a node also sends back its
76 configuration values. The requesting node then checks that the remote
77 node is configured the same way as itself. */
79 static void make_config(struct dlm_ls *ls, struct rcom_config *rf)
81 rf->rf_lvblen = ls->ls_lvblen;
82 rf->rf_lsflags = ls->ls_exflags;
85 static int check_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
87 struct rcom_config *rf = (struct rcom_config *) rc->rc_buf;
89 if ((rc->rc_header.h_version & 0xFFFF0000) != DLM_HEADER_MAJOR) {
90 log_error(ls, "version mismatch: %x nodeid %d: %x",
91 DLM_HEADER_MAJOR | DLM_HEADER_MINOR, nodeid,
92 rc->rc_header.h_version);
96 if (rf->rf_lvblen != ls->ls_lvblen ||
97 rf->rf_lsflags != ls->ls_exflags) {
98 log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x",
99 ls->ls_lvblen, ls->ls_exflags,
100 nodeid, rf->rf_lvblen, rf->rf_lsflags);
106 static void allow_sync_reply(struct dlm_ls *ls, uint64_t *new_seq)
108 spin_lock(&ls->ls_rcom_spin);
109 *new_seq = ++ls->ls_rcom_seq;
110 set_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
111 spin_unlock(&ls->ls_rcom_spin);
114 static void disallow_sync_reply(struct dlm_ls *ls)
116 spin_lock(&ls->ls_rcom_spin);
117 clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
118 clear_bit(LSFL_RCOM_READY, &ls->ls_flags);
119 spin_unlock(&ls->ls_rcom_spin);
122 int dlm_rcom_status(struct dlm_ls *ls, int nodeid)
125 struct dlm_mhandle *mh;
128 ls->ls_recover_nodeid = nodeid;
130 if (nodeid == dlm_our_nodeid()) {
131 rc = (struct dlm_rcom *) ls->ls_recover_buf;
132 rc->rc_result = dlm_recover_status(ls);
136 error = create_rcom(ls, nodeid, DLM_RCOM_STATUS, 0, &rc, &mh);
140 allow_sync_reply(ls, &rc->rc_id);
141 memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size);
143 send_rcom(ls, mh, rc);
145 error = dlm_wait_function(ls, &rcom_response);
146 disallow_sync_reply(ls);
150 rc = (struct dlm_rcom *) ls->ls_recover_buf;
152 if (rc->rc_result == -ESRCH) {
153 /* we pretend the remote lockspace exists with 0 status */
154 log_debug(ls, "remote node %d not ready", nodeid);
157 error = check_config(ls, rc, nodeid);
158 /* the caller looks at rc_result for the remote recovery status */
163 static void receive_rcom_status(struct dlm_ls *ls, struct dlm_rcom *rc_in)
166 struct dlm_mhandle *mh;
167 int error, nodeid = rc_in->rc_header.h_nodeid;
169 error = create_rcom(ls, nodeid, DLM_RCOM_STATUS_REPLY,
170 sizeof(struct rcom_config), &rc, &mh);
173 rc->rc_id = rc_in->rc_id;
174 rc->rc_seq_reply = rc_in->rc_seq;
175 rc->rc_result = dlm_recover_status(ls);
176 make_config(ls, (struct rcom_config *) rc->rc_buf);
178 send_rcom(ls, mh, rc);
181 static void receive_sync_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
183 spin_lock(&ls->ls_rcom_spin);
184 if (!test_bit(LSFL_RCOM_WAIT, &ls->ls_flags) ||
185 rc_in->rc_id != ls->ls_rcom_seq) {
186 log_debug(ls, "reject reply %d from %d seq %llx expect %llx",
187 rc_in->rc_type, rc_in->rc_header.h_nodeid,
188 (unsigned long long)rc_in->rc_id,
189 (unsigned long long)ls->ls_rcom_seq);
192 memcpy(ls->ls_recover_buf, rc_in, rc_in->rc_header.h_length);
193 set_bit(LSFL_RCOM_READY, &ls->ls_flags);
194 clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
