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/blkpg.h>
33 #include <linux/cpumask.h>
35 #include "drbd_protocol.h"
37 #include <asm/unaligned.h>
38 #include <linux/drbd_limits.h>
39 #include <linux/kthread.h>
41 #include <net/genetlink.h>
44 // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info);
45 // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info);
47 int drbd_adm_new_minor(struct sk_buff
*skb
, struct genl_info
*info
);
48 int drbd_adm_del_minor(struct sk_buff
*skb
, struct genl_info
*info
);
50 int drbd_adm_new_resource(struct sk_buff
*skb
, struct genl_info
*info
);
51 int drbd_adm_del_resource(struct sk_buff
*skb
, struct genl_info
*info
);
52 int drbd_adm_down(struct sk_buff
*skb
, struct genl_info
*info
);
54 int drbd_adm_set_role(struct sk_buff
*skb
, struct genl_info
*info
);
55 int drbd_adm_attach(struct sk_buff
*skb
, struct genl_info
*info
);
56 int drbd_adm_disk_opts(struct sk_buff
*skb
, struct genl_info
*info
);
57 int drbd_adm_detach(struct sk_buff
*skb
, struct genl_info
*info
);
58 int drbd_adm_connect(struct sk_buff
*skb
, struct genl_info
*info
);
59 int drbd_adm_net_opts(struct sk_buff
*skb
, struct genl_info
*info
);
60 int drbd_adm_resize(struct sk_buff
*skb
, struct genl_info
*info
);
61 int drbd_adm_start_ov(struct sk_buff
*skb
, struct genl_info
*info
);
62 int drbd_adm_new_c_uuid(struct sk_buff
*skb
, struct genl_info
*info
);
63 int drbd_adm_disconnect(struct sk_buff
*skb
, struct genl_info
*info
);
64 int drbd_adm_invalidate(struct sk_buff
*skb
, struct genl_info
*info
);
65 int drbd_adm_invalidate_peer(struct sk_buff
*skb
, struct genl_info
*info
);
66 int drbd_adm_pause_sync(struct sk_buff
*skb
, struct genl_info
*info
);
67 int drbd_adm_resume_sync(struct sk_buff
*skb
, struct genl_info
*info
);
68 int drbd_adm_suspend_io(struct sk_buff
*skb
, struct genl_info
*info
);
69 int drbd_adm_resume_io(struct sk_buff
*skb
, struct genl_info
*info
);
70 int drbd_adm_outdate(struct sk_buff
*skb
, struct genl_info
*info
);
71 int drbd_adm_resource_opts(struct sk_buff
*skb
, struct genl_info
*info
);
72 int drbd_adm_get_status(struct sk_buff
*skb
, struct genl_info
*info
);
73 int drbd_adm_get_timeout_type(struct sk_buff
*skb
, struct genl_info
*info
);
75 int drbd_adm_get_status_all(struct sk_buff
*skb
, struct netlink_callback
*cb
);
77 #include <linux/drbd_genl_api.h>
79 #include <linux/genl_magic_func.h>
81 /* used blkdev_get_by_path, to claim our meta data device(s) */
82 static char *drbd_m_holder
= "Hands off! this is DRBD's meta data device.";
84 static void drbd_adm_send_reply(struct sk_buff
*skb
, struct genl_info
*info
)
86 genlmsg_end(skb
, genlmsg_data(nlmsg_data(nlmsg_hdr(skb
))));
87 if (genlmsg_reply(skb
, info
))
88 printk(KERN_ERR
"drbd: error sending genl reply\n");
91 /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only
92 * reason it could fail was no space in skb, and there are 4k available. */
93 int drbd_msg_put_info(struct sk_buff
*skb
, const char *info
)
98 if (!info
|| !info
[0])
101 nla
= nla_nest_start(skb
, DRBD_NLA_CFG_REPLY
);
105 err
= nla_put_string(skb
, T_info_text
, info
);
107 nla_nest_cancel(skb
, nla
);
110 nla_nest_end(skb
, nla
);
114 /* This would be a good candidate for a "pre_doit" hook,
115 * and per-family private info->pointers.
116 * But we need to stay compatible with older kernels.
117 * If it returns successfully, adm_ctx members are valid.
119 * At this point, we still rely on the global genl_lock().
120 * If we want to avoid that, and allow "genl_family.parallel_ops", we may need
121 * to add additional synchronization against object destruction/modification.
123 #define DRBD_ADM_NEED_MINOR 1
124 #define DRBD_ADM_NEED_RESOURCE 2
125 #define DRBD_ADM_NEED_CONNECTION 4
126 static int drbd_adm_prepare(struct drbd_config_context
*adm_ctx
,
127 struct sk_buff
*skb
, struct genl_info
*info
, unsigned flags
)
129 struct drbd_genlmsghdr
*d_in
= info
->userhdr
;
130 const u8 cmd
= info
->genlhdr
->cmd
;
133 memset(adm_ctx
, 0, sizeof(*adm_ctx
));
135 /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */
136 if (cmd
!= DRBD_ADM_GET_STATUS
&& !capable(CAP_NET_ADMIN
))
139 adm_ctx
->reply_skb
= genlmsg_new(NLMSG_GOODSIZE
, GFP_KERNEL
);
140 if (!adm_ctx
->reply_skb
) {
145 adm_ctx
->reply_dh
= genlmsg_put_reply(adm_ctx
->reply_skb
,
146 info
, &drbd_genl_family
, 0, cmd
);
147 /* put of a few bytes into a fresh skb of >= 4k will always succeed.
149 if (!adm_ctx
->reply_dh
) {
154 adm_ctx
->reply_dh
->minor
= d_in
->minor
;
155 adm_ctx
->reply_dh
->ret_code
= NO_ERROR
;
157 adm_ctx
->volume
= VOLUME_UNSPECIFIED
;
158 if (info
->attrs
[DRBD_NLA_CFG_CONTEXT
]) {
160 /* parse and validate only */
161 err
= drbd_cfg_context_from_attrs(NULL
, info
);
165 /* It was present, and valid,
166 * copy it over to the reply skb. */
167 err
= nla_put_nohdr(adm_ctx
->reply_skb
,
168 info
->attrs
[DRBD_NLA_CFG_CONTEXT
]->nla_len
,
169 info
->attrs
[DRBD_NLA_CFG_CONTEXT
]);
173 /* and assign stuff to the adm_ctx */
174 nla
= nested_attr_tb
[__nla_type(T_ctx_volume
)];
176 adm_ctx
->volume
= nla_get_u32(nla
);
177 nla
= nested_attr_tb
[__nla_type(T_ctx_resource_name
)];
179 adm_ctx
->resource_name
= nla_data(nla
);
180 adm_ctx
->my_addr
= nested_attr_tb
[__nla_type(T_ctx_my_addr
)];
181 adm_ctx
->peer_addr
= nested_attr_tb
[__nla_type(T_ctx_peer_addr
)];
182 if ((adm_ctx
->my_addr
&&
183 nla_len(adm_ctx
->my_addr
) > sizeof(adm_ctx
->connection
->my_addr
)) ||
184 (adm_ctx
->peer_addr
&&
185 nla_len(adm_ctx
->peer_addr
) > sizeof(adm_ctx
->connection
->peer_addr
))) {
191 adm_ctx
->minor
= d_in
->minor
;
192 adm_ctx
->device
= minor_to_device(d_in
->minor
);
194 /* We are protected by the global genl_lock().
195 * But we may explicitly drop it/retake it in drbd_adm_set_role(),
196 * so make sure this object stays around. */
198 kref_get(&adm_ctx
->device
->kref
);
200 if (adm_ctx
->resource_name
) {
201 adm_ctx
->resource
= drbd_find_resource(adm_ctx
->resource_name
);
204 if (!adm_ctx
->device
&& (flags
& DRBD_ADM_NEED_MINOR
)) {
205 drbd_msg_put_info(adm_ctx
->reply_skb
, "unknown minor");
206 return ERR_MINOR_INVALID
;
208 if (!adm_ctx
->resource
&& (flags
& DRBD_ADM_NEED_RESOURCE
)) {
209 drbd_msg_put_info(adm_ctx
->reply_skb
, "unknown resource");
210 if (adm_ctx
->resource_name
)
211 return ERR_RES_NOT_KNOWN
;
212 return ERR_INVALID_REQUEST
;
215 if (flags
& DRBD_ADM_NEED_CONNECTION
) {
216 if (adm_ctx
->resource
) {
217 drbd_msg_put_info(adm_ctx
->reply_skb
, "no resource name expected");
218 return ERR_INVALID_REQUEST
;
220 if (adm_ctx
->device
) {
221 drbd_msg_put_info(adm_ctx
->reply_skb
, "no minor number expected");
222 return ERR_INVALID_REQUEST
;
224 if (adm_ctx
->my_addr
&& adm_ctx
->peer_addr
)
225 adm_ctx
->connection
= conn_get_by_addrs(nla_data(adm_ctx
->my_addr
),
226 nla_len(adm_ctx
->my_addr
),
227 nla_data(adm_ctx
->peer_addr
),
228 nla_len(adm_ctx
->peer_addr
));
229 if (!adm_ctx
->connection
) {
230 drbd_msg_put_info(adm_ctx
->reply_skb
, "unknown connection");
231 return ERR_INVALID_REQUEST
;
235 /* some more paranoia, if the request was over-determined */
236 if (adm_ctx
->device
&& adm_ctx
->resource
&&
237 adm_ctx
->device
->resource
!= adm_ctx
->resource
) {
238 pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n",
239 adm_ctx
->minor
, adm_ctx
->resource
->name
,
240 adm_ctx
->device
->resource
->name
);
241 drbd_msg_put_info(adm_ctx
->reply_skb
, "minor exists in different resource");
242 return ERR_INVALID_REQUEST
;
244 if (adm_ctx
->device
&&
245 adm_ctx
->volume
!= VOLUME_UNSPECIFIED
&&
246 adm_ctx
->volume
!= adm_ctx
->device
->vnr
) {
247 pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n",
248 adm_ctx
->minor
, adm_ctx
->volume
,
249 adm_ctx
->device
->vnr
,
250 adm_ctx
->device
->resource
->name
);
251 drbd_msg_put_info(adm_ctx
->reply_skb
, "minor exists as different volume");
252 return ERR_INVALID_REQUEST
;
255 /* still, provide adm_ctx->resource always, if possible. */
256 if (!adm_ctx
->resource
) {
257 adm_ctx
->resource
= adm_ctx
->device
? adm_ctx
->device
->resource
258 : adm_ctx
->connection
? adm_ctx
->connection
->resource
: NULL
;
259 if (adm_ctx
->resource
)
260 kref_get(&adm_ctx
->resource
->kref
);
266 nlmsg_free(adm_ctx
->reply_skb
);
267 adm_ctx
->reply_skb
= NULL
;
271 static int drbd_adm_finish(struct drbd_config_context
*adm_ctx
,
272 struct genl_info
*info
, int retcode
)
274 if (adm_ctx
->device
) {
275 kref_put(&adm_ctx
->device
->kref
, drbd_destroy_device
);
276 adm_ctx
->device
= NULL
;
278 if (adm_ctx
->connection
) {
279 kref_put(&adm_ctx
->connection
->kref
, &drbd_destroy_connection
);
280 adm_ctx
->connection
= NULL
;
282 if (adm_ctx
->resource
) {
283 kref_put(&adm_ctx
->resource
->kref
, drbd_destroy_resource
);
284 adm_ctx
->resource
= NULL
;
287 if (!adm_ctx
->reply_skb
)
290 adm_ctx
->reply_dh
->ret_code
= retcode
;
291 drbd_adm_send_reply(adm_ctx
->reply_skb
, info
);
295 static void setup_khelper_env(struct drbd_connection
*connection
, char **envp
)
299 /* FIXME: A future version will not allow this case. */
300 if (connection
->my_addr_len
== 0 || connection
->peer_addr_len
== 0)
303 switch (((struct sockaddr
*)&connection
->peer_addr
)->sa_family
) {
306 snprintf(envp
[4], 60, "DRBD_PEER_ADDRESS=%pI6",
307 &((struct sockaddr_in6
*)&connection
->peer_addr
)->sin6_addr
);
311 snprintf(envp
[4], 60, "DRBD_PEER_ADDRESS=%pI4",
312 &((struct sockaddr_in
*)&connection
->peer_addr
)->sin_addr
);
316 snprintf(envp
[4], 60, "DRBD_PEER_ADDRESS=%pI4",
317 &((struct sockaddr_in
*)&connection
->peer_addr
)->sin_addr
);
319 snprintf(envp
[3], 20, "DRBD_PEER_AF=%s", afs
);
322 int drbd_khelper(struct drbd_device
*device
, char *cmd
)
324 char *envp
[] = { "HOME=/",
326 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
327 (char[20]) { }, /* address family */
328 (char[60]) { }, /* address */
331 char *argv
[] = {usermode_helper
, cmd
, mb
, NULL
};
332 struct drbd_connection
*connection
= first_peer_device(device
)->connection
;
336 if (current
== connection
->worker
.task
)
337 set_bit(CALLBACK_PENDING
, &connection
->flags
);
339 snprintf(mb
, 12, "minor-%d", device_to_minor(device
));
340 setup_khelper_env(connection
, envp
);
342 /* The helper may take some time.
343 * write out any unsynced meta data changes now */
344 drbd_md_sync(device
);
346 drbd_info(device
, "helper command: %s %s %s\n", usermode_helper
, cmd
, mb
);
347 sib
.sib_reason
= SIB_HELPER_PRE
;
348 sib
.helper_name
= cmd
;
349 drbd_bcast_event(device
, &sib
);
350 ret
= call_usermodehelper(usermode_helper
, argv
, envp
, UMH_WAIT_PROC
);
352 drbd_warn(device
, "helper command: %s %s %s exit code %u (0x%x)\n",
353 usermode_helper
, cmd
, mb
,
354 (ret
>> 8) & 0xff, ret
);
356 drbd_info(device
, "helper command: %s %s %s exit code %u (0x%x)\n",
357 usermode_helper
, cmd
, mb
,
358 (ret
>> 8) & 0xff, ret
);
359 sib
.sib_reason
= SIB_HELPER_POST
;
360 sib
.helper_exit_code
= ret
;
361 drbd_bcast_event(device
, &sib
);
363 if (current
== connection
->worker
.task
)
364 clear_bit(CALLBACK_PENDING
, &connection
->flags
);
366 if (ret
< 0) /* Ignore any ERRNOs we got. */
372 static int conn_khelper(struct drbd_connection
*connection
, char *cmd
)
374 char *envp
[] = { "HOME=/",
376 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
377 (char[20]) { }, /* address family */
378 (char[60]) { }, /* address */
380 char *resource_name
= connection
->resource
->name
;
381 char *argv
[] = {usermode_helper
, cmd
, resource_name
, NULL
};
384 setup_khelper_env(connection
, envp
);
385 conn_md_sync(connection
);
387 drbd_info(connection
, "helper command: %s %s %s\n", usermode_helper
, cmd
, resource_name
);
388 /* TODO: conn_bcast_event() ?? */
390 ret
= call_usermodehelper(usermode_helper
, argv
, envp
, UMH_WAIT_PROC
);
392 drbd_warn(connection
, "helper command: %s %s %s exit code %u (0x%x)\n",
393 usermode_helper
, cmd
, resource_name
,
394 (ret
>> 8) & 0xff, ret
);
396 drbd_info(connection
, "helper command: %s %s %s exit code %u (0x%x)\n",
397 usermode_helper
, cmd
, resource_name
,
398 (ret
>> 8) & 0xff, ret
);
399 /* TODO: conn_bcast_event() ?? */
401 if (ret
< 0) /* Ignore any ERRNOs we got. */
407 static enum drbd_fencing_p
highest_fencing_policy(struct drbd_connection
*connection
)
409 enum drbd_fencing_p fp
= FP_NOT_AVAIL
;
410 struct drbd_peer_device
*peer_device
;
414 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
) {
415 struct drbd_device
*device
= peer_device
->device
;
416 if (get_ldev_if_state(device
, D_CONSISTENT
)) {
417 struct disk_conf
*disk_conf
=
418 rcu_dereference(peer_device
->device
->ldev
->disk_conf
);
419 fp
= max_t(enum drbd_fencing_p
, fp
, disk_conf
->fencing
);
425 if (fp
== FP_NOT_AVAIL
) {
426 /* IO Suspending works on the whole resource.
