projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drbd: store in-core bitmap little endian, regardless of architecture
[deliverable/linux.git] / drivers / block / drbd / drbd_nl.c
CommitLineData
b411b363
PR		 1/*
2 drbd_nl.c
3
4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
9
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
13 any later version.
14
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
19
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
23
24 */
25
b411b363
PR		 26#include <linux/module.h>
27#include <linux/drbd.h>
28#include <linux/in.h>
29#include <linux/fs.h>
30#include <linux/file.h>
31#include <linux/slab.h>
32#include <linux/connector.h>
33#include <linux/blkpg.h>
34#include <linux/cpumask.h>
35#include "drbd_int.h"
265be2d0 36#include "drbd_req.h"
b411b363
PR		 37#include "drbd_wrappers.h"
38#include <asm/unaligned.h>
39#include <linux/drbd_tag_magic.h>
40#include <linux/drbd_limits.h>
87f7be4c
PR		 41#include <linux/compiler.h>
42#include <linux/kthread.h>
b411b363
PR		 43
44static unsigned short *tl_add_blob(unsigned short *, enum drbd_tags, const void *, int);
45static unsigned short *tl_add_str(unsigned short *, enum drbd_tags, const char *);
46static unsigned short *tl_add_int(unsigned short *, enum drbd_tags, const void *);
47
48/* see get_sb_bdev and bd_claim */
49static char *drbd_m_holder = "Hands off! this is DRBD's meta data device.";
50
51/* Generate the tag_list to struct functions */
52#define NL_PACKET(name, number, fields) \
53static int name ## _from_tags(struct drbd_conf *mdev, \
54 unsigned short *tags, struct name *arg) __attribute__ ((unused)); \
55static int name ## _from_tags(struct drbd_conf *mdev, \
56 unsigned short *tags, struct name *arg) \
57{ \
58 int tag; \
59 int dlen; \
60 \
61 while ((tag = get_unaligned(tags++)) != TT_END) { \
62 dlen = get_unaligned(tags++); \
63 switch (tag_number(tag)) { \
64 fields \
65 default: \
66 if (tag & T_MANDATORY) { \
67 dev_err(DEV, "Unknown tag: %d\n", tag_number(tag)); \
68 return 0; \
69 } \
70 } \
71 tags = (unsigned short *)((char *)tags + dlen); \
72 } \
73 return 1; \
74}
75#define NL_INTEGER(pn, pr, member) \
76 case pn: /* D_ASSERT( tag_type(tag) == TT_INTEGER ); */ \
77 arg->member = get_unaligned((int *)(tags)); \
78 break;
79#define NL_INT64(pn, pr, member) \
80 case pn: /* D_ASSERT( tag_type(tag) == TT_INT64 ); */ \
81 arg->member = get_unaligned((u64 *)(tags)); \
82 break;
83#define NL_BIT(pn, pr, member) \
84 case pn: /* D_ASSERT( tag_type(tag) == TT_BIT ); */ \
85 arg->member = *(char *)(tags) ? 1 : 0; \
86 break;
87#define NL_STRING(pn, pr, member, len) \
88 case pn: /* D_ASSERT( tag_type(tag) == TT_STRING ); */ \
89 if (dlen > len) { \
90 dev_err(DEV, "arg too long: %s (%u wanted, max len: %u bytes)\n", \
91 #member, dlen, (unsigned int)len); \
92 return 0; \
93 } \
94 arg->member ## _len = dlen; \
95 memcpy(arg->member, tags, min_t(size_t, dlen, len)); \
96 break;
97#include "linux/drbd_nl.h"
98
99/* Generate the struct to tag_list functions */
100#define NL_PACKET(name, number, fields) \
101static unsigned short* \
102name ## _to_tags(struct drbd_conf *mdev, \
103 struct name *arg, unsigned short *tags) __attribute__ ((unused)); \
104static unsigned short* \
105name ## _to_tags(struct drbd_conf *mdev, \
106 struct name *arg, unsigned short *tags) \
107{ \
108 fields \
109 return tags; \
110}
111
112#define NL_INTEGER(pn, pr, member) \
113 put_unaligned(pn | pr | TT_INTEGER, tags++); \
114 put_unaligned(sizeof(int), tags++); \
115 put_unaligned(arg->member, (int *)tags); \
116 tags = (unsigned short *)((char *)tags+sizeof(int));
117#define NL_INT64(pn, pr, member) \
118 put_unaligned(pn | pr | TT_INT64, tags++); \
119 put_unaligned(sizeof(u64), tags++); \
120 put_unaligned(arg->member, (u64 *)tags); \
121 tags = (unsigned short *)((char *)tags+sizeof(u64));
122#define NL_BIT(pn, pr, member) \
123 put_unaligned(pn | pr | TT_BIT, tags++); \
124 put_unaligned(sizeof(char), tags++); \
125 *(char *)tags = arg->member; \
126 tags = (unsigned short *)((char *)tags+sizeof(char));
127#define NL_STRING(pn, pr, member, len) \
128 put_unaligned(pn | pr | TT_STRING, tags++); \
129 put_unaligned(arg->member ## _len, tags++); \
130 memcpy(tags, arg->member, arg->member ## _len); \
131 tags = (unsigned short *)((char *)tags + arg->member ## _len);
132#include "linux/drbd_nl.h"
133
134void drbd_bcast_ev_helper(struct drbd_conf *mdev, char *helper_name);
135void drbd_nl_send_reply(struct cn_msg *, int);
136
137int drbd_khelper(struct drbd_conf *mdev, char *cmd)
138{
139 char *envp[] = { "HOME=/",
140 "TERM=linux",
141 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
142 NULL, /* Will be set to address family */
143 NULL, /* Will be set to address */
144 NULL };
145
146 char mb[12], af[20], ad[60], *afs;
147 char *argv[] = {usermode_helper, cmd, mb, NULL };
148 int ret;
149
150 snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev));
151
152 if (get_net_conf(mdev)) {
153 switch (((struct sockaddr *)mdev->net_conf->peer_addr)->sa_family) {
154 case AF_INET6:
155 afs = "ipv6";
156 snprintf(ad, 60, "DRBD_PEER_ADDRESS=%pI6",
157 &((struct sockaddr_in6 *)mdev->net_conf->peer_addr)->sin6_addr);
158 break;
159 case AF_INET:
160 afs = "ipv4";
161 snprintf(ad, 60, "DRBD_PEER_ADDRESS=%pI4",
162 &((struct sockaddr_in *)mdev->net_conf->peer_addr)->sin_addr);
163 break;
164 default:
165 afs = "ssocks";
166 snprintf(ad, 60, "DRBD_PEER_ADDRESS=%pI4",
167 &((struct sockaddr_in *)mdev->net_conf->peer_addr)->sin_addr);
168 }
169 snprintf(af, 20, "DRBD_PEER_AF=%s", afs);
170 envp[3]=af;
171 envp[4]=ad;
172 put_net_conf(mdev);
173 }
174
1090c056
LE		 175 /* The helper may take some time.
176 * write out any unsynced meta data changes now */
177 drbd_md_sync(mdev);
178
b411b363
PR		 179 dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb);
180
181 drbd_bcast_ev_helper(mdev, cmd);
182 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
183 if (ret)
184 dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
185 usermode_helper, cmd, mb,
186 (ret >> 8) & 0xff, ret);
187 else
188 dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
189 usermode_helper, cmd, mb,
190 (ret >> 8) & 0xff, ret);
191
192 if (ret < 0) /* Ignore any ERRNOs we got. */
193 ret = 0;
194
195 return ret;
196}
197
198enum drbd_disk_state drbd_try_outdate_peer(struct drbd_conf *mdev)
199{
200 char *ex_to_string;
201 int r;
202 enum drbd_disk_state nps;
203 enum drbd_fencing_p fp;
204
205 D_ASSERT(mdev->state.pdsk == D_UNKNOWN);
206
207 if (get_ldev_if_state(mdev, D_CONSISTENT)) {
208 fp = mdev->ldev->dc.fencing;
209 put_ldev(mdev);
210 } else {
211 dev_warn(DEV, "Not fencing peer, I'm not even Consistent myself.\n");
fb22c402
PR		 212 nps = mdev->state.pdsk;
213 goto out;
b411b363
PR		 214 }
215
b411b363
PR		 216 r = drbd_khelper(mdev, "fence-peer");
217
218 switch ((r>>8) & 0xff) {
219 case 3: /* peer is inconsistent */
220 ex_to_string = "peer is inconsistent or worse";
221 nps = D_INCONSISTENT;
222 break;
223 case 4: /* peer got outdated, or was already outdated */
224 ex_to_string = "peer was fenced";
225 nps = D_OUTDATED;
226 break;
227 case 5: /* peer was down */
228 if (mdev->state.disk == D_UP_TO_DATE) {
229 /* we will(have) create(d) a new UUID anyways... */
230 ex_to_string = "peer is unreachable, assumed to be dead";
231 nps = D_OUTDATED;
232 } else {
233 ex_to_string = "peer unreachable, doing nothing since disk != UpToDate";
234 nps = mdev->state.pdsk;
235 }
236 break;
237 case 6: /* Peer is primary, voluntarily outdate myself.
238 * This is useful when an unconnected R_SECONDARY is asked to
239 * become R_PRIMARY, but finds the other peer being active. */
240 ex_to_string = "peer is active";
241 dev_warn(DEV, "Peer is primary, outdating myself.\n");
242 nps = D_UNKNOWN;
243 _drbd_request_state(mdev, NS(disk, D_OUTDATED), CS_WAIT_COMPLETE);
244 break;
245 case 7:
246 if (fp != FP_STONITH)
247 dev_err(DEV, "fence-peer() = 7 && fencing != Stonith !!!\n");
248 ex_to_string = "peer was stonithed";
249 nps = D_OUTDATED;
250 break;
251 default:
252 /* The script is broken ... */
253 nps = D_UNKNOWN;
254 dev_err(DEV, "fence-peer helper broken, returned %d\n", (r>>8)&0xff);
255 return nps;
256 }
257
258 dev_info(DEV, "fence-peer helper returned %d (%s)\n",
259 (r>>8) & 0xff, ex_to_string);
fb22c402
PR		 260
261out:
262 if (mdev->state.susp_fen && nps >= D_UNKNOWN) {
263 /* The handler was not successful... unfreeze here, the
264 state engine can not unfreeze... */
265 _drbd_request_state(mdev, NS(susp_fen, 0), CS_VERBOSE);
266 }
267
b411b363
PR		 268 return nps;
269}
270
87f7be4c
PR		 271static int _try_outdate_peer_async(void *data)
272{
273 struct drbd_conf *mdev = (struct drbd_conf *)data;
274 enum drbd_disk_state nps;
275
276 nps = drbd_try_outdate_peer(mdev);
277 drbd_request_state(mdev, NS(pdsk, nps));
278
279 return 0;
280}
281
282void drbd_try_outdate_peer_async(struct drbd_conf *mdev)
283{
284 struct task_struct *opa;
285
286 opa = kthread_run(_try_outdate_peer_async, mdev, "drbd%d_a_helper", mdev_to_minor(mdev));
287 if (IS_ERR(opa))
288 dev_err(DEV, "out of mem, failed to invoke fence-peer helper\n");
289}
b411b363 290
bf885f8a
AG		 291enum drbd_state_rv
292drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
b411b363
PR		 293{
294 const int max_tries = 4;
bf885f8a 295 enum drbd_state_rv rv = SS_UNKNOWN_ERROR;
b411b363
PR		 296 int try = 0;
297 int forced = 0;
298 union drbd_state mask, val;
299 enum drbd_disk_state nps;
300
301 if (new_role == R_PRIMARY)
302 request_ping(mdev); /* Detect a dead peer ASAP */
303
304 mutex_lock(&mdev->state_mutex);
305
306 mask.i = 0; mask.role = R_MASK;
307 val.i = 0; val.role = new_role;
308
309 while (try++ < max_tries) {
bf885f8a 310 rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE);
b411b363
PR		 311
312 /* in case we first succeeded to outdate,
313 * but now suddenly could establish a connection */
bf885f8a 314 if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) {
b411b363
PR		 315 val.pdsk = 0;
316 mask.pdsk = 0;
317 continue;
318 }
319
bf885f8a 320 if (rv == SS_NO_UP_TO_DATE_DISK && force &&
d10a33c6
PR		 321 (mdev->state.disk < D_UP_TO_DATE &&
322 mdev->state.disk >= D_INCONSISTENT)) {
b411b363
PR		 323 mask.disk = D_MASK;
324 val.disk = D_UP_TO_DATE;
325 forced = 1;
326 continue;
327 }
328
bf885f8a 329 if (rv == SS_NO_UP_TO_DATE_DISK &&
b411b363
PR		 330 mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) {
331 D_ASSERT(mdev->state.pdsk == D_UNKNOWN);
332 nps = drbd_try_outdate_peer(mdev);
333
334 if (nps == D_OUTDATED || nps == D_INCONSISTENT) {
335 val.disk = D_UP_TO_DATE;
336 mask.disk = D_MASK;
337 }
338
339 val.pdsk = nps;
340 mask.pdsk = D_MASK;
341
342 continue;
343 }
344
bf885f8a 345 if (rv == SS_NOTHING_TO_DO)
b411b363 346 goto fail;
bf885f8a 347 if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) {
b411b363
PR		 348 nps = drbd_try_outdate_peer(mdev);
349
350 if (force && nps > D_OUTDATED) {
351 dev_warn(DEV, "Forced into split brain situation!\n");
352 nps = D_OUTDATED;
353 }
354
355 mask.pdsk = D_MASK;
356 val.pdsk = nps;
357
358 continue;
359 }
bf885f8a 360 if (rv == SS_TWO_PRIMARIES) {
b411b363
PR		 361 /* Maybe the peer is detected as dead very soon...
