Merge tag 'pm-5.11-rc1-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
[platform/kernel/linux-starfive.git] / fs / dlm / rcom.c
1 // SPDX-License-Identifier: GPL-2.0-only
2 /******************************************************************************
3 *******************************************************************************
4 **
5 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
6 **  Copyright (C) 2005-2008 Red Hat, Inc.  All rights reserved.
7 **
8 **
9 *******************************************************************************
10 ******************************************************************************/
11
12 #include "dlm_internal.h"
13 #include "lockspace.h"
14 #include "member.h"
15 #include "lowcomms.h"
16 #include "midcomms.h"
17 #include "rcom.h"
18 #include "recover.h"
19 #include "dir.h"
20 #include "config.h"
21 #include "memory.h"
22 #include "lock.h"
23 #include "util.h"
24
25 static int rcom_response(struct dlm_ls *ls)
26 {
27         return test_bit(LSFL_RCOM_READY, &ls->ls_flags);
28 }
29
30 static int create_rcom(struct dlm_ls *ls, int to_nodeid, int type, int len,
31                        struct dlm_rcom **rc_ret, struct dlm_mhandle **mh_ret)
32 {
33         struct dlm_rcom *rc;
34         struct dlm_mhandle *mh;
35         char *mb;
36         int mb_len = sizeof(struct dlm_rcom) + len;
37
38         mh = dlm_lowcomms_get_buffer(to_nodeid, mb_len, GFP_NOFS, &mb);
39         if (!mh) {
40                 log_print("create_rcom to %d type %d len %d ENOBUFS",
41                           to_nodeid, type, len);
42                 return -ENOBUFS;
43         }
44         memset(mb, 0, mb_len);
45
46         rc = (struct dlm_rcom *) mb;
47
48         rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
49         rc->rc_header.h_lockspace = ls->ls_global_id;
50         rc->rc_header.h_nodeid = dlm_our_nodeid();
51         rc->rc_header.h_length = mb_len;
52         rc->rc_header.h_cmd = DLM_RCOM;
53
54         rc->rc_type = type;
55
56         spin_lock(&ls->ls_recover_lock);
57         rc->rc_seq = ls->ls_recover_seq;
58         spin_unlock(&ls->ls_recover_lock);
59
60         *mh_ret = mh;
61         *rc_ret = rc;
62         return 0;
63 }
64
65 static void send_rcom(struct dlm_ls *ls, struct dlm_mhandle *mh,
66                       struct dlm_rcom *rc)
67 {
68         dlm_rcom_out(rc);
69         dlm_lowcomms_commit_buffer(mh);
70 }
71
72 static void set_rcom_status(struct dlm_ls *ls, struct rcom_status *rs,
73                             uint32_t flags)
74 {
75         rs->rs_flags = cpu_to_le32(flags);
76 }
77
78 /* When replying to a status request, a node also sends back its
79    configuration values.  The requesting node then checks that the remote
80    node is configured the same way as itself. */
81
82 static void set_rcom_config(struct dlm_ls *ls, struct rcom_config *rf,
83                             uint32_t num_slots)
84 {
85         rf->rf_lvblen = cpu_to_le32(ls->ls_lvblen);
86         rf->rf_lsflags = cpu_to_le32(ls->ls_exflags);
87
88         rf->rf_our_slot = cpu_to_le16(ls->ls_slot);
89         rf->rf_num_slots = cpu_to_le16(num_slots);
90         rf->rf_generation =  cpu_to_le32(ls->ls_generation);
91 }
92
93 static int check_rcom_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
94 {
95         struct rcom_config *rf = (struct rcom_config *) rc->rc_buf;
96
97         if ((rc->rc_header.h_version & 0xFFFF0000) != DLM_HEADER_MAJOR) {
98                 log_error(ls, "version mismatch: %x nodeid %d: %x",
