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 "drbd_wrappers.h"
38 #include <asm/unaligned.h>
39 #include <linux/drbd_limits.h>
40 #include <linux/kthread.h>
42 #include <net/genetlink.h>
45 // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info);
46 // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info);
48 int drbd_adm_new_minor(struct sk_buff
*skb
, struct genl_info
*info
);
49 int drbd_adm_del_minor(struct sk_buff
*skb
, struct genl_info
*info
);
51 int drbd_adm_new_resource(struct sk_buff
*skb
, struct genl_info
*info
);
52 int drbd_adm_del_resource(struct sk_buff
*skb
, struct genl_info
*info
);
53 int drbd_adm_down(struct sk_buff
*skb
, struct genl_info
*info
);
55 int drbd_adm_set_role(struct sk_buff
*skb
, struct genl_info
*info
);
56 int drbd_adm_attach(struct sk_buff
*skb
, struct genl_info
*info
);
57 int drbd_adm_disk_opts(struct sk_buff
*skb
, struct genl_info
*info
);
58 int drbd_adm_detach(struct sk_buff
*skb
, struct genl_info
*info
);
59 int drbd_adm_connect(struct sk_buff
*skb
, struct genl_info
*info
);
60 int drbd_adm_net_opts(struct sk_buff
*skb
, struct genl_info
*info
);
61 int drbd_adm_resize(struct sk_buff
*skb
, struct genl_info
*info
);
62 int drbd_adm_start_ov(struct sk_buff
*skb
, struct genl_info
*info
);
63 int drbd_adm_new_c_uuid(struct sk_buff
*skb
, struct genl_info
*info
);
64 int drbd_adm_disconnect(struct sk_buff
*skb
, struct genl_info
*info
);
65 int drbd_adm_invalidate(struct sk_buff
*skb
, struct genl_info
*info
);
66 int drbd_adm_invalidate_peer(struct sk_buff
*skb
, struct genl_info
*info
);
67 int drbd_adm_pause_sync(struct sk_buff
*skb
, struct genl_info
*info
);
68 int drbd_adm_resume_sync(struct sk_buff
*skb
, struct genl_info
*info
);
69 int drbd_adm_suspend_io(struct sk_buff
*skb
, struct genl_info
*info
);
70 int drbd_adm_resume_io(struct sk_buff
*skb
, struct genl_info
*info
);
71 int drbd_adm_outdate(struct sk_buff
*skb
, struct genl_info
*info
);
72 int drbd_adm_resource_opts(struct sk_buff
*skb
, struct genl_info
*info
);
73 int drbd_adm_get_status(struct sk_buff
*skb
, struct genl_info
*info
);
74 int drbd_adm_get_timeout_type(struct sk_buff
*skb
, struct genl_info
*info
);
76 int drbd_adm_get_status_all(struct sk_buff
*skb
, struct netlink_callback
*cb
);
78 #include <linux/drbd_genl_api.h>
80 #include <linux/genl_magic_func.h>
82 /* used blkdev_get_by_path, to claim our meta data device(s) */
83 static char *drbd_m_holder
= "Hands off! this is DRBD's meta data device.";
85 /* Configuration is strictly serialized, because generic netlink message
86 * processing is strictly serialized by the genl_lock().
87 * Which means we can use one static global drbd_config_context struct.
89 static struct drbd_config_context
{
90 /* assigned from drbd_genlmsghdr */
92 /* assigned from request attributes, if present */
94 #define VOLUME_UNSPECIFIED (-1U)
95 /* pointer into the request skb,
96 * limited lifetime! */
98 struct nlattr
*my_addr
;
99 struct nlattr
*peer_addr
;
102 struct sk_buff
*reply_skb
;
103 /* pointer into reply buffer */
104 struct drbd_genlmsghdr
*reply_dh
;
105 /* resolved from attributes, if possible */
106 struct drbd_device
*device
;
107 struct drbd_resource
*resource
;
108 struct drbd_connection
*connection
;
111 static void drbd_adm_send_reply(struct sk_buff
*skb
, struct genl_info
*info
)
113 genlmsg_end(skb
, genlmsg_data(nlmsg_data(nlmsg_hdr(skb
))));
114 if (genlmsg_reply(skb
, info
))
115 printk(KERN_ERR
"drbd: error sending genl reply\n");
118 /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only
119 * reason it could fail was no space in skb, and there are 4k available. */
120 int drbd_msg_put_info(const char *info
)
122 struct sk_buff
*skb
= adm_ctx
.reply_skb
;
126 if (!info
|| !info
[0])
129 nla
= nla_nest_start(skb
, DRBD_NLA_CFG_REPLY
);
133 err
= nla_put_string(skb
, T_info_text
, info
);
135 nla_nest_cancel(skb
, nla
);
138 nla_nest_end(skb
, nla
);
142 /* This would be a good candidate for a "pre_doit" hook,
143 * and per-family private info->pointers.
144 * But we need to stay compatible with older kernels.
145 * If it returns successfully, adm_ctx members are valid.
147 #define DRBD_ADM_NEED_MINOR 1
148 #define DRBD_ADM_NEED_RESOURCE 2
149 #define DRBD_ADM_NEED_CONNECTION 4
150 static int drbd_adm_prepare(struct sk_buff
*skb
, struct genl_info
*info
,
153 struct drbd_genlmsghdr
*d_in
= info
->userhdr
;
154 const u8 cmd
= info
->genlhdr
->cmd
;
157 memset(&adm_ctx
, 0, sizeof(adm_ctx
));
159 /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */
160 if (cmd
!= DRBD_ADM_GET_STATUS
&& !capable(CAP_NET_ADMIN
))
163 adm_ctx
.reply_skb
= genlmsg_new(NLMSG_GOODSIZE
, GFP_KERNEL
);
164 if (!adm_ctx
.reply_skb
) {
169 adm_ctx
.reply_dh
= genlmsg_put_reply(adm_ctx
.reply_skb
,
170 info
, &drbd_genl_family
, 0, cmd
);
171 /* put of a few bytes into a fresh skb of >= 4k will always succeed.
173 if (!adm_ctx
.reply_dh
) {
178 adm_ctx
.reply_dh
->minor
= d_in
->minor
;
179 adm_ctx
.reply_dh
->ret_code
= NO_ERROR
;
181 adm_ctx
.volume
= VOLUME_UNSPECIFIED
;
182 if (info
->attrs
[DRBD_NLA_CFG_CONTEXT
]) {
184 /* parse and validate only */
185 err
= drbd_cfg_context_from_attrs(NULL
, info
);
189 /* It was present, and valid,
190 * copy it over to the reply skb. */
191 err
= nla_put_nohdr(adm_ctx
.reply_skb
,
192 info
->attrs
[DRBD_NLA_CFG_CONTEXT
]->nla_len
,
193 info
->attrs
[DRBD_NLA_CFG_CONTEXT
]);
197 /* and assign stuff to the global adm_ctx */
198 nla
= nested_attr_tb
[__nla_type(T_ctx_volume
)];
200 adm_ctx
.volume
= nla_get_u32(nla
);
201 nla
= nested_attr_tb
[__nla_type(T_ctx_resource_name
)];
203 adm_ctx
.resource_name
= nla_data(nla
);
204 adm_ctx
.my_addr
= nested_attr_tb
[__nla_type(T_ctx_my_addr
)];
205 adm_ctx
.peer_addr
= nested_attr_tb
[__nla_type(T_ctx_peer_addr
)];
206 if ((adm_ctx
.my_addr
&&
207 nla_len(adm_ctx
.my_addr
) > sizeof(adm_ctx
.connection
->my_addr
)) ||
208 (adm_ctx
.peer_addr
&&
209 nla_len(adm_ctx
.peer_addr
) > sizeof(adm_ctx
.connection
->peer_addr
))) {
215 adm_ctx
.minor
= d_in
->minor
;
216 adm_ctx
.device
= minor_to_device(d_in
->minor
);
217 if (adm_ctx
.resource_name
) {
218 adm_ctx
.resource
= drbd_find_resource(adm_ctx
.resource_name
);
219 if (adm_ctx
.resource
) {
220 adm_ctx
.connection
= first_connection(adm_ctx
.resource
);
221 kref_get(&adm_ctx
.connection
->kref
);
225 if (!adm_ctx
.device
&& (flags
& DRBD_ADM_NEED_MINOR
)) {
226 drbd_msg_put_info("unknown minor");
227 return ERR_MINOR_INVALID
;
229 if (!adm_ctx
.resource
&& (flags
& DRBD_ADM_NEED_RESOURCE
)) {
230 drbd_msg_put_info("unknown resource");
231 if (adm_ctx
.resource_name
)
232 return ERR_RES_NOT_KNOWN
;
233 return ERR_INVALID_REQUEST
;
236 if (flags
& DRBD_ADM_NEED_CONNECTION
) {
237 if (adm_ctx
.connection
&& !(flags
& DRBD_ADM_NEED_RESOURCE
)) {
238 drbd_msg_put_info("no resource name expected");
239 return ERR_INVALID_REQUEST
;
241 if (adm_ctx
.device
) {
242 drbd_msg_put_info("no minor number expected");
243 return ERR_INVALID_REQUEST
;
245 if (adm_ctx
.my_addr
&& adm_ctx
.peer_addr
)
246 adm_ctx
.connection
= conn_get_by_addrs(nla_data(adm_ctx
.my_addr
),
247 nla_len(adm_ctx
.my_addr
),
248 nla_data(adm_ctx
.peer_addr
),
249 nla_len(adm_ctx
.peer_addr
));
250 if (!adm_ctx
.connection
) {
251 drbd_msg_put_info("unknown connection");
252 return ERR_INVALID_REQUEST
;
256 /* some more paranoia, if the request was over-determined */
257 if (adm_ctx
.device
&& adm_ctx
.resource
&&
258 adm_ctx
.device
->resource
!= adm_ctx
.resource
) {
259 pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n",
260 adm_ctx
.minor
, adm_ctx
.resource
->name
,
261 adm_ctx
.device
->resource
->name
);
262 drbd_msg_put_info("minor exists in different resource");
263 return ERR_INVALID_REQUEST
;
265 if (adm_ctx
.device
&&
266 adm_ctx
.volume
!= VOLUME_UNSPECIFIED
&&
267 adm_ctx
.volume
!= adm_ctx
.device
->vnr
) {
268 pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n",
269 adm_ctx
.minor
, adm_ctx
.volume
,
271 adm_ctx
.device
->resource
->name
);
272 drbd_msg_put_info("minor exists as different volume");
273 return ERR_INVALID_REQUEST
;
279 nlmsg_free(adm_ctx
.reply_skb
);
280 adm_ctx
.reply_skb
= NULL
;
284 static int drbd_adm_finish(struct genl_info
*info
, int retcode
)
286 if (adm_ctx
.connection
) {
287 kref_put(&adm_ctx
.connection
->kref
, drbd_destroy_connection
);
288 adm_ctx
.connection
= NULL
;
290 if (adm_ctx
.resource
) {
291 kref_put(&adm_ctx
.resource
->kref
, drbd_destroy_resource
);
292 adm_ctx
.resource
= NULL
;
295 if (!adm_ctx
.reply_skb
)
298 adm_ctx
.reply_dh
->ret_code
= retcode
;
299 drbd_adm_send_reply(adm_ctx
.reply_skb
, info
);
303 static void setup_khelper_env(struct drbd_connection
*connection
, char **envp
)
307 /* FIXME: A future version will not allow this case. */
308 if (connection
->my_addr_len
== 0 || connection
->peer_addr_len
== 0)
311 switch (((struct sockaddr
*)&connection
->peer_addr
)->sa_family
) {
314 snprintf(envp
[4], 60, "DRBD_PEER_ADDRESS=%pI6",
315 &((struct sockaddr_in6
*)&connection
->peer_addr
)->sin6_addr
);
319 snprintf(envp
[4], 60, "DRBD_PEER_ADDRESS=%pI4",
320 &((struct sockaddr_in
*)&connection
->peer_addr
)->sin_addr
);
324 snprintf(envp
[4], 60, "DRBD_PEER_ADDRESS=%pI4",
325 &((struct sockaddr_in
*)&connection
->peer_addr
)->sin_addr
);
327 snprintf(envp
[3], 20, "DRBD_PEER_AF=%s", afs
);
330 int drbd_khelper(struct drbd_device
*device
, char *cmd
)
332 char *envp
[] = { "HOME=/",
334 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
335 (char[20]) { }, /* address family */
336 (char[60]) { }, /* address */
339 char *argv
[] = {usermode_helper
, cmd
, mb
, NULL
};
340 struct drbd_connection
*connection
= first_peer_device(device
)->connection
;
344 if (current
== connection
->worker
.task
)
345 set_bit(CALLBACK_PENDING
, &connection
->flags
);
347 snprintf(mb
, 12, "minor-%d", device_to_minor(device
));
348 setup_khelper_env(connection
, envp
);
350 /* The helper may take some time.
351 * write out any unsynced meta data changes now */
352 drbd_md_sync(device
);
354 drbd_info(device
, "helper command: %s %s %s\n", usermode_helper
, cmd
, mb
);
355 sib
.sib_reason
= SIB_HELPER_PRE
;
356 sib
.helper_name
= cmd
;
357 drbd_bcast_event(device
, &sib
);
358 ret
= call_usermodehelper(usermode_helper
, argv
, envp
, UMH_WAIT_PROC
);
360 drbd_warn(device
, "helper command: %s %s %s exit code %u (0x%x)\n",
361 usermode_helper
, cmd
, mb
,
362 (ret
>> 8) & 0xff, ret
);
364 drbd_info(device
, "helper command: %s %s %s exit code %u (0x%x)\n",
365 usermode_helper
, cmd
, mb
,
366 (ret
>> 8) & 0xff, ret
);
367 sib
.sib_reason
= SIB_HELPER_POST
;
368 sib
.helper_exit_code
= ret
;
369 drbd_bcast_event(device
, &sib
);
371 if (current
== connection
->worker
.task
)
372 clear_bit(CALLBACK_PENDING
, &connection
->flags
);
374 if (ret
< 0) /* Ignore any ERRNOs we got. */
380 static int conn_khelper(struct drbd_connection
*connection
, char *cmd
)
382 char *envp
[] = { "HOME=/",
384 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
385 (char[20]) { }, /* address family */
386 (char[60]) { }, /* address */
388 char *resource_name
= connection
->resource
->name
;
389 char *argv
[] = {usermode_helper
, cmd
, resource_name
, NULL
};
392 setup_khelper_env(connection
, envp
);
393 conn_md_sync(connection
);
395 drbd_info(connection
, "helper command: %s %s %s\n", usermode_helper
, cmd
, resource_name
);
396 /* TODO: conn_bcast_event() ?? */
398 ret
= call_usermodehelper(usermode_helper
, argv
, envp
, UMH_WAIT_PROC
);
400 drbd_warn(connection
, "helper command: %s %s %s exit code %u (0x%x)\n",
401 usermode_helper
, cmd
, resource_name
,
402 (ret
>> 8) & 0xff, ret
);
404 drbd_info(connection
, "helper command: %s %s %s exit code %u (0x%x)\n",
405 usermode_helper
, cmd
, resource_name
,
406 (ret
>> 8) & 0xff, ret
);
407 /* TODO: conn_bcast_event() ?? */
409 if (ret
< 0) /* Ignore any ERRNOs we got. */
415 static enum drbd_fencing_p
highest_fencing_policy(struct drbd_connection
*connection
)
417 enum drbd_fencing_p fp
= FP_NOT_AVAIL
;
418 struct drbd_peer_device
*peer_device
;
422 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
) {
423 struct drbd_device
*device
= peer_device
->device
;
424 if (get_ldev_if_state(device
, D_CONSISTENT
)) {
425 struct disk_conf
*disk_conf
=
426 rcu_dereference(peer_device
->device
->ldev
->disk_conf
);
427 fp
= max_t(enum drbd_fencing_p
, fp
, disk_conf
->fencing
);
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 if (connection
->cstate
>= C_WF_REPORT_PARAMS
) {
446 drbd_err(connection
, "Expected cstate < C_WF_REPORT_PARAMS\n");
450 spin_lock_irq(&connection
->req_lock
);
451 connect_cnt
= connection
->connect_cnt
;
452 spin_unlock_irq(&connection
->req_lock
);
454 fp
= highest_fencing_policy(connection
);
457 drbd_warn(connection
, "Not fencing peer, I'm not even Consistent myself.\n");
464 r
= conn_khelper(connection
, "fence-peer");
466 switch ((r
>>8) & 0xff) {
467 case 3: /* peer is inconsistent */
468 ex_to_string
= "peer is inconsistent or worse";
470 val
.pdsk
= D_INCONSISTENT
;
472 case 4: /* peer got outdated, or was already outdated */
473 ex_to_string
= "peer was fenced";
475 val
.pdsk
= D_OUTDATED
;
477 case 5: /* peer was down */
478 if (conn_highest_disk(connection
) == D_UP_TO_DATE
) {
479 /* we will(have) create(d) a new UUID anyways... */
480 ex_to_string
= "peer is unreachable, assumed to be dead";
482 val
.pdsk
= D_OUTDATED
;
484 ex_to_string
= "peer unreachable, doing nothing since disk != UpToDate";
487 case 6: /* Peer is primary, voluntarily outdate myself.
