4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
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>.
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)
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.
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.
26 #include <linux/module.h>
27 #include <linux/drbd.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>
37 #include "drbd_wrappers.h"
38 #include <asm/unaligned.h>
39 #include <linux/drbd_tag_magic.h>
40 #include <linux/drbd_limits.h>
41 #include <linux/compiler.h>
42 #include <linux/kthread.h>
44 static unsigned short *tl_add_blob(unsigned short *, enum drbd_tags
, const void *, int);
45 static unsigned short *tl_add_str(unsigned short *, enum drbd_tags
, const char *);
46 static unsigned short *tl_add_int(unsigned short *, enum drbd_tags
, const void *);
48 /* see get_sb_bdev and bd_claim */
49 static char *drbd_m_holder
= "Hands off! this is DRBD's meta data device.";
51 /* Generate the tag_list to struct functions */
52 #define NL_PACKET(name, number, fields) \
53 static int name ## _from_tags(struct drbd_conf *mdev, \
54 unsigned short *tags, struct name *arg) __attribute__ ((unused)); \
55 static int name ## _from_tags(struct drbd_conf *mdev, \
56 unsigned short *tags, struct name *arg) \
61 while ((tag = get_unaligned(tags++)) != TT_END) { \
62 dlen = get_unaligned(tags++); \
63 switch (tag_number(tag)) { \
66 if (tag & T_MANDATORY) { \
67 dev_err(DEV, "Unknown tag: %d\n", tag_number(tag)); \
71 tags = (unsigned short *)((char *)tags + dlen); \
75 #define NL_INTEGER(pn, pr, member) \
76 case pn: /* D_ASSERT( tag_type(tag) == TT_INTEGER ); */ \
77 arg->member = get_unaligned((int *)(tags)); \
79 #define NL_INT64(pn, pr, member) \
80 case pn: /* D_ASSERT( tag_type(tag) == TT_INT64 ); */ \
81 arg->member = get_unaligned((u64 *)(tags)); \
83 #define NL_BIT(pn, pr, member) \
84 case pn: /* D_ASSERT( tag_type(tag) == TT_BIT ); */ \
85 arg->member = *(char *)(tags) ? 1 : 0; \
87 #define NL_STRING(pn, pr, member, len) \
88 case pn: /* D_ASSERT( tag_type(tag) == TT_STRING ); */ \
90 dev_err(DEV, "arg too long: %s (%u wanted, max len: %u bytes)\n", \
91 #member, dlen, (unsigned int)len); \
94 arg->member ## _len = dlen; \
95 memcpy(arg->member, tags, min_t(size_t, dlen, len)); \
97 #include "linux/drbd_nl.h"
99 /* Generate the struct to tag_list functions */
100 #define NL_PACKET(name, number, fields) \
101 static unsigned short* \
102 name ## _to_tags(struct drbd_conf *mdev, \
103 struct name *arg, unsigned short *tags) __attribute__ ((unused)); \
104 static unsigned short* \
105 name ## _to_tags(struct drbd_conf *mdev, \
106 struct name *arg, unsigned short *tags) \
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"
134 void drbd_bcast_ev_helper(struct drbd_conf
*mdev
, char *helper_name
);
135 void drbd_nl_send_reply(struct cn_msg
*, int);
137 int drbd_khelper(struct drbd_conf
*mdev
, char *cmd
)
139 char *envp
[] = { "HOME=/",
141 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
142 NULL
, /* Will be set to address family */
143 NULL
, /* Will be set to address */
146 char mb
[12], af
[20], ad
[60], *afs
;
147 char *argv
[] = {usermode_helper
, cmd
, mb
, NULL
};
150 snprintf(mb
, 12, "minor-%d", mdev_to_minor(mdev
));
152 if (get_net_conf(mdev
)) {
153 switch (((struct sockaddr
*)mdev
->net_conf
->peer_addr
)->sa_family
) {
156 snprintf(ad
, 60, "DRBD_PEER_ADDRESS=%pI6",
157 &((struct sockaddr_in6
*)mdev
->net_conf
->peer_addr
)->sin6_addr
);
161 snprintf(ad
, 60, "DRBD_PEER_ADDRESS=%pI4",
162 &((struct sockaddr_in
*)mdev
->net_conf
->peer_addr
)->sin_addr
);
166 snprintf(ad
, 60, "DRBD_PEER_ADDRESS=%pI4",
167 &((struct sockaddr_in
*)mdev
->net_conf
->peer_addr
)->sin_addr
);
169 snprintf(af
, 20, "DRBD_PEER_AF=%s", afs
);
175 /* The helper may take some time.
176 * write out any unsynced meta data changes now */
179 dev_info(DEV
, "helper command: %s %s %s\n", usermode_helper
, cmd
, mb
);
181 drbd_bcast_ev_helper(mdev
, cmd
);
182 ret
= call_usermodehelper(usermode_helper
, argv
, envp
, 1);
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
);
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
);
192 if (ret
< 0) /* Ignore any ERRNOs we got. */
198 enum drbd_disk_state
drbd_try_outdate_peer(struct drbd_conf
*mdev
)
202 enum drbd_disk_state nps
;
203 enum drbd_fencing_p fp
;
205 D_ASSERT(mdev
->state
.pdsk
== D_UNKNOWN
);
207 if (get_ldev_if_state(mdev
, D_CONSISTENT
)) {
208 fp
= mdev
->ldev
->dc
.fencing
;
211 dev_warn(DEV
, "Not fencing peer, I'm not even Consistent myself.\n");
212 nps
= mdev
->state
.pdsk
;
216 r
= drbd_khelper(mdev
, "fence-peer");
218 switch ((r
>>8) & 0xff) {
219 case 3: /* peer is inconsistent */
220 ex_to_string
= "peer is inconsistent or worse";
221 nps
= D_INCONSISTENT
;
223 case 4: /* peer got outdated, or was already outdated */
224 ex_to_string
= "peer was fenced";
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";
233 ex_to_string
= "peer unreachable, doing nothing since disk != UpToDate";
234 nps
= mdev
->state
.pdsk
;
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");
243 _drbd_request_state(mdev
, NS(disk
, D_OUTDATED
), CS_WAIT_COMPLETE
);
246 if (fp
!= FP_STONITH
)
247 dev_err(DEV
, "fence-peer() = 7 && fencing != Stonith !!!\n");
248 ex_to_string
= "peer was stonithed";
252 /* The script is broken ... */
254 dev_err(DEV
, "fence-peer helper broken, returned %d\n", (r
>>8)&0xff);
258 dev_info(DEV
, "fence-peer helper returned %d (%s)\n",
259 (r
>>8) & 0xff, ex_to_string
);
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
);
271 static int _try_outdate_peer_async(void *data
)
273 struct drbd_conf
*mdev
= (struct drbd_conf
*)data
;
274 enum drbd_disk_state nps
;
276 nps
= drbd_try_outdate_peer(mdev
);
277 drbd_request_state(mdev
, NS(pdsk
, nps
));
282 void drbd_try_outdate_peer_async(struct drbd_conf
*mdev
)
284 struct task_struct
*opa
;
286 opa
= kthread_run(_try_outdate_peer_async
, mdev
, "drbd%d_a_helper", mdev_to_minor(mdev
));
288 dev_err(DEV
, "out of mem, failed to invoke fence-peer helper\n");
292 drbd_set_role(struct drbd_conf
*mdev
, enum drbd_role new_role
, int force
)
294 const int max_tries
= 4;
295 enum drbd_state_rv rv
= SS_UNKNOWN_ERROR
;
298 union drbd_state mask
, val
;
299 enum drbd_disk_state nps
;
301 if (new_role
== R_PRIMARY
)
302 request_ping(mdev
); /* Detect a dead peer ASAP */
304 mutex_lock(&mdev
->state_mutex
);
306 mask
.i
= 0; mask
.role
= R_MASK
;
307 val
.i
= 0; val
.role
= new_role
;
309 while (try++ < max_tries
) {
310 rv
= _drbd_request_state(mdev
, mask
, val
, CS_WAIT_COMPLETE
);
312 /* in case we first succeeded to outdate,
313 * but now suddenly could establish a connection */
314 if (rv
== SS_CW_FAILED_BY_PEER
&& mask
.pdsk
!= 0) {
320 if (rv
== SS_NO_UP_TO_DATE_DISK
&& force
&&
321 (mdev
->state
.disk
< D_UP_TO_DATE
&&
322 mdev
->state
.disk
>= D_INCONSISTENT
)) {
324 val
.disk
= D_UP_TO_DATE
;
329 if (rv
== SS_NO_UP_TO_DATE_DISK
&&
330 mdev
->state
.disk
== D_CONSISTENT
&& mask
.pdsk
== 0) {
331 D_ASSERT(mdev
->state
.pdsk
== D_UNKNOWN
);
332 nps
= drbd_try_outdate_peer(mdev
);
334 if (nps
== D_OUTDATED
|| nps
== D_INCONSISTENT
) {
335 val
.disk
= D_UP_TO_DATE
;
345 if (rv
== SS_NOTHING_TO_DO
)
347 if (rv
== SS_PRIMARY_NOP
&& mask
.pdsk
== 0) {
348 nps
= drbd_try_outdate_peer(mdev
);
350 if (force
&& nps
> D_OUTDATED
) {
351 dev_warn(DEV
, "Forced into split brain situation!\n");
360 if (rv
== SS_TWO_PRIMARIES
) {
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);
369 if (rv
< SS_SUCCESS
) {
370 rv
= _drbd_request_state(mdev
, mask
, val
,
371 CS_VERBOSE
+ CS_WAIT_COMPLETE
);
382 dev_warn(DEV
, "Forced to consider local data as UpToDate!\n");
384 /* Wait until nothing is on the fly :) */
385 wait_event(mdev
->misc_wait
, atomic_read(&mdev
->ap_pending_cnt
) == 0);
387 if (new_role
== R_SECONDARY
) {
388 set_disk_ro(mdev
->vdisk
, TRUE
);
389 if (get_ldev(mdev
)) {
390 mdev
->ldev
->md
.uuid
[UI_CURRENT
] &= ~(u64
)1;
394 if (get_net_conf(mdev
)) {
395 mdev
->net_conf
->want_lose
= 0;
398 set_disk_ro(mdev
->vdisk
, FALSE
);
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
);
405 mdev
->ldev
->md
.uuid
[UI_CURRENT
] |= (u64
)1;
410 if ((new_role
== R_SECONDARY
) && get_ldev(mdev
)) {
411 drbd_al_to_on_disk_bm(mdev
);
415 if (mdev
->state
.conn
>= C_WF_REPORT_PARAMS
) {
416 /* if this was forced, we should consider sync */
418 drbd_send_uuids(mdev
);
419 drbd_send_state(mdev
);
424 kobject_uevent(&disk_to_dev(mdev
->vdisk
)->kobj
, KOBJ_CHANGE
);
426 mutex_unlock(&mdev
->state_mutex
);
430 static struct drbd_conf
*ensure_mdev(int minor
, int create
)
432 struct drbd_conf
*mdev
;
434 if (minor
>= minor_count
)
437 mdev
= minor_to_mdev(minor
);
439 if (!mdev
&& create
) {
440 struct gendisk
*disk
= NULL
;
441 mdev
= drbd_new_device(minor
);
443 spin_lock_irq(&drbd_pp_lock
);
444 if (minor_table
[minor
] == NULL
) {
445 minor_table
[minor
] = mdev
;
448 } /* else: we lost the race */
449 spin_unlock_irq(&drbd_pp_lock
);
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! */
455 else /* we lost the race above */
456 drbd_free_mdev(mdev
);
458 mdev
= minor_to_mdev(minor
);
464 static int drbd_nl_primary(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
465 struct drbd_nl_cfg_reply
*reply
)
467 struct primary primary_args
;
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
;
476 drbd_set_role(mdev
, R_PRIMARY
, primary_args
.primary_force
);
481 static int drbd_nl_secondary(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
482 struct drbd_nl_cfg_reply
*reply
)
484 reply
->ret_code
= drbd_set_role(mdev
, R_SECONDARY
, 0);
489 /* initializes the md.*_offset members, so we are able to find
490 * the on disk meta data */
491 static void drbd_md_set_sector_offsets(struct drbd_conf
*mdev
,
492 struct drbd_backing_dev
*bdev
)
494 sector_t md_size_sect
= 0;
495 switch (bdev
->dc
.meta_dev_idx
) {
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
;
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
;
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);
521 /* plus the "drbd meta data super block",
522 * and the activity log; */
523 md_size_sect
+= MD_BM_OFFSET
;
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
;
532 char *ppsize(char *buf
, unsigned long long size
)
534 /* Needs 9 bytes at max. */
535 static char units
[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
537 while (size
>= 10000) {
539 size
= (size
>> 10) + !!(size
& (1<<9));
542 sprintf(buf
, "%lu %cB", (long)size
, units
[base
]);
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.