99                           DLM_HEADER_MAJOR | DLM_HEADER_MINOR, nodeid,
100                           rc->rc_header.h_version);
101                 return -EPROTO;
102         }
103
104         if (le32_to_cpu(rf->rf_lvblen) != ls->ls_lvblen ||
105             le32_to_cpu(rf->rf_lsflags) != ls->ls_exflags) {
106                 log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x",
107                           ls->ls_lvblen, ls->ls_exflags, nodeid,
108                           le32_to_cpu(rf->rf_lvblen),
109                           le32_to_cpu(rf->rf_lsflags));
110                 return -EPROTO;
111         }
112         return 0;
113 }
114
115 static void allow_sync_reply(struct dlm_ls *ls, uint64_t *new_seq)
116 {
117         spin_lock(&ls->ls_rcom_spin);
118         *new_seq = ++ls->ls_rcom_seq;
119         set_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
120         spin_unlock(&ls->ls_rcom_spin);
121 }
122
123 static void disallow_sync_reply(struct dlm_ls *ls)
124 {
125         spin_lock(&ls->ls_rcom_spin);
126         clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
127         clear_bit(LSFL_RCOM_READY, &ls->ls_flags);
128         spin_unlock(&ls->ls_rcom_spin);
129 }
130
131 /*
132  * low nodeid gathers one slot value at a time from each node.
133  * it sets need_slots=0, and saves rf_our_slot returned from each
134  * rcom_config.
135  *
136  * other nodes gather all slot values at once from the low nodeid.
137  * they set need_slots=1, and ignore the rf_our_slot returned from each
138  * rcom_config.  they use the rf_num_slots returned from the low
139  * node's rcom_config.
140  */
141
142 int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags)
143 {
144         struct dlm_rcom *rc;
145         struct dlm_mhandle *mh;
146         int error = 0;
147
148         ls->ls_recover_nodeid = nodeid;
149
150         if (nodeid == dlm_our_nodeid()) {
151                 rc = ls->ls_recover_buf;
152                 rc->rc_result = dlm_recover_status(ls);
153                 goto out;
154         }
155
156 retry:
157         error = create_rcom(ls, nodeid, DLM_RCOM_STATUS,
158                             sizeof(struct rcom_status), &rc, &mh);
159         if (error)
160                 goto out;
161
162         set_rcom_status(ls, (struct rcom_status *)rc->rc_buf, status_flags);
163
164         allow_sync_reply(ls, &rc->rc_id);
165         memset(ls->ls_recover_buf, 0, LOWCOMMS_MAX_TX_BUFFER_LEN);
166
167         send_rcom(ls, mh, rc);
168
169         error = dlm_wait_function(ls, &rcom_response);
170         disallow_sync_reply(ls);
171         if (error == -ETIMEDOUT)
172                 goto retry;
173         if (error)
174                 goto out;
175
176         rc = ls->ls_recover_buf;
177
178         if (rc->rc_result == -ESRCH) {
179                 /* we pretend the remote lockspace exists with 0 status */
180                 log_debug(ls, "remote node %d not ready", nodeid);
181                 rc->rc_result = 0;
182                 error = 0;
183         } else {
184                 error = check_rcom_config(ls, rc, nodeid);
185         }
186
187         /* the caller looks at rc_result for the remote recovery status */
188  out:
189         return error;
190 }
191
192 static void receive_rcom_status(struct dlm_ls *ls, struct dlm_rcom *rc_in)
193 {
194         struct dlm_rcom *rc;
195         struct dlm_mhandle *mh;
196         struct rcom_status *rs;
197         uint32_t status;
198         int nodeid = rc_in->rc_header.h_nodeid;
199         int len = sizeof(struct rcom_config);
