xref: /openbmc/linux/fs/dlm/rcom.c (revision e868d61272caa648214046a096e5a6bfc068dc8c)
1 /******************************************************************************
2 *******************************************************************************
3 **
4 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
5 **  Copyright (C) 2005 Red Hat, Inc.  All rights reserved.
6 **
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.
10 **
11 *******************************************************************************
12 ******************************************************************************/
13 
14 #include "dlm_internal.h"
15 #include "lockspace.h"
16 #include "member.h"
17 #include "lowcomms.h"
18 #include "midcomms.h"
19 #include "rcom.h"
20 #include "recover.h"
21 #include "dir.h"
22 #include "config.h"
23 #include "memory.h"
24 #include "lock.h"
25 #include "util.h"
26 
27 
28 static int rcom_response(struct dlm_ls *ls)
29 {
30 	return test_bit(LSFL_RCOM_READY, &ls->ls_flags);
31 }
32 
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)
35 {
36 	struct dlm_rcom *rc;
37 	struct dlm_mhandle *mh;
38 	char *mb;
39 	int mb_len = sizeof(struct dlm_rcom) + len;
40 
41 	mh = dlm_lowcomms_get_buffer(to_nodeid, mb_len, GFP_KERNEL, &mb);
42 	if (!mh) {
43 		log_print("create_rcom to %d type %d len %d ENOBUFS",
44 			  to_nodeid, type, len);
45 		return -ENOBUFS;
46 	}
47 	memset(mb, 0, mb_len);
48 
49 	rc = (struct dlm_rcom *) mb;
50 
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;
56 
57 	rc->rc_type = type;
58 
59 	spin_lock(&ls->ls_recover_lock);
60 	rc->rc_seq = ls->ls_recover_seq;
61 	spin_unlock(&ls->ls_recover_lock);
62 
63 	*mh_ret = mh;
64 	*rc_ret = rc;
65 	return 0;
66 }
67 
68 static void send_rcom(struct dlm_ls *ls, struct dlm_mhandle *mh,
69 		      struct dlm_rcom *rc)
70 {
71 	dlm_rcom_out(rc);
72 	dlm_lowcomms_commit_buffer(mh);
73 }
74 
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. */
78 
79 static void make_config(struct dlm_ls *ls, struct rcom_config *rf)
80 {
81 	rf->rf_lvblen = ls->ls_lvblen;
82 	rf->rf_lsflags = ls->ls_exflags;
83 }
84 
85 static int check_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
86 {
87 	struct rcom_config *rf = (struct rcom_config *) rc->rc_buf;
88 
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);
93 		return -EINVAL;
94 	}
95 
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);
101 		return -EINVAL;
102 	}
103 	return 0;
104 }
105 
106 static void allow_sync_reply(struct dlm_ls *ls, uint64_t *new_seq)
107 {
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);
112 }
113 
114 static void disallow_sync_reply(struct dlm_ls *ls)
115 {
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);
120 }
121 
122 int dlm_rcom_status(struct dlm_ls *ls, int nodeid)
123 {
124 	struct dlm_rcom *rc;
125 	struct dlm_mhandle *mh;
126 	int error = 0;
127 
128 	ls->ls_recover_nodeid = nodeid;
129 
130 	if (nodeid == dlm_our_nodeid()) {
131 		rc = (struct dlm_rcom *) ls->ls_recover_buf;
132 		rc->rc_result = dlm_recover_status(ls);
133 		goto out;
134 	}
135 
136 	error = create_rcom(ls, nodeid, DLM_RCOM_STATUS, 0, &rc, &mh);
137 	if (error)
138 		goto out;
139 
140 	allow_sync_reply(ls, &rc->rc_id);
141 	memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size);
142 
143 	send_rcom(ls, mh, rc);
144 
145 	error = dlm_wait_function(ls, &rcom_response);
146 	disallow_sync_reply(ls);
147 	if (error)
148 		goto out;
149 
150 	rc = (struct dlm_rcom *) ls->ls_recover_buf;
151 
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);
155 		rc->rc_result = 0;
156 	} else
157 		error = check_config(ls, rc, nodeid);
158 	/* the caller looks at rc_result for the remote recovery status */
159  out:
160 	return error;
161 }
162 
163 static void receive_rcom_status(struct dlm_ls *ls, struct dlm_rcom *rc_in)
164 {