560 void drbd_suspend_io(struct drbd_conf
*mdev
)
562 set_bit(SUSPEND_IO
, &mdev
->flags
);
563 if (is_susp(mdev
->state
))
565 wait_event(mdev
->misc_wait
, !atomic_read(&mdev
->ap_bio_cnt
));
568 void drbd_resume_io(struct drbd_conf
*mdev
)
570 clear_bit(SUSPEND_IO
, &mdev
->flags
);
571 wake_up(&mdev
->misc_wait
);
575 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
576 * @mdev: DRBD device.
578 * Returns 0 on success, negative return values indicate errors.
579 * You should call drbd_md_sync() after calling this function.
581 enum determine_dev_size
drbd_determin_dev_size(struct drbd_conf
*mdev
, enum dds_flags flags
) __must_hold(local
)
583 sector_t prev_first_sect
, prev_size
; /* previous meta location */
588 int md_moved
, la_size_changed
;
589 enum determine_dev_size rv
= unchanged
;
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.
597 * Suspend IO right here.
598 * still lock the act_log to not trigger ASSERTs there.
600 drbd_suspend_io(mdev
);
602 /* no wait necessary anymore, actually we could assert that */
603 wait_event(mdev
->al_wait
, lc_try_lock(mdev
->act_log
));
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
;
609 /* TODO: should only be some assert here, not (re)init... */
610 drbd_md_set_sector_offsets(mdev
, mdev
->ldev
);
612 size
= drbd_new_dev_size(mdev
, mdev
->ldev
, flags
& DDSF_FORCED
);
614 if (drbd_get_capacity(mdev
->this_bdev
) != size
||
615 drbd_bm_capacity(mdev
) != size
) {
617 err
= drbd_bm_resize(mdev
, size
, !(flags
& DDSF_NO_RESYNC
));
619 /* currently there is only one error: ENOMEM! */
620 size
= drbd_bm_capacity(mdev
)>>1;
622 dev_err(DEV
, "OUT OF MEMORY! "
623 "Could not allocate bitmap!\n");
625 dev_err(DEV
, "BM resizing failed. "
626 "Leaving size unchanged at size = %lu KB\n",
627 (unsigned long)size
);
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);
637 if (rv
== dev_size_error
)
640 la_size_changed
= (la_size
!= mdev
->ldev
->md
.la_size_sect
);
642 md_moved
= prev_first_sect
!= drbd_md_first_sector(mdev
->ldev
)
643 || prev_size
!= mdev
->ldev
->md
.md_size_sect
;
645 if (la_size_changed
|| md_moved
) {
646 drbd_al_shrink(mdev
); /* All extents inactive. */
647 dev_info(DEV
, "Writing the whole bitmap, %s\n",
648 la_size_changed
&& md_moved
? "size changed and md moved" :
649 la_size_changed
? "size changed" : "md moved");
650 rv
= drbd_bitmap_io(mdev
, &drbd_bm_write
, "size changed"); /* does drbd_resume_io() ! */
651 drbd_md_mark_dirty(mdev
);
659 lc_unlock(mdev
->act_log
);
660 wake_up(&mdev
->al_wait
);
661 drbd_resume_io(mdev
);
667 drbd_new_dev_size(struct drbd_conf
*mdev
, struct drbd_backing_dev
*bdev
, int assume_peer_has_space
)
669 sector_t p_size
= mdev
->p_size
; /* partner's disk size. */
670 sector_t la_size
= bdev
->md
.la_size_sect
; /* last agreed size. */
671 sector_t m_size
; /* my size */
672 sector_t u_size
= bdev
->dc
.disk_size
; /* size requested by user. */
675 m_size
= drbd_get_max_capacity(bdev
);
677 if (mdev
->state
.conn
< C_CONNECTED
&& assume_peer_has_space
) {
678 dev_warn(DEV
, "Resize while not connected was forced by the user!\n");
682 if (p_size
&& m_size
) {
683 size
= min_t(sector_t
, p_size
, m_size
);
687 if (m_size
&& m_size
< size
)
689 if (p_size
&& p_size
< size
)
700 dev_err(DEV
, "Both nodes diskless!\n");
704 dev_err(DEV
, "Requested disk size is too big (%lu > %lu)\n",
705 (unsigned long)u_size
>>1, (unsigned long)size
>>1);
714 * drbd_check_al_size() - Ensures that the AL is of the right size
715 * @mdev: DRBD device.
717 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
718 * failed, and 0 on success. You should call drbd_md_sync() after you called
721 static int drbd_check_al_size(struct drbd_conf
*mdev
)
723 struct lru_cache
*n
, *t
;
724 struct lc_element
*e
;
728 ERR_IF(mdev
->sync_conf
.al_extents
< 7)
729 mdev
->sync_conf
.al_extents
= 127;
732 mdev
->act_log
->nr_elements
== mdev
->sync_conf
.al_extents
)
737 n
= lc_create("act_log", drbd_al_ext_cache
,
738 mdev
->sync_conf
.al_extents
, sizeof(struct lc_element
), 0);
741 dev_err(DEV
, "Cannot allocate act_log lru!\n");
744 spin_lock_irq(&mdev
->al_lock
);
746 for (i
= 0; i
< t
->nr_elements
; i
++) {
747 e
= lc_element_by_index(t
, i
);
749 dev_err(DEV
, "refcnt(%d)==%d\n",
750 e
->lc_number
, e
->refcnt
);
756 spin_unlock_irq(&mdev
->al_lock
);
758 dev_err(DEV
, "Activity log still in use!\n");
765 drbd_md_mark_dirty(mdev
); /* we changed mdev->act_log->nr_elemens */
769 void drbd_setup_queue_param(struct drbd_conf
*mdev
, unsigned int max_bio_size
) __must_hold(local
)
771 struct request_queue
* const q
= mdev
->rq_queue
;
772 struct request_queue
* const b
= mdev
->ldev
->backing_bdev
->bd_disk
->queue
;
773 int max_segments
= mdev
->ldev
->dc
.max_bio_bvecs
;
774 int max_hw_sectors
= min(queue_max_hw_sectors(b
), max_bio_size
>> 9);
776 blk_queue_logical_block_size(q
, 512);
777 blk_queue_max_hw_sectors(q
, max_hw_sectors
);
778 /* This is the workaround for "bio would need to, but cannot, be split" */
779 blk_queue_max_segments(q
, max_segments
? max_segments
: BLK_MAX_SEGMENTS
);
780 blk_queue_segment_boundary(q
, PAGE_CACHE_SIZE
-1);
781 blk_queue_stack_limits(q
, b
);
783 dev_info(DEV
, "max BIO size = %u\n", queue_max_hw_sectors(q
) << 9);
785 if (q
->backing_dev_info
.ra_pages
!= b
->backing_dev_info
.ra_pages
) {
786 dev_info(DEV
, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
787 q
->backing_dev_info
.ra_pages
,
788 b
->backing_dev_info
.ra_pages
);
789 q
->backing_dev_info
.ra_pages
= b
->backing_dev_info
.ra_pages
;
793 /* serialize deconfig (worker exiting, doing cleanup)
794 * and reconfig (drbdsetup disk, drbdsetup net)
796 * Wait for a potentially exiting worker, then restart it,
797 * or start a new one. Flush any pending work, there may still be an
798 * after_state_change queued.
800 static void drbd_reconfig_start(struct drbd_conf
*mdev
)
802 wait_event(mdev
->state_wait
, !test_and_set_bit(CONFIG_PENDING
, &mdev
->flags
));
803 wait_event(mdev
->state_wait
, !test_bit(DEVICE_DYING
, &mdev
->flags
));
804 drbd_thread_start(&mdev
->worker
);
805 drbd_flush_workqueue(mdev
);
808 /* if still unconfigured, stops worker again.
809 * if configured now, clears CONFIG_PENDING.