488 * This is useful when an unconnected R_SECONDARY is asked to
489 * become R_PRIMARY, but finds the other peer being active. */
490 ex_to_string
= "peer is active";
491 drbd_warn(connection
, "Peer is primary, outdating myself.\n");
493 val
.disk
= D_OUTDATED
;
496 if (fp
!= FP_STONITH
)
497 drbd_err(connection
, "fence-peer() = 7 && fencing != Stonith !!!\n");
498 ex_to_string
= "peer was stonithed";
500 val
.pdsk
= D_OUTDATED
;
503 /* The script is broken ... */
504 drbd_err(connection
, "fence-peer helper broken, returned %d\n", (r
>>8)&0xff);
505 return false; /* Eventually leave IO frozen */
508 drbd_info(connection
, "fence-peer helper returned %d (%s)\n",
509 (r
>>8) & 0xff, ex_to_string
);
514 conn_request_state(connection, mask, val, CS_VERBOSE);
515 here, because we might were able to re-establish the connection in the
517 spin_lock_irq(&connection
->req_lock
);
518 if (connection
->cstate
< C_WF_REPORT_PARAMS
&& !test_bit(STATE_SENT
, &connection
->flags
)) {
519 if (connection
->connect_cnt
!= connect_cnt
)
520 /* In case the connection was established and droped
521 while the fence-peer handler was running, ignore it */
522 drbd_info(connection
, "Ignoring fence-peer exit code\n");
524 _conn_request_state(connection
, mask
, val
, CS_VERBOSE
);
526 spin_unlock_irq(&connection
->req_lock
);
528 return conn_highest_pdsk(connection
) <= D_OUTDATED
;
531 static int _try_outdate_peer_async(void *data
)
533 struct drbd_connection
*connection
= (struct drbd_connection
*)data
;
535 conn_try_outdate_peer(connection
);
537 kref_put(&connection
->kref
, drbd_destroy_connection
);
541 void conn_try_outdate_peer_async(struct drbd_connection
*connection
)
543 struct task_struct
*opa
;
545 kref_get(&connection
->kref
);
546 opa
= kthread_run(_try_outdate_peer_async
, connection
, "drbd_async_h");
548 drbd_err(connection
, "out of mem, failed to invoke fence-peer helper\n");
549 kref_put(&connection
->kref
, drbd_destroy_connection
);
554 drbd_set_role(struct drbd_device
*device
, enum drbd_role new_role
, int force
)
556 const int max_tries
= 4;
557 enum drbd_state_rv rv
= SS_UNKNOWN_ERROR
;
561 union drbd_state mask
, val
;
563 if (new_role
== R_PRIMARY
) {
564 struct drbd_connection
*connection
;
566 /* Detect dead peers as soon as possible. */
569 for_each_connection(connection
, device
->resource
)
570 request_ping(connection
);
574 mutex_lock(device
->state_mutex
);
576 mask
.i
= 0; mask
.role
= R_MASK
;
577 val
.i
= 0; val
.role
= new_role
;
579 while (try++ < max_tries
) {
580 rv
= _drbd_request_state(device
, mask
, val
, CS_WAIT_COMPLETE
);
582 /* in case we first succeeded to outdate,
583 * but now suddenly could establish a connection */
584 if (rv
== SS_CW_FAILED_BY_PEER
&& mask
.pdsk
!= 0) {
590 if (rv
== SS_NO_UP_TO_DATE_DISK
&& force
&&
591 (device
->state
.disk
< D_UP_TO_DATE
&&
592 device
->state
.disk
>= D_INCONSISTENT
)) {
594 val
.disk
= D_UP_TO_DATE
;
599 if (rv
== SS_NO_UP_TO_DATE_DISK
&&
600 device
->state
.disk
== D_CONSISTENT
&& mask
.pdsk
== 0) {
601 D_ASSERT(device
, device
->state
.pdsk
== D_UNKNOWN
);
603 if (conn_try_outdate_peer(first_peer_device(device
)->connection
)) {
604 val
.disk
= D_UP_TO_DATE
;
610 if (rv
== SS_NOTHING_TO_DO
)
612 if (rv
== SS_PRIMARY_NOP
&& mask
.pdsk
== 0) {
613 if (!conn_try_outdate_peer(first_peer_device(device
)->connection
) && force
) {
614 drbd_warn(device
, "Forced into split brain situation!\n");
616 val
.pdsk
= D_OUTDATED
;
621 if (rv
== SS_TWO_PRIMARIES
) {
622 /* Maybe the peer is detected as dead very soon...
623 retry at most once more in this case. */
626 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
627 timeo
= nc
? (nc
->ping_timeo
+ 1) * HZ
/ 10 : 1;
629 schedule_timeout_interruptible(timeo
);
634 if (rv
< SS_SUCCESS
) {
635 rv
= _drbd_request_state(device
, mask
, val
,
636 CS_VERBOSE
+ CS_WAIT_COMPLETE
);
647 drbd_warn(device
, "Forced to consider local data as UpToDate!\n");
649 /* Wait until nothing is on the fly :) */
650 wait_event(device
->misc_wait
, atomic_read(&device
->ap_pending_cnt
) == 0);
652 /* FIXME also wait for all pending P_BARRIER_ACK? */
654 if (new_role
== R_SECONDARY
) {
655 set_disk_ro(device
->vdisk
, true);
656 if (get_ldev(device
)) {
657 device
->ldev
->md
.uuid
[UI_CURRENT
] &= ~(u64
)1;
661 mutex_lock(&first_peer_device(device
)->connection
->conf_update
);
662 nc
= first_peer_device(device
)->connection
->net_conf
;
664 nc
->discard_my_data
= 0; /* without copy; single bit op is atomic */
665 mutex_unlock(&first_peer_device(device
)->connection
->conf_update
);
667 set_disk_ro(device
->vdisk
, false);
668 if (get_ldev(device
)) {
669 if (((device
->state
.conn
< C_CONNECTED
||
670 device
->state
.pdsk
<= D_FAILED
)
671 && device
->ldev
->md
.uuid
[UI_BITMAP
] == 0) || forced
)
672 drbd_uuid_new_current(device
);
674 device
->ldev
->md
.uuid
[UI_CURRENT
] |= (u64
)1;
679 /* writeout of activity log covered areas of the bitmap
680 * to stable storage done in after state change already */
682 if (device
->state
.conn
>= C_WF_REPORT_PARAMS
) {
683 /* if this was forced, we should consider sync */
685 drbd_send_uuids(device
);
686 drbd_send_current_state(device
);
689 drbd_md_sync(device
);
691 kobject_uevent(&disk_to_dev(device
->vdisk
)->kobj
, KOBJ_CHANGE
);
693 mutex_unlock(device
->state_mutex
);
697 static const char *from_attrs_err_to_txt(int err
)
699 return err
== -ENOMSG
? "required attribute missing" :
700 err
== -EOPNOTSUPP
? "unknown mandatory attribute" :
701 err
== -EEXIST
? "can not change invariant setting" :
702 "invalid attribute value";
705 int drbd_adm_set_role(struct sk_buff
*skb
, struct genl_info
*info
)
707 struct set_role_parms parms
;
709 enum drbd_ret_code retcode
;
711 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
712 if (!adm_ctx
.reply_skb
)
714 if (retcode
!= NO_ERROR
)
717 memset(&parms
, 0, sizeof(parms
));
718 if (info
->attrs
[DRBD_NLA_SET_ROLE_PARMS
]) {
719 err
= set_role_parms_from_attrs(&parms
, info
);
721 retcode
= ERR_MANDATORY_TAG
;
722 drbd_msg_put_info(from_attrs_err_to_txt(err
));
727 if (info
->genlhdr
->cmd
== DRBD_ADM_PRIMARY
)
728 retcode
= drbd_set_role(adm_ctx
.device
, R_PRIMARY
, parms
.assume_uptodate
);
730 retcode
= drbd_set_role(adm_ctx
.device
, R_SECONDARY
, 0);
732 drbd_adm_finish(info
, retcode
);
736 /* Initializes the md.*_offset members, so we are able to find
737 * the on disk meta data.
739 * We currently have two possible layouts:
741 * |----------- md_size_sect ------------------|
742 * [ 4k superblock ][ activity log ][ Bitmap ]
744 * | bm_offset = al_offset + X |
745 * ==> bitmap sectors = md_size_sect - bm_offset
748 * |----------- md_size_sect ------------------|
749 * [data.....][ Bitmap ][ activity log ][ 4k superblock ]
751 * | bm_offset = al_offset - Y |
752 * ==> bitmap sectors = Y = al_offset - bm_offset
754 * Activity log size used to be fixed 32kB,
755 * but is about to become configurable.
757 static void drbd_md_set_sector_offsets(struct drbd_device
*device
,
758 struct drbd_backing_dev
*bdev
)
760 sector_t md_size_sect
= 0;
761 unsigned int al_size_sect
= bdev
->md
.al_size_4k
* 8;
763 bdev
->md
.md_offset
= drbd_md_ss(bdev
);
765 switch (bdev
->md
.meta_dev_idx
) {
767 /* v07 style fixed size indexed meta data */
768 bdev
->md
.md_size_sect
= MD_128MB_SECT
;
769 bdev
->md
.al_offset
= MD_4kB_SECT
;
770 bdev
->md
.bm_offset
= MD_4kB_SECT
+ al_size_sect
;
772 case DRBD_MD_INDEX_FLEX_EXT
:
773 /* just occupy the full device; unit: sectors */
774 bdev
->md
.md_size_sect
= drbd_get_capacity(bdev
->md_bdev
);
775 bdev
->md
.al_offset
= MD_4kB_SECT
;
776 bdev
->md
.bm_offset
= MD_4kB_SECT
+ al_size_sect
;
778 case DRBD_MD_INDEX_INTERNAL
:
779 case DRBD_MD_INDEX_FLEX_INT
:
780 /* al size is still fixed */
781 bdev
->md
.al_offset
= -al_size_sect
;
782 /* we need (slightly less than) ~ this much bitmap sectors: */
783 md_size_sect
= drbd_get_capacity(bdev
->backing_bdev
);
784 md_size_sect
= ALIGN(md_size_sect
, BM_SECT_PER_EXT
);
785 md_size_sect
= BM_SECT_TO_EXT(md_size_sect
);
786 md_size_sect
= ALIGN(md_size_sect
, 8);
788 /* plus the "drbd meta data super block",
789 * and the activity log; */
790 md_size_sect
+= MD_4kB_SECT
+ al_size_sect
;
792 bdev
->md
.md_size_sect
= md_size_sect
;
793 /* bitmap offset is adjusted by 'super' block size */
794 bdev
->md
.bm_offset
= -md_size_sect
+ MD_4kB_SECT
;
799 /* input size is expected to be in KB */
800 char *ppsize(char *buf
, unsigned long long size
)
802 /* Needs 9 bytes at max including trailing NUL:
803 * -1ULL ==> "16384 EB" */
804 static char units
[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
806 while (size
>= 10000 && base
< sizeof(units
)-1) {
808 size
= (size
>> 10) + !!(size
& (1<<9));
811 sprintf(buf
, "%u %cB", (unsigned)size
, units
[base
]);
816 /* there is still a theoretical deadlock when called from receiver
817 * on an D_INCONSISTENT R_PRIMARY:
818 * remote READ does inc_ap_bio, receiver would need to receive answer
819 * packet from remote to dec_ap_bio again.
820 * receiver receive_sizes(), comes here,
821 * waits for ap_bio_cnt == 0. -> deadlock.
822 * but this cannot happen, actually, because:
823 * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable
824 * (not connected, or bad/no disk on peer):
825 * see drbd_fail_request_early, ap_bio_cnt is zero.
826 * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
827 * peer may not initiate a resize.
829 /* Note these are not to be confused with
830 * drbd_adm_suspend_io/drbd_adm_resume_io,
831 * which are (sub) state changes triggered by admin (drbdsetup),
832 * and can be long lived.
833 * This changes an device->flag, is triggered by drbd internals,
834 * and should be short-lived. */
835 void drbd_suspend_io(struct drbd_device
*device
)
837 set_bit(SUSPEND_IO
, &device
->flags
);
838 if (drbd_suspended(device
))
840 wait_event(device
->misc_wait
, !atomic_read(&device
->ap_bio_cnt
));
843 void drbd_resume_io(struct drbd_device
*device
)
845 clear_bit(SUSPEND_IO
, &device
->flags
);
846 wake_up(&device
->misc_wait
);
850 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
851 * @device: DRBD device.
853 * Returns 0 on success, negative return values indicate errors.
854 * You should call drbd_md_sync() after calling this function.
856 enum determine_dev_size
857 drbd_determine_dev_size(struct drbd_device
*device
, enum dds_flags flags
, struct resize_parms
*rs
) __must_hold(local
)
859 sector_t prev_first_sect
, prev_size
; /* previous meta location */
860 sector_t la_size_sect
, u_size
;
861 struct drbd_md
*md
= &device
->ldev
->md
;
862 u32 prev_al_stripe_size_4k
;
868 int md_moved
, la_size_changed
;
869 enum determine_dev_size rv
= DS_UNCHANGED
;
872 * application request passes inc_ap_bio,
873 * but then cannot get an AL-reference.
874 * this function later may wait on ap_bio_cnt == 0. -> deadlock.
877 * Suspend IO right here.
878 * still lock the act_log to not trigger ASSERTs there.