362 retry at most once more in this case. */
363 __set_current_state(TASK_INTERRUPTIBLE);
364 schedule_timeout((mdev->net_conf->ping_timeo+1)*HZ/10);
365 if (try < max_tries)
366 try = max_tries - 1;
367 continue;
368 }
bf885f8a
AG		 369 if (rv < SS_SUCCESS) {
370 rv = _drbd_request_state(mdev, mask, val,
b411b363 371 CS_VERBOSE + CS_WAIT_COMPLETE);
bf885f8a 372 if (rv < SS_SUCCESS)
b411b363
PR		 373 goto fail;
374 }
375 break;
376 }
377
bf885f8a 378 if (rv < SS_SUCCESS)
b411b363
PR		 379 goto fail;
380
381 if (forced)
382 dev_warn(DEV, "Forced to consider local data as UpToDate!\n");
383
384 /* Wait until nothing is on the fly :) */
385 wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0);
386
387 if (new_role == R_SECONDARY) {
81e84650 388 set_disk_ro(mdev->vdisk, true);
b411b363
PR		 389 if (get_ldev(mdev)) {
390 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
391 put_ldev(mdev);
392 }
393 } else {
394 if (get_net_conf(mdev)) {
395 mdev->net_conf->want_lose = 0;
396 put_net_conf(mdev);
397 }
81e84650 398 set_disk_ro(mdev->vdisk, false);
b411b363
PR		 399 if (get_ldev(mdev)) {
400 if (((mdev->state.conn < C_CONNECTED ||
401 mdev->state.pdsk <= D_FAILED)
402 && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced)
403 drbd_uuid_new_current(mdev);
404
405 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
406 put_ldev(mdev);
407 }
408 }
409
410 if ((new_role == R_SECONDARY) && get_ldev(mdev)) {
411 drbd_al_to_on_disk_bm(mdev);
412 put_ldev(mdev);
413 }
414
415 if (mdev->state.conn >= C_WF_REPORT_PARAMS) {
416 /* if this was forced, we should consider sync */
417 if (forced)
418 drbd_send_uuids(mdev);
419 drbd_send_state(mdev);
420 }
421
422 drbd_md_sync(mdev);
423
424 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
425 fail:
426 mutex_unlock(&mdev->state_mutex);
bf885f8a 427 return rv;
b411b363
PR		 428}
429
ef50a3e3
LE		 430static struct drbd_conf *ensure_mdev(int minor, int create)
431{
432 struct drbd_conf *mdev;
433
434 if (minor >= minor_count)
435 return NULL;
436
437 mdev = minor_to_mdev(minor);
438
439 if (!mdev && create) {
440 struct gendisk *disk = NULL;
441 mdev = drbd_new_device(minor);
442
443 spin_lock_irq(&drbd_pp_lock);
444 if (minor_table[minor] == NULL) {
445 minor_table[minor] = mdev;
446 disk = mdev->vdisk;
447 mdev = NULL;
448 } /* else: we lost the race */
449 spin_unlock_irq(&drbd_pp_lock);
450
451 if (disk) /* we won the race above */
452 /* in case we ever add a drbd_delete_device(),
453 * don't forget the del_gendisk! */
454 add_disk(disk);
455 else /* we lost the race above */
456 drbd_free_mdev(mdev);
457
458 mdev = minor_to_mdev(minor);
459 }
460
461 return mdev;
462}
b411b363
PR		 463
464static int drbd_nl_primary(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
465 struct drbd_nl_cfg_reply *reply)
466{
467 struct primary primary_args;
468
469 memset(&primary_args, 0, sizeof(struct primary));
470 if (!primary_from_tags(mdev, nlp->tag_list, &primary_args)) {
471 reply->ret_code = ERR_MANDATORY_TAG;
472 return 0;
473 }
474
475 reply->ret_code =
1f552430 476 drbd_set_role(mdev, R_PRIMARY, primary_args.primary_force);
b411b363
PR		 477
478 return 0;
479}
480
481static int drbd_nl_secondary(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
482 struct drbd_nl_cfg_reply *reply)
483{
484 reply->ret_code = drbd_set_role(mdev, R_SECONDARY, 0);
485
486 return 0;
487}
488
489/* initializes the md.*_offset members, so we are able to find
490 * the on disk meta data */
491static void drbd_md_set_sector_offsets(struct drbd_conf *mdev,
492 struct drbd_backing_dev *bdev)
493{
494 sector_t md_size_sect = 0;
495 switch (bdev->dc.meta_dev_idx) {
496 default:
497 /* v07 style fixed size indexed meta data */
498 bdev->md.md_size_sect = MD_RESERVED_SECT;
499 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
500 bdev->md.al_offset = MD_AL_OFFSET;
501 bdev->md.bm_offset = MD_BM_OFFSET;
502 break;
503 case DRBD_MD_INDEX_FLEX_EXT:
504 /* just occupy the full device; unit: sectors */
505 bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev);
506 bdev->md.md_offset = 0;
507 bdev->md.al_offset = MD_AL_OFFSET;
508 bdev->md.bm_offset = MD_BM_OFFSET;
509 break;
510 case DRBD_MD_INDEX_INTERNAL:
511 case DRBD_MD_INDEX_FLEX_INT:
512 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
513 /* al size is still fixed */
514 bdev->md.al_offset = -MD_AL_MAX_SIZE;
515 /* we need (slightly less than) ~ this much bitmap sectors: */
516 md_size_sect = drbd_get_capacity(bdev->backing_bdev);
517 md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT);
518 md_size_sect = BM_SECT_TO_EXT(md_size_sect);
519 md_size_sect = ALIGN(md_size_sect, 8);
520
521 /* plus the "drbd meta data super block",
522 * and the activity log; */
523 md_size_sect += MD_BM_OFFSET;
524
525 bdev->md.md_size_sect = md_size_sect;
526 /* bitmap offset is adjusted by 'super' block size */
527 bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET;
528 break;
529 }
530}
531
532char *ppsize(char *buf, unsigned long long size)
533{
534 /* Needs 9 bytes at max. */
535 static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
536 int base = 0;
537 while (size >= 10000) {
538 /* shift + round */
539 size = (size >> 10) + !!(size & (1<<9));
540 base++;
541 }
542 sprintf(buf, "%lu %cB", (long)size, units[base]);
543
544 return buf;
545}
546
547/* there is still a theoretical deadlock when called from receiver
548 * on an D_INCONSISTENT R_PRIMARY:
549 * remote READ does inc_ap_bio, receiver would need to receive answer
550 * packet from remote to dec_ap_bio again.
551 * receiver receive_sizes(), comes here,
552 * waits for ap_bio_cnt == 0. -> deadlock.
553 * but this cannot happen, actually, because:
554 * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable
555 * (not connected, or bad/no disk on peer):
556 * see drbd_fail_request_early, ap_bio_cnt is zero.
557 * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
558 * peer may not initiate a resize.
559 */
560void drbd_suspend_io(struct drbd_conf *mdev)
561{
562 set_bit(SUSPEND_IO, &mdev->flags);
fb22c402 563 if (is_susp(mdev->state))
265be2d0 564 return;
b411b363
PR		 565 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt));
566}
567
568void drbd_resume_io(struct drbd_conf *mdev)
569{
570 clear_bit(SUSPEND_IO, &mdev->flags);
571 wake_up(&mdev->misc_wait);
572}
573
574/**
575 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
576 * @mdev: DRBD device.
577 *
578 * Returns 0 on success, negative return values indicate errors.
579 * You should call drbd_md_sync() after calling this function.
580 */
d845030f 581enum determine_dev_size drbd_determin_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local)
b411b363
PR		 582{
583 sector_t prev_first_sect, prev_size; /* previous meta location */
584 sector_t la_size;
585 sector_t size;
586 char ppb[10];
587
588 int md_moved, la_size_changed;
589 enum determine_dev_size rv = unchanged;
590
591 /* race:
592 * application request passes inc_ap_bio,
593 * but then cannot get an AL-reference.
594 * this function later may wait on ap_bio_cnt == 0. -> deadlock.
595 *
596 * to avoid that:
597 * Suspend IO right here.
598 * still lock the act_log to not trigger ASSERTs there.
599 */
600 drbd_suspend_io(mdev);
601
602 /* no wait necessary anymore, actually we could assert that */
603 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
604
605 prev_first_sect = drbd_md_first_sector(mdev->ldev);
606 prev_size = mdev->ldev->md.md_size_sect;
607 la_size = mdev->ldev->md.la_size_sect;
608
609 /* TODO: should only be some assert here, not (re)init... */
610 drbd_md_set_sector_offsets(mdev, mdev->ldev);
611
d845030f 612 size = drbd_new_dev_size(mdev, mdev->ldev, flags & DDSF_FORCED);
b411b363
PR		 613
614 if (drbd_get_capacity(mdev->this_bdev) != size ||
615 drbd_bm_capacity(mdev) != size) {
616 int err;
02d9a94b 617 err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC));
b411b363
PR		 618 if (unlikely(err)) {
619 /* currently there is only one error: ENOMEM! */
620 size = drbd_bm_capacity(mdev)>>1;
621 if (size == 0) {
622 dev_err(DEV, "OUT OF MEMORY! "
623 "Could not allocate bitmap!\n");
624 } else {
625 dev_err(DEV, "BM resizing failed. "
626 "Leaving size unchanged at size = %lu KB\n",
627 (unsigned long)size);
628 }
629 rv = dev_size_error;
630 }
631 /* racy, see comments above. */
632 drbd_set_my_capacity(mdev, size);
633 mdev->ldev->md.la_size_sect = size;
634 dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1),
635 (unsigned long long)size>>1);
636 }
637 if (rv == dev_size_error)
638 goto out;
639
640 la_size_changed = (la_size != mdev->ldev->md.la_size_sect);
641
642 md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev)
643 || prev_size != mdev->ldev->md.md_size_sect;
644
645 if (la_size_changed || md_moved) {
24dccabb
AG		 646 int err;
647
b411b363
PR		 648 drbd_al_shrink(mdev); /* All extents inactive. */
649 dev_info(DEV, "Writing the whole bitmap, %s\n",
650 la_size_changed && md_moved ? "size changed and md moved" :
651 la_size_changed ? "size changed" : "md moved");
24dccabb
AG		 652 err = drbd_bitmap_io(mdev, &drbd_bm_write, "size changed"); /* does drbd_resume_io() ! */
653 if (err) {
654 rv = dev_size_error;
655 goto out;
656 }
b411b363
PR		 657 drbd_md_mark_dirty(mdev);
658 }
659
660 if (size > la_size)
661 rv = grew;
662 if (size < la_size)
663 rv = shrunk;
664out:
665 lc_unlock(mdev->act_log);
666 wake_up(&mdev->al_wait);
667 drbd_resume_io(mdev);
668
669 return rv;
670}
671
672sector_t
a393db6f 673drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, int assume_peer_has_space)
b411b363
PR		 674{
675 sector_t p_size = mdev->p_size; /* partner's disk size. */
676 sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */
677 sector_t m_size; /* my size */
678 sector_t u_size = bdev->dc.disk_size; /* size requested by user. */
679 sector_t size = 0;
680
681 m_size = drbd_get_max_capacity(bdev);
682
a393db6f
PR		 683 if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) {
684 dev_warn(DEV, "Resize while not connected was forced by the user!\n");
685 p_size = m_size;
686 }
687
b411b363
PR		 688 if (p_size && m_size) {
689 size = min_t(sector_t, p_size, m_size);
690 } else {
691 if (la_size) {
692 size = la_size;
693 if (m_size && m_size < size)
694 size = m_size;
695 if (p_size && p_size < size)
696 size = p_size;
697 } else {
698 if (m_size)
699 size = m_size;
700 if (p_size)
701 size = p_size;
702 }
703 }
704
705 if (size == 0)
706 dev_err(DEV, "Both nodes diskless!\n");
707
708 if (u_size) {
709 if (u_size > size)
710 dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n",
711 (unsigned long)u_size>>1, (unsigned long)size>>1);
712 else
713 size = u_size;
714 }
715
716 return size;
717}
718
719/**
720 * drbd_check_al_size() - Ensures that the AL is of the right size
721 * @mdev: DRBD device.