165 	struct dlm_rcom *rc;
166 	struct dlm_mhandle *mh;
167 	int error, nodeid = rc_in->rc_header.h_nodeid;
168 
169 	error = create_rcom(ls, nodeid, DLM_RCOM_STATUS_REPLY,
170 			    sizeof(struct rcom_config), &rc, &mh);
171 	if (error)
172 		return;
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);
177 
178 	send_rcom(ls, mh, rc);
179 }
180 
181 static void receive_sync_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
182 {
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);
190 		goto out;
191 	}
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);
196  out:
197 	spin_unlock(&ls->ls_rcom_spin);
198 }
199 
200 static void receive_rcom_status_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
201 {
202 	receive_sync_reply(ls, rc_in);
203 }
204 
205 int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
206 {
207 	struct dlm_rcom *rc;
208 	struct dlm_mhandle *mh;
209 	int error = 0, len = sizeof(struct dlm_rcom);
210 
211 	ls->ls_recover_nodeid = nodeid;
212 
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);
217 		goto out;
218 	}
219 
220 	error = create_rcom(ls, nodeid, DLM_RCOM_NAMES, last_len, &rc, &mh);
221 	if (error)
222 		goto out;
223 	memcpy(rc->rc_buf, last_name, last_len);
224 
225 	allow_sync_reply(ls, &rc->rc_id);
226 	memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size);
227 
228 	send_rcom(ls, mh, rc);
229 
230 	error = dlm_wait_function(ls, &rcom_response);
231 	disallow_sync_reply(ls);
232  out:
233 	return error;
234 }
235 
236 static void receive_rcom_names(struct dlm_ls *ls, struct dlm_rcom *rc_in)
237 {
238 	struct dlm_rcom *rc;
239 	struct dlm_mhandle *mh;
240 	int error, inlen, outlen, nodeid;
241 
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);
245 
246 	error = create_rcom(ls, nodeid, DLM_RCOM_NAMES_REPLY, outlen, &rc, &mh);
247 	if (error)
248 		return;
249 	rc->rc_id = rc_in->rc_id;
250 	rc->rc_seq_reply = rc_in->rc_seq;
251 
252 	dlm_copy_master_names(ls, rc_in->rc_buf, inlen, rc->rc_buf, outlen,
253 			      nodeid);
254 	send_rcom(ls, mh, rc);
255 }
256 
257 static void receive_rcom_names_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
258 {
259 	receive_sync_reply(ls, rc_in);
260 }
261 
262 int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
263 {
264 	struct dlm_rcom *rc;
265 	struct dlm_mhandle *mh;
266 	struct dlm_ls *ls = r->res_ls;
267 	int error;
268 
269 	error = create_rcom(ls, dir_nodeid, DLM_RCOM_LOOKUP, r->res_length,
270 			    &rc, &mh);
271 	if (error)
272 		goto out;
273 	memcpy(rc->rc_buf, r->res_name, r->res_length);
274 	rc->rc_id = (unsigned long) r;
275 
276 	send_rcom(ls, mh, rc);
277  out:
278 	return error;
279 }
280 
281 static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
282 {
283 	struct dlm_rcom *rc;
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);
287 
288 	error = create_rcom(ls, nodeid, DLM_RCOM_LOOKUP_REPLY, 0, &rc, &mh);
289 	if (error)
290 		return;
291 
292 	error = dlm_dir_lookup(ls, nodeid, rc_in->rc_buf, len, &ret_nodeid);
293 	if (error)
294 		ret_nodeid = error;
295 	rc->rc_result = ret_nodeid;
296 	rc->rc_id = rc_in->rc_id;
297 	rc->rc_seq_reply = rc_in->rc_seq;
298 
299 	send_rcom(ls, mh, rc);
300 }
301 
302 static void receive_rcom_lookup_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
303 {
304 	dlm_recover_master_reply(ls, rc_in);
305 }
306 
307 static void pack_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb,
308 			   struct rcom_lock *rl)
309 {
310 	memset(rl, 0, sizeof(*rl));
311 
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;
321 
322 	if (lkb->lkb_bastaddr)
323 		rl->rl_asts |= AST_BAST;
324 	if (lkb->lkb_astaddr)
325 		rl->rl_asts |= AST_COMP;
326 
327 	rl->rl_namelen = r->res_length;
328 	memcpy(rl->rl_name, r->res_name, r->res_length);
329 
330 	/* FIXME: might we have an lvb without DLM_LKF_VALBLK set ?