880 drbd_suspend_io(device
);
881 buffer
= drbd_md_get_buffer(device
); /* Lock meta-data IO */
883 drbd_resume_io(device
);
887 /* no wait necessary anymore, actually we could assert that */
888 wait_event(device
->al_wait
, lc_try_lock(device
->act_log
));
890 prev_first_sect
= drbd_md_first_sector(device
->ldev
);
891 prev_size
= device
->ldev
->md
.md_size_sect
;
892 la_size_sect
= device
->ldev
->md
.la_size_sect
;
895 /* rs is non NULL if we should change the AL layout only */
897 prev_al_stripes
= md
->al_stripes
;
898 prev_al_stripe_size_4k
= md
->al_stripe_size_4k
;
900 md
->al_stripes
= rs
->al_stripes
;
901 md
->al_stripe_size_4k
= rs
->al_stripe_size
/ 4;
902 md
->al_size_4k
= (u64
)rs
->al_stripes
* rs
->al_stripe_size
/ 4;
905 drbd_md_set_sector_offsets(device
, device
->ldev
);
908 u_size
= rcu_dereference(device
->ldev
->disk_conf
)->disk_size
;
910 size
= drbd_new_dev_size(device
, device
->ldev
, u_size
, flags
& DDSF_FORCED
);
912 if (size
< la_size_sect
) {
913 if (rs
&& u_size
== 0) {
914 /* Remove "rs &&" later. This check should always be active, but
915 right now the receiver expects the permissive behavior */
916 drbd_warn(device
, "Implicit shrink not allowed. "
917 "Use --size=%llus for explicit shrink.\n",
918 (unsigned long long)size
);
919 rv
= DS_ERROR_SHRINK
;
922 rv
= DS_ERROR_SPACE_MD
;
923 if (rv
!= DS_UNCHANGED
)
927 if (drbd_get_capacity(device
->this_bdev
) != size
||
928 drbd_bm_capacity(device
) != size
) {
930 err
= drbd_bm_resize(device
, size
, !(flags
& DDSF_NO_RESYNC
));
932 /* currently there is only one error: ENOMEM! */
933 size
= drbd_bm_capacity(device
)>>1;
935 drbd_err(device
, "OUT OF MEMORY! "
936 "Could not allocate bitmap!\n");
938 drbd_err(device
, "BM resizing failed. "
939 "Leaving size unchanged at size = %lu KB\n",
940 (unsigned long)size
);
944 /* racy, see comments above. */
945 drbd_set_my_capacity(device
, size
);
946 device
->ldev
->md
.la_size_sect
= size
;
947 drbd_info(device
, "size = %s (%llu KB)\n", ppsize(ppb
, size
>>1),
948 (unsigned long long)size
>>1);
953 la_size_changed
= (la_size_sect
!= device
->ldev
->md
.la_size_sect
);
955 md_moved
= prev_first_sect
!= drbd_md_first_sector(device
->ldev
)
956 || prev_size
!= device
->ldev
->md
.md_size_sect
;
958 if (la_size_changed
|| md_moved
|| rs
) {
961 drbd_al_shrink(device
); /* All extents inactive. */
963 prev_flags
= md
->flags
;
964 md
->flags
&= ~MDF_PRIMARY_IND
;
965 drbd_md_write(device
, buffer
);
967 drbd_info(device
, "Writing the whole bitmap, %s\n",
968 la_size_changed
&& md_moved
? "size changed and md moved" :
969 la_size_changed
? "size changed" : "md moved");
970 /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */
971 drbd_bitmap_io(device
, md_moved
? &drbd_bm_write_all
: &drbd_bm_write
,
972 "size changed", BM_LOCKED_MASK
);
973 drbd_initialize_al(device
, buffer
);
975 md
->flags
= prev_flags
;
976 drbd_md_write(device
, buffer
);
979 drbd_info(device
, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n",
980 md
->al_stripes
, md
->al_stripe_size_4k
* 4);
983 if (size
> la_size_sect
)
984 rv
= la_size_sect
? DS_GREW
: DS_GREW_FROM_ZERO
;
985 if (size
< la_size_sect
)
991 md
->al_stripes
= prev_al_stripes
;
992 md
->al_stripe_size_4k
= prev_al_stripe_size_4k
;
993 md
->al_size_4k
= (u64
)prev_al_stripes
* prev_al_stripe_size_4k
;
995 drbd_md_set_sector_offsets(device
, device
->ldev
);
998 lc_unlock(device
->act_log
);
999 wake_up(&device
->al_wait
);
1000 drbd_md_put_buffer(device
);
1001 drbd_resume_io(device
);
1007 drbd_new_dev_size(struct drbd_device
*device
, struct drbd_backing_dev
*bdev
,
1008 sector_t u_size
, int assume_peer_has_space
)
1010 sector_t p_size
= device
->p_size
; /* partner's disk size. */
1011 sector_t la_size_sect
= bdev
->md
.la_size_sect
; /* last agreed size. */
1012 sector_t m_size
; /* my size */
1015 m_size
= drbd_get_max_capacity(bdev
);
1017 if (device
->state
.conn
< C_CONNECTED
&& assume_peer_has_space
) {
1018 drbd_warn(device
, "Resize while not connected was forced by the user!\n");
1022 if (p_size
&& m_size
) {
1023 size
= min_t(sector_t
, p_size
, m_size
);
1026 size
= la_size_sect
;
1027 if (m_size
&& m_size
< size
)
1029 if (p_size
&& p_size
< size
)
1040 drbd_err(device
, "Both nodes diskless!\n");
1044 drbd_err(device
, "Requested disk size is too big (%lu > %lu)\n",
1045 (unsigned long)u_size
>>1, (unsigned long)size
>>1);
1054 * drbd_check_al_size() - Ensures that the AL is of the right size
1055 * @device: DRBD device.
1057 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
1058 * failed, and 0 on success. You should call drbd_md_sync() after you called
1061 static int drbd_check_al_size(struct drbd_device
*device
, struct disk_conf
*dc
)
1063 struct lru_cache
*n
, *t
;
1064 struct lc_element
*e
;
1065 unsigned int in_use
;
1068 if (device
->act_log
&&
1069 device
->act_log
->nr_elements
== dc
->al_extents
)
1073 t
= device
->act_log
;
1074 n
= lc_create("act_log", drbd_al_ext_cache
, AL_UPDATES_PER_TRANSACTION
,
1075 dc
->al_extents
, sizeof(struct lc_element
), 0);
1078 drbd_err(device
, "Cannot allocate act_log lru!\n");
1081 spin_lock_irq(&device
->al_lock
);
1083 for (i
= 0; i
< t
->nr_elements
; i
++) {
1084 e
= lc_element_by_index(t
, i
);
1086 drbd_err(device
, "refcnt(%d)==%d\n",
1087 e
->lc_number
, e
->refcnt
);
1088 in_use
+= e
->refcnt
;
1092 device
->act_log
= n
;
1093 spin_unlock_irq(&device
->al_lock
);
1095 drbd_err(device
, "Activity log still in use!\n");
1102 drbd_md_mark_dirty(device
); /* we changed device->act_log->nr_elemens */
1106 static void drbd_setup_queue_param(struct drbd_device
*device
, unsigned int max_bio_size
)
1108 struct request_queue
* const q
= device
->rq_queue
;
1109 unsigned int max_hw_sectors
= max_bio_size
>> 9;
1110 unsigned int max_segments
= 0;
1112 if (get_ldev_if_state(device
, D_ATTACHING
)) {
1113 struct request_queue
* const b
= device
->ldev
->backing_bdev
->bd_disk
->queue
;
1115 max_hw_sectors
= min(queue_max_hw_sectors(b
), max_bio_size
>> 9);
1117 max_segments
= rcu_dereference(device
->ldev
->disk_conf
)->max_bio_bvecs
;
1122 blk_queue_logical_block_size(q
, 512);
1123 blk_queue_max_hw_sectors(q
, max_hw_sectors
);
1124 /* This is the workaround for "bio would need to, but cannot, be split" */
1125 blk_queue_max_segments(q
, max_segments
? max_segments
: BLK_MAX_SEGMENTS
);
1126 blk_queue_segment_boundary(q
, PAGE_CACHE_SIZE
-1);
1128 if (get_ldev_if_state(device
, D_ATTACHING
)) {
1129 struct request_queue
* const b
= device
->ldev
->backing_bdev
->bd_disk
->queue
;
1131 blk_queue_stack_limits(q
, b
);
1133 if (q
->backing_dev_info
.ra_pages
!= b
->backing_dev_info
.ra_pages
) {
1134 drbd_info(device
, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
1135 q
->backing_dev_info
.ra_pages
,
1136 b
->backing_dev_info
.ra_pages
);
1137 q
->backing_dev_info
.ra_pages
= b
->backing_dev_info
.ra_pages
;
1143 void drbd_reconsider_max_bio_size(struct drbd_device
*device
)
1145 unsigned int now
, new, local
, peer
;
1147 now
= queue_max_hw_sectors(device
->rq_queue
) << 9;
1148 local
= device
->local_max_bio_size
; /* Eventually last known value, from volatile memory */
1149 peer
= device
->peer_max_bio_size
; /* Eventually last known value, from meta data */
1151 if (get_ldev_if_state(device
, D_ATTACHING
)) {
1152 local
= queue_max_hw_sectors(device
->ldev
->backing_bdev
->bd_disk
->queue
) << 9;
1153 device
->local_max_bio_size
= local
;
1156 local
= min(local
, DRBD_MAX_BIO_SIZE
);
1158 /* We may ignore peer limits if the peer is modern enough.
1159 Because new from 8.3.8 onwards the peer can use multiple
1160 BIOs for a single peer_request */
1161 if (device
->state
.conn
>= C_WF_REPORT_PARAMS
) {
1162 if (first_peer_device(device
)->connection
->agreed_pro_version
< 94)
1163 peer
= min(device
->peer_max_bio_size
, DRBD_MAX_SIZE_H80_PACKET
);
1164 /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */
1165 else if (first_peer_device(device
)->connection
->agreed_pro_version
== 94)
1166 peer
= DRBD_MAX_SIZE_H80_PACKET
;
1167 else if (first_peer_device(device
)->connection
->agreed_pro_version
< 100)
1168 peer
= DRBD_MAX_BIO_SIZE_P95
; /* drbd 8.3.8 onwards, before 8.4.0 */
1170 peer
= DRBD_MAX_BIO_SIZE
;
1173 new = min(local
, peer
);
1175 if (device
->state
.role
== R_PRIMARY
&& new < now
)
1176 drbd_err(device
, "ASSERT FAILED new < now; (%u < %u)\n", new, now
);
1179 drbd_info(device
, "max BIO size = %u\n", new);
1181 drbd_setup_queue_param(device
, new);
1184 /* Starts the worker thread */
1185 static void conn_reconfig_start(struct drbd_connection
*connection
)
1187 drbd_thread_start(&connection
->worker
);
1188 conn_flush_workqueue(connection
);
1191 /* if still unconfigured, stops worker again. */
1192 static void conn_reconfig_done(struct drbd_connection
*connection
)
1195 spin_lock_irq(&connection
->req_lock
);
1196 stop_threads
= conn_all_vols_unconf(connection
) &&
1197 connection
->cstate
== C_STANDALONE
;
1198 spin_unlock_irq(&connection
->req_lock
);
1200 /* asender is implicitly stopped by receiver
1201 * in conn_disconnect() */
1202 drbd_thread_stop(&connection
->receiver
);
1203 drbd_thread_stop(&connection
->worker
);
1207 /* Make sure IO is suspended before calling this function(). */
1208 static void drbd_suspend_al(struct drbd_device
*device
)
1212 if (!lc_try_lock(device
->act_log
)) {
1213 drbd_warn(device
, "Failed to lock al in drbd_suspend_al()\n");
1217 drbd_al_shrink(device
);
1218 spin_lock_irq(&first_peer_device(device
)->connection
->req_lock
);
1219 if (device
->state
.conn
< C_CONNECTED
)
1220 s
= !test_and_set_bit(AL_SUSPENDED
, &device
->flags
);
1221 spin_unlock_irq(&first_peer_device(device
)->connection
->req_lock
);
1222 lc_unlock(device
->act_log
);
1225 drbd_info(device
, "Suspended AL updates\n");
1229 static bool should_set_defaults(struct genl_info
*info
)
1231 unsigned flags
= ((struct drbd_genlmsghdr
*)info
->userhdr
)->flags
;
1232 return 0 != (flags
& DRBD_GENL_F_SET_DEFAULTS
);
1235 static unsigned int drbd_al_extents_max(struct drbd_backing_dev
*bdev
)
1237 /* This is limited by 16 bit "slot" numbers,
1238 * and by available on-disk context storage.
1240 * Also (u16)~0 is special (denotes a "free" extent).
1242 * One transaction occupies one 4kB on-disk block,
1243 * we have n such blocks in the on disk ring buffer,
1244 * the "current" transaction may fail (n-1),
1245 * and there is 919 slot numbers context information per transaction.
1247 * 72 transaction blocks amounts to more than 2**16 context slots,
1248 * so cap there first.