722 *
723 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
724 * failed, and 0 on success. You should call drbd_md_sync() after you called
725 * this function.
726 */
727static int drbd_check_al_size(struct drbd_conf *mdev)
728{
729 struct lru_cache *n, *t;
730 struct lc_element *e;
731 unsigned int in_use;
732 int i;
733
734 ERR_IF(mdev->sync_conf.al_extents < 7)
735 mdev->sync_conf.al_extents = 127;
736
737 if (mdev->act_log &&
738 mdev->act_log->nr_elements == mdev->sync_conf.al_extents)
739 return 0;
740
741 in_use = 0;
742 t = mdev->act_log;
743 n = lc_create("act_log", drbd_al_ext_cache,
744 mdev->sync_conf.al_extents, sizeof(struct lc_element), 0);
745
746 if (n == NULL) {
747 dev_err(DEV, "Cannot allocate act_log lru!\n");
748 return -ENOMEM;
749 }
750 spin_lock_irq(&mdev->al_lock);
751 if (t) {
752 for (i = 0; i < t->nr_elements; i++) {
753 e = lc_element_by_index(t, i);
754 if (e->refcnt)
755 dev_err(DEV, "refcnt(%d)==%d\n",
756 e->lc_number, e->refcnt);
757 in_use += e->refcnt;
758 }
759 }
760 if (!in_use)
761 mdev->act_log = n;
762 spin_unlock_irq(&mdev->al_lock);
763 if (in_use) {
764 dev_err(DEV, "Activity log still in use!\n");
765 lc_destroy(n);
766 return -EBUSY;
767 } else {
768 if (t)
769 lc_destroy(t);
770 }
771 drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */
772 return 0;
773}
774
1816a2b4 775void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size) __must_hold(local)
b411b363
PR		 776{
777 struct request_queue * const q = mdev->rq_queue;
778 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
779 int max_segments = mdev->ldev->dc.max_bio_bvecs;
1816a2b4 780 int max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9);
b411b363 781
b411b363 782 blk_queue_logical_block_size(q, 512);
1816a2b4
LE		 783 blk_queue_max_hw_sectors(q, max_hw_sectors);
784 /* This is the workaround for "bio would need to, but cannot, be split" */
785 blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS);
786 blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1);
787 blk_queue_stack_limits(q, b);
b411b363 788
1816a2b4 789 dev_info(DEV, "max BIO size = %u\n", queue_max_hw_sectors(q) << 9);
b411b363
PR		 790
791 if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) {
792 dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
793 q->backing_dev_info.ra_pages,
794 b->backing_dev_info.ra_pages);
795 q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages;
796 }
797}
798
799/* serialize deconfig (worker exiting, doing cleanup)
800 * and reconfig (drbdsetup disk, drbdsetup net)
801 *
c518d04f
LE		 802 * Wait for a potentially exiting worker, then restart it,
803 * or start a new one. Flush any pending work, there may still be an
804 * after_state_change queued.
b411b363
PR		 805 */
806static void drbd_reconfig_start(struct drbd_conf *mdev)
807{
6c6c7951 808 wait_event(mdev->state_wait, !test_and_set_bit(CONFIG_PENDING, &mdev->flags));
b411b363
PR		 809 wait_event(mdev->state_wait, !test_bit(DEVICE_DYING, &mdev->flags));
810 drbd_thread_start(&mdev->worker);
c518d04f 811 drbd_flush_workqueue(mdev);
b411b363
PR		 812}
813
814/* if still unconfigured, stops worker again.
815 * if configured now, clears CONFIG_PENDING.
816 * wakes potential waiters */
817static void drbd_reconfig_done(struct drbd_conf *mdev)
818{
819 spin_lock_irq(&mdev->req_lock);
820 if (mdev->state.disk == D_DISKLESS &&
821 mdev->state.conn == C_STANDALONE &&
822 mdev->state.role == R_SECONDARY) {
823 set_bit(DEVICE_DYING, &mdev->flags);
824 drbd_thread_stop_nowait(&mdev->worker);
825 } else
826 clear_bit(CONFIG_PENDING, &mdev->flags);
827 spin_unlock_irq(&mdev->req_lock);
828 wake_up(&mdev->state_wait);
829}
830
0778286a
PR		 831/* Make sure IO is suspended before calling this function(). */
832static void drbd_suspend_al(struct drbd_conf *mdev)
833{
834 int s = 0;
835
836 if (lc_try_lock(mdev->act_log)) {
837 drbd_al_shrink(mdev);
838 lc_unlock(mdev->act_log);
839 } else {
840 dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n");
841 return;
842 }
843
844 spin_lock_irq(&mdev->req_lock);
845 if (mdev->state.conn < C_CONNECTED)
846 s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags);
847
848 spin_unlock_irq(&mdev->req_lock);
849
850 if (s)
851 dev_info(DEV, "Suspended AL updates\n");
852}
853
b411b363
PR		 854/* does always return 0;
855 * interesting return code is in reply->ret_code */
856static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
857 struct drbd_nl_cfg_reply *reply)
858{
116676ca 859 enum drbd_ret_code retcode;
b411b363
PR		 860 enum determine_dev_size dd;
861 sector_t max_possible_sectors;
862 sector_t min_md_device_sectors;
863 struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */
e525fd89 864 struct block_device *bdev;
b411b363
PR		 865 struct lru_cache *resync_lru = NULL;
866 union drbd_state ns, os;
1816a2b4 867 unsigned int max_bio_size;
f2024e7c 868 enum drbd_state_rv rv;
b411b363
PR		 869 int cp_discovered = 0;
870 int logical_block_size;
871
872 drbd_reconfig_start(mdev);
873
874 /* if you want to reconfigure, please tear down first */
875 if (mdev->state.disk > D_DISKLESS) {
876 retcode = ERR_DISK_CONFIGURED;
877 goto fail;
878 }
82f59cc6
LE		 879 /* It may just now have detached because of IO error. Make sure
880 * drbd_ldev_destroy is done already, we may end up here very fast,
881 * e.g. if someone calls attach from the on-io-error handler,
882 * to realize a "hot spare" feature (not that I'd recommend that) */
883 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
b411b363
PR		 884
885 /* allocation not in the IO path, cqueue thread context */
886 nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL);
887 if (!nbc) {
888 retcode = ERR_NOMEM;
889 goto fail;
890 }
891
892 nbc->dc.disk_size = DRBD_DISK_SIZE_SECT_DEF;
893 nbc->dc.on_io_error = DRBD_ON_IO_ERROR_DEF;
894 nbc->dc.fencing = DRBD_FENCING_DEF;
895 nbc->dc.max_bio_bvecs = DRBD_MAX_BIO_BVECS_DEF;
896
897 if (!disk_conf_from_tags(mdev, nlp->tag_list, &nbc->dc)) {
898 retcode = ERR_MANDATORY_TAG;
899 goto fail;
900 }
901
902 if (nbc->dc.meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) {
903 retcode = ERR_MD_IDX_INVALID;
904 goto fail;
905 }
906
47ff2d0a
PR		 907 if (get_net_conf(mdev)) {
908 int prot = mdev->net_conf->wire_protocol;
909 put_net_conf(mdev);
910 if (nbc->dc.fencing == FP_STONITH && prot == DRBD_PROT_A) {
911 retcode = ERR_STONITH_AND_PROT_A;
912 goto fail;
913 }
914 }
915
d4d77629
TH		 916 bdev = blkdev_get_by_path(nbc->dc.backing_dev,
917 FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev);
e525fd89 918 if (IS_ERR(bdev)) {
b411b363 919 dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.backing_dev,
e525fd89 920 PTR_ERR(bdev));
b411b363
PR		 921 retcode = ERR_OPEN_DISK;
922 goto fail;
923 }
e525fd89
TH		 924 nbc->backing_bdev = bdev;
925
926 /*
927 * meta_dev_idx >= 0: external fixed size, possibly multiple
928 * drbd sharing one meta device. TODO in that case, paranoia
929 * check that [md_bdev, meta_dev_idx] is not yet used by some
930 * other drbd minor! (if you use drbd.conf + drbdadm, that
931 * should check it for you already; but if you don't, or
932 * someone fooled it, we need to double check here)
933 */
d4d77629
TH		 934 bdev = blkdev_get_by_path(nbc->dc.meta_dev,
935 FMODE_READ | FMODE_WRITE | FMODE_EXCL,
936 (nbc->dc.meta_dev_idx < 0) ?
937 (void *)mdev : (void *)drbd_m_holder);
e525fd89 938 if (IS_ERR(bdev)) {
b411b363 939 dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.meta_dev,
e525fd89 940 PTR_ERR(bdev));
b411b363
PR		 941 retcode = ERR_OPEN_MD_DISK;
942 goto fail;
943 }
e525fd89 944 nbc->md_bdev = bdev;
b411b363 945
e525fd89
TH		 946 if ((nbc->backing_bdev == nbc->md_bdev) !=
947 (nbc->dc.meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
948 nbc->dc.meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) {
949 retcode = ERR_MD_IDX_INVALID;
b411b363
PR		 950 goto fail;
951 }
952
953 resync_lru = lc_create("resync", drbd_bm_ext_cache,
954 61, sizeof(struct bm_extent),
955 offsetof(struct bm_extent, lce));
956 if (!resync_lru) {
957 retcode = ERR_NOMEM;
e525fd89 958 goto fail;
b411b363
PR		 959 }
960
961 /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */
962 drbd_md_set_sector_offsets(mdev, nbc);
963
964 if (drbd_get_max_capacity(nbc) < nbc->dc.disk_size) {
965 dev_err(DEV, "max capacity %llu smaller than disk size %llu\n",
966 (unsigned long long) drbd_get_max_capacity(nbc),
967 (unsigned long long) nbc->dc.disk_size);
968 retcode = ERR_DISK_TO_SMALL;
e525fd89 969 goto fail;
b411b363
PR		 970 }
971
972 if (nbc->dc.meta_dev_idx < 0) {
973 max_possible_sectors = DRBD_MAX_SECTORS_FLEX;
974 /* at least one MB, otherwise it does not make sense */
975 min_md_device_sectors = (2<<10);
976 } else {
977 max_possible_sectors = DRBD_MAX_SECTORS;
978 min_md_device_sectors = MD_RESERVED_SECT * (nbc->dc.meta_dev_idx + 1);
979 }
980
b411b363
PR		 981 if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) {
982 retcode = ERR_MD_DISK_TO_SMALL;
983 dev_warn(DEV, "refusing attach: md-device too small, "
984 "at least %llu sectors needed for this meta-disk type\n",
985 (unsigned long long) min_md_device_sectors);
e525fd89 986 goto fail;
b411b363
PR		 987 }
988
989 /* Make sure the new disk is big enough
990 * (we may currently be R_PRIMARY with no local disk...) */
991 if (drbd_get_max_capacity(nbc) <
992 drbd_get_capacity(mdev->this_bdev)) {
993 retcode = ERR_DISK_TO_SMALL;
e525fd89 994 goto fail;
b411b363
PR		 995 }
996
997 nbc->known_size = drbd_get_capacity(nbc->backing_bdev);
998
1352994b
LE		 999 if (nbc->known_size > max_possible_sectors) {
1000 dev_warn(DEV, "==> truncating very big lower level device "
1001 "to currently maximum possible %llu sectors <==\n",
1002 (unsigned long long) max_possible_sectors);
1003 if (nbc->dc.meta_dev_idx >= 0)
1004 dev_warn(DEV, "==>> using internal or flexible "
1005 "meta data may help <<==\n");
1006 }
1007
b411b363
PR		 1008 drbd_suspend_io(mdev);
1009 /* also wait for the last barrier ack. */
fb22c402 1010 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || is_susp(mdev->state));
b411b363
PR		 1011 /* and for any other previously queued work */
1012 drbd_flush_workqueue(mdev);
1013
f2024e7c
AG		 1014 rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE);
1015 retcode = rv; /* FIXME: Type mismatch. */
b411b363 1016 drbd_resume_io(mdev);
f2024e7c 1017 if (rv < SS_SUCCESS)
e525fd89 1018 goto fail;
b411b363
PR		 1019
1020 if (!get_ldev_if_state(mdev, D_ATTACHING))
1021 goto force_diskless;
1022
1023 drbd_md_set_sector_offsets(mdev, nbc);
1024
4aa83b7b
LE		 1025 /* allocate a second IO page if logical_block_size != 512 */
1026 logical_block_size = bdev_logical_block_size(nbc->md_bdev);
1027 if (logical_block_size == 0)
1028 logical_block_size = MD_SECTOR_SIZE;
1029
1030 if (logical_block_size != MD_SECTOR_SIZE) {
1031 if (!mdev->md_io_tmpp) {
1032 struct page *page = alloc_page(GFP_NOIO);
1033 if (!page)
1034 goto force_diskless_dec;
1035
1036 dev_warn(DEV, "Meta data's bdev logical_block_size = %d != %d\n",
1037 logical_block_size, MD_SECTOR_SIZE);
1038 dev_warn(DEV, "Workaround engaged (has performance impact).\n");
1039
1040 mdev->md_io_tmpp = page;
1041 }
1042 }
1043
b411b363
PR		 1044 if (!mdev->bitmap) {
1045 if (drbd_bm_init(mdev)) {
1046 retcode = ERR_NOMEM;
1047 goto force_diskless_dec;
1048 }
1049 }
1050
1051 retcode = drbd_md_read(mdev, nbc);
1052 if (retcode != NO_ERROR)
1053 goto force_diskless_dec;
1054
1055 if (mdev->state.conn < C_CONNECTED &&
1056 mdev->state.role == R_PRIMARY &&
1057 (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) {
1058 dev_err(DEV, "Can only attach to data with current UUID=%016llX\n",
1059 (unsigned long long)mdev->ed_uuid);
1060 retcode = ERR_DATA_NOT_CURRENT;
1061 goto force_diskless_dec;
1062 }
1063
1064 /* Since we are diskless, fix the activity log first... */
1065 if (drbd_check_al_size(mdev)) {
1066 retcode = ERR_NOMEM;
1067 goto force_diskless_dec;
1068 }
1069
1070 /* Prevent shrinking of consistent devices ! */
1071 if (drbd_md_test_flag(nbc, MDF_CONSISTENT) &&
a393db6f 1072 drbd_new_dev_size(mdev, nbc, 0) < nbc->md.la_size_sect) {
b411b363
PR		 1073 dev_warn(DEV, "refusing to truncate a consistent device\n");
1074 retcode = ERR_DISK_TO_SMALL;
1075 goto force_diskless_dec;
1076 }
1077
1078 if (!drbd_al_read_log(mdev, nbc)) {
1079 retcode = ERR_IO_MD_DISK;
1080 goto force_diskless_dec;
1081 }
1082
b411b363
PR		 1083 /* Reset the "barriers don't work" bits here, then force meta data to
1084 * be written, to ensure we determine if barriers are supported. */
1085 if (nbc->dc.no_md_flush)
a8a4e51e 1086 set_bit(MD_NO_FUA, &mdev->flags);
b411b363 1087 else
a8a4e51e 1088 clear_bit(MD_NO_FUA, &mdev->flags);
b411b363
PR		 1089
1090 /* Point of no return reached.