810 * wakes potential waiters */
811 static void drbd_reconfig_done(struct drbd_conf
*mdev
)
813 spin_lock_irq(&mdev
->req_lock
);
814 if (mdev
->state
.disk
== D_DISKLESS
&&
815 mdev
->state
.conn
== C_STANDALONE
&&
816 mdev
->state
.role
== R_SECONDARY
) {
817 set_bit(DEVICE_DYING
, &mdev
->flags
);
818 drbd_thread_stop_nowait(&mdev
->worker
);
820 clear_bit(CONFIG_PENDING
, &mdev
->flags
);
821 spin_unlock_irq(&mdev
->req_lock
);
822 wake_up(&mdev
->state_wait
);
825 /* Make sure IO is suspended before calling this function(). */
826 static void drbd_suspend_al(struct drbd_conf
*mdev
)
830 if (lc_try_lock(mdev
->act_log
)) {
831 drbd_al_shrink(mdev
);
832 lc_unlock(mdev
->act_log
);
834 dev_warn(DEV
, "Failed to lock al in drbd_suspend_al()\n");
838 spin_lock_irq(&mdev
->req_lock
);
839 if (mdev
->state
.conn
< C_CONNECTED
)
840 s
= !test_and_set_bit(AL_SUSPENDED
, &mdev
->flags
);
842 spin_unlock_irq(&mdev
->req_lock
);
845 dev_info(DEV
, "Suspended AL updates\n");
848 /* does always return 0;
849 * interesting return code is in reply->ret_code */
850 static int drbd_nl_disk_conf(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
851 struct drbd_nl_cfg_reply
*reply
)
853 enum drbd_ret_code retcode
;
854 enum determine_dev_size dd
;
855 sector_t max_possible_sectors
;
856 sector_t min_md_device_sectors
;
857 struct drbd_backing_dev
*nbc
= NULL
; /* new_backing_conf */
858 struct block_device
*bdev
;
859 struct lru_cache
*resync_lru
= NULL
;
860 union drbd_state ns
, os
;
861 unsigned int max_bio_size
;
863 int cp_discovered
= 0;
864 int logical_block_size
;
866 drbd_reconfig_start(mdev
);
868 /* if you want to reconfigure, please tear down first */
869 if (mdev
->state
.disk
> D_DISKLESS
) {
870 retcode
= ERR_DISK_CONFIGURED
;
873 /* It may just now have detached because of IO error. Make sure
874 * drbd_ldev_destroy is done already, we may end up here very fast,
875 * e.g. if someone calls attach from the on-io-error handler,
876 * to realize a "hot spare" feature (not that I'd recommend that) */
877 wait_event(mdev
->misc_wait
, !atomic_read(&mdev
->local_cnt
));
879 /* allocation not in the IO path, cqueue thread context */
880 nbc
= kzalloc(sizeof(struct drbd_backing_dev
), GFP_KERNEL
);
886 nbc
->dc
.disk_size
= DRBD_DISK_SIZE_SECT_DEF
;
887 nbc
->dc
.on_io_error
= DRBD_ON_IO_ERROR_DEF
;
888 nbc
->dc
.fencing
= DRBD_FENCING_DEF
;
889 nbc
->dc
.max_bio_bvecs
= DRBD_MAX_BIO_BVECS_DEF
;
891 if (!disk_conf_from_tags(mdev
, nlp
->tag_list
, &nbc
->dc
)) {
892 retcode
= ERR_MANDATORY_TAG
;
896 if (nbc
->dc
.meta_dev_idx
< DRBD_MD_INDEX_FLEX_INT
) {
897 retcode
= ERR_MD_IDX_INVALID
;
901 if (get_net_conf(mdev
)) {
902 int prot
= mdev
->net_conf
->wire_protocol
;
904 if (nbc
->dc
.fencing
== FP_STONITH
&& prot
== DRBD_PROT_A
) {
905 retcode
= ERR_STONITH_AND_PROT_A
;
910 bdev
= blkdev_get_by_path(nbc
->dc
.backing_dev
,
911 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
, mdev
);
913 dev_err(DEV
, "open(\"%s\") failed with %ld\n", nbc
->dc
.backing_dev
,
915 retcode
= ERR_OPEN_DISK
;
918 nbc
->backing_bdev
= bdev
;
921 * meta_dev_idx >= 0: external fixed size, possibly multiple
922 * drbd sharing one meta device. TODO in that case, paranoia
923 * check that [md_bdev, meta_dev_idx] is not yet used by some
924 * other drbd minor! (if you use drbd.conf + drbdadm, that
925 * should check it for you already; but if you don't, or
926 * someone fooled it, we need to double check here)
928 bdev
= blkdev_get_by_path(nbc
->dc
.meta_dev
,
929 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
,
930 (nbc
->dc
.meta_dev_idx
< 0) ?
931 (void *)mdev
: (void *)drbd_m_holder
);
933 dev_err(DEV
, "open(\"%s\") failed with %ld\n", nbc
->dc
.meta_dev
,
935 retcode
= ERR_OPEN_MD_DISK
;
940 if ((nbc
->backing_bdev
== nbc
->md_bdev
) !=
941 (nbc
->dc
.meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
942 nbc
->dc
.meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)) {
943 retcode
= ERR_MD_IDX_INVALID
;
947 resync_lru
= lc_create("resync", drbd_bm_ext_cache
,
948 61, sizeof(struct bm_extent
),
949 offsetof(struct bm_extent
, lce
));
955 /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */
956 drbd_md_set_sector_offsets(mdev
, nbc
);
958 if (drbd_get_max_capacity(nbc
) < nbc
->dc
.disk_size
) {
959 dev_err(DEV
, "max capacity %llu smaller than disk size %llu\n",
960 (unsigned long long) drbd_get_max_capacity(nbc
),
961 (unsigned long long) nbc
->dc
.disk_size
);
962 retcode
= ERR_DISK_TO_SMALL
;
966 if (nbc
->dc
.meta_dev_idx
< 0) {
967 max_possible_sectors
= DRBD_MAX_SECTORS_FLEX
;
968 /* at least one MB, otherwise it does not make sense */
969 min_md_device_sectors
= (2<<10);
971 max_possible_sectors
= DRBD_MAX_SECTORS
;
972 min_md_device_sectors
= MD_RESERVED_SECT
* (nbc
->dc
.meta_dev_idx
+ 1);
975 if (drbd_get_capacity(nbc
->md_bdev
) < min_md_device_sectors
) {
976 retcode
= ERR_MD_DISK_TO_SMALL
;
977 dev_warn(DEV
, "refusing attach: md-device too small, "
978 "at least %llu sectors needed for this meta-disk type\n",
979 (unsigned long long) min_md_device_sectors
);
983 /* Make sure the new disk is big enough
984 * (we may currently be R_PRIMARY with no local disk...) */
985 if (drbd_get_max_capacity(nbc
) <
986 drbd_get_capacity(mdev
->this_bdev
)) {
987 retcode
= ERR_DISK_TO_SMALL
;
991 nbc
->known_size
= drbd_get_capacity(nbc
->backing_bdev
);
993 if (nbc
->known_size
> max_possible_sectors
) {
994 dev_warn(DEV
, "==> truncating very big lower level device "
995 "to currently maximum possible %llu sectors <==\n",
996 (unsigned long long) max_possible_sectors
);
997 if (nbc
->dc
.meta_dev_idx
>= 0)
998 dev_warn(DEV
, "==>> using internal or flexible "
999 "meta data may help <<==\n");
1002 drbd_suspend_io(mdev
);
1003 /* also wait for the last barrier ack. */
1004 wait_event(mdev
->misc_wait
, !atomic_read(&mdev
->ap_pending_cnt
) || is_susp(mdev
->state
));
1005 /* and for any other previously queued work */
1006 drbd_flush_workqueue(mdev
);
1008 retcode
= _drbd_request_state(mdev
, NS(disk
, D_ATTACHING
), CS_VERBOSE
);
1009 drbd_resume_io(mdev
);
1010 if (retcode
< SS_SUCCESS
)
1013 if (!get_ldev_if_state(mdev
, D_ATTACHING
))
1014 goto force_diskless
;
1016 drbd_md_set_sector_offsets(mdev
, nbc
);
1018 /* allocate a second IO page if logical_block_size != 512 */
1019 logical_block_size
= bdev_logical_block_size(nbc
->md_bdev
);
1020 if (logical_block_size
== 0)
1021 logical_block_size
= MD_SECTOR_SIZE
;
1023 if (logical_block_size
!= MD_SECTOR_SIZE
) {
1024 if (!mdev
->md_io_tmpp
) {
1025 struct page
*page
= alloc_page(GFP_NOIO
);
1027 goto force_diskless_dec
;
1029 dev_warn(DEV
, "Meta data's bdev logical_block_size = %d != %d\n",
1030 logical_block_size
, MD_SECTOR_SIZE
);
1031 dev_warn(DEV
, "Workaround engaged (has performance impact).\n");
1033 mdev
->md_io_tmpp
= page
;
1037 if (!mdev
->bitmap
) {
1038 if (drbd_bm_init(mdev
)) {
1039 retcode
= ERR_NOMEM
;
1040 goto force_diskless_dec
;
1044 retcode
= drbd_md_read(mdev
, nbc
);
1045 if (retcode
!= NO_ERROR
)
1046 goto force_diskless_dec
;
1048 if (mdev
->state
.conn
< C_CONNECTED
&&
1049 mdev
->state
.role
== R_PRIMARY
&&
1050 (mdev
->ed_uuid
& ~((u64
)1)) != (nbc
->md
.uuid
[UI_CURRENT
] & ~((u64
)1))) {
1051 dev_err(DEV
, "Can only attach to data with current UUID=%016llX\n",
1052 (unsigned long long)mdev
->ed_uuid
);
1053 retcode
= ERR_DATA_NOT_CURRENT
;
1054 goto force_diskless_dec
;
1057 /* Since we are diskless, fix the activity log first... */
1058 if (drbd_check_al_size(mdev
)) {
1059 retcode
= ERR_NOMEM
;
1060 goto force_diskless_dec
;
1063 /* Prevent shrinking of consistent devices ! */
1064 if (drbd_md_test_flag(nbc
, MDF_CONSISTENT
) &&
1065 drbd_new_dev_size(mdev
, nbc
, 0) < nbc
->md
.la_size_sect
) {
1066 dev_warn(DEV
, "refusing to truncate a consistent device\n");
1067 retcode
= ERR_DISK_TO_SMALL
;
1068 goto force_diskless_dec
;
1071 if (!drbd_al_read_log(mdev
, nbc
)) {
1072 retcode
= ERR_IO_MD_DISK
;
1073 goto force_diskless_dec
;
1076 /* Reset the "barriers don't work" bits here, then force meta data to
1077 * be written, to ensure we determine if barriers are supported. */
1078 if (nbc
->dc
.no_md_flush
)
1079 set_bit(MD_NO_FUA
, &mdev
->flags
);
1081 clear_bit(MD_NO_FUA
, &mdev
->flags
);
1083 /* Point of no return reached.
1084 * Devices and memory are no longer released by error cleanup below.
1085 * now mdev takes over responsibility, and the state engine should
1086 * clean it up somewhere. */
1087 D_ASSERT(mdev
->ldev
== NULL
);
1089 mdev
->resync
= resync_lru
;
1093 mdev
->write_ordering
= WO_bdev_flush
;
1094 drbd_bump_write_ordering(mdev
, WO_bdev_flush
);
1096 if (drbd_md_test_flag(mdev
->ldev
, MDF_CRASHED_PRIMARY
))
1097 set_bit(CRASHED_PRIMARY
, &mdev
->flags
);
1099 clear_bit(CRASHED_PRIMARY
, &mdev
->flags
);
1101 if (drbd_md_test_flag(mdev
->ldev
, MDF_PRIMARY_IND
) &&
1102 !(mdev
->state
.role
== R_PRIMARY
&& mdev
->state
.susp_nod
)) {
1103 set_bit(CRASHED_PRIMARY
, &mdev
->flags
);
1112 max_bio_size
= DRBD_MAX_BIO_SIZE
;
1113 if (mdev
->state
.conn
== C_CONNECTED
) {
1114 /* We are Primary, Connected, and now attach a new local
1115 * backing store. We must not increase the user visible maximum
1116 * bio size on this device to something the peer may not be
1117 * able to handle. */
1118 if (mdev
->agreed_pro_version
< 94)
1119 max_bio_size
= queue_max_hw_sectors(mdev
->rq_queue
) << 9;
1120 else if (mdev
->agreed_pro_version
== 94)
1121 max_bio_size
= DRBD_MAX_SIZE_H80_PACKET
;
1122 /* else: drbd 8.3.9 and later, stay with default */
1125 drbd_setup_queue_param(mdev
, max_bio_size
);
1127 /* If I am currently not R_PRIMARY,
1128 * but meta data primary indicator is set,
1129 * I just now recover from a hard crash,
1130 * and have been R_PRIMARY before that crash.
