projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drbd: Allow to change data-integrity-alg on the fly
[deliverable/linux.git] / drivers / block / drbd / drbd_nl.c
CommitLineData
b411b363
PR		 1/*
2 drbd_nl.c
3
4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg.
5
6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH.
7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>.
8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>.
9
10 drbd is free software; you can redistribute it and/or modify
11 it under the terms of the GNU General Public License as published by
12 the Free Software Foundation; either version 2, or (at your option)
13 any later version.
14
15 drbd is distributed in the hope that it will be useful,
16 but WITHOUT ANY WARRANTY; without even the implied warranty of
17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 GNU General Public License for more details.
19
20 You should have received a copy of the GNU General Public License
21 along with drbd; see the file COPYING. If not, write to
22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
23
24 */
25
b411b363
PR		 26#include <linux/module.h>
27#include <linux/drbd.h>
28#include <linux/in.h>
29#include <linux/fs.h>
30#include <linux/file.h>
31#include <linux/slab.h>
b411b363
PR		 32#include <linux/blkpg.h>
33#include <linux/cpumask.h>
34#include "drbd_int.h"
265be2d0 35#include "drbd_req.h"
b411b363
PR		 36#include "drbd_wrappers.h"
37#include <asm/unaligned.h>
b411b363 38#include <linux/drbd_limits.h>
87f7be4c 39#include <linux/kthread.h>
b411b363 40
3b98c0c2
LE		 41#include <net/genetlink.h>
42
43/* .doit */
44// int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info);
45// int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info);
46
47int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info);
48int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info);
49
50int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info);
51int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info);
85f75dd7 52int drbd_adm_down(struct sk_buff *skb, struct genl_info *info);
3b98c0c2
LE		 53
54int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info);
55int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info);
f399002e 56int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info);
3b98c0c2
LE		 57int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info);
58int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info);
f399002e 59int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info);
3b98c0c2
LE		 60int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info);
61int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info);
62int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info);
63int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info);
64int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info);
65int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info);
66int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info);
67int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info);
68int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info);
69int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info);
70int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info);
f399002e 71int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info);
3b98c0c2
LE		 72int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info);
73int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info);
74/* .dumpit */
75int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb);
76
77#include <linux/drbd_genl_api.h>
78#include <linux/genl_magic_func.h>
79
80/* used blkdev_get_by_path, to claim our meta data device(s) */
b411b363
PR		 81static char *drbd_m_holder = "Hands off! this is DRBD's meta data device.";
82
3b98c0c2
LE		 83/* Configuration is strictly serialized, because generic netlink message
84 * processing is strictly serialized by the genl_lock().
85 * Which means we can use one static global drbd_config_context struct.
86 */
87static struct drbd_config_context {
88 /* assigned from drbd_genlmsghdr */
89 unsigned int minor;
90 /* assigned from request attributes, if present */
91 unsigned int volume;
92#define VOLUME_UNSPECIFIED (-1U)
93 /* pointer into the request skb,
94 * limited lifetime! */
95 char *conn_name;
96
97 /* reply buffer */
98 struct sk_buff *reply_skb;
99 /* pointer into reply buffer */
100 struct drbd_genlmsghdr *reply_dh;
101 /* resolved from attributes, if possible */
102 struct drbd_conf *mdev;
103 struct drbd_tconn *tconn;
104} adm_ctx;
105
106static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info)
107{
108 genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb))));
109 if (genlmsg_reply(skb, info))
110 printk(KERN_ERR "drbd: error sending genl reply\n");
b411b363 111}
3b98c0c2
LE		 112
113/* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only
114 * reason it could fail was no space in skb, and there are 4k available. */
8432b314 115int drbd_msg_put_info(const char *info)
3b98c0c2
LE		 116{
117 struct sk_buff *skb = adm_ctx.reply_skb;
118 struct nlattr *nla;
119 int err = -EMSGSIZE;
120
121 if (!info || !info[0])
122 return 0;
123
124 nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY);
125 if (!nla)
126 return err;
127
128 err = nla_put_string(skb, T_info_text, info);
129 if (err) {
130 nla_nest_cancel(skb, nla);
131 return err;
132 } else
133 nla_nest_end(skb, nla);
134 return 0;
b411b363
PR		 135}
136
3b98c0c2
LE		 137/* This would be a good candidate for a "pre_doit" hook,
138 * and per-family private info->pointers.
139 * But we need to stay compatible with older kernels.
140 * If it returns successfully, adm_ctx members are valid.
141 */
142#define DRBD_ADM_NEED_MINOR 1
143#define DRBD_ADM_NEED_CONN 2
144static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info,
145 unsigned flags)
146{
147 struct drbd_genlmsghdr *d_in = info->userhdr;
148 const u8 cmd = info->genlhdr->cmd;
149 int err;
150
151 memset(&adm_ctx, 0, sizeof(adm_ctx));
152
153 /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */
154 if (cmd != DRBD_ADM_GET_STATUS
155 && security_netlink_recv(skb, CAP_SYS_ADMIN))
156 return -EPERM;
157
158 adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
159 if (!adm_ctx.reply_skb)
160 goto fail;
161
162 adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb,
163 info, &drbd_genl_family, 0, cmd);
164 /* put of a few bytes into a fresh skb of >= 4k will always succeed.
165 * but anyways */
166 if (!adm_ctx.reply_dh)
167 goto fail;
168
169 adm_ctx.reply_dh->minor = d_in->minor;
170 adm_ctx.reply_dh->ret_code = NO_ERROR;
171
172 if (info->attrs[DRBD_NLA_CFG_CONTEXT]) {
173 struct nlattr *nla;
174 /* parse and validate only */
f399002e 175 err = drbd_cfg_context_from_attrs(NULL, info);
3b98c0c2
LE		 176 if (err)
177 goto fail;
178
179 /* It was present, and valid,
180 * copy it over to the reply skb. */
181 err = nla_put_nohdr(adm_ctx.reply_skb,
182 info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len,
183 info->attrs[DRBD_NLA_CFG_CONTEXT]);
184 if (err)
185 goto fail;
186
187 /* and assign stuff to the global adm_ctx */
188 nla = nested_attr_tb[__nla_type(T_ctx_volume)];
189 adm_ctx.volume = nla ? nla_get_u32(nla) : VOLUME_UNSPECIFIED;
190 nla = nested_attr_tb[__nla_type(T_ctx_conn_name)];
191 if (nla)
192 adm_ctx.conn_name = nla_data(nla);
193 } else
194 adm_ctx.volume = VOLUME_UNSPECIFIED;
195
196 adm_ctx.minor = d_in->minor;
197 adm_ctx.mdev = minor_to_mdev(d_in->minor);
0ace9dfa 198 adm_ctx.tconn = conn_get_by_name(adm_ctx.conn_name);
3b98c0c2
LE		 199
200 if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) {
201 drbd_msg_put_info("unknown minor");
202 return ERR_MINOR_INVALID;
203 }
204 if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_CONN)) {
205 drbd_msg_put_info("unknown connection");
206 return ERR_INVALID_REQUEST;
207 }
208
209 /* some more paranoia, if the request was over-determined */
527f4b24
LE		 210 if (adm_ctx.mdev && adm_ctx.tconn &&
211 adm_ctx.mdev->tconn != adm_ctx.tconn) {
212 pr_warning("request: minor=%u, conn=%s; but that minor belongs to connection %s\n",
213 adm_ctx.minor, adm_ctx.conn_name, adm_ctx.mdev->tconn->name);
214 drbd_msg_put_info("minor exists in different connection");
215 return ERR_INVALID_REQUEST;
216 }
3b98c0c2
LE		 217 if (adm_ctx.mdev &&
218 adm_ctx.volume != VOLUME_UNSPECIFIED &&
219 adm_ctx.volume != adm_ctx.mdev->vnr) {
220 pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n",
221 adm_ctx.minor, adm_ctx.volume,
222 adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name);
527f4b24 223 drbd_msg_put_info("minor exists as different volume");
3b98c0c2
LE		 224 return ERR_INVALID_REQUEST;
225 }
0ace9dfa 226
3b98c0c2
LE		 227 return NO_ERROR;
228
229fail:
230 nlmsg_free(adm_ctx.reply_skb);
231 adm_ctx.reply_skb = NULL;
232 return -ENOMEM;
233}
234
235static int drbd_adm_finish(struct genl_info *info, int retcode)
236{
237 struct nlattr *nla;
238 const char *conn_name = NULL;
239
0ace9dfa
PR		 240 if (adm_ctx.tconn) {
241 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
242 adm_ctx.tconn = NULL;
243 }
244
3b98c0c2
LE		 245 if (!adm_ctx.reply_skb)
246 return -ENOMEM;
247
248 adm_ctx.reply_dh->ret_code = retcode;
249
250 nla = info->attrs[DRBD_NLA_CFG_CONTEXT];
251 if (nla) {
252 nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name));
253 if (nla)
254 conn_name = nla_data(nla);
255 }
256
257 drbd_adm_send_reply(adm_ctx.reply_skb, info);
258 return 0;
259}
b411b363 260
6b75dced 261static void setup_khelper_env(struct drbd_tconn *tconn, char **envp)
b411b363 262{
6b75dced 263 char *afs;
44ed167d 264 struct net_conf *nc;
b411b363 265
44ed167d
PR		 266 rcu_read_lock();
267 nc = rcu_dereference(tconn->net_conf);
268 if (nc) {
269 switch (((struct sockaddr *)nc->peer_addr)->sa_family) {
b411b363
PR		 270 case AF_INET6:
271 afs = "ipv6";
6b75dced 272 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6",
44ed167d 273 &((struct sockaddr_in6 *)nc->peer_addr)->sin6_addr);
b411b363
PR		 274 break;
275 case AF_INET:
276 afs = "ipv4";
6b75dced 277 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4",
44ed167d 278 &((struct sockaddr_in *)nc->peer_addr)->sin_addr);
b411b363
PR		 279 break;
280 default:
281 afs = "ssocks";
6b75dced 282 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4",
44ed167d 283 &((struct sockaddr_in *)nc->peer_addr)->sin_addr);
b411b363 284 }
6b75dced 285 snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs);
b411b363 286 }
44ed167d 287 rcu_read_unlock();
6b75dced
PR		 288}
289
290int drbd_khelper(struct drbd_conf *mdev, char *cmd)
291{
292 char *envp[] = { "HOME=/",
293 "TERM=linux",
294 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
295 (char[20]) { }, /* address family */
296 (char[60]) { }, /* address */
297 NULL };
298 char mb[12];
299 char *argv[] = {usermode_helper, cmd, mb, NULL };
300 struct sib_info sib;
301 int ret;
302
303 snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev));
304 setup_khelper_env(mdev->tconn, envp);
b411b363 305
1090c056
LE		 306 /* The helper may take some time.
307 * write out any unsynced meta data changes now */
308 drbd_md_sync(mdev);
309
b411b363 310 dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb);
3b98c0c2
LE		 311 sib.sib_reason = SIB_HELPER_PRE;
312 sib.helper_name = cmd;
313 drbd_bcast_event(mdev, &sib);
b411b363
PR		 314 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
315 if (ret)
316 dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
317 usermode_helper, cmd, mb,
318 (ret >> 8) & 0xff, ret);
319 else
320 dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
321 usermode_helper, cmd, mb,
322 (ret >> 8) & 0xff, ret);
3b98c0c2
LE		 323 sib.sib_reason = SIB_HELPER_POST;
324 sib.helper_exit_code = ret;
325 drbd_bcast_event(mdev, &sib);
b411b363
PR		 326
327 if (ret < 0) /* Ignore any ERRNOs we got. */
328 ret = 0;
329
330 return ret;
331}
332
6b75dced
PR		 333static void conn_md_sync(struct drbd_tconn *tconn)
334{
335 struct drbd_conf *mdev;
e90285e0 336 int vnr;
6b75dced 337
d3fcb490 338 down_read(&drbd_cfg_rwsem);
e90285e0 339 idr_for_each_entry(&tconn->volumes, mdev, vnr)
6b75dced 340 drbd_md_sync(mdev);
d3fcb490 341 up_read(&drbd_cfg_rwsem);
6b75dced
PR		 342}
343
344int conn_khelper(struct drbd_tconn *tconn, char *cmd)
345{
346 char *envp[] = { "HOME=/",
347 "TERM=linux",
348 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
349 (char[20]) { }, /* address family */
350 (char[60]) { }, /* address */
351 NULL };
352 char *argv[] = {usermode_helper, cmd, tconn->name, NULL };
353 int ret;
354
355 setup_khelper_env(tconn, envp);
356 conn_md_sync(tconn);
357
358 conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name);
359 /* TODO: conn_bcast_event() ?? */
360
361 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
362 if (ret)
363 conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n",
364 usermode_helper, cmd, tconn->name,
365 (ret >> 8) & 0xff, ret);
366 else
367 conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n",
368 usermode_helper, cmd, tconn->name,
369 (ret >> 8) & 0xff, ret);
370 /* TODO: conn_bcast_event() ?? */
371
372 if (ret < 0) /* Ignore any ERRNOs we got. */
373 ret = 0;
374
375 return ret;
376}
377
cb703454 378static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn)
b411b363 379{
cb703454
PR		 380 enum drbd_fencing_p fp = FP_NOT_AVAIL;
381 struct drbd_conf *mdev;
382 int vnr;
383
695d08fa 384 rcu_read_lock();
cb703454
PR		 385 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
386 if (get_ldev_if_state(mdev, D_CONSISTENT)) {
387 fp = max_t(enum drbd_fencing_p, fp, mdev->ldev->dc.fencing);
388 put_ldev(mdev);
389 }
390 }
695d08fa 391 rcu_read_unlock();
cb703454
PR		 392
393 return fp;
394}
395
396bool conn_try_outdate_peer(struct drbd_tconn *tconn)
397{
398 union drbd_state mask = { };
399 union drbd_state val = { };
400 enum drbd_fencing_p fp;
b411b363
PR		 401 char *ex_to_string;
402 int r;
b411b363 403
cb703454
PR		 404 if (tconn->cstate >= C_WF_REPORT_PARAMS) {
405 conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n");
406 return false;
407 }
b411b363 408
cb703454
PR		 409 fp = highest_fencing_policy(tconn);
410 switch (fp) {
411 case FP_NOT_AVAIL:
412 conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n");
fb22c402 413 goto out;
cb703454
PR		 414 case FP_DONT_CARE:
415 return true;
416 default: ;
b411b363
PR		 417 }
418
cb703454 419 r = conn_khelper(tconn, "fence-peer");
b411b363
PR		 420
421 switch ((r>>8) & 0xff) {
422 case 3: /* peer is inconsistent */
423 ex_to_string = "peer is inconsistent or worse";
cb703454
PR		 424 mask.pdsk = D_MASK;
425 val.pdsk = D_INCONSISTENT;
b411b363
PR		 426 break;
427 case 4: /* peer got outdated, or was already outdated */
428 ex_to_string = "peer was fenced";
cb703454
PR		 429 mask.pdsk = D_MASK;
430 val.pdsk = D_OUTDATED;
b411b363
PR		 431 break;
432 case 5: /* peer was down */
cb703454 433 if (conn_highest_disk(tconn) == D_UP_TO_DATE) {
b411b363
PR		 434 /* we will(have) create(d) a new UUID anyways... */
435 ex_to_string = "peer is unreachable, assumed to be dead";
cb703454
PR		 436 mask.pdsk = D_MASK;
437 val.pdsk = D_OUTDATED;
b411b363
PR		 438 } else {
439 ex_to_string = "peer unreachable, doing nothing since disk != UpToDate";
b411b363
PR		 440 }
441 break;
442 case 6: /* Peer is primary, voluntarily outdate myself.