1091 * Devices and memory are no longer released by error cleanup below.
1092 * now mdev takes over responsibility, and the state engine should
1093 * clean it up somewhere. */
1094 D_ASSERT(mdev->ldev == NULL);
1095 mdev->ldev = nbc;
1096 mdev->resync = resync_lru;
1097 nbc = NULL;
1098 resync_lru = NULL;
1099
2451fc3b
PR		 1100 mdev->write_ordering = WO_bdev_flush;
1101 drbd_bump_write_ordering(mdev, WO_bdev_flush);
b411b363
PR		 1102
1103 if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY))
1104 set_bit(CRASHED_PRIMARY, &mdev->flags);
1105 else
1106 clear_bit(CRASHED_PRIMARY, &mdev->flags);
1107
894c6a94 1108 if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
fb22c402 1109 !(mdev->state.role == R_PRIMARY && mdev->state.susp_nod)) {
b411b363
PR		 1110 set_bit(CRASHED_PRIMARY, &mdev->flags);
1111 cp_discovered = 1;
1112 }
1113
1114 mdev->send_cnt = 0;
1115 mdev->recv_cnt = 0;
1116 mdev->read_cnt = 0;
1117 mdev->writ_cnt = 0;
1118
1816a2b4 1119 max_bio_size = DRBD_MAX_BIO_SIZE;
8979d9c9
LE		 1120 if (mdev->state.conn == C_CONNECTED) {
1121 /* We are Primary, Connected, and now attach a new local
1122 * backing store. We must not increase the user visible maximum
1123 * bio size on this device to something the peer may not be
1124 * able to handle. */
1125 if (mdev->agreed_pro_version < 94)
1816a2b4 1126 max_bio_size = queue_max_hw_sectors(mdev->rq_queue) << 9;
8979d9c9 1127 else if (mdev->agreed_pro_version == 94)
1816a2b4 1128 max_bio_size = DRBD_MAX_SIZE_H80_PACKET;
8979d9c9
LE		 1129 /* else: drbd 8.3.9 and later, stay with default */
1130 }
1131
1816a2b4 1132 drbd_setup_queue_param(mdev, max_bio_size);
b411b363
PR		 1133
1134 /* If I am currently not R_PRIMARY,
1135 * but meta data primary indicator is set,
1136 * I just now recover from a hard crash,
1137 * and have been R_PRIMARY before that crash.
1138 *
1139 * Now, if I had no connection before that crash
1140 * (have been degraded R_PRIMARY), chances are that
1141 * I won't find my peer now either.
1142 *
1143 * In that case, and _only_ in that case,
1144 * we use the degr-wfc-timeout instead of the default,
1145 * so we can automatically recover from a crash of a
1146 * degraded but active "cluster" after a certain timeout.
1147 */
1148 clear_bit(USE_DEGR_WFC_T, &mdev->flags);
1149 if (mdev->state.role != R_PRIMARY &&
1150 drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
1151 !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND))
1152 set_bit(USE_DEGR_WFC_T, &mdev->flags);
1153
a393db6f 1154 dd = drbd_determin_dev_size(mdev, 0);
b411b363
PR		 1155 if (dd == dev_size_error) {
1156 retcode = ERR_NOMEM_BITMAP;
1157 goto force_diskless_dec;
1158 } else if (dd == grew)
1159 set_bit(RESYNC_AFTER_NEG, &mdev->flags);
1160
1161 if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) {
1162 dev_info(DEV, "Assuming that all blocks are out of sync "
1163 "(aka FullSync)\n");
1164 if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, "set_n_write from attaching")) {
1165 retcode = ERR_IO_MD_DISK;
1166 goto force_diskless_dec;
1167 }
1168 } else {
1169 if (drbd_bitmap_io(mdev, &drbd_bm_read, "read from attaching") < 0) {
1170 retcode = ERR_IO_MD_DISK;
1171 goto force_diskless_dec;
1172 }
1173 }
1174
1175 if (cp_discovered) {
1176 drbd_al_apply_to_bm(mdev);
1177 drbd_al_to_on_disk_bm(mdev);
1178 }
1179
0778286a
PR		 1180 if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev))
1181 drbd_suspend_al(mdev); /* IO is still suspended here... */
1182
b411b363
PR		 1183 spin_lock_irq(&mdev->req_lock);
1184 os = mdev->state;
1185 ns.i = os.i;
1186 /* If MDF_CONSISTENT is not set go into inconsistent state,
1187 otherwise investigate MDF_WasUpToDate...
1188 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1189 otherwise into D_CONSISTENT state.
1190 */
1191 if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) {
1192 if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE))
1193 ns.disk = D_CONSISTENT;
1194 else
1195 ns.disk = D_OUTDATED;
1196 } else {
1197 ns.disk = D_INCONSISTENT;
1198 }
1199
1200 if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED))
1201 ns.pdsk = D_OUTDATED;
1202
1203 if ( ns.disk == D_CONSISTENT &&
1204 (ns.pdsk == D_OUTDATED || mdev->ldev->dc.fencing == FP_DONT_CARE))
1205 ns.disk = D_UP_TO_DATE;
1206
1207 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1208 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1209 this point, because drbd_request_state() modifies these
1210 flags. */
1211
1212 /* In case we are C_CONNECTED postpone any decision on the new disk
1213 state after the negotiation phase. */
1214 if (mdev->state.conn == C_CONNECTED) {
1215 mdev->new_state_tmp.i = ns.i;
1216 ns.i = os.i;
1217 ns.disk = D_NEGOTIATING;
dc66c74d
PR		 1218
1219 /* We expect to receive up-to-date UUIDs soon.
1220 To avoid a race in receive_state, free p_uuid while
1221 holding req_lock. I.e. atomic with the state change */
1222 kfree(mdev->p_uuid);
1223 mdev->p_uuid = NULL;
b411b363
PR		 1224 }
1225
1226 rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
1227 ns = mdev->state;
1228 spin_unlock_irq(&mdev->req_lock);
1229
1230 if (rv < SS_SUCCESS)
1231 goto force_diskless_dec;
1232
1233 if (mdev->state.role == R_PRIMARY)
1234 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
1235 else
1236 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
1237
1238 drbd_md_mark_dirty(mdev);
1239 drbd_md_sync(mdev);
1240
1241 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
1242 put_ldev(mdev);
1243 reply->ret_code = retcode;
1244 drbd_reconfig_done(mdev);
1245 return 0;
1246
1247 force_diskless_dec:
1248 put_ldev(mdev);
1249 force_diskless:
82f59cc6 1250 drbd_force_state(mdev, NS(disk, D_FAILED));
b411b363 1251 drbd_md_sync(mdev);
b411b363
PR		 1252 fail:
1253 if (nbc) {
e525fd89
TH		 1254 if (nbc->backing_bdev)
1255 blkdev_put(nbc->backing_bdev,
1256 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
1257 if (nbc->md_bdev)
1258 blkdev_put(nbc->md_bdev,
1259 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
b411b363
PR		 1260 kfree(nbc);
1261 }
1262 lc_destroy(resync_lru);
1263
1264 reply->ret_code = retcode;
1265 drbd_reconfig_done(mdev);
1266 return 0;
1267}
1268
82f59cc6
LE		 1269/* Detaching the disk is a process in multiple stages. First we need to lock
1270 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1271 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1272 * internal references as well.