427 Do it only for one device. */
429 peer_device
= idr_get_next(&connection
->peer_devices
, &vnr
);
430 drbd_change_state(peer_device
->device
, CS_VERBOSE
| CS_HARD
, NS(susp_fen
, 0));
436 bool conn_try_outdate_peer(struct drbd_connection
*connection
)
438 unsigned int connect_cnt
;
439 union drbd_state mask
= { };
440 union drbd_state val
= { };
441 enum drbd_fencing_p fp
;
445 spin_lock_irq(&connection
->resource
->req_lock
);
446 if (connection
->cstate
>= C_WF_REPORT_PARAMS
) {
447 drbd_err(connection
, "Expected cstate < C_WF_REPORT_PARAMS\n");
448 spin_unlock_irq(&connection
->resource
->req_lock
);
452 connect_cnt
= connection
->connect_cnt
;
453 spin_unlock_irq(&connection
->resource
->req_lock
);
455 fp
= highest_fencing_policy(connection
);
458 drbd_warn(connection
, "Not fencing peer, I'm not even Consistent myself.\n");
465 r
= conn_khelper(connection
, "fence-peer");
467 switch ((r
>>8) & 0xff) {
468 case 3: /* peer is inconsistent */
469 ex_to_string
= "peer is inconsistent or worse";
471 val
.pdsk
= D_INCONSISTENT
;
473 case 4: /* peer got outdated, or was already outdated */
474 ex_to_string
= "peer was fenced";
476 val
.pdsk
= D_OUTDATED
;
478 case 5: /* peer was down */
479 if (conn_highest_disk(connection
) == D_UP_TO_DATE
) {
480 /* we will(have) create(d) a new UUID anyways... */
481 ex_to_string
= "peer is unreachable, assumed to be dead";
483 val
.pdsk
= D_OUTDATED
;
485 ex_to_string
= "peer unreachable, doing nothing since disk != UpToDate";
488 case 6: /* Peer is primary, voluntarily outdate myself.
489 * This is useful when an unconnected R_SECONDARY is asked to
490 * become R_PRIMARY, but finds the other peer being active. */
491 ex_to_string
= "peer is active";
492 drbd_warn(connection
, "Peer is primary, outdating myself.\n");
494 val
.disk
= D_OUTDATED
;
497 if (fp
!= FP_STONITH
)
498 drbd_err(connection
, "fence-peer() = 7 && fencing != Stonith !!!\n");
499 ex_to_string
= "peer was stonithed";
501 val
.pdsk
= D_OUTDATED
;
504 /* The script is broken ... */
505 drbd_err(connection
, "fence-peer helper broken, returned %d\n", (r
>>8)&0xff);
506 return false; /* Eventually leave IO frozen */
509 drbd_info(connection
, "fence-peer helper returned %d (%s)\n",
510 (r
>>8) & 0xff, ex_to_string
);
515 conn_request_state(connection, mask, val, CS_VERBOSE);
516 here, because we might were able to re-establish the connection in the
518 spin_lock_irq(&connection
->resource
->req_lock
);
519 if (connection
->cstate
< C_WF_REPORT_PARAMS
&& !test_bit(STATE_SENT
, &connection
->flags
)) {
520 if (connection
->connect_cnt
!= connect_cnt
)
521 /* In case the connection was established and droped
522 while the fence-peer handler was running, ignore it */
523 drbd_info(connection
, "Ignoring fence-peer exit code\n");
525 _conn_request_state(connection
, mask
, val
, CS_VERBOSE
);
527 spin_unlock_irq(&connection
->resource
->req_lock
);
529 return conn_highest_pdsk(connection
) <= D_OUTDATED
;
532 static int _try_outdate_peer_async(void *data
)
534 struct drbd_connection
*connection
= (struct drbd_connection
*)data
;
536 conn_try_outdate_peer(connection
);
538 kref_put(&connection
->kref
, drbd_destroy_connection
);
542 void conn_try_outdate_peer_async(struct drbd_connection
*connection
)
544 struct task_struct
*opa
;
546 kref_get(&connection
->kref
);
547 opa
= kthread_run(_try_outdate_peer_async
, connection
, "drbd_async_h");
549 drbd_err(connection
, "out of mem, failed to invoke fence-peer helper\n");
550 kref_put(&connection
->kref
, drbd_destroy_connection
);
555 drbd_set_role(struct drbd_device
*const device
, enum drbd_role new_role
, int force
)
557 struct drbd_peer_device
*const peer_device
= first_peer_device(device
);
558 struct drbd_connection
*const connection
= peer_device
? peer_device
->connection
: NULL
;
559 const int max_tries
= 4;
560 enum drbd_state_rv rv
= SS_UNKNOWN_ERROR
;
564 union drbd_state mask
, val
;
566 if (new_role
== R_PRIMARY
) {
567 struct drbd_connection
*connection
;
569 /* Detect dead peers as soon as possible. */
572 for_each_connection(connection
, device
->resource
)
573 request_ping(connection
);
577 mutex_lock(device
->state_mutex
);
579 mask
.i
= 0; mask
.role
= R_MASK
;
580 val
.i
= 0; val
.role
= new_role
;
582 while (try++ < max_tries
) {
583 rv
= _drbd_request_state(device
, mask
, val
, CS_WAIT_COMPLETE
);
585 /* in case we first succeeded to outdate,
586 * but now suddenly could establish a connection */
587 if (rv
== SS_CW_FAILED_BY_PEER
&& mask
.pdsk
!= 0) {
593 if (rv
== SS_NO_UP_TO_DATE_DISK
&& force
&&
594 (device
->state
.disk
< D_UP_TO_DATE
&&
595 device
->state
.disk
>= D_INCONSISTENT
)) {
597 val
.disk
= D_UP_TO_DATE
;
602 if (rv
== SS_NO_UP_TO_DATE_DISK
&&
603 device
->state
.disk
== D_CONSISTENT
&& mask
.pdsk
== 0) {
604 D_ASSERT(device
, device
->state
.pdsk
== D_UNKNOWN
);
606 if (conn_try_outdate_peer(connection
)) {
607 val
.disk
= D_UP_TO_DATE
;
613 if (rv
== SS_NOTHING_TO_DO
)
615 if (rv
== SS_PRIMARY_NOP
&& mask
.pdsk
== 0) {
616 if (!conn_try_outdate_peer(connection
) && force
) {
617 drbd_warn(device
, "Forced into split brain situation!\n");
619 val
.pdsk
= D_OUTDATED
;
624 if (rv
== SS_TWO_PRIMARIES
) {
625 /* Maybe the peer is detected as dead very soon...
626 retry at most once more in this case. */
629 nc
= rcu_dereference(connection
->net_conf
);
630 timeo
= nc
? (nc
->ping_timeo
+ 1) * HZ
/ 10 : 1;
632 schedule_timeout_interruptible(timeo
);
637 if (rv
< SS_SUCCESS
) {
638 rv
= _drbd_request_state(device
, mask
, val
,
639 CS_VERBOSE
+ CS_WAIT_COMPLETE
);
650 drbd_warn(device
, "Forced to consider local data as UpToDate!\n");
652 /* Wait until nothing is on the fly :) */
653 wait_event(device
->misc_wait
, atomic_read(&device
->ap_pending_cnt
) == 0);
655 /* FIXME also wait for all pending P_BARRIER_ACK? */
657 if (new_role
== R_SECONDARY
) {
658 if (get_ldev(device
)) {
659 device
->ldev
->md
.uuid
[UI_CURRENT
] &= ~(u64
)1;
663 mutex_lock(&device
->resource
->conf_update
);
664 nc
= connection
->net_conf
;
666 nc
->discard_my_data
= 0; /* without copy; single bit op is atomic */
667 mutex_unlock(&device
->resource
->conf_update
);
669 if (get_ldev(device
)) {
670 if (((device
->state
.conn
< C_CONNECTED
||
671 device
->state
.pdsk
<= D_FAILED
)
672 && device
->ldev
->md
.uuid
[UI_BITMAP
] == 0) || forced
)
673 drbd_uuid_new_current(device
);
675 device
->ldev
->md
.uuid
[UI_CURRENT
] |= (u64
)1;
680 /* writeout of activity log covered areas of the bitmap
681 * to stable storage done in after state change already */
683 if (device
->state
.conn
>= C_WF_REPORT_PARAMS
) {
684 /* if this was forced, we should consider sync */
686 drbd_send_uuids(peer_device
);
687 drbd_send_current_state(peer_device
);
690 drbd_md_sync(device
);
691 set_disk_ro(device
->vdisk
, new_role
== R_SECONDARY
);
692 kobject_uevent(&disk_to_dev(device
->vdisk
)->kobj
, KOBJ_CHANGE
);
694 mutex_unlock(device
->state_mutex
);
698 static const char *from_attrs_err_to_txt(int err
)
700 return err
== -ENOMSG
? "required attribute missing" :
701 err
== -EOPNOTSUPP
? "unknown mandatory attribute" :
702 err
== -EEXIST
? "can not change invariant setting" :
703 "invalid attribute value";
706 int drbd_adm_set_role(struct sk_buff
*skb
, struct genl_info
*info
)
708 struct drbd_config_context adm_ctx
;
709 struct set_role_parms parms
;
711 enum drbd_ret_code retcode
;
713 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
714 if (!adm_ctx
.reply_skb
)
716 if (retcode
!= NO_ERROR
)
719 memset(&parms
, 0, sizeof(parms
));
720 if (info
->attrs
[DRBD_NLA_SET_ROLE_PARMS
]) {
721 err
= set_role_parms_from_attrs(&parms
, info
);
723 retcode
= ERR_MANDATORY_TAG
;
724 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
729 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
731 if (info
->genlhdr
->cmd
== DRBD_ADM_PRIMARY
)
732 retcode
= drbd_set_role(adm_ctx
.device
, R_PRIMARY
, parms
.assume_uptodate
);
734 retcode
= drbd_set_role(adm_ctx
.device
, R_SECONDARY
, 0);
736 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
739 drbd_adm_finish(&adm_ctx
, info
, retcode
);
743 /* Initializes the md.*_offset members, so we are able to find
744 * the on disk meta data.
746 * We currently have two possible layouts:
748 * |----------- md_size_sect ------------------|
749 * [ 4k superblock ][ activity log ][ Bitmap ]
751 * | bm_offset = al_offset + X |
752 * ==> bitmap sectors = md_size_sect - bm_offset
755 * |----------- md_size_sect ------------------|
756 * [data.....][ Bitmap ][ activity log ][ 4k superblock ]
758 * | bm_offset = al_offset - Y |
759 * ==> bitmap sectors = Y = al_offset - bm_offset
761 * Activity log size used to be fixed 32kB,
762 * but is about to become configurable.
764 static void drbd_md_set_sector_offsets(struct drbd_device
*device
,
765 struct drbd_backing_dev
*bdev
)
767 sector_t md_size_sect
= 0;
768 unsigned int al_size_sect
= bdev
->md
.al_size_4k
* 8;
770 bdev
->md
.md_offset
= drbd_md_ss(bdev
);
772 switch (bdev
->md
.meta_dev_idx
) {
774 /* v07 style fixed size indexed meta data */
775 bdev
->md
.md_size_sect
= MD_128MB_SECT
;
776 bdev
->md
.al_offset
= MD_4kB_SECT
;
777 bdev
->md
.bm_offset
= MD_4kB_SECT
+ al_size_sect
;
779 case DRBD_MD_INDEX_FLEX_EXT
:
780 /* just occupy the full device; unit: sectors */
781 bdev
->md
.md_size_sect
= drbd_get_capacity(bdev
->md_bdev
);
782 bdev
->md
.al_offset
= MD_4kB_SECT
;
783 bdev
->md
.bm_offset
= MD_4kB_SECT
+ al_size_sect
;
785 case DRBD_MD_INDEX_INTERNAL
:
786 case DRBD_MD_INDEX_FLEX_INT
:
787 /* al size is still fixed */
788 bdev
->md
.al_offset
= -al_size_sect
;
789 /* we need (slightly less than) ~ this much bitmap sectors: */
790 md_size_sect
= drbd_get_capacity(bdev
->backing_bdev
);
791 md_size_sect
= ALIGN(md_size_sect
, BM_SECT_PER_EXT
);
792 md_size_sect
= BM_SECT_TO_EXT(md_size_sect
);
793 md_size_sect
= ALIGN(md_size_sect
, 8);
795 /* plus the "drbd meta data super block",
796 * and the activity log; */
797 md_size_sect
+= MD_4kB_SECT
+ al_size_sect
;
799 bdev
->md
.md_size_sect
= md_size_sect
;
800 /* bitmap offset is adjusted by 'super' block size */
801 bdev
->md
.bm_offset
= -md_size_sect
+ MD_4kB_SECT
;
806 /* input size is expected to be in KB */
807 char *ppsize(char *buf
, unsigned long long size
)
809 /* Needs 9 bytes at max including trailing NUL:
810 * -1ULL ==> "16384 EB" */
811 static char units
[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
813 while (size
>= 10000 && base
< sizeof(units
)-1) {
815 size
= (size
>> 10) + !!(size
& (1<<9));
818 sprintf(buf
, "%u %cB", (unsigned)size
, units
[base
]);
823 /* there is still a theoretical deadlock when called from receiver
824 * on an D_INCONSISTENT R_PRIMARY:
825 * remote READ does inc_ap_bio, receiver would need to receive answer
826 * packet from remote to dec_ap_bio again.
827 * receiver receive_sizes(), comes here,
828 * waits for ap_bio_cnt == 0. -> deadlock.
829 * but this cannot happen, actually, because:
830 * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable
831 * (not connected, or bad/no disk on peer):
832 * see drbd_fail_request_early, ap_bio_cnt is zero.
833 * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
834 * peer may not initiate a resize.
836 /* Note these are not to be confused with
837 * drbd_adm_suspend_io/drbd_adm_resume_io,
838 * which are (sub) state changes triggered by admin (drbdsetup),
839 * and can be long lived.
840 * This changes an device->flag, is triggered by drbd internals,
841 * and should be short-lived. */
842 void drbd_suspend_io(struct drbd_device
*device
)
844 set_bit(SUSPEND_IO
, &device
->flags
);
845 if (drbd_suspended(device
))
847 wait_event(device
->misc_wait
, !atomic_read(&device
->ap_bio_cnt
));
850 void drbd_resume_io(struct drbd_device
*device
)
852 clear_bit(SUSPEND_IO
, &device
->flags
);
853 wake_up(&device
->misc_wait
);
857 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
858 * @device: DRBD device.
860 * Returns 0 on success, negative return values indicate errors.
861 * You should call drbd_md_sync() after calling this function.
863 enum determine_dev_size
864 drbd_determine_dev_size(struct drbd_device
*device
, enum dds_flags flags
, struct resize_parms
*rs
) __must_hold(local
)
866 sector_t prev_first_sect
, prev_size
; /* previous meta location */
867 sector_t la_size_sect
, u_size
;
868 struct drbd_md
*md
= &device
->ldev
->md
;
869 u32 prev_al_stripe_size_4k
;
875 int md_moved
, la_size_changed
;
876 enum determine_dev_size rv
= DS_UNCHANGED
;
879 * application request passes inc_ap_bio,
880 * but then cannot get an AL-reference.
881 * this function later may wait on ap_bio_cnt == 0. -> deadlock.
884 * Suspend IO right here.
885 * still lock the act_log to not trigger ASSERTs there.