1132 * Now, if I had no connection before that crash
1133 * (have been degraded R_PRIMARY), chances are that
1134 * I won't find my peer now either.
1136 * In that case, and _only_ in that case,
1137 * we use the degr-wfc-timeout instead of the default,
1138 * so we can automatically recover from a crash of a
1139 * degraded but active "cluster" after a certain timeout.
1141 clear_bit(USE_DEGR_WFC_T
, &mdev
->flags
);
1142 if (mdev
->state
.role
!= R_PRIMARY
&&
1143 drbd_md_test_flag(mdev
->ldev
, MDF_PRIMARY_IND
) &&
1144 !drbd_md_test_flag(mdev
->ldev
, MDF_CONNECTED_IND
))
1145 set_bit(USE_DEGR_WFC_T
, &mdev
->flags
);
1147 dd
= drbd_determin_dev_size(mdev
, 0);
1148 if (dd
== dev_size_error
) {
1149 retcode
= ERR_NOMEM_BITMAP
;
1150 goto force_diskless_dec
;
1151 } else if (dd
== grew
)
1152 set_bit(RESYNC_AFTER_NEG
, &mdev
->flags
);
1154 if (drbd_md_test_flag(mdev
->ldev
, MDF_FULL_SYNC
)) {
1155 dev_info(DEV
, "Assuming that all blocks are out of sync "
1156 "(aka FullSync)\n");
1157 if (drbd_bitmap_io(mdev
, &drbd_bmio_set_n_write
, "set_n_write from attaching")) {
1158 retcode
= ERR_IO_MD_DISK
;
1159 goto force_diskless_dec
;
1162 if (drbd_bitmap_io(mdev
, &drbd_bm_read
, "read from attaching") < 0) {
1163 retcode
= ERR_IO_MD_DISK
;
1164 goto force_diskless_dec
;
1168 if (cp_discovered
) {
1169 drbd_al_apply_to_bm(mdev
);
1170 drbd_al_to_on_disk_bm(mdev
);
1173 if (_drbd_bm_total_weight(mdev
) == drbd_bm_bits(mdev
))
1174 drbd_suspend_al(mdev
); /* IO is still suspended here... */
1176 spin_lock_irq(&mdev
->req_lock
);
1179 /* If MDF_CONSISTENT is not set go into inconsistent state,
1180 otherwise investigate MDF_WasUpToDate...
1181 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1182 otherwise into D_CONSISTENT state.
1184 if (drbd_md_test_flag(mdev
->ldev
, MDF_CONSISTENT
)) {
1185 if (drbd_md_test_flag(mdev
->ldev
, MDF_WAS_UP_TO_DATE
))
1186 ns
.disk
= D_CONSISTENT
;
1188 ns
.disk
= D_OUTDATED
;
1190 ns
.disk
= D_INCONSISTENT
;
1193 if (drbd_md_test_flag(mdev
->ldev
, MDF_PEER_OUT_DATED
))
1194 ns
.pdsk
= D_OUTDATED
;
1196 if ( ns
.disk
== D_CONSISTENT
&&
1197 (ns
.pdsk
== D_OUTDATED
|| mdev
->ldev
->dc
.fencing
== FP_DONT_CARE
))
1198 ns
.disk
= D_UP_TO_DATE
;
1200 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1201 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1202 this point, because drbd_request_state() modifies these
1205 /* In case we are C_CONNECTED postpone any decision on the new disk
1206 state after the negotiation phase. */
1207 if (mdev
->state
.conn
== C_CONNECTED
) {
1208 mdev
->new_state_tmp
.i
= ns
.i
;
1210 ns
.disk
= D_NEGOTIATING
;
1212 /* We expect to receive up-to-date UUIDs soon.
1213 To avoid a race in receive_state, free p_uuid while
1214 holding req_lock. I.e. atomic with the state change */
1215 kfree(mdev
->p_uuid
);
1216 mdev
->p_uuid
= NULL
;
1219 rv
= _drbd_set_state(mdev
, ns
, CS_VERBOSE
, NULL
);
1221 spin_unlock_irq(&mdev
->req_lock
);
1223 if (rv
< SS_SUCCESS
)
1224 goto force_diskless_dec
;
1226 if (mdev
->state
.role
== R_PRIMARY
)
1227 mdev
->ldev
->md
.uuid
[UI_CURRENT
] |= (u64
)1;
1229 mdev
->ldev
->md
.uuid
[UI_CURRENT
] &= ~(u64
)1;
1231 drbd_md_mark_dirty(mdev
);
1234 kobject_uevent(&disk_to_dev(mdev
->vdisk
)->kobj
, KOBJ_CHANGE
);
1236 reply
->ret_code
= retcode
;
1237 drbd_reconfig_done(mdev
);
1243 drbd_force_state(mdev
, NS(disk
, D_FAILED
));
1247 if (nbc
->backing_bdev
)
1248 blkdev_put(nbc
->backing_bdev
,
1249 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
);
1251 blkdev_put(nbc
->md_bdev
,
1252 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
);
1255 lc_destroy(resync_lru
);
1257 reply
->ret_code
= retcode
;
1258 drbd_reconfig_done(mdev
);
1262 /* Detaching the disk is a process in multiple stages. First we need to lock
1263 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1264 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1265 * internal references as well.
1266 * Only then we have finally detached. */
1267 static int drbd_nl_detach(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1268 struct drbd_nl_cfg_reply
*reply
)
1270 drbd_suspend_io(mdev
); /* so no-one is stuck in drbd_al_begin_io */
1271 reply
->ret_code
= drbd_request_state(mdev
, NS(disk
, D_DISKLESS
));
1272 if (mdev
->state
.disk
== D_DISKLESS
)
1273 wait_event(mdev
->misc_wait
, !atomic_read(&mdev
->local_cnt
));
1274 drbd_resume_io(mdev
);
1278 static int drbd_nl_net_conf(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1279 struct drbd_nl_cfg_reply
*reply
)
1282 enum drbd_ret_code retcode
;
1283 struct net_conf
*new_conf
= NULL
;
1284 struct crypto_hash
*tfm
= NULL
;
1285 struct crypto_hash
*integrity_w_tfm
= NULL
;
1286 struct crypto_hash
*integrity_r_tfm
= NULL
;
1287 struct hlist_head
*new_tl_hash
= NULL
;
1288 struct hlist_head
*new_ee_hash
= NULL
;
1289 struct drbd_conf
*odev
;
1290 char hmac_name
[CRYPTO_MAX_ALG_NAME
];
1291 void *int_dig_out
= NULL
;
1292 void *int_dig_in
= NULL
;
1293 void *int_dig_vv
= NULL
;
1294 struct sockaddr
*new_my_addr
, *new_peer_addr
, *taken_addr
;
1296 drbd_reconfig_start(mdev
);
1298 if (mdev
->state
.conn
> C_STANDALONE
) {
1299 retcode
= ERR_NET_CONFIGURED
;
1303 /* allocation not in the IO path, cqueue thread context */
1304 new_conf
= kzalloc(sizeof(struct net_conf
), GFP_KERNEL
);
1306 retcode
= ERR_NOMEM
;
1310 new_conf
->timeout
= DRBD_TIMEOUT_DEF
;
1311 new_conf
->try_connect_int
= DRBD_CONNECT_INT_DEF
;
1312 new_conf
->ping_int
= DRBD_PING_INT_DEF
;
1313 new_conf
->max_epoch_size
= DRBD_MAX_EPOCH_SIZE_DEF
;
1314 new_conf
->max_buffers
= DRBD_MAX_BUFFERS_DEF
;
1315 new_conf
->unplug_watermark
= DRBD_UNPLUG_WATERMARK_DEF
;
1316 new_conf
->sndbuf_size
= DRBD_SNDBUF_SIZE_DEF
;
1317 new_conf
->rcvbuf_size
= DRBD_RCVBUF_SIZE_DEF
;
1318 new_conf
->ko_count
= DRBD_KO_COUNT_DEF
;
1319 new_conf
->after_sb_0p
= DRBD_AFTER_SB_0P_DEF
;
1320 new_conf
->after_sb_1p
= DRBD_AFTER_SB_1P_DEF
;
1321 new_conf
->after_sb_2p
= DRBD_AFTER_SB_2P_DEF
;
1322 new_conf
->want_lose
= 0;
1323 new_conf
->two_primaries
= 0;
1324 new_conf
->wire_protocol
= DRBD_PROT_C
;
1325 new_conf
->ping_timeo
= DRBD_PING_TIMEO_DEF
;
1326 new_conf
->rr_conflict
= DRBD_RR_CONFLICT_DEF
;
1327 new_conf
->on_congestion
= DRBD_ON_CONGESTION_DEF
;
1328 new_conf
->cong_extents
= DRBD_CONG_EXTENTS_DEF
;
1330 if (!net_conf_from_tags(mdev
, nlp
->tag_list
, new_conf
)) {
1331 retcode
= ERR_MANDATORY_TAG
;
1335 if (new_conf
->two_primaries
1336 && (new_conf
->wire_protocol
!= DRBD_PROT_C
)) {
1337 retcode
= ERR_NOT_PROTO_C
;
1341 if (get_ldev(mdev
)) {
1342 enum drbd_fencing_p fp
= mdev
->ldev
->dc
.fencing
;
1344 if (new_conf
->wire_protocol
== DRBD_PROT_A
&& fp
== FP_STONITH
) {
1345 retcode
= ERR_STONITH_AND_PROT_A
;
1350 if (new_conf
->on_congestion
!= OC_BLOCK
&& new_conf
->wire_protocol
!= DRBD_PROT_A
) {
1351 retcode
= ERR_CONG_NOT_PROTO_A
;
1355 if (mdev
->state
.role
== R_PRIMARY
&& new_conf
->want_lose
) {
1356 retcode
= ERR_DISCARD
;
1362 new_my_addr
= (struct sockaddr
*)&new_conf
->my_addr
;
1363 new_peer_addr
= (struct sockaddr
*)&new_conf
->peer_addr
;
1364 for (i
= 0; i
< minor_count
; i
++) {
1365 odev
= minor_to_mdev(i
);
1366 if (!odev
|| odev
== mdev
)
1368 if (get_net_conf(odev
)) {
1369 taken_addr
= (struct sockaddr
*)&odev
->net_conf
->my_addr
;
1370 if (new_conf
->my_addr_len
== odev
->net_conf
->my_addr_len
&&
1371 !memcmp(new_my_addr
, taken_addr
, new_conf
->my_addr_len
))
1372 retcode
= ERR_LOCAL_ADDR
;
1374 taken_addr
= (struct sockaddr
*)&odev
->net_conf
->peer_addr
;
1375 if (new_conf
->peer_addr_len
== odev
->net_conf
->peer_addr_len
&&
1376 !memcmp(new_peer_addr
, taken_addr
, new_conf
->peer_addr_len
))
1377 retcode
= ERR_PEER_ADDR
;
1380 if (retcode
!= NO_ERROR
)
1385 if (new_conf
->cram_hmac_alg
[0] != 0) {
1386 snprintf(hmac_name
, CRYPTO_MAX_ALG_NAME
, "hmac(%s)",
1387 new_conf
->cram_hmac_alg
);
1388 tfm
= crypto_alloc_hash(hmac_name
, 0, CRYPTO_ALG_ASYNC
);
1391 retcode
= ERR_AUTH_ALG
;
1395 if (!drbd_crypto_is_hash(crypto_hash_tfm(tfm
))) {
1396 retcode
= ERR_AUTH_ALG_ND