1273 * Only then we have finally detached. */
b411b363
PR		 1274static int drbd_nl_detach(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1275 struct drbd_nl_cfg_reply *reply)
1276{
82f59cc6 1277 drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */
b411b363 1278 reply->ret_code = drbd_request_state(mdev, NS(disk, D_DISKLESS));
82f59cc6
LE		 1279 if (mdev->state.disk == D_DISKLESS)
1280 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
1281 drbd_resume_io(mdev);
b411b363
PR		 1282 return 0;
1283}
1284
1285static int drbd_nl_net_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1286 struct drbd_nl_cfg_reply *reply)
1287{
1288 int i, ns;
116676ca 1289 enum drbd_ret_code retcode;
b411b363
PR		 1290 struct net_conf *new_conf = NULL;
1291 struct crypto_hash *tfm = NULL;
1292 struct crypto_hash *integrity_w_tfm = NULL;
1293 struct crypto_hash *integrity_r_tfm = NULL;
1294 struct hlist_head *new_tl_hash = NULL;
1295 struct hlist_head *new_ee_hash = NULL;
1296 struct drbd_conf *odev;
1297 char hmac_name[CRYPTO_MAX_ALG_NAME];
1298 void *int_dig_out = NULL;
1299 void *int_dig_in = NULL;
1300 void *int_dig_vv = NULL;
1301 struct sockaddr *new_my_addr, *new_peer_addr, *taken_addr;
1302
1303 drbd_reconfig_start(mdev);
1304
1305 if (mdev->state.conn > C_STANDALONE) {
1306 retcode = ERR_NET_CONFIGURED;
1307 goto fail;
1308 }
1309
1310 /* allocation not in the IO path, cqueue thread context */
2db4e42e 1311 new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
b411b363
PR		 1312 if (!new_conf) {
1313 retcode = ERR_NOMEM;
1314 goto fail;
1315 }
1316
b411b363
PR		 1317 new_conf->timeout = DRBD_TIMEOUT_DEF;
1318 new_conf->try_connect_int = DRBD_CONNECT_INT_DEF;
1319 new_conf->ping_int = DRBD_PING_INT_DEF;
1320 new_conf->max_epoch_size = DRBD_MAX_EPOCH_SIZE_DEF;
1321 new_conf->max_buffers = DRBD_MAX_BUFFERS_DEF;
1322 new_conf->unplug_watermark = DRBD_UNPLUG_WATERMARK_DEF;
1323 new_conf->sndbuf_size = DRBD_SNDBUF_SIZE_DEF;
1324 new_conf->rcvbuf_size = DRBD_RCVBUF_SIZE_DEF;
1325 new_conf->ko_count = DRBD_KO_COUNT_DEF;
1326 new_conf->after_sb_0p = DRBD_AFTER_SB_0P_DEF;
1327 new_conf->after_sb_1p = DRBD_AFTER_SB_1P_DEF;
1328 new_conf->after_sb_2p = DRBD_AFTER_SB_2P_DEF;
1329 new_conf->want_lose = 0;
1330 new_conf->two_primaries = 0;
1331 new_conf->wire_protocol = DRBD_PROT_C;
1332 new_conf->ping_timeo = DRBD_PING_TIMEO_DEF;
1333 new_conf->rr_conflict = DRBD_RR_CONFLICT_DEF;
422028b1
PR		 1334 new_conf->on_congestion = DRBD_ON_CONGESTION_DEF;
1335 new_conf->cong_extents = DRBD_CONG_EXTENTS_DEF;
b411b363
PR		 1336
1337 if (!net_conf_from_tags(mdev, nlp->tag_list, new_conf)) {
1338 retcode = ERR_MANDATORY_TAG;
1339 goto fail;
1340 }
1341
1342 if (new_conf->two_primaries
1343 && (new_conf->wire_protocol != DRBD_PROT_C)) {
1344 retcode = ERR_NOT_PROTO_C;
1345 goto fail;
47ff2d0a
PR		 1346 }
1347
1348 if (get_ldev(mdev)) {
1349 enum drbd_fencing_p fp = mdev->ldev->dc.fencing;
1350 put_ldev(mdev);
1351 if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) {
1352 retcode = ERR_STONITH_AND_PROT_A;
1353 goto fail;
1354 }
1355 }
b411b363 1356
422028b1
PR		 1357 if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A) {
1358 retcode = ERR_CONG_NOT_PROTO_A;
1359 goto fail;
1360 }
1361
b411b363
PR		 1362 if (mdev->state.role == R_PRIMARY && new_conf->want_lose) {
1363 retcode = ERR_DISCARD;
1364 goto fail;
1365 }
1366
1367 retcode = NO_ERROR;
1368
1369 new_my_addr = (struct sockaddr *)&new_conf->my_addr;
1370 new_peer_addr = (struct sockaddr *)&new_conf->peer_addr;
1371 for (i = 0; i < minor_count; i++) {
1372 odev = minor_to_mdev(i);
1373 if (!odev || odev == mdev)
1374 continue;
1375 if (get_net_conf(odev)) {
1376 taken_addr = (struct sockaddr *)&odev->net_conf->my_addr;
1377 if (new_conf->my_addr_len == odev->net_conf->my_addr_len &&
1378 !memcmp(new_my_addr, taken_addr, new_conf->my_addr_len))
1379 retcode = ERR_LOCAL_ADDR;
1380
1381 taken_addr = (struct sockaddr *)&odev->net_conf->peer_addr;
1382 if (new_conf->peer_addr_len == odev->net_conf->peer_addr_len &&
1383 !memcmp(new_peer_addr, taken_addr, new_conf->peer_addr_len))
1384 retcode = ERR_PEER_ADDR;
1385
1386 put_net_conf(odev);
1387 if (retcode != NO_ERROR)
1388 goto fail;
1389 }
1390 }
1391
1392 if (new_conf->cram_hmac_alg[0] != 0) {
1393 snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)",
1394 new_conf->cram_hmac_alg);
1395 tfm = crypto_alloc_hash(hmac_name, 0, CRYPTO_ALG_ASYNC);
1396 if (IS_ERR(tfm)) {
1397 tfm = NULL;
1398 retcode = ERR_AUTH_ALG;
1399 goto fail;
1400 }
1401
0798219f 1402 if (!drbd_crypto_is_hash(crypto_hash_tfm(tfm))) {
b411b363
PR		 1403 retcode = ERR_AUTH_ALG_ND;
1404 goto fail;
1405 }
1406 }
1407
1408 if (new_conf->integrity_alg[0]) {
1409 integrity_w_tfm = crypto_alloc_hash(new_conf->integrity_alg, 0, CRYPTO_ALG_ASYNC);
1410 if (IS_ERR(integrity_w_tfm)) {
1411 integrity_w_tfm = NULL;
1412 retcode=ERR_INTEGRITY_ALG;
1413 goto fail;
1414 }
1415
1416 if (!drbd_crypto_is_hash(crypto_hash_tfm(integrity_w_tfm))) {
1417 retcode=ERR_INTEGRITY_ALG_ND;
1418 goto fail;
1419 }
1420
1421 integrity_r_tfm = crypto_alloc_hash(new_conf->integrity_alg, 0, CRYPTO_ALG_ASYNC);
1422 if (IS_ERR(integrity_r_tfm)) {
1423 integrity_r_tfm = NULL;
1424 retcode=ERR_INTEGRITY_ALG;
1425 goto fail;
1426 }
1427 }
1428
1429 ns = new_conf->max_epoch_size/8;
1430 if (mdev->tl_hash_s != ns) {
1431 new_tl_hash = kzalloc(ns*sizeof(void *), GFP_KERNEL);
1432 if (!new_tl_hash) {
1433 retcode = ERR_NOMEM;
1434 goto fail;
1435 }
1436 }
1437
1438 ns = new_conf->max_buffers/8;
1439 if (new_conf->two_primaries && (mdev->ee_hash_s != ns)) {
1440 new_ee_hash = kzalloc(ns*sizeof(void *), GFP_KERNEL);
1441 if (!new_ee_hash) {
1442 retcode = ERR_NOMEM;
1443 goto fail;
1444 }
1445 }
1446
1447 ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0;
1448
1449 if (integrity_w_tfm) {
1450 i = crypto_hash_digestsize(integrity_w_tfm);
1451 int_dig_out = kmalloc(i, GFP_KERNEL);
1452 if (!int_dig_out) {
1453 retcode = ERR_NOMEM;
1454 goto fail;
1455 }
1456 int_dig_in = kmalloc(i, GFP_KERNEL);
1457 if (!int_dig_in) {
1458 retcode = ERR_NOMEM;
1459 goto fail;
1460 }
1461 int_dig_vv = kmalloc(i, GFP_KERNEL);
1462 if (!int_dig_vv) {
1463 retcode = ERR_NOMEM;
1464 goto fail;
1465 }
1466 }
1467
1468 if (!mdev->bitmap) {
1469 if(drbd_bm_init(mdev)) {
1470 retcode = ERR_NOMEM;
1471 goto fail;
1472 }
1473 }
1474
f70b3511 1475 drbd_flush_workqueue(mdev);
b411b363
PR		 1476 spin_lock_irq(&mdev->req_lock);
1477 if (mdev->net_conf != NULL) {
1478 retcode = ERR_NET_CONFIGURED;
1479 spin_unlock_irq(&mdev->req_lock);
1480 goto fail;
1481 }
1482 mdev->net_conf = new_conf;
1483
1484 mdev->send_cnt = 0;
1485 mdev->recv_cnt = 0;
1486
1487 if (new_tl_hash) {
1488 kfree(mdev->tl_hash);
1489 mdev->tl_hash_s = mdev->net_conf->max_epoch_size/8;
1490 mdev->tl_hash = new_tl_hash;
1491 }
1492
1493 if (new_ee_hash) {
1494 kfree(mdev->ee_hash);
1495 mdev->ee_hash_s = mdev->net_conf->max_buffers/8;
1496 mdev->ee_hash = new_ee_hash;
1497 }
1498
1499 crypto_free_hash(mdev->cram_hmac_tfm);
1500 mdev->cram_hmac_tfm = tfm;
1501
1502 crypto_free_hash(mdev->integrity_w_tfm);
1503 mdev->integrity_w_tfm = integrity_w_tfm;
1504
1505 crypto_free_hash(mdev->integrity_r_tfm);
1506 mdev->integrity_r_tfm = integrity_r_tfm;
1507
1508 kfree(mdev->int_dig_out);
1509 kfree(mdev->int_dig_in);
1510 kfree(mdev->int_dig_vv);
1511 mdev->int_dig_out=int_dig_out;
1512 mdev->int_dig_in=int_dig_in;
1513 mdev->int_dig_vv=int_dig_vv;
f70b3511 1514 retcode = _drbd_set_state(_NS(mdev, conn, C_UNCONNECTED), CS_VERBOSE, NULL);
b411b363
PR		 1515 spin_unlock_irq(&mdev->req_lock);
1516
b411b363
PR		 1517 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
1518 reply->ret_code = retcode;
1519 drbd_reconfig_done(mdev);
1520 return 0;
1521
1522fail:
1523 kfree(int_dig_out);
1524 kfree(int_dig_in);
1525 kfree(int_dig_vv);
1526 crypto_free_hash(tfm);
1527 crypto_free_hash(integrity_w_tfm);
1528 crypto_free_hash(integrity_r_tfm);
1529 kfree(new_tl_hash);
1530 kfree(new_ee_hash);
1531 kfree(new_conf);
1532
1533 reply->ret_code = retcode;
1534 drbd_reconfig_done(mdev);
1535 return 0;
1536}
1537
1538static int drbd_nl_disconnect(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1539 struct drbd_nl_cfg_reply *reply)
1540{
1541 int retcode;
2561b9c1
PR		 1542 struct disconnect dc;
1543
1544 memset(&dc, 0, sizeof(struct disconnect));
1545 if (!disconnect_from_tags(mdev, nlp->tag_list, &dc)) {
1546 retcode = ERR_MANDATORY_TAG;
1547 goto fail;
1548 }
1549
1550 if (dc.force) {
1551 spin_lock_irq(&mdev->req_lock);
1552 if (mdev->state.conn >= C_WF_CONNECTION)
1553 _drbd_set_state(_NS(mdev, conn, C_DISCONNECTING), CS_HARD, NULL);
1554 spin_unlock_irq(&mdev->req_lock);
1555 goto done;
1556 }
b411b363
PR		 1557
1558 retcode = _drbd_request_state(mdev, NS(conn, C_DISCONNECTING), CS_ORDERED);
1559
1560 if (retcode == SS_NOTHING_TO_DO)
1561 goto done;
1562 else if (retcode == SS_ALREADY_STANDALONE)
1563 goto done;
1564 else if (retcode == SS_PRIMARY_NOP) {
1565 /* Our statche checking code wants to see the peer outdated. */
1566 retcode = drbd_request_state(mdev, NS2(conn, C_DISCONNECTING,
1567 pdsk, D_OUTDATED));
1568 } else if (retcode == SS_CW_FAILED_BY_PEER) {
1569 /* The peer probably wants to see us outdated. */
1570 retcode = _drbd_request_state(mdev, NS2(conn, C_DISCONNECTING,
1571 disk, D_OUTDATED),
1572 CS_ORDERED);
1573 if (retcode == SS_IS_DISKLESS || retcode == SS_LOWER_THAN_OUTDATED) {
1574 drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
1575 retcode = SS_SUCCESS;
1576 }
1577 }
1578
1579 if (retcode < SS_SUCCESS)
1580 goto fail;
1581
1582 if (wait_event_interruptible(mdev->state_wait,
1583 mdev->state.conn != C_DISCONNECTING)) {
1584 /* Do not test for mdev->state.conn == C_STANDALONE, since
1585 someone else might connect us in the mean time! */
1586 retcode = ERR_INTR;
1587 goto fail;
1588 }
1589
1590 done:
1591 retcode = NO_ERROR;
1592 fail:
1593 drbd_md_sync(mdev);
1594 reply->ret_code = retcode;
1595 return 0;
1596}
1597
1598void resync_after_online_grow(struct drbd_conf *mdev)
1599{
1600 int iass; /* I am sync source */
1601
1602 dev_info(DEV, "Resync of new storage after online grow\n");
1603 if (mdev->state.role != mdev->state.peer)
1604 iass = (mdev->state.role == R_PRIMARY);
1605 else
1606 iass = test_bit(DISCARD_CONCURRENT, &mdev->flags);
1607
1608 if (iass)
1609 drbd_start_resync(mdev, C_SYNC_SOURCE);
1610 else
1611 _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE);
1612}
1613
1614static int drbd_nl_resize(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1615 struct drbd_nl_cfg_reply *reply)
1616{
1617 struct resize rs;
1618 int retcode = NO_ERROR;
b411b363 1619 enum determine_dev_size dd;
6495d2c6 1620 enum dds_flags ddsf;
b411b363
PR		 1621
1622 memset(&rs, 0, sizeof(struct resize));
1623 if (!resize_from_tags(mdev, nlp->tag_list, &rs)) {
1624 retcode = ERR_MANDATORY_TAG;
1625 goto fail;
1626 }
1627
1628 if (mdev->state.conn > C_CONNECTED) {
1629 retcode = ERR_RESIZE_RESYNC;
1630 goto fail;
1631 }
1632
1633 if (mdev->state.role == R_SECONDARY &&
1634 mdev->state.peer == R_SECONDARY) {
1635 retcode = ERR_NO_PRIMARY;
1636 goto fail;
1637 }
1638
1639 if (!get_ldev(mdev)) {
1640 retcode = ERR_NO_DISK;
1641 goto fail;
1642 }
1643
6495d2c6
PR		 1644 if (rs.no_resync && mdev->agreed_pro_version < 93) {
1645 retcode = ERR_NEED_APV_93;
1646 goto fail;
1647 }
1648
087c2492 1649 if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev))
b411b363 1650 mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
b411b363
PR		 1651
1652 mdev->ldev->dc.disk_size = (sector_t)rs.resize_size;
6495d2c6
PR		 1653 ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0);
1654 dd = drbd_determin_dev_size(mdev, ddsf);
b411b363
PR		 1655 drbd_md_sync(mdev);
1656 put_ldev(mdev);
1657 if (dd == dev_size_error) {
1658 retcode = ERR_NOMEM_BITMAP;
1659 goto fail;
1660 }
1661
087c2492 1662 if (mdev->state.conn == C_CONNECTED) {