331 	   If so, receive_rcom_lock_args() won't take this copy. */
332 
333 	if (lkb->lkb_lvbptr)
334 		memcpy(rl->rl_lvb, lkb->lkb_lvbptr, r->res_ls->ls_lvblen);
335 }
336 
337 int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
338 {
339 	struct dlm_ls *ls = r->res_ls;
340 	struct dlm_rcom *rc;
341 	struct dlm_mhandle *mh;
342 	struct rcom_lock *rl;
343 	int error, len = sizeof(struct rcom_lock);
344 
345 	if (lkb->lkb_lvbptr)
346 		len += ls->ls_lvblen;
347 
348 	error = create_rcom(ls, r->res_nodeid, DLM_RCOM_LOCK, len, &rc, &mh);
349 	if (error)
350 		goto out;
351 
352 	rl = (struct rcom_lock *) rc->rc_buf;
353 	pack_rcom_lock(r, lkb, rl);
354 	rc->rc_id = (unsigned long) r;
355 
356 	send_rcom(ls, mh, rc);
357  out:
358 	return error;
359 }
360 
361 static void receive_rcom_lock(struct dlm_ls *ls, struct dlm_rcom *rc_in)
362 {
363 	struct dlm_rcom *rc;
364 	struct dlm_mhandle *mh;
365 	int error, nodeid = rc_in->rc_header.h_nodeid;
366 
367 	dlm_recover_master_copy(ls, rc_in);
368 
369 	error = create_rcom(ls, nodeid, DLM_RCOM_LOCK_REPLY,
370 			    sizeof(struct rcom_lock), &rc, &mh);
371 	if (error)
372 		return;
373 
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 */
376 
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;
380 
381 	send_rcom(ls, mh, rc);
382 }
383 
384 static void receive_rcom_lock_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
385 {
386 	dlm_recover_process_copy(ls, rc_in);
387 }
388 
389 static int send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
390 {
391 	struct dlm_rcom *rc;
392 	struct rcom_config *rf;
393 	struct dlm_mhandle *mh;
394 	char *mb;
395 	int mb_len = sizeof(struct dlm_rcom) + sizeof(struct rcom_config);
396 
397 	mh = dlm_lowcomms_get_buffer(nodeid, mb_len, GFP_KERNEL, &mb);
398 	if (!mh)
399 		return -ENOBUFS;
400 	memset(mb, 0, mb_len);
401 
402 	rc = (struct dlm_rcom *) mb;
403 
404 	rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
405 	rc->rc_header.h_lockspace = rc_in->rc_header.h_lockspace;
406 	rc->rc_header.h_nodeid = dlm_our_nodeid();
407 	rc->rc_header.h_length = mb_len;
408 	rc->rc_header.h_cmd = DLM_RCOM;
409 
410 	rc->rc_type = DLM_RCOM_STATUS_REPLY;
411 	rc->rc_id = rc_in->rc_id;
412 	rc->rc_seq_reply = rc_in->rc_seq;
413 	rc->rc_result = -ESRCH;
414 
415 	rf = (struct rcom_config *) rc->rc_buf;
416 	rf->rf_lvblen = -1;
417 
418 	dlm_rcom_out(rc);
419 	dlm_lowcomms_commit_buffer(mh);
420 
421 	return 0;
422 }
423 
424 static int is_old_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
425 {
426 	uint64_t seq;
427 	int rv = 0;
428 