1250 const unsigned int max_al_nr
= DRBD_AL_EXTENTS_MAX
;
1251 const unsigned int sufficient_on_disk
=
1252 (max_al_nr
+ AL_CONTEXT_PER_TRANSACTION
-1)
1253 /AL_CONTEXT_PER_TRANSACTION
;
1255 unsigned int al_size_4k
= bdev
->md
.al_size_4k
;
1257 if (al_size_4k
> sufficient_on_disk
)
1260 return (al_size_4k
- 1) * AL_CONTEXT_PER_TRANSACTION
;
1263 int drbd_adm_disk_opts(struct sk_buff
*skb
, struct genl_info
*info
)
1265 enum drbd_ret_code retcode
;
1266 struct drbd_device
*device
;
1267 struct disk_conf
*new_disk_conf
, *old_disk_conf
;
1268 struct fifo_buffer
*old_plan
= NULL
, *new_plan
= NULL
;
1271 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
1272 if (!adm_ctx
.reply_skb
)
1274 if (retcode
!= NO_ERROR
)
1277 device
= adm_ctx
.device
;
1279 /* we also need a disk
1280 * to change the options on */
1281 if (!get_ldev(device
)) {
1282 retcode
= ERR_NO_DISK
;
1286 new_disk_conf
= kmalloc(sizeof(struct disk_conf
), GFP_KERNEL
);
1287 if (!new_disk_conf
) {
1288 retcode
= ERR_NOMEM
;
1292 mutex_lock(&first_peer_device(device
)->connection
->conf_update
);
1293 old_disk_conf
= device
->ldev
->disk_conf
;
1294 *new_disk_conf
= *old_disk_conf
;
1295 if (should_set_defaults(info
))
1296 set_disk_conf_defaults(new_disk_conf
);
1298 err
= disk_conf_from_attrs_for_change(new_disk_conf
, info
);
1299 if (err
&& err
!= -ENOMSG
) {
1300 retcode
= ERR_MANDATORY_TAG
;
1301 drbd_msg_put_info(from_attrs_err_to_txt(err
));
1305 if (!expect(new_disk_conf
->resync_rate
>= 1))
1306 new_disk_conf
->resync_rate
= 1;
1308 if (new_disk_conf
->al_extents
< DRBD_AL_EXTENTS_MIN
)
1309 new_disk_conf
->al_extents
= DRBD_AL_EXTENTS_MIN
;
1310 if (new_disk_conf
->al_extents
> drbd_al_extents_max(device
->ldev
))
1311 new_disk_conf
->al_extents
= drbd_al_extents_max(device
->ldev
);
1313 if (new_disk_conf
->c_plan_ahead
> DRBD_C_PLAN_AHEAD_MAX
)
1314 new_disk_conf
->c_plan_ahead
= DRBD_C_PLAN_AHEAD_MAX
;
1316 fifo_size
= (new_disk_conf
->c_plan_ahead
* 10 * SLEEP_TIME
) / HZ
;
1317 if (fifo_size
!= device
->rs_plan_s
->size
) {
1318 new_plan
= fifo_alloc(fifo_size
);
1320 drbd_err(device
, "kmalloc of fifo_buffer failed");
1321 retcode
= ERR_NOMEM
;
1326 drbd_suspend_io(device
);
1327 wait_event(device
->al_wait
, lc_try_lock(device
->act_log
));
1328 drbd_al_shrink(device
);
1329 err
= drbd_check_al_size(device
, new_disk_conf
);
1330 lc_unlock(device
->act_log
);
1331 wake_up(&device
->al_wait
);
1332 drbd_resume_io(device
);
1335 retcode
= ERR_NOMEM
;
1339 write_lock_irq(&global_state_lock
);
1340 retcode
= drbd_resync_after_valid(device
, new_disk_conf
->resync_after
);
1341 if (retcode
== NO_ERROR
) {
1342 rcu_assign_pointer(device
->ldev
->disk_conf
, new_disk_conf
);
1343 drbd_resync_after_changed(device
);
1345 write_unlock_irq(&global_state_lock
);
1347 if (retcode
!= NO_ERROR
)
1351 old_plan
= device
->rs_plan_s
;
1352 rcu_assign_pointer(device
->rs_plan_s
, new_plan
);
1355 mutex_unlock(&first_peer_device(device
)->connection
->conf_update
);
1357 if (new_disk_conf
->al_updates
)
1358 device
->ldev
->md
.flags
&= ~MDF_AL_DISABLED
;
1360 device
->ldev
->md
.flags
|= MDF_AL_DISABLED
;
1362 if (new_disk_conf
->md_flushes
)
1363 clear_bit(MD_NO_FUA
, &device
->flags
);
1365 set_bit(MD_NO_FUA
, &device
->flags
);
1367 drbd_bump_write_ordering(first_peer_device(device
)->connection
, WO_bdev_flush
);
1369 drbd_md_sync(device
);
1371 if (device
->state
.conn
>= C_CONNECTED
)
1372 drbd_send_sync_param(device
);
1375 kfree(old_disk_conf
);
1377 mod_timer(&device
->request_timer
, jiffies
+ HZ
);
1381 mutex_unlock(&first_peer_device(device
)->connection
->conf_update
);
1383 kfree(new_disk_conf
);
1388 drbd_adm_finish(info
, retcode
);
1392 int drbd_adm_attach(struct sk_buff
*skb
, struct genl_info
*info
)
1394 struct drbd_device
*device
;
1396 enum drbd_ret_code retcode
;
1397 enum determine_dev_size dd
;
1398 sector_t max_possible_sectors
;
1399 sector_t min_md_device_sectors
;
1400 struct drbd_backing_dev
*nbc
= NULL
; /* new_backing_conf */
1401 struct disk_conf
*new_disk_conf
= NULL
;
1402 struct block_device
*bdev
;
1403 struct lru_cache
*resync_lru
= NULL
;
1404 struct fifo_buffer
*new_plan
= NULL
;
1405 union drbd_state ns
, os
;
1406 enum drbd_state_rv rv
;
1407 struct net_conf
*nc
;
1409 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
1410 if (!adm_ctx
.reply_skb
)
1412 if (retcode
!= NO_ERROR
)
1415 device
= adm_ctx
.device
;
1416 conn_reconfig_start(first_peer_device(device
)->connection
);
1418 /* if you want to reconfigure, please tear down first */
1419 if (device
->state
.disk
> D_DISKLESS
) {
1420 retcode
= ERR_DISK_CONFIGURED
;
1423 /* It may just now have detached because of IO error. Make sure
1424 * drbd_ldev_destroy is done already, we may end up here very fast,
1425 * e.g. if someone calls attach from the on-io-error handler,
1426 * to realize a "hot spare" feature (not that I'd recommend that) */
1427 wait_event(device
->misc_wait
, !atomic_read(&device
->local_cnt
));
1429 /* make sure there is no leftover from previous force-detach attempts */
1430 clear_bit(FORCE_DETACH
, &device
->flags
);
1431 clear_bit(WAS_IO_ERROR
, &device
->flags
);
1432 clear_bit(WAS_READ_ERROR
, &device
->flags
);
1434 /* and no leftover from previously aborted resync or verify, either */
1435 device
->rs_total
= 0;
1436 device
->rs_failed
= 0;
1437 atomic_set(&device
->rs_pending_cnt
, 0);
1439 /* allocation not in the IO path, drbdsetup context */
1440 nbc
= kzalloc(sizeof(struct drbd_backing_dev
), GFP_KERNEL
);
1442 retcode
= ERR_NOMEM
;
1445 spin_lock_init(&nbc
->md
.uuid_lock
);
1447 new_disk_conf
= kzalloc(sizeof(struct disk_conf
), GFP_KERNEL
);
1448 if (!new_disk_conf
) {
1449 retcode
= ERR_NOMEM
;
1452 nbc
->disk_conf
= new_disk_conf
;
1454 set_disk_conf_defaults(new_disk_conf
);
1455 err
= disk_conf_from_attrs(new_disk_conf
, info
);
1457 retcode
= ERR_MANDATORY_TAG
;
1458 drbd_msg_put_info(from_attrs_err_to_txt(err
));
1462 if (new_disk_conf
->c_plan_ahead
> DRBD_C_PLAN_AHEAD_MAX
)
1463 new_disk_conf
->c_plan_ahead
= DRBD_C_PLAN_AHEAD_MAX
;
1465 new_plan
= fifo_alloc((new_disk_conf
->c_plan_ahead
* 10 * SLEEP_TIME
) / HZ
);
1467 retcode
= ERR_NOMEM
;
1471 if (new_disk_conf
->meta_dev_idx
< DRBD_MD_INDEX_FLEX_INT
) {
1472 retcode
= ERR_MD_IDX_INVALID
;
1476 write_lock_irq(&global_state_lock
);
1477 retcode
= drbd_resync_after_valid(device
, new_disk_conf
->resync_after
);
1478 write_unlock_irq(&global_state_lock
);
1479 if (retcode
!= NO_ERROR
)
1483 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
1485 if (new_disk_conf
->fencing
== FP_STONITH
&& nc
->wire_protocol
== DRBD_PROT_A
) {
1487 retcode
= ERR_STONITH_AND_PROT_A
;
1493 bdev
= blkdev_get_by_path(new_disk_conf
->backing_dev
,
1494 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
, device
);
1496 drbd_err(device
, "open(\"%s\") failed with %ld\n", new_disk_conf
->backing_dev
,
1498 retcode
= ERR_OPEN_DISK
;
1501 nbc
->backing_bdev
= bdev
;
1504 * meta_dev_idx >= 0: external fixed size, possibly multiple
1505 * drbd sharing one meta device. TODO in that case, paranoia
1506 * check that [md_bdev, meta_dev_idx] is not yet used by some
1507 * other drbd minor! (if you use drbd.conf + drbdadm, that
1508 * should check it for you already; but if you don't, or
1509 * someone fooled it, we need to double check here)
1511 bdev
= blkdev_get_by_path(new_disk_conf
->meta_dev
,
1512 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
,
1513 (new_disk_conf
->meta_dev_idx
< 0) ?
1514 (void *)device
: (void *)drbd_m_holder
);
1516 drbd_err(device
, "open(\"%s\") failed with %ld\n", new_disk_conf
->meta_dev
,
1518 retcode
= ERR_OPEN_MD_DISK
;
1521 nbc
->md_bdev
= bdev
;
1523 if ((nbc
->backing_bdev
== nbc
->md_bdev
) !=
1524 (new_disk_conf
->meta_dev_idx
== DRBD_MD_INDEX_INTERNAL
||
1525 new_disk_conf
->meta_dev_idx
== DRBD_MD_INDEX_FLEX_INT
)) {
1526 retcode
= ERR_MD_IDX_INVALID
;
1530 resync_lru
= lc_create("resync", drbd_bm_ext_cache
,
1531 1, 61, sizeof(struct bm_extent
),
1532 offsetof(struct bm_extent
, lce
));
1534 retcode
= ERR_NOMEM
;
1538 /* Read our meta data super block early.
1539 * This also sets other on-disk offsets. */
1540 retcode
= drbd_md_read(device
, nbc
);
1541 if (retcode
!= NO_ERROR
)
1544 if (new_disk_conf
->al_extents
< DRBD_AL_EXTENTS_MIN
)
1545 new_disk_conf
->al_extents
= DRBD_AL_EXTENTS_MIN
;
1546 if (new_disk_conf
->al_extents
> drbd_al_extents_max(nbc
))
1547 new_disk_conf
->al_extents
= drbd_al_extents_max(nbc
);
1549 if (drbd_get_max_capacity(nbc
) < new_disk_conf
->disk_size
) {
1550 drbd_err(device
, "max capacity %llu smaller than disk size %llu\n",
1551 (unsigned long long) drbd_get_max_capacity(nbc
),
1552 (unsigned long long) new_disk_conf
->disk_size
);
1553 retcode
= ERR_DISK_TOO_SMALL
;
1557 if (new_disk_conf
->meta_dev_idx
< 0) {
1558 max_possible_sectors
= DRBD_MAX_SECTORS_FLEX
;
1559 /* at least one MB, otherwise it does not make sense */
1560 min_md_device_sectors
= (2<<10);
1562 max_possible_sectors
= DRBD_MAX_SECTORS
;
1563 min_md_device_sectors
= MD_128MB_SECT
* (new_disk_conf
->meta_dev_idx
+ 1);
1566 if (drbd_get_capacity(nbc
->md_bdev
) < min_md_device_sectors
) {
1567 retcode
= ERR_MD_DISK_TOO_SMALL
;
1568 drbd_warn(device
, "refusing attach: md-device too small, "
1569 "at least %llu sectors needed for this meta-disk type\n",
1570 (unsigned long long) min_md_device_sectors
);
1574 /* Make sure the new disk is big enough
1575 * (we may currently be R_PRIMARY with no local disk...) */
1576 if (drbd_get_max_capacity(nbc
) <
1577 drbd_get_capacity(device
->this_bdev
)) {
1578 retcode
= ERR_DISK_TOO_SMALL
;
1582 nbc
->known_size
= drbd_get_capacity(nbc
->backing_bdev
);
1584 if (nbc
->known_size
> max_possible_sectors
) {
1585 drbd_warn(device
, "==> truncating very big lower level device "
1586 "to currently maximum possible %llu sectors <==\n",
1587 (unsigned long long) max_possible_sectors
);
1588 if (new_disk_conf
->meta_dev_idx
>= 0)
1589 drbd_warn(device
, "==>> using internal or flexible "
1590 "meta data may help <<==\n");
1593 drbd_suspend_io(device
);
1594 /* also wait for the last barrier ack. */
1595 /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171
1596 * We need a way to either ignore barrier acks for barriers sent before a device
1597 * was attached, or a way to wait for all pending barrier acks to come in.
1598 * As barriers are counted per resource,
1599 * we'd need to suspend io on all devices of a resource.
1601 wait_event(device
->misc_wait
, !atomic_read(&device
->ap_pending_cnt
) || drbd_suspended(device
));
1602 /* and for any other previously queued work */
1603 drbd_flush_workqueue(device
);
1605 rv
= _drbd_request_state(device
, NS(disk
, D_ATTACHING
), CS_VERBOSE
);
1606 retcode
= rv
; /* FIXME: Type mismatch. */
1607 drbd_resume_io(device
);
1608 if (rv
< SS_SUCCESS
)
1611 if (!get_ldev_if_state(device
, D_ATTACHING
))
1612 goto force_diskless
;
1614 if (!device
->bitmap
) {
1615 if (drbd_bm_init(device
)) {
1616 retcode
= ERR_NOMEM
;
1617 goto force_diskless_dec
;
1621 if (device
->state
.conn
< C_CONNECTED
&&
1622 device
->state
.role
== R_PRIMARY
&&
1623 (device
->ed_uuid
& ~((u64
)1)) != (nbc
->md
.uuid
[UI_CURRENT
] & ~((u64
)1))) {
1624 drbd_err(device
, "Can only attach to data with current UUID=%016llX\n",
1625 (unsigned long long)device
->ed_uuid
);
1626 retcode
= ERR_DATA_NOT_CURRENT
;
1627 goto force_diskless_dec
;
1630 /* Since we are diskless, fix the activity log first... */
1631 if (drbd_check_al_size(device
, new_disk_conf
)) {
1632 retcode
= ERR_NOMEM
;
1633 goto force_diskless_dec
;
1636 /* Prevent shrinking of consistent devices ! */
1637 if (drbd_md_test_flag(nbc
, MDF_CONSISTENT
) &&
1638 drbd_new_dev_size(device
, nbc
, nbc
->disk_conf
->disk_size
, 0) < nbc
->md
.la_size_sect
) {
1639 drbd_warn(device
, "refusing to truncate a consistent device\n");
1640 retcode
= ERR_DISK_TOO_SMALL
;
1641 goto force_diskless_dec
;
1644 /* Reset the "barriers don't work" bits here, then force meta data to
1645 * be written, to ensure we determine if barriers are supported. */
1646 if (new_disk_conf
->md_flushes
)
1647 clear_bit(MD_NO_FUA
, &device
->flags
);
1649 set_bit(MD_NO_FUA
, &device
->flags
);
1651 /* Point of no return reached.
1652 * Devices and memory are no longer released by error cleanup below.
1653 * now device takes over responsibility, and the state engine should
1654 * clean it up somewhere. */
1655 D_ASSERT(device
, device
->ldev
== NULL
);
1657 device
->resync
= resync_lru
;
1658 device
->rs_plan_s
= new_plan
;
1661 new_disk_conf
= NULL
;
1664 drbd_bump_write_ordering(first_peer_device(device
)->connection
, WO_bdev_flush
);
1666 if (drbd_md_test_flag(device
->ldev
, MDF_CRASHED_PRIMARY
))
1667 set_bit(CRASHED_PRIMARY
, &device
->flags
);
1669 clear_bit(CRASHED_PRIMARY
, &device
->flags
);
1671 if (drbd_md_test_flag(device
->ldev
, MDF_PRIMARY_IND
) &&
1672 !(device
->state
.role
== R_PRIMARY
&&
1673 first_peer_device(device
)->connection
->susp_nod
))
1674 set_bit(CRASHED_PRIMARY
, &device
->flags
);
1676 device
->send_cnt
= 0;
1677 device
->recv_cnt
= 0;
1678 device
->read_cnt
= 0;
1679 device
->writ_cnt
= 0;
1681 drbd_reconsider_max_bio_size(device
);
1683 /* If I am currently not R_PRIMARY,
1684 * but meta data primary indicator is set,
1685 * I just now recover from a hard crash,
1686 * and have been R_PRIMARY before that crash.
1688 * Now, if I had no connection before that crash
1689 * (have been degraded R_PRIMARY), chances are that
1690 * I won't find my peer now either.
1692 * In that case, and _only_ in that case,
1693 * we use the degr-wfc-timeout instead of the default,
1694 * so we can automatically recover from a crash of a
1695 * degraded but active "cluster" after a certain timeout.
1697 clear_bit(USE_DEGR_WFC_T
, &device
->flags
);
1698 if (device
->state
.role
!= R_PRIMARY
&&
1699 drbd_md_test_flag(device
->ldev
, MDF_PRIMARY_IND
) &&
1700 !drbd_md_test_flag(device
->ldev
, MDF_CONNECTED_IND
))
1701 set_bit(USE_DEGR_WFC_T
, &device
->flags
);
1703 dd
= drbd_determine_dev_size(device
, 0, NULL
);
1704 if (dd
<= DS_ERROR
) {
1705 retcode
= ERR_NOMEM_BITMAP
;
1706 goto force_diskless_dec
;
1707 } else if (dd
== DS_GREW
)
1708 set_bit(RESYNC_AFTER_NEG
, &device
->flags
);
1710 if (drbd_md_test_flag(device
->ldev
, MDF_FULL_SYNC
) ||
1711 (test_bit(CRASHED_PRIMARY
, &device
->flags
) &&
1712 drbd_md_test_flag(device
->ldev
, MDF_AL_DISABLED
))) {
1713 drbd_info(device
, "Assuming that all blocks are out of sync "
1714 "(aka FullSync)\n");
1715 if (drbd_bitmap_io(device
, &drbd_bmio_set_n_write
,
1716 "set_n_write from attaching", BM_LOCKED_MASK
)) {
1717 retcode
= ERR_IO_MD_DISK
;
1718 goto force_diskless_dec
;
1721 if (drbd_bitmap_io(device
, &drbd_bm_read
,
1722 "read from attaching", BM_LOCKED_MASK
)) {
1723 retcode
= ERR_IO_MD_DISK
;
1724 goto force_diskless_dec
;
1728 if (_drbd_bm_total_weight(device
) == drbd_bm_bits(device
))
1729 drbd_suspend_al(device
); /* IO is still suspended here... */
1731 spin_lock_irq(&first_peer_device(device
)->connection
->req_lock
);
1732 os
= drbd_read_state(device
);
1734 /* If MDF_CONSISTENT is not set go into inconsistent state,
1735 otherwise investigate MDF_WasUpToDate...