b411b363
PR		 1663 if (dd == grew)
1664 set_bit(RESIZE_PENDING, &mdev->flags);
1665
1666 drbd_send_uuids(mdev);
6495d2c6 1667 drbd_send_sizes(mdev, 1, ddsf);
b411b363
PR		 1668 }
1669
1670 fail:
1671 reply->ret_code = retcode;
1672 return 0;
1673}
1674
1675static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1676 struct drbd_nl_cfg_reply *reply)
1677{
1678 int retcode = NO_ERROR;
1679 int err;
1680 int ovr; /* online verify running */
1681 int rsr; /* re-sync running */
1682 struct crypto_hash *verify_tfm = NULL;
1683 struct crypto_hash *csums_tfm = NULL;
1684 struct syncer_conf sc;
1685 cpumask_var_t new_cpu_mask;
778f271d
PR		 1686 int *rs_plan_s = NULL;
1687 int fifo_size;
b411b363
PR		 1688
1689 if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) {
1690 retcode = ERR_NOMEM;
1691 goto fail;
1692 }
1693
1694 if (nlp->flags & DRBD_NL_SET_DEFAULTS) {
1695 memset(&sc, 0, sizeof(struct syncer_conf));
1696 sc.rate = DRBD_RATE_DEF;
1697 sc.after = DRBD_AFTER_DEF;
1698 sc.al_extents = DRBD_AL_EXTENTS_DEF;
265be2d0 1699 sc.on_no_data = DRBD_ON_NO_DATA_DEF;
9a31d716
PR		 1700 sc.c_plan_ahead = DRBD_C_PLAN_AHEAD_DEF;
1701 sc.c_delay_target = DRBD_C_DELAY_TARGET_DEF;
1702 sc.c_fill_target = DRBD_C_FILL_TARGET_DEF;
0f0601f4
LE		 1703 sc.c_max_rate = DRBD_C_MAX_RATE_DEF;
1704 sc.c_min_rate = DRBD_C_MIN_RATE_DEF;
b411b363
PR		 1705 } else
1706 memcpy(&sc, &mdev->sync_conf, sizeof(struct syncer_conf));
1707
1708 if (!syncer_conf_from_tags(mdev, nlp->tag_list, &sc)) {
1709 retcode = ERR_MANDATORY_TAG;
1710 goto fail;
1711 }
1712
1713 /* re-sync running */
1714 rsr = ( mdev->state.conn == C_SYNC_SOURCE ||
1715 mdev->state.conn == C_SYNC_TARGET ||
1716 mdev->state.conn == C_PAUSED_SYNC_S ||
1717 mdev->state.conn == C_PAUSED_SYNC_T );
1718
1719 if (rsr && strcmp(sc.csums_alg, mdev->sync_conf.csums_alg)) {
1720 retcode = ERR_CSUMS_RESYNC_RUNNING;
1721 goto fail;
1722 }
1723
1724 if (!rsr && sc.csums_alg[0]) {
1725 csums_tfm = crypto_alloc_hash(sc.csums_alg, 0, CRYPTO_ALG_ASYNC);
1726 if (IS_ERR(csums_tfm)) {
1727 csums_tfm = NULL;
1728 retcode = ERR_CSUMS_ALG;
1729 goto fail;
1730 }
1731
1732 if (!drbd_crypto_is_hash(crypto_hash_tfm(csums_tfm))) {
1733 retcode = ERR_CSUMS_ALG_ND;
1734 goto fail;
1735 }
1736 }
1737
1738 /* online verify running */
1739 ovr = (mdev->state.conn == C_VERIFY_S || mdev->state.conn == C_VERIFY_T);
1740
1741 if (ovr) {
1742 if (strcmp(sc.verify_alg, mdev->sync_conf.verify_alg)) {
1743 retcode = ERR_VERIFY_RUNNING;
1744 goto fail;
1745 }
1746 }
1747
1748 if (!ovr && sc.verify_alg[0]) {
1749 verify_tfm = crypto_alloc_hash(sc.verify_alg, 0, CRYPTO_ALG_ASYNC);
1750 if (IS_ERR(verify_tfm)) {
1751 verify_tfm = NULL;
1752 retcode = ERR_VERIFY_ALG;
1753 goto fail;
1754 }
1755
1756 if (!drbd_crypto_is_hash(crypto_hash_tfm(verify_tfm))) {
1757 retcode = ERR_VERIFY_ALG_ND;
1758 goto fail;
1759 }
1760 }
1761
1762 /* silently ignore cpu mask on UP kernel */
1763 if (nr_cpu_ids > 1 && sc.cpu_mask[0] != 0) {
1764 err = __bitmap_parse(sc.cpu_mask, 32, 0,
1765 cpumask_bits(new_cpu_mask), nr_cpu_ids);
1766 if (err) {
1767 dev_warn(DEV, "__bitmap_parse() failed with %d\n", err);
1768 retcode = ERR_CPU_MASK_PARSE;
1769 goto fail;
1770 }
1771 }
1772
1773 ERR_IF (sc.rate < 1) sc.rate = 1;
1774 ERR_IF (sc.al_extents < 7) sc.al_extents = 127; /* arbitrary minimum */
1775#define AL_MAX ((MD_AL_MAX_SIZE-1) * AL_EXTENTS_PT)
1776 if (sc.al_extents > AL_MAX) {
1777 dev_err(DEV, "sc.al_extents > %d\n", AL_MAX);
1778 sc.al_extents = AL_MAX;
1779 }
1780#undef AL_MAX
1781
ef50a3e3
LE		 1782 /* to avoid spurious errors when configuring minors before configuring
1783 * the minors they depend on: if necessary, first create the minor we
1784 * depend on */
1785 if (sc.after >= 0)
1786 ensure_mdev(sc.after, 1);
1787
b411b363
PR		 1788 /* most sanity checks done, try to assign the new sync-after
1789 * dependency. need to hold the global lock in there,
1790 * to avoid a race in the dependency loop check. */
1791 retcode = drbd_alter_sa(mdev, sc.after);
1792 if (retcode != NO_ERROR)
1793 goto fail;
1794
778f271d
PR		 1795 fifo_size = (sc.c_plan_ahead * 10 * SLEEP_TIME) / HZ;
1796 if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) {
1797 rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL);
1798 if (!rs_plan_s) {
1799 dev_err(DEV, "kmalloc of fifo_buffer failed");
1800 retcode = ERR_NOMEM;
1801 goto fail;
1802 }
1803 }
1804
b411b363
PR		 1805 /* ok, assign the rest of it as well.
1806 * lock against receive_SyncParam() */
1807 spin_lock(&mdev->peer_seq_lock);
1808 mdev->sync_conf = sc;
1809
1810 if (!rsr) {
1811 crypto_free_hash(mdev->csums_tfm);
1812 mdev->csums_tfm = csums_tfm;
1813 csums_tfm = NULL;
1814 }
1815
1816 if (!ovr) {
1817 crypto_free_hash(mdev->verify_tfm);
1818 mdev->verify_tfm = verify_tfm;
1819 verify_tfm = NULL;
1820 }
778f271d
PR		 1821
1822 if (fifo_size != mdev->rs_plan_s.size) {
1823 kfree(mdev->rs_plan_s.values);
1824 mdev->rs_plan_s.values = rs_plan_s;
1825 mdev->rs_plan_s.size = fifo_size;
1826 mdev->rs_planed = 0;
1827 rs_plan_s = NULL;
1828 }
1829
b411b363
PR		 1830 spin_unlock(&mdev->peer_seq_lock);
1831
1832 if (get_ldev(mdev)) {
1833 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
1834 drbd_al_shrink(mdev);
1835 err = drbd_check_al_size(mdev);
1836 lc_unlock(mdev->act_log);
1837 wake_up(&mdev->al_wait);
1838
1839 put_ldev(mdev);
1840 drbd_md_sync(mdev);
1841
1842 if (err) {
1843 retcode = ERR_NOMEM;
1844 goto fail;
1845 }
1846 }
1847
1848 if (mdev->state.conn >= C_CONNECTED)
1849 drbd_send_sync_param(mdev, &sc);
1850
1851 if (!cpumask_equal(mdev->cpu_mask, new_cpu_mask)) {
1852 cpumask_copy(mdev->cpu_mask, new_cpu_mask);
1853 drbd_calc_cpu_mask(mdev);
1854 mdev->receiver.reset_cpu_mask = 1;
1855 mdev->asender.reset_cpu_mask = 1;
1856 mdev->worker.reset_cpu_mask = 1;
1857 }
1858
1859 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
1860fail:
778f271d 1861 kfree(rs_plan_s);
b411b363
PR		 1862 free_cpumask_var(new_cpu_mask);
1863 crypto_free_hash(csums_tfm);
1864 crypto_free_hash(verify_tfm);
1865 reply->ret_code = retcode;
1866 return 0;
1867}
1868
1869static int drbd_nl_invalidate(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1870 struct drbd_nl_cfg_reply *reply)
1871{
1872 int retcode;
1873
1874 retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED);
1875
1876 if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION)
1877 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
1878
1879 while (retcode == SS_NEED_CONNECTION) {
1880 spin_lock_irq(&mdev->req_lock);
1881 if (mdev->state.conn < C_CONNECTED)
1882 retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL);
1883 spin_unlock_irq(&mdev->req_lock);
1884
1885 if (retcode != SS_NEED_CONNECTION)
1886 break;
1887
1888 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
1889 }
1890
1891 reply->ret_code = retcode;
1892 return 0;
1893}
1894
0778286a
PR		 1895static int drbd_bmio_set_susp_al(struct drbd_conf *mdev)
1896{
1897 int rv;
1898
1899 rv = drbd_bmio_set_n_write(mdev);
1900 drbd_suspend_al(mdev);
1901 return rv;
1902}
1903
b411b363
PR		 1904static int drbd_nl_invalidate_peer(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1905 struct drbd_nl_cfg_reply *reply)
1906{
0778286a 1907 int retcode;
b411b363 1908
0778286a
PR		 1909 retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED);
1910
1911 if (retcode < SS_SUCCESS) {
1912 if (retcode == SS_NEED_CONNECTION && mdev->state.role == R_PRIMARY) {
1913 /* The peer will get a resync upon connect anyways. Just make that
1914 into a full resync. */
1915 retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT));
1916 if (retcode >= SS_SUCCESS) {
1917 /* open coded drbd_bitmap_io() */
1918 if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al,
1919 "set_n_write from invalidate_peer"))
1920 retcode = ERR_IO_MD_DISK;
1921 }
1922 } else
1923 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S));
1924 }
b411b363 1925
0778286a 1926 reply->ret_code = retcode;
b411b363
PR		 1927 return 0;
1928}
1929
1930static int drbd_nl_pause_sync(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1931 struct drbd_nl_cfg_reply *reply)
1932{
1933 int retcode = NO_ERROR;
1934
1935 if (drbd_request_state(mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO)
1936 retcode = ERR_PAUSE_IS_SET;
1937
1938 reply->ret_code = retcode;
1939 return 0;
1940}
1941
1942static int drbd_nl_resume_sync(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1943 struct drbd_nl_cfg_reply *reply)
1944{
1945 int retcode = NO_ERROR;
1946
1947 if (drbd_request_state(mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO)
1948 retcode = ERR_PAUSE_IS_CLEAR;
1949
1950 reply->ret_code = retcode;
1951 return 0;
1952}
1953
1954static int drbd_nl_suspend_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1955 struct drbd_nl_cfg_reply *reply)
1956{
1957 reply->ret_code = drbd_request_state(mdev, NS(susp, 1));
1958
1959 return 0;
1960}
1961
1962static int drbd_nl_resume_io(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1963 struct drbd_nl_cfg_reply *reply)
1964{
43a5182c
PR		 1965 if (test_bit(NEW_CUR_UUID, &mdev->flags)) {
1966 drbd_uuid_new_current(mdev);
1967 clear_bit(NEW_CUR_UUID, &mdev->flags);
43a5182c 1968 }
265be2d0 1969 drbd_suspend_io(mdev);
fb22c402 1970 reply->ret_code = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0));
265be2d0
PR		 1971 if (reply->ret_code == SS_SUCCESS) {
1972 if (mdev->state.conn < C_CONNECTED)
1973 tl_clear(mdev);
1974 if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED)
1975 tl_restart(mdev, fail_frozen_disk_io);
1976 }
1977 drbd_resume_io(mdev);
1978
b411b363
PR		 1979 return 0;
1980}
1981
1982static int drbd_nl_outdate(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1983 struct drbd_nl_cfg_reply *reply)
1984{
1985 reply->ret_code = drbd_request_state(mdev, NS(disk, D_OUTDATED));
1986 return 0;
1987}
1988
1989static int drbd_nl_get_config(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
1990 struct drbd_nl_cfg_reply *reply)
1991{
1992 unsigned short *tl;
1993
1994 tl = reply->tag_list;
1995
1996 if (get_ldev(mdev)) {
1997 tl = disk_conf_to_tags(mdev, &mdev->ldev->dc, tl);
1998 put_ldev(mdev);
1999 }
2000
2001 if (get_net_conf(mdev)) {
2002 tl = net_conf_to_tags(mdev, mdev->net_conf, tl);
2003 put_net_conf(mdev);
2004 }
2005 tl = syncer_conf_to_tags(mdev, &mdev->sync_conf, tl);
2006
2007 put_unaligned(TT_END, tl++); /* Close the tag list */
2008
2009 return (int)((char *)tl - (char *)reply->tag_list);
2010}
2011
2012static int drbd_nl_get_state(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
2013 struct drbd_nl_cfg_reply *reply)
2014{
2015 unsigned short *tl = reply->tag_list;
2016 union drbd_state s = mdev->state;
2017 unsigned long rs_left;
2018 unsigned int res;
2019
2020 tl = get_state_to_tags(mdev, (struct get_state *)&s, tl);
2021
2022 /* no local ref, no bitmap, no syncer progress. */
2023 if (s.conn >= C_SYNC_SOURCE && s.conn <= C_PAUSED_SYNC_T) {
2024 if (get_ldev(mdev)) {
2025 drbd_get_syncer_progress(mdev, &rs_left, &res);
2026 tl = tl_add_int(tl, T_sync_progress, &res);
2027 put_ldev(mdev);
2028 }
2029 }
2030 put_unaligned(TT_END, tl++); /* Close the tag list */
2031
2032 return (int)((char *)tl - (char *)reply->tag_list);
2033}
2034
2035static int drbd_nl_get_uuids(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
2036 struct drbd_nl_cfg_reply *reply)
2037{
2038 unsigned short *tl;
2039
2040 tl = reply->tag_list;
2041
2042 if (get_ldev(mdev)) {
2043 tl = tl_add_blob(tl, T_uuids, mdev->ldev->md.uuid, UI_SIZE*sizeof(u64));
2044 tl = tl_add_int(tl, T_uuids_flags, &mdev->ldev->md.flags);
2045 put_ldev(mdev);
2046 }
2047 put_unaligned(TT_END, tl++); /* Close the tag list */
2048
2049 return (int)((char *)tl - (char *)reply->tag_list);
2050}
2051
2052/**
2053 * drbd_nl_get_timeout_flag() - Used by drbdsetup to find out which timeout value to use
2054 * @mdev: DRBD device.