443 * This is useful when an unconnected R_SECONDARY is asked to
444 * become R_PRIMARY, but finds the other peer being active. */
445 ex_to_string = "peer is active";
cb703454
PR		 446 conn_warn(tconn, "Peer is primary, outdating myself.\n");
447 mask.disk = D_MASK;
448 val.disk = D_OUTDATED;
b411b363
PR		 449 break;
450 case 7:
451 if (fp != FP_STONITH)
cb703454 452 conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n");
b411b363 453 ex_to_string = "peer was stonithed";
cb703454
PR		 454 mask.pdsk = D_MASK;
455 val.pdsk = D_OUTDATED;
b411b363
PR		 456 break;
457 default:
458 /* The script is broken ... */
cb703454
PR		 459 conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff);
460 return false; /* Eventually leave IO frozen */
b411b363
PR		 461 }
462
cb703454
PR		 463 conn_info(tconn, "fence-peer helper returned %d (%s)\n",
464 (r>>8) & 0xff, ex_to_string);
fb22c402 465
cb703454 466 out:
fb22c402 467
cb703454
PR		 468 /* Not using
469 conn_request_state(tconn, mask, val, CS_VERBOSE);
470 here, because we might were able to re-establish the connection in the
471 meantime. */
472 spin_lock_irq(&tconn->req_lock);
473 if (tconn->cstate < C_WF_REPORT_PARAMS)
474 _conn_request_state(tconn, mask, val, CS_VERBOSE);
475 spin_unlock_irq(&tconn->req_lock);
476
477 return conn_highest_pdsk(tconn) <= D_OUTDATED;
b411b363
PR		 478}
479
87f7be4c
PR		 480static int _try_outdate_peer_async(void *data)
481{
cb703454 482 struct drbd_tconn *tconn = (struct drbd_tconn *)data;
87f7be4c 483
cb703454 484 conn_try_outdate_peer(tconn);
87f7be4c 485
9dc9fbb3 486 kref_put(&tconn->kref, &conn_destroy);
87f7be4c
PR		 487 return 0;
488}
489
cb703454 490void conn_try_outdate_peer_async(struct drbd_tconn *tconn)
87f7be4c
PR		 491{
492 struct task_struct *opa;
493
9dc9fbb3 494 kref_get(&tconn->kref);
cb703454 495 opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h");
9dc9fbb3 496 if (IS_ERR(opa)) {
cb703454 497 conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n");
9dc9fbb3
PR		 498 kref_put(&tconn->kref, &conn_destroy);
499 }
87f7be4c 500}
b411b363 501
bf885f8a
AG		 502enum drbd_state_rv
503drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
b411b363
PR		 504{
505 const int max_tries = 4;
bf885f8a 506 enum drbd_state_rv rv = SS_UNKNOWN_ERROR;
44ed167d 507 struct net_conf *nc;
b411b363
PR		 508 int try = 0;
509 int forced = 0;
510 union drbd_state mask, val;
b411b363
PR		 511
512 if (new_role == R_PRIMARY)
0625ac19 513 request_ping(mdev->tconn); /* Detect a dead peer ASAP */
b411b363 514
8410da8f 515 mutex_lock(mdev->state_mutex);
b411b363
PR		 516
517 mask.i = 0; mask.role = R_MASK;
518 val.i = 0; val.role = new_role;
519
520 while (try++ < max_tries) {
bf885f8a 521 rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE);
b411b363
PR		 522
523 /* in case we first succeeded to outdate,
524 * but now suddenly could establish a connection */
bf885f8a 525 if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) {
b411b363
PR		 526 val.pdsk = 0;
527 mask.pdsk = 0;
528 continue;
529 }
530
bf885f8a 531 if (rv == SS_NO_UP_TO_DATE_DISK && force &&
d10a33c6
PR		 532 (mdev->state.disk < D_UP_TO_DATE &&
533 mdev->state.disk >= D_INCONSISTENT)) {
b411b363
PR		 534 mask.disk = D_MASK;
535 val.disk = D_UP_TO_DATE;
536 forced = 1;
537 continue;
538 }
539
bf885f8a 540 if (rv == SS_NO_UP_TO_DATE_DISK &&
b411b363
PR		 541 mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) {
542 D_ASSERT(mdev->state.pdsk == D_UNKNOWN);
b411b363 543
cb703454 544 if (conn_try_outdate_peer(mdev->tconn)) {
b411b363
PR		 545 val.disk = D_UP_TO_DATE;
546 mask.disk = D_MASK;
547 }
b411b363
PR		 548 continue;
549 }
550
bf885f8a 551 if (rv == SS_NOTHING_TO_DO)
3b98c0c2 552 goto out;
bf885f8a 553 if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) {
cb703454 554 if (!conn_try_outdate_peer(mdev->tconn) && force) {
b411b363 555 dev_warn(DEV, "Forced into split brain situation!\n");
cb703454
PR		 556 mask.pdsk = D_MASK;
557 val.pdsk = D_OUTDATED;
b411b363 558
cb703454 559 }
b411b363
PR		 560 continue;
561 }
bf885f8a 562 if (rv == SS_TWO_PRIMARIES) {
b411b363
PR		 563 /* Maybe the peer is detected as dead very soon...
564 retry at most once more in this case. */
44ed167d
PR		 565 int timeo;
566 rcu_read_lock();
567 nc = rcu_dereference(mdev->tconn->net_conf);
568 timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1;
569 rcu_read_unlock();
570 schedule_timeout_interruptible(timeo);
b411b363
PR		 571 if (try < max_tries)
572 try = max_tries - 1;
573 continue;
574 }
bf885f8a
AG		 575 if (rv < SS_SUCCESS) {
576 rv = _drbd_request_state(mdev, mask, val,
b411b363 577 CS_VERBOSE + CS_WAIT_COMPLETE);
bf885f8a 578 if (rv < SS_SUCCESS)
3b98c0c2 579 goto out;
b411b363
PR		 580 }
581 break;
582 }
583
bf885f8a 584 if (rv < SS_SUCCESS)
3b98c0c2 585 goto out;
b411b363
PR		 586
587 if (forced)
588 dev_warn(DEV, "Forced to consider local data as UpToDate!\n");
589
590 /* Wait until nothing is on the fly :) */
591 wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0);
592
593 if (new_role == R_SECONDARY) {
81e84650 594 set_disk_ro(mdev->vdisk, true);
b411b363
PR		 595 if (get_ldev(mdev)) {
596 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
597 put_ldev(mdev);
598 }
599 } else {
91fd4dad
PR		 600 mutex_lock(&mdev->tconn->net_conf_update);
601 nc = mdev->tconn->net_conf;
44ed167d 602 if (nc)
91fd4dad
PR		 603 nc->want_lose = 0; /* without copy; single bit op is atomic */
604 mutex_unlock(&mdev->tconn->net_conf_update);
605
81e84650 606 set_disk_ro(mdev->vdisk, false);
b411b363
PR		 607 if (get_ldev(mdev)) {
608 if (((mdev->state.conn < C_CONNECTED ||
609 mdev->state.pdsk <= D_FAILED)
610 && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced)
611 drbd_uuid_new_current(mdev);
612
613 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
614 put_ldev(mdev);
615 }
616 }
617
19f843aa
LE		 618 /* writeout of activity log covered areas of the bitmap
619 * to stable storage done in after state change already */
b411b363
PR		 620
621 if (mdev->state.conn >= C_WF_REPORT_PARAMS) {
622 /* if this was forced, we should consider sync */
623 if (forced)
624 drbd_send_uuids(mdev);
625 drbd_send_state(mdev);
626 }
627
628 drbd_md_sync(mdev);
629
630 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
3b98c0c2 631out:
8410da8f 632 mutex_unlock(mdev->state_mutex);
bf885f8a 633 return rv;
b411b363
PR		 634}
635
3b98c0c2 636static const char *from_attrs_err_to_txt(int err)
b411b363 637{
3b98c0c2
LE		 638 return err == -ENOMSG ? "required attribute missing" :
639 err == -EOPNOTSUPP ? "unknown mandatory attribute" :
f399002e 640 err == -EEXIST ? "can not change invariant setting" :
3b98c0c2 641 "invalid attribute value";
b411b363
PR		 642}
643
3b98c0c2 644int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info)
b411b363 645{
3b98c0c2
LE		 646 struct set_role_parms parms;
647 int err;
648 enum drbd_ret_code retcode;
b411b363 649
3b98c0c2
LE		 650 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
651 if (!adm_ctx.reply_skb)
652 return retcode;
653 if (retcode != NO_ERROR)
654 goto out;
655
656 memset(&parms, 0, sizeof(parms));
657 if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) {
f399002e 658 err = set_role_parms_from_attrs(&parms, info);
3b98c0c2
LE		 659 if (err) {
660 retcode = ERR_MANDATORY_TAG;
661 drbd_msg_put_info(from_attrs_err_to_txt(err));
662 goto out;
663 }
664 }
665
666 if (info->genlhdr->cmd == DRBD_ADM_PRIMARY)
667 retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate);
668 else
669 retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0);
670out:
671 drbd_adm_finish(info, retcode);
b411b363
PR		 672 return 0;
673}
674
675/* initializes the md.*_offset members, so we are able to find
676 * the on disk meta data */
677static void drbd_md_set_sector_offsets(struct drbd_conf *mdev,
678 struct drbd_backing_dev *bdev)
679{
680 sector_t md_size_sect = 0;
681 switch (bdev->dc.meta_dev_idx) {
682 default:
683 /* v07 style fixed size indexed meta data */
684 bdev->md.md_size_sect = MD_RESERVED_SECT;
685 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
686 bdev->md.al_offset = MD_AL_OFFSET;
687 bdev->md.bm_offset = MD_BM_OFFSET;
688 break;
689 case DRBD_MD_INDEX_FLEX_EXT:
690 /* just occupy the full device; unit: sectors */
691 bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev);
692 bdev->md.md_offset = 0;
693 bdev->md.al_offset = MD_AL_OFFSET;
694 bdev->md.bm_offset = MD_BM_OFFSET;
695 break;
696 case DRBD_MD_INDEX_INTERNAL:
697 case DRBD_MD_INDEX_FLEX_INT:
698 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
699 /* al size is still fixed */
7ad651b5 700 bdev->md.al_offset = -MD_AL_SECTORS;
b411b363
PR		 701 /* we need (slightly less than) ~ this much bitmap sectors: */
702 md_size_sect = drbd_get_capacity(bdev->backing_bdev);
703 md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT);
704 md_size_sect = BM_SECT_TO_EXT(md_size_sect);
705 md_size_sect = ALIGN(md_size_sect, 8);
706
707 /* plus the "drbd meta data super block",
708 * and the activity log; */
709 md_size_sect += MD_BM_OFFSET;
710
711 bdev->md.md_size_sect = md_size_sect;
712 /* bitmap offset is adjusted by 'super' block size */
713 bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET;
714 break;
715 }
716}
717
4b0715f0 718/* input size is expected to be in KB */
b411b363
PR		 719char *ppsize(char *buf, unsigned long long size)
720{
4b0715f0
LE		 721 /* Needs 9 bytes at max including trailing NUL:
722 * -1ULL ==> "16384 EB" */
b411b363
PR		 723 static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
724 int base = 0;
4b0715f0 725 while (size >= 10000 && base < sizeof(units)-1) {
b411b363
PR		 726 /* shift + round */
727 size = (size >> 10) + !!(size & (1<<9));
728 base++;
729 }
4b0715f0 730 sprintf(buf, "%u %cB", (unsigned)size, units[base]);
b411b363
PR		 731
732 return buf;
733}
734
735/* there is still a theoretical deadlock when called from receiver
736 * on an D_INCONSISTENT R_PRIMARY:
737 * remote READ does inc_ap_bio, receiver would need to receive answer
738 * packet from remote to dec_ap_bio again.
739 * receiver receive_sizes(), comes here,
740 * waits for ap_bio_cnt == 0. -> deadlock.
741 * but this cannot happen, actually, because:
742 * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable
743 * (not connected, or bad/no disk on peer):
744 * see drbd_fail_request_early, ap_bio_cnt is zero.
745 * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
746 * peer may not initiate a resize.
747 */
3b98c0c2
LE		 748/* Note these are not to be confused with
749 * drbd_adm_suspend_io/drbd_adm_resume_io,
750 * which are (sub) state changes triggered by admin (drbdsetup),
751 * and can be long lived.
752 * This changes an mdev->flag, is triggered by drbd internals,
753 * and should be short-lived. */
b411b363
PR		 754void drbd_suspend_io(struct drbd_conf *mdev)
755{
756 set_bit(SUSPEND_IO, &mdev->flags);
2aebfabb 757 if (drbd_suspended(mdev))
265be2d0 758 return;
b411b363
PR		 759 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt));
760}
761
762void drbd_resume_io(struct drbd_conf *mdev)
763{
764 clear_bit(SUSPEND_IO, &mdev->flags);
765 wake_up(&mdev->misc_wait);
766}
767
768/**
769 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
770 * @mdev: DRBD device.
771 *
772 * Returns 0 on success, negative return values indicate errors.
773 * You should call drbd_md_sync() after calling this function.
774 */
24c4830c 775enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local)
b411b363
PR		 776{
777 sector_t prev_first_sect, prev_size; /* previous meta location */
778 sector_t la_size;
779 sector_t size;
780 char ppb[10];
781
782 int md_moved, la_size_changed;
783 enum determine_dev_size rv = unchanged;
784
785 /* race:
786 * application request passes inc_ap_bio,
787 * but then cannot get an AL-reference.
788 * this function later may wait on ap_bio_cnt == 0. -> deadlock.
789 *
790 * to avoid that:
791 * Suspend IO right here.
792 * still lock the act_log to not trigger ASSERTs there.
793 */
794 drbd_suspend_io(mdev);
795
796 /* no wait necessary anymore, actually we could assert that */
797 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
798
799 prev_first_sect = drbd_md_first_sector(mdev->ldev);
800 prev_size = mdev->ldev->md.md_size_sect;
801 la_size = mdev->ldev->md.la_size_sect;
802
803 /* TODO: should only be some assert here, not (re)init... */
804 drbd_md_set_sector_offsets(mdev, mdev->ldev);
805
d845030f 806 size = drbd_new_dev_size(mdev, mdev->ldev, flags & DDSF_FORCED);
b411b363
PR		 807
808 if (drbd_get_capacity(mdev->this_bdev) != size ||
809 drbd_bm_capacity(mdev) != size) {
810 int err;
02d9a94b 811 err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC));
b411b363
PR		 812 if (unlikely(err)) {
813 /* currently there is only one error: ENOMEM! */
814 size = drbd_bm_capacity(mdev)>>1;
815 if (size == 0) {
816 dev_err(DEV, "OUT OF MEMORY! "
817 "Could not allocate bitmap!\n");
818 } else {
819 dev_err(DEV, "BM resizing failed. "
820 "Leaving size unchanged at size = %lu KB\n",
821 (unsigned long)size);
822 }
823 rv = dev_size_error;
824 }
825 /* racy, see comments above. */
826 drbd_set_my_capacity(mdev, size);
827 mdev->ldev->md.la_size_sect = size;
828 dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1),
829 (unsigned long long)size>>1);
830 }
831 if (rv == dev_size_error)
832 goto out;
833
834 la_size_changed = (la_size != mdev->ldev->md.la_size_sect);
835
836 md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev)
837 || prev_size != mdev->ldev->md.md_size_sect;
838
839 if (la_size_changed || md_moved) {
24dccabb
AG		 840 int err;
841
b411b363
PR		 842 drbd_al_shrink(mdev); /* All extents inactive. */
843 dev_info(DEV, "Writing the whole bitmap, %s\n",
844 la_size_changed && md_moved ? "size changed and md moved" :
845 la_size_changed ? "size changed" : "md moved");
20ceb2b2
LE		 846 /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */
847 err = drbd_bitmap_io(mdev, &drbd_bm_write,
848 "size changed", BM_LOCKED_MASK);
24dccabb
AG		 849 if (err) {
850 rv = dev_size_error;
851 goto out;
852 }
b411b363
PR		 853 drbd_md_mark_dirty(mdev);
854 }
855
856 if (size > la_size)
857 rv = grew;
858 if (size < la_size)
859 rv = shrunk;
860out:
861 lc_unlock(mdev->act_log);
862 wake_up(&mdev->al_wait);
863 drbd_resume_io(mdev);
864
865 return rv;
866}
867
868sector_t
a393db6f 869drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, int assume_peer_has_space)
b411b363
PR		 870{
871 sector_t p_size = mdev->p_size; /* partner's disk size. */
872 sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */
873 sector_t m_size; /* my size */
874 sector_t u_size = bdev->dc.disk_size; /* size requested by user. */
875 sector_t size = 0;
876
877 m_size = drbd_get_max_capacity(bdev);
878
a393db6f
PR		 879 if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) {
880 dev_warn(DEV, "Resize while not connected was forced by the user!\n");
881 p_size = m_size;
882 }
883
b411b363
PR		 884 if (p_size && m_size) {
885 size = min_t(sector_t, p_size, m_size);
886 } else {
887 if (la_size) {
888 size = la_size;
889 if (m_size && m_size < size)
890 size = m_size;
891 if (p_size && p_size < size)
892 size = p_size;
893 } else {
894 if (m_size)
895 size = m_size;
896 if (p_size)
897 size = p_size;
898 }
899 }
900
901 if (size == 0)
902 dev_err(DEV, "Both nodes diskless!\n");
903
904 if (u_size) {
905 if (u_size > size)
906 dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n",
907 (unsigned long)u_size>>1, (unsigned long)size>>1);
908 else
909 size = u_size;
910 }
911
912 return size;
913}
914
915/**
916 * drbd_check_al_size() - Ensures that the AL is of the right size
917 * @mdev: DRBD device.