1736 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1737 otherwise into D_CONSISTENT state.
1739 if (drbd_md_test_flag(device
->ldev
, MDF_CONSISTENT
)) {
1740 if (drbd_md_test_flag(device
->ldev
, MDF_WAS_UP_TO_DATE
))
1741 ns
.disk
= D_CONSISTENT
;
1743 ns
.disk
= D_OUTDATED
;
1745 ns
.disk
= D_INCONSISTENT
;
1748 if (drbd_md_test_flag(device
->ldev
, MDF_PEER_OUT_DATED
))
1749 ns
.pdsk
= D_OUTDATED
;
1752 if (ns
.disk
== D_CONSISTENT
&&
1753 (ns
.pdsk
== D_OUTDATED
|| rcu_dereference(device
->ldev
->disk_conf
)->fencing
== FP_DONT_CARE
))
1754 ns
.disk
= D_UP_TO_DATE
;
1756 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1757 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1758 this point, because drbd_request_state() modifies these
1761 if (rcu_dereference(device
->ldev
->disk_conf
)->al_updates
)
1762 device
->ldev
->md
.flags
&= ~MDF_AL_DISABLED
;
1764 device
->ldev
->md
.flags
|= MDF_AL_DISABLED
;
1768 /* In case we are C_CONNECTED postpone any decision on the new disk
1769 state after the negotiation phase. */
1770 if (device
->state
.conn
== C_CONNECTED
) {
1771 device
->new_state_tmp
.i
= ns
.i
;
1773 ns
.disk
= D_NEGOTIATING
;
1775 /* We expect to receive up-to-date UUIDs soon.
1776 To avoid a race in receive_state, free p_uuid while
1777 holding req_lock. I.e. atomic with the state change */
1778 kfree(device
->p_uuid
);
1779 device
->p_uuid
= NULL
;
1782 rv
= _drbd_set_state(device
, ns
, CS_VERBOSE
, NULL
);
1783 spin_unlock_irq(&first_peer_device(device
)->connection
->req_lock
);
1785 if (rv
< SS_SUCCESS
)
1786 goto force_diskless_dec
;
1788 mod_timer(&device
->request_timer
, jiffies
+ HZ
);
1790 if (device
->state
.role
== R_PRIMARY
)
1791 device
->ldev
->md
.uuid
[UI_CURRENT
] |= (u64
)1;
1793 device
->ldev
->md
.uuid
[UI_CURRENT
] &= ~(u64
)1;
1795 drbd_md_mark_dirty(device
);
1796 drbd_md_sync(device
);
1798 kobject_uevent(&disk_to_dev(device
->vdisk
)->kobj
, KOBJ_CHANGE
);
1800 conn_reconfig_done(first_peer_device(device
)->connection
);
1801 drbd_adm_finish(info
, retcode
);
1807 drbd_force_state(device
, NS(disk
, D_DISKLESS
));
1808 drbd_md_sync(device
);
1810 conn_reconfig_done(first_peer_device(device
)->connection
);
1812 if (nbc
->backing_bdev
)
1813 blkdev_put(nbc
->backing_bdev
,
1814 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
);
1816 blkdev_put(nbc
->md_bdev
,
1817 FMODE_READ
| FMODE_WRITE
| FMODE_EXCL
);
1820 kfree(new_disk_conf
);
1821 lc_destroy(resync_lru
);
1825 drbd_adm_finish(info
, retcode
);
1829 static int adm_detach(struct drbd_device
*device
, int force
)
1831 enum drbd_state_rv retcode
;
1835 set_bit(FORCE_DETACH
, &device
->flags
);
1836 drbd_force_state(device
, NS(disk
, D_FAILED
));
1837 retcode
= SS_SUCCESS
;
1841 drbd_suspend_io(device
); /* so no-one is stuck in drbd_al_begin_io */
1842 drbd_md_get_buffer(device
); /* make sure there is no in-flight meta-data IO */
1843 retcode
= drbd_request_state(device
, NS(disk
, D_FAILED
));
1844 drbd_md_put_buffer(device
);
1845 /* D_FAILED will transition to DISKLESS. */
1846 ret
= wait_event_interruptible(device
->misc_wait
,
1847 device
->state
.disk
!= D_FAILED
);
1848 drbd_resume_io(device
);
1849 if ((int)retcode
== (int)SS_IS_DISKLESS
)
1850 retcode
= SS_NOTHING_TO_DO
;
1857 /* Detaching the disk is a process in multiple stages. First we need to lock
1858 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1859 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1860 * internal references as well.
1861 * Only then we have finally detached. */
1862 int drbd_adm_detach(struct sk_buff
*skb
, struct genl_info
*info
)
1864 enum drbd_ret_code retcode
;
1865 struct detach_parms parms
= { };
1868 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
1869 if (!adm_ctx
.reply_skb
)
1871 if (retcode
!= NO_ERROR
)
1874 if (info
->attrs
[DRBD_NLA_DETACH_PARMS
]) {
1875 err
= detach_parms_from_attrs(&parms
, info
);
1877 retcode
= ERR_MANDATORY_TAG
;
1878 drbd_msg_put_info(from_attrs_err_to_txt(err
));
1883 retcode
= adm_detach(adm_ctx
.device
, parms
.force_detach
);
1885 drbd_adm_finish(info
, retcode
);
1889 static bool conn_resync_running(struct drbd_connection
*connection
)
1891 struct drbd_peer_device
*peer_device
;
1896 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
) {
1897 struct drbd_device
*device
= peer_device
->device
;
1898 if (device
->state
.conn
== C_SYNC_SOURCE
||
1899 device
->state
.conn
== C_SYNC_TARGET
||
1900 device
->state
.conn
== C_PAUSED_SYNC_S
||
1901 device
->state
.conn
== C_PAUSED_SYNC_T
) {
1911 static bool conn_ov_running(struct drbd_connection
*connection
)
1913 struct drbd_peer_device
*peer_device
;
1918 idr_for_each_entry(&connection
->peer_devices
, peer_device
, vnr
) {
1919 struct drbd_device
*device
= peer_device
->device
;
1920 if (device
->state
.conn
== C_VERIFY_S
||
1921 device
->state
.conn
== C_VERIFY_T
) {
1931 static enum drbd_ret_code
1932 _check_net_options(struct drbd_connection
*connection
, struct net_conf
*old_net_conf
, struct net_conf
*new_net_conf
)
1934 struct drbd_peer_device
*peer_device
;
1937 if (old_net_conf
&& connection
->cstate
== C_WF_REPORT_PARAMS
&& connection
->agreed_pro_version
< 100) {
1938 if (new_net_conf
->wire_protocol
!= old_net_conf
->wire_protocol
)
1939 return ERR_NEED_APV_100
;
1941 if (new_net_conf
->two_primaries
!= old_net_conf
->two_primaries
)
1942 return ERR_NEED_APV_100
;
1944 if (strcmp(new_net_conf
->integrity_alg
, old_net_conf
->integrity_alg
))
1945 return ERR_NEED_APV_100
;
1948 if (!new_net_conf
->two_primaries
&&
1949 conn_highest_role(connection
) == R_PRIMARY
&&
1950 conn_highest_peer(connection
) == R_PRIMARY
)
1951 return ERR_NEED_ALLOW_TWO_PRI
;
1953 if (new_net_conf
->two_primaries
&&
1954 (new_net_conf
->wire_protocol
!= DRBD_PROT_C
))
1955 return ERR_NOT_PROTO_C
;
1957 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
1958 struct drbd_device
*device
= peer_device
->device
;
1959 if (get_ldev(device
)) {
1960 enum drbd_fencing_p fp
= rcu_dereference(device
->ldev
->disk_conf
)->fencing
;
1962 if (new_net_conf
->wire_protocol
== DRBD_PROT_A
&& fp
== FP_STONITH
)
1963 return ERR_STONITH_AND_PROT_A
;
1965 if (device
->state
.role
== R_PRIMARY
&& new_net_conf
->discard_my_data
)
1966 return ERR_DISCARD_IMPOSSIBLE
;
1969 if (new_net_conf
->on_congestion
!= OC_BLOCK
&& new_net_conf
->wire_protocol
!= DRBD_PROT_A
)
1970 return ERR_CONG_NOT_PROTO_A
;
1975 static enum drbd_ret_code
1976 check_net_options(struct drbd_connection
*connection
, struct net_conf
*new_net_conf
)
1978 static enum drbd_ret_code rv
;
1979 struct drbd_peer_device
*peer_device
;
1983 rv
= _check_net_options(connection
, rcu_dereference(connection
->net_conf
), new_net_conf
);
1986 /* connection->volumes protected by genl_lock() here */
1987 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
1988 struct drbd_device
*device
= peer_device
->device
;
1989 if (!device
->bitmap
) {
1990 if (drbd_bm_init(device
))
1999 struct crypto_hash
*verify_tfm
;
2000 struct crypto_hash
*csums_tfm
;
2001 struct crypto_hash
*cram_hmac_tfm
;
2002 struct crypto_hash
*integrity_tfm
;
2006 alloc_hash(struct crypto_hash
**tfm
, char *tfm_name
, int err_alg
)
2011 *tfm
= crypto_alloc_hash(tfm_name
, 0, CRYPTO_ALG_ASYNC
);
2020 static enum drbd_ret_code
2021 alloc_crypto(struct crypto
*crypto
, struct net_conf
*new_net_conf
)
2023 char hmac_name
[CRYPTO_MAX_ALG_NAME
];
2024 enum drbd_ret_code rv
;
2026 rv
= alloc_hash(&crypto
->csums_tfm
, new_net_conf
->csums_alg
,
2030 rv
= alloc_hash(&crypto
->verify_tfm
, new_net_conf
->verify_alg
,
2034 rv
= alloc_hash(&crypto
->integrity_tfm
, new_net_conf
->integrity_alg
,
2038 if (new_net_conf
->cram_hmac_alg
[0] != 0) {
2039 snprintf(hmac_name
, CRYPTO_MAX_ALG_NAME
, "hmac(%s)",
2040 new_net_conf
->cram_hmac_alg
);
2042 rv
= alloc_hash(&crypto
->cram_hmac_tfm
, hmac_name
,
2049 static void free_crypto(struct crypto
*crypto
)
2051 crypto_free_hash(crypto
->cram_hmac_tfm
);
2052 crypto_free_hash(crypto
->integrity_tfm
);
2053 crypto_free_hash(crypto
->csums_tfm
);
2054 crypto_free_hash(crypto
->verify_tfm
);
2057 int drbd_adm_net_opts(struct sk_buff
*skb
, struct genl_info
*info
)
2059 enum drbd_ret_code retcode
;
2060 struct drbd_connection
*connection
;
2061 struct net_conf
*old_net_conf
, *new_net_conf
= NULL
;
2063 int ovr
; /* online verify running */
2064 int rsr
; /* re-sync running */
2065 struct crypto crypto
= { };
2067 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_CONNECTION
);
2068 if (!adm_ctx
.reply_skb
)
2070 if (retcode
!= NO_ERROR
)
2073 connection
= adm_ctx
.connection
;
2075 new_net_conf
= kzalloc(sizeof(struct net_conf
), GFP_KERNEL
);
2076 if (!new_net_conf
) {
2077 retcode
= ERR_NOMEM
;
2081 conn_reconfig_start(connection
);
2083 mutex_lock(&connection
->data
.mutex
);
2084 mutex_lock(&connection
->conf_update
);
2085 old_net_conf
= connection
->net_conf
;
2087 if (!old_net_conf
) {
2088 drbd_msg_put_info("net conf missing, try connect");
2089 retcode
= ERR_INVALID_REQUEST
;
2093 *new_net_conf
= *old_net_conf
;
2094 if (should_set_defaults(info
))
2095 set_net_conf_defaults(new_net_conf
);
2097 err
= net_conf_from_attrs_for_change(new_net_conf
, info
);
2098 if (err
&& err
!= -ENOMSG
) {
2099 retcode
= ERR_MANDATORY_TAG
;
2100 drbd_msg_put_info(from_attrs_err_to_txt(err
));
2104 retcode
= check_net_options(connection
, new_net_conf
);
2105 if (retcode
!= NO_ERROR
)
2108 /* re-sync running */
2109 rsr
= conn_resync_running(connection
);
2110 if (rsr
&& strcmp(new_net_conf
->csums_alg
, old_net_conf
->csums_alg
)) {
2111 retcode
= ERR_CSUMS_RESYNC_RUNNING
;
2115 /* online verify running */
2116 ovr
= conn_ov_running(connection
);
2117 if (ovr
&& strcmp(new_net_conf
->verify_alg
, old_net_conf
->verify_alg
)) {
2118 retcode
= ERR_VERIFY_RUNNING
;
2122 retcode
= alloc_crypto(&crypto
, new_net_conf
);
2123 if (retcode
!= NO_ERROR
)
2126 rcu_assign_pointer(connection
->net_conf
, new_net_conf
);
2129 crypto_free_hash(connection
->csums_tfm
);
2130 connection
->csums_tfm
= crypto
.csums_tfm
;
2131 crypto
.csums_tfm
= NULL
;
2134 crypto_free_hash(connection
->verify_tfm
);
2135 connection
->verify_tfm
= crypto
.verify_tfm
;
2136 crypto
.verify_tfm
= NULL
;
2139 crypto_free_hash(connection
->integrity_tfm
);
2140 connection
->integrity_tfm
= crypto
.integrity_tfm
;
2141 if (connection
->cstate
>= C_WF_REPORT_PARAMS
&& connection
->agreed_pro_version
>= 100)
2142 /* Do this without trying to take connection->data.mutex again. */
2143 __drbd_send_protocol(connection
, P_PROTOCOL_UPDATE
);
2145 crypto_free_hash(connection
->cram_hmac_tfm
);
2146 connection
->cram_hmac_tfm
= crypto
.cram_hmac_tfm
;
2148 mutex_unlock(&connection
->conf_update
);
2149 mutex_unlock(&connection
->data
.mutex
);
2151 kfree(old_net_conf
);
2153 if (connection
->cstate
>= C_WF_REPORT_PARAMS
)
2154 drbd_send_sync_param(minor_to_device(conn_lowest_minor(connection
)));
2159 mutex_unlock(&connection
->conf_update
);
2160 mutex_unlock(&connection
->data
.mutex
);
2161 free_crypto(&crypto
);
2162 kfree(new_net_conf
);
2164 conn_reconfig_done(connection
);
2166 drbd_adm_finish(info
, retcode
);
2170 int drbd_adm_connect(struct sk_buff
*skb
, struct genl_info
*info
)
2172 struct drbd_peer_device
*peer_device
;
2173 struct net_conf
*old_net_conf
, *new_net_conf
= NULL
;
2174 struct crypto crypto
= { };
2175 struct drbd_resource
*resource
;
2176 struct drbd_connection
*connection
;
2177 enum drbd_ret_code retcode
;
2181 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_RESOURCE
);
2183 if (!adm_ctx
.reply_skb
)
2185 if (retcode
!= NO_ERROR
)
2187 if (!(adm_ctx
.my_addr
&& adm_ctx
.peer_addr
)) {
2188 drbd_msg_put_info("connection endpoint(s) missing");
2189 retcode
= ERR_INVALID_REQUEST
;
2193 /* No need for _rcu here. All reconfiguration is
2194 * strictly serialized on genl_lock(). We are protected against
2195 * concurrent reconfiguration/addition/deletion */
2196 for_each_resource(resource
, &drbd_resources
) {
2197 for_each_connection(connection
, resource
) {
2198 if (nla_len(adm_ctx
.my_addr
) == connection