887 drbd_suspend_io(device
);
888 buffer
= drbd_md_get_buffer(device
); /* Lock meta-data IO */
890 drbd_resume_io(device
);
894 /* no wait necessary anymore, actually we could assert that */
895 wait_event(device
->al_wait
, lc_try_lock(device
->act_log
));
897 prev_first_sect
= drbd_md_first_sector(device
->ldev
);
898 prev_size
= device
->ldev
->md
.md_size_sect
;
899 la_size_sect
= device
->ldev
->md
.la_size_sect
;
902 /* rs is non NULL if we should change the AL layout only */
904 prev_al_stripes
= md
->al_stripes
;
905 prev_al_stripe_size_4k
= md
->al_stripe_size_4k
;
907 md
->al_stripes
= rs
->al_stripes
;
908 md
->al_stripe_size_4k
= rs
->al_stripe_size
/ 4;
909 md
->al_size_4k
= (u64
)rs
->al_stripes
* rs
->al_stripe_size
/ 4;
912 drbd_md_set_sector_offsets(device
, device
->ldev
);
915 u_size
= rcu_dereference(device
->ldev
->disk_conf
)->disk_size
;
917 size
= drbd_new_dev_size(device
, device
->ldev
, u_size
, flags
& DDSF_FORCED
);
919 if (size
< la_size_sect
) {
920 if (rs
&& u_size
== 0) {
921 /* Remove "rs &&" later. This check should always be active, but
922 right now the receiver expects the permissive behavior */
923 drbd_warn(device
, "Implicit shrink not allowed. "
924 "Use --size=%llus for explicit shrink.\n",
925 (unsigned long long)size
);
926 rv
= DS_ERROR_SHRINK
;
929 rv
= DS_ERROR_SPACE_MD
;
930 if (rv
!= DS_UNCHANGED
)
934 if (drbd_get_capacity(device
->this_bdev
) != size
||
935 drbd_bm_capacity(device
) != size
) {
937 err
= drbd_bm_resize(device
, size
, !(flags
& DDSF_NO_RESYNC
));
939 /* currently there is only one error: ENOMEM! */
940 size
= drbd_bm_capacity(device
)>>1;
942 drbd_err(device
, "OUT OF MEMORY! "
943 "Could not allocate bitmap!\n");
945 drbd_err(device
, "BM resizing failed. "
946 "Leaving size unchanged at size = %lu KB\n",
947 (unsigned long)size
);
951 /* racy, see comments above. */
952 drbd_set_my_capacity(device
, size
);
953 device
->ldev
->md
.la_size_sect
= size
;
954 drbd_info(device
, "size = %s (%llu KB)\n", ppsize(ppb
, size
>>1),
955 (unsigned long long)size
>>1);
960 la_size_changed
= (la_size_sect
!= device
->ldev
->md
.la_size_sect
);
962 md_moved
= prev_first_sect
!= drbd_md_first_sector(device
->ldev
)
963 || prev_size
!= device
->ldev
->md
.md_size_sect
;
965 if (la_size_changed
|| md_moved
|| rs
) {
968 /* We do some synchronous IO below, which may take some time.
969 * Clear the timer, to avoid scary "timer expired!" messages,
970 * "Superblock" is written out at least twice below, anyways. */
971 del_timer(&device
->md_sync_timer
);
972 drbd_al_shrink(device
); /* All extents inactive. */
974 prev_flags
= md
->flags
;
975 md
->flags
&= ~MDF_PRIMARY_IND
;
976 drbd_md_write(device
, buffer
);
978 drbd_info(device
, "Writing the whole bitmap, %s\n",
979 la_size_changed
&& md_moved
? "size changed and md moved" :
980 la_size_changed
? "size changed" : "md moved");
981 /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */
982 drbd_bitmap_io(device
, md_moved
? &drbd_bm_write_all
: &drbd_bm_write
,
983 "size changed", BM_LOCKED_MASK
);
984 drbd_initialize_al(device
, buffer
);
986 md
->flags
= prev_flags
;
987 drbd_md_write(device
, buffer
);
990 drbd_info(device
, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n",
991 md
->al_stripes
, md
->al_stripe_size_4k
* 4);
994 if (size
> la_size_sect
)
995 rv
= la_size_sect
? DS_GREW
: DS_GREW_FROM_ZERO
;
996 if (size
< la_size_sect
)
1002 md
->al_stripes
= prev_al_stripes
;
1003 md
->al_stripe_size_4k
= prev_al_stripe_size_4k
;
1004 md
->al_size_4k
= (u64
)prev_al_stripes
* prev_al_stripe_size_4k
;
1006 drbd_md_set_sector_offsets(device
, device
->ldev
);
1009 lc_unlock(device
->act_log
);
1010 wake_up(&device
->al_wait
);
1011 drbd_md_put_buffer(device
);
1012 drbd_resume_io(device
);
1018 drbd_new_dev_size(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
,
1019 sector_t u_size
, int assume_peer_has_space
)
1021 sector_t p_size
= device
->p_size
; /* partner's disk size. */
1022 sector_t la_size_sect
= bdev
->md
.la_size_sect
; /* last agreed size. */
1023 sector_t m_size
; /* my size */
1026 m_size
= drbd_get_max_capacity(bdev
);
1028 if (device
->state
.conn
< C_CONNECTED
&& assume_peer_has_space
) {
1029 drbd_warn(device
, "Resize while not connected was forced by the user!\n");
1033 if (p_size
&& m_size
) {
1034 size
= min_t(sector_t
, p_size
, m_size
);
1037 size
= la_size_sect
;
1038 if (m_size
&& m_size
< size
)
1040 if (p_size
&& p_size
< size
)
1051 drbd_err(device
, "Both nodes diskless!\n");
1055 drbd_err(device
, "Requested disk size is too big (%lu > %lu)\n",
1056 (unsigned long)u_size
>>1, (unsigned long)size
>>1);
1065 * drbd_check_al_size() - Ensures that the AL is of the right size
1066 * @device: DRBD device.
1068 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
1069 * failed, and 0 on success. You should call drbd_md_sync() after you called
1072 static int drbd_check_al_size(struct drbd_device
*device
, struct disk_conf
*dc
)
1074 struct lru_cache
*n
, *t
;
1075 struct lc_element
*e
;
1076 unsigned int in_use
;
1079 if (device
->act_log
&&
1080 device
->act_log
->nr_elements
== dc
->al_extents
)
1084 t
= device
->act_log
;
1085 n
= lc_create("act_log", drbd_al_ext_cache
, AL_UPDATES_PER_TRANSACTION
,
1086 dc
->al_extents
, sizeof(struct lc_element
), 0);
1089 drbd_err(device
, "Cannot allocate act_log lru!\n");
1092 spin_lock_irq(&device
->al_lock
);
1094 for (i
= 0; i
< t
->nr_elements
; i
++) {
1095 e
= lc_element_by_index(t
, i
);
1097 drbd_err(device
, "refcnt(%d)==%d\n",
1098 e
->lc_number
, e
->refcnt
);
1099 in_use
+= e
->refcnt
;
1103 device
->act_log
= n
;
1104 spin_unlock_irq(&device
->al_lock
);
1106 drbd_err(device
, "Activity log still in use!\n");
1113 drbd_md_mark_dirty(device
); /* we changed device->act_log->nr_elemens */
1117 static void drbd_setup_queue_param(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
,
1118 unsigned int max_bio_size
)
1120 struct request_queue
* const q
= device
->rq_queue
;
1121 unsigned int max_hw_sectors
= max_bio_size
>> 9;
1122 unsigned int max_segments
= 0;
1123 struct request_queue
*b
= NULL
;
1126 b
= bdev
->backing_bdev
->bd_disk
->queue
;
1128 max_hw_sectors
= min(queue_max_hw_sectors(b
), max_bio_size
>> 9);
1130 max_segments
= rcu_dereference(device
->ldev
->disk_conf
)->max_bio_bvecs
;
1133 blk_set_stacking_limits(&q
->limits
);
1134 blk_queue_max_write_same_sectors(q
, 0);
1137 blk_queue_logical_block_size(q
, 512);
1138 blk_queue_max_hw_sectors(q
, max_hw_sectors
);
1139 /* This is the workaround for "bio would need to, but cannot, be split" */
1140 blk_queue_max_segments(q
, max_segments
? max_segments
: BLK_MAX_SEGMENTS
);
1141 blk_queue_segment_boundary(q
, PAGE_CACHE_SIZE
-1);
1144 struct drbd_connection
*connection
= first_peer_device(device
)->connection
;
1146 if (blk_queue_discard(b
) &&
1147 (connection
->cstate
< C_CONNECTED
|| connection
->agreed_features
& FF_TRIM
)) {
1148 /* For now, don't allow more than one activity log extent worth of data
1149 * to be discarded in one go. We may need to rework drbd_al_begin_io()
1150 * to allow for even larger discard ranges */
1151 q
->limits
.max_discard_sectors
= DRBD_MAX_DISCARD_SECTORS
;
1153 queue_flag_set_unlocked(QUEUE_FLAG_DISCARD
, q
);
1154 /* REALLY? Is stacking secdiscard "legal"? */
1155 if (blk_queue_secdiscard(b
))
1156 queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD
, q
);
1158 q
->limits
.max_discard_sectors
= 0;
1159 queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD
, q
);
1160 queue_flag_clear_unlocked(QUEUE_FLAG_SECDISCARD
, q
);
1163 blk_queue_stack_limits(q
, b
);
1165 if (q
->backing_dev_info
.ra_pages
!= b
->backing_dev_info
.ra_pages
) {
1166 drbd_info(device
, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
1167 q
->backing_dev_info
.ra_pages
,
1168 b
->backing_dev_info
.ra_pages
);
1169 q
->backing_dev_info
.ra_pages
= b
->backing_dev_info
.ra_pages
;
1174 void drbd_reconsider_max_bio_size(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
)
1176 unsigned int now
, new, local
, peer
;
1178 now
= queue_max_hw_sectors(device
->rq_queue
) << 9;
1179 local
= device
->local_max_bio_size
; /* Eventually last known value, from volatile memory */
1180 peer
= device
->peer_max_bio_size
; /* Eventually last known value, from meta data */
1183 local
= queue_max_hw_sectors(bdev
->backing_bdev
->bd_disk
->queue
) << 9;
1184 device
->local_max_bio_size
= local
;
1186 local
= min(local
, DRBD_MAX_BIO_SIZE
);
1188 /* We may ignore peer limits if the peer is modern enough.
1189 Because new from 8.3.8 onwards the peer can use multiple
1190 BIOs for a single peer_request */
1191 if (device
->state
.conn
>= C_WF_REPORT_PARAMS
) {
1192 if (first_peer_device(device
)->connection
->agreed_pro_version
< 94)
1193 peer
= min(device
->peer_max_bio_size
, DRBD_MAX_SIZE_H80_PACKET
);
1194 /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */
1195 else if (first_peer_device(device
)->connection
->agreed_pro_version
== 94)
1196 peer
= DRBD_MAX_SIZE_H80_PACKET
;
1197 else if (first_peer_device(device
)->connection
->agreed_pro_version
< 100)
1198 peer
= DRBD_MAX_BIO_SIZE_P95
; /* drbd 8.3.8 onwards, before 8.4.0 */
1200 peer
= DRBD_MAX_BIO_SIZE
;
1202 /* We may later detach and re-attach on a disconnected Primary.
1203 * Avoid this setting to jump back in that case.
1204 * We want to store what we know the peer DRBD can handle,
1205 * not what the peer IO backend can handle. */
1206 if (peer
> device
->peer_max_bio_size
)
1207 device
->peer_max_bio_size
= peer
;
1209 new = min(local
, peer
);
1211 if (device
->state
.role
== R_PRIMARY
&& new < now
)
1212 drbd_err(device
, "ASSERT FAILED new < now; (%u < %u)\n", new, now
);
1215 drbd_info(device
, "max BIO size = %u\n", new);
1217 drbd_setup_queue_param(device
, bdev
, new);
1220 /* Starts the worker thread */
1221 static void conn_reconfig_start(struct drbd_connection
*connection
)
1223 drbd_thread_start(&connection
->worker
);
1224 drbd_flush_workqueue(&connection
->sender_work
);
1227 /* if still unconfigured, stops worker again. */
1228 static void conn_reconfig_done(struct drbd_connection
*connection
)
1231 spin_lock_irq(&connection
->resource
->req_lock
);
1232 stop_threads
= conn_all_vols_unconf(connection
) &&
1233 connection
->cstate
== C_STANDALONE
;
1234 spin_unlock_irq(&connection
->resource
->req_lock
);
1236 /* asender is implicitly stopped by receiver
1237 * in conn_disconnect() */
1238 drbd_thread_stop(&connection
->receiver
);
1239 drbd_thread_stop(&connection
->worker
);
1243 /* Make sure IO is suspended before calling this function(). */
1244 static void drbd_suspend_al(struct drbd_device
*device
)
1248 if (!lc_try_lock(device
->act_log
)) {
1249 drbd_warn(device
, "Failed to lock al in drbd_suspend_al()\n");
1253 drbd_al_shrink(device
);
1254 spin_lock_irq(&device
->resource
->req_lock
);
1255 if (device
->state
.conn
< C_CONNECTED
)
1256 s
= !test_and_set_bit(AL_SUSPENDED
, &device
->flags
);
1257 spin_unlock_irq(&device
->resource
->req_lock
);
1258 lc_unlock(device
->act_log
);
1261 drbd_info(device
, "Suspended AL updates\n");
1265 static bool should_set_defaults(struct genl_info
*info
)
1267 unsigned flags
= ((struct drbd_genlmsghdr
*)info
->userhdr
)->flags
;
1268 return 0 != (flags
& DRBD_GENL_F_SET_DEFAULTS
);
1271 static unsigned int drbd_al_extents_max(struct drbd_backing_dev
*bdev
)
1273 /* This is limited by 16 bit "slot" numbers,
1274 * and by available on-disk context storage.
1276 * Also (u16)~0 is special (denotes a "free" extent).
1278 * One transaction occupies one 4kB on-disk block,
1279 * we have n such blocks in the on disk ring buffer,
1280 * the "current" transaction may fail (n-1),
1281 * and there is 919 slot numbers context information per transaction.
1283 * 72 transaction blocks amounts to more than 2**16 context slots,
1284 * so cap there first.