918 *
919 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
920 * failed, and 0 on success. You should call drbd_md_sync() after you called
921 * this function.
922 */
f399002e 923static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc)
b411b363
PR		 924{
925 struct lru_cache *n, *t;
926 struct lc_element *e;
927 unsigned int in_use;
928 int i;
929
f399002e
LE		 930 if (!expect(dc->al_extents >= DRBD_AL_EXTENTS_MIN))
931 dc->al_extents = DRBD_AL_EXTENTS_MIN;
b411b363
PR		 932
933 if (mdev->act_log &&
f399002e 934 mdev->act_log->nr_elements == dc->al_extents)
b411b363
PR		 935 return 0;
936
937 in_use = 0;
938 t = mdev->act_log;
7ad651b5 939 n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION,
f399002e 940 dc->al_extents, sizeof(struct lc_element), 0);
b411b363
PR		 941
942 if (n == NULL) {
943 dev_err(DEV, "Cannot allocate act_log lru!\n");
944 return -ENOMEM;
945 }
946 spin_lock_irq(&mdev->al_lock);
947 if (t) {
948 for (i = 0; i < t->nr_elements; i++) {
949 e = lc_element_by_index(t, i);
950 if (e->refcnt)
951 dev_err(DEV, "refcnt(%d)==%d\n",
952 e->lc_number, e->refcnt);
953 in_use += e->refcnt;
954 }
955 }
956 if (!in_use)
957 mdev->act_log = n;
958 spin_unlock_irq(&mdev->al_lock);
959 if (in_use) {
960 dev_err(DEV, "Activity log still in use!\n");
961 lc_destroy(n);
962 return -EBUSY;
963 } else {
964 if (t)
965 lc_destroy(t);
966 }
967 drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */
968 return 0;
969}
970
99432fcc 971static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size)
b411b363
PR		 972{
973 struct request_queue * const q = mdev->rq_queue;
99432fcc
PR		 974 int max_hw_sectors = max_bio_size >> 9;
975 int max_segments = 0;
976
977 if (get_ldev_if_state(mdev, D_ATTACHING)) {
978 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
979
980 max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9);
981 max_segments = mdev->ldev->dc.max_bio_bvecs;
982 put_ldev(mdev);
983 }
b411b363 984
b411b363 985 blk_queue_logical_block_size(q, 512);
1816a2b4
LE		 986 blk_queue_max_hw_sectors(q, max_hw_sectors);
987 /* This is the workaround for "bio would need to, but cannot, be split" */
988 blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS);
989 blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1);
b411b363 990
99432fcc
PR		 991 if (get_ldev_if_state(mdev, D_ATTACHING)) {
992 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
993
994 blk_queue_stack_limits(q, b);
995
996 if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) {
997 dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
998 q->backing_dev_info.ra_pages,
999 b->backing_dev_info.ra_pages);
1000 q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages;
1001 }
1002 put_ldev(mdev);
1003 }
1004}
1005
1006void drbd_reconsider_max_bio_size(struct drbd_conf *mdev)
1007{
1008 int now, new, local, peer;
1009
1010 now = queue_max_hw_sectors(mdev->rq_queue) << 9;
1011 local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */
1012 peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */
b411b363 1013
99432fcc
PR		 1014 if (get_ldev_if_state(mdev, D_ATTACHING)) {
1015 local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9;
1016 mdev->local_max_bio_size = local;
1017 put_ldev(mdev);
b411b363 1018 }
99432fcc
PR		 1019
1020 /* We may ignore peer limits if the peer is modern enough.
1021 Because new from 8.3.8 onwards the peer can use multiple
1022 BIOs for a single peer_request */
1023 if (mdev->state.conn >= C_CONNECTED) {
31890f4a 1024 if (mdev->tconn->agreed_pro_version < 94)
99432fcc 1025 peer = mdev->peer_max_bio_size;
31890f4a 1026 else if (mdev->tconn->agreed_pro_version == 94)
99432fcc
PR		 1027 peer = DRBD_MAX_SIZE_H80_PACKET;
1028 else /* drbd 8.3.8 onwards */
1029 peer = DRBD_MAX_BIO_SIZE;
1030 }
1031
1032 new = min_t(int, local, peer);
1033
1034 if (mdev->state.role == R_PRIMARY && new < now)
1035 dev_err(DEV, "ASSERT FAILED new < now; (%d < %d)\n", new, now);
1036
1037 if (new != now)
1038 dev_info(DEV, "max BIO size = %u\n", new);
1039
1040 drbd_setup_queue_param(mdev, new);
b411b363
PR		 1041}
1042
a18e9d1e 1043/* Starts the worker thread */
0e29d163 1044static void conn_reconfig_start(struct drbd_tconn *tconn)
b411b363 1045{
0e29d163
PR		 1046 drbd_thread_start(&tconn->worker);
1047 conn_flush_workqueue(tconn);
b411b363
PR		 1048}
1049
a18e9d1e 1050/* if still unconfigured, stops worker again. */
0e29d163 1051static void conn_reconfig_done(struct drbd_tconn *tconn)
b411b363 1052{
0e29d163 1053 spin_lock_irq(&tconn->req_lock);
a18e9d1e 1054 if (conn_all_vols_unconf(tconn))
0e29d163 1055 drbd_thread_stop_nowait(&tconn->worker);
0e29d163 1056 spin_unlock_irq(&tconn->req_lock);
b411b363
PR		 1057}
1058
0778286a
PR		 1059/* Make sure IO is suspended before calling this function(). */
1060static void drbd_suspend_al(struct drbd_conf *mdev)
1061{
1062 int s = 0;
1063
61610420 1064 if (!lc_try_lock(mdev->act_log)) {
0778286a
PR		 1065 dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n");
1066 return;
1067 }
1068
61610420 1069 drbd_al_shrink(mdev);
87eeee41 1070 spin_lock_irq(&mdev->tconn->req_lock);
0778286a
PR		 1071 if (mdev->state.conn < C_CONNECTED)
1072 s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags);
87eeee41 1073 spin_unlock_irq(&mdev->tconn->req_lock);
61610420 1074 lc_unlock(mdev->act_log);
0778286a
PR		 1075
1076 if (s)
1077 dev_info(DEV, "Suspended AL updates\n");
1078}
1079
5979e361
LE		 1080
1081static bool should_set_defaults(struct genl_info *info)
1082{
1083 unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags;
1084 return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS);
1085}
1086
1087/* Maybe we should we generate these functions
1088 * from the drbd_genl.h magic as well?
1089 * That way we would not "accidentally forget" to add defaults here. */
1090
1091#define RESET_ARRAY_FIELD(field) do { \
1092 memset(field, 0, sizeof(field)); \
1093 field ## _len = 0; \
1094} while (0)
1095void drbd_set_res_opts_default(struct res_opts *r)
1096{
1097 RESET_ARRAY_FIELD(r->cpu_mask);
1098 r->on_no_data = DRBD_ON_NO_DATA_DEF;
1099}
1100
1101static void drbd_set_net_conf_defaults(struct net_conf *nc)
1102{
1103 /* Do NOT (re)set those fields marked as GENLA_F_INVARIANT
1104 * in drbd_genl.h, they can only be change with disconnect/reconnect */
1105 RESET_ARRAY_FIELD(nc->shared_secret);
1106
1107 RESET_ARRAY_FIELD(nc->cram_hmac_alg);
1108 RESET_ARRAY_FIELD(nc->integrity_alg);
1109 RESET_ARRAY_FIELD(nc->verify_alg);
1110 RESET_ARRAY_FIELD(nc->csums_alg);
1111#undef RESET_ARRAY_FIELD
1112
1113 nc->wire_protocol = DRBD_PROTOCOL_DEF;
1114 nc->try_connect_int = DRBD_CONNECT_INT_DEF;
1115 nc->timeout = DRBD_TIMEOUT_DEF;
1116 nc->ping_int = DRBD_PING_INT_DEF;
1117 nc->ping_timeo = DRBD_PING_TIMEO_DEF;
1118 nc->sndbuf_size = DRBD_SNDBUF_SIZE_DEF;
1119 nc->rcvbuf_size = DRBD_RCVBUF_SIZE_DEF;
1120 nc->ko_count = DRBD_KO_COUNT_DEF;
1121 nc->max_buffers = DRBD_MAX_BUFFERS_DEF;
1122 nc->max_epoch_size = DRBD_MAX_EPOCH_SIZE_DEF;
1123 nc->unplug_watermark = DRBD_UNPLUG_WATERMARK_DEF;
1124 nc->after_sb_0p = DRBD_AFTER_SB_0P_DEF;
1125 nc->after_sb_1p = DRBD_AFTER_SB_1P_DEF;
1126 nc->after_sb_2p = DRBD_AFTER_SB_2P_DEF;
1127 nc->rr_conflict = DRBD_RR_CONFLICT_DEF;
1128 nc->on_congestion = DRBD_ON_CONGESTION_DEF;
1129 nc->cong_fill = DRBD_CONG_FILL_DEF;
1130 nc->cong_extents = DRBD_CONG_EXTENTS_DEF;
1131 nc->two_primaries = 0;
1132 nc->no_cork = 0;
1133 nc->always_asbp = 0;
1134 nc->use_rle = 0;
1135}
1136
1137static void drbd_set_disk_conf_defaults(struct disk_conf *dc)
1138{
1139 /* Do NOT (re)set those fields marked as GENLA_F_INVARIANT
1140 * in drbd_genl.h, they can only be change with detach/reattach */
1141 dc->on_io_error = DRBD_ON_IO_ERROR_DEF;
1142 dc->fencing = DRBD_FENCING_DEF;
1143 dc->resync_rate = DRBD_RATE_DEF;
1144 dc->resync_after = DRBD_AFTER_DEF;
1145 dc->al_extents = DRBD_AL_EXTENTS_DEF;
1146 dc->c_plan_ahead = DRBD_C_PLAN_AHEAD_DEF;
1147 dc->c_delay_target = DRBD_C_DELAY_TARGET_DEF;
1148 dc->c_fill_target = DRBD_C_FILL_TARGET_DEF;
1149 dc->c_max_rate = DRBD_C_MAX_RATE_DEF;
1150 dc->c_min_rate = DRBD_C_MIN_RATE_DEF;
1151 dc->no_disk_barrier = 0;
1152 dc->no_disk_flush = 0;
1153 dc->no_disk_drain = 0;
1154 dc->no_md_flush = 0;
1155}
1156
1157
f399002e
LE		 1158int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info)
1159{
1160 enum drbd_ret_code retcode;
1161 struct drbd_conf *mdev;
5ecc72c3 1162 struct disk_conf *new_disk_conf;
f399002e
LE		 1163 int err, fifo_size;
1164 int *rs_plan_s = NULL;
1165
1166 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1167 if (!adm_ctx.reply_skb)
1168 return retcode;
1169 if (retcode != NO_ERROR)
1170 goto out;
1171
1172 mdev = adm_ctx.mdev;
1173
1174 /* we also need a disk
1175 * to change the options on */
1176 if (!get_ldev(mdev)) {
1177 retcode = ERR_NO_DISK;
1178 goto out;
1179 }
1180
1181/* FIXME freeze IO, cluster wide.
1182 *
1183 * We should make sure no-one uses
1184 * some half-updated struct when we
1185 * assign it later. */
1186
5ecc72c3
LE		 1187 new_disk_conf = kmalloc(sizeof(*new_disk_conf), GFP_KERNEL);
1188 if (!new_disk_conf) {
f399002e
LE		 1189 retcode = ERR_NOMEM;
1190 goto fail;
1191 }
1192
5ecc72c3 1193 memcpy(new_disk_conf, &mdev->ldev->dc, sizeof(*new_disk_conf));
5979e361 1194 if (should_set_defaults(info))
5ecc72c3 1195 drbd_set_disk_conf_defaults(new_disk_conf);
5979e361 1196
5ecc72c3 1197 err = disk_conf_from_attrs_for_change(new_disk_conf, info);
f399002e
LE		 1198 if (err) {
1199 retcode = ERR_MANDATORY_TAG;
1200 drbd_msg_put_info(from_attrs_err_to_txt(err));
1201 }
1202
5ecc72c3
LE		 1203 if (!expect(new_disk_conf->resync_rate >= 1))
1204 new_disk_conf->resync_rate = 1;
f399002e
LE		 1205
1206 /* clip to allowed range */
5ecc72c3
LE		 1207 if (!expect(new_disk_conf->al_extents >= DRBD_AL_EXTENTS_MIN))
1208 new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN;
1209 if (!expect(new_disk_conf->al_extents <= DRBD_AL_EXTENTS_MAX))
1210 new_disk_conf->al_extents = DRBD_AL_EXTENTS_MAX;
f399002e
LE		 1211
1212 /* most sanity checks done, try to assign the new sync-after
1213 * dependency. need to hold the global lock in there,
1214 * to avoid a race in the dependency loop check. */
5ecc72c3 1215 retcode = drbd_alter_sa(mdev, new_disk_conf->resync_after);
f399002e
LE		 1216 if (retcode != NO_ERROR)
1217 goto fail;
1218
5ecc72c3 1219 fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ;
f399002e
LE		 1220 if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) {
1221 rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL);
1222 if (!rs_plan_s) {
1223 dev_err(DEV, "kmalloc of fifo_buffer failed");
1224 retcode = ERR_NOMEM;
1225 goto fail;
1226 }
1227 }
1228
1229 if (fifo_size != mdev->rs_plan_s.size) {
1230 kfree(mdev->rs_plan_s.values);
1231 mdev->rs_plan_s.values = rs_plan_s;
1232 mdev->rs_plan_s.size = fifo_size;
1233 mdev->rs_planed = 0;
1234 rs_plan_s = NULL;
1235 }
1236
1237 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
1238 drbd_al_shrink(mdev);
5ecc72c3 1239 err = drbd_check_al_size(mdev, new_disk_conf);
f399002e
LE		 1240 lc_unlock(mdev->act_log);
1241 wake_up(&mdev->al_wait);
1242
1243 if (err) {
1244 retcode = ERR_NOMEM;
1245 goto fail;
1246 }
1247
1248 /* FIXME
1249 * To avoid someone looking at a half-updated struct, we probably
1250 * should have a rw-semaphor on net_conf and disk_conf.