->my_addr_len
&&
2199 !memcmp(nla_data(adm_ctx
.my_addr
), &connection
->my_addr
,
2200 connection
->my_addr_len
)) {
2201 retcode
= ERR_LOCAL_ADDR
;
2205 if (nla_len(adm_ctx
.peer_addr
) == connection
->peer_addr_len
&&
2206 !memcmp(nla_data(adm_ctx
.peer_addr
), &connection
->peer_addr
,
2207 connection
->peer_addr_len
)) {
2208 retcode
= ERR_PEER_ADDR
;
2214 connection
= adm_ctx
.connection
;
2215 conn_reconfig_start(connection
);
2217 if (connection
->cstate
> C_STANDALONE
) {
2218 retcode
= ERR_NET_CONFIGURED
;
2222 /* allocation not in the IO path, drbdsetup / netlink process context */
2223 new_net_conf
= kzalloc(sizeof(*new_net_conf
), GFP_KERNEL
);
2224 if (!new_net_conf
) {
2225 retcode
= ERR_NOMEM
;
2229 set_net_conf_defaults(new_net_conf
);
2231 err
= net_conf_from_attrs(new_net_conf
, info
);
2232 if (err
&& err
!= -ENOMSG
) {
2233 retcode
= ERR_MANDATORY_TAG
;
2234 drbd_msg_put_info(from_attrs_err_to_txt(err
));
2238 retcode
= check_net_options(connection
, new_net_conf
);
2239 if (retcode
!= NO_ERROR
)
2242 retcode
= alloc_crypto(&crypto
, new_net_conf
);
2243 if (retcode
!= NO_ERROR
)
2246 ((char *)new_net_conf
->shared_secret
)[SHARED_SECRET_MAX
-1] = 0;
2248 conn_flush_workqueue(connection
);
2250 mutex_lock(&connection
->conf_update
);
2251 old_net_conf
= connection
->net_conf
;
2253 retcode
= ERR_NET_CONFIGURED
;
2254 mutex_unlock(&connection
->conf_update
);
2257 rcu_assign_pointer(connection
->net_conf
, new_net_conf
);
2259 conn_free_crypto(connection
);
2260 connection
->cram_hmac_tfm
= crypto
.cram_hmac_tfm
;
2261 connection
->integrity_tfm
= crypto
.integrity_tfm
;
2262 connection
->csums_tfm
= crypto
.csums_tfm
;
2263 connection
->verify_tfm
= crypto
.verify_tfm
;
2265 connection
->my_addr_len
= nla_len(adm_ctx
.my_addr
);
2266 memcpy(&connection
->my_addr
, nla_data(adm_ctx
.my_addr
), connection
->my_addr_len
);
2267 connection
->peer_addr_len
= nla_len(adm_ctx
.peer_addr
);
2268 memcpy(&connection
->peer_addr
, nla_data(adm_ctx
.peer_addr
), connection
->peer_addr_len
);
2270 mutex_unlock(&connection
->conf_update
);
2273 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
2274 struct drbd_device
*device
= peer_device
->device
;
2275 device
->send_cnt
= 0;
2276 device
->recv_cnt
= 0;
2280 retcode
= conn_request_state(connection
, NS(conn
, C_UNCONNECTED
), CS_VERBOSE
);
2282 conn_reconfig_done(connection
);
2283 drbd_adm_finish(info
, retcode
);
2287 free_crypto(&crypto
);
2288 kfree(new_net_conf
);
2290 conn_reconfig_done(connection
);
2292 drbd_adm_finish(info
, retcode
);
2296 static enum drbd_state_rv
conn_try_disconnect(struct drbd_connection
*connection
, bool force
)
2298 enum drbd_state_rv rv
;
2300 rv
= conn_request_state(connection
, NS(conn
, C_DISCONNECTING
),
2301 force
? CS_HARD
: 0);
2304 case SS_NOTHING_TO_DO
:
2306 case SS_ALREADY_STANDALONE
:
2308 case SS_PRIMARY_NOP
:
2309 /* Our state checking code wants to see the peer outdated. */
2310 rv
= conn_request_state(connection
, NS2(conn
, C_DISCONNECTING
, pdsk
, D_OUTDATED
), 0);
2312 if (rv
== SS_OUTDATE_WO_CONN
) /* lost connection before graceful disconnect succeeded */
2313 rv
= conn_request_state(connection
, NS(conn
, C_DISCONNECTING
), CS_VERBOSE
);
2316 case SS_CW_FAILED_BY_PEER
:
2317 /* The peer probably wants to see us outdated. */
2318 rv
= conn_request_state(connection
, NS2(conn
, C_DISCONNECTING
,
2319 disk
, D_OUTDATED
), 0);
2320 if (rv
== SS_IS_DISKLESS
|| rv
== SS_LOWER_THAN_OUTDATED
) {
2321 rv
= conn_request_state(connection
, NS(conn
, C_DISCONNECTING
),
2326 /* no special handling necessary */
2329 if (rv
>= SS_SUCCESS
) {
2330 enum drbd_state_rv rv2
;
2331 /* No one else can reconfigure the network while I am here.
2332 * The state handling only uses drbd_thread_stop_nowait(),
2333 * we want to really wait here until the receiver is no more.
2335 drbd_thread_stop(&connection
->receiver
);
2337 /* Race breaker. This additional state change request may be
2338 * necessary, if this was a forced disconnect during a receiver
2339 * restart. We may have "killed" the receiver thread just
2340 * after drbdd_init() returned. Typically, we should be
2341 * C_STANDALONE already, now, and this becomes a no-op.
2343 rv2
= conn_request_state(connection
, NS(conn
, C_STANDALONE
),
2344 CS_VERBOSE
| CS_HARD
);
2345 if (rv2
< SS_SUCCESS
)
2346 drbd_err(connection
,
2347 "unexpected rv2=%d in conn_try_disconnect()\n",
2353 int drbd_adm_disconnect(struct sk_buff
*skb
, struct genl_info
*info
)
2355 struct disconnect_parms parms
;
2356 struct drbd_connection
*connection
;
2357 enum drbd_state_rv rv
;
2358 enum drbd_ret_code retcode
;
2361 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_CONNECTION
);
2362 if (!adm_ctx
.reply_skb
)
2364 if (retcode
!= NO_ERROR
)
2367 connection
= adm_ctx
.connection
;
2368 memset(&parms
, 0, sizeof(parms
));
2369 if (info
->attrs
[DRBD_NLA_DISCONNECT_PARMS
]) {
2370 err
= disconnect_parms_from_attrs(&parms
, info
);
2372 retcode
= ERR_MANDATORY_TAG
;
2373 drbd_msg_put_info(from_attrs_err_to_txt(err
));
2378 rv
= conn_try_disconnect(connection
, parms
.force_disconnect
);
2379 if (rv
< SS_SUCCESS
)
2380 retcode
= rv
; /* FIXME: Type mismatch. */
2384 drbd_adm_finish(info
, retcode
);
2388 void resync_after_online_grow(struct drbd_device
*device
)
2390 int iass
; /* I am sync source */
2392 drbd_info(device
, "Resync of new storage after online grow\n");
2393 if (device
->state
.role
!= device
->state
.peer
)
2394 iass
= (device
->state
.role
== R_PRIMARY
);
2396 iass
= test_bit(RESOLVE_CONFLICTS
, &first_peer_device(device
)->connection
->flags
);
2399 drbd_start_resync(device
, C_SYNC_SOURCE
);
2401 _drbd_request_state(device
, NS(conn
, C_WF_SYNC_UUID
), CS_VERBOSE
+ CS_SERIALIZE
);
2404 int drbd_adm_resize(struct sk_buff
*skb
, struct genl_info
*info
)
2406 struct disk_conf
*old_disk_conf
, *new_disk_conf
= NULL
;
2407 struct resize_parms rs
;
2408 struct drbd_device
*device
;
2409 enum drbd_ret_code retcode
;
2410 enum determine_dev_size dd
;
2411 bool change_al_layout
= false;
2412 enum dds_flags ddsf
;
2416 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2417 if (!adm_ctx
.reply_skb
)
2419 if (retcode
!= NO_ERROR
)
2422 device
= adm_ctx
.device
;
2423 if (!get_ldev(device
)) {
2424 retcode
= ERR_NO_DISK
;
2428 memset(&rs
, 0, sizeof(struct resize_parms
));
2429 rs
.al_stripes
= device
->ldev
->md
.al_stripes
;
2430 rs
.al_stripe_size
= device
->ldev
->md
.al_stripe_size_4k
* 4;
2431 if (info
->attrs
[DRBD_NLA_RESIZE_PARMS
]) {
2432 err
= resize_parms_from_attrs(&rs
, info
);
2434 retcode
= ERR_MANDATORY_TAG
;
2435 drbd_msg_put_info(from_attrs_err_to_txt(err
));
2440 if (device
->state
.conn
> C_CONNECTED
) {
2441 retcode
= ERR_RESIZE_RESYNC
;
2445 if (device
->state
.role
== R_SECONDARY
&&
2446 device
->state
.peer
== R_SECONDARY
) {
2447 retcode
= ERR_NO_PRIMARY
;
2451 if (rs
.no_resync
&& first_peer_device(device
)->connection
->agreed_pro_version
< 93) {
2452 retcode
= ERR_NEED_APV_93
;
2457 u_size
= rcu_dereference(device
->ldev
->disk_conf
)->disk_size
;
2459 if (u_size
!= (sector_t
)rs
.resize_size
) {
2460 new_disk_conf
= kmalloc(sizeof(struct disk_conf
), GFP_KERNEL
);
2461 if (!new_disk_conf
) {
2462 retcode
= ERR_NOMEM
;
2467 if (device
->ldev
->md
.al_stripes
!= rs
.al_stripes
||
2468 device
->ldev
->md
.al_stripe_size_4k
!= rs
.al_stripe_size
/ 4) {
2469 u32 al_size_k
= rs
.al_stripes
* rs
.al_stripe_size
;
2471 if (al_size_k
> (16 * 1024 * 1024)) {
2472 retcode
= ERR_MD_LAYOUT_TOO_BIG
;
2476 if (al_size_k
< MD_32kB_SECT
/2) {
2477 retcode
= ERR_MD_LAYOUT_TOO_SMALL
;
2481 if (device
->state
.conn
!= C_CONNECTED
) {
2482 retcode
= ERR_MD_LAYOUT_CONNECTED
;
2486 change_al_layout
= true;
2489 if (device
->ldev
->known_size
!= drbd_get_capacity(device
->ldev
->backing_bdev
))
2490 device
->ldev
->known_size
= drbd_get_capacity(device
->ldev
->backing_bdev
);
2492 if (new_disk_conf
) {
2493 mutex_lock(&first_peer_device(device
)->connection
->conf_update
);
2494 old_disk_conf
= device
->ldev
->disk_conf
;
2495 *new_disk_conf
= *old_disk_conf
;
2496 new_disk_conf
->disk_size
= (sector_t
)rs
.resize_size
;
2497 rcu_assign_pointer(device
->ldev
->disk_conf
, new_disk_conf
);
2498 mutex_unlock(&first_peer_device(device
)->connection
->conf_update
);
2500 kfree(old_disk_conf
);
2503 ddsf
= (rs
.resize_force
? DDSF_FORCED
: 0) | (rs
.no_resync
? DDSF_NO_RESYNC
: 0);
2504 dd
= drbd_determine_dev_size(device
, ddsf
, change_al_layout
? &rs
: NULL
);
2505 drbd_md_sync(device
);
2507 if (dd
== DS_ERROR
) {
2508 retcode
= ERR_NOMEM_BITMAP
;
2510 } else if (dd
== DS_ERROR_SPACE_MD
) {
2511 retcode
= ERR_MD_LAYOUT_NO_FIT
;
2513 } else if (dd
== DS_ERROR_SHRINK
) {
2514 retcode
= ERR_IMPLICIT_SHRINK
;
2518 if (device
->state
.conn
== C_CONNECTED
) {
2520 set_bit(RESIZE_PENDING
, &device
->flags
);
2522 drbd_send_uuids(device
);
2523 drbd_send_sizes(device
, 1, ddsf
);
2527 drbd_adm_finish(info
, retcode
);
2535 int drbd_adm_resource_opts(struct sk_buff
*skb
, struct genl_info
*info
)
2537 enum drbd_ret_code retcode
;
2538 struct res_opts res_opts
;
2541 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_RESOURCE
);
2542 if (!adm_ctx
.reply_skb
)
2544 if (retcode
!= NO_ERROR
)
2547 res_opts
= adm_ctx
.resource
->res_opts
;
2548 if (should_set_defaults(info
))
2549 set_res_opts_defaults(&res_opts
);
2551 err
= res_opts_from_attrs(&res_opts
, info
);
2552 if (err
&& err
!= -ENOMSG
) {
2553 retcode
= ERR_MANDATORY_TAG
;
2554 drbd_msg_put_info(from_attrs_err_to_txt(err
));
2558 err
= set_resource_options(adm_ctx
.resource
, &res_opts
);
2560 retcode
= ERR_INVALID_REQUEST
;
2562 retcode
= ERR_NOMEM
;
2566 drbd_adm_finish(info
, retcode
);
2570 int drbd_adm_invalidate(struct sk_buff
*skb
, struct genl_info
*info
)
2572 struct drbd_device
*device
;
2573 int retcode
; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2575 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2576 if (!adm_ctx
.reply_skb
)
2578 if (retcode
!= NO_ERROR
)
2581 device
= adm_ctx
.device
;
2583 /* If there is still bitmap IO pending, probably because of a previous
2584 * resync just being finished, wait for it before requesting a new resync.
2585 * Also wait for it's after_state_ch(). */
2586 drbd_suspend_io(device
);
2587 wait_event(device
->misc_wait
, !test_bit(BITMAP_IO
, &device
->flags
));
2588 drbd_flush_workqueue(device
);
2590 /* If we happen to be C_STANDALONE R_SECONDARY, just change to
2591 * D_INCONSISTENT, and set all bits in the bitmap. Otherwise,
2592 * try to start a resync handshake as sync target for full sync.
2594 if (device
->state
.conn
== C_STANDALONE
&& device
->state
.role
== R_SECONDARY
) {
2595 retcode
= drbd_request_state(device
, NS(disk
, D_INCONSISTENT
));
2596 if (retcode
>= SS_SUCCESS
) {
2597 if (drbd_bitmap_io(device
, &drbd_bmio_set_n_write
,
2598 "set_n_write from invalidate", BM_LOCKED_MASK
))
2599 retcode
= ERR_IO_MD_DISK
;
2602 retcode
= drbd_request_state(device
, NS(conn
, C_STARTING_SYNC_T
));
2603 drbd_resume_io(device
);
2606 drbd_adm_finish(info
, retcode
);
2610 static int drbd_adm_simple_request_state(struct sk_buff
*skb
, struct genl_info
*info
,
2611 union drbd_state mask
, union drbd_state val
)
2613 enum drbd_ret_code retcode
;
2615 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2616 if (!adm_ctx
.reply_skb
)
2618 if (retcode
!= NO_ERROR
)
2621 retcode
= drbd_request_state(adm_ctx
.device
, mask
, val
);
2623 drbd_adm_finish(info
, retcode
);
2627 static int drbd_bmio_set_susp_al(struct drbd_device
*device
)
2631 rv
= drbd_bmio_set_n_write(device
);
2632 drbd_suspend_al(device
);
2636 int drbd_adm_invalidate_peer(struct sk_buff
*skb
, struct genl_info
*info
)
2638 int retcode
; /* drbd_ret_code, drbd_state_rv */
2639 struct drbd_device
*device
;
2641 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2642 if (!adm_ctx
.reply_skb
)
2644 if (retcode
!= NO_ERROR
)
2647 device
= adm_ctx
.device
;
2649 /* If there is still bitmap IO pending, probably because of a previous
2650 * resync just being finished, wait for it before requesting a new resync.