;
1401 if (new_conf
->integrity_alg
[0]) {
1402 integrity_w_tfm
= crypto_alloc_hash(new_conf
->integrity_alg
, 0, CRYPTO_ALG_ASYNC
);
1403 if (IS_ERR(integrity_w_tfm
)) {
1404 integrity_w_tfm
= NULL
;
1405 retcode
=ERR_INTEGRITY_ALG
;
1409 if (!drbd_crypto_is_hash(crypto_hash_tfm(integrity_w_tfm
))) {
1410 retcode
=ERR_INTEGRITY_ALG_ND
;
1414 integrity_r_tfm
= crypto_alloc_hash(new_conf
->integrity_alg
, 0, CRYPTO_ALG_ASYNC
);
1415 if (IS_ERR(integrity_r_tfm
)) {
1416 integrity_r_tfm
= NULL
;
1417 retcode
=ERR_INTEGRITY_ALG
;
1422 ns
= new_conf
->max_epoch_size
/8;
1423 if (mdev
->tl_hash_s
!= ns
) {
1424 new_tl_hash
= kzalloc(ns
*sizeof(void *), GFP_KERNEL
);
1426 retcode
= ERR_NOMEM
;
1431 ns
= new_conf
->max_buffers
/8;
1432 if (new_conf
->two_primaries
&& (mdev
->ee_hash_s
!= ns
)) {
1433 new_ee_hash
= kzalloc(ns
*sizeof(void *), GFP_KERNEL
);
1435 retcode
= ERR_NOMEM
;
1440 ((char *)new_conf
->shared_secret
)[SHARED_SECRET_MAX
-1] = 0;
1442 if (integrity_w_tfm
) {
1443 i
= crypto_hash_digestsize(integrity_w_tfm
);
1444 int_dig_out
= kmalloc(i
, GFP_KERNEL
);
1446 retcode
= ERR_NOMEM
;
1449 int_dig_in
= kmalloc(i
, GFP_KERNEL
);
1451 retcode
= ERR_NOMEM
;
1454 int_dig_vv
= kmalloc(i
, GFP_KERNEL
);
1456 retcode
= ERR_NOMEM
;
1461 if (!mdev
->bitmap
) {
1462 if(drbd_bm_init(mdev
)) {
1463 retcode
= ERR_NOMEM
;
1468 drbd_flush_workqueue(mdev
);
1469 spin_lock_irq(&mdev
->req_lock
);
1470 if (mdev
->net_conf
!= NULL
) {
1471 retcode
= ERR_NET_CONFIGURED
;
1472 spin_unlock_irq(&mdev
->req_lock
);
1475 mdev
->net_conf
= new_conf
;
1481 kfree(mdev
->tl_hash
);
1482 mdev
->tl_hash_s
= mdev
->net_conf
->max_epoch_size
/8;
1483 mdev
->tl_hash
= new_tl_hash
;
1487 kfree(mdev
->ee_hash
);
1488 mdev
->ee_hash_s
= mdev
->net_conf
->max_buffers
/8;
1489 mdev
->ee_hash
= new_ee_hash
;
1492 crypto_free_hash(mdev
->cram_hmac_tfm
);
1493 mdev
->cram_hmac_tfm
= tfm
;
1495 crypto_free_hash(mdev
->integrity_w_tfm
);
1496 mdev
->integrity_w_tfm
= integrity_w_tfm
;
1498 crypto_free_hash(mdev
->integrity_r_tfm
);
1499 mdev
->integrity_r_tfm
= integrity_r_tfm
;
1501 kfree(mdev
->int_dig_out
);
1502 kfree(mdev
->int_dig_in
);
1503 kfree(mdev
->int_dig_vv
);
1504 mdev
->int_dig_out
=int_dig_out
;
1505 mdev
->int_dig_in
=int_dig_in
;
1506 mdev
->int_dig_vv
=int_dig_vv
;
1507 retcode
= _drbd_set_state(_NS(mdev
, conn
, C_UNCONNECTED
), CS_VERBOSE
, NULL
);
1508 spin_unlock_irq(&mdev
->req_lock
);
1510 kobject_uevent(&disk_to_dev(mdev
->vdisk
)->kobj
, KOBJ_CHANGE
);
1511 reply
->ret_code
= retcode
;
1512 drbd_reconfig_done(mdev
);
1519 crypto_free_hash(tfm
);
1520 crypto_free_hash(integrity_w_tfm
);
1521 crypto_free_hash(integrity_r_tfm
);
1526 reply
->ret_code
= retcode
;
1527 drbd_reconfig_done(mdev
);
1531 static int drbd_nl_disconnect(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1532 struct drbd_nl_cfg_reply
*reply
)
1535 struct disconnect dc
;
1537 memset(&dc
, 0, sizeof(struct disconnect
));
1538 if (!disconnect_from_tags(mdev
, nlp
->tag_list
, &dc
)) {
1539 retcode
= ERR_MANDATORY_TAG
;
1544 spin_lock_irq(&mdev
->req_lock
);
1545 if (mdev
->state
.conn
>= C_WF_CONNECTION
)
1546 _drbd_set_state(_NS(mdev
, conn
, C_DISCONNECTING
), CS_HARD
, NULL
);
1547 spin_unlock_irq(&mdev
->req_lock
);
1551 retcode
= _drbd_request_state(mdev
, NS(conn
, C_DISCONNECTING
), CS_ORDERED
);
1553 if (retcode
== SS_NOTHING_TO_DO
)
1555 else if (retcode
== SS_ALREADY_STANDALONE
)
1557 else if (retcode
== SS_PRIMARY_NOP
) {
1558 /* Our statche checking code wants to see the peer outdated. */
1559 retcode
= drbd_request_state(mdev
, NS2(conn
, C_DISCONNECTING
,
1561 } else if (retcode
== SS_CW_FAILED_BY_PEER
) {
1562 /* The peer probably wants to see us outdated. */
1563 retcode
= _drbd_request_state(mdev
, NS2(conn
, C_DISCONNECTING
,
1566 if (retcode
== SS_IS_DISKLESS
|| retcode
== SS_LOWER_THAN_OUTDATED
) {
1567 drbd_force_state(mdev
, NS(conn
, C_DISCONNECTING
));
1568 retcode
= SS_SUCCESS
;
1572 if (retcode
< SS_SUCCESS
)
1575 if (wait_event_interruptible(mdev
->state_wait
,
1576 mdev
->state
.conn
!= C_DISCONNECTING
)) {
1577 /* Do not test for mdev->state.conn == C_STANDALONE, since
1578 someone else might connect us in the mean time! */
1587 reply
->ret_code
= retcode
;
1591 void resync_after_online_grow(struct drbd_conf
*mdev
)
1593 int iass
; /* I am sync source */
1595 dev_info(DEV
, "Resync of new storage after online grow\n");
1596 if (mdev
->state
.role
!= mdev
->state
.peer
)
1597 iass
= (mdev
->state
.role
== R_PRIMARY
);
1599 iass
= test_bit(DISCARD_CONCURRENT
, &mdev
->flags
);
1602 drbd_start_resync(mdev
, C_SYNC_SOURCE
);
1604 _drbd_request_state(mdev
, NS(conn
, C_WF_SYNC_UUID
), CS_VERBOSE
+ CS_SERIALIZE
);
1607 static int drbd_nl_resize(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1608 struct drbd_nl_cfg_reply
*reply
)
1611 int retcode
= NO_ERROR
;
1612 enum determine_dev_size dd
;
1613 enum dds_flags ddsf
;
1615 memset(&rs
, 0, sizeof(struct resize
));
1616 if (!resize_from_tags(mdev
, nlp
->tag_list
, &rs
)) {
1617 retcode
= ERR_MANDATORY_TAG
;
1621 if (mdev
->state
.conn
> C_CONNECTED
) {
1622 retcode
= ERR_RESIZE_RESYNC
;
1626 if (mdev
->state
.role
== R_SECONDARY
&&
1627 mdev
->state
.peer
== R_SECONDARY
) {
1628 retcode
= ERR_NO_PRIMARY
;
1632 if (!get_ldev(mdev
)) {
1633 retcode
= ERR_NO_DISK
;
1637 if (rs
.no_resync
&& mdev
->agreed_pro_version
< 93) {
1638 retcode
= ERR_NEED_APV_93
;
1642 if (mdev
->ldev
->known_size
!= drbd_get_capacity(mdev
->ldev
->backing_bdev
))
1643 mdev
->ldev
->known_size
= drbd_get_capacity(mdev
->ldev
->backing_bdev
);
1645 mdev
->ldev
->dc
.disk_size
= (sector_t
)rs
.resize_size
;
1646 ddsf
= (rs
.resize_force
? DDSF_FORCED
: 0) | (rs
.no_resync
? DDSF_NO_RESYNC
: 0);
1647 dd
= drbd_determin_dev_size(mdev
, ddsf
);
1650 if (dd
== dev_size_error
) {
1651 retcode
= ERR_NOMEM_BITMAP
;
1655 if (mdev
->state
.conn
== C_CONNECTED
) {
1657 set_bit(RESIZE_PENDING
, &mdev
->flags
);
1659 drbd_send_uuids(mdev
);
1660 drbd_send_sizes(mdev
, 1, ddsf
);
1664 reply
->ret_code
= retcode
;
1668 static int drbd_nl_syncer_conf(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1669 struct drbd_nl_cfg_reply
*reply
)
1671 int retcode
= NO_ERROR
;
1673 int ovr
; /* online verify running */
1674 int rsr
; /* re-sync running */
1675 struct crypto_hash
*verify_tfm
= NULL
;
1676 struct crypto_hash
*csums_tfm
= NULL
;
1677 struct syncer_conf sc
;
1678 cpumask_var_t new_cpu_mask
;
1679 int *rs_plan_s
= NULL
;
1682 if (!zalloc_cpumask_var(&new_cpu_mask
, GFP_KERNEL
)) {
1683 retcode
= ERR_NOMEM
;
1687 if (nlp
->flags
& DRBD_NL_SET_DEFAULTS
) {
1688 memset(&sc
, 0, sizeof(struct syncer_conf
));
1689 sc
.rate
= DRBD_RATE_DEF
;
1690 sc
.after
= DRBD_AFTER_DEF
;
1691 sc
.al_extents
= DRBD_AL_EXTENTS_DEF
;
1692 sc
.on_no_data
= DRBD_ON_NO_DATA_DEF
;
1693 sc
.c_plan_ahead
= DRBD_C_PLAN_AHEAD_DEF
;
1694 sc
.c_delay_target
= DRBD_C_DELAY_TARGET_DEF
;
1695 sc
.c_fill_target
= DRBD_C_FILL_TARGET_DEF
;
1696 sc
.c_max_rate
= DRBD_C_MAX_RATE_DEF
;
1697 sc
.c_min_rate
= DRBD_C_MIN_RATE_DEF
;
1699 memcpy(&sc
, &mdev
->sync_conf
, sizeof(struct syncer_conf
));
1701 if (!syncer_conf_from_tags(mdev
, nlp
->tag_list
, &sc
)) {
1702 retcode
= ERR_MANDATORY_TAG
;
1706 /* re-sync running */
1707 rsr
= ( mdev
->state
.conn
== C_SYNC_SOURCE
||
1708 mdev
->state
.conn
== C_SYNC_TARGET
||
1709 mdev
->state
.conn
== C_PAUSED_SYNC_S
||
1710 mdev
->state
.conn
== C_PAUSED_SYNC_T
);
1712 if (rsr
&& strcmp(sc
.csums_alg
, mdev
->sync_conf
.csums_alg
)) {
1713 retcode
= ERR_CSUMS_RESYNC_RUNNING
;
1717 if (!rsr
&& sc
.csums_alg
[0]) {
1718 csums_tfm
= crypto_alloc_hash(sc
.csums_alg
, 0, CRYPTO_ALG_ASYNC
);
1719 if (IS_ERR(csums_tfm
)) {
1721 retcode
= ERR_CSUMS_ALG
;
1725 if (!drbd_crypto_is_hash(crypto_hash_tfm(csums_tfm
))) {
1726 retcode
= ERR_CSUMS_ALG_ND
;
1731 /* online verify running */
1732 ovr
= (mdev
->state
.conn
== C_VERIFY_S
|| mdev
->state
.conn
== C_VERIFY_T
);
1735 if (strcmp(sc
.verify_alg
, mdev
->sync_conf
.verify_alg
)) {
1736 retcode
= ERR_VERIFY_RUNNING
;
1741 if (!ovr
&& sc