200         int num_slots = 0;
201         int error;
202
203         if (!dlm_slots_version(&rc_in->rc_header)) {
204                 status = dlm_recover_status(ls);
205                 goto do_create;
206         }
207
208         rs = (struct rcom_status *)rc_in->rc_buf;
209
210         if (!(le32_to_cpu(rs->rs_flags) & DLM_RSF_NEED_SLOTS)) {
211                 status = dlm_recover_status(ls);
212                 goto do_create;
213         }
214
215         spin_lock(&ls->ls_recover_lock);
216         status = ls->ls_recover_status;
217         num_slots = ls->ls_num_slots;
218         spin_unlock(&ls->ls_recover_lock);
219         len += num_slots * sizeof(struct rcom_slot);
220
221  do_create:
222         error = create_rcom(ls, nodeid, DLM_RCOM_STATUS_REPLY,
223                             len, &rc, &mh);
224         if (error)
225                 return;
226
227         rc->rc_id = rc_in->rc_id;
228         rc->rc_seq_reply = rc_in->rc_seq;
229         rc->rc_result = status;
230
231         set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, num_slots);
232
233         if (!num_slots)
234                 goto do_send;
235
236         spin_lock(&ls->ls_recover_lock);
237         if (ls->ls_num_slots != num_slots) {
238                 spin_unlock(&ls->ls_recover_lock);
239                 log_debug(ls, "receive_rcom_status num_slots %d to %d",
240                           num_slots, ls->ls_num_slots);
241                 rc->rc_result = 0;
242                 set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, 0);
243                 goto do_send;
244         }
245
246         dlm_slots_copy_out(ls, rc);
247         spin_unlock(&ls->ls_recover_lock);
248
249  do_send:
250         send_rcom(ls, mh, rc);
251 }
252
253 static void receive_sync_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
254 {
255         spin_lock(&ls->ls_rcom_spin);
256         if (!test_bit(LSFL_RCOM_WAIT, &ls->ls_flags) ||
257             rc_in->rc_id != ls->ls_rcom_seq) {
258                 log_debug(ls, "reject reply %d from %d seq %llx expect %llx",
259                           rc_in->rc_type, rc_in->rc_header.h_nodeid,
260                           (unsigned long long)rc_in->rc_id,
261                           (unsigned long long)ls->ls_rcom_seq);
262                 goto out;
263         }
264         memcpy(ls->ls_recover_buf, rc_in, rc_in->rc_header.h_length);
265         set_bit(LSFL_RCOM_READY, &ls->ls_flags);
266         clear_bit(LSFL_RCOM_WAIT, &ls->ls_flags);
267         wake_up(&ls->ls_wait_general);
268  out:
269         spin_unlock(&ls->ls_rcom_spin);
270 }
271
272 int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
273 {
274         struct dlm_rcom *rc;
275         struct dlm_mhandle *mh;
276         int error = 0;
277
278         ls->ls_recover_nodeid = nodeid;
279
280 retry:
281         error = create_rcom(ls, nodeid, DLM_RCOM_NAMES, last_len, &rc, &mh);
282         if (error)
283                 goto out;
284         memcpy(rc->rc_buf, last_name, last_len);
285
286         allow_sync_reply(ls, &rc->rc_id);
287         memset(ls->ls_recover_buf, 0, LOWCOMMS_MAX_TX_BUFFER_LEN);
288
289         send_rcom(ls, mh, rc);
290
291         error = dlm_wait_function(ls, &rcom_response);
292         disallow_sync_reply(ls);
293         if (error == -ETIMEDOUT)
294                 goto retry;
295  out:
296         return error;
297 }
298
299 static void receive_rcom_names(struct dlm_ls *ls, struct dlm_rcom *rc_in)
300 {
301         struct dlm_rcom *rc;