1251 */
5ecc72c3 1252 mdev->ldev->dc = *new_disk_conf;
f399002e
LE		 1253
1254 drbd_md_sync(mdev);
1255
1256
1257 if (mdev->state.conn >= C_CONNECTED)
1258 drbd_send_sync_param(mdev);
1259
1260 fail:
1261 put_ldev(mdev);
5ecc72c3 1262 kfree(new_disk_conf);
f399002e
LE		 1263 kfree(rs_plan_s);
1264 out:
1265 drbd_adm_finish(info, retcode);
1266 return 0;
1267}
1268
3b98c0c2 1269int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info)
b411b363 1270{
3b98c0c2
LE		 1271 struct drbd_conf *mdev;
1272 int err;
116676ca 1273 enum drbd_ret_code retcode;
b411b363
PR		 1274 enum determine_dev_size dd;
1275 sector_t max_possible_sectors;
1276 sector_t min_md_device_sectors;
1277 struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */
e525fd89 1278 struct block_device *bdev;
b411b363
PR		 1279 struct lru_cache *resync_lru = NULL;
1280 union drbd_state ns, os;
f2024e7c 1281 enum drbd_state_rv rv;
44ed167d 1282 struct net_conf *nc;
b411b363 1283 int cp_discovered = 0;
b411b363 1284
3b98c0c2
LE		 1285 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1286 if (!adm_ctx.reply_skb)
1287 return retcode;
1288 if (retcode != NO_ERROR)
40cbf085 1289 goto finish;
3b98c0c2
LE		 1290
1291 mdev = adm_ctx.mdev;
0e29d163 1292 conn_reconfig_start(mdev->tconn);
b411b363
PR		 1293
1294 /* if you want to reconfigure, please tear down first */
1295 if (mdev->state.disk > D_DISKLESS) {
1296 retcode = ERR_DISK_CONFIGURED;
1297 goto fail;
1298 }
82f59cc6
LE		 1299 /* It may just now have detached because of IO error. Make sure
1300 * drbd_ldev_destroy is done already, we may end up here very fast,
1301 * e.g. if someone calls attach from the on-io-error handler,
1302 * to realize a "hot spare" feature (not that I'd recommend that) */
1303 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
b411b363 1304
3b98c0c2 1305 /* allocation not in the IO path, drbdsetup context */
b411b363
PR		 1306 nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL);
1307 if (!nbc) {
1308 retcode = ERR_NOMEM;
1309 goto fail;
1310 }
1311
5979e361 1312 drbd_set_disk_conf_defaults(&nbc->dc);
f399002e
LE		 1313
1314 err = disk_conf_from_attrs(&nbc->dc, info);
3b98c0c2 1315 if (err) {
b411b363 1316 retcode = ERR_MANDATORY_TAG;
3b98c0c2 1317 drbd_msg_put_info(from_attrs_err_to_txt(err));
b411b363
PR		 1318 goto fail;
1319 }
1320
3b98c0c2 1321 if ((int)nbc->dc.meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) {
b411b363
PR		 1322 retcode = ERR_MD_IDX_INVALID;
1323 goto fail;
1324 }
1325
44ed167d
PR		 1326 rcu_read_lock();
1327 nc = rcu_dereference(mdev->tconn->net_conf);
1328 if (nc) {
1329 if (nbc->dc.fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) {
1330 rcu_read_unlock();
47ff2d0a
PR		 1331 retcode = ERR_STONITH_AND_PROT_A;
1332 goto fail;
1333 }
1334 }
44ed167d 1335 rcu_read_unlock();
47ff2d0a 1336
d4d77629
TH		 1337 bdev = blkdev_get_by_path(nbc->dc.backing_dev,
1338 FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev);
e525fd89 1339 if (IS_ERR(bdev)) {
b411b363 1340 dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.backing_dev,
e525fd89 1341 PTR_ERR(bdev));
b411b363
PR		 1342 retcode = ERR_OPEN_DISK;
1343 goto fail;
1344 }
e525fd89
TH		 1345 nbc->backing_bdev = bdev;
1346
1347 /*
1348 * meta_dev_idx >= 0: external fixed size, possibly multiple
1349 * drbd sharing one meta device. TODO in that case, paranoia
1350 * check that [md_bdev, meta_dev_idx] is not yet used by some
1351 * other drbd minor! (if you use drbd.conf + drbdadm, that
1352 * should check it for you already; but if you don't, or
1353 * someone fooled it, we need to double check here)
1354 */
d4d77629
TH		 1355 bdev = blkdev_get_by_path(nbc->dc.meta_dev,
1356 FMODE_READ | FMODE_WRITE | FMODE_EXCL,
3b98c0c2 1357 ((int)nbc->dc.meta_dev_idx < 0) ?
d4d77629 1358 (void *)mdev : (void *)drbd_m_holder);
e525fd89 1359 if (IS_ERR(bdev)) {
b411b363 1360 dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.meta_dev,
e525fd89 1361 PTR_ERR(bdev));
b411b363
PR		 1362 retcode = ERR_OPEN_MD_DISK;
1363 goto fail;
1364 }
e525fd89 1365 nbc->md_bdev = bdev;
b411b363 1366
e525fd89
TH		 1367 if ((nbc->backing_bdev == nbc->md_bdev) !=
1368 (nbc->dc.meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1369 nbc->dc.meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) {
1370 retcode = ERR_MD_IDX_INVALID;
b411b363
PR		 1371 goto fail;
1372 }
1373
1374 resync_lru = lc_create("resync", drbd_bm_ext_cache,
46a15bc3 1375 1, 61, sizeof(struct bm_extent),
b411b363
PR		 1376 offsetof(struct bm_extent, lce));
1377 if (!resync_lru) {
1378 retcode = ERR_NOMEM;
e525fd89 1379 goto fail;
b411b363
PR		 1380 }
1381
1382 /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */
1383 drbd_md_set_sector_offsets(mdev, nbc);
1384
1385 if (drbd_get_max_capacity(nbc) < nbc->dc.disk_size) {
1386 dev_err(DEV, "max capacity %llu smaller than disk size %llu\n",
1387 (unsigned long long) drbd_get_max_capacity(nbc),
1388 (unsigned long long) nbc->dc.disk_size);
1389 retcode = ERR_DISK_TO_SMALL;
e525fd89 1390 goto fail;
b411b363
PR		 1391 }
1392
3b98c0c2 1393 if ((int)nbc->dc.meta_dev_idx < 0) {
b411b363
PR		 1394 max_possible_sectors = DRBD_MAX_SECTORS_FLEX;
1395 /* at least one MB, otherwise it does not make sense */
1396 min_md_device_sectors = (2<<10);
1397 } else {
1398 max_possible_sectors = DRBD_MAX_SECTORS;
1399 min_md_device_sectors = MD_RESERVED_SECT * (nbc->dc.meta_dev_idx + 1);
1400 }
1401
b411b363
PR		 1402 if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) {
1403 retcode = ERR_MD_DISK_TO_SMALL;
1404 dev_warn(DEV, "refusing attach: md-device too small, "
1405 "at least %llu sectors needed for this meta-disk type\n",
1406 (unsigned long long) min_md_device_sectors);
e525fd89 1407 goto fail;
b411b363
PR		 1408 }
1409
1410 /* Make sure the new disk is big enough
1411 * (we may currently be R_PRIMARY with no local disk...) */
1412 if (drbd_get_max_capacity(nbc) <
1413 drbd_get_capacity(mdev->this_bdev)) {
1414 retcode = ERR_DISK_TO_SMALL;
e525fd89 1415 goto fail;
b411b363
PR		 1416 }
1417
1418 nbc->known_size = drbd_get_capacity(nbc->backing_bdev);
1419
1352994b
LE		 1420 if (nbc->known_size > max_possible_sectors) {
1421 dev_warn(DEV, "==> truncating very big lower level device "
1422 "to currently maximum possible %llu sectors <==\n",
1423 (unsigned long long) max_possible_sectors);
3b98c0c2 1424 if ((int)nbc->dc.meta_dev_idx >= 0)
1352994b
LE		 1425 dev_warn(DEV, "==>> using internal or flexible "
1426 "meta data may help <<==\n");
1427 }
1428
b411b363
PR		 1429 drbd_suspend_io(mdev);
1430 /* also wait for the last barrier ack. */
2aebfabb 1431 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev));
b411b363 1432 /* and for any other previously queued work */
a21e9298 1433 drbd_flush_workqueue(mdev);
b411b363 1434
f2024e7c
AG		 1435 rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE);
1436 retcode = rv; /* FIXME: Type mismatch. */
b411b363 1437 drbd_resume_io(mdev);
f2024e7c 1438 if (rv < SS_SUCCESS)
e525fd89 1439 goto fail;
b411b363
PR		 1440
1441 if (!get_ldev_if_state(mdev, D_ATTACHING))
1442 goto force_diskless;
1443
1444 drbd_md_set_sector_offsets(mdev, nbc);
1445
1446 if (!mdev->bitmap) {
1447 if (drbd_bm_init(mdev)) {
1448 retcode = ERR_NOMEM;
1449 goto force_diskless_dec;
1450 }
1451 }
1452
1453 retcode = drbd_md_read(mdev, nbc);
1454 if (retcode != NO_ERROR)
1455 goto force_diskless_dec;
1456
1457 if (mdev->state.conn < C_CONNECTED &&
1458 mdev->state.role == R_PRIMARY &&
1459 (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) {
1460 dev_err(DEV, "Can only attach to data with current UUID=%016llX\n",
1461 (unsigned long long)mdev->ed_uuid);
1462 retcode = ERR_DATA_NOT_CURRENT;
1463 goto force_diskless_dec;
1464 }
1465
1466 /* Since we are diskless, fix the activity log first... */
f399002e 1467 if (drbd_check_al_size(mdev, &nbc->dc)) {
b411b363
PR		 1468 retcode = ERR_NOMEM;
1469 goto force_diskless_dec;
1470 }
1471
1472 /* Prevent shrinking of consistent devices ! */
1473 if (drbd_md_test_flag(nbc, MDF_CONSISTENT) &&
a393db6f 1474 drbd_new_dev_size(mdev, nbc, 0) < nbc->md.la_size_sect) {
b411b363
PR		 1475 dev_warn(DEV, "refusing to truncate a consistent device\n");
1476 retcode = ERR_DISK_TO_SMALL;
1477 goto force_diskless_dec;
1478 }
1479
1480 if (!drbd_al_read_log(mdev, nbc)) {
1481 retcode = ERR_IO_MD_DISK;
1482 goto force_diskless_dec;
1483 }
1484
b411b363
PR		 1485 /* Reset the "barriers don't work" bits here, then force meta data to
1486 * be written, to ensure we determine if barriers are supported. */
1487 if (nbc->dc.no_md_flush)
a8a4e51e 1488 set_bit(MD_NO_FUA, &mdev->flags);
b411b363 1489 else
a8a4e51e 1490 clear_bit(MD_NO_FUA, &mdev->flags);
b411b363
PR		 1491
1492 /* Point of no return reached.
1493 * Devices and memory are no longer released by error cleanup below.
1494 * now mdev takes over responsibility, and the state engine should
1495 * clean it up somewhere. */
1496 D_ASSERT(mdev->ldev == NULL);
1497 mdev->ldev = nbc;
1498 mdev->resync = resync_lru;
1499 nbc = NULL;
1500 resync_lru = NULL;
1501
2451fc3b
PR		 1502 mdev->write_ordering = WO_bdev_flush;
1503 drbd_bump_write_ordering(mdev, WO_bdev_flush);
b411b363
PR		 1504
1505 if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY))
1506 set_bit(CRASHED_PRIMARY, &mdev->flags);
1507 else
1508 clear_bit(CRASHED_PRIMARY, &mdev->flags);
1509
894c6a94 1510 if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
da9fbc27 1511 !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod)) {
b411b363
PR		 1512 set_bit(CRASHED_PRIMARY, &mdev->flags);
1513 cp_discovered = 1;
1514 }
1515
1516 mdev->send_cnt = 0;
1517 mdev->recv_cnt = 0;
1518 mdev->read_cnt = 0;
1519 mdev->writ_cnt = 0;
1520
99432fcc 1521 drbd_reconsider_max_bio_size(mdev);
b411b363
PR		 1522
1523 /* If I am currently not R_PRIMARY,
1524 * but meta data primary indicator is set,
1525 * I just now recover from a hard crash,
1526 * and have been R_PRIMARY before that crash.
1527 *
1528 * Now, if I had no connection before that crash
1529 * (have been degraded R_PRIMARY), chances are that
1530 * I won't find my peer now either.
1531 *
1532 * In that case, and _only_ in that case,
1533 * we use the degr-wfc-timeout instead of the default,
1534 * so we can automatically recover from a crash of a
1535 * degraded but active "cluster" after a certain timeout.
1536 */
1537 clear_bit(USE_DEGR_WFC_T, &mdev->flags);
1538 if (mdev->state.role != R_PRIMARY &&
1539 drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
1540 !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND))
1541 set_bit(USE_DEGR_WFC_T, &mdev->flags);
1542
24c4830c 1543 dd = drbd_determine_dev_size(mdev, 0);
b411b363
PR		 1544 if (dd == dev_size_error) {
1545 retcode = ERR_NOMEM_BITMAP;
1546 goto force_diskless_dec;
1547 } else if (dd == grew)
1548 set_bit(RESYNC_AFTER_NEG, &mdev->flags);
1549
1550 if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) {
1551 dev_info(DEV, "Assuming that all blocks are out of sync "
1552 "(aka FullSync)\n");
20ceb2b2
LE		 1553 if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write,
1554 "set_n_write from attaching", BM_LOCKED_MASK)) {
b411b363
PR		 1555 retcode = ERR_IO_MD_DISK;
1556 goto force_diskless_dec;
1557 }
1558 } else {
20ceb2b2 1559 if (drbd_bitmap_io(mdev, &drbd_bm_read,
22ab6a30 1560 "read from attaching", BM_LOCKED_MASK)) {
b411b363
PR		 1561 retcode = ERR_IO_MD_DISK;
1562 goto force_diskless_dec;
1563 }
1564 }
1565
1566 if (cp_discovered) {
1567 drbd_al_apply_to_bm(mdev);
20ceb2b2
LE		 1568 if (drbd_bitmap_io(mdev, &drbd_bm_write,
1569 "crashed primary apply AL", BM_LOCKED_MASK)) {
19f843aa
LE		 1570 retcode = ERR_IO_MD_DISK;
1571 goto force_diskless_dec;
1572 }
b411b363
PR		 1573 }
1574
0778286a
PR		 1575 if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev))
1576 drbd_suspend_al(mdev); /* IO is still suspended here... */
1577
87eeee41 1578 spin_lock_irq(&mdev->tconn->req_lock);
78bae59b
PR		 1579 os = drbd_read_state(mdev);
1580 ns = os;
b411b363
PR		 1581 /* If MDF_CONSISTENT is not set go into inconsistent state,
1582 otherwise investigate MDF_WasUpToDate...
1583 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1584 otherwise into D_CONSISTENT state.
1585 */
1586 if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) {
1587 if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE))
1588 ns.disk = D_CONSISTENT;
1589 else
1590 ns.disk = D_OUTDATED;
1591 } else {
1592 ns.disk = D_INCONSISTENT;
1593 }
1594
1595 if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED))
1596 ns.pdsk = D_OUTDATED;
1597
1598 if ( ns.disk == D_CONSISTENT &&
1599 (ns.pdsk == D_OUTDATED || mdev->ldev->dc.fencing == FP_DONT_CARE))
1600 ns.disk = D_UP_TO_DATE;
1601
1602 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1603 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1604 this point, because drbd_request_state() modifies these
1605 flags. */
1606
1607 /* In case we are C_CONNECTED postpone any decision on the new disk
1608 state after the negotiation phase. */
1609 if (mdev->state.conn == C_CONNECTED) {
1610 mdev->new_state_tmp.i = ns.i;
1611 ns.i = os.i;
1612 ns.disk = D_NEGOTIATING;
dc66c74d
PR		 1613
1614 /* We expect to receive up-to-date UUIDs soon.