1286 const unsigned int max_al_nr
= DRBD_AL_EXTENTS_MAX
;
1287 const unsigned int sufficient_on_disk
=
1288 (max_al_nr
+ AL_CONTEXT_PER_TRANSACTION
-1)
1289 /AL_CONTEXT_PER_TRANSACTION
;
1291 unsigned int al_size_4k
= bdev
->md
.al_size_4k
;
1293 if (al_size_4k
> sufficient_on_disk
)
1296 return (al_size_4k
- 1) * AL_CONTEXT_PER_TRANSACTION
;
1299 static bool write_ordering_changed(struct disk_conf
*a
, struct disk_conf
*b
)
1301 return a
->disk_barrier
!= b
->disk_barrier
||
1302 a
->disk_flushes
!= b
->disk_flushes
||
1303 a
->disk_drain
!= b
->disk_drain
;
1306 int drbd_adm_disk_opts(struct sk_buff
*skb
, struct genl_info
*info
)
1308 struct drbd_config_context adm_ctx
;
1309 enum drbd_ret_code retcode
;
1310 struct drbd_device
*device
;
1311 struct disk_conf
*new_disk_conf
, *old_disk_conf
;
1312 struct fifo_buffer
*old_plan
= NULL
, *new_plan
= NULL
;
1315 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
1316 if (!adm_ctx
.reply_skb
)
1318 if (retcode
!= NO_ERROR
)
1321 device
= adm_ctx
.device
;
1322 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
1324 /* we also need a disk
1325 * to change the options on */
1326 if (!get_ldev(device
)) {
1327 retcode
= ERR_NO_DISK
;
1331 new_disk_conf
= kmalloc(sizeof(struct disk_conf
), GFP_KERNEL
);
1332 if (!new_disk_conf
) {
1333 retcode
= ERR_NOMEM
;
1337 mutex_lock(&device
->resource
->conf_update
);
1338 old_disk_conf
= device
->ldev
->disk_conf
;
1339 *new_disk_conf
= *old_disk_conf
;
1340 if (should_set_defaults(info
))
1341 set_disk_conf_defaults(new_disk_conf
);
1343 err
= disk_conf_from_attrs_for_change(new_disk_conf
, info
);
1344 if (err
&& err
!= -ENOMSG
) {
1345 retcode
= ERR_MANDATORY_TAG
;
1346 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
1350 if (!expect(new_disk_conf
->resync_rate
>= 1))
1351 new_disk_conf
->resync_rate
= 1;
1353 if (new_disk_conf
->al_extents
< DRBD_AL_EXTENTS_MIN
)
1354 new_disk_conf
->al_extents
= DRBD_AL_EXTENTS_MIN
;
1355 if (new_disk_conf
->al_extents
> drbd_al_extents_max(device
->ldev
))
1356 new_disk_conf
->al_extents
= drbd_al_extents_max(device
->ldev
);
1358 if (new_disk_conf
->c_plan_ahead
> DRBD_C_PLAN_AHEAD_MAX
)
1359 new_disk_conf
->c_plan_ahead
= DRBD_C_PLAN_AHEAD_MAX
;
1361 fifo_size
= (new_disk_conf
->c_plan_ahead
* 10 * SLEEP_TIME
) / HZ
;
1362 if (fifo_size
!= device
->rs_plan_s
->size
) {
1363 new_plan
= fifo_alloc(fifo_size
);
1365 drbd_err(device
, "kmalloc of fifo_buffer failed");
1366 retcode
= ERR_NOMEM
;
1371 drbd_suspend_io(device
);
1372 wait_event(device
->al_wait
, lc_try_lock(device
->act_log
));
1373 drbd_al_shrink(device
);
1374 err
= drbd_check_al_size(device
, new_disk_conf
);
1375 lc_unlock(device
->act_log
);
1376 wake_up(&device
->al_wait
);
1377 drbd_resume_io(device
);
1380 retcode
= ERR_NOMEM
;
1384 write_lock_irq(&global_state_lock
);
1385 retcode
= drbd_resync_after_valid(device
, new_disk_conf
->resync_after
);
1386 if (retcode
== NO_ERROR
) {
1387 rcu_assign_pointer(device
->ldev
->disk_conf
, new_disk_conf
);
1388 drbd_resync_after_changed(device
);
1390 write_unlock_irq(&global_state_lock
);
1392 if (retcode
!= NO_ERROR
)
1396 old_plan
= device
->rs_plan_s
;
1397 rcu_assign_pointer(device
->rs_plan_s
, new_plan
);
1400 mutex_unlock(&device
->resource
->conf_update
);
1402 if (new_disk_conf
->al_updates
)
1403 device
->ldev
->md
.flags
&= ~MDF_AL_DISABLED
;
1405 device
->ldev
->md
.flags
|= MDF_AL_DISABLED
;
1407 if (new_disk_conf
->md_flushes
)
1408 clear_bit(MD_NO_FUA
, &device
->flags
);
1410 set_bit(MD_NO_FUA
, &device
->flags
);
1412 if (write_ordering_changed(old_disk_conf
, new_disk_conf
))
1413 drbd_bump_write_ordering(device
->resource
, NULL
, WO_bdev_flush
);
1415 drbd_md_sync(device
);
1417 if (device
->state
.conn
>= C_CONNECTED
) {
1418 struct drbd_peer_device
*peer_device
;
1420 for_each_peer_device(peer_device
, device
)
1421 drbd_send_sync_param(peer_device
);
1425 kfree(old_disk_conf
);
1427 mod_timer(&device
->request_timer
, jiffies
+ HZ
);
1431 mutex_unlock(&device
->resource
->conf_update
);
1433 kfree(new_disk_conf
);
1438 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
1440 drbd_adm_finish(&adm_ctx
, info
, retcode
);
1444 int drbd_adm_attach(struct sk_buff
*skb
, struct genl_info
*info
)
1446 struct drbd_config_context adm_ctx
;
1447 struct drbd_device
*device
;
1448 struct drbd_peer_device
*peer_device
;
1449 struct drbd_connection
*connection
;
1451 enum drbd_ret_code retcode
;
1452 enum determine_dev_size dd
;
1453 sector_t max_possible_sectors
;
1454 sector_t min_md_device_sectors
;
1455 struct drbd_backing_dev
*nbc
= NULL
; /* new_backing_conf */
1456 struct disk_conf
*new_disk_conf
= NULL
;
1457 struct block_device
*bdev
;
1458 struct lru_cache
*resync_lru
= NULL
;
1459 struct fifo_buffer
*new_plan
= NULL
;
1460 union drbd_state ns
, os
;
1461 enum drbd_state_rv rv
;
1462 struct net_conf
*nc
;
1464 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
1465 if (!adm_ctx
.reply_skb
)
1467 if (retcode
!= NO_ERROR
)
1470 device
= adm_ctx
.device
;
1471 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
1472 peer_device
= first_peer_device(device
);
1473 connection
= peer_device
? peer_device
->connection
: NULL
;
1474 conn_reconfig_start(connection
);
1476 /* if you want to reconfigure, please tear down first */
1477 if (device
->state
.disk
> D_DISKLESS
) {
1478 retcode
= ERR_DISK_CONFIGURED
;
1481 /* It may just now have detached because of IO error. Make sure
1482 * drbd_ldev_destroy is done already, we may end up here very fast,
1483 * e.g. if someone calls attach from the on-io-error handler,
1484 * to realize a "hot spare" feature (not that I'd recommend that) */
1485 wait_event(device
->misc_wait
, !atomic_read(&device
->local_cnt
));
1487 /* make sure there is no leftover from previous force-detach attempts */
1488 clear_bit(FORCE_DETACH
, &device
->flags
);
1489 clear_bit(WAS_IO_ERROR
, &device
->flags
);
1490 clear_bit(WAS_READ_ERROR
, &device
->flags
);
1492 /* and no leftover from previously aborted resync or verify, either */
1493 device
->rs_total
= 0;
1494 device
->rs_failed
= 0;
1495 atomic_set(&device
->rs_pending_cnt
, 0);
1497 /* allocation not in the IO path, drbdsetup context */
1498 nbc
= kzalloc(sizeof(struct drbd_backing_dev
), GFP_KERNEL
);
1500 retcode
= ERR_NOMEM
;
1503 spin_lock_init(&nbc
->md
.uuid_lock
);
1505 new_disk_conf
= kzalloc(sizeof(struct disk_conf
), GFP_KERNEL
);
1506 if (!new_disk_conf
) {
1507 retcode
= ERR_NOMEM
;
1510 nbc
->disk_conf
= new_disk_conf
;
1512 set_disk_conf_defaults(new_disk_conf
);
1513 err
= disk_conf_from_attrs(new_disk_conf
, info
);
1515 retcode
= ERR_MANDATORY_TAG
;
1516 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
1520 if (new_disk_conf
->c_plan_ahead
> DRBD_C_PLAN_AHEAD_MAX
)
1521 new_disk_conf
->c_plan_ahead
= DRBD_C_PLAN_AHEAD_MAX
;
1523 new_plan
= fifo_alloc((new_disk_conf
->c_plan_ahead
* 10 * SLEEP_TIME
) / HZ
);
1525 retcode
= ERR_NOMEM
;
1529 if (new_disk_conf
->meta_dev_idx
< DRBD_MD_INDEX_FLEX_INT
) {
1530 retcode
= ERR_MD_IDX_INVALID
;
1534 write_lock_irq(&global_state_lock
);
1535 retcode
= drbd_resync_after_valid(device
, new_disk_conf
->resync_after
);
1536 write_unlock_irq(&global_state_lock
);
1537 if (retcode
!= NO_ERROR
)
1541 nc
= rcu_dereference(connection
->net_conf
);
1543 if (new_disk_conf
->fencing
== FP_STONITH
&& nc
->wire_protocol
== DRBD_PROT_A
) {
1545 retcode
= ERR_STONITH_AND_PROT_A
;
1551 bdev
= blkdev_get_by_path(new_disk_conf
->backing_dev
,
1552 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
, device
);
1554 drbd_err(device
, "open(\"%s\") failed with %ld\n", new_disk_conf
->backing_dev
,
1556 retcode
= ERR_OPEN_DISK
;
1559 nbc
->backing_bdev
= bdev
;
1562 * meta_dev_idx >= 0: external fixed size, possibly multiple
1563 * drbd sharing one meta device. TODO in that case, paranoia
1564 * check that [md_bdev, meta_dev_idx] is not yet used by some
1565 * other drbd minor! (if you use drbd.conf + drbdadm, that
1566 * should check it for you already; but if you don't, or
1567 * someone fooled it, we need to double check here)
1569 bdev
= blkdev_get_by_path(new_disk_conf
->meta_dev
,
1570 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
,
1571 (new_disk_conf
->meta_dev_idx
< 0) ?
1572 (void *)device
: (void *)drbd_m_holder
);
1574 drbd_err(device
, "open(\"%s\") failed with %ld\n", new_disk_conf
->meta_dev
,
1576 retcode
= ERR_OPEN_MD_DISK
;
1579 nbc
->md_bdev
= bdev
;
1581 if ((nbc
->backing_bdev
== nbc
->md_bdev
) !=
1582 (new_disk_conf
->meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1583 new_disk_conf
->meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)) {
1584 retcode
= ERR_MD_IDX_INVALID
;
1588 resync_lru
= lc_create("resync", drbd_bm_ext_cache
,
1589 1, 61, sizeof(struct bm_extent
),
1590 offsetof(struct bm_extent
, lce
));
1592 retcode
= ERR_NOMEM
;
1596 /* Read our meta data super block early.
1597 * This also sets other on-disk offsets. */
1598 retcode
= drbd_md_read(device
, nbc
);
1599 if (retcode
!= NO_ERROR
)
1602 if (new_disk_conf
->al_extents
< DRBD_AL_EXTENTS_MIN
)
1603 new_disk_conf
->al_extents
= DRBD_AL_EXTENTS_MIN
;
1604 if (new_disk_conf
->al_extents
> drbd_al_extents_max(nbc
))
1605 new_disk_conf
->al_extents
= drbd_al_extents_max(nbc
);
1607 if (drbd_get_max_capacity(nbc
) < new_disk_conf
->disk_size
) {
1608 drbd_err(device
, "max capacity %llu smaller than disk size %llu\n",
1609 (unsigned long long) drbd_get_max_capacity(nbc
),
1610 (unsigned long long) new_disk_conf
->disk_size
);
1611 retcode
= ERR_DISK_TOO_SMALL
;
1615 if (new_disk_conf
->meta_dev_idx
< 0) {
1616 max_possible_sectors
= DRBD_MAX_SECTORS_FLEX
;
1617 /* at least one MB, otherwise it does not make sense */
1618 min_md_device_sectors
= (2<<10);
1620 max_possible_sectors
= DRBD_MAX_SECTORS
;
1621 min_md_device_sectors
= MD_128MB_SECT
* (new_disk_conf
->meta_dev_idx
+ 1);
1624 if (drbd_get_capacity(nbc
->md_bdev
) < min_md_device_sectors
) {
1625 retcode
= ERR_MD_DISK_TOO_SMALL
;
1626 drbd_warn(device
, "refusing attach: md-device too small, "
1627 "at least %llu sectors needed for this meta-disk type\n",
1628 (unsigned long long) min_md_device_sectors
);
1632 /* Make sure the new disk is big enough
1633 * (we may currently be R_PRIMARY with no local disk...) */
1634 if (drbd_get_max_capacity(nbc
) <
1635 drbd_get_capacity(device
->this_bdev
)) {
1636 retcode
= ERR_DISK_TOO_SMALL
;
1640 nbc
->known_size
= drbd_get_capacity(nbc
->backing_bdev
);
1642 if (nbc
->known_size
> max_possible_sectors
) {
1643 drbd_warn(device
, "==> truncating very big lower level device "
1644 "to currently maximum possible %llu sectors <==\n",
1645 (unsigned long long) max_possible_sectors
);
1646 if (new_disk_conf
->meta_dev_idx
>= 0)
1647 drbd_warn(device
, "==>> using internal or flexible "
1648 "meta data may help <<==\n");
1651 drbd_suspend_io(device
);
1652 /* also wait for the last barrier ack. */
1653 /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171
1654 * We need a way to either ignore barrier acks for barriers sent before a device
1655 * was attached, or a way to wait for all pending barrier acks to come in.
1656 * As barriers are counted per resource,
1657 * we'd need to suspend io on all devices of a resource.
1659 wait_event(device
->misc_wait
, !atomic_read(&device
->ap_pending_cnt
) || drbd_suspended(device
));
1660 /* and for any other previously queued work */
1661 drbd_flush_workqueue(&connection
->sender_work
);
1663 rv
= _drbd_request_state(device
, NS(disk
, D_ATTACHING
), CS_VERBOSE
);
1664 retcode
= rv
; /* FIXME: Type mismatch. */
1665 drbd_resume_io(device
);
1666 if (rv
< SS_SUCCESS
)
1669 if (!get_ldev_if_state(device
, D_ATTACHING
))
1670 goto force_diskless
;
1672 if (!device
->bitmap
) {
1673 if (drbd_bm_init(device
)) {
1674 retcode
= ERR_NOMEM
;
1675 goto force_diskless_dec
;
1679 if (device
->state
.conn
< C_CONNECTED
&&
1680 device
->state
.role
== R_PRIMARY
&& device
->ed_uuid
&&
1681 (device
->ed_uuid
& ~((u64
)1)) != (nbc
->md
.uuid
[UI_CURRENT
] & ~((u64
)1))) {
1682 drbd_err(device
, "Can only attach to data with current UUID=%016llX\n",
1683 (unsigned long long)device
->ed_uuid
);
1684 retcode
= ERR_DATA_NOT_CURRENT
;
1685 goto force_diskless_dec
;
1688 /* Since we are diskless, fix the activity log first... */
1689 if (drbd_check_al_size(device
, new_disk_conf
)) {
1690 retcode
= ERR_NOMEM
;
1691 goto force_diskless_dec
;
1694 /* Prevent shrinking of consistent devices ! */
1695 if (drbd_md_test_flag(nbc
, MDF_CONSISTENT
) &&
1696 drbd_new_dev_size(device
, nbc
, nbc
->disk_conf
->disk_size
, 0) < nbc
->md
.la_size_sect
) {
1697 drbd_warn(device
, "refusing to truncate a consistent device\n");
1698 retcode
= ERR_DISK_TOO_SMALL
;
1699 goto force_diskless_dec
;
1702 /* Reset the "barriers don't work" bits here, then force meta data to
1703 * be written, to ensure we determine if barriers are supported. */
1704 if (new_disk_conf
->md_flushes
)
1705 clear_bit(MD_NO_FUA
, &device
->flags
);
1707 set_bit(MD_NO_FUA
, &device
->flags
);
1709 /* Point of no return reached.
1710 * Devices and memory are no longer released by error cleanup below.
1711 * now device takes over responsibility, and the state engine should
1712 * clean it up somewhere. */
1713 D_ASSERT(device
, device
->ldev
== NULL
);
1715 device
->resync
= resync_lru
;
1716 device
->rs_plan_s
= new_plan
;
1719 new_disk_conf
= NULL
;
1722 drbd_bump_write_ordering(device
->resource
, device
->ldev
, WO_bdev_flush
);
1724 if (drbd_md_test_flag(device
->ldev
, MDF_CRASHED_PRIMARY
))
1725 set_bit(CRASHED_PRIMARY
, &device
->flags
);
1727 clear_bit(CRASHED_PRIMARY
, &device
->flags
);
1729 if (drbd_md_test_flag(device
->ldev
, MDF_PRIMARY_IND
) &&
1730 !(device
->state
.role
== R_PRIMARY
&& device
->resource
->susp_nod
))
1731 set_bit(CRASHED_PRIMARY
, &device
->flags
);
1733 device
->send_cnt
= 0;
1734 device
->recv_cnt
= 0;
1735 device
->read_cnt
= 0;
1736 device
->writ_cnt
= 0;
1738 drbd_reconsider_max_bio_size(device
, device
->ldev
);
1740 /* If I am currently not R_PRIMARY,
1741 * but meta data primary indicator is set,
1742 * I just now recover from a hard crash,
1743 * and have been R_PRIMARY before that crash.
1745 * Now, if I had no connection before that crash
1746 * (have been degraded R_PRIMARY), chances are that
1747 * I won't find my peer now either.
1749 * In that case, and _only_ in that case,
1750 * we use the degr-wfc-timeout instead of the default,
1751 * so we can automatically recover from a crash of a
1752 * degraded but active "cluster" after a certain timeout.