302         struct dlm_mhandle *mh;
303         int error, inlen, outlen, nodeid;
304
305         nodeid = rc_in->rc_header.h_nodeid;
306         inlen = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
307         outlen = LOWCOMMS_MAX_TX_BUFFER_LEN - sizeof(struct dlm_rcom);
308
309         error = create_rcom(ls, nodeid, DLM_RCOM_NAMES_REPLY, outlen, &rc, &mh);
310         if (error)
311                 return;
312         rc->rc_id = rc_in->rc_id;
313         rc->rc_seq_reply = rc_in->rc_seq;
314
315         dlm_copy_master_names(ls, rc_in->rc_buf, inlen, rc->rc_buf, outlen,
316                               nodeid);
317         send_rcom(ls, mh, rc);
318 }
319
320 int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
321 {
322         struct dlm_rcom *rc;
323         struct dlm_mhandle *mh;
324         struct dlm_ls *ls = r->res_ls;
325         int error;
326
327         error = create_rcom(ls, dir_nodeid, DLM_RCOM_LOOKUP, r->res_length,
328                             &rc, &mh);
329         if (error)
330                 goto out;
331         memcpy(rc->rc_buf, r->res_name, r->res_length);
332         rc->rc_id = (unsigned long) r->res_id;
333
334         send_rcom(ls, mh, rc);
335  out:
336         return error;
337 }
338
339 static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
340 {
341         struct dlm_rcom *rc;
342         struct dlm_mhandle *mh;
343         int error, ret_nodeid, nodeid = rc_in->rc_header.h_nodeid;
344         int len = rc_in->rc_header.h_length - sizeof(struct dlm_rcom);
345
346         error = create_rcom(ls, nodeid, DLM_RCOM_LOOKUP_REPLY, 0, &rc, &mh);
347         if (error)
348                 return;
349
350         /* Old code would send this special id to trigger a debug dump. */
351         if (rc_in->rc_id == 0xFFFFFFFF) {
352                 log_error(ls, "receive_rcom_lookup dump from %d", nodeid);
353                 dlm_dump_rsb_name(ls, rc_in->rc_buf, len);
354                 return;
355         }
356
357         error = dlm_master_lookup(ls, nodeid, rc_in->rc_buf, len,
358                                   DLM_LU_RECOVER_MASTER, &ret_nodeid, NULL);
359         if (error)
360                 ret_nodeid = error;
361         rc->rc_result = ret_nodeid;
362         rc->rc_id = rc_in->rc_id;
363         rc->rc_seq_reply = rc_in->rc_seq;
364
365         send_rcom(ls, mh, rc);
366 }
367
368 static void receive_rcom_lookup_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
369 {
370         dlm_recover_master_reply(ls, rc_in);
371 }
372
373 static void pack_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb,
374                            struct rcom_lock *rl)
375 {
376         memset(rl, 0, sizeof(*rl));
377
378         rl->rl_ownpid = cpu_to_le32(lkb->lkb_ownpid);
379         rl->rl_lkid = cpu_to_le32(lkb->lkb_id);
380         rl->rl_exflags = cpu_to_le32(lkb->lkb_exflags);
381         rl->rl_flags = cpu_to_le32(lkb->lkb_flags);
382         rl->rl_lvbseq = cpu_to_le32(lkb->lkb_lvbseq);
383         rl->rl_rqmode = lkb->lkb_rqmode;
384         rl->rl_grmode = lkb->lkb_grmode;
385         rl->rl_status = lkb->lkb_status;
386         rl->rl_wait_type = cpu_to_le16(lkb->lkb_wait_type);
387
388         if (lkb->lkb_bastfn)
389                 rl->rl_asts |= DLM_CB_BAST;
390         if (lkb->lkb_astfn)
391                 rl->rl_asts |= DLM_CB_CAST;
392
393         rl->rl_namelen = cpu_to_le16(r->res_length);
394         memcpy(rl->rl_name, r->res_name, r->res_length);
395
396         /* FIXME: might we have an lvb without DLM_LKF_VALBLK set ?