.verify_alg
[0]) {
1742 verify_tfm
= crypto_alloc_hash(sc
.verify_alg
, 0, CRYPTO_ALG_ASYNC
);
1743 if (IS_ERR(verify_tfm
)) {
1745 retcode
= ERR_VERIFY_ALG
;
1749 if (!drbd_crypto_is_hash(crypto_hash_tfm(verify_tfm
))) {
1750 retcode
= ERR_VERIFY_ALG_ND
;
1755 /* silently ignore cpu mask on UP kernel */
1756 if (nr_cpu_ids
> 1 && sc
.cpu_mask
[0] != 0) {
1757 err
= __bitmap_parse(sc
.cpu_mask
, 32, 0,
1758 cpumask_bits(new_cpu_mask
), nr_cpu_ids
);
1760 dev_warn(DEV
, "__bitmap_parse() failed with %d\n", err
);
1761 retcode
= ERR_CPU_MASK_PARSE
;
1766 ERR_IF (sc
.rate
< 1) sc
.rate
= 1;
1767 ERR_IF (sc
.al_extents
< 7) sc
.al_extents
= 127; /* arbitrary minimum */
1768 #define AL_MAX ((MD_AL_MAX_SIZE-1) * AL_EXTENTS_PT)
1769 if (sc
.al_extents
> AL_MAX
) {
1770 dev_err(DEV
, "sc.al_extents > %d\n", AL_MAX
);
1771 sc
.al_extents
= AL_MAX
;
1775 /* to avoid spurious errors when configuring minors before configuring
1776 * the minors they depend on: if necessary, first create the minor we
1779 ensure_mdev(sc
.after
, 1);
1781 /* most sanity checks done, try to assign the new sync-after
1782 * dependency. need to hold the global lock in there,
1783 * to avoid a race in the dependency loop check. */
1784 retcode
= drbd_alter_sa(mdev
, sc
.after
);
1785 if (retcode
!= NO_ERROR
)
1788 fifo_size
= (sc
.c_plan_ahead
* 10 * SLEEP_TIME
) / HZ
;
1789 if (fifo_size
!= mdev
->rs_plan_s
.size
&& fifo_size
> 0) {
1790 rs_plan_s
= kzalloc(sizeof(int) * fifo_size
, GFP_KERNEL
);
1792 dev_err(DEV
, "kmalloc of fifo_buffer failed");
1793 retcode
= ERR_NOMEM
;
1798 /* ok, assign the rest of it as well.
1799 * lock against receive_SyncParam() */
1800 spin_lock(&mdev
->peer_seq_lock
);
1801 mdev
->sync_conf
= sc
;
1804 crypto_free_hash(mdev
->csums_tfm
);
1805 mdev
->csums_tfm
= csums_tfm
;
1810 crypto_free_hash(mdev
->verify_tfm
);
1811 mdev
->verify_tfm
= verify_tfm
;
1815 if (fifo_size
!= mdev
->rs_plan_s
.size
) {
1816 kfree(mdev
->rs_plan_s
.values
);
1817 mdev
->rs_plan_s
.values
= rs_plan_s
;
1818 mdev
->rs_plan_s
.size
= fifo_size
;
1819 mdev
->rs_planed
= 0;
1823 spin_unlock(&mdev
->peer_seq_lock
);
1825 if (get_ldev(mdev
)) {
1826 wait_event(mdev
->al_wait
, lc_try_lock(mdev
->act_log
));
1827 drbd_al_shrink(mdev
);
1828 err
= drbd_check_al_size(mdev
);
1829 lc_unlock(mdev
->act_log
);
1830 wake_up(&mdev
->al_wait
);
1836 retcode
= ERR_NOMEM
;
1841 if (mdev
->state
.conn
>= C_CONNECTED
)
1842 drbd_send_sync_param(mdev
, &sc
);
1844 if (!cpumask_equal(mdev
->cpu_mask
, new_cpu_mask
)) {
1845 cpumask_copy(mdev
->cpu_mask
, new_cpu_mask
);
1846 drbd_calc_cpu_mask(mdev
);
1847 mdev
->receiver
.reset_cpu_mask
= 1;
1848 mdev
->asender
.reset_cpu_mask
= 1;
1849 mdev
->worker
.reset_cpu_mask
= 1;
1852 kobject_uevent(&disk_to_dev(mdev
->vdisk
)->kobj
, KOBJ_CHANGE
);
1855 free_cpumask_var(new_cpu_mask
);
1856 crypto_free_hash(csums_tfm
);
1857 crypto_free_hash(verify_tfm
);
1858 reply
->ret_code
= retcode
;
1862 static int drbd_nl_invalidate(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1863 struct drbd_nl_cfg_reply
*reply
)
1867 retcode
= _drbd_request_state(mdev
, NS(conn
, C_STARTING_SYNC_T
), CS_ORDERED
);
1869 if (retcode
< SS_SUCCESS
&& retcode
!= SS_NEED_CONNECTION
)
1870 retcode
= drbd_request_state(mdev
, NS(conn
, C_STARTING_SYNC_T
));
1872 while (retcode
== SS_NEED_CONNECTION
) {
1873 spin_lock_irq(&mdev
->req_lock
);
1874 if (mdev
->state
.conn
< C_CONNECTED
)
1875 retcode
= _drbd_set_state(_NS(mdev
, disk
, D_INCONSISTENT
), CS_VERBOSE
, NULL
);
1876 spin_unlock_irq(&mdev
->req_lock
);
1878 if (retcode
!= SS_NEED_CONNECTION
)
1881 retcode
= drbd_request_state(mdev
, NS(conn
, C_STARTING_SYNC_T
));
1884 reply
->ret_code
= retcode
;
1888 static int drbd_bmio_set_susp_al(struct drbd_conf
*mdev
)
1892 rv
= drbd_bmio_set_n_write(mdev
);
1893 drbd_suspend_al(mdev
);
1897 static int drbd_nl_invalidate_peer(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1898 struct drbd_nl_cfg_reply
*reply
)
1902 retcode
= _drbd_request_state(mdev
, NS(conn
, C_STARTING_SYNC_S
), CS_ORDERED
);
1904 if (retcode
< SS_SUCCESS
) {
1905 if (retcode
== SS_NEED_CONNECTION
&& mdev
->state
.role
== R_PRIMARY
) {
1906 /* The peer will get a resync upon connect anyways. Just make that
1907 into a full resync. */
1908 retcode
= drbd_request_state(mdev
, NS(pdsk
, D_INCONSISTENT
));
1909 if (retcode
>= SS_SUCCESS
) {
1910 /* open coded drbd_bitmap_io() */
1911 if (drbd_bitmap_io(mdev
, &drbd_bmio_set_susp_al
,
1912 "set_n_write from invalidate_peer"))
1913 retcode
= ERR_IO_MD_DISK
;
1916 retcode
= drbd_request_state(mdev
, NS(conn
, C_STARTING_SYNC_S
));
1919 reply
->ret_code
= retcode
;
1923 static int drbd_nl_pause_sync(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1924 struct drbd_nl_cfg_reply
*reply
)
1926 int retcode
= NO_ERROR
;
1928 if (drbd_request_state(mdev
, NS(user_isp
, 1)) == SS_NOTHING_TO_DO
)
1929 retcode
= ERR_PAUSE_IS_SET
;
1931 reply
->ret_code
= retcode
;
1935 static int drbd_nl_resume_sync(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1936 struct drbd_nl_cfg_reply
*reply
)
1938 int retcode
= NO_ERROR
;
1940 if (drbd_request_state(mdev
, NS(user_isp
, 0)) == SS_NOTHING_TO_DO
)
1941 retcode
= ERR_PAUSE_IS_CLEAR
;
1943 reply
->ret_code
= retcode
;
1947 static int drbd_nl_suspend_io(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1948 struct drbd_nl_cfg_reply
*reply
)
1950 reply
->ret_code
= drbd_request_state(mdev
, NS(susp
, 1));
1955 static int drbd_nl_resume_io(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1956 struct drbd_nl_cfg_reply
*reply
)
1958 if (test_bit(NEW_CUR_UUID
, &mdev
->flags
)) {
1959 drbd_uuid_new_current(mdev
);
1960 clear_bit(NEW_CUR_UUID
, &mdev
->flags
);
1962 drbd_suspend_io(mdev
);
1963 reply
->ret_code
= drbd_request_state(mdev
, NS3(susp
, 0, susp_nod
, 0, susp_fen
, 0));
1964 if (reply
->ret_code
== SS_SUCCESS
) {
1965 if (mdev
->state
.conn
< C_CONNECTED
)
1967 if (mdev
->state
.disk
== D_DISKLESS
|| mdev
->state
.disk
== D_FAILED
)
1968 tl_restart(mdev
, fail_frozen_disk_io
);
1970 drbd_resume_io(mdev
);
1975 static int drbd_nl_outdate(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1976 struct drbd_nl_cfg_reply
*reply
)
1978 reply
->ret_code
= drbd_request_state(mdev
, NS(disk
, D_OUTDATED
));
1982 static int drbd_nl_get_config(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
1983 struct drbd_nl_cfg_reply
*reply
)
1987 tl
= reply
->tag_list
;
1989 if (get_ldev(mdev
)) {
1990 tl
= disk_conf_to_tags(mdev
, &mdev
->ldev
->dc
, tl
);
1994 if (get_net_conf(mdev
)) {
1995 tl
= net_conf_to_tags(mdev
, mdev
->net_conf
, tl
);
1998 tl
= syncer_conf_to_tags(mdev
, &mdev
->sync_conf
, tl
);
2000 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2002 return (int)((char *)tl
- (char *)reply
->tag_list
);
2005 static int drbd_nl_get_state(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
2006 struct drbd_nl_cfg_reply
*reply
)
2008 unsigned short *tl
= reply
->tag_list
;
2009 union drbd_state s
= mdev
->state
;
2010 unsigned long rs_left
;
2013 tl
= get_state_to_tags(mdev
, (struct get_state
*)&s
, tl
);
2015 /* no local ref, no bitmap, no syncer progress. */
2016 if (s
.conn
>= C_SYNC_SOURCE
&& s
.conn
<= C_PAUSED_SYNC_T
) {
2017 if (get_ldev(mdev
)) {
2018 drbd_get_syncer_progress(mdev
, &rs_left
, &res
);
2019 tl
= tl_add_int(tl
, T_sync_progress
, &res
);
2023 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2025 return (int)((char *)tl
- (char *)reply
->tag_list
);
2028 static int drbd_nl_get_uuids(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
2029 struct drbd_nl_cfg_reply
*reply
)
2033 tl
= reply
->tag_list
;
2035 if (get_ldev(mdev
)) {
2036 tl
= tl_add_blob(tl
, T_uuids
, mdev
->ldev
->md
.uuid
, UI_SIZE
*sizeof(u64
));
2037 tl
= tl_add_int(tl
, T_uuids_flags
, &mdev
->ldev
->md
.flags
);
2040 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2042 return (int)((char *)tl
- (char *)reply
->tag_list
);
2046 * drbd_nl_get_timeout_flag() - Used by drbdsetup to find out which timeout value to use
2047 * @mdev: DRBD device.