1754 clear_bit(USE_DEGR_WFC_T
, &device
->flags
);
1755 if (device
->state
.role
!= R_PRIMARY
&&
1756 drbd_md_test_flag(device
->ldev
, MDF_PRIMARY_IND
) &&
1757 !drbd_md_test_flag(device
->ldev
, MDF_CONNECTED_IND
))
1758 set_bit(USE_DEGR_WFC_T
, &device
->flags
);
1760 dd
= drbd_determine_dev_size(device
, 0, NULL
);
1761 if (dd
<= DS_ERROR
) {
1762 retcode
= ERR_NOMEM_BITMAP
;
1763 goto force_diskless_dec
;
1764 } else if (dd
== DS_GREW
)
1765 set_bit(RESYNC_AFTER_NEG
, &device
->flags
);
1767 if (drbd_md_test_flag(device
->ldev
, MDF_FULL_SYNC
) ||
1768 (test_bit(CRASHED_PRIMARY
, &device
->flags
) &&
1769 drbd_md_test_flag(device
->ldev
, MDF_AL_DISABLED
))) {
1770 drbd_info(device
, "Assuming that all blocks are out of sync "
1771 "(aka FullSync)\n");
1772 if (drbd_bitmap_io(device
, &drbd_bmio_set_n_write
,
1773 "set_n_write from attaching", BM_LOCKED_MASK
)) {
1774 retcode
= ERR_IO_MD_DISK
;
1775 goto force_diskless_dec
;
1778 if (drbd_bitmap_io(device
, &drbd_bm_read
,
1779 "read from attaching", BM_LOCKED_MASK
)) {
1780 retcode
= ERR_IO_MD_DISK
;
1781 goto force_diskless_dec
;
1785 if (_drbd_bm_total_weight(device
) == drbd_bm_bits(device
))
1786 drbd_suspend_al(device
); /* IO is still suspended here... */
1788 spin_lock_irq(&device
->resource
->req_lock
);
1789 os
= drbd_read_state(device
);
1791 /* If MDF_CONSISTENT is not set go into inconsistent state,
1792 otherwise investigate MDF_WasUpToDate...
1793 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1794 otherwise into D_CONSISTENT state.
1796 if (drbd_md_test_flag(device
->ldev
, MDF_CONSISTENT
)) {
1797 if (drbd_md_test_flag(device
->ldev
, MDF_WAS_UP_TO_DATE
))
1798 ns
.disk
= D_CONSISTENT
;
1800 ns
.disk
= D_OUTDATED
;
1802 ns
.disk
= D_INCONSISTENT
;
1805 if (drbd_md_test_flag(device
->ldev
, MDF_PEER_OUT_DATED
))
1806 ns
.pdsk
= D_OUTDATED
;
1809 if (ns
.disk
== D_CONSISTENT
&&
1810 (ns
.pdsk
== D_OUTDATED
|| rcu_dereference(device
->ldev
->disk_conf
)->fencing
== FP_DONT_CARE
))
1811 ns
.disk
= D_UP_TO_DATE
;
1813 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1814 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1815 this point, because drbd_request_state() modifies these
1818 if (rcu_dereference(device
->ldev
->disk_conf
)->al_updates
)
1819 device
->ldev
->md
.flags
&= ~MDF_AL_DISABLED
;
1821 device
->ldev
->md
.flags
|= MDF_AL_DISABLED
;
1825 /* In case we are C_CONNECTED postpone any decision on the new disk
1826 state after the negotiation phase. */
1827 if (device
->state
.conn
== C_CONNECTED
) {
1828 device
->new_state_tmp
.i
= ns
.i
;
1830 ns
.disk
= D_NEGOTIATING
;
1832 /* We expect to receive up-to-date UUIDs soon.
1833 To avoid a race in receive_state, free p_uuid while
1834 holding req_lock. I.e. atomic with the state change */
1835 kfree(device
->p_uuid
);
1836 device
->p_uuid
= NULL
;
1839 rv
= _drbd_set_state(device
, ns
, CS_VERBOSE
, NULL
);
1840 spin_unlock_irq(&device
->resource
->req_lock
);
1842 if (rv
< SS_SUCCESS
)
1843 goto force_diskless_dec
;
1845 mod_timer(&device
->request_timer
, jiffies
+ HZ
);
1847 if (device
->state
.role
== R_PRIMARY
)
1848 device
->ldev
->md
.uuid
[UI_CURRENT
] |= (u64
)1;
1850 device
->ldev
->md
.uuid
[UI_CURRENT
] &= ~(u64
)1;
1852 drbd_md_mark_dirty(device
);
1853 drbd_md_sync(device
);
1855 kobject_uevent(&disk_to_dev(device
->vdisk
)->kobj
, KOBJ_CHANGE
);
1857 conn_reconfig_done(connection
);
1858 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
1859 drbd_adm_finish(&adm_ctx
, info
, retcode
);
1865 drbd_force_state(device
, NS(disk
, D_DISKLESS
));
1866 drbd_md_sync(device
);
1868 conn_reconfig_done(connection
);
1870 if (nbc
->backing_bdev
)
1871 blkdev_put(nbc
->backing_bdev
,
1872 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
);
1874 blkdev_put(nbc
->md_bdev
,
1875 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
);
1878 kfree(new_disk_conf
);
1879 lc_destroy(resync_lru
);
1881 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
1883 drbd_adm_finish(&adm_ctx
, info
, retcode
);
1887 static int adm_detach(struct drbd_device
*device
, int force
)
1889 enum drbd_state_rv retcode
;
1893 set_bit(FORCE_DETACH
, &device
->flags
);
1894 drbd_force_state(device
, NS(disk
, D_FAILED
));
1895 retcode
= SS_SUCCESS
;
1899 drbd_suspend_io(device
); /* so no-one is stuck in drbd_al_begin_io */
1900 drbd_md_get_buffer(device
); /* make sure there is no in-flight meta-data IO */
1901 retcode
= drbd_request_state(device
, NS(disk
, D_FAILED
));
1902 drbd_md_put_buffer(device
);
1903 /* D_FAILED will transition to DISKLESS. */
1904 ret
= wait_event_interruptible(device
->misc_wait
,
1905 device
->state
.disk
!= D_FAILED
);
1906 drbd_resume_io(device
);
1907 if ((int)retcode
== (int)SS_IS_DISKLESS
)
1908 retcode
= SS_NOTHING_TO_DO
;
1915 /* Detaching the disk is a process in multiple stages. First we need to lock
1916 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1917 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1918 * internal references as well.
1919 * Only then we have finally detached. */
1920 int drbd_adm_detach(struct sk_buff
*skb
, struct genl_info
*info
)
1922 struct drbd_config_context adm_ctx
;
1923 enum drbd_ret_code retcode
;
1924 struct detach_parms parms
= { };
1927 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
1928 if (!adm_ctx
.reply_skb
)
1930 if (retcode
!= NO_ERROR
)
1933 if (info
->attrs
[DRBD_NLA_DETACH_PARMS
]) {
1934 err
= detach_parms_from_attrs(&parms
, info
);
1936 retcode
= ERR_MANDATORY_TAG
;
1937 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
1942 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
1943 retcode
= adm_detach(adm_ctx
.device
, parms
.force_detach
);
1944 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
1946 drbd_adm_finish(&adm_ctx
, info
, retcode
);
1950 static bool conn_resync_running(struct drbd_connection
*connection
)
1952 struct drbd_peer_device
*peer_device
;
1957 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
) {
1958 struct drbd_device
*device
= peer_device
->device
;
1959 if (device
->state
.conn
== C_SYNC_SOURCE
||
1960 device
->state
.conn
== C_SYNC_TARGET
||
1961 device
->state
.conn
== C_PAUSED_SYNC_S
||
1962 device
->state
.conn
== C_PAUSED_SYNC_T
) {
1972 static bool conn_ov_running(struct drbd_connection
*connection
)
1974 struct drbd_peer_device
*peer_device
;
1979 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
) {
1980 struct drbd_device
*device
= peer_device
->device
;
1981 if (device
->state
.conn
== C_VERIFY_S
||
1982 device
->state
.conn
== C_VERIFY_T
) {
1992 static enum drbd_ret_code
1993 _check_net_options(struct drbd_connection
*connection
, struct net_conf
*old_net_conf
, struct net_conf
*new_net_conf
)
1995 struct drbd_peer_device
*peer_device
;
1998 if (old_net_conf
&& connection
->cstate
== C_WF_REPORT_PARAMS
&& connection
->agreed_pro_version
< 100) {
1999 if (new_net_conf
->wire_protocol
!= old_net_conf
->wire_protocol
)
2000 return ERR_NEED_APV_100
;
2002 if (new_net_conf
->two_primaries
!= old_net_conf
->two_primaries
)
2003 return ERR_NEED_APV_100
;
2005 if (strcmp(new_net_conf
->integrity_alg
, old_net_conf
->integrity_alg
))
2006 return ERR_NEED_APV_100
;
2009 if (!new_net_conf
->two_primaries
&&
2010 conn_highest_role(connection
) == R_PRIMARY
&&
2011 conn_highest_peer(connection
) == R_PRIMARY
)
2012 return ERR_NEED_ALLOW_TWO_PRI
;
2014 if (new_net_conf
->two_primaries
&&
2015 (new_net_conf
->wire_protocol
!= DRBD_PROT_C
))
2016 return ERR_NOT_PROTO_C
;
2018 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
2019 struct drbd_device
*device
= peer_device
->device
;
2020 if (get_ldev(device
)) {
2021 enum drbd_fencing_p fp
= rcu_dereference(device
->ldev
->disk_conf
)->fencing
;
2023 if (new_net_conf
->wire_protocol
== DRBD_PROT_A
&& fp
== FP_STONITH
)
2024 return ERR_STONITH_AND_PROT_A
;
2026 if (device
->state
.role
== R_PRIMARY
&& new_net_conf
->discard_my_data
)
2027 return ERR_DISCARD_IMPOSSIBLE
;
2030 if (new_net_conf
->on_congestion
!= OC_BLOCK
&& new_net_conf
->wire_protocol
!= DRBD_PROT_A
)
2031 return ERR_CONG_NOT_PROTO_A
;
2036 static enum drbd_ret_code
2037 check_net_options(struct drbd_connection
*connection
, struct net_conf
*new_net_conf
)
2039 static enum drbd_ret_code rv
;
2040 struct drbd_peer_device
*peer_device
;
2044 rv
= _check_net_options(connection
, rcu_dereference(connection
->net_conf
), new_net_conf
);
2047 /* connection->volumes protected by genl_lock() here */
2048 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
2049 struct drbd_device
*device
= peer_device
->device
;
2050 if (!device
->bitmap
) {
2051 if (drbd_bm_init(device
))
2060 struct crypto_hash
*verify_tfm
;
2061 struct crypto_hash
*csums_tfm
;
2062 struct crypto_hash
*cram_hmac_tfm
;
2063 struct crypto_hash
*integrity_tfm
;
2067 alloc_hash(struct crypto_hash
**tfm
, char *tfm_name
, int err_alg
)
2072 *tfm
= crypto_alloc_hash(tfm_name
, 0, CRYPTO_ALG_ASYNC
);
2081 static enum drbd_ret_code
2082 alloc_crypto(struct crypto
*crypto
, struct net_conf
*new_net_conf
)
2084 char hmac_name
[CRYPTO_MAX_ALG_NAME
];
2085 enum drbd_ret_code rv
;
2087 rv
= alloc_hash(&crypto
->csums_tfm
, new_net_conf
->csums_alg
,
2091 rv
= alloc_hash(&crypto
->verify_tfm
, new_net_conf
->verify_alg
,
2095 rv
= alloc_hash(&crypto
->integrity_tfm
, new_net_conf
->integrity_alg
,
2099 if (new_net_conf
->cram_hmac_alg
[0] != 0) {
2100 snprintf(hmac_name
, CRYPTO_MAX_ALG_NAME
, "hmac(%s)",
2101 new_net_conf
->cram_hmac_alg
);
2103 rv
= alloc_hash(&crypto
->cram_hmac_tfm
, hmac_name
,
2110 static void free_crypto(struct crypto
*crypto
)
2112 crypto_free_hash(crypto
->cram_hmac_tfm
);
2113 crypto_free_hash(crypto
->integrity_tfm
);
2114 crypto_free_hash(crypto
->csums_tfm
);
2115 crypto_free_hash(crypto
->verify_tfm
);
2118 int drbd_adm_net_opts(struct sk_buff
*skb
, struct genl_info
*info
)
2120 struct drbd_config_context adm_ctx
;
2121 enum drbd_ret_code retcode
;
2122 struct drbd_connection
*connection
;
2123 struct net_conf
*old_net_conf
, *new_net_conf
= NULL
;
2125 int ovr
; /* online verify running */
2126 int rsr
; /* re-sync running */
2127 struct crypto crypto
= { };
2129 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_CONNECTION
);
2130 if (!adm_ctx
.reply_skb
)
2132 if (retcode
!= NO_ERROR
)
2135 connection
= adm_ctx
.connection
;
2136 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2138 new_net_conf
= kzalloc(sizeof(struct net_conf
), GFP_KERNEL
);
2139 if (!new_net_conf
) {
2140 retcode
= ERR_NOMEM
;
2144 conn_reconfig_start(connection
);
2146 mutex_lock(&connection
->data
.mutex
);
2147 mutex_lock(&connection
->resource
->conf_update
);
2148 old_net_conf
= connection
->net_conf
;
2150 if (!old_net_conf
) {
2151 drbd_msg_put_info(adm_ctx
.reply_skb
, "net conf missing, try connect");
2152 retcode
= ERR_INVALID_REQUEST
;
2156 *new_net_conf
= *old_net_conf
;
2157 if (should_set_defaults(info
))
2158 set_net_conf_defaults(new_net_conf
);
2160 err
= net_conf_from_attrs_for_change(new_net_conf
, info
);
2161 if (err
&& err
!= -ENOMSG
) {
2162 retcode
= ERR_MANDATORY_TAG
;
2163 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
2167 retcode
= check_net_options(connection
, new_net_conf
);
2168 if (retcode
!= NO_ERROR
)
2171 /* re-sync running */
2172 rsr
= conn_resync_running(connection
);
2173 if (rsr
&& strcmp(new_net_conf
->csums_alg
, old_net_conf
->csums_alg
)) {
2174 retcode
= ERR_CSUMS_RESYNC_RUNNING
;
2178 /* online verify running */
2179 ovr
= conn_ov_running(connection
);
2180 if (ovr
&& strcmp(new_net_conf
->verify_alg
, old_net_conf
->verify_alg
)) {
2181 retcode
= ERR_VERIFY_RUNNING
;
2185 retcode
= alloc_crypto(&crypto
, new_net_conf
);
2186 if (retcode
!= NO_ERROR
)
2189 rcu_assign_pointer(connection
->net_conf
, new_net_conf
);
2192 crypto_free_hash(connection
->csums_tfm
);
2193 connection
->csums_tfm
= crypto
.csums_tfm
;
2194 crypto
.csums_tfm
= NULL
;
2197 crypto_free_hash(connection
->verify_tfm
);
2198 connection
->verify_tfm
= crypto
.verify_tfm
;
2199 crypto
.verify_tfm
= NULL
;
2202 crypto_free_hash(connection
->integrity_tfm
);
2203 connection
->integrity_tfm
= crypto
.integrity_tfm
;
2204 if (connection
->cstate
>= C_WF_REPORT_PARAMS
&& connection
->agreed_pro_version
>= 100)
2205 /* Do this without trying to take connection->data.mutex again. */
2206 __drbd_send_protocol(connection
, P_PROTOCOL_UPDATE
);
2208 crypto_free_hash(connection
->cram_hmac_tfm
);
2209 connection
->cram_hmac_tfm
= crypto
.cram_hmac_tfm
;
2211 mutex_unlock(&connection
->resource
->conf_update
);
2212 mutex_unlock(&connection
->data
.mutex
);
2214 kfree(old_net_conf
);
2216 if (connection
->cstate
>= C_WF_REPORT_PARAMS
) {
2217 struct drbd_peer_device
*peer_device
;
2220 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
)
2221 drbd_send_sync_param(peer_device
);
2227 mutex_unlock(&connection
->resource
->conf_update
);
2228 mutex_unlock(&connection
->data
.mutex
);
2229 free_crypto(&crypto
);
2230 kfree(new_net_conf
);
2232 conn_reconfig_done(connection
);
2234 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2236 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2240 int drbd_adm_connect(struct sk_buff
*skb
, struct genl_info
*info
)
2242 struct drbd_config_context adm_ctx
;
2243 struct drbd_peer_device
*peer_device
;
2244 struct net_conf
*old_net_conf
, *new_net_conf
= NULL
;
2245 struct crypto crypto
= { };
2246 struct drbd_resource
*resource
;
2247 struct drbd_connection
*connection
;
2248 enum drbd_ret_code retcode
;
2252 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_RESOURCE
);
2254 if (!adm_ctx
.reply_skb
)
2256 if (retcode
!= NO_ERROR
)
2258 if (!(adm_ctx
.my_addr
&& adm_ctx
.peer_addr
)) {
2259 drbd_msg_put_info(adm_ctx
.reply_skb
, "connection endpoint(s) missing");
2260 retcode
= ERR_INVALID_REQUEST
;
2264 /* No need for _rcu here. All reconfiguration is