1615 To avoid a race in receive_state, free p_uuid while
1616 holding req_lock. I.e. atomic with the state change */
1617 kfree(mdev->p_uuid);
1618 mdev->p_uuid = NULL;
b411b363
PR		 1619 }
1620
1621 rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
87eeee41 1622 spin_unlock_irq(&mdev->tconn->req_lock);
b411b363
PR		 1623
1624 if (rv < SS_SUCCESS)
1625 goto force_diskless_dec;
1626
1627 if (mdev->state.role == R_PRIMARY)
1628 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
1629 else
1630 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
1631
1632 drbd_md_mark_dirty(mdev);
1633 drbd_md_sync(mdev);
1634
1635 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
1636 put_ldev(mdev);
0e29d163 1637 conn_reconfig_done(mdev->tconn);
3b98c0c2 1638 drbd_adm_finish(info, retcode);
b411b363
PR		 1639 return 0;
1640
1641 force_diskless_dec:
1642 put_ldev(mdev);
1643 force_diskless:
82f59cc6 1644 drbd_force_state(mdev, NS(disk, D_FAILED));
b411b363 1645 drbd_md_sync(mdev);
b411b363 1646 fail:
40cbf085 1647 conn_reconfig_done(mdev->tconn);
b411b363 1648 if (nbc) {
e525fd89
TH		 1649 if (nbc->backing_bdev)
1650 blkdev_put(nbc->backing_bdev,
1651 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
1652 if (nbc->md_bdev)
1653 blkdev_put(nbc->md_bdev,
1654 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
b411b363
PR		 1655 kfree(nbc);
1656 }
1657 lc_destroy(resync_lru);
1658
40cbf085 1659 finish:
3b98c0c2 1660 drbd_adm_finish(info, retcode);
b411b363
PR		 1661 return 0;
1662}
1663
85f75dd7
LE		 1664static int adm_detach(struct drbd_conf *mdev)
1665{
19f83c76 1666 enum drbd_state_rv retcode;
85f75dd7
LE		 1667 drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */
1668 retcode = drbd_request_state(mdev, NS(disk, D_DISKLESS));
1669 wait_event(mdev->misc_wait,
1670 mdev->state.disk != D_DISKLESS ||
1671 !atomic_read(&mdev->local_cnt));
1672 drbd_resume_io(mdev);
1673 return retcode;
1674}
1675
82f59cc6
LE		 1676/* Detaching the disk is a process in multiple stages. First we need to lock
1677 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1678 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1679 * internal references as well.
1680 * Only then we have finally detached. */
3b98c0c2 1681int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info)
b411b363 1682{
9a0d9d03 1683 enum drbd_ret_code retcode;
3b98c0c2
LE		 1684
1685 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1686 if (!adm_ctx.reply_skb)
1687 return retcode;
1688 if (retcode != NO_ERROR)
1689 goto out;
1690
85f75dd7 1691 retcode = adm_detach(adm_ctx.mdev);
3b98c0c2
LE		 1692out:
1693 drbd_adm_finish(info, retcode);
b411b363
PR		 1694 return 0;
1695}
1696
f399002e
LE		 1697static bool conn_resync_running(struct drbd_tconn *tconn)
1698{
1699 struct drbd_conf *mdev;
695d08fa 1700 bool rv = false;
f399002e
LE		 1701 int vnr;
1702
695d08fa 1703 rcu_read_lock();
f399002e
LE		 1704 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1705 if (mdev->state.conn == C_SYNC_SOURCE ||
1706 mdev->state.conn == C_SYNC_TARGET ||
1707 mdev->state.conn == C_PAUSED_SYNC_S ||
695d08fa
PR		 1708 mdev->state.conn == C_PAUSED_SYNC_T) {
1709 rv = true;
1710 break;
1711 }
f399002e 1712 }
695d08fa
PR		 1713 rcu_read_unlock();
1714
1715 return rv;
f399002e
LE		 1716}
1717
1718static bool conn_ov_running(struct drbd_tconn *tconn)
1719{
1720 struct drbd_conf *mdev;
695d08fa 1721 bool rv = false;
f399002e
LE		 1722 int vnr;
1723
695d08fa 1724 rcu_read_lock();
f399002e
LE		 1725 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1726 if (mdev->state.conn == C_VERIFY_S ||
695d08fa
PR		 1727 mdev->state.conn == C_VERIFY_T) {
1728 rv = true;
1729 break;
1730 }
f399002e 1731 }
695d08fa
PR		 1732 rcu_read_unlock();
1733
1734 return rv;
f399002e
LE		 1735}
1736
cd64397c 1737static enum drbd_ret_code
44ed167d 1738_check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf)
cd64397c
PR		 1739{
1740 struct drbd_conf *mdev;
1741 int i;
1742
44ed167d 1743 if (old_conf && tconn->agreed_pro_version < 100 &&
b032b6fa 1744 tconn->cstate == C_WF_REPORT_PARAMS &&
44ed167d 1745 new_conf->wire_protocol != old_conf->wire_protocol)
b032b6fa
PR		 1746 return ERR_NEED_APV_100;
1747
cd64397c
PR		 1748 if (new_conf->two_primaries &&
1749 (new_conf->wire_protocol != DRBD_PROT_C))
1750 return ERR_NOT_PROTO_C;
1751
cd64397c
PR		 1752 idr_for_each_entry(&tconn->volumes, mdev, i) {
1753 if (get_ldev(mdev)) {
1754 enum drbd_fencing_p fp = mdev->ldev->dc.fencing;
1755 put_ldev(mdev);
44ed167d 1756 if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH)
cd64397c 1757 return ERR_STONITH_AND_PROT_A;
cd64397c 1758 }
44ed167d 1759 if (mdev->state.role == R_PRIMARY && new_conf->want_lose)
cd64397c 1760 return ERR_DISCARD;
cd64397c 1761 }
cd64397c
PR		 1762
1763 if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A)
1764 return ERR_CONG_NOT_PROTO_A;
1765
1766 return NO_ERROR;
1767}
1768
44ed167d
PR		 1769static enum drbd_ret_code
1770check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf)
1771{
1772 static enum drbd_ret_code rv;
1773 struct drbd_conf *mdev;
1774 int i;
1775
1776 rcu_read_lock();
1777 rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf);
1778 rcu_read_unlock();
1779
1780 /* tconn->volumes protected by genl_lock() here */
1781 idr_for_each_entry(&tconn->volumes, mdev, i) {
1782 if (!mdev->bitmap) {
1783 if(drbd_bm_init(mdev))
1784 return ERR_NOMEM;
1785 }
1786 }
1787
1788 return rv;
1789}
1790
0fd0ea06
PR		 1791struct crypto {
1792 struct crypto_hash *verify_tfm;
1793 struct crypto_hash *csums_tfm;
1794 struct crypto_hash *cram_hmac_tfm;
8d412fc6 1795 struct crypto_hash *integrity_tfm;
0fd0ea06
PR		 1796 void *int_dig_in;
1797 void *int_dig_vv;
1798};
1799
1800static int
4b6ad6d4 1801alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg)
0fd0ea06
PR		 1802{
1803 if (!tfm_name[0])
1804 return NO_ERROR;
1805
1806 *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC);
1807 if (IS_ERR(*tfm)) {
1808 *tfm = NULL;
1809 return err_alg;
1810 }
1811
0fd0ea06
PR		 1812 return NO_ERROR;
1813}
1814
1815static enum drbd_ret_code
1816alloc_crypto(struct crypto *crypto, struct net_conf *new_conf)
1817{
1818 char hmac_name[CRYPTO_MAX_ALG_NAME];
1819 enum drbd_ret_code rv;
1820 int hash_size;
1821
4b6ad6d4
AG		 1822 rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg,
1823 ERR_CSUMS_ALG);
0fd0ea06
PR		 1824 if (rv != NO_ERROR)
1825 return rv;
4b6ad6d4
AG		 1826 rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg,
1827 ERR_VERIFY_ALG);
0fd0ea06
PR		 1828 if (rv != NO_ERROR)
1829 return rv;
4b6ad6d4
AG		 1830 rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg,
1831 ERR_INTEGRITY_ALG);
0fd0ea06
PR		 1832 if (rv != NO_ERROR)
1833 return rv;
0fd0ea06
PR		 1834 if (new_conf->cram_hmac_alg[0] != 0) {
1835 snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)",
1836 new_conf->cram_hmac_alg);
1837
4b6ad6d4
AG		 1838 rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name,
1839 ERR_AUTH_ALG);
0fd0ea06 1840 }
8d412fc6
AG		 1841 if (crypto->integrity_tfm) {
1842 hash_size = crypto_hash_digestsize(crypto->integrity_tfm);
0fd0ea06
PR		 1843 crypto->int_dig_in = kmalloc(hash_size, GFP_KERNEL);
1844 if (!crypto->int_dig_in)
1845 return ERR_NOMEM;
1846 crypto->int_dig_vv = kmalloc(hash_size, GFP_KERNEL);
1847 if (!crypto->int_dig_vv)
1848 return ERR_NOMEM;
1849 }
1850
1851 return rv;
1852}
1853
1854static void free_crypto(struct crypto *crypto)
1855{
1856 kfree(crypto->int_dig_in);
1857 kfree(crypto->int_dig_vv);
1858 crypto_free_hash(crypto->cram_hmac_tfm);
8d412fc6 1859 crypto_free_hash(crypto->integrity_tfm);
0fd0ea06
PR		 1860 crypto_free_hash(crypto->csums_tfm);
1861 crypto_free_hash(crypto->verify_tfm);
1862}
1863
f399002e
LE		 1864int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info)
1865{
1866 enum drbd_ret_code retcode;
1867 struct drbd_tconn *tconn;
44ed167d 1868 struct net_conf *old_conf, *new_conf = NULL;
f399002e
LE		 1869 int err;
1870 int ovr; /* online verify running */
1871 int rsr; /* re-sync running */
0fd0ea06 1872 struct crypto crypto = { };
88104ca4 1873 bool change_integrity_alg;
f399002e
LE		 1874
1875 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
1876 if (!adm_ctx.reply_skb)
1877 return retcode;
1878 if (retcode != NO_ERROR)
1879 goto out;
1880
1881 tconn = adm_ctx.tconn;
1882
1883 new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
1884 if (!new_conf) {
1885 retcode = ERR_NOMEM;
1886 goto out;
1887 }
1888
44ed167d
PR		 1889 conn_reconfig_start(tconn);
1890
88104ca4 1891 mutex_lock(&tconn->data.mutex);
91fd4dad
PR		 1892 mutex_lock(&tconn->net_conf_update);
1893 old_conf = tconn->net_conf;
44ed167d
PR		 1894
1895 if (!old_conf) {
f399002e
LE		 1896 drbd_msg_put_info("net conf missing, try connect");
1897 retcode = ERR_INVALID_REQUEST;
91fd4dad 1898 goto fail;
f399002e
LE		 1899 }
1900
44ed167d 1901 *new_conf = *old_conf;
5979e361
LE		 1902 if (should_set_defaults(info))
1903 drbd_set_net_conf_defaults(new_conf);
f399002e 1904
f399002e
LE		 1905 err = net_conf_from_attrs_for_change(new_conf, info);
1906 if (err) {
1907 retcode = ERR_MANDATORY_TAG;
1908 drbd_msg_put_info(from_attrs_err_to_txt(err));
1909 goto fail;
1910 }
1911
cd64397c
PR		 1912 retcode = check_net_options(tconn, new_conf);
1913 if (retcode != NO_ERROR)
1914 goto fail;
1915
f399002e
LE		 1916 /* re-sync running */
1917 rsr = conn_resync_running(tconn);
0fd0ea06 1918 if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) {
f399002e 1919 retcode = ERR_CSUMS_RESYNC_RUNNING;
91fd4dad 1920 goto fail;
f399002e
LE		 1921 }
1922
f399002e
LE		 1923 /* online verify running */
1924 ovr = conn_ov_running(tconn);
0fd0ea06
PR		 1925 if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) {
1926 retcode = ERR_VERIFY_RUNNING;
1927 goto fail;
f399002e
LE		 1928 }
1929
88104ca4
AG		 1930 change_integrity_alg = strcmp(old_conf->integrity_alg,
1931 new_conf->integrity_alg);
1932
0fd0ea06
PR		 1933 retcode = alloc_crypto(&crypto, new_conf);
1934 if (retcode != NO_ERROR)
1935 goto fail;
f399002e 1936
44ed167d 1937 rcu_assign_pointer(tconn->net_conf, new_conf);
f399002e
LE		 1938
1939 if (!rsr) {
1940 crypto_free_hash(tconn->csums_tfm);
0fd0ea06
PR		 1941 tconn->csums_tfm = crypto.csums_tfm;
1942 crypto.csums_tfm = NULL;
f399002e
LE		 1943 }
1944 if (!ovr) {
1945 crypto_free_hash(tconn->verify_tfm);
0fd0ea06
PR		 1946 tconn->verify_tfm = crypto.verify_tfm;
1947 crypto.verify_tfm = NULL;
f399002e
LE		 1948 }
1949
0fd0ea06
PR		 1950 kfree(tconn->int_dig_in);
1951 tconn->int_dig_in = crypto.int_dig_in;
1952 kfree(tconn->int_dig_vv);
1953 tconn->int_dig_vv = crypto.int_dig_vv;
8d412fc6
AG		 1954 crypto_free_hash(tconn->integrity_tfm);
1955 tconn->integrity_tfm = crypto.integrity_tfm;
88104ca4
AG		 1956 if (change_integrity_alg) {
1957 /* Do this without trying to take tconn->data.mutex again. */
1958 if (__drbd_send_protocol(tconn))
1959 goto fail;
1960 }
0fd0ea06
PR		 1961
1962 /* FIXME Changing cram_hmac while the connection is established is useless */
1963 crypto_free_hash(tconn->cram_hmac_tfm);
1964 tconn->cram_hmac_tfm = crypto.cram_hmac_tfm;
1965
91fd4dad 1966 mutex_unlock(&tconn->net_conf_update);
88104ca4 1967 mutex_unlock(&tconn->data.mutex);
91fd4dad
PR		 1968 synchronize_rcu();
1969 kfree(old_conf);
1970
f399002e
LE		 1971 if (tconn->cstate >= C_WF_REPORT_PARAMS)
1972 drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn)));
1973
91fd4dad
PR		 1974 goto done;
1975
f399002e 1976 fail:
91fd4dad 1977 mutex_unlock(&tconn->net_conf_update);
88104ca4 1978 mutex_unlock(&tconn->data.mutex);
0fd0ea06 1979 free_crypto(&crypto);
f399002e 1980 kfree(new_conf);
91fd4dad 1981 done:
f399002e
LE		 1982 conn_reconfig_done(tconn);
1983 out:
1984 drbd_adm_finish(info, retcode);
1985 return 0;
1986}
1987
3b98c0c2 1988int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info)
b411b363 1989{
3b98c0c2 1990 struct drbd_conf *mdev;
44ed167d 1991 struct net_conf *old_conf, *new_conf = NULL;
0fd0ea06 1992 struct crypto crypto = { };
80883197 1993 struct drbd_tconn *oconn;
3b98c0c2 1994 struct drbd_tconn *tconn;
b411b363 1995 struct sockaddr *new_my_addr, *new_peer_addr, *taken_addr;
3b98c0c2
LE		 1996 enum drbd_ret_code retcode;
1997 int i;
1998 int err;
b411b363 1999
3b98c0c2
LE		 2000 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
2001 if (!adm_ctx.reply_skb)
2002 return retcode;
2003 if (retcode != NO_ERROR)
2004 goto out;
2005
2006 tconn = adm_ctx.tconn;