2048 * @nlp: Netlink/connector packet from drbdsetup
2049 * @reply: Reply packet for drbdsetup
2051 static int drbd_nl_get_timeout_flag(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
2052 struct drbd_nl_cfg_reply
*reply
)
2057 tl
= reply
->tag_list
;
2059 rv
= mdev
->state
.pdsk
== D_OUTDATED
? UT_PEER_OUTDATED
:
2060 test_bit(USE_DEGR_WFC_T
, &mdev
->flags
) ? UT_DEGRADED
: UT_DEFAULT
;
2062 tl
= tl_add_blob(tl
, T_use_degraded
, &rv
, sizeof(rv
));
2063 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2065 return (int)((char *)tl
- (char *)reply
->tag_list
);
2068 static int drbd_nl_start_ov(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
2069 struct drbd_nl_cfg_reply
*reply
)
2071 /* default to resume from last known position, if possible */
2072 struct start_ov args
=
2073 { .start_sector
= mdev
->ov_start_sector
};
2075 if (!start_ov_from_tags(mdev
, nlp
->tag_list
, &args
)) {
2076 reply
->ret_code
= ERR_MANDATORY_TAG
;
2079 /* w_make_ov_request expects position to be aligned */
2080 mdev
->ov_start_sector
= args
.start_sector
& ~BM_SECT_PER_BIT
;
2081 reply
->ret_code
= drbd_request_state(mdev
,NS(conn
,C_VERIFY_S
));
2086 static int drbd_nl_new_c_uuid(struct drbd_conf
*mdev
, struct drbd_nl_cfg_req
*nlp
,
2087 struct drbd_nl_cfg_reply
*reply
)
2089 int retcode
= NO_ERROR
;
2090 int skip_initial_sync
= 0;
2093 struct new_c_uuid args
;
2095 memset(&args
, 0, sizeof(struct new_c_uuid
));
2096 if (!new_c_uuid_from_tags(mdev
, nlp
->tag_list
, &args
)) {
2097 reply
->ret_code
= ERR_MANDATORY_TAG
;
2101 mutex_lock(&mdev
->state_mutex
); /* Protects us against serialized state changes. */
2103 if (!get_ldev(mdev
)) {
2104 retcode
= ERR_NO_DISK
;
2108 /* this is "skip initial sync", assume to be clean */
2109 if (mdev
->state
.conn
== C_CONNECTED
&& mdev
->agreed_pro_version
>= 90 &&
2110 mdev
->ldev
->md
.uuid
[UI_CURRENT
] == UUID_JUST_CREATED
&& args
.clear_bm
) {
2111 dev_info(DEV
, "Preparing to skip initial sync\n");
2112 skip_initial_sync
= 1;
2113 } else if (mdev
->state
.conn
!= C_STANDALONE
) {
2114 retcode
= ERR_CONNECTED
;
2118 drbd_uuid_set(mdev
, UI_BITMAP
, 0); /* Rotate UI_BITMAP to History 1, etc... */
2119 drbd_uuid_new_current(mdev
); /* New current, previous to UI_BITMAP */
2121 if (args
.clear_bm
) {
2122 err
= drbd_bitmap_io(mdev
, &drbd_bmio_clear_n_write
, "clear_n_write from new_c_uuid");
2124 dev_err(DEV
, "Writing bitmap failed with %d\n",err
);
2125 retcode
= ERR_IO_MD_DISK
;
2127 if (skip_initial_sync
) {
2128 drbd_send_uuids_skip_initial_sync(mdev
);
2129 _drbd_uuid_set(mdev
, UI_BITMAP
, 0);
2130 spin_lock_irq(&mdev
->req_lock
);
2131 _drbd_set_state(_NS2(mdev
, disk
, D_UP_TO_DATE
, pdsk
, D_UP_TO_DATE
),
2133 spin_unlock_irq(&mdev
->req_lock
);
2141 mutex_unlock(&mdev
->state_mutex
);
2143 reply
->ret_code
= retcode
;
2147 struct cn_handler_struct
{
2148 int (*function
)(struct drbd_conf
*,
2149 struct drbd_nl_cfg_req
*,
2150 struct drbd_nl_cfg_reply
*);
2151 int reply_body_size
;
2154 static struct cn_handler_struct cnd_table
[] = {
2155 [ P_primary
] = { &drbd_nl_primary
, 0 },
2156 [ P_secondary
] = { &drbd_nl_secondary
, 0 },
2157 [ P_disk_conf
] = { &drbd_nl_disk_conf
, 0 },
2158 [ P_detach
] = { &drbd_nl_detach
, 0 },
2159 [ P_net_conf
] = { &drbd_nl_net_conf
, 0 },
2160 [ P_disconnect
] = { &drbd_nl_disconnect
, 0 },
2161 [ P_resize
] = { &drbd_nl_resize
, 0 },
2162 [ P_syncer_conf
] = { &drbd_nl_syncer_conf
, 0 },
2163 [ P_invalidate
] = { &drbd_nl_invalidate
, 0 },
2164 [ P_invalidate_peer
] = { &drbd_nl_invalidate_peer
, 0 },
2165 [ P_pause_sync
] = { &drbd_nl_pause_sync
, 0 },
2166 [ P_resume_sync
] = { &drbd_nl_resume_sync
, 0 },
2167 [ P_suspend_io
] = { &drbd_nl_suspend_io
, 0 },
2168 [ P_resume_io
] = { &drbd_nl_resume_io
, 0 },
2169 [ P_outdate
] = { &drbd_nl_outdate
, 0 },
2170 [ P_get_config
] = { &drbd_nl_get_config
,
2171 sizeof(struct syncer_conf_tag_len_struct
) +
2172 sizeof(struct disk_conf_tag_len_struct
) +
2173 sizeof(struct net_conf_tag_len_struct
) },
2174 [ P_get_state
] = { &drbd_nl_get_state
,
2175 sizeof(struct get_state_tag_len_struct
) +
2176 sizeof(struct sync_progress_tag_len_struct
) },
2177 [ P_get_uuids
] = { &drbd_nl_get_uuids
,
2178 sizeof(struct get_uuids_tag_len_struct
) },
2179 [ P_get_timeout_flag
] = { &drbd_nl_get_timeout_flag
,
2180 sizeof(struct get_timeout_flag_tag_len_struct
)},
2181 [ P_start_ov
] = { &drbd_nl_start_ov
, 0 },
2182 [ P_new_c_uuid
] = { &drbd_nl_new_c_uuid
, 0 },
2185 static void drbd_connector_callback(struct cn_msg
*req
, struct netlink_skb_parms
*nsp
)
2187 struct drbd_nl_cfg_req
*nlp
= (struct drbd_nl_cfg_req
*)req
->data
;
2188 struct cn_handler_struct
*cm
;
2189 struct cn_msg
*cn_reply
;
2190 struct drbd_nl_cfg_reply
*reply
;
2191 struct drbd_conf
*mdev
;
2193 int reply_size
= sizeof(struct cn_msg
)
2194 + sizeof(struct drbd_nl_cfg_reply
)
2195 + sizeof(short int);
2197 if (!try_module_get(THIS_MODULE
)) {
2198 printk(KERN_ERR
"drbd: try_module_get() failed!\n");
2202 if (!cap_raised(nsp
->eff_cap
, CAP_SYS_ADMIN
)) {
2207 mdev
= ensure_mdev(nlp
->drbd_minor
,
2208 (nlp
->flags
& DRBD_NL_CREATE_DEVICE
));
2210 retcode
= ERR_MINOR_INVALID
;
2214 if (nlp
->packet_type
>= P_nl_after_last_packet
||
2215 nlp
->packet_type
== P_return_code_only
) {
2216 retcode
= ERR_PACKET_NR
;
2220 cm
= cnd_table
+ nlp
->packet_type
;
2222 /* This may happen if packet number is 0: */
2223 if (cm
->function
== NULL
) {
2224 retcode
= ERR_PACKET_NR
;
2228 reply_size
+= cm
->reply_body_size
;
2230 /* allocation not in the IO path, cqueue thread context */
2231 cn_reply
= kzalloc(reply_size
, GFP_KERNEL
);
2233 retcode
= ERR_NOMEM
;
2236 reply
= (struct drbd_nl_cfg_reply
*) cn_reply
->data
;
2238 reply
->packet_type
=
2239 cm
->reply_body_size
? nlp
->packet_type
: P_return_code_only
;
2240 reply
->minor
= nlp
->drbd_minor
;
2241 reply
->ret_code
= NO_ERROR
; /* Might by modified by cm->function. */
2242 /* reply->tag_list; might be modified by cm->function. */
2244 rr
= cm
->function(mdev
, nlp
, reply
);
2246 cn_reply
->id
= req
->id
;
2247 cn_reply
->seq
= req
->seq
;
2248 cn_reply
->ack
= req
->ack
+ 1;
2249 cn_reply
->len
= sizeof(struct drbd_nl_cfg_reply
) + rr
;
2250 cn_reply
->flags
= 0;
2252 rr
= cn_netlink_send(cn_reply
, CN_IDX_DRBD
, GFP_KERNEL
);
2253 if (rr
&& rr
!= -ESRCH
)
2254 printk(KERN_INFO
"drbd: cn_netlink_send()=%d\n", rr
);
2257 module_put(THIS_MODULE
);
2260 drbd_nl_send_reply(req
, retcode
);
2261 module_put(THIS_MODULE
);
2264 static atomic_t drbd_nl_seq
= ATOMIC_INIT(2); /* two. */
2266 static unsigned short *
2267 __tl_add_blob(unsigned short *tl
, enum drbd_tags tag
, const void *data
,
2268 unsigned short len
, int nul_terminated
)
2270 unsigned short l
= tag_descriptions
[tag_number(tag
)].max_len
;
2271 len
= (len
< l
) ? len
: l
;
2272 put_unaligned(tag
, tl
++);
2273 put_unaligned(len
, tl
++);
2274 memcpy(tl
, data
, len
);
2275 tl
= (unsigned short*)((char*)tl
+ len
);
2277 *((char*)tl
- 1) = 0;
2281 static unsigned short *
2282 tl_add_blob(unsigned short *tl
, enum drbd_tags tag
, const void *data
, int len
)
2284 return __tl_add_blob(tl
, tag
, data
, len
, 0);
2287 static unsigned short *
2288 tl_add_str(unsigned short *tl
, enum drbd_tags tag
, const char *str
)
2290 return __tl_add_blob(tl
, tag
, str
, strlen(str
)+1, 0);
2293 static unsigned short *