2265 * strictly serialized on genl_lock(). We are protected against
2266 * concurrent reconfiguration/addition/deletion */
2267 for_each_resource(resource
, &drbd_resources
) {
2268 for_each_connection(connection
, resource
) {
2269 if (nla_len(adm_ctx
.my_addr
) == connection
->my_addr_len
&&
2270 !memcmp(nla_data(adm_ctx
.my_addr
), &connection
->my_addr
,
2271 connection
->my_addr_len
)) {
2272 retcode
= ERR_LOCAL_ADDR
;
2276 if (nla_len(adm_ctx
.peer_addr
) == connection
->peer_addr_len
&&
2277 !memcmp(nla_data(adm_ctx
.peer_addr
), &connection
->peer_addr
,
2278 connection
->peer_addr_len
)) {
2279 retcode
= ERR_PEER_ADDR
;
2285 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2286 connection
= first_connection(adm_ctx
.resource
);
2287 conn_reconfig_start(connection
);
2289 if (connection
->cstate
> C_STANDALONE
) {
2290 retcode
= ERR_NET_CONFIGURED
;
2294 /* allocation not in the IO path, drbdsetup / netlink process context */
2295 new_net_conf
= kzalloc(sizeof(*new_net_conf
), GFP_KERNEL
);
2296 if (!new_net_conf
) {
2297 retcode
= ERR_NOMEM
;
2301 set_net_conf_defaults(new_net_conf
);
2303 err
= net_conf_from_attrs(new_net_conf
, info
);
2304 if (err
&& err
!= -ENOMSG
) {
2305 retcode
= ERR_MANDATORY_TAG
;
2306 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
2310 retcode
= check_net_options(connection
, new_net_conf
);
2311 if (retcode
!= NO_ERROR
)
2314 retcode
= alloc_crypto(&crypto
, new_net_conf
);
2315 if (retcode
!= NO_ERROR
)
2318 ((char *)new_net_conf
->shared_secret
)[SHARED_SECRET_MAX
-1] = 0;
2320 drbd_flush_workqueue(&connection
->sender_work
);
2322 mutex_lock(&adm_ctx
.resource
->conf_update
);
2323 old_net_conf
= connection
->net_conf
;
2325 retcode
= ERR_NET_CONFIGURED
;
2326 mutex_unlock(&adm_ctx
.resource
->conf_update
);
2329 rcu_assign_pointer(connection
->net_conf
, new_net_conf
);
2331 conn_free_crypto(connection
);
2332 connection
->cram_hmac_tfm
= crypto
.cram_hmac_tfm
;
2333 connection
->integrity_tfm
= crypto
.integrity_tfm
;
2334 connection
->csums_tfm
= crypto
.csums_tfm
;
2335 connection
->verify_tfm
= crypto
.verify_tfm
;
2337 connection
->my_addr_len
= nla_len(adm_ctx
.my_addr
);
2338 memcpy(&connection
->my_addr
, nla_data(adm_ctx
.my_addr
), connection
->my_addr_len
);
2339 connection
->peer_addr_len
= nla_len(adm_ctx
.peer_addr
);
2340 memcpy(&connection
->peer_addr
, nla_data(adm_ctx
.peer_addr
), connection
->peer_addr_len
);
2342 mutex_unlock(&adm_ctx
.resource
->conf_update
);
2345 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
2346 struct drbd_device
*device
= peer_device
->device
;
2347 device
->send_cnt
= 0;
2348 device
->recv_cnt
= 0;
2352 retcode
= conn_request_state(connection
, NS(conn
, C_UNCONNECTED
), CS_VERBOSE
);
2354 conn_reconfig_done(connection
);
2355 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2356 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2360 free_crypto(&crypto
);
2361 kfree(new_net_conf
);
2363 conn_reconfig_done(connection
);
2364 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2366 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2370 static enum drbd_state_rv
conn_try_disconnect(struct drbd_connection
*connection
, bool force
)
2372 enum drbd_state_rv rv
;
2374 rv
= conn_request_state(connection
, NS(conn
, C_DISCONNECTING
),
2375 force
? CS_HARD
: 0);
2378 case SS_NOTHING_TO_DO
:
2380 case SS_ALREADY_STANDALONE
:
2382 case SS_PRIMARY_NOP
:
2383 /* Our state checking code wants to see the peer outdated. */
2384 rv
= conn_request_state(connection
, NS2(conn
, C_DISCONNECTING
, pdsk
, D_OUTDATED
), 0);
2386 if (rv
== SS_OUTDATE_WO_CONN
) /* lost connection before graceful disconnect succeeded */
2387 rv
= conn_request_state(connection
, NS(conn
, C_DISCONNECTING
), CS_VERBOSE
);
2390 case SS_CW_FAILED_BY_PEER
:
2391 /* The peer probably wants to see us outdated. */
2392 rv
= conn_request_state(connection
, NS2(conn
, C_DISCONNECTING
,
2393 disk
, D_OUTDATED
), 0);
2394 if (rv
== SS_IS_DISKLESS
|| rv
== SS_LOWER_THAN_OUTDATED
) {
2395 rv
= conn_request_state(connection
, NS(conn
, C_DISCONNECTING
),
2400 /* no special handling necessary */
2403 if (rv
>= SS_SUCCESS
) {
2404 enum drbd_state_rv rv2
;
2405 /* No one else can reconfigure the network while I am here.
2406 * The state handling only uses drbd_thread_stop_nowait(),
2407 * we want to really wait here until the receiver is no more.
2409 drbd_thread_stop(&connection
->receiver
);
2411 /* Race breaker. This additional state change request may be
2412 * necessary, if this was a forced disconnect during a receiver
2413 * restart. We may have "killed" the receiver thread just
2414 * after drbd_receiver() returned. Typically, we should be
2415 * C_STANDALONE already, now, and this becomes a no-op.
2417 rv2
= conn_request_state(connection
, NS(conn
, C_STANDALONE
),
2418 CS_VERBOSE
| CS_HARD
);
2419 if (rv2
< SS_SUCCESS
)
2420 drbd_err(connection
,
2421 "unexpected rv2=%d in conn_try_disconnect()\n",
2427 int drbd_adm_disconnect(struct sk_buff
*skb
, struct genl_info
*info
)
2429 struct drbd_config_context adm_ctx
;
2430 struct disconnect_parms parms
;
2431 struct drbd_connection
*connection
;
2432 enum drbd_state_rv rv
;
2433 enum drbd_ret_code retcode
;
2436 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_CONNECTION
);
2437 if (!adm_ctx
.reply_skb
)
2439 if (retcode
!= NO_ERROR
)
2442 connection
= adm_ctx
.connection
;
2443 memset(&parms
, 0, sizeof(parms
));
2444 if (info
->attrs
[DRBD_NLA_DISCONNECT_PARMS
]) {
2445 err
= disconnect_parms_from_attrs(&parms
, info
);
2447 retcode
= ERR_MANDATORY_TAG
;
2448 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
2453 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2454 rv
= conn_try_disconnect(connection
, parms
.force_disconnect
);
2455 if (rv
< SS_SUCCESS
)
2456 retcode
= rv
; /* FIXME: Type mismatch. */
2459 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2461 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2465 void resync_after_online_grow(struct drbd_device
*device
)
2467 int iass
; /* I am sync source */
2469 drbd_info(device
, "Resync of new storage after online grow\n");
2470 if (device
->state
.role
!= device
->state
.peer
)
2471 iass
= (device
->state
.role
== R_PRIMARY
);
2473 iass
= test_bit(RESOLVE_CONFLICTS
, &first_peer_device(device
)->connection
->flags
);
2476 drbd_start_resync(device
, C_SYNC_SOURCE
);
2478 _drbd_request_state(device
, NS(conn
, C_WF_SYNC_UUID
), CS_VERBOSE
+ CS_SERIALIZE
);
2481 int drbd_adm_resize(struct sk_buff
*skb
, struct genl_info
*info
)
2483 struct drbd_config_context adm_ctx
;
2484 struct disk_conf
*old_disk_conf
, *new_disk_conf
= NULL
;
2485 struct resize_parms rs
;
2486 struct drbd_device
*device
;
2487 enum drbd_ret_code retcode
;
2488 enum determine_dev_size dd
;
2489 bool change_al_layout
= false;
2490 enum dds_flags ddsf
;
2494 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2495 if (!adm_ctx
.reply_skb
)
2497 if (retcode
!= NO_ERROR
)
2500 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2501 device
= adm_ctx
.device
;
2502 if (!get_ldev(device
)) {
2503 retcode
= ERR_NO_DISK
;
2507 memset(&rs
, 0, sizeof(struct resize_parms
));
2508 rs
.al_stripes
= device
->ldev
->md
.al_stripes
;
2509 rs
.al_stripe_size
= device
->ldev
->md
.al_stripe_size_4k
* 4;
2510 if (info
->attrs
[DRBD_NLA_RESIZE_PARMS
]) {
2511 err
= resize_parms_from_attrs(&rs
, info
);
2513 retcode
= ERR_MANDATORY_TAG
;
2514 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
2519 if (device
->state
.conn
> C_CONNECTED
) {
2520 retcode
= ERR_RESIZE_RESYNC
;
2524 if (device
->state
.role
== R_SECONDARY
&&
2525 device
->state
.peer
== R_SECONDARY
) {
2526 retcode
= ERR_NO_PRIMARY
;
2530 if (rs
.no_resync
&& first_peer_device(device
)->connection
->agreed_pro_version
< 93) {
2531 retcode
= ERR_NEED_APV_93
;
2536 u_size
= rcu_dereference(device
->ldev
->disk_conf
)->disk_size
;
2538 if (u_size
!= (sector_t
)rs
.resize_size
) {
2539 new_disk_conf
= kmalloc(sizeof(struct disk_conf
), GFP_KERNEL
);
2540 if (!new_disk_conf
) {
2541 retcode
= ERR_NOMEM
;
2546 if (device
->ldev
->md
.al_stripes
!= rs
.al_stripes
||
2547 device
->ldev
->md
.al_stripe_size_4k
!= rs
.al_stripe_size
/ 4) {
2548 u32 al_size_k
= rs
.al_stripes
* rs
.al_stripe_size
;
2550 if (al_size_k
> (16 * 1024 * 1024)) {
2551 retcode
= ERR_MD_LAYOUT_TOO_BIG
;
2555 if (al_size_k
< MD_32kB_SECT
/2) {
2556 retcode
= ERR_MD_LAYOUT_TOO_SMALL
;
2560 if (device
->state
.conn
!= C_CONNECTED
&& !rs
.resize_force
) {
2561 retcode
= ERR_MD_LAYOUT_CONNECTED
;
2565 change_al_layout
= true;
2568 if (device
->ldev
->known_size
!= drbd_get_capacity(device
->ldev
->backing_bdev
))
2569 device
->ldev
->known_size
= drbd_get_capacity(device
->ldev
->backing_bdev
);
2571 if (new_disk_conf
) {
2572 mutex_lock(&device
->resource
->conf_update
);
2573 old_disk_conf
= device
->ldev
->disk_conf
;
2574 *new_disk_conf
= *old_disk_conf
;
2575 new_disk_conf
->disk_size
= (sector_t
)rs
.resize_size
;
2576 rcu_assign_pointer(device
->ldev
->disk_conf
, new_disk_conf
);
2577 mutex_unlock(&device
->resource
->conf_update
);
2579 kfree(old_disk_conf
);
2582 ddsf
= (rs
.resize_force
? DDSF_FORCED
: 0) | (rs
.no_resync
? DDSF_NO_RESYNC
: 0);
2583 dd
= drbd_determine_dev_size(device
, ddsf
, change_al_layout
? &rs
: NULL
);
2584 drbd_md_sync(device
);
2586 if (dd
== DS_ERROR
) {
2587 retcode
= ERR_NOMEM_BITMAP
;
2589 } else if (dd
== DS_ERROR_SPACE_MD
) {
2590 retcode
= ERR_MD_LAYOUT_NO_FIT
;
2592 } else if (dd
== DS_ERROR_SHRINK
) {
2593 retcode
= ERR_IMPLICIT_SHRINK
;
2597 if (device
->state
.conn
== C_CONNECTED
) {
2599 set_bit(RESIZE_PENDING
, &device
->flags
);
2601 drbd_send_uuids(first_peer_device(device
));
2602 drbd_send_sizes(first_peer_device(device
), 1, ddsf
);
2606 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2608 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2616 int drbd_adm_resource_opts(struct sk_buff
*skb
, struct genl_info
*info
)
2618 struct drbd_config_context adm_ctx
;
2619 enum drbd_ret_code retcode
;
2620 struct res_opts res_opts
;
2623 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_RESOURCE
);
2624 if (!adm_ctx
.reply_skb
)
2626 if (retcode
!= NO_ERROR
)
2629 res_opts
= adm_ctx
.resource
->res_opts
;
2630 if (should_set_defaults(info
))
2631 set_res_opts_defaults(&res_opts
);
2633 err
= res_opts_from_attrs(&res_opts
, info
);
2634 if (err
&& err
!= -ENOMSG
) {
2635 retcode
= ERR_MANDATORY_TAG
;
2636 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
2640 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2641 err
= set_resource_options(adm_ctx
.resource
, &res_opts
);
2643 retcode
= ERR_INVALID_REQUEST
;
2645 retcode
= ERR_NOMEM
;
2647 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2650 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2654 int drbd_adm_invalidate(struct sk_buff
*skb
, struct genl_info
*info
)
2656 struct drbd_config_context adm_ctx
;
2657 struct drbd_device
*device
;
2658 int retcode
; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2660 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2661 if (!adm_ctx
.reply_skb
)
2663 if (retcode
!= NO_ERROR
)
2666 device
= adm_ctx
.device
;
2667 if (!get_ldev(device
)) {
2668 retcode
= ERR_NO_DISK
;
2672 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2674 /* If there is still bitmap IO pending, probably because of a previous
2675 * resync just being finished, wait for it before requesting a new resync.
2676 * Also wait for it's after_state_ch(). */
2677 drbd_suspend_io(device
);
2678 wait_event(device
->misc_wait
, !test_bit(BITMAP_IO
, &device
->flags
));
2679 drbd_flush_workqueue(&first_peer_device(device
)->connection
->sender_work
);
2681 /* If we happen to be C_STANDALONE R_SECONDARY, just change to
2682 * D_INCONSISTENT, and set all bits in the bitmap. Otherwise,
2683 * try to start a resync handshake as sync target for full sync.
2685 if (device
->state
.conn
== C_STANDALONE
&& device
->state
.role
== R_SECONDARY
) {
2686 retcode
= drbd_request_state(device
, NS(disk
, D_INCONSISTENT
));
2687 if (retcode
>= SS_SUCCESS
) {
2688 if (drbd_bitmap_io(device
, &drbd_bmio_set_n_write
,
2689 "set_n_write from invalidate", BM_LOCKED_MASK
))
2690 retcode
= ERR_IO_MD_DISK
;
2693 retcode
= drbd_request_state(device
, NS(conn
, C_STARTING_SYNC_T
));
2694 drbd_resume_io(device
);
2695 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2698 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2702 static int drbd_adm_simple_request_state(struct sk_buff
*skb
, struct genl_info
*info
,
2703 union drbd_state mask
, union drbd_state val
)
2705 struct drbd_config_context adm_ctx
;
2706 enum drbd_ret_code retcode
;
2708 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2709 if (!adm_ctx
.reply_skb
)
2711 if (retcode
!= NO_ERROR
)
2714 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2715 retcode
= drbd_request_state(adm_ctx
.device
, mask
, val
);
2716 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2718 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2722 static int drbd_bmio_set_susp_al(struct drbd_device
*device
) __must_hold(local
)
2726 rv
= drbd_bmio_set_n_write(device
);
2727 drbd_suspend_al(device
);
2731 int drbd_adm_invalidate_peer(struct sk_buff
*skb
, struct genl_info
*info
)
2733 struct drbd_config_context adm_ctx
;
2734 int retcode
; /* drbd_ret_code, drbd_state_rv */
2735 struct drbd_device
*device
;
2737 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2738 if (!adm_ctx
.reply_skb
)
2740 if (retcode
!= NO_ERROR
)
2743 device
= adm_ctx
.device
;
2744 if (!get_ldev(device
)) {
2745 retcode
= ERR_NO_DISK
;
2749 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2751 /* If there is still bitmap IO pending, probably because of a previous
2752 * resync just being finished, wait for it before requesting a new resync.