80883197 2007 conn_reconfig_start(tconn);
b411b363 2008
80883197 2009 if (tconn->cstate > C_STANDALONE) {
b411b363
PR		 2010 retcode = ERR_NET_CONFIGURED;
2011 goto fail;
2012 }
2013
2014 /* allocation not in the IO path, cqueue thread context */
5979e361 2015 new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL);
b411b363
PR		 2016 if (!new_conf) {
2017 retcode = ERR_NOMEM;
2018 goto fail;
2019 }
2020
5979e361 2021 drbd_set_net_conf_defaults(new_conf);
f399002e
LE		 2022
2023 err = net_conf_from_attrs(new_conf, info);
3b98c0c2 2024 if (err) {
b411b363 2025 retcode = ERR_MANDATORY_TAG;
3b98c0c2 2026 drbd_msg_put_info(from_attrs_err_to_txt(err));
b411b363
PR		 2027 goto fail;
2028 }
2029
cd64397c
PR		 2030 retcode = check_net_options(tconn, new_conf);
2031 if (retcode != NO_ERROR)
422028b1 2032 goto fail;
422028b1 2033
b411b363
PR		 2034 retcode = NO_ERROR;
2035
2036 new_my_addr = (struct sockaddr *)&new_conf->my_addr;
2037 new_peer_addr = (struct sockaddr *)&new_conf->peer_addr;
543cc10b 2038
ef356262 2039 /* No need to take drbd_cfg_rwsem here. All reconfiguration is
543cc10b
LE		 2040 * strictly serialized on genl_lock(). We are protected against
2041 * concurrent reconfiguration/addition/deletion */
80883197 2042 list_for_each_entry(oconn, &drbd_tconns, all_tconn) {
44ed167d 2043 struct net_conf *nc;
80883197 2044 if (oconn == tconn)
b411b363 2045 continue;
44ed167d
PR		 2046
2047 rcu_read_lock();
2048 nc = rcu_dereference(oconn->net_conf);
2049 if (nc) {
2050 taken_addr = (struct sockaddr *)&nc->my_addr;
2051 if (new_conf->my_addr_len == nc->my_addr_len &&
b411b363
PR		 2052 !memcmp(new_my_addr, taken_addr, new_conf->my_addr_len))
2053 retcode = ERR_LOCAL_ADDR;
2054
44ed167d
PR		 2055 taken_addr = (struct sockaddr *)&nc->peer_addr;
2056 if (new_conf->peer_addr_len == nc->peer_addr_len &&
b411b363
PR		 2057 !memcmp(new_peer_addr, taken_addr, new_conf->peer_addr_len))
2058 retcode = ERR_PEER_ADDR;
b411b363 2059 }
44ed167d
PR		 2060 rcu_read_unlock();
2061 if (retcode != NO_ERROR)
2062 goto fail;
b411b363
PR		 2063 }
2064
0fd0ea06
PR		 2065 retcode = alloc_crypto(&crypto, new_conf);
2066 if (retcode != NO_ERROR)
2067 goto fail;
b411b363 2068
b411b363
PR		 2069 ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0;
2070
80883197 2071 conn_flush_workqueue(tconn);
91fd4dad
PR		 2072
2073 mutex_lock(&tconn->net_conf_update);
2074 old_conf = tconn->net_conf;
2075 if (old_conf) {
b411b363 2076 retcode = ERR_NET_CONFIGURED;
91fd4dad 2077 mutex_unlock(&tconn->net_conf_update);
b411b363
PR		 2078 goto fail;
2079 }
44ed167d 2080 rcu_assign_pointer(tconn->net_conf, new_conf);
b411b363 2081
91fd4dad 2082 conn_free_crypto(tconn);
0fd0ea06
PR		 2083 tconn->int_dig_in = crypto.int_dig_in;
2084 tconn->int_dig_vv = crypto.int_dig_vv;
2085 tconn->cram_hmac_tfm = crypto.cram_hmac_tfm;
8d412fc6 2086 tconn->integrity_tfm = crypto.integrity_tfm;
0fd0ea06
PR		 2087 tconn->csums_tfm = crypto.csums_tfm;
2088 tconn->verify_tfm = crypto.verify_tfm;
b411b363 2089
91fd4dad
PR		 2090 mutex_unlock(&tconn->net_conf_update);
2091
695d08fa 2092 rcu_read_lock();
80883197
PR		 2093 idr_for_each_entry(&tconn->volumes, mdev, i) {
2094 mdev->send_cnt = 0;
2095 mdev->recv_cnt = 0;
80883197 2096 }
695d08fa 2097 rcu_read_unlock();
5ee743e9
LE		 2098
2099 retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE);
2100
80883197 2101 conn_reconfig_done(tconn);
3b98c0c2 2102 drbd_adm_finish(info, retcode);
b411b363
PR		 2103 return 0;
2104
2105fail:
0fd0ea06 2106 free_crypto(&crypto);
b411b363
PR		 2107 kfree(new_conf);
2108
80883197 2109 conn_reconfig_done(tconn);
3b98c0c2
LE		 2110out:
2111 drbd_adm_finish(info, retcode);
b411b363
PR		 2112 return 0;
2113}
2114
85f75dd7
LE		 2115static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force)
2116{
2117 enum drbd_state_rv rv;
2118 if (force) {
2119 spin_lock_irq(&tconn->req_lock);
d9cc6e23 2120 rv = _conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
85f75dd7 2121 spin_unlock_irq(&tconn->req_lock);
d9cc6e23 2122 return rv;
85f75dd7
LE		 2123 }
2124
2125 rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), 0);
2126
2127 switch (rv) {
2128 case SS_NOTHING_TO_DO:
2129 case SS_ALREADY_STANDALONE:
2130 return SS_SUCCESS;
2131 case SS_PRIMARY_NOP:
2132 /* Our state checking code wants to see the peer outdated. */
2133 rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING,
2134 pdsk, D_OUTDATED), CS_VERBOSE);
2135 break;
2136 case SS_CW_FAILED_BY_PEER:
2137 /* The peer probably wants to see us outdated. */
2138 rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING,
2139 disk, D_OUTDATED), 0);
2140 if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) {
2141 conn_request_state(tconn, NS(conn, C_DISCONNECTING), CS_HARD);
2142 rv = SS_SUCCESS;
2143 }
2144 break;
2145 default:;
2146 /* no special handling necessary */
2147 }
2148
2149 return rv;
2150}
2151
3b98c0c2 2152int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info)
b411b363 2153{
3b98c0c2
LE		 2154 struct disconnect_parms parms;
2155 struct drbd_tconn *tconn;
85f75dd7 2156 enum drbd_state_rv rv;
3b98c0c2
LE		 2157 enum drbd_ret_code retcode;
2158 int err;
2561b9c1 2159
3b98c0c2
LE		 2160 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
2161 if (!adm_ctx.reply_skb)
2162 return retcode;
2163 if (retcode != NO_ERROR)
2561b9c1 2164 goto fail;
3b98c0c2
LE		 2165
2166 tconn = adm_ctx.tconn;
2167 memset(&parms, 0, sizeof(parms));
2168 if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) {
f399002e 2169 err = disconnect_parms_from_attrs(&parms, info);
3b98c0c2
LE		 2170 if (err) {
2171 retcode = ERR_MANDATORY_TAG;
2172 drbd_msg_put_info(from_attrs_err_to_txt(err));
2173 goto fail;
2174 }
2561b9c1
PR		 2175 }
2176
85f75dd7
LE		 2177 rv = conn_try_disconnect(tconn, parms.force_disconnect);
2178 if (rv < SS_SUCCESS)
b411b363
PR		 2179 goto fail;
2180
d9cc6e23
LE		 2181 /* No one else can reconfigure the network while I am here.
2182 * The state handling only uses drbd_thread_stop_nowait(),
2183 * we want to really wait here until the receiver is no more. */
2184 drbd_thread_stop(&tconn->receiver);
df24aa45 2185 if (wait_event_interruptible(tconn->ping_wait,
d9cc6e23 2186 tconn->cstate == C_STANDALONE)) {
b411b363
PR		 2187 retcode = ERR_INTR;
2188 goto fail;
2189 }
2190
b411b363
PR		 2191 retcode = NO_ERROR;
2192 fail:
3b98c0c2 2193 drbd_adm_finish(info, retcode);
b411b363
PR		 2194 return 0;
2195}
2196
2197void resync_after_online_grow(struct drbd_conf *mdev)
2198{
2199 int iass; /* I am sync source */
2200
2201 dev_info(DEV, "Resync of new storage after online grow\n");
2202 if (mdev->state.role != mdev->state.peer)
2203 iass = (mdev->state.role == R_PRIMARY);
2204 else
25703f83 2205 iass = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags);
b411b363
PR		 2206
2207 if (iass)
2208 drbd_start_resync(mdev, C_SYNC_SOURCE);
2209 else
2210 _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE);
2211}
2212
3b98c0c2 2213int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
b411b363 2214{
3b98c0c2
LE		 2215 struct resize_parms rs;
2216 struct drbd_conf *mdev;
2217 enum drbd_ret_code retcode;
b411b363 2218 enum determine_dev_size dd;
6495d2c6 2219 enum dds_flags ddsf;
3b98c0c2 2220 int err;
b411b363 2221
3b98c0c2
LE		 2222 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2223 if (!adm_ctx.reply_skb)
2224 return retcode;
2225 if (retcode != NO_ERROR)
b411b363 2226 goto fail;
3b98c0c2
LE		 2227
2228 memset(&rs, 0, sizeof(struct resize_parms));
2229 if (info->attrs[DRBD_NLA_RESIZE_PARMS]) {
f399002e 2230 err = resize_parms_from_attrs(&rs, info);
3b98c0c2
LE		 2231 if (err) {
2232 retcode = ERR_MANDATORY_TAG;
2233 drbd_msg_put_info(from_attrs_err_to_txt(err));
2234 goto fail;
2235 }
b411b363
PR		 2236 }
2237
3b98c0c2 2238 mdev = adm_ctx.mdev;
b411b363
PR		 2239 if (mdev->state.conn > C_CONNECTED) {
2240 retcode = ERR_RESIZE_RESYNC;
2241 goto fail;
2242 }
2243
2244 if (mdev->state.role == R_SECONDARY &&
2245 mdev->state.peer == R_SECONDARY) {
2246 retcode = ERR_NO_PRIMARY;
2247 goto fail;
2248 }
2249
2250 if (!get_ldev(mdev)) {
2251 retcode = ERR_NO_DISK;
2252 goto fail;
2253 }
2254
31890f4a 2255 if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) {
6495d2c6
PR		 2256 retcode = ERR_NEED_APV_93;
2257 goto fail;
2258 }
2259
087c2492 2260 if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev))
b411b363 2261 mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
b411b363
PR		 2262
2263 mdev->ldev->dc.disk_size = (sector_t)rs.resize_size;
6495d2c6 2264 ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0);
24c4830c 2265 dd = drbd_determine_dev_size(mdev, ddsf);
b411b363
PR		 2266 drbd_md_sync(mdev);
2267 put_ldev(mdev);
2268 if (dd == dev_size_error) {
2269 retcode = ERR_NOMEM_BITMAP;
2270 goto fail;
2271 }
2272
087c2492 2273 if (mdev->state.conn == C_CONNECTED) {
b411b363
PR		 2274 if (dd == grew)
2275 set_bit(RESIZE_PENDING, &mdev->flags);
2276
2277 drbd_send_uuids(mdev);
6495d2c6 2278 drbd_send_sizes(mdev, 1, ddsf);
b411b363
PR		 2279 }
2280
2281 fail:
3b98c0c2 2282 drbd_adm_finish(info, retcode);
b411b363
PR		 2283 return 0;
2284}
2285
f399002e 2286int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info)
b411b363 2287{
3b98c0c2 2288 enum drbd_ret_code retcode;
b411b363 2289 cpumask_var_t new_cpu_mask;
f399002e 2290 struct drbd_tconn *tconn;
778f271d 2291 int *rs_plan_s = NULL;
b57a1e27 2292 struct res_opts res_opts;
f399002e 2293 int err;
b411b363 2294
f399002e 2295 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
3b98c0c2
LE		 2296 if (!adm_ctx.reply_skb)
2297 return retcode;
2298 if (retcode != NO_ERROR)
2299 goto fail;
f399002e 2300 tconn = adm_ctx.tconn;
3b98c0c2 2301
b411b363
PR		 2302 if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) {
2303 retcode = ERR_NOMEM;
3b98c0c2 2304 drbd_msg_put_info("unable to allocate cpumask");
b411b363
PR		 2305 goto fail;
2306 }
2307
b57a1e27 2308 res_opts = tconn->res_opts;
5979e361 2309 if (should_set_defaults(info))
b57a1e27 2310 drbd_set_res_opts_default(&res_opts);
b411b363 2311
b57a1e27 2312 err = res_opts_from_attrs(&res_opts, info);
3b98c0c2 2313 if (err) {
b411b363 2314 retcode = ERR_MANDATORY_TAG;
3b98c0c2 2315 drbd_msg_put_info(from_attrs_err_to_txt(err));
b411b363
PR		 2316 goto fail;
2317 }
2318
b411b363 2319 /* silently ignore cpu mask on UP kernel */
b57a1e27
LE		 2320 if (nr_cpu_ids > 1 && res_opts.cpu_mask[0] != 0) {
2321 err = __bitmap_parse(res_opts.cpu_mask, 32, 0,
b411b363
PR		 2322 cpumask_bits(new_cpu_mask), nr_cpu_ids);
2323 if (err) {
f399002e 2324 conn_warn(tconn, "__bitmap_parse() failed with %d\n", err);
b411b363
PR		 2325 retcode = ERR_CPU_MASK_PARSE;
2326 goto fail;
2327 }
2328 }
2329
b411b363 2330
b57a1e27 2331 tconn->res_opts = res_opts;
b411b363 2332
f399002e
LE		 2333 if (!cpumask_equal(tconn->cpu_mask, new_cpu_mask)) {
2334 cpumask_copy(tconn->cpu_mask, new_cpu_mask);
2335 drbd_calc_cpu_mask(tconn);
2336 tconn->receiver.reset_cpu_mask = 1;
2337 tconn->asender.reset_cpu_mask = 1;
2338 tconn->worker.reset_cpu_mask = 1;
b411b363
PR		 2339 }
2340
b411b363 2341fail:
778f271d 2342 kfree(rs_plan_s);
b411b363 2343 free_cpumask_var(new_cpu_mask);
3b98c0c2
LE		 2344
2345 drbd_adm_finish(info, retcode);
b411b363
PR		 2346 return 0;
2347}
2348
3b98c0c2 2349int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info)
b411b363 2350{
3b98c0c2
LE		 2351 struct drbd_conf *mdev;
2352 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2353
2354 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2355 if (!adm_ctx.reply_skb)
2356 return retcode;
2357 if (retcode != NO_ERROR)
2358 goto out;
2359
2360 mdev = adm_ctx.mdev;
b411b363 2361
194bfb32
LE		 2362 /* If there is still bitmap IO pending, probably because of a previous
2363 * resync just being finished, wait for it before requesting a new resync. */
2364 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
2365
b411b363
PR		 2366 retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED);
2367
2368 if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION)
2369 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
2370
2371 while (retcode == SS_NEED_CONNECTION) {
87eeee41 2372 spin_lock_irq(&mdev->tconn->req_lock);
b411b363
PR		 2373 if (mdev->state.conn < C_CONNECTED)
2374 retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL);
87eeee41 2375 spin_unlock_irq(&mdev->tconn->req_lock);
b411b363
PR		 2376
2377 if (retcode != SS_NEED_CONNECTION)