195 wake_up(&ls->ls_wait_general);
197 spin_unlock(&ls->ls_rcom_spin);
200 static void receive_rcom_status_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
202 receive_sync_reply(ls, rc_in);
205 int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
208 struct dlm_mhandle *mh;
209 int error = 0, len = sizeof(struct dlm_rcom);
211 ls->ls_recover_nodeid = nodeid;
213 if (nodeid == dlm_our_nodeid()) {
214 dlm_copy_master_names(ls, last_name, last_len,
215 ls->ls_recover_buf + len,
216 dlm_config.ci_buffer_size - len, nodeid);
220 error = create_rcom(ls, nodeid, DLM_RCOM_NAMES, last_len, &rc, &mh);
223 memcpy(rc->rc_buf, last_name, last_len);
225 allow_sync_reply(ls, &rc->rc_id);
226 memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size);
228 send_rcom(ls, mh, rc);
230 error = dlm_wait_function(ls, &rcom_response);
231 disallow_sync_reply(ls);
236 static void receive_rcom_names(struct dlm_ls *ls, struct dlm_rcom *rc_in)
239 struct dlm_mhandle *mh;
240 int error, inlen, outlen, nodeid;
242 nodeid = rc_in->rc_header.h_nodeid;
243 inlen = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
244 outlen = dlm_config.ci_buffer_size - sizeof(struct dlm_rcom);
246 error = create_rcom(ls, nodeid, DLM_RCOM_NAMES_REPLY, outlen, &rc, &mh);
249 rc->rc_id = rc_in->rc_id;
250 rc->rc_seq_reply = rc_in->rc_seq;
252 dlm_copy_master_names(ls, rc_in->rc_buf, inlen, rc->rc_buf, outlen,
254 send_rcom(ls, mh, rc);
257 static void receive_rcom_names_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
259 receive_sync_reply(ls, rc_in);
262 int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
265 struct dlm_mhandle *mh;
266 struct dlm_ls *ls = r->res_ls;
269 error = create_rcom(ls, dir_nodeid, DLM_RCOM_LOOKUP, r->res_length,
273 memcpy(rc->rc_buf, r->res_name, r->res_length);
274 rc->rc_id = (unsigned long) r;
276 send_rcom(ls, mh, rc);
281 static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
284 struct dlm_mhandle *mh;
285 int error, ret_nodeid, nodeid = rc_in->rc_header.h_nodeid;
286 int len = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
288 error = create_rcom(ls, nodeid, DLM_RCOM_LOOKUP_REPLY, 0, &rc, &mh);
292 error = dlm_dir_lookup(ls, nodeid, rc_in->rc_buf, len, &ret_nodeid);
295 rc->rc_result = ret_nodeid;
296 rc->rc_id = rc_in->rc_id;
297 rc->rc_seq_reply = rc_in->rc_seq;
299 send_rcom(ls, mh, rc);
302 static void receive_rcom_lookup_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
304 dlm_recover_master_reply(ls, rc_in);
307 static void pack_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb,
308 struct rcom_lock *rl)
310 memset(rl, 0, sizeof(*rl));
312 rl->rl_ownpid = lkb->lkb_ownpid;
313 rl->rl_lkid = lkb->lkb_id;
314 rl->rl_exflags = lkb->lkb_exflags;
315 rl->rl_flags = lkb->lkb_flags;
316 rl->rl_lvbseq = lkb->lkb_lvbseq;
317 rl->rl_rqmode = lkb->lkb_rqmode;
318 rl->rl_grmode = lkb->lkb_grmode;
319 rl->rl_status = lkb->lkb_status;
320 rl->rl_wait_type = lkb->lkb_wait_type;
322 if (lkb->lkb_bastaddr)
323 rl->rl_asts |= AST_BAST;
324 if (lkb->lkb_astaddr)
325 rl->rl_asts |= AST_COMP;
327 rl->rl_namelen = r->res_length;
328 memcpy(rl->rl_name, r->res_name, r->res_length);
330 /* FIXME: might we have an lvb without DLM_LKF_VALBLK set ?