2753 * Also wait for it's after_state_ch(). */
2754 drbd_suspend_io(device
);
2755 wait_event(device
->misc_wait
, !test_bit(BITMAP_IO
, &device
->flags
));
2756 drbd_flush_workqueue(&first_peer_device(device
)->connection
->sender_work
);
2758 /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits
2759 * in the bitmap. Otherwise, try to start a resync handshake
2760 * as sync source for full sync.
2762 if (device
->state
.conn
== C_STANDALONE
&& device
->state
.role
== R_PRIMARY
) {
2763 /* The peer will get a resync upon connect anyways. Just make that
2764 into a full resync. */
2765 retcode
= drbd_request_state(device
, NS(pdsk
, D_INCONSISTENT
));
2766 if (retcode
>= SS_SUCCESS
) {
2767 if (drbd_bitmap_io(device
, &drbd_bmio_set_susp_al
,
2768 "set_n_write from invalidate_peer",
2769 BM_LOCKED_SET_ALLOWED
))
2770 retcode
= ERR_IO_MD_DISK
;
2773 retcode
= drbd_request_state(device
, NS(conn
, C_STARTING_SYNC_S
));
2774 drbd_resume_io(device
);
2775 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2778 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2782 int drbd_adm_pause_sync(struct sk_buff
*skb
, struct genl_info
*info
)
2784 struct drbd_config_context adm_ctx
;
2785 enum drbd_ret_code retcode
;
2787 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2788 if (!adm_ctx
.reply_skb
)
2790 if (retcode
!= NO_ERROR
)
2793 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2794 if (drbd_request_state(adm_ctx
.device
, NS(user_isp
, 1)) == SS_NOTHING_TO_DO
)
2795 retcode
= ERR_PAUSE_IS_SET
;
2796 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2798 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2802 int drbd_adm_resume_sync(struct sk_buff
*skb
, struct genl_info
*info
)
2804 struct drbd_config_context adm_ctx
;
2805 union drbd_dev_state s
;
2806 enum drbd_ret_code retcode
;
2808 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2809 if (!adm_ctx
.reply_skb
)
2811 if (retcode
!= NO_ERROR
)
2814 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2815 if (drbd_request_state(adm_ctx
.device
, NS(user_isp
, 0)) == SS_NOTHING_TO_DO
) {
2816 s
= adm_ctx
.device
->state
;
2817 if (s
.conn
== C_PAUSED_SYNC_S
|| s
.conn
== C_PAUSED_SYNC_T
) {
2818 retcode
= s
.aftr_isp
? ERR_PIC_AFTER_DEP
:
2819 s
.peer_isp
? ERR_PIC_PEER_DEP
: ERR_PAUSE_IS_CLEAR
;
2821 retcode
= ERR_PAUSE_IS_CLEAR
;
2824 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2826 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2830 int drbd_adm_suspend_io(struct sk_buff
*skb
, struct genl_info
*info
)
2832 return drbd_adm_simple_request_state(skb
, info
, NS(susp
, 1));
2835 int drbd_adm_resume_io(struct sk_buff
*skb
, struct genl_info
*info
)
2837 struct drbd_config_context adm_ctx
;
2838 struct drbd_device
*device
;
2839 int retcode
; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2841 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
2842 if (!adm_ctx
.reply_skb
)
2844 if (retcode
!= NO_ERROR
)
2847 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
2848 device
= adm_ctx
.device
;
2849 if (test_bit(NEW_CUR_UUID
, &device
->flags
)) {
2850 drbd_uuid_new_current(device
);
2851 clear_bit(NEW_CUR_UUID
, &device
->flags
);
2853 drbd_suspend_io(device
);
2854 retcode
= drbd_request_state(device
, NS3(susp
, 0, susp_nod
, 0, susp_fen
, 0));
2855 if (retcode
== SS_SUCCESS
) {
2856 if (device
->state
.conn
< C_CONNECTED
)
2857 tl_clear(first_peer_device(device
)->connection
);
2858 if (device
->state
.disk
== D_DISKLESS
|| device
->state
.disk
== D_FAILED
)
2859 tl_restart(first_peer_device(device
)->connection
, FAIL_FROZEN_DISK_IO
);
2861 drbd_resume_io(device
);
2862 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
2864 drbd_adm_finish(&adm_ctx
, info
, retcode
);
2868 int drbd_adm_outdate(struct sk_buff
*skb
, struct genl_info
*info
)
2870 return drbd_adm_simple_request_state(skb
, info
, NS(disk
, D_OUTDATED
));
2873 static int nla_put_drbd_cfg_context(struct sk_buff
*skb
,
2874 struct drbd_resource
*resource
,
2875 struct drbd_connection
*connection
,
2876 struct drbd_device
*device
)
2879 nla
= nla_nest_start(skb
, DRBD_NLA_CFG_CONTEXT
);
2881 goto nla_put_failure
;
2883 nla_put_u32(skb
, T_ctx_volume
, device
->vnr
))
2884 goto nla_put_failure
;
2885 if (nla_put_string(skb
, T_ctx_resource_name
, resource
->name
))
2886 goto nla_put_failure
;
2888 if (connection
->my_addr_len
&&
2889 nla_put(skb
, T_ctx_my_addr
, connection
->my_addr_len
, &connection
->my_addr
))
2890 goto nla_put_failure
;
2891 if (connection
->peer_addr_len
&&
2892 nla_put(skb
, T_ctx_peer_addr
, connection
->peer_addr_len
, &connection
->peer_addr
))
2893 goto nla_put_failure
;
2895 nla_nest_end(skb
, nla
);
2900 nla_nest_cancel(skb
, nla
);
2905 * Return the connection of @resource if @resource has exactly one connection.
2907 static struct drbd_connection
*the_only_connection(struct drbd_resource
*resource
)
2909 struct list_head
*connections
= &resource
->connections
;
2911 if (list_empty(connections
) || connections
->next
->next
!= connections
)
2913 return list_first_entry(&resource
->connections
, struct drbd_connection
, connections
);
2916 int nla_put_status_info(struct sk_buff
*skb
, struct drbd_device
*device
,
2917 const struct sib_info
*sib
)
2919 struct drbd_resource
*resource
= device
->resource
;
2920 struct state_info
*si
= NULL
; /* for sizeof(si->member); */
2924 int exclude_sensitive
;
2926 /* If sib != NULL, this is drbd_bcast_event, which anyone can listen
2927 * to. So we better exclude_sensitive information.
2929 * If sib == NULL, this is drbd_adm_get_status, executed synchronously
2930 * in the context of the requesting user process. Exclude sensitive
2931 * information, unless current has superuser.
2933 * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and
2934 * relies on the current implementation of netlink_dump(), which
2935 * executes the dump callback successively from netlink_recvmsg(),
2936 * always in the context of the receiving process */
2937 exclude_sensitive
= sib
|| !capable(CAP_SYS_ADMIN
);
2939 got_ldev
= get_ldev(device
);
2941 /* We need to add connection name and volume number information still.
2942 * Minor number is in drbd_genlmsghdr. */
2943 if (nla_put_drbd_cfg_context(skb
, resource
, the_only_connection(resource
), device
))
2944 goto nla_put_failure
;
2946 if (res_opts_to_skb(skb
, &device
->resource
->res_opts
, exclude_sensitive
))
2947 goto nla_put_failure
;
2951 struct disk_conf
*disk_conf
;
2953 disk_conf
= rcu_dereference(device
->ldev
->disk_conf
);
2954 err
= disk_conf_to_skb(skb
, disk_conf
, exclude_sensitive
);
2957 struct net_conf
*nc
;
2959 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
2961 err
= net_conf_to_skb(skb
, nc
, exclude_sensitive
);
2965 goto nla_put_failure
;
2967 nla
= nla_nest_start(skb
, DRBD_NLA_STATE_INFO
);
2969 goto nla_put_failure
;
2970 if (nla_put_u32(skb
, T_sib_reason
, sib
? sib
->sib_reason
: SIB_GET_STATUS_REPLY
) ||
2971 nla_put_u32(skb
, T_current_state
, device
->state
.i
) ||
2972 nla_put_u64(skb
, T_ed_uuid
, device
->ed_uuid
) ||
2973 nla_put_u64(skb
, T_capacity
, drbd_get_capacity(device
->this_bdev
)) ||
2974 nla_put_u64(skb
, T_send_cnt
, device
->send_cnt
) ||
2975 nla_put_u64(skb
, T_recv_cnt
, device
->recv_cnt
) ||
2976 nla_put_u64(skb
, T_read_cnt
, device
->read_cnt
) ||
2977 nla_put_u64(skb
, T_writ_cnt
, device
->writ_cnt
) ||
2978 nla_put_u64(skb
, T_al_writ_cnt
, device
->al_writ_cnt
) ||
2979 nla_put_u64(skb
, T_bm_writ_cnt
, device
->bm_writ_cnt
) ||
2980 nla_put_u32(skb
, T_ap_bio_cnt
, atomic_read(&device
->ap_bio_cnt
)) ||
2981 nla_put_u32(skb
, T_ap_pending_cnt
, atomic_read(&device
->ap_pending_cnt
)) ||
2982 nla_put_u32(skb
, T_rs_pending_cnt
, atomic_read(&device
->rs_pending_cnt
)))
2983 goto nla_put_failure
;
2988 spin_lock_irq(&device
->ldev
->md
.uuid_lock
);
2989 err
= nla_put(skb
, T_uuids
, sizeof(si
->uuids
), device
->ldev
->md
.uuid
);
2990 spin_unlock_irq(&device
->ldev
->md
.uuid_lock
);
2993 goto nla_put_failure
;
2995 if (nla_put_u32(skb
, T_disk_flags
, device
->ldev
->md
.flags
) ||
2996 nla_put_u64(skb
, T_bits_total
, drbd_bm_bits(device
)) ||
2997 nla_put_u64(skb
, T_bits_oos
, drbd_bm_total_weight(device
)))
2998 goto nla_put_failure
;
2999 if (C_SYNC_SOURCE
<= device
->state
.conn
&&
3000 C_PAUSED_SYNC_T
>= device
->state
.conn
) {
3001 if (nla_put_u64(skb
, T_bits_rs_total
, device
->rs_total
) ||
3002 nla_put_u64(skb
, T_bits_rs_failed
, device
->rs_failed
))
3003 goto nla_put_failure
;
3008 switch(sib
->sib_reason
) {
3009 case SIB_SYNC_PROGRESS
:
3010 case SIB_GET_STATUS_REPLY
:
3012 case SIB_STATE_CHANGE
:
3013 if (nla_put_u32(skb
, T_prev_state
, sib
->os
.i
) ||
3014 nla_put_u32(skb
, T_new_state
, sib
->ns
.i
))
3015 goto nla_put_failure
;
3017 case SIB_HELPER_POST
:
3018 if (nla_put_u32(skb
, T_helper_exit_code
,
3019 sib
->helper_exit_code
))
3020 goto nla_put_failure
;
3022 case SIB_HELPER_PRE
:
3023 if (nla_put_string(skb
, T_helper
, sib
->helper_name
))
3024 goto nla_put_failure
;
3028 nla_nest_end(skb
, nla
);
3038 int drbd_adm_get_status(struct sk_buff
*skb
, struct genl_info
*info
)
3040 struct drbd_config_context adm_ctx
;
3041 enum drbd_ret_code retcode
;
3044 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
3045 if (!adm_ctx
.reply_skb
)
3047 if (retcode
!= NO_ERROR
)
3050 err
= nla_put_status_info(adm_ctx
.reply_skb
, adm_ctx
.device
, NULL
);
3052 nlmsg_free(adm_ctx
.reply_skb
);
3056 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3060 static int get_one_status(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3062 struct drbd_device
*device
;
3063 struct drbd_genlmsghdr
*dh
;
3064 struct drbd_resource
*pos
= (struct drbd_resource
*)cb
->args
[0];
3065 struct drbd_resource
*resource
= NULL
;
3066 struct drbd_resource
*tmp
;
3067 unsigned volume
= cb
->args
[1];
3069 /* Open coded, deferred, iteration:
3070 * for_each_resource_safe(resource, tmp, &drbd_resources) {
3071 * connection = "first connection of resource or undefined";
3072 * idr_for_each_entry(&resource->devices, device, i) {
3076 * where resource is cb->args[0];
3077 * and i is cb->args[1];
3079 * cb->args[2] indicates if we shall loop over all resources,
3080 * or just dump all volumes of a single resource.
3082 * This may miss entries inserted after this dump started,
3083 * or entries deleted before they are reached.
3085 * We need to make sure the device won't disappear while
3086 * we are looking at it, and revalidate our iterators
3087 * on each iteration.
3090 /* synchronize with conn_create()/drbd_destroy_connection() */
3092 /* revalidate iterator position */
3093 for_each_resource_rcu(tmp
, &drbd_resources
) {
3095 /* first iteration */
3107 device
= idr_get_next(&resource
->devices
, &volume
);
3109 /* No more volumes to dump on this resource.
3110 * Advance resource iterator. */
3111 pos
= list_entry_rcu(resource
->resources
.next
,
3112 struct drbd_resource
, resources
);
3113 /* Did we dump any volume of this resource yet? */
3115 /* If we reached the end of the list,
3116 * or only a single resource dump was requested,
3118 if (&pos
->resources
== &drbd_resources
|| cb
->args
[2])
3126 dh
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
,
3127 cb
->nlh
->nlmsg_seq
, &drbd_genl_family
,
3128 NLM_F_MULTI
, DRBD_ADM_GET_STATUS
);
3133 /* This is a connection without a single volume.
3134 * Suprisingly enough, it may have a network
3136 struct drbd_connection
*connection
;
3139 dh
->ret_code
= NO_ERROR
;
3140 connection
= the_only_connection(resource
);
3141 if (nla_put_drbd_cfg_context(skb
, resource
, connection
, NULL
))
3144 struct net_conf
*nc
;
3146 nc
= rcu_dereference(connection
->net_conf
);
3147 if (nc
&& net_conf_to_skb(skb
, nc
, 1) != 0)
3153 D_ASSERT(device
, device
->vnr
== volume
);
3154 D_ASSERT(device
, device
->resource
== resource
);
3156 dh
->minor
= device_to_minor(device
);
3157 dh
->ret_code
= NO_ERROR
;
3159 if (nla_put_status_info(skb
, device
, NULL
)) {
3161 genlmsg_cancel(skb
, dh
);
3165 genlmsg_end(skb
, dh
);
3170 /* where to start the next iteration */
3171 cb
->args
[0] = (long)pos
;
3172 cb
->args
[1] = (pos
== resource
) ? volume
+ 1 : 0;
3174 /* No more resources/volumes/minors found results in an empty skb.
3175 * Which will terminate the dump. */
3180 * Request status of all resources, or of all volumes within a single resource.
3182 * This is a dump, as the answer may not fit in a single reply skb otherwise.
3183 * Which means we cannot use the family->attrbuf or other such members, because
3184 * dump is NOT protected by the genl_lock(). During dump, we only have access
3185 * to the incoming skb, and need to opencode "parsing" of the nlattr payload.
3187 * Once things are setup properly, we call into get_one_status().