2055 * @nlp: Netlink/connector packet from drbdsetup
2056 * @reply: Reply packet for drbdsetup
2057 */
2058static int drbd_nl_get_timeout_flag(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
2059 struct drbd_nl_cfg_reply *reply)
2060{
2061 unsigned short *tl;
2062 char rv;
2063
2064 tl = reply->tag_list;
2065
2066 rv = mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED :
2067 test_bit(USE_DEGR_WFC_T, &mdev->flags) ? UT_DEGRADED : UT_DEFAULT;
2068
2069 tl = tl_add_blob(tl, T_use_degraded, &rv, sizeof(rv));
2070 put_unaligned(TT_END, tl++); /* Close the tag list */
2071
2072 return (int)((char *)tl - (char *)reply->tag_list);
2073}
2074
2075static int drbd_nl_start_ov(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
2076 struct drbd_nl_cfg_reply *reply)
2077{
2078 /* default to resume from last known position, if possible */
2079 struct start_ov args =
2080 { .start_sector = mdev->ov_start_sector };
2081
2082 if (!start_ov_from_tags(mdev, nlp->tag_list, &args)) {
2083 reply->ret_code = ERR_MANDATORY_TAG;
2084 return 0;
2085 }
2086 /* w_make_ov_request expects position to be aligned */
2087 mdev->ov_start_sector = args.start_sector & ~BM_SECT_PER_BIT;
2088 reply->ret_code = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
2089 return 0;
2090}
2091
2092
2093static int drbd_nl_new_c_uuid(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp,
2094 struct drbd_nl_cfg_reply *reply)
2095{
2096 int retcode = NO_ERROR;
2097 int skip_initial_sync = 0;
2098 int err;
2099
2100 struct new_c_uuid args;
2101
2102 memset(&args, 0, sizeof(struct new_c_uuid));
2103 if (!new_c_uuid_from_tags(mdev, nlp->tag_list, &args)) {
2104 reply->ret_code = ERR_MANDATORY_TAG;
2105 return 0;
2106 }
2107
2108 mutex_lock(&mdev->state_mutex); /* Protects us against serialized state changes. */
2109
2110 if (!get_ldev(mdev)) {
2111 retcode = ERR_NO_DISK;
2112 goto out;
2113 }
2114
2115 /* this is "skip initial sync", assume to be clean */
2116 if (mdev->state.conn == C_CONNECTED && mdev->agreed_pro_version >= 90 &&
2117 mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) {
2118 dev_info(DEV, "Preparing to skip initial sync\n");
2119 skip_initial_sync = 1;
2120 } else if (mdev->state.conn != C_STANDALONE) {
2121 retcode = ERR_CONNECTED;
2122 goto out_dec;
2123 }
2124
2125 drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */
2126 drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */
2127
2128 if (args.clear_bm) {
2129 err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, "clear_n_write from new_c_uuid");
2130 if (err) {
2131 dev_err(DEV, "Writing bitmap failed with %d\n",err);
2132 retcode = ERR_IO_MD_DISK;
2133 }
2134 if (skip_initial_sync) {
2135 drbd_send_uuids_skip_initial_sync(mdev);
2136 _drbd_uuid_set(mdev, UI_BITMAP, 0);
2137 spin_lock_irq(&mdev->req_lock);
2138 _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE),
2139 CS_VERBOSE, NULL);
2140 spin_unlock_irq(&mdev->req_lock);
2141 }
2142 }
2143
2144 drbd_md_sync(mdev);
2145out_dec:
2146 put_ldev(mdev);
2147out:
2148 mutex_unlock(&mdev->state_mutex);
2149
2150 reply->ret_code = retcode;
2151 return 0;
2152}
2153
b411b363
PR		 2154struct cn_handler_struct {
2155 int (*function)(struct drbd_conf *,
2156 struct drbd_nl_cfg_req *,
2157 struct drbd_nl_cfg_reply *);
2158 int reply_body_size;
2159};
2160
2161static struct cn_handler_struct cnd_table[] = {
2162 [ P_primary ] = { &drbd_nl_primary, 0 },
2163 [ P_secondary ] = { &drbd_nl_secondary, 0 },
2164 [ P_disk_conf ] = { &drbd_nl_disk_conf, 0 },
2165 [ P_detach ] = { &drbd_nl_detach, 0 },
2166 [ P_net_conf ] = { &drbd_nl_net_conf, 0 },
2167 [ P_disconnect ] = { &drbd_nl_disconnect, 0 },
2168 [ P_resize ] = { &drbd_nl_resize, 0 },
2169 [ P_syncer_conf ] = { &drbd_nl_syncer_conf, 0 },
2170 [ P_invalidate ] = { &drbd_nl_invalidate, 0 },
2171 [ P_invalidate_peer ] = { &drbd_nl_invalidate_peer, 0 },
2172 [ P_pause_sync ] = { &drbd_nl_pause_sync, 0 },
2173 [ P_resume_sync ] = { &drbd_nl_resume_sync, 0 },
2174 [ P_suspend_io ] = { &drbd_nl_suspend_io, 0 },
2175 [ P_resume_io ] = { &drbd_nl_resume_io, 0 },
2176 [ P_outdate ] = { &drbd_nl_outdate, 0 },
2177 [ P_get_config ] = { &drbd_nl_get_config,
2178 sizeof(struct syncer_conf_tag_len_struct) +
2179 sizeof(struct disk_conf_tag_len_struct) +
2180 sizeof(struct net_conf_tag_len_struct) },
2181 [ P_get_state ] = { &drbd_nl_get_state,
2182 sizeof(struct get_state_tag_len_struct) +
2183 sizeof(struct sync_progress_tag_len_struct) },
2184 [ P_get_uuids ] = { &drbd_nl_get_uuids,
2185 sizeof(struct get_uuids_tag_len_struct) },
2186 [ P_get_timeout_flag ] = { &drbd_nl_get_timeout_flag,
2187 sizeof(struct get_timeout_flag_tag_len_struct)},
2188 [ P_start_ov ] = { &drbd_nl_start_ov, 0 },
2189 [ P_new_c_uuid ] = { &drbd_nl_new_c_uuid, 0 },
2190};
2191
9f5180e5 2192static void drbd_connector_callback(struct cn_msg *req, struct netlink_skb_parms *nsp)
b411b363
PR		 2193{
2194 struct drbd_nl_cfg_req *nlp = (struct drbd_nl_cfg_req *)req->data;
2195 struct cn_handler_struct *cm;
2196 struct cn_msg *cn_reply;
2197 struct drbd_nl_cfg_reply *reply;
2198 struct drbd_conf *mdev;
2199 int retcode, rr;
2200 int reply_size = sizeof(struct cn_msg)
2201 + sizeof(struct drbd_nl_cfg_reply)
2202 + sizeof(short int);
2203
2204 if (!try_module_get(THIS_MODULE)) {
2205 printk(KERN_ERR "drbd: try_module_get() failed!\n");
2206 return;
2207 }
2208
9f5180e5
PR		 2209 if (!cap_raised(nsp->eff_cap, CAP_SYS_ADMIN)) {
2210 retcode = ERR_PERM;
2211 goto fail;
2212 }
2213
ef50a3e3
LE		 2214 mdev = ensure_mdev(nlp->drbd_minor,
2215 (nlp->flags & DRBD_NL_CREATE_DEVICE));
b411b363
PR		 2216 if (!mdev) {
2217 retcode = ERR_MINOR_INVALID;
2218 goto fail;
2219 }
2220
42ff269d
LE		 2221 if (nlp->packet_type >= P_nl_after_last_packet ||
2222 nlp->packet_type == P_return_code_only) {
b411b363
PR		 2223 retcode = ERR_PACKET_NR;
2224 goto fail;
2225 }
2226
2227 cm = cnd_table + nlp->packet_type;
2228
2229 /* This may happen if packet number is 0: */
2230 if (cm->function == NULL) {
2231 retcode = ERR_PACKET_NR;
2232 goto fail;
2233 }
2234
2235 reply_size += cm->reply_body_size;
2236
2237 /* allocation not in the IO path, cqueue thread context */
3e3a7766 2238 cn_reply = kzalloc(reply_size, GFP_KERNEL);
b411b363
PR		 2239 if (!cn_reply) {
2240 retcode = ERR_NOMEM;
2241 goto fail;
2242 }
2243 reply = (struct drbd_nl_cfg_reply *) cn_reply->data;
2244
2245 reply->packet_type =
42ff269d 2246 cm->reply_body_size ? nlp->packet_type : P_return_code_only;
b411b363
PR		 2247 reply->minor = nlp->drbd_minor;
2248 reply->ret_code = NO_ERROR; /* Might by modified by cm->function. */
2249 /* reply->tag_list; might be modified by cm->function. */
2250
2251 rr = cm->function(mdev, nlp, reply);
2252
2253 cn_reply->id = req->id;
2254 cn_reply->seq = req->seq;
2255 cn_reply->ack = req->ack + 1;
2256 cn_reply->len = sizeof(struct drbd_nl_cfg_reply) + rr;
2257 cn_reply->flags = 0;
2258
b411b363
PR		 2259 rr = cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_KERNEL);
2260 if (rr && rr != -ESRCH)
2261 printk(KERN_INFO "drbd: cn_netlink_send()=%d\n", rr);
2262
2263 kfree(cn_reply);
2264 module_put(THIS_MODULE);
2265 return;
2266 fail:
2267 drbd_nl_send_reply(req, retcode);
2268 module_put(THIS_MODULE);
2269}
2270
2271static atomic_t drbd_nl_seq = ATOMIC_INIT(2); /* two. */
2272
2273static unsigned short *
2274__tl_add_blob(unsigned short *tl, enum drbd_tags tag, const void *data,
2275 unsigned short len, int nul_terminated)
2276{
2277 unsigned short l = tag_descriptions[tag_number(tag)].max_len;
2278 len = (len < l) ? len : l;
2279 put_unaligned(tag, tl++);
2280 put_unaligned(len, tl++);
2281 memcpy(tl, data, len);
2282 tl = (unsigned short*)((char*)tl + len);
2283 if (nul_terminated)
2284 *((char*)tl - 1) = 0;
2285 return tl;
2286}
2287
2288static unsigned short *
2289tl_add_blob(unsigned short *tl, enum drbd_tags tag, const void *data, int len)
2290{
2291 return __tl_add_blob(tl, tag, data, len, 0);
2292}
2293
2294static unsigned short *
2295tl_add_str(unsigned short *tl, enum drbd_tags tag, const char *str)
2296{
2297 return __tl_add_blob(tl, tag, str, strlen(str)+1, 0);
2298}
2299
2300static unsigned short *
2301tl_add_int(unsigned short *tl, enum drbd_tags tag, const void *val)
2302{
2303 put_unaligned(tag, tl++);
2304 switch(tag_type(tag)) {
2305 case TT_INTEGER:
2306 put_unaligned(sizeof(int), tl++);