2294 tl_add_int(unsigned short *tl
, enum drbd_tags tag
, const void *val
)
2296 put_unaligned(tag
, tl
++);
2297 switch(tag_type(tag
)) {
2299 put_unaligned(sizeof(int), tl
++);
2300 put_unaligned(*(int *)val
, (int *)tl
);
2301 tl
= (unsigned short*)((char*)tl
+sizeof(int));
2304 put_unaligned(sizeof(u64
), tl
++);
2305 put_unaligned(*(u64
*)val
, (u64
*)tl
);
2306 tl
= (unsigned short*)((char*)tl
+sizeof(u64
));
2309 /* someone did something stupid. */
2315 void drbd_bcast_state(struct drbd_conf
*mdev
, union drbd_state state
)
2317 char buffer
[sizeof(struct cn_msg
)+
2318 sizeof(struct drbd_nl_cfg_reply
)+
2319 sizeof(struct get_state_tag_len_struct
)+
2321 struct cn_msg
*cn_reply
= (struct cn_msg
*) buffer
;
2322 struct drbd_nl_cfg_reply
*reply
=
2323 (struct drbd_nl_cfg_reply
*)cn_reply
->data
;
2324 unsigned short *tl
= reply
->tag_list
;
2326 /* dev_warn(DEV, "drbd_bcast_state() got called\n"); */
2328 tl
= get_state_to_tags(mdev
, (struct get_state
*)&state
, tl
);
2330 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2332 cn_reply
->id
.idx
= CN_IDX_DRBD
;
2333 cn_reply
->id
.val
= CN_VAL_DRBD
;
2335 cn_reply
->seq
= atomic_add_return(1, &drbd_nl_seq
);
2336 cn_reply
->ack
= 0; /* not used here. */
2337 cn_reply
->len
= sizeof(struct drbd_nl_cfg_reply
) +
2338 (int)((char *)tl
- (char *)reply
->tag_list
);
2339 cn_reply
->flags
= 0;
2341 reply
->packet_type
= P_get_state
;
2342 reply
->minor
= mdev_to_minor(mdev
);
2343 reply
->ret_code
= NO_ERROR
;
2345 cn_netlink_send(cn_reply
, CN_IDX_DRBD
, GFP_NOIO
);
2348 void drbd_bcast_ev_helper(struct drbd_conf
*mdev
, char *helper_name
)
2350 char buffer
[sizeof(struct cn_msg
)+
2351 sizeof(struct drbd_nl_cfg_reply
)+
2352 sizeof(struct call_helper_tag_len_struct
)+
2354 struct cn_msg
*cn_reply
= (struct cn_msg
*) buffer
;
2355 struct drbd_nl_cfg_reply
*reply
=
2356 (struct drbd_nl_cfg_reply
*)cn_reply
->data
;
2357 unsigned short *tl
= reply
->tag_list
;
2359 /* dev_warn(DEV, "drbd_bcast_state() got called\n"); */
2361 tl
= tl_add_str(tl
, T_helper
, helper_name
);
2362 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2364 cn_reply
->id
.idx
= CN_IDX_DRBD
;
2365 cn_reply
->id
.val
= CN_VAL_DRBD
;
2367 cn_reply
->seq
= atomic_add_return(1, &drbd_nl_seq
);
2368 cn_reply
->ack
= 0; /* not used here. */
2369 cn_reply
->len
= sizeof(struct drbd_nl_cfg_reply
) +
2370 (int)((char *)tl
- (char *)reply
->tag_list
);
2371 cn_reply
->flags
= 0;
2373 reply
->packet_type
= P_call_helper
;
2374 reply
->minor
= mdev_to_minor(mdev
);
2375 reply
->ret_code
= NO_ERROR
;
2377 cn_netlink_send(cn_reply
, CN_IDX_DRBD
, GFP_NOIO
);
2380 void drbd_bcast_ee(struct drbd_conf
*mdev
,
2381 const char *reason
, const int dgs
,
2382 const char* seen_hash
, const char* calc_hash
,
2383 const struct drbd_epoch_entry
* e
)
2385 struct cn_msg
*cn_reply
;
2386 struct drbd_nl_cfg_reply
*reply
;
2393 if (!reason
|| !reason
[0])
2396 /* apparently we have to memcpy twice, first to prepare the data for the
2397 * struct cn_msg, then within cn_netlink_send from the cn_msg to the
2399 /* receiver thread context, which is not in the writeout path (of this node),
2400 * but may be in the writeout path of the _other_ node.
2401 * GFP_NOIO to avoid potential "distributed deadlock". */
2403 sizeof(struct cn_msg
)+
2404 sizeof(struct drbd_nl_cfg_reply
)+
2405 sizeof(struct dump_ee_tag_len_struct
)+
2410 dev_err(DEV
, "could not kmalloc buffer for drbd_bcast_ee, sector %llu, size %u\n",
2411 (unsigned long long)e
->sector
, e
->size
);
2415 reply
= (struct drbd_nl_cfg_reply
*)cn_reply
->data
;
2416 tl
= reply
->tag_list
;
2418 tl
= tl_add_str(tl
, T_dump_ee_reason
, reason
);
2419 tl
= tl_add_blob(tl
, T_seen_digest
, seen_hash
, dgs
);
2420 tl
= tl_add_blob(tl
, T_calc_digest
, calc_hash
, dgs
);
2421 tl
= tl_add_int(tl
, T_ee_sector
, &e
->sector
);
2422 tl
= tl_add_int(tl
, T_ee_block_id
, &e
->block_id
);
2424 /* dump the first 32k */
2425 len
= min_t(unsigned, e
->size
, 32 << 10);
2426 put_unaligned(T_ee_data
, tl
++);
2427 put_unaligned(len
, tl
++);
2430 page_chain_for_each(page
) {
2431 void *d
= kmap_atomic(page
, KM_USER0
);
2432 unsigned l
= min_t(unsigned, len
, PAGE_SIZE
);
2434 kunmap_atomic(d
, KM_USER0
);
2435 tl
= (unsigned short*)((char*)tl
+ l
);
2440 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2442 cn_reply
->id
.idx
= CN_IDX_DRBD
;
2443 cn_reply
->id
.val
= CN_VAL_DRBD
;
2445 cn_reply
->seq
= atomic_add_return(1,&drbd_nl_seq
);
2446 cn_reply
->ack
= 0; // not used here.
2447 cn_reply
->len
= sizeof(struct drbd_nl_cfg_reply
) +
2448 (int)((char*)tl
- (char*)reply
->tag_list
);
2449 cn_reply
->flags
= 0;
2451 reply
->packet_type
= P_dump_ee
;
2452 reply
->minor
= mdev_to_minor(mdev
);
2453 reply
->ret_code
= NO_ERROR
;
2455 cn_netlink_send(cn_reply
, CN_IDX_DRBD
, GFP_NOIO
);
2459 void drbd_bcast_sync_progress(struct drbd_conf
*mdev
)
2461 char buffer
[sizeof(struct cn_msg
)+
2462 sizeof(struct drbd_nl_cfg_reply
)+
2463 sizeof(struct sync_progress_tag_len_struct
)+
2465 struct cn_msg
*cn_reply
= (struct cn_msg
*) buffer
;
2466 struct drbd_nl_cfg_reply
*reply
=
2467 (struct drbd_nl_cfg_reply
*)cn_reply
->data
;
2468 unsigned short *tl
= reply
->tag_list
;
2469 unsigned long rs_left
;
2472 /* no local ref, no bitmap, no syncer progress, no broadcast. */
2473 if (!get_ldev(mdev
))
2475 drbd_get_syncer_progress(mdev
, &rs_left
, &res
);
2478 tl
= tl_add_int(tl
, T_sync_progress
, &res
);
2479 put_unaligned(TT_END
, tl
++); /* Close the tag list */
2481 cn_reply
->id
.idx
= CN_IDX_DRBD
;
2482 cn_reply
->id
.val
= CN_VAL_DRBD
;
2484 cn_reply
->seq
= atomic_add_return(1, &drbd_nl_seq
);
2485 cn_reply
->ack
= 0; /* not used here. */
2486 cn_reply
->len
= sizeof(struct drbd_nl_cfg_reply
) +
2487 (int)((char *)tl
- (char *)reply
->tag_list
);
2488 cn_reply
->flags
= 0;
2490 reply
->packet_type
= P_sync_progress
;
2491 reply
->minor
= mdev_to_minor(mdev
);
2492 reply
->ret_code
= NO_ERROR
;
2494 cn_netlink_send(cn_reply
, CN_IDX_DRBD
, GFP_NOIO
);
2497 int __init
drbd_nl_init(void)
2499 static struct cb_id cn_id_drbd
;
2502 cn_id_drbd
.val
= CN_VAL_DRBD
;
2504 cn_id_drbd
.idx
= cn_idx
;
2505 err
= cn_add_callback(&cn_id_drbd
, "cn_drbd", &drbd_connector_callback
);
2508 cn_idx
= (cn_idx
+ CN_IDX_STEP
);
2512 printk(KERN_ERR
"drbd: cn_drbd failed to register\n");
2519 void drbd_nl_cleanup(void)
2521 static struct cb_id cn_id_drbd
;
2523 cn_id_drbd
.idx
= cn_idx
;
2524 cn_id_drbd
.val
= CN_VAL_DRBD
;
2526 cn_del_callback(&cn_id_drbd
);
2529 void drbd_nl_send_reply(struct cn_msg
*req
, int ret_code
)
2531 char buffer
[sizeof(struct cn_msg
)+sizeof(struct drbd_nl_cfg_reply
)];
2532 struct cn_msg
*cn_reply
= (struct cn_msg
*) buffer
;
2533 struct drbd_nl_cfg_reply
*reply
=
2534 (struct drbd_nl_cfg_reply
*)cn_reply
->data
;
2537 memset(buffer
, 0, sizeof(buffer
));
2538 cn_reply
->id
= req
->id
;
2540 cn_reply
->seq
= req
->seq
;
2541 cn_reply
->ack
= req
->ack
+ 1;
2542 cn_reply
->len
= sizeof(struct drbd_nl_cfg_reply
);
2543 cn_reply
->flags
= 0;
2545 reply
->packet_type
= P_return_code_only
;
2546 reply
->minor
= ((struct drbd_nl_cfg_req
*)req
->data
)->drbd_minor
;
2547 reply
->ret_code
= ret_code
;
2549 rr
= cn_netlink_send(cn_reply
, CN_IDX_DRBD
, GFP_NOIO
);
2550 if (rr
&& rr
!= -ESRCH
)
2551 printk(KERN_INFO
"drbd: cn_netlink_send()=%d\n", rr
);