397            If so, receive_rcom_lock_args() won't take this copy. */
398
399         if (lkb->lkb_lvbptr)
400                 memcpy(rl->rl_lvb, lkb->lkb_lvbptr, r->res_ls->ls_lvblen);
401 }
402
403 int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
404 {
405         struct dlm_ls *ls = r->res_ls;
406         struct dlm_rcom *rc;
407         struct dlm_mhandle *mh;
408         struct rcom_lock *rl;
409         int error, len = sizeof(struct rcom_lock);
410
411         if (lkb->lkb_lvbptr)
412                 len += ls->ls_lvblen;
413
414         error = create_rcom(ls, r->res_nodeid, DLM_RCOM_LOCK, len, &rc, &mh);
415         if (error)
416                 goto out;
417
418         rl = (struct rcom_lock *) rc->rc_buf;
419         pack_rcom_lock(r, lkb, rl);
420         rc->rc_id = (unsigned long) r;
421
422         send_rcom(ls, mh, rc);
423  out:
424         return error;
425 }
426
427 /* needs at least dlm_rcom + rcom_lock */
428 static void receive_rcom_lock(struct dlm_ls *ls, struct dlm_rcom *rc_in)
429 {
430         struct dlm_rcom *rc;
431         struct dlm_mhandle *mh;
432         int error, nodeid = rc_in->rc_header.h_nodeid;
433
434         dlm_recover_master_copy(ls, rc_in);
435
436         error = create_rcom(ls, nodeid, DLM_RCOM_LOCK_REPLY,
437                             sizeof(struct rcom_lock), &rc, &mh);
438         if (error)
439                 return;
440
441         /* We send back the same rcom_lock struct we received, but
442            dlm_recover_master_copy() has filled in rl_remid and rl_result */
443
444         memcpy(rc->rc_buf, rc_in->rc_buf, sizeof(struct rcom_lock));
445         rc->rc_id = rc_in->rc_id;
446         rc->rc_seq_reply = rc_in->rc_seq;
447
448         send_rcom(ls, mh, rc);
449 }
450
451 /* If the lockspace doesn't exist then still send a status message
452    back; it's possible that it just doesn't have its global_id yet. */
453
454 int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
455 {
456         struct dlm_rcom *rc;
457         struct rcom_config *rf;
458         struct dlm_mhandle *mh;
459         char *mb;
460         int mb_len = sizeof(struct dlm_rcom) + sizeof(struct rcom_config);
461
462         mh = dlm_lowcomms_get_buffer(nodeid, mb_len, GFP_NOFS, &mb);
463         if (!mh)
464                 return -ENOBUFS;
465         memset(mb, 0, mb_len);
466
467         rc = (struct dlm_rcom *) mb;
468
469         rc->rc_header.h_version = (DLM_HEADER_MAJOR | DLM_HEADER_MINOR);
470         rc->rc_header.h_lockspace = rc_in->rc_header.h_lockspace;
471         rc->rc_header.h_nodeid = dlm_our_nodeid();
472         rc->rc_header.h_length = mb_len;
473         rc->rc_header.h_cmd = DLM_RCOM;
474
475         rc->rc_type = DLM_RCOM_STATUS_REPLY;
476         rc->rc_id = rc_in->rc_id;
477         rc->rc_seq_reply = rc_in->rc_seq;
478         rc->rc_result = -ESRCH;
479
480         rf = (struct rcom_config *) rc->rc_buf;
481         rf->rf_lvblen = cpu_to_le32(~0U);
482
483         dlm_rcom_out(rc);
484         dlm_lowcomms_commit_buffer(mh);
485
486         return 0;
487 }
488
489 /*
490  * Ignore messages for stage Y before we set
491  * recover_status bit for stage X:
492  *
493  * recover_status = 0
494  *
495  * dlm_recover_members()
496  * - send nothing
497  * - recv nothing
498  * - ignore NAMES, NAMES_REPLY
499  * - ignore LOOKUP, LOOKUP_REPLY
500  * - ignore LOCK, LOCK_REPLY
501  *
502  * recover_status |= NODES
503  *
504  * dlm_recover_members_wait()
505  *
506  * dlm_recover_directory()
507  * - send NAMES
508  * - recv NAMES_REPLY
509  * - ignore LOOKUP, LOOKUP_REPLY
510  * - ignore LOCK, LOCK_REPLY
511  *
512  * recover_status |= DIR
513  *
514  * dlm_recover_directory_wait()
515  *
516  * dlm_recover_masters()
517  * - send LOOKUP
518  * - recv LOOKUP_REPLY
519  *
520  * dlm_recover_locks()
521  * - send LOCKS
522  * - recv LOCKS_REPLY