429 	switch (rc->rc_type) {
430 	case DLM_RCOM_STATUS_REPLY:
431 	case DLM_RCOM_NAMES_REPLY:
432 	case DLM_RCOM_LOOKUP_REPLY:
433 	case DLM_RCOM_LOCK_REPLY:
434 		spin_lock(&ls->ls_recover_lock);
435 		seq = ls->ls_recover_seq;
436 		spin_unlock(&ls->ls_recover_lock);
437 		if (rc->rc_seq_reply != seq) {
438 			log_debug(ls, "ignoring old reply %x from %d "
439 				      "seq_reply %llx expect %llx",
440 				      rc->rc_type, rc->rc_header.h_nodeid,
441 				      (unsigned long long)rc->rc_seq_reply,
442 				      (unsigned long long)seq);
443 			rv = 1;
444 		}
445 	}
446 	return rv;
447 }
448 
449 /* Called by dlm_recvd; corresponds to dlm_receive_message() but special
450    recovery-only comms are sent through here. */
451 
452 void dlm_receive_rcom(struct dlm_header *hd, int nodeid)
453 {
454 	struct dlm_rcom *rc = (struct dlm_rcom *) hd;
455 	struct dlm_ls *ls;
456 
457 	dlm_rcom_in(rc);
458 
459 	/* If the lockspace doesn't exist then still send a status message
460 	   back; it's possible that it just doesn't have its global_id yet. */
461 
462 	ls = dlm_find_lockspace_global(hd->h_lockspace);
463 	if (!ls) {
464 		log_print("lockspace %x from %d type %x not found",
465 			  hd->h_lockspace, nodeid, rc->rc_type);
466 		if (rc->rc_type == DLM_RCOM_STATUS)
467 			send_ls_not_ready(nodeid, rc);
468 		return;
469 	}
470 
471 	if (dlm_recovery_stopped(ls) && (rc->rc_type != DLM_RCOM_STATUS)) {
472 		log_debug(ls, "ignoring recovery message %x from %d",
473 			  rc->rc_type, nodeid);
474 		goto out;
475 	}
476 
477 	if (is_old_reply(ls, rc))
478 		goto out;
479 
480 	if (nodeid != rc->rc_header.h_nodeid) {
481 		log_error(ls, "bad rcom nodeid %d from %d",
482 			  rc->rc_header.h_nodeid, nodeid);
483 		goto out;
484 	}
485 
486 	switch (rc->rc_type) {
487 	case DLM_RCOM_STATUS:
488 		receive_rcom_status(ls, rc);
489 		break;
490 
491 	case DLM_RCOM_NAMES:
492 		receive_rcom_names(ls, rc);
493 		break;
494 
495 	case DLM_RCOM_LOOKUP:
496 		receive_rcom_lookup(ls, rc);
497 		break;
498 
499 	case DLM_RCOM_LOCK:
500 		receive_rcom_lock(ls, rc);
501 		break;
502 
503 	case DLM_RCOM_STATUS_REPLY:
504 		receive_rcom_status_reply(ls, rc);
505 		break;
506 
507 	case DLM_RCOM_NAMES_REPLY:
508 		receive_rcom_names_reply(ls, rc);
509 		break;
510 
511 	case DLM_RCOM_LOOKUP_REPLY:
512 		receive_rcom_lookup_reply(ls, rc);
513 		break;
514 
515 	case DLM_RCOM_LOCK_REPLY:
516 		receive_rcom_lock_reply(ls, rc);
517 		break;
518 
519 	default:
520 		DLM_ASSERT(0, printk("rc_type=%x\n", rc->rc_type););
521 	}
522  out:
523 	dlm_put_lockspace(ls);
524 }
525 
526