2651 * Also wait for it's after_state_ch(). */
2652 drbd_suspend_io(device
);
2653 wait_event(device
->misc_wait
, !test_bit(BITMAP_IO
, &device
->flags
));
2654 drbd_flush_workqueue(device
);
2656 /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits
2657 * in the bitmap. Otherwise, try to start a resync handshake
2658 * as sync source for full sync.
2660 if (device
->state
.conn
== C_STANDALONE
&& device
->state
.role
== R_PRIMARY
) {
2661 /* The peer will get a resync upon connect anyways. Just make that
2662 into a full resync. */
2663 retcode
= drbd_request_state(device
, NS(pdsk
, D_INCONSISTENT
));
2664 if (retcode
>= SS_SUCCESS
) {
2665 if (drbd_bitmap_io(device
, &drbd_bmio_set_susp_al
,
2666 "set_n_write from invalidate_peer",
2667 BM_LOCKED_SET_ALLOWED
))
2668 retcode
= ERR_IO_MD_DISK
;
2671 retcode
= drbd_request_state(device
, NS(conn
, C_STARTING_SYNC_S
));
2672 drbd_resume_io(device
);
2675 drbd_adm_finish(info
, retcode
);
2679 int drbd_adm_pause_sync(struct sk_buff
*skb
, struct genl_info
*info
)
2681 enum drbd_ret_code retcode
;
2683 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2684 if (!adm_ctx
.reply_skb
)
2686 if (retcode
!= NO_ERROR
)
2689 if (drbd_request_state(adm_ctx
.device
, NS(user_isp
, 1)) == SS_NOTHING_TO_DO
)
2690 retcode
= ERR_PAUSE_IS_SET
;
2692 drbd_adm_finish(info
, retcode
);
2696 int drbd_adm_resume_sync(struct sk_buff
*skb
, struct genl_info
*info
)
2698 union drbd_dev_state s
;
2699 enum drbd_ret_code retcode
;
2701 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2702 if (!adm_ctx
.reply_skb
)
2704 if (retcode
!= NO_ERROR
)
2707 if (drbd_request_state(adm_ctx
.device
, NS(user_isp
, 0)) == SS_NOTHING_TO_DO
) {
2708 s
= adm_ctx
.device
->state
;
2709 if (s
.conn
== C_PAUSED_SYNC_S
|| s
.conn
== C_PAUSED_SYNC_T
) {
2710 retcode
= s
.aftr_isp
? ERR_PIC_AFTER_DEP
:
2711 s
.peer_isp
? ERR_PIC_PEER_DEP
: ERR_PAUSE_IS_CLEAR
;
2713 retcode
= ERR_PAUSE_IS_CLEAR
;
2718 drbd_adm_finish(info
, retcode
);
2722 int drbd_adm_suspend_io(struct sk_buff
*skb
, struct genl_info
*info
)
2724 return drbd_adm_simple_request_state(skb
, info
, NS(susp
, 1));
2727 int drbd_adm_resume_io(struct sk_buff
*skb
, struct genl_info
*info
)
2729 struct drbd_device
*device
;
2730 int retcode
; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2732 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2733 if (!adm_ctx
.reply_skb
)
2735 if (retcode
!= NO_ERROR
)
2738 device
= adm_ctx
.device
;
2739 if (test_bit(NEW_CUR_UUID
, &device
->flags
)) {
2740 drbd_uuid_new_current(device
);
2741 clear_bit(NEW_CUR_UUID
, &device
->flags
);
2743 drbd_suspend_io(device
);
2744 retcode
= drbd_request_state(device
, NS3(susp
, 0, susp_nod
, 0, susp_fen
, 0));
2745 if (retcode
== SS_SUCCESS
) {
2746 if (device
->state
.conn
< C_CONNECTED
)
2747 tl_clear(first_peer_device(device
)->connection
);
2748 if (device
->state
.disk
== D_DISKLESS
|| device
->state
.disk
== D_FAILED
)
2749 tl_restart(first_peer_device(device
)->connection
, FAIL_FROZEN_DISK_IO
);
2751 drbd_resume_io(device
);
2754 drbd_adm_finish(info
, retcode
);
2758 int drbd_adm_outdate(struct sk_buff
*skb
, struct genl_info
*info
)
2760 return drbd_adm_simple_request_state(skb
, info
, NS(disk
, D_OUTDATED
));
2763 static int nla_put_drbd_cfg_context(struct sk_buff
*skb
,
2764 struct drbd_resource
*resource
,
2765 struct drbd_connection
*connection
,
2766 struct drbd_device
*device
)
2769 nla
= nla_nest_start(skb
, DRBD_NLA_CFG_CONTEXT
);
2771 goto nla_put_failure
;
2773 nla_put_u32(skb
, T_ctx_volume
, device
->vnr
))
2774 goto nla_put_failure
;
2775 if (nla_put_string(skb
, T_ctx_resource_name
, connection
->resource
->name
))
2776 goto nla_put_failure
;
2778 if (connection
->my_addr_len
&&
2779 nla_put(skb
, T_ctx_my_addr
, connection
->my_addr_len
, &connection
->my_addr
))
2780 goto nla_put_failure
;
2781 if (connection
->peer_addr_len
&&
2782 nla_put(skb
, T_ctx_peer_addr
, connection
->peer_addr_len
, &connection
->peer_addr
))
2783 goto nla_put_failure
;
2785 nla_nest_end(skb
, nla
);
2790 nla_nest_cancel(skb
, nla
);
2795 * Return the connection of @resource if @resource has exactly one connection.
2797 static struct drbd_connection
*the_only_connection(struct drbd_resource
*resource
)
2799 struct list_head
*connections
= &resource
->connections
;
2801 if (list_empty(connections
) || connections
->next
->next
!= connections
)
2803 return list_first_entry(&resource
->connections
, struct drbd_connection
, connections
);
2806 int nla_put_status_info(struct sk_buff
*skb
, struct drbd_device
*device
,
2807 const struct sib_info
*sib
)
2809 struct drbd_resource
*resource
= device
->resource
;
2810 struct state_info
*si
= NULL
; /* for sizeof(si->member); */
2814 int exclude_sensitive
;
2816 /* If sib != NULL, this is drbd_bcast_event, which anyone can listen
2817 * to. So we better exclude_sensitive information.
2819 * If sib == NULL, this is drbd_adm_get_status, executed synchronously
2820 * in the context of the requesting user process. Exclude sensitive
2821 * information, unless current has superuser.
2823 * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and
2824 * relies on the current implementation of netlink_dump(), which
2825 * executes the dump callback successively from netlink_recvmsg(),
2826 * always in the context of the receiving process */
2827 exclude_sensitive
= sib
|| !capable(CAP_SYS_ADMIN
);
2829 got_ldev
= get_ldev(device
);
2831 /* We need to add connection name and volume number information still.
2832 * Minor number is in drbd_genlmsghdr. */
2833 if (nla_put_drbd_cfg_context(skb
, resource
, the_only_connection(resource
), device
))
2834 goto nla_put_failure
;
2836 if (res_opts_to_skb(skb
, &device
->resource
->res_opts
, exclude_sensitive
))
2837 goto nla_put_failure
;
2841 struct disk_conf
*disk_conf
;
2843 disk_conf
= rcu_dereference(device
->ldev
->disk_conf
);
2844 err
= disk_conf_to_skb(skb
, disk_conf
, exclude_sensitive
);
2847 struct net_conf
*nc
;
2849 nc
= rcu_dereference(first_peer_device(device
)->connection
->net_conf
);
2851 err
= net_conf_to_skb(skb
, nc
, exclude_sensitive
);
2855 goto nla_put_failure
;
2857 nla
= nla_nest_start(skb
, DRBD_NLA_STATE_INFO
);
2859 goto nla_put_failure
;
2860 if (nla_put_u32(skb
, T_sib_reason
, sib
? sib
->sib_reason
: SIB_GET_STATUS_REPLY
) ||
2861 nla_put_u32(skb
, T_current_state
, device
->state
.i
) ||
2862 nla_put_u64(skb
, T_ed_uuid
, device
->ed_uuid
) ||
2863 nla_put_u64(skb
, T_capacity
, drbd_get_capacity(device
->this_bdev
)) ||
2864 nla_put_u64(skb
, T_send_cnt
, device
->send_cnt
) ||
2865 nla_put_u64(skb
, T_recv_cnt
, device
->recv_cnt
) ||
2866 nla_put_u64(skb
, T_read_cnt
, device
->read_cnt
) ||
2867 nla_put_u64(skb
, T_writ_cnt
, device
->writ_cnt
) ||
2868 nla_put_u64(skb
, T_al_writ_cnt
, device
->al_writ_cnt
) ||
2869 nla_put_u64(skb
, T_bm_writ_cnt
, device
->bm_writ_cnt
) ||
2870 nla_put_u32(skb
, T_ap_bio_cnt
, atomic_read(&device
->ap_bio_cnt
)) ||
2871 nla_put_u32(skb
, T_ap_pending_cnt
, atomic_read(&device
->ap_pending_cnt
)) ||
2872 nla_put_u32(skb
, T_rs_pending_cnt
, atomic_read(&device
->rs_pending_cnt
)))
2873 goto nla_put_failure
;
2878 spin_lock_irq(&device
->ldev
->md
.uuid_lock
);
2879 err
= nla_put(skb
, T_uuids
, sizeof(si
->uuids
), device
->ldev
->md
.uuid
);
2880 spin_unlock_irq(&device
->ldev
->md
.uuid_lock
);
2883 goto nla_put_failure
;
2885 if (nla_put_u32(skb
, T_disk_flags
, device
->ldev
->md
.flags
) ||
2886 nla_put_u64(skb
, T_bits_total
, drbd_bm_bits(device
)) ||
2887 nla_put_u64(skb
, T_bits_oos
, drbd_bm_total_weight(device
)))
2888 goto nla_put_failure
;
2889 if (C_SYNC_SOURCE
<= device
->state
.conn
&&
2890 C_PAUSED_SYNC_T
>= device
->state
.conn
) {
2891 if (nla_put_u64(skb
, T_bits_rs_total
, device
->rs_total
) ||
2892 nla_put_u64(skb
, T_bits_rs_failed
, device
->rs_failed
))
2893 goto nla_put_failure
;
2898 switch(sib
->sib_reason
) {
2899 case SIB_SYNC_PROGRESS
:
2900 case SIB_GET_STATUS_REPLY
:
2902 case SIB_STATE_CHANGE
:
2903 if (nla_put_u32(skb
, T_prev_state
, sib
->os
.i
) ||
2904 nla_put_u32(skb
, T_new_state
, sib
->ns
.i
))
2905 goto nla_put_failure
;
2907 case SIB_HELPER_POST
:
2908 if (nla_put_u32(skb
, T_helper_exit_code
,
2909 sib
->helper_exit_code
))
2910 goto nla_put_failure
;
2912 case SIB_HELPER_PRE
:
2913 if (nla_put_string(skb
, T_helper
, sib
->helper_name
))
2914 goto nla_put_failure
;
2918 nla_nest_end(skb
, nla
);
2928 int drbd_adm_get_status(struct sk_buff
*skb
, struct genl_info
*info
)
2930 enum drbd_ret_code retcode
;
2933 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
2934 if (!adm_ctx
.reply_skb
)
2936 if (retcode
!= NO_ERROR
)
2939 err
= nla_put_status_info(adm_ctx
.reply_skb
, adm_ctx
.device
, NULL
);
2941 nlmsg_free(adm_ctx
.reply_skb
);
2945 drbd_adm_finish(info
, retcode
);
2949 static int get_one_status(struct sk_buff
*skb
, struct netlink_callback
*cb
)
2951 struct drbd_device
*device
;
2952 struct drbd_genlmsghdr
*dh
;
2953 struct drbd_resource
*pos
= (struct drbd_resource
*)cb
->args
[0];
2954 struct drbd_resource
*resource
= NULL
;
2955 struct drbd_resource
*tmp
;
2956 unsigned volume
= cb
->args
[1];
2958 /* Open coded, deferred, iteration:
2959 * for_each_resource_safe(resource, tmp, &drbd_resources) {
2960 * connection = "first connection of resource or undefined";
2961 * idr_for_each_entry(&resource->devices, device, i) {
2965 * where resource is cb->args[0];
2966 * and i is cb->args[1];
2968 * cb->args[2] indicates if we shall loop over all resources,
2969 * or just dump all volumes of a single resource.
2971 * This may miss entries inserted after this dump started,
2972 * or entries deleted before they are reached.
2974 * We need to make sure the device won't disappear while
2975 * we are looking at it, and revalidate our iterators
2976 * on each iteration.
2979 /* synchronize with conn_create()/drbd_destroy_connection() */
2981 /* revalidate iterator position */
2982 for_each_resource_rcu(tmp
, &drbd_resources
) {
2984 /* first iteration */
2996 device
= idr_get_next(&resource
->devices
, &volume
);
2998 /* No more volumes to dump on this resource.
2999 * Advance resource iterator. */
3000 pos
= list_entry_rcu(resource
->resources
.next
,
3001 struct drbd_resource
, resources
);
3002 /* Did we dump any volume of this resource yet? */
3004 /* If we reached the end of the list,
3005 * or only a single resource dump was requested,
3007 if (&pos
->resources
== &drbd_resources
|| cb
->args
[2])
3015 dh
= genlmsg_put(skb
, NETLINK_CB(cb
->skb
).portid
,
3016 cb
->nlh
->nlmsg_seq
, &drbd_genl_family
,
3017 NLM_F_MULTI
, DRBD_ADM_GET_STATUS
);
3022 /* This is a connection without a single volume.
3023 * Suprisingly enough, it may have a network
3025 struct drbd_connection
*connection
;
3028 dh
->ret_code
= NO_ERROR
;
3029 connection
= the_only_connection(resource
);
3030 if (nla_put_drbd_cfg_context(skb
, resource
, connection
, NULL
))
3033 struct net_conf
*nc
;
3035 nc
= rcu_dereference(connection
->net_conf
);
3036 if (nc
&& net_conf_to_skb(skb
, nc
, 1) != 0)
3042 D_ASSERT(device
, device
->vnr
== volume
);
3043 D_ASSERT(device
, device
->resource
== resource
);
3045 dh
->minor
= device_to_minor(device
);
3046 dh
->ret_code
= NO_ERROR
;
3048 if (nla_put_status_info(skb
, device
, NULL
)) {
3050 genlmsg_cancel(skb
, dh
);
3054 genlmsg_end(skb
, dh
);
3059 /* where to start the next iteration */
3060 cb
->args
[0] = (long)pos
;
3061 cb
->args
[1] = (pos
== resource
) ? volume
+ 1 : 0;
3063 /* No more resources/volumes/minors found results in an empty skb.
3064 * Which will terminate the dump. */
3069 * Request status of all resources, or of all volumes within a single resource.
3071 * This is a dump, as the answer may not fit in a single reply skb otherwise.
3072 * Which means we cannot use the family->attrbuf or other such members, because
3073 * dump is NOT protected by the genl_lock(). During dump, we only have access
3074 * to the incoming skb, and need to opencode "parsing" of the nlattr payload.
3076 * Once things are setup properly, we call into get_one_status().