2378 break;
2379
2380 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
2381 }
2382
3b98c0c2
LE		 2383out:
2384 drbd_adm_finish(info, retcode);
b411b363
PR		 2385 return 0;
2386}
2387
0778286a
PR		 2388static int drbd_bmio_set_susp_al(struct drbd_conf *mdev)
2389{
2390 int rv;
2391
2392 rv = drbd_bmio_set_n_write(mdev);
2393 drbd_suspend_al(mdev);
2394 return rv;
2395}
2396
3b98c0c2
LE		 2397static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info,
2398 union drbd_state mask, union drbd_state val)
b411b363 2399{
3b98c0c2 2400 enum drbd_ret_code retcode;
194bfb32 2401
3b98c0c2
LE		 2402 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2403 if (!adm_ctx.reply_skb)
2404 return retcode;
2405 if (retcode != NO_ERROR)
2406 goto out;
b411b363 2407
3b98c0c2
LE		 2408 retcode = drbd_request_state(adm_ctx.mdev, mask, val);
2409out:
2410 drbd_adm_finish(info, retcode);
b411b363
PR		 2411 return 0;
2412}
2413
3b98c0c2 2414int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info)
b411b363 2415{
3b98c0c2
LE		 2416 return drbd_adm_simple_request_state(skb, info, NS(conn, C_STARTING_SYNC_S));
2417}
b411b363 2418
3b98c0c2
LE		 2419int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info)
2420{
2421 enum drbd_ret_code retcode;
2422
2423 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2424 if (!adm_ctx.reply_skb)
2425 return retcode;
2426 if (retcode != NO_ERROR)
2427 goto out;
b411b363 2428
3b98c0c2
LE		 2429 if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO)
2430 retcode = ERR_PAUSE_IS_SET;
2431out:
2432 drbd_adm_finish(info, retcode);
b411b363
PR		 2433 return 0;
2434}
2435
3b98c0c2 2436int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info)
b411b363 2437{
da9fbc27 2438 union drbd_dev_state s;
3b98c0c2
LE		 2439 enum drbd_ret_code retcode;
2440
2441 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2442 if (!adm_ctx.reply_skb)
2443 return retcode;
2444 if (retcode != NO_ERROR)
2445 goto out;
b411b363 2446
3b98c0c2
LE		 2447 if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) {
2448 s = adm_ctx.mdev->state;
cd88d030
PR		 2449 if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) {
2450 retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP :
2451 s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR;
2452 } else {
2453 retcode = ERR_PAUSE_IS_CLEAR;
2454 }
2455 }
b411b363 2456
3b98c0c2
LE		 2457out:
2458 drbd_adm_finish(info, retcode);
b411b363
PR		 2459 return 0;
2460}
2461
3b98c0c2 2462int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info)
b411b363 2463{
3b98c0c2 2464 return drbd_adm_simple_request_state(skb, info, NS(susp, 1));
b411b363
PR		 2465}
2466
3b98c0c2 2467int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info)
b411b363 2468{
3b98c0c2
LE		 2469 struct drbd_conf *mdev;
2470 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2471
2472 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2473 if (!adm_ctx.reply_skb)
2474 return retcode;
2475 if (retcode != NO_ERROR)
2476 goto out;
2477
2478 mdev = adm_ctx.mdev;
43a5182c
PR		 2479 if (test_bit(NEW_CUR_UUID, &mdev->flags)) {
2480 drbd_uuid_new_current(mdev);
2481 clear_bit(NEW_CUR_UUID, &mdev->flags);
43a5182c 2482 }
265be2d0 2483 drbd_suspend_io(mdev);
3b98c0c2
LE		 2484 retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0));
2485 if (retcode == SS_SUCCESS) {
265be2d0 2486 if (mdev->state.conn < C_CONNECTED)
2f5cdd0b 2487 tl_clear(mdev->tconn);
265be2d0 2488 if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED)
2f5cdd0b 2489 tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO);
265be2d0
PR		 2490 }
2491 drbd_resume_io(mdev);
2492
3b98c0c2
LE		 2493out:
2494 drbd_adm_finish(info, retcode);
b411b363
PR		 2495 return 0;
2496}
2497
3b98c0c2 2498int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info)
b411b363 2499{
3b98c0c2 2500 return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED));
b411b363
PR		 2501}
2502
543cc10b
LE		 2503int nla_put_drbd_cfg_context(struct sk_buff *skb, const char *conn_name, unsigned vnr)
2504{
2505 struct nlattr *nla;
2506 nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT);
2507 if (!nla)
2508 goto nla_put_failure;
2509 if (vnr != VOLUME_UNSPECIFIED)
2510 NLA_PUT_U32(skb, T_ctx_volume, vnr);
2511 NLA_PUT_STRING(skb, T_ctx_conn_name, conn_name);
2512 nla_nest_end(skb, nla);
2513 return 0;
2514
2515nla_put_failure:
2516 if (nla)
2517 nla_nest_cancel(skb, nla);
2518 return -EMSGSIZE;
2519}
2520
3b98c0c2
LE		 2521int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev,
2522 const struct sib_info *sib)
b411b363 2523{
3b98c0c2 2524 struct state_info *si = NULL; /* for sizeof(si->member); */
44ed167d 2525 struct net_conf *nc;
3b98c0c2
LE		 2526 struct nlattr *nla;
2527 int got_ldev;
3b98c0c2
LE		 2528 int err = 0;
2529 int exclude_sensitive;
2530
2531 /* If sib != NULL, this is drbd_bcast_event, which anyone can listen
2532 * to. So we better exclude_sensitive information.
2533 *
2534 * If sib == NULL, this is drbd_adm_get_status, executed synchronously
2535 * in the context of the requesting user process. Exclude sensitive
2536 * information, unless current has superuser.
2537 *
2538 * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and
2539 * relies on the current implementation of netlink_dump(), which
2540 * executes the dump callback successively from netlink_recvmsg(),
2541 * always in the context of the receiving process */
2542 exclude_sensitive = sib || !capable(CAP_SYS_ADMIN);
2543
2544 got_ldev = get_ldev(mdev);
3b98c0c2
LE		 2545
2546 /* We need to add connection name and volume number information still.
2547 * Minor number is in drbd_genlmsghdr. */
543cc10b 2548 if (nla_put_drbd_cfg_context(skb, mdev->tconn->name, mdev->vnr))
3b98c0c2 2549 goto nla_put_failure;
3b98c0c2 2550
f399002e
LE		 2551 if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive))
2552 goto nla_put_failure;
2553
3b98c0c2
LE		 2554 if (got_ldev)
2555 if (disk_conf_to_skb(skb, &mdev->ldev->dc, exclude_sensitive))
2556 goto nla_put_failure;
44ed167d
PR		 2557
2558 rcu_read_lock();
2559 nc = rcu_dereference(mdev->tconn->net_conf);
2560 if (nc)
2561 err = net_conf_to_skb(skb, nc, exclude_sensitive);
2562 rcu_read_unlock();
2563 if (err)
2564 goto nla_put_failure;
3b98c0c2 2565
3b98c0c2
LE		 2566 nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO);
2567 if (!nla)
2568 goto nla_put_failure;
2569 NLA_PUT_U32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY);
2570 NLA_PUT_U32(skb, T_current_state, mdev->state.i);
2571 NLA_PUT_U64(skb, T_ed_uuid, mdev->ed_uuid);
2572 NLA_PUT_U64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev));
2573
2574 if (got_ldev) {
2575 NLA_PUT_U32(skb, T_disk_flags, mdev->ldev->md.flags);
2576 NLA_PUT(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid);
2577 NLA_PUT_U64(skb, T_bits_total, drbd_bm_bits(mdev));
2578 NLA_PUT_U64(skb, T_bits_oos, drbd_bm_total_weight(mdev));
2579 if (C_SYNC_SOURCE <= mdev->state.conn &&
2580 C_PAUSED_SYNC_T >= mdev->state.conn) {
2581 NLA_PUT_U64(skb, T_bits_rs_total, mdev->rs_total);
2582 NLA_PUT_U64(skb, T_bits_rs_failed, mdev->rs_failed);
2583 }
b411b363
PR		 2584 }
2585
3b98c0c2
LE		 2586 if (sib) {
2587 switch(sib->sib_reason) {
2588 case SIB_SYNC_PROGRESS:
2589 case SIB_GET_STATUS_REPLY:
2590 break;
2591 case SIB_STATE_CHANGE:
2592 NLA_PUT_U32(skb, T_prev_state, sib->os.i);
2593 NLA_PUT_U32(skb, T_new_state, sib->ns.i);
2594 break;
2595 case SIB_HELPER_POST:
2596 NLA_PUT_U32(skb,
2597 T_helper_exit_code, sib->helper_exit_code);
2598 /* fall through */
2599 case SIB_HELPER_PRE:
2600 NLA_PUT_STRING(skb, T_helper, sib->helper_name);
2601 break;
2602 }
b411b363 2603 }
3b98c0c2 2604 nla_nest_end(skb, nla);
b411b363 2605
3b98c0c2
LE		 2606 if (0)
2607nla_put_failure:
2608 err = -EMSGSIZE;
2609 if (got_ldev)
2610 put_ldev(mdev);
3b98c0c2 2611 return err;
b411b363
PR		 2612}
2613
3b98c0c2 2614int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info)
b411b363 2615{
3b98c0c2
LE		 2616 enum drbd_ret_code retcode;
2617 int err;
b411b363 2618
3b98c0c2
LE		 2619 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2620 if (!adm_ctx.reply_skb)
2621 return retcode;
2622 if (retcode != NO_ERROR)
2623 goto out;
b411b363 2624
3b98c0c2
LE		 2625 err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL);
2626 if (err) {
2627 nlmsg_free(adm_ctx.reply_skb);
2628 return err;
b411b363 2629 }
3b98c0c2
LE		 2630out:
2631 drbd_adm_finish(info, retcode);
2632 return 0;
b411b363
PR		 2633}
2634
71932efc 2635int get_one_status(struct sk_buff *skb, struct netlink_callback *cb)
b411b363 2636{
3b98c0c2
LE		 2637 struct drbd_conf *mdev;
2638 struct drbd_genlmsghdr *dh;
543cc10b
LE		 2639 struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0];
2640 struct drbd_tconn *tconn = NULL;
2641 struct drbd_tconn *tmp;
2642 unsigned volume = cb->args[1];
2643
2644 /* Open coded, deferred, iteration:
2645 * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) {
2646 * idr_for_each_entry(&tconn->volumes, mdev, i) {
2647 * ...
2648 * }
2649 * }
2650 * where tconn is cb->args[0];
2651 * and i is cb->args[1];
2652 *
71932efc
LE		 2653 * cb->args[2] indicates if we shall loop over all resources,
2654 * or just dump all volumes of a single resource.
2655 *
3b98c0c2
LE		 2656 * This may miss entries inserted after this dump started,
2657 * or entries deleted before they are reached.
543cc10b
LE		 2658 *
2659 * We need to make sure the mdev won't disappear while
2660 * we are looking at it, and revalidate our iterators
2661 * on each iteration.
2662 */
3b98c0c2 2663
9dc9fbb3 2664 /* synchronize with conn_create()/conn_destroy() */
ef356262 2665 down_read(&drbd_cfg_rwsem);
543cc10b
LE		 2666 /* revalidate iterator position */
2667 list_for_each_entry(tmp, &drbd_tconns, all_tconn) {
2668 if (pos == NULL) {
2669 /* first iteration */
2670 pos = tmp;
2671 tconn = pos;
2672 break;
2673 }
2674 if (tmp == pos) {
2675 tconn = pos;
2676 break;
2677 }
2678 }
2679 if (tconn) {
71932efc 2680next_tconn:
543cc10b
LE		 2681 mdev = idr_get_next(&tconn->volumes, &volume);
2682 if (!mdev) {
2683 /* No more volumes to dump on this tconn.
2684 * Advance tconn iterator. */
2685 pos = list_entry(tconn->all_tconn.next,
2686 struct drbd_tconn, all_tconn);
71932efc 2687 /* Did we dump any volume on this tconn yet? */
543cc10b 2688 if (volume != 0) {
71932efc
LE		 2689 /* If we reached the end of the list,
2690 * or only a single resource dump was requested,
2691 * we are done. */
2692 if (&pos->all_tconn == &drbd_tconns || cb->args[2])
2693 goto out;
543cc10b 2694 volume = 0;
71932efc 2695 tconn = pos;
543cc10b
LE		 2696 goto next_tconn;
2697 }
2698 }
2699
3b98c0c2
LE		 2700 dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid,
2701 cb->nlh->nlmsg_seq, &drbd_genl_family,
2702 NLM_F_MULTI, DRBD_ADM_GET_STATUS);
2703 if (!dh)
543cc10b
LE		 2704 goto out;
2705
2706 if (!mdev) {
2707 /* this is a tconn without a single volume */
2708 dh->minor = -1U;
2709 dh->ret_code = NO_ERROR;
2710 if (nla_put_drbd_cfg_context(skb, tconn->name, VOLUME_UNSPECIFIED))
2711 genlmsg_cancel(skb, dh);
2712 else
2713 genlmsg_end(skb, dh);
2714 goto out;
2715 }
3b98c0c2 2716
543cc10b
LE		 2717 D_ASSERT(mdev->vnr == volume);
2718 D_ASSERT(mdev->tconn == tconn);
3b98c0c2 2719
543cc10b 2720 dh->minor = mdev_to_minor(mdev);
3b98c0c2
LE		 2721 dh->ret_code = NO_ERROR;
2722
2723 if (nla_put_status_info(skb, mdev, NULL)) {
2724 genlmsg_cancel(skb, dh);
543cc10b 2725 goto out;
3b98c0c2
LE		 2726 }
2727 genlmsg_end(skb, dh);
2728 }
b411b363 2729
543cc10b 2730out:
ef356262 2731 up_read(&drbd_cfg_rwsem);
543cc10b
LE		 2732 /* where to start the next iteration */
2733 cb->args[0] = (long)pos;
2734 cb->args[1] = (pos == tconn) ? volume + 1 : 0;
b411b363 2735
543cc10b
LE		 2736 /* No more tconns/volumes/minors found results in an empty skb.
2737 * Which will terminate the dump. */
3b98c0c2 2738 return skb->len;
b411b363
PR		 2739}
2740
71932efc
LE		 2741/*
2742 * Request status of all resources, or of all volumes within a single resource.
2743 *
2744 * This is a dump, as the answer may not fit in a single reply skb otherwise.
2745 * Which means we cannot use the family->attrbuf or other such members, because
2746 * dump is NOT protected by the genl_lock(). During dump, we only have access
2747 * to the incoming skb, and need to opencode "parsing" of the nlattr payload.
2748 *
2749 * Once things are setup properly, we call into get_one_status().
