projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drbd: disambiguation, s/DISCARD_CONCURRENT/RESOLVE_CONFLICTS/
[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
789c1b62
AG		 50int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info);
51int drbd_adm_del_resource(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>
01b39b50 78#include "drbd_nla.h"
3b98c0c2
LE		 79#include <linux/genl_magic_func.h>
80
81/* used blkdev_get_by_path, to claim our meta data device(s) */
b411b363
PR		 82static char *drbd_m_holder = "Hands off! this is DRBD's meta data device.";
83
3b98c0c2
LE		 84/* Configuration is strictly serialized, because generic netlink message
85 * processing is strictly serialized by the genl_lock().
86 * Which means we can use one static global drbd_config_context struct.
87 */
88static struct drbd_config_context {
89 /* assigned from drbd_genlmsghdr */
90 unsigned int minor;
91 /* assigned from request attributes, if present */
92 unsigned int volume;
93#define VOLUME_UNSPECIFIED (-1U)
94 /* pointer into the request skb,
95 * limited lifetime! */
7c3063cc 96 char *resource_name;
089c075d
AG		 97 struct nlattr *my_addr;
98 struct nlattr *peer_addr;
3b98c0c2
LE		 99
100 /* reply buffer */
101 struct sk_buff *reply_skb;
102 /* pointer into reply buffer */
103 struct drbd_genlmsghdr *reply_dh;
104 /* resolved from attributes, if possible */
105 struct drbd_conf *mdev;
106 struct drbd_tconn *tconn;
107} adm_ctx;
108
109static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info)
110{
111 genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb))));
112 if (genlmsg_reply(skb, info))
113 printk(KERN_ERR "drbd: error sending genl reply\n");
b411b363 114}
3b98c0c2
LE		 115
116/* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only
117 * reason it could fail was no space in skb, and there are 4k available. */
8432b314 118int drbd_msg_put_info(const char *info)
3b98c0c2
LE		 119{
120 struct sk_buff *skb = adm_ctx.reply_skb;
121 struct nlattr *nla;
122 int err = -EMSGSIZE;
123
124 if (!info || !info[0])
125 return 0;
126
127 nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY);
128 if (!nla)
129 return err;
130
131 err = nla_put_string(skb, T_info_text, info);
132 if (err) {
133 nla_nest_cancel(skb, nla);
134 return err;
135 } else
136 nla_nest_end(skb, nla);
137 return 0;
b411b363
PR		 138}
139
3b98c0c2
LE		 140/* This would be a good candidate for a "pre_doit" hook,
141 * and per-family private info->pointers.
142 * But we need to stay compatible with older kernels.
143 * If it returns successfully, adm_ctx members are valid.
144 */
145#define DRBD_ADM_NEED_MINOR 1
44e52cfa 146#define DRBD_ADM_NEED_RESOURCE 2
089c075d 147#define DRBD_ADM_NEED_CONNECTION 4
3b98c0c2
LE		 148static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info,
149 unsigned flags)
150{
151 struct drbd_genlmsghdr *d_in = info->userhdr;
152 const u8 cmd = info->genlhdr->cmd;
153 int err;
154
155 memset(&adm_ctx, 0, sizeof(adm_ctx));
156
157 /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */
158 if (cmd != DRBD_ADM_GET_STATUS
159 && security_netlink_recv(skb, CAP_SYS_ADMIN))
160 return -EPERM;
161
162 adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
1e2a2551
AG		 163 if (!adm_ctx.reply_skb) {
164 err = -ENOMEM;
3b98c0c2 165 goto fail;
1e2a2551 166 }
3b98c0c2
LE		 167
168 adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb,
169 info, &drbd_genl_family, 0, cmd);
170 /* put of a few bytes into a fresh skb of >= 4k will always succeed.