523  *
524  * recover_status |= LOCKS
525  *
526  * dlm_recover_locks_wait()
527  *
528  * recover_status |= DONE
529  */
530
531 /* Called by dlm_recv; corresponds to dlm_receive_message() but special
532    recovery-only comms are sent through here. */
533
534 void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
535 {
536         int lock_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_lock);
537         int stop, reply = 0, names = 0, lookup = 0, lock = 0;
538         uint32_t status;
539         uint64_t seq;
540
541         switch (rc->rc_type) {
542         case DLM_RCOM_STATUS_REPLY:
543                 reply = 1;
544                 break;
545         case DLM_RCOM_NAMES:
546                 names = 1;
547                 break;
548         case DLM_RCOM_NAMES_REPLY:
549                 names = 1;
550                 reply = 1;
551                 break;
552         case DLM_RCOM_LOOKUP:
553                 lookup = 1;
554                 break;
555         case DLM_RCOM_LOOKUP_REPLY:
556                 lookup = 1;
557                 reply = 1;
558                 break;
559         case DLM_RCOM_LOCK:
560                 lock = 1;
561                 break;
562         case DLM_RCOM_LOCK_REPLY:
563                 lock = 1;
564                 reply = 1;
565                 break;
566         }
567
568         spin_lock(&ls->ls_recover_lock);
569         status = ls->ls_recover_status;
570         stop = test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
571         seq = ls->ls_recover_seq;
572         spin_unlock(&ls->ls_recover_lock);
573
574         if (stop && (rc->rc_type != DLM_RCOM_STATUS))
575                 goto ignore;
576
577         if (reply && (rc->rc_seq_reply != seq))
578                 goto ignore;
579
580         if (!(status & DLM_RS_NODES) && (names || lookup || lock))
581                 goto ignore;
582
583         if (!(status & DLM_RS_DIR) && (lookup || lock))
584                 goto ignore;
585
586         switch (rc->rc_type) {
587         case DLM_RCOM_STATUS:
588                 receive_rcom_status(ls, rc);
589                 break;
590
591         case DLM_RCOM_NAMES:
592                 receive_rcom_names(ls, rc);
593                 break;
594
595         case DLM_RCOM_LOOKUP:
596                 receive_rcom_lookup(ls, rc);
597                 break;
598
599         case DLM_RCOM_LOCK:
600                 if (rc->rc_header.h_length < lock_size)
601                         goto Eshort;
602                 receive_rcom_lock(ls, rc);
603                 break;
604
605         case DLM_RCOM_STATUS_REPLY:
606                 receive_sync_reply(ls, rc);
607                 break;
608
609         case DLM_RCOM_NAMES_REPLY:
610                 receive_sync_reply(ls, rc);
611                 break;
612
613         case DLM_RCOM_LOOKUP_REPLY:
614                 receive_rcom_lookup_reply(ls, rc);
615                 break;
616
617         case DLM_RCOM_LOCK_REPLY:
618                 if (rc->rc_header.h_length < lock_size)
619                         goto Eshort;
620                 dlm_recover_process_copy(ls, rc);
621                 break;
622
623         default:
624                 log_error(ls, "receive_rcom bad type %d", rc->rc_type);
625         }
626         return;
627
628 ignore:
629         log_limit(ls, "dlm_receive_rcom ignore msg %d "
630                   "from %d %llu %llu recover seq %llu sts %x gen %u",
631                    rc->rc_type,
632                    nodeid,
633                    (unsigned long long)rc->rc_seq,
634                    (unsigned long long)rc->rc_seq_reply,
635                    (unsigned long long)seq,
636                    status, ls->ls_generation);
637         return;
638 Eshort:
639         log_error(ls, "recovery message %d from %d is too short",
640                   rc->rc_type, nodeid);
641 }
642