2307 put_unaligned(*(int *)val, (int *)tl);
2308 tl = (unsigned short*)((char*)tl+sizeof(int));
2309 break;
2310 case TT_INT64:
2311 put_unaligned(sizeof(u64), tl++);
2312 put_unaligned(*(u64 *)val, (u64 *)tl);
2313 tl = (unsigned short*)((char*)tl+sizeof(u64));
2314 break;
2315 default:
2316 /* someone did something stupid. */
2317 ;
2318 }
2319 return tl;
2320}
2321
2322void drbd_bcast_state(struct drbd_conf *mdev, union drbd_state state)
2323{
2324 char buffer[sizeof(struct cn_msg)+
2325 sizeof(struct drbd_nl_cfg_reply)+
2326 sizeof(struct get_state_tag_len_struct)+
2327 sizeof(short int)];
2328 struct cn_msg *cn_reply = (struct cn_msg *) buffer;
2329 struct drbd_nl_cfg_reply *reply =
2330 (struct drbd_nl_cfg_reply *)cn_reply->data;
2331 unsigned short *tl = reply->tag_list;
2332
2333 /* dev_warn(DEV, "drbd_bcast_state() got called\n"); */
2334
2335 tl = get_state_to_tags(mdev, (struct get_state *)&state, tl);
2336
2337 put_unaligned(TT_END, tl++); /* Close the tag list */
2338
2339 cn_reply->id.idx = CN_IDX_DRBD;
2340 cn_reply->id.val = CN_VAL_DRBD;
2341
2342 cn_reply->seq = atomic_add_return(1, &drbd_nl_seq);
2343 cn_reply->ack = 0; /* not used here. */
2344 cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
2345 (int)((char *)tl - (char *)reply->tag_list);
2346 cn_reply->flags = 0;
2347
2348 reply->packet_type = P_get_state;
2349 reply->minor = mdev_to_minor(mdev);
2350 reply->ret_code = NO_ERROR;
2351
b411b363
PR		 2352 cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
2353}
2354
2355void drbd_bcast_ev_helper(struct drbd_conf *mdev, char *helper_name)
2356{
2357 char buffer[sizeof(struct cn_msg)+
2358 sizeof(struct drbd_nl_cfg_reply)+
2359 sizeof(struct call_helper_tag_len_struct)+
2360 sizeof(short int)];
2361 struct cn_msg *cn_reply = (struct cn_msg *) buffer;
2362 struct drbd_nl_cfg_reply *reply =
2363 (struct drbd_nl_cfg_reply *)cn_reply->data;
2364 unsigned short *tl = reply->tag_list;
2365
2366 /* dev_warn(DEV, "drbd_bcast_state() got called\n"); */
2367
2368 tl = tl_add_str(tl, T_helper, helper_name);
2369 put_unaligned(TT_END, tl++); /* Close the tag list */
2370
2371 cn_reply->id.idx = CN_IDX_DRBD;
2372 cn_reply->id.val = CN_VAL_DRBD;
2373
2374 cn_reply->seq = atomic_add_return(1, &drbd_nl_seq);
2375 cn_reply->ack = 0; /* not used here. */
2376 cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
2377 (int)((char *)tl - (char *)reply->tag_list);
2378 cn_reply->flags = 0;
2379
2380 reply->packet_type = P_call_helper;
2381 reply->minor = mdev_to_minor(mdev);
2382 reply->ret_code = NO_ERROR;
2383
b411b363
PR		 2384 cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
2385}
2386
2387void drbd_bcast_ee(struct drbd_conf *mdev,
2388 const char *reason, const int dgs,
2389 const char* seen_hash, const char* calc_hash,
2390 const struct drbd_epoch_entry* e)
2391{
2392 struct cn_msg *cn_reply;
2393 struct drbd_nl_cfg_reply *reply;
b411b363 2394 unsigned short *tl;
45bb912b
LE		 2395 struct page *page;
2396 unsigned len;
b411b363
PR		 2397
2398 if (!e)
2399 return;
2400 if (!reason || !reason[0])
2401 return;
2402
2403 /* apparently we have to memcpy twice, first to prepare the data for the
2404 * struct cn_msg, then within cn_netlink_send from the cn_msg to the
2405 * netlink skb. */
2406 /* receiver thread context, which is not in the writeout path (of this node),
2407 * but may be in the writeout path of the _other_ node.
2408 * GFP_NOIO to avoid potential "distributed deadlock". */
3e3a7766 2409 cn_reply = kzalloc(
b411b363
PR		 2410 sizeof(struct cn_msg)+
2411 sizeof(struct drbd_nl_cfg_reply)+
2412 sizeof(struct dump_ee_tag_len_struct)+
2413 sizeof(short int),
2414 GFP_NOIO);
2415
2416 if (!cn_reply) {
2417 dev_err(DEV, "could not kmalloc buffer for drbd_bcast_ee, sector %llu, size %u\n",
2418 (unsigned long long)e->sector, e->size);
2419 return;
2420 }
2421
2422 reply = (struct drbd_nl_cfg_reply*)cn_reply->data;
2423 tl = reply->tag_list;
2424
2425 tl = tl_add_str(tl, T_dump_ee_reason, reason);
2426 tl = tl_add_blob(tl, T_seen_digest, seen_hash, dgs);
2427 tl = tl_add_blob(tl, T_calc_digest, calc_hash, dgs);
2428 tl = tl_add_int(tl, T_ee_sector, &e->sector);
2429 tl = tl_add_int(tl, T_ee_block_id, &e->block_id);
2430
3129b1b9
LE		 2431 /* dump the first 32k */
2432 len = min_t(unsigned, e->size, 32 << 10);
b411b363 2433 put_unaligned(T_ee_data, tl++);
3129b1b9 2434 put_unaligned(len, tl++);
b411b363 2435
45bb912b
LE		 2436 page = e->pages;
2437 page_chain_for_each(page) {
2438 void *d = kmap_atomic(page, KM_USER0);
2439 unsigned l = min_t(unsigned, len, PAGE_SIZE);
2440 memcpy(tl, d, l);
2441 kunmap_atomic(d, KM_USER0);
2442 tl = (unsigned short*)((char*)tl + l);
2443 len -= l;
3129b1b9
LE		 2444 if (len == 0)
2445 break;
b411b363
PR		 2446 }
2447 put_unaligned(TT_END, tl++); /* Close the tag list */
2448
2449 cn_reply->id.idx = CN_IDX_DRBD;
2450 cn_reply->id.val = CN_VAL_DRBD;
2451
2452 cn_reply->seq = atomic_add_return(1,&drbd_nl_seq);
2453 cn_reply->ack = 0; // not used here.
2454 cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
2455 (int)((char*)tl - (char*)reply->tag_list);
2456 cn_reply->flags = 0;
2457
2458 reply->packet_type = P_dump_ee;
2459 reply->minor = mdev_to_minor(mdev);
2460 reply->ret_code = NO_ERROR;
2461
b411b363
PR		 2462 cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
2463 kfree(cn_reply);
2464}
2465
2466void drbd_bcast_sync_progress(struct drbd_conf *mdev)
2467{
2468 char buffer[sizeof(struct cn_msg)+
2469 sizeof(struct drbd_nl_cfg_reply)+
2470 sizeof(struct sync_progress_tag_len_struct)+
2471 sizeof(short int)];
2472 struct cn_msg *cn_reply = (struct cn_msg *) buffer;
2473 struct drbd_nl_cfg_reply *reply =
2474 (struct drbd_nl_cfg_reply *)cn_reply->data;
2475 unsigned short *tl = reply->tag_list;
2476 unsigned long rs_left;
2477 unsigned int res;
2478
2479 /* no local ref, no bitmap, no syncer progress, no broadcast. */
2480 if (!get_ldev(mdev))
2481 return;
2482 drbd_get_syncer_progress(mdev, &rs_left, &res);
2483 put_ldev(mdev);
2484
2485 tl = tl_add_int(tl, T_sync_progress, &res);
2486 put_unaligned(TT_END, tl++); /* Close the tag list */
2487
2488 cn_reply->id.idx = CN_IDX_DRBD;
2489 cn_reply->id.val = CN_VAL_DRBD;
2490
2491 cn_reply->seq = atomic_add_return(1, &drbd_nl_seq);
2492 cn_reply->ack = 0; /* not used here. */
2493 cn_reply->len = sizeof(struct drbd_nl_cfg_reply) +
2494 (int)((char *)tl - (char *)reply->tag_list);
2495 cn_reply->flags = 0;
2496
2497 reply->packet_type = P_sync_progress;
2498 reply->minor = mdev_to_minor(mdev);
2499 reply->ret_code = NO_ERROR;
2500
b411b363
PR		 2501 cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
2502}
2503
2504int __init drbd_nl_init(void)
2505{
2506 static struct cb_id cn_id_drbd;
2507 int err, try=10;
2508
2509 cn_id_drbd.val = CN_VAL_DRBD;
2510 do {
2511 cn_id_drbd.idx = cn_idx;
2512 err = cn_add_callback(&cn_id_drbd, "cn_drbd", &drbd_connector_callback);
2513 if (!err)
2514 break;
2515 cn_idx = (cn_idx + CN_IDX_STEP);
2516 } while (try--);
2517
2518 if (err) {
2519 printk(KERN_ERR "drbd: cn_drbd failed to register\n");
2520 return err;
2521 }
2522
2523 return 0;
2524}
2525
2526void drbd_nl_cleanup(void)
2527{
2528 static struct cb_id cn_id_drbd;
2529
2530 cn_id_drbd.idx = cn_idx;
2531 cn_id_drbd.val = CN_VAL_DRBD;
2532
2533 cn_del_callback(&cn_id_drbd);
2534}
2535
2536void drbd_nl_send_reply(struct cn_msg *req, int ret_code)
2537{
2538 char buffer[sizeof(struct cn_msg)+sizeof(struct drbd_nl_cfg_reply)];
2539 struct cn_msg *cn_reply = (struct cn_msg *) buffer;
2540 struct drbd_nl_cfg_reply *reply =
2541 (struct drbd_nl_cfg_reply *)cn_reply->data;
2542 int rr;
2543
3e3a7766 2544 memset(buffer, 0, sizeof(buffer));
b411b363
PR		 2545 cn_reply->id = req->id;
2546
2547 cn_reply->seq = req->seq;
2548 cn_reply->ack = req->ack + 1;
2549 cn_reply->len = sizeof(struct drbd_nl_cfg_reply);
2550 cn_reply->flags = 0;
2551
42ff269d 2552 reply->packet_type = P_return_code_only;
b411b363
PR		 2553 reply->minor = ((struct drbd_nl_cfg_req *)req->data)->drbd_minor;
2554 reply->ret_code = ret_code;
2555
b411b363
PR		 2556 rr = cn_netlink_send(cn_reply, CN_IDX_DRBD, GFP_NOIO);
2557 if (rr && rr != -ESRCH)
2558 printk(KERN_INFO "drbd: cn_netlink_send()=%d\n", rr);
2559}
2560
Linux kernel - Deliverables
This page took 0.185116 seconds and 5 git commands to generate.