171 * but anyways */
1e2a2551
AG		 172 if (!adm_ctx.reply_dh) {
173 err = -ENOMEM;
3b98c0c2 174 goto fail;
1e2a2551 175 }
3b98c0c2
LE		 176
177 adm_ctx.reply_dh->minor = d_in->minor;
178 adm_ctx.reply_dh->ret_code = NO_ERROR;
179
089c075d 180 adm_ctx.volume = VOLUME_UNSPECIFIED;
3b98c0c2
LE		 181 if (info->attrs[DRBD_NLA_CFG_CONTEXT]) {
182 struct nlattr *nla;
183 /* parse and validate only */
f399002e 184 err = drbd_cfg_context_from_attrs(NULL, info);
3b98c0c2
LE		 185 if (err)
186 goto fail;
187
188 /* It was present, and valid,
189 * copy it over to the reply skb. */
190 err = nla_put_nohdr(adm_ctx.reply_skb,
191 info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len,
192 info->attrs[DRBD_NLA_CFG_CONTEXT]);
193 if (err)
194 goto fail;
195
196 /* and assign stuff to the global adm_ctx */
197 nla = nested_attr_tb[__nla_type(T_ctx_volume)];
089c075d
AG		 198 if (nla)
199 adm_ctx.volume = nla_get_u32(nla);
7c3063cc 200 nla = nested_attr_tb[__nla_type(T_ctx_resource_name)];
3b98c0c2 201 if (nla)
7c3063cc 202 adm_ctx.resource_name = nla_data(nla);
089c075d
AG		 203 adm_ctx.my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)];
204 adm_ctx.peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)];
205 if ((adm_ctx.my_addr &&
206 nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.tconn->my_addr)) ||
207 (adm_ctx.peer_addr &&
208 nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.tconn->peer_addr))) {
209 err = -EINVAL;
210 goto fail;
211 }
212 }
3b98c0c2
LE		 213
214 adm_ctx.minor = d_in->minor;
215 adm_ctx.mdev = minor_to_mdev(d_in->minor);
7c3063cc 216 adm_ctx.tconn = conn_get_by_name(adm_ctx.resource_name);
3b98c0c2
LE		 217
218 if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) {
219 drbd_msg_put_info("unknown minor");
220 return ERR_MINOR_INVALID;
221 }
44e52cfa
AG		 222 if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_RESOURCE)) {
223 drbd_msg_put_info("unknown resource");
3b98c0c2
LE		 224 return ERR_INVALID_REQUEST;
225 }
226
089c075d
AG		 227 if (flags & DRBD_ADM_NEED_CONNECTION) {
228 if (adm_ctx.tconn && !(flags & DRBD_ADM_NEED_RESOURCE)) {
229 drbd_msg_put_info("no resource name expected");
230 return ERR_INVALID_REQUEST;
231 }
232 if (adm_ctx.mdev) {
233 drbd_msg_put_info("no minor number expected");
234 return ERR_INVALID_REQUEST;
235 }
236 if (adm_ctx.my_addr && adm_ctx.peer_addr)
237 adm_ctx.tconn = conn_get_by_addrs(nla_data(adm_ctx.my_addr),
238 nla_len(adm_ctx.my_addr),
239 nla_data(adm_ctx.peer_addr),
240 nla_len(adm_ctx.peer_addr));
241 if (!adm_ctx.tconn) {
242 drbd_msg_put_info("unknown connection");
243 return ERR_INVALID_REQUEST;
244 }
245 }
246
3b98c0c2 247 /* some more paranoia, if the request was over-determined */
527f4b24
LE		 248 if (adm_ctx.mdev && adm_ctx.tconn &&
249 adm_ctx.mdev->tconn != adm_ctx.tconn) {
44e52cfa 250 pr_warning("request: minor=%u, resource=%s; but that minor belongs to connection %s\n",
7c3063cc
AG		 251 adm_ctx.minor, adm_ctx.resource_name,
252 adm_ctx.mdev->tconn->name);
44e52cfa 253 drbd_msg_put_info("minor exists in different resource");
527f4b24
LE		 254 return ERR_INVALID_REQUEST;
255 }
3b98c0c2
LE		 256 if (adm_ctx.mdev &&
257 adm_ctx.volume != VOLUME_UNSPECIFIED &&
258 adm_ctx.volume != adm_ctx.mdev->vnr) {
259 pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n",
260 adm_ctx.minor, adm_ctx.volume,
261 adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name);
527f4b24 262 drbd_msg_put_info("minor exists as different volume");
3b98c0c2
LE		 263 return ERR_INVALID_REQUEST;
264 }
0ace9dfa 265
3b98c0c2
LE		 266 return NO_ERROR;
267
268fail:
269 nlmsg_free(adm_ctx.reply_skb);
270 adm_ctx.reply_skb = NULL;
1e2a2551 271 return err;
3b98c0c2
LE		 272}
273
274static int drbd_adm_finish(struct genl_info *info, int retcode)
275{
0ace9dfa
PR		 276 if (adm_ctx.tconn) {
277 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
278 adm_ctx.tconn = NULL;
279 }
280
3b98c0c2
LE		 281 if (!adm_ctx.reply_skb)
282 return -ENOMEM;
283
284 adm_ctx.reply_dh->ret_code = retcode;
3b98c0c2
LE		 285 drbd_adm_send_reply(adm_ctx.reply_skb, info);
286 return 0;
287}
b411b363 288
6b75dced 289static void setup_khelper_env(struct drbd_tconn *tconn, char **envp)
b411b363 290{
6b75dced 291 char *afs;
b411b363 292
089c075d
AG		 293 /* FIXME: A future version will not allow this case. */
294 if (tconn->my_addr_len == 0 || tconn->peer_addr_len == 0)
295 return;
296
297 switch (((struct sockaddr *)&tconn->peer_addr)->sa_family) {
298 case AF_INET6:
299 afs = "ipv6";
300 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6",
301 &((struct sockaddr_in6 *)&tconn->peer_addr)->sin6_addr);
302 break;
303 case AF_INET:
304 afs = "ipv4";
305 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4",
306 &((struct sockaddr_in *)&tconn->peer_addr)->sin_addr);
307 break;
308 default:
309 afs = "ssocks";
310 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4",
311 &((struct sockaddr_in *)&tconn->peer_addr)->sin_addr);
b411b363 312 }
089c075d 313 snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs);
6b75dced
PR		 314}
315
316int drbd_khelper(struct drbd_conf *mdev, char *cmd)
317{
318 char *envp[] = { "HOME=/",
319 "TERM=linux",
320 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
321 (char[20]) { }, /* address family */
322 (char[60]) { }, /* address */
323 NULL };
324 char mb[12];
325 char *argv[] = {usermode_helper, cmd, mb, NULL };
6f3465ed 326 struct drbd_tconn *tconn = mdev->tconn;
6b75dced
PR		 327 struct sib_info sib;
328 int ret;
329
6f3465ed
LE		 330 if (current == tconn->worker.task)
331 set_bit(CALLBACK_PENDING, &tconn->flags);
332
6b75dced 333 snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev));
6f3465ed 334 setup_khelper_env(tconn, envp);
b411b363 335
1090c056
LE		 336 /* The helper may take some time.