3189 int drbd_adm_get_status_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3191 const unsigned hdrlen
= GENL_HDRLEN
+ GENL_MAGIC_FAMILY_HDRSZ
;
3193 const char *resource_name
;
3194 struct drbd_resource
*resource
;
3197 /* Is this a followup call? */
3199 /* ... of a single resource dump,
3200 * and the resource iterator has been advanced already? */
3201 if (cb
->args
[2] && cb
->args
[2] != cb
->args
[0])
3202 return 0; /* DONE. */
3206 /* First call (from netlink_dump_start). We need to figure out
3207 * which resource(s) the user wants us to dump. */
3208 nla
= nla_find(nlmsg_attrdata(cb
->nlh
, hdrlen
),
3209 nlmsg_attrlen(cb
->nlh
, hdrlen
),
3210 DRBD_NLA_CFG_CONTEXT
);
3212 /* No explicit context given. Dump all. */
3215 maxtype
= ARRAY_SIZE(drbd_cfg_context_nl_policy
) - 1;
3216 nla
= drbd_nla_find_nested(maxtype
, nla
, __nla_type(T_ctx_resource_name
));
3218 return PTR_ERR(nla
);
3219 /* context given, but no name present? */
3222 resource_name
= nla_data(nla
);
3223 if (!*resource_name
)
3225 resource
= drbd_find_resource(resource_name
);
3229 kref_put(&resource
->kref
, drbd_destroy_resource
); /* get_one_status() revalidates the resource */
3231 /* prime iterators, and set "filter" mode mark:
3232 * only dump this connection. */
3233 cb
->args
[0] = (long)resource
;
3234 /* cb->args[1] = 0; passed in this way. */
3235 cb
->args
[2] = (long)resource
;
3238 return get_one_status(skb
, cb
);
3241 int drbd_adm_get_timeout_type(struct sk_buff
*skb
, struct genl_info
*info
)
3243 struct drbd_config_context adm_ctx
;
3244 enum drbd_ret_code retcode
;
3245 struct timeout_parms tp
;
3248 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
3249 if (!adm_ctx
.reply_skb
)
3251 if (retcode
!= NO_ERROR
)
3255 adm_ctx
.device
->state
.pdsk
== D_OUTDATED
? UT_PEER_OUTDATED
:
3256 test_bit(USE_DEGR_WFC_T
, &adm_ctx
.device
->flags
) ? UT_DEGRADED
:
3259 err
= timeout_parms_to_priv_skb(adm_ctx
.reply_skb
, &tp
);
3261 nlmsg_free(adm_ctx
.reply_skb
);
3265 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3269 int drbd_adm_start_ov(struct sk_buff
*skb
, struct genl_info
*info
)
3271 struct drbd_config_context adm_ctx
;
3272 struct drbd_device
*device
;
3273 enum drbd_ret_code retcode
;
3274 struct start_ov_parms parms
;
3276 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
3277 if (!adm_ctx
.reply_skb
)
3279 if (retcode
!= NO_ERROR
)
3282 device
= adm_ctx
.device
;
3284 /* resume from last known position, if possible */
3285 parms
.ov_start_sector
= device
->ov_start_sector
;
3286 parms
.ov_stop_sector
= ULLONG_MAX
;
3287 if (info
->attrs
[DRBD_NLA_START_OV_PARMS
]) {
3288 int err
= start_ov_parms_from_attrs(&parms
, info
);
3290 retcode
= ERR_MANDATORY_TAG
;
3291 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
3295 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
3297 /* w_make_ov_request expects position to be aligned */
3298 device
->ov_start_sector
= parms
.ov_start_sector
& ~(BM_SECT_PER_BIT
-1);
3299 device
->ov_stop_sector
= parms
.ov_stop_sector
;
3301 /* If there is still bitmap IO pending, e.g. previous resync or verify
3302 * just being finished, wait for it before requesting a new resync. */
3303 drbd_suspend_io(device
);
3304 wait_event(device
->misc_wait
, !test_bit(BITMAP_IO
, &device
->flags
));
3305 retcode
= drbd_request_state(device
, NS(conn
, C_VERIFY_S
));
3306 drbd_resume_io(device
);
3308 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
3310 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3315 int drbd_adm_new_c_uuid(struct sk_buff
*skb
, struct genl_info
*info
)
3317 struct drbd_config_context adm_ctx
;
3318 struct drbd_device
*device
;
3319 enum drbd_ret_code retcode
;
3320 int skip_initial_sync
= 0;
3322 struct new_c_uuid_parms args
;
3324 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
3325 if (!adm_ctx
.reply_skb
)
3327 if (retcode
!= NO_ERROR
)
3330 device
= adm_ctx
.device
;
3331 memset(&args
, 0, sizeof(args
));
3332 if (info
->attrs
[DRBD_NLA_NEW_C_UUID_PARMS
]) {
3333 err
= new_c_uuid_parms_from_attrs(&args
, info
);
3335 retcode
= ERR_MANDATORY_TAG
;
3336 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
3341 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
3342 mutex_lock(device
->state_mutex
); /* Protects us against serialized state changes. */
3344 if (!get_ldev(device
)) {
3345 retcode
= ERR_NO_DISK
;
3349 /* this is "skip initial sync", assume to be clean */
3350 if (device
->state
.conn
== C_CONNECTED
&&
3351 first_peer_device(device
)->connection
->agreed_pro_version
>= 90 &&
3352 device
->ldev
->md
.uuid
[UI_CURRENT
] == UUID_JUST_CREATED
&& args
.clear_bm
) {
3353 drbd_info(device
, "Preparing to skip initial sync\n");
3354 skip_initial_sync
= 1;
3355 } else if (device
->state
.conn
!= C_STANDALONE
) {
3356 retcode
= ERR_CONNECTED
;
3360 drbd_uuid_set(device
, UI_BITMAP
, 0); /* Rotate UI_BITMAP to History 1, etc... */
3361 drbd_uuid_new_current(device
); /* New current, previous to UI_BITMAP */
3363 if (args
.clear_bm
) {
3364 err
= drbd_bitmap_io(device
, &drbd_bmio_clear_n_write
,
3365 "clear_n_write from new_c_uuid", BM_LOCKED_MASK
);
3367 drbd_err(device
, "Writing bitmap failed with %d\n", err
);
3368 retcode
= ERR_IO_MD_DISK
;
3370 if (skip_initial_sync
) {
3371 drbd_send_uuids_skip_initial_sync(first_peer_device(device
));
3372 _drbd_uuid_set(device
, UI_BITMAP
, 0);
3373 drbd_print_uuids(device
, "cleared bitmap UUID");
3374 spin_lock_irq(&device
->resource
->req_lock
);
3375 _drbd_set_state(_NS2(device
, disk
, D_UP_TO_DATE
, pdsk
, D_UP_TO_DATE
),
3377 spin_unlock_irq(&device
->resource
->req_lock
);
3381 drbd_md_sync(device
);
3385 mutex_unlock(device
->state_mutex
);
3386 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
3388 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3392 static enum drbd_ret_code
3393 drbd_check_resource_name(struct drbd_config_context
*adm_ctx
)
3395 const char *name
= adm_ctx
->resource_name
;
3396 if (!name
|| !name
[0]) {
3397 drbd_msg_put_info(adm_ctx
->reply_skb
, "resource name missing");
3398 return ERR_MANDATORY_TAG
;
3400 /* if we want to use these in sysfs/configfs/debugfs some day,
3401 * we must not allow slashes */
3402 if (strchr(name
, '/')) {
3403 drbd_msg_put_info(adm_ctx
->reply_skb
, "invalid resource name");
3404 return ERR_INVALID_REQUEST
;
3409 int drbd_adm_new_resource(struct sk_buff
*skb
, struct genl_info
*info
)
3411 struct drbd_config_context adm_ctx
;
3412 enum drbd_ret_code retcode
;
3413 struct res_opts res_opts
;
3416 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, 0);
3417 if (!adm_ctx
.reply_skb
)
3419 if (retcode
!= NO_ERROR
)
3422 set_res_opts_defaults(&res_opts
);
3423 err
= res_opts_from_attrs(&res_opts
, info
);
3424 if (err
&& err
!= -ENOMSG
) {
3425 retcode
= ERR_MANDATORY_TAG
;
3426 drbd_msg_put_info(adm_ctx
.reply_skb
, from_attrs_err_to_txt(err
));
3430 retcode
= drbd_check_resource_name(&adm_ctx
);
3431 if (retcode
!= NO_ERROR
)
3434 if (adm_ctx
.resource
) {
3435 if (info
->nlhdr
->nlmsg_flags
& NLM_F_EXCL
) {
3436 retcode
= ERR_INVALID_REQUEST
;
3437 drbd_msg_put_info(adm_ctx
.reply_skb
, "resource exists");
3439 /* else: still NO_ERROR */
3443 /* not yet safe for genl_family.parallel_ops */
3444 if (!conn_create(adm_ctx
.resource_name
, &res_opts
))
3445 retcode
= ERR_NOMEM
;
3447 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3451 int drbd_adm_new_minor(struct sk_buff
*skb
, struct genl_info
*info
)
3453 struct drbd_config_context adm_ctx
;
3454 struct drbd_genlmsghdr
*dh
= info
->userhdr
;
3455 enum drbd_ret_code retcode
;
3457 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_RESOURCE
);
3458 if (!adm_ctx
.reply_skb
)
3460 if (retcode
!= NO_ERROR
)
3463 if (dh
->minor
> MINORMASK
) {
3464 drbd_msg_put_info(adm_ctx
.reply_skb
, "requested minor out of range");
3465 retcode
= ERR_INVALID_REQUEST
;
3468 if (adm_ctx
.volume
> DRBD_VOLUME_MAX
) {
3469 drbd_msg_put_info(adm_ctx
.reply_skb
, "requested volume id out of range");
3470 retcode
= ERR_INVALID_REQUEST
;
3474 /* drbd_adm_prepare made sure already
3475 * that first_peer_device(device)->connection and device->vnr match the request. */
3476 if (adm_ctx
.device
) {
3477 if (info
->nlhdr
->nlmsg_flags
& NLM_F_EXCL
)
3478 retcode
= ERR_MINOR_EXISTS
;
3479 /* else: still NO_ERROR */
3483 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
3484 retcode
= drbd_create_device(&adm_ctx
, dh
->minor
);
3485 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
3487 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3491 static enum drbd_ret_code
adm_del_minor(struct drbd_device
*device
)
3493 if (device
->state
.disk
== D_DISKLESS
&&
3494 /* no need to be device->state.conn == C_STANDALONE &&
3495 * we may want to delete a minor from a live replication group.
3497 device
->state
.role
== R_SECONDARY
) {
3498 _drbd_request_state(device
, NS(conn
, C_WF_REPORT_PARAMS
),
3499 CS_VERBOSE
+ CS_WAIT_COMPLETE
);
3500 drbd_delete_device(device
);
3503 return ERR_MINOR_CONFIGURED
;
3506 int drbd_adm_del_minor(struct sk_buff
*skb
, struct genl_info
*info
)
3508 struct drbd_config_context adm_ctx
;
3509 enum drbd_ret_code retcode
;
3511 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_MINOR
);
3512 if (!adm_ctx
.reply_skb
)
3514 if (retcode
!= NO_ERROR
)
3517 mutex_lock(&adm_ctx
.resource
->adm_mutex
);
3518 retcode
= adm_del_minor(adm_ctx
.device
);
3519 mutex_unlock(&adm_ctx
.resource
->adm_mutex
);
3521 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3525 int drbd_adm_down(struct sk_buff
*skb
, struct genl_info
*info
)
3527 struct drbd_config_context adm_ctx
;
3528 struct drbd_resource
*resource
;
3529 struct drbd_connection
*connection
;
3530 struct drbd_device
*device
;
3531 int retcode
; /* enum drbd_ret_code rsp. enum drbd_state_rv */
3534 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_RESOURCE
);
3535 if (!adm_ctx
.reply_skb
)
3537 if (retcode
!= NO_ERROR
)
3540 resource
= adm_ctx
.resource
;
3541 mutex_lock(&resource
->adm_mutex
);
3543 for_each_connection(connection
, resource
) {
3544 struct drbd_peer_device
*peer_device
;
3546 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
3547 retcode
= drbd_set_role(peer_device
->device
, R_SECONDARY
, 0);
3548 if (retcode
< SS_SUCCESS
) {
3549 drbd_msg_put_info(adm_ctx
.reply_skb
, "failed to demote");
3554 retcode
= conn_try_disconnect(connection
, 0);
3555 if (retcode
< SS_SUCCESS
) {
3556 drbd_msg_put_info(adm_ctx
.reply_skb
, "failed to disconnect");
3562 idr_for_each_entry(&resource
->devices
, device
, i
) {
3563 retcode
= adm_detach(device
, 0);
3564 if (retcode
< SS_SUCCESS
|| retcode
> NO_ERROR
) {
3565 drbd_msg_put_info(adm_ctx
.reply_skb
, "failed to detach");
3570 /* If we reach this, all volumes (of this connection) are Secondary,
3571 * Disconnected, Diskless, aka Unconfigured. Make sure all threads have
3572 * actually stopped, state handling only does drbd_thread_stop_nowait(). */
3573 for_each_connection(connection
, resource
)
3574 drbd_thread_stop(&connection
->worker
);
3576 /* Now, nothing can fail anymore */
3578 /* delete volumes */
3579 idr_for_each_entry(&resource
->devices
, device
, i
) {
3580 retcode
= adm_del_minor(device
);
3581 if (retcode
!= NO_ERROR
) {
3582 /* "can not happen" */
3583 drbd_msg_put_info(adm_ctx
.reply_skb
, "failed to delete volume");
3588 list_del_rcu(&resource
->resources
);
3590 drbd_free_resource(resource
);
3593 mutex_unlock(&resource
->adm_mutex
);
3595 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3599 int drbd_adm_del_resource(struct sk_buff
*skb
, struct genl_info
*info
)
3601 struct drbd_config_context adm_ctx
;
3602 struct drbd_resource
*resource
;
3603 struct drbd_connection
*connection
;
3604 enum drbd_ret_code retcode
;
3606 retcode
= drbd_adm_prepare(&adm_ctx
, skb
, info
, DRBD_ADM_NEED_RESOURCE
);
3607 if (!adm_ctx
.reply_skb
)
3609 if (retcode
!= NO_ERROR
)
3612 resource
= adm_ctx
.resource
;
3613 mutex_lock(&resource
->adm_mutex
);
3614 for_each_connection(connection
, resource
) {
3615 if (connection
->cstate
> C_STANDALONE
) {
3616 retcode
= ERR_NET_CONFIGURED
;
3620 if (!idr_is_empty(&resource
->devices
)) {
3621 retcode
= ERR_RES_IN_USE
;
3625 list_del_rcu(&resource
->resources
);
3626 for_each_connection(connection
, resource
)
3627 drbd_thread_stop(&connection
->worker
);
3629 drbd_free_resource(resource
);
3632 mutex_unlock(&resource
->adm_mutex
);
3634 drbd_adm_finish(&adm_ctx
, info
, retcode
);
3638 void drbd_bcast_event(struct drbd_device
*device
, const struct sib_info
*sib
)
3640 static atomic_t drbd_genl_seq
= ATOMIC_INIT(2); /* two. */
3641 struct sk_buff
*msg
;
3642 struct drbd_genlmsghdr
*d_out
;
3646 seq
= atomic_inc_return(&drbd_genl_seq
);
3647 msg
= genlmsg_new(NLMSG_GOODSIZE
, GFP_NOIO
);
3652 d_out
= genlmsg_put(msg
, 0, seq
, &drbd_genl_family
, 0, DRBD_EVENT
);
3653 if (!d_out
) /* cannot happen, but anyways. */
3654 goto nla_put_failure
;
3655 d_out
->minor
= device_to_minor(device
);
3656 d_out
->ret_code
= NO_ERROR
;
3658 if (nla_put_status_info(msg
, device
, sib
))
3659 goto nla_put_failure
;
3660 genlmsg_end(msg
, d_out
);
3661 err
= drbd_genl_multicast_events(msg
, 0);
3662 /* msg has been consumed or freed in netlink_broadcast() */
3663 if (err
&& err
!= -ESRCH
)
3671 drbd_err(device
, "Error %d while broadcasting event. "
3672 "Event seq:%u sib_reason:%u\n",
3673 err
, seq
, sib
->sib_reason
);