331 If so, receive_rcom_lock_args() won't take this copy. */
334 memcpy(rl->rl_lvb, lkb->lkb_lvbptr, r->res_ls->ls_lvblen);
337 int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
339 struct dlm_ls *ls = r->res_ls;
341 struct dlm_mhandle *mh;
342 struct rcom_lock *rl;
343 int error, len = sizeof(struct rcom_lock);
346 len += ls->ls_lvblen;
348 error = create_rcom(ls, r->res_nodeid, DLM_RCOM_LOCK, len, &rc, &mh);
352 rl = (struct rcom_lock *) rc->rc_buf;
353 pack_rcom_lock(r, lkb, rl);
354 rc->rc_id = (unsigned long) r;
356 send_rcom(ls, mh, rc);
361 static void receive_rcom_lock(struct dlm_ls *ls, struct dlm_rcom *rc_in)
364 struct dlm_mhandle *mh;
365 int error, nodeid = rc_in->rc_header.h_nodeid;
367 dlm_recover_master_copy(ls, rc_in);
369 error = create_rcom(ls, nodeid, DLM_RCOM_LOCK_REPLY,
370 sizeof(struct rcom_lock), &rc, &mh);
374 /* We send back the same rcom_lock struct we received, but
375 dlm_recover_master_copy() has filled in rl_remid and rl_result */
377 memcpy(rc->rc_buf, rc_in->rc_buf, sizeof(struct rcom_lock));
378 rc->rc_id = rc_in->rc_id;
379 rc->rc_seq_reply = rc_in->rc_seq;
381 send_rcom(ls, mh, rc);
384 static void receive_rcom_lock_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
386 dlm_recover_process_copy(ls, rc_in);
389 static int send_ls_not_ready(struct dlm_ls *ls, int nodeid,
390 struct dlm_rcom *rc_in)
393 struct rcom_config *rf;
394 struct dlm_mhandle *mh;
396 int mb_len = sizeof(struct dlm_rcom) + sizeof(struct rcom_config);
398 mh = dlm_lowcomms_get_buffer(nodeid, mb_len, ls->ls_allocation, &mb);
401 memset(mb, 0, mb_len);
403 rc = (struct dlm_rcom *) mb;
405 rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
406 rc->rc_header.h_lockspace = rc_in->rc_header.h_lockspace;
407 rc->rc_header.h_nodeid = dlm_our_nodeid();
408 rc->rc_header.h_length = mb_len;
409 rc->rc_header.h_cmd = DLM_RCOM;
411 rc->rc_type = DLM_RCOM_STATUS_REPLY;
412 rc->rc_id = rc_in->rc_id;
413 rc->rc_seq_reply = rc_in->rc_seq;
414 rc->rc_result = -ESRCH;
416 rf = (struct rcom_config *) rc->rc_buf;
420 dlm_lowcomms_commit_buffer(mh);
425 static int is_old_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
430 switch (rc->rc_type) {
431 case DLM_RCOM_STATUS_REPLY:
432 case DLM_RCOM_NAMES_REPLY:
433 case DLM_RCOM_LOOKUP_REPLY:
434 case DLM_RCOM_LOCK_REPLY:
435 spin_lock(&ls->ls_recover_lock);
436 seq = ls->ls_recover_seq;
437 spin_unlock(&ls->ls_recover_lock);
438 if (rc->rc_seq_reply != seq) {
439 log_debug(ls, "ignoring old reply %x from %d "
440 "seq_reply %llx expect %llx",
441 rc->rc_type, rc->rc_header.h_nodeid,
442 (unsigned long long)rc->rc_seq_reply,
443 (unsigned long long)seq);
450 /* Called by dlm_recvd; corresponds to dlm_receive_message() but special
451 recovery-only comms are sent through here. */
453 void dlm_receive_rcom(struct dlm_header *hd, int nodeid)
455 struct dlm_rcom *rc = (struct dlm_rcom *) hd;
460 /* If the lockspace doesn't exist then still send a status message
461 back; it's possible that it just doesn't have its global_id yet. */
463 ls = dlm_find_lockspace_global(hd->h_lockspace);
465 log_print("lockspace %x from %d type %x not found",
466 hd->h_lockspace, nodeid, rc->rc_type);
467 if (rc->rc_type == DLM_RCOM_STATUS)
468 send_ls_not_ready(ls, nodeid, rc);
472 if (dlm_recovery_stopped(ls) && (rc->rc_type != DLM_RCOM_STATUS)) {
473 log_debug(ls, "ignoring recovery message %x from %d",
474 rc->rc_type, nodeid);
478 if (is_old_reply(ls, rc))
481 if (nodeid != rc->rc_header.h_nodeid) {
482 log_error(ls, "bad rcom nodeid %d from %d",
483 rc->rc_header.h_nodeid, nodeid);
487 switch (rc->rc_type) {
488 case DLM_RCOM_STATUS:
489 receive_rcom_status(ls, rc);
493 receive_rcom_names(ls, rc);
496 case DLM_RCOM_LOOKUP:
497 receive_rcom_lookup(ls, rc);
501 receive_rcom_lock(ls, rc);
504 case DLM_RCOM_STATUS_REPLY:
505 receive_rcom_status_reply(ls, rc);
508 case DLM_RCOM_NAMES_REPLY:
509 receive_rcom_names_reply(ls, rc);
512 case DLM_RCOM_LOOKUP_REPLY:
513 receive_rcom_lookup_reply(ls, rc);
516 case DLM_RCOM_LOCK_REPLY:
517 receive_rcom_lock_reply(ls, rc);
521 DLM_ASSERT(0, printk("rc_type=%x\n", rc->rc_type););
524 dlm_put_lockspace(ls);