3078 int drbd_adm_get_status_all(struct sk_buff
*skb
, struct netlink_callback
*cb
)
3080 const unsigned hdrlen
= GENL_HDRLEN
+ GENL_MAGIC_FAMILY_HDRSZ
;
3082 const char *resource_name
;
3083 struct drbd_resource
*resource
;
3086 /* Is this a followup call? */
3088 /* ... of a single resource dump,
3089 * and the resource iterator has been advanced already? */
3090 if (cb
->args
[2] && cb
->args
[2] != cb
->args
[0])
3091 return 0; /* DONE. */
3095 /* First call (from netlink_dump_start). We need to figure out
3096 * which resource(s) the user wants us to dump. */
3097 nla
= nla_find(nlmsg_attrdata(cb
->nlh
, hdrlen
),
3098 nlmsg_attrlen(cb
->nlh
, hdrlen
),
3099 DRBD_NLA_CFG_CONTEXT
);
3101 /* No explicit context given. Dump all. */
3104 maxtype
= ARRAY_SIZE(drbd_cfg_context_nl_policy
) - 1;
3105 nla
= drbd_nla_find_nested(maxtype
, nla
, __nla_type(T_ctx_resource_name
));
3107 return PTR_ERR(nla
);
3108 /* context given, but no name present? */
3111 resource_name
= nla_data(nla
);
3112 if (!*resource_name
)
3114 resource
= drbd_find_resource(resource_name
);
3118 kref_put(&resource
->kref
, drbd_destroy_resource
); /* get_one_status() revalidates the resource */
3120 /* prime iterators, and set "filter" mode mark:
3121 * only dump this connection. */
3122 cb
->args
[0] = (long)resource
;
3123 /* cb->args[1] = 0; passed in this way. */
3124 cb
->args
[2] = (long)resource
;
3127 return get_one_status(skb
, cb
);
3130 int drbd_adm_get_timeout_type(struct sk_buff
*skb
, struct genl_info
*info
)
3132 enum drbd_ret_code retcode
;
3133 struct timeout_parms tp
;
3136 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
3137 if (!adm_ctx
.reply_skb
)
3139 if (retcode
!= NO_ERROR
)
3143 adm_ctx
.device
->state
.pdsk
== D_OUTDATED
? UT_PEER_OUTDATED
:
3144 test_bit(USE_DEGR_WFC_T
, &adm_ctx
.device
->flags
) ? UT_DEGRADED
:
3147 err
= timeout_parms_to_priv_skb(adm_ctx
.reply_skb
, &tp
);
3149 nlmsg_free(adm_ctx
.reply_skb
);
3153 drbd_adm_finish(info
, retcode
);
3157 int drbd_adm_start_ov(struct sk_buff
*skb
, struct genl_info
*info
)
3159 struct drbd_device
*device
;
3160 enum drbd_ret_code retcode
;
3161 struct start_ov_parms parms
;
3163 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
3164 if (!adm_ctx
.reply_skb
)
3166 if (retcode
!= NO_ERROR
)
3169 device
= adm_ctx
.device
;
3171 /* resume from last known position, if possible */
3172 parms
.ov_start_sector
= device
->ov_start_sector
;
3173 parms
.ov_stop_sector
= ULLONG_MAX
;
3174 if (info
->attrs
[DRBD_NLA_START_OV_PARMS
]) {
3175 int err
= start_ov_parms_from_attrs(&parms
, info
);
3177 retcode
= ERR_MANDATORY_TAG
;
3178 drbd_msg_put_info(from_attrs_err_to_txt(err
));
3182 /* w_make_ov_request expects position to be aligned */
3183 device
->ov_start_sector
= parms
.ov_start_sector
& ~(BM_SECT_PER_BIT
-1);
3184 device
->ov_stop_sector
= parms
.ov_stop_sector
;
3186 /* If there is still bitmap IO pending, e.g. previous resync or verify
3187 * just being finished, wait for it before requesting a new resync. */
3188 drbd_suspend_io(device
);
3189 wait_event(device
->misc_wait
, !test_bit(BITMAP_IO
, &device
->flags
));
3190 retcode
= drbd_request_state(device
, NS(conn
, C_VERIFY_S
));
3191 drbd_resume_io(device
);
3193 drbd_adm_finish(info
, retcode
);
3198 int drbd_adm_new_c_uuid(struct sk_buff
*skb
, struct genl_info
*info
)
3200 struct drbd_device
*device
;
3201 enum drbd_ret_code retcode
;
3202 int skip_initial_sync
= 0;
3204 struct new_c_uuid_parms args
;
3206 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
3207 if (!adm_ctx
.reply_skb
)
3209 if (retcode
!= NO_ERROR
)
3212 device
= adm_ctx
.device
;
3213 memset(&args
, 0, sizeof(args
));
3214 if (info
->attrs
[DRBD_NLA_NEW_C_UUID_PARMS
]) {
3215 err
= new_c_uuid_parms_from_attrs(&args
, info
);
3217 retcode
= ERR_MANDATORY_TAG
;
3218 drbd_msg_put_info(from_attrs_err_to_txt(err
));
3223 mutex_lock(device
->state_mutex
); /* Protects us against serialized state changes. */
3225 if (!get_ldev(device
)) {
3226 retcode
= ERR_NO_DISK
;
3230 /* this is "skip initial sync", assume to be clean */
3231 if (device
->state
.conn
== C_CONNECTED
&&
3232 first_peer_device(device
)->connection
->agreed_pro_version
>= 90 &&
3233 device
->ldev
->md
.uuid
[UI_CURRENT
] == UUID_JUST_CREATED
&& args
.clear_bm
) {
3234 drbd_info(device
, "Preparing to skip initial sync\n");
3235 skip_initial_sync
= 1;
3236 } else if (device
->state
.conn
!= C_STANDALONE
) {
3237 retcode
= ERR_CONNECTED
;
3241 drbd_uuid_set(device
, UI_BITMAP
, 0); /* Rotate UI_BITMAP to History 1, etc... */
3242 drbd_uuid_new_current(device
); /* New current, previous to UI_BITMAP */
3244 if (args
.clear_bm
) {
3245 err
= drbd_bitmap_io(device
, &drbd_bmio_clear_n_write
,
3246 "clear_n_write from new_c_uuid", BM_LOCKED_MASK
);
3248 drbd_err(device
, "Writing bitmap failed with %d\n", err
);
3249 retcode
= ERR_IO_MD_DISK
;
3251 if (skip_initial_sync
) {
3252 drbd_send_uuids_skip_initial_sync(device
);
3253 _drbd_uuid_set(device
, UI_BITMAP
, 0);
3254 drbd_print_uuids(device
, "cleared bitmap UUID");
3255 spin_lock_irq(&first_peer_device(device
)->connection
->req_lock
);
3256 _drbd_set_state(_NS2(device
, disk
, D_UP_TO_DATE
, pdsk
, D_UP_TO_DATE
),
3258 spin_unlock_irq(&first_peer_device(device
)->connection
->req_lock
);
3262 drbd_md_sync(device
);
3266 mutex_unlock(device
->state_mutex
);
3268 drbd_adm_finish(info
, retcode
);
3272 static enum drbd_ret_code
3273 drbd_check_resource_name(const char *name
)
3275 if (!name
|| !name
[0]) {
3276 drbd_msg_put_info("resource name missing");
3277 return ERR_MANDATORY_TAG
;
3279 /* if we want to use these in sysfs/configfs/debugfs some day,
3280 * we must not allow slashes */
3281 if (strchr(name
, '/')) {
3282 drbd_msg_put_info("invalid resource name");
3283 return ERR_INVALID_REQUEST
;
3288 int drbd_adm_new_resource(struct sk_buff
*skb
, struct genl_info
*info
)
3290 enum drbd_ret_code retcode
;
3291 struct res_opts res_opts
;
3294 retcode
= drbd_adm_prepare(skb
, info
, 0);
3295 if (!adm_ctx
.reply_skb
)
3297 if (retcode
!= NO_ERROR
)
3300 set_res_opts_defaults(&res_opts
);
3301 err
= res_opts_from_attrs(&res_opts
, info
);
3302 if (err
&& err
!= -ENOMSG
) {
3303 retcode
= ERR_MANDATORY_TAG
;
3304 drbd_msg_put_info(from_attrs_err_to_txt(err
));
3308 retcode
= drbd_check_resource_name(adm_ctx
.resource_name
);
3309 if (retcode
!= NO_ERROR
)
3312 if (adm_ctx
.resource
) {
3313 if (info
->nlhdr
->nlmsg_flags
& NLM_F_EXCL
) {
3314 retcode
= ERR_INVALID_REQUEST
;
3315 drbd_msg_put_info("resource exists");
3317 /* else: still NO_ERROR */
3321 if (!conn_create(adm_ctx
.resource_name
, &res_opts
))
3322 retcode
= ERR_NOMEM
;
3324 drbd_adm_finish(info
, retcode
);
3328 int drbd_adm_new_minor(struct sk_buff
*skb
, struct genl_info
*info
)
3330 struct drbd_genlmsghdr
*dh
= info
->userhdr
;
3331 enum drbd_ret_code retcode
;
3333 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_RESOURCE
);
3334 if (!adm_ctx
.reply_skb
)
3336 if (retcode
!= NO_ERROR
)
3339 if (dh
->minor
> MINORMASK
) {
3340 drbd_msg_put_info("requested minor out of range");
3341 retcode
= ERR_INVALID_REQUEST
;
3344 if (adm_ctx
.volume
> DRBD_VOLUME_MAX
) {
3345 drbd_msg_put_info("requested volume id out of range");
3346 retcode
= ERR_INVALID_REQUEST
;
3350 /* drbd_adm_prepare made sure already
3351 * that first_peer_device(device)->connection and device->vnr match the request. */
3352 if (adm_ctx
.device
) {
3353 if (info
->nlhdr
->nlmsg_flags
& NLM_F_EXCL
)
3354 retcode
= ERR_MINOR_EXISTS
;
3355 /* else: still NO_ERROR */
3359 retcode
= drbd_create_device(adm_ctx
.resource
, dh
->minor
, adm_ctx
.volume
);
3361 drbd_adm_finish(info
, retcode
);
3365 static enum drbd_ret_code
adm_del_minor(struct drbd_device
*device
)
3367 if (device
->state
.disk
== D_DISKLESS
&&
3368 /* no need to be device->state.conn == C_STANDALONE &&
3369 * we may want to delete a minor from a live replication group.
3371 device
->state
.role
== R_SECONDARY
) {
3372 _drbd_request_state(device
, NS(conn
, C_WF_REPORT_PARAMS
),
3373 CS_VERBOSE
+ CS_WAIT_COMPLETE
);
3374 drbd_delete_device(device
);
3377 return ERR_MINOR_CONFIGURED
;
3380 int drbd_adm_del_minor(struct sk_buff
*skb
, struct genl_info
*info
)
3382 enum drbd_ret_code retcode
;
3384 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_MINOR
);
3385 if (!adm_ctx
.reply_skb
)
3387 if (retcode
!= NO_ERROR
)
3390 retcode
= adm_del_minor(adm_ctx
.device
);
3392 drbd_adm_finish(info
, retcode
);
3396 int drbd_adm_down(struct sk_buff
*skb
, struct genl_info
*info
)
3398 struct drbd_resource
*resource
;
3399 struct drbd_connection
*connection
;
3400 struct drbd_device
*device
;
3401 int retcode
; /* enum drbd_ret_code rsp. enum drbd_state_rv */
3404 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_RESOURCE
);
3405 if (!adm_ctx
.reply_skb
)
3407 if (retcode
!= NO_ERROR
)
3410 resource
= adm_ctx
.resource
;
3412 for_each_connection(connection
, resource
) {
3413 struct drbd_peer_device
*peer_device
;
3415 idr_for_each_entry(&connection
->peer_devices
, peer_device
, i
) {
3416 retcode
= drbd_set_role(peer_device
->device
, R_SECONDARY
, 0);
3417 if (retcode
< SS_SUCCESS
) {
3418 drbd_msg_put_info("failed to demote");
3423 retcode
= conn_try_disconnect(connection
, 0);
3424 if (retcode
< SS_SUCCESS
) {
3425 drbd_msg_put_info("failed to disconnect");
3431 idr_for_each_entry(&resource
->devices
, device
, i
) {
3432 retcode
= adm_detach(device
, 0);
3433 if (retcode
< SS_SUCCESS
|| retcode
> NO_ERROR
) {
3434 drbd_msg_put_info("failed to detach");
3439 /* If we reach this, all volumes (of this connection) are Secondary,
3440 * Disconnected, Diskless, aka Unconfigured. Make sure all threads have
3441 * actually stopped, state handling only does drbd_thread_stop_nowait(). */
3442 for_each_connection(connection
, resource
)
3443 drbd_thread_stop(&connection
->worker
);
3445 /* Now, nothing can fail anymore */
3447 /* delete volumes */
3448 idr_for_each_entry(&resource
->devices
, device
, i
) {
3449 retcode
= adm_del_minor(device
);
3450 if (retcode
!= NO_ERROR
) {
3451 /* "can not happen" */
3452 drbd_msg_put_info("failed to delete volume");
3457 list_del_rcu(&resource
->resources
);
3459 drbd_free_resource(resource
);
3463 drbd_adm_finish(info
, retcode
);
3467 int drbd_adm_del_resource(struct sk_buff
*skb
, struct genl_info
*info
)
3469 struct drbd_resource
*resource
;
3470 struct drbd_connection
*connection
;
3471 enum drbd_ret_code retcode
;
3473 retcode
= drbd_adm_prepare(skb
, info
, DRBD_ADM_NEED_RESOURCE
);
3474 if (!adm_ctx
.reply_skb
)
3476 if (retcode
!= NO_ERROR
)
3479 resource
= adm_ctx
.resource
;
3480 for_each_connection(connection
, resource
) {
3481 if (connection
->cstate
> C_STANDALONE
) {
3482 retcode
= ERR_NET_CONFIGURED
;
3486 if (!idr_is_empty(&resource
->devices
)) {
3487 retcode
= ERR_RES_IN_USE
;
3491 list_del_rcu(&resource
->resources
);
3492 for_each_connection(connection
, resource
)
3493 drbd_thread_stop(&connection
->worker
);
3495 drbd_free_resource(resource
);
3498 drbd_adm_finish(info
, retcode
);
3502 void drbd_bcast_event(struct drbd_device
*device
, const struct sib_info
*sib
)
3504 static atomic_t drbd_genl_seq
= ATOMIC_INIT(2); /* two. */
3505 struct sk_buff
*msg
;
3506 struct drbd_genlmsghdr
*d_out
;
3510 if (sib
->sib_reason
== SIB_SYNC_PROGRESS
) {
3511 if (time_after(jiffies
, device
->rs_last_bcast
+ HZ
))
3512 device
->rs_last_bcast
= jiffies
;
3517 seq
= atomic_inc_return(&drbd_genl_seq
);
3518 msg
= genlmsg_new(NLMSG_GOODSIZE
, GFP_NOIO
);
3523 d_out
= genlmsg_put(msg
, 0, seq
, &drbd_genl_family
, 0, DRBD_EVENT
);
3524 if (!d_out
) /* cannot happen, but anyways. */
3525 goto nla_put_failure
;
3526 d_out
->minor
= device_to_minor(device
);
3527 d_out
->ret_code
= NO_ERROR
;
3529 if (nla_put_status_info(msg
, device
, sib
))
3530 goto nla_put_failure
;
3531 genlmsg_end(msg
, d_out
);
3532 err
= drbd_genl_multicast_events(msg
, 0);
3533 /* msg has been consumed or freed in netlink_broadcast() */
3534 if (err
&& err
!= -ESRCH
)
3542 drbd_err(device
, "Error %d while broadcasting event. "
3543 "Event seq:%u sib_reason:%u\n",
3544 err
, seq
, sib
->sib_reason
);