2750 */
2751int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb)
2752{
2753 const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ;
2754 struct nlattr *nla;
2755 const char *conn_name;
2756 struct drbd_tconn *tconn;
2757
2758 /* Is this a followup call? */
2759 if (cb->args[0]) {
2760 /* ... of a single resource dump,
2761 * and the resource iterator has been advanced already? */
2762 if (cb->args[2] && cb->args[2] != cb->args[0])
2763 return 0; /* DONE. */
2764 goto dump;
2765 }
2766
2767 /* First call (from netlink_dump_start). We need to figure out
2768 * which resource(s) the user wants us to dump. */
2769 nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen),
2770 nlmsg_attrlen(cb->nlh, hdrlen),
2771 DRBD_NLA_CFG_CONTEXT);
2772
2773 /* No explicit context given. Dump all. */
2774 if (!nla)
2775 goto dump;
2776 nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name));
2777 /* context given, but no name present? */
2778 if (!nla)
2779 return -EINVAL;
2780 conn_name = nla_data(nla);
0ace9dfa
PR		 2781 tconn = conn_get_by_name(conn_name);
2782
71932efc
LE		 2783 if (!tconn)
2784 return -ENODEV;
2785
0ace9dfa
PR		 2786 kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */
2787
71932efc
LE		 2788 /* prime iterators, and set "filter" mode mark:
2789 * only dump this tconn. */
2790 cb->args[0] = (long)tconn;
2791 /* cb->args[1] = 0; passed in this way. */
2792 cb->args[2] = (long)tconn;
2793
2794dump:
2795 return get_one_status(skb, cb);
2796}
2797
3b98c0c2 2798int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info)
b411b363 2799{
3b98c0c2
LE		 2800 enum drbd_ret_code retcode;
2801 struct timeout_parms tp;
2802 int err;
b411b363 2803
3b98c0c2
LE		 2804 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2805 if (!adm_ctx.reply_skb)
2806 return retcode;
2807 if (retcode != NO_ERROR)
2808 goto out;
b411b363 2809
3b98c0c2
LE		 2810 tp.timeout_type =
2811 adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED :
2812 test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED :
2813 UT_DEFAULT;
b411b363 2814
3b98c0c2
LE		 2815 err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp);
2816 if (err) {
2817 nlmsg_free(adm_ctx.reply_skb);
2818 return err;
2819 }
2820out:
2821 drbd_adm_finish(info, retcode);
2822 return 0;
b411b363
PR		 2823}
2824
3b98c0c2 2825int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info)
b411b363 2826{
3b98c0c2
LE		 2827 struct drbd_conf *mdev;
2828 enum drbd_ret_code retcode;
b411b363 2829
3b98c0c2
LE		 2830 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2831 if (!adm_ctx.reply_skb)
2832 return retcode;
2833 if (retcode != NO_ERROR)
2834 goto out;
873b0d5f 2835
3b98c0c2
LE		 2836 mdev = adm_ctx.mdev;
2837 if (info->attrs[DRBD_NLA_START_OV_PARMS]) {
2838 /* resume from last known position, if possible */
2839 struct start_ov_parms parms =
2840 { .ov_start_sector = mdev->ov_start_sector };
f399002e 2841 int err = start_ov_parms_from_attrs(&parms, info);
3b98c0c2
LE		 2842 if (err) {
2843 retcode = ERR_MANDATORY_TAG;
2844 drbd_msg_put_info(from_attrs_err_to_txt(err));
2845 goto out;
2846 }
2847 /* w_make_ov_request expects position to be aligned */
2848 mdev->ov_start_sector = parms.ov_start_sector & ~BM_SECT_PER_BIT;
2849 }
873b0d5f
LE		 2850 /* If there is still bitmap IO pending, e.g. previous resync or verify
2851 * just being finished, wait for it before requesting a new resync. */
2852 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
3b98c0c2
LE		 2853 retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
2854out:
2855 drbd_adm_finish(info, retcode);
b411b363
PR		 2856 return 0;
2857}
2858
2859
3b98c0c2 2860int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info)
b411b363 2861{
3b98c0c2
LE		 2862 struct drbd_conf *mdev;
2863 enum drbd_ret_code retcode;
b411b363
PR		 2864 int skip_initial_sync = 0;
2865 int err;
3b98c0c2 2866 struct new_c_uuid_parms args;
b411b363 2867
3b98c0c2
LE		 2868 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2869 if (!adm_ctx.reply_skb)
2870 return retcode;
2871 if (retcode != NO_ERROR)
2872 goto out_nolock;
b411b363 2873
3b98c0c2
LE		 2874 mdev = adm_ctx.mdev;
2875 memset(&args, 0, sizeof(args));
2876 if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) {
f399002e 2877 err = new_c_uuid_parms_from_attrs(&args, info);
3b98c0c2
LE		 2878 if (err) {
2879 retcode = ERR_MANDATORY_TAG;
2880 drbd_msg_put_info(from_attrs_err_to_txt(err));
2881 goto out_nolock;
2882 }
b411b363
PR		 2883 }
2884
8410da8f 2885 mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */
b411b363
PR		 2886
2887 if (!get_ldev(mdev)) {
2888 retcode = ERR_NO_DISK;
2889 goto out;
2890 }
2891
2892 /* this is "skip initial sync", assume to be clean */
31890f4a 2893 if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 &&
b411b363
PR		 2894 mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) {
2895 dev_info(DEV, "Preparing to skip initial sync\n");
2896 skip_initial_sync = 1;
2897 } else if (mdev->state.conn != C_STANDALONE) {
2898 retcode = ERR_CONNECTED;
2899 goto out_dec;
2900 }
2901
2902 drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */
2903 drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */
2904
2905 if (args.clear_bm) {
20ceb2b2
LE		 2906 err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write,
2907 "clear_n_write from new_c_uuid", BM_LOCKED_MASK);
b411b363
PR		 2908 if (err) {
2909 dev_err(DEV, "Writing bitmap failed with %d\n",err);
2910 retcode = ERR_IO_MD_DISK;
2911 }
2912 if (skip_initial_sync) {
2913 drbd_send_uuids_skip_initial_sync(mdev);
2914 _drbd_uuid_set(mdev, UI_BITMAP, 0);
62b0da3a 2915 drbd_print_uuids(mdev, "cleared bitmap UUID");
87eeee41 2916 spin_lock_irq(&mdev->tconn->req_lock);
b411b363
PR		 2917 _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE),
2918 CS_VERBOSE, NULL);
87eeee41 2919 spin_unlock_irq(&mdev->tconn->req_lock);
b411b363
PR		 2920 }
2921 }
2922
2923 drbd_md_sync(mdev);
2924out_dec:
2925 put_ldev(mdev);
2926out:
8410da8f 2927 mutex_unlock(mdev->state_mutex);
3b98c0c2
LE		 2928out_nolock:
2929 drbd_adm_finish(info, retcode);
774b3055
PR		 2930 return 0;
2931}
2932
3b98c0c2
LE		 2933static enum drbd_ret_code
2934drbd_check_conn_name(const char *name)
774b3055 2935{
3b98c0c2
LE		 2936 if (!name || !name[0]) {
2937 drbd_msg_put_info("connection name missing");
2938 return ERR_MANDATORY_TAG;
774b3055 2939 }
3b98c0c2
LE		 2940 /* if we want to use these in sysfs/configfs/debugfs some day,
2941 * we must not allow slashes */
2942 if (strchr(name, '/')) {
2943 drbd_msg_put_info("invalid connection name");
2944 return ERR_INVALID_REQUEST;
774b3055 2945 }
3b98c0c2 2946 return NO_ERROR;
774b3055
PR		 2947}
2948
3b98c0c2 2949int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info)
b411b363 2950{
3b98c0c2 2951 enum drbd_ret_code retcode;
9f5180e5 2952
3b98c0c2
LE		 2953 retcode = drbd_adm_prepare(skb, info, 0);
2954 if (!adm_ctx.reply_skb)
2955 return retcode;
2956 if (retcode != NO_ERROR)
2957 goto out;
b411b363 2958
3b98c0c2
LE		 2959 retcode = drbd_check_conn_name(adm_ctx.conn_name);
2960 if (retcode != NO_ERROR)
2961 goto out;
b411b363 2962
3b98c0c2 2963 if (adm_ctx.tconn) {
38f19616
LE		 2964 if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) {
2965 retcode = ERR_INVALID_REQUEST;
2966 drbd_msg_put_info("connection exists");
2967 }
2968 /* else: still NO_ERROR */
3b98c0c2 2969 goto out;
b411b363
PR		 2970 }
2971
9dc9fbb3 2972 if (!conn_create(adm_ctx.conn_name))
b411b363 2973 retcode = ERR_NOMEM;
3b98c0c2
LE		 2974out:
2975 drbd_adm_finish(info, retcode);
2976 return 0;
b411b363
PR		 2977}
2978
3b98c0c2 2979int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info)
b411b363 2980{
3b98c0c2
LE		 2981 struct drbd_genlmsghdr *dh = info->userhdr;
2982 enum drbd_ret_code retcode;
b411b363 2983
3b98c0c2
LE		 2984 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
2985 if (!adm_ctx.reply_skb)
2986 return retcode;
2987 if (retcode != NO_ERROR)
2988 goto out;
b411b363 2989
3b98c0c2
LE		 2990 /* FIXME drop minor_count parameter, limit to MINORMASK */
2991 if (dh->minor >= minor_count) {
2992 drbd_msg_put_info("requested minor out of range");
2993 retcode = ERR_INVALID_REQUEST;
2994 goto out;
b411b363 2995 }
0c8e36d9 2996 if (adm_ctx.volume > DRBD_VOLUME_MAX) {
3b98c0c2
LE		 2997 drbd_msg_put_info("requested volume id out of range");
2998 retcode = ERR_INVALID_REQUEST;
2999 goto out;
b411b363 3000 }
b411b363 3001
38f19616
LE		 3002 /* drbd_adm_prepare made sure already
3003 * that mdev->tconn and mdev->vnr match the request. */
3004 if (adm_ctx.mdev) {
3005 if (info->nlhdr->nlmsg_flags & NLM_F_EXCL)
3006 retcode = ERR_MINOR_EXISTS;
3007 /* else: still NO_ERROR */
3008 goto out;
3009 }
3010
d3fcb490 3011 down_write(&drbd_cfg_rwsem);
3b98c0c2 3012 retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume);
d3fcb490 3013 up_write(&drbd_cfg_rwsem);
3b98c0c2
LE		 3014out:
3015 drbd_adm_finish(info, retcode);
3016 return 0;
b411b363
PR		 3017}
3018
85f75dd7
LE		 3019static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev)
3020{
3021 if (mdev->state.disk == D_DISKLESS &&
3022 /* no need to be mdev->state.conn == C_STANDALONE &&
3023 * we may want to delete a minor from a live replication group.
3024 */
3025 mdev->state.role == R_SECONDARY) {
ff370e5a 3026 drbd_delete_device(mdev);
85f75dd7
LE		 3027 return NO_ERROR;
3028 } else
3029 return ERR_MINOR_CONFIGURED;
3030}
3031
3b98c0c2 3032int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info)
b411b363 3033{
3b98c0c2 3034 enum drbd_ret_code retcode;
b411b363 3035
3b98c0c2
LE		 3036 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
3037 if (!adm_ctx.reply_skb)
3038 return retcode;
3039 if (retcode != NO_ERROR)
3040 goto out;
b411b363 3041
ef356262 3042 down_write(&drbd_cfg_rwsem);
85f75dd7 3043 retcode = adm_delete_minor(adm_ctx.mdev);
ef356262 3044 up_write(&drbd_cfg_rwsem);
85f75dd7
LE		 3045out:
3046 drbd_adm_finish(info, retcode);
3047 return 0;
3048}
3049
3050int drbd_adm_down(struct sk_buff *skb, struct genl_info *info)
3051{
3052 enum drbd_ret_code retcode;
3053 enum drbd_state_rv rv;
3054 struct drbd_conf *mdev;
3055 unsigned i;
3056
3057 retcode = drbd_adm_prepare(skb, info, 0);
3058 if (!adm_ctx.reply_skb)
3059 return retcode;
3060 if (retcode != NO_ERROR)
3061 goto out;
3062
3063 if (!adm_ctx.tconn) {
3064 retcode = ERR_CONN_NOT_KNOWN;
3065 goto out;
3066 }
3067
ef356262 3068 down_read(&drbd_cfg_rwsem);
85f75dd7
LE		 3069 /* demote */
3070 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3071 retcode = drbd_set_role(mdev, R_SECONDARY, 0);
3072 if (retcode < SS_SUCCESS) {
3073 drbd_msg_put_info("failed to demote");
3074 goto out_unlock;
3075 }
3076 }
3077
3078 /* disconnect */
3079 rv = conn_try_disconnect(adm_ctx.tconn, 0);
3080 if (rv < SS_SUCCESS) {
3081 retcode = rv; /* enum type mismatch! */
3082 drbd_msg_put_info("failed to disconnect");
3083 goto out_unlock;
3084 }
3085
d9cc6e23
LE		 3086 /* Make sure the network threads have actually stopped,
3087 * state handling only does drbd_thread_stop_nowait(). */
3088 drbd_thread_stop(&adm_ctx.tconn->receiver);
3089
85f75dd7
LE		 3090 /* detach */
3091 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3092 rv = adm_detach(mdev);
3093 if (rv < SS_SUCCESS) {
3094 retcode = rv; /* enum type mismatch! */
3095 drbd_msg_put_info("failed to detach");
3096 goto out_unlock;
3097 }
3098 }
ef356262 3099 up_read(&drbd_cfg_rwsem);
85f75dd7
LE		 3100
3101 /* delete volumes */
ef356262 3102 down_write(&drbd_cfg_rwsem);
85f75dd7
LE		 3103 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3104 retcode = adm_delete_minor(mdev);
3105 if (retcode != NO_ERROR) {
3106 /* "can not happen" */
3107 drbd_msg_put_info("failed to delete volume");
ef356262
PR		 3108 up_write(&drbd_cfg_rwsem);
3109 goto out;
85f75dd7
LE		 3110 }
3111 }
3112
85f75dd7
LE		 3113 /* delete connection */
3114 if (conn_lowest_minor(adm_ctx.tconn) < 0) {
d9cc6e23 3115 drbd_thread_stop(&adm_ctx.tconn->worker);
9dc9fbb3
PR		 3116 list_del(&adm_ctx.tconn->all_tconn);
3117 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
3118
85f75dd7
LE		 3119 retcode = NO_ERROR;
3120 } else {
3121 /* "can not happen" */
3122 retcode = ERR_CONN_IN_USE;
3123 drbd_msg_put_info("failed to delete connection");
85f75dd7 3124 }
ef356262
PR		 3125
3126 up_write(&drbd_cfg_rwsem);
3127 goto out;
85f75dd7 3128out_unlock:
ef356262 3129 up_read(&drbd_cfg_rwsem);
3b98c0c2
LE		 3130out:
3131 drbd_adm_finish(info, retcode);
3132 return 0;
b411b363
PR		 3133}
3134
3b98c0c2 3135int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info)
b411b363 3136{
3b98c0c2 3137 enum drbd_ret_code retcode;
b411b363 3138
3b98c0c2
LE		 3139 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
3140 if (!adm_ctx.reply_skb)
3141 return retcode;
3142 if (retcode != NO_ERROR)
3143 goto out;
3144
ef356262 3145 down_write(&drbd_cfg_rwsem);
3b98c0c2 3146 if (conn_lowest_minor(adm_ctx.tconn) < 0) {
9dc9fbb3
PR		 3147 list_del(&adm_ctx.tconn->all_tconn);
3148 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
3149
3b98c0c2
LE		 3150 retcode = NO_ERROR;
3151 } else {
3152 retcode = ERR_CONN_IN_USE;
b411b363 3153 }
ef356262 3154 up_write(&drbd_cfg_rwsem);
b411b363 3155
3b98c0c2
LE		 3156out:
3157 drbd_adm_finish(info, retcode);
b411b363
PR		 3158 return 0;
3159}
3160
3b98c0c2 3161void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib)
b411b363 3162{
3b98c0c2
LE		 3163 static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */
3164 struct sk_buff *msg;
3165 struct drbd_genlmsghdr *d_out;
3166 unsigned seq;
3167 int err = -ENOMEM;
3168
3169 seq = atomic_inc_return(&drbd_genl_seq);
3170 msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO);
3171 if (!msg)
3172 goto failed;
3173
3174 err = -EMSGSIZE;
3175 d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT);
3176 if (!d_out) /* cannot happen, but anyways. */
3177 goto nla_put_failure;
3178 d_out->minor = mdev_to_minor(mdev);
3179 d_out->ret_code = 0;
3180
3181 if (nla_put_status_info(msg, mdev, sib))
3182 goto nla_put_failure;
3183 genlmsg_end(msg, d_out);
3184 err = drbd_genl_multicast_events(msg, 0);
3185 /* msg has been consumed or freed in netlink_broadcast() */
3186 if (err && err != -ESRCH)
3187 goto failed;
b411b363 3188
3b98c0c2 3189 return;
b411b363 3190
3b98c0c2
LE		 3191nla_put_failure:
3192 nlmsg_free(msg);
3193failed:
3194 dev_err(DEV, "Error %d while broadcasting event. "
3195 "Event seq:%u sib_reason:%u\n",
3196 err, seq, sib->sib_reason);
b411b363 3197}
Linux kernel - Deliverables
This page took 0.287754 seconds and 5 git commands to generate.