337 * write out any unsynced meta data changes now */
338 drbd_md_sync(mdev);
339
b411b363 340 dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb);
3b98c0c2
LE		 341 sib.sib_reason = SIB_HELPER_PRE;
342 sib.helper_name = cmd;
343 drbd_bcast_event(mdev, &sib);
b411b363
PR		 344 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
345 if (ret)
346 dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
347 usermode_helper, cmd, mb,
348 (ret >> 8) & 0xff, ret);
349 else
350 dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
351 usermode_helper, cmd, mb,
352 (ret >> 8) & 0xff, ret);
3b98c0c2
LE		 353 sib.sib_reason = SIB_HELPER_POST;
354 sib.helper_exit_code = ret;
355 drbd_bcast_event(mdev, &sib);
b411b363 356
6f3465ed
LE		 357 if (current == tconn->worker.task)
358 clear_bit(CALLBACK_PENDING, &tconn->flags);
359
b411b363
PR		 360 if (ret < 0) /* Ignore any ERRNOs we got. */
361 ret = 0;
362
363 return ret;
364}
365
6b75dced
PR		 366static void conn_md_sync(struct drbd_tconn *tconn)
367{
368 struct drbd_conf *mdev;
e90285e0 369 int vnr;
6b75dced 370
c141ebda
PR		 371 rcu_read_lock();
372 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
373 kref_get(&mdev->kref);
374 rcu_read_unlock();
6b75dced 375 drbd_md_sync(mdev);
c141ebda
PR		 376 kref_put(&mdev->kref, &drbd_minor_destroy);
377 rcu_read_lock();
378 }
379 rcu_read_unlock();
6b75dced
PR		 380}
381
382int conn_khelper(struct drbd_tconn *tconn, char *cmd)
383{
384 char *envp[] = { "HOME=/",
385 "TERM=linux",
386 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
387 (char[20]) { }, /* address family */
388 (char[60]) { }, /* address */
389 NULL };
390 char *argv[] = {usermode_helper, cmd, tconn->name, NULL };
391 int ret;
392
393 setup_khelper_env(tconn, envp);
394 conn_md_sync(tconn);
395
396 conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name);
397 /* TODO: conn_bcast_event() ?? */
398
399 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
400 if (ret)
401 conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n",
402 usermode_helper, cmd, tconn->name,
403 (ret >> 8) & 0xff, ret);
404 else
405 conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n",
406 usermode_helper, cmd, tconn->name,
407 (ret >> 8) & 0xff, ret);
408 /* TODO: conn_bcast_event() ?? */
409
410 if (ret < 0) /* Ignore any ERRNOs we got. */
411 ret = 0;
412
413 return ret;
414}
415
cb703454 416static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn)
b411b363 417{
cb703454
PR		 418 enum drbd_fencing_p fp = FP_NOT_AVAIL;
419 struct drbd_conf *mdev;
420 int vnr;
421
695d08fa 422 rcu_read_lock();
cb703454
PR		 423 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
424 if (get_ldev_if_state(mdev, D_CONSISTENT)) {
daeda1cc
PR		 425 fp = max_t(enum drbd_fencing_p, fp,
426 rcu_dereference(mdev->ldev->disk_conf)->fencing);
cb703454
PR		 427 put_ldev(mdev);
428 }
429 }
695d08fa 430 rcu_read_unlock();
cb703454
PR		 431
432 return fp;
433}
434
435bool conn_try_outdate_peer(struct drbd_tconn *tconn)
436{
437 union drbd_state mask = { };
438 union drbd_state val = { };
439 enum drbd_fencing_p fp;
b411b363
PR		 440 char *ex_to_string;
441 int r;
b411b363 442
cb703454
PR		 443 if (tconn->cstate >= C_WF_REPORT_PARAMS) {
444 conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n");
445 return false;
446 }
b411b363 447
cb703454
PR		 448 fp = highest_fencing_policy(tconn);
449 switch (fp) {
450 case FP_NOT_AVAIL:
451 conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n");
fb22c402 452 goto out;
cb703454
PR		 453 case FP_DONT_CARE:
454 return true;
455 default: ;
b411b363
PR		 456 }
457
cb703454 458 r = conn_khelper(tconn, "fence-peer");
b411b363
PR		 459
460 switch ((r>>8) & 0xff) {
461 case 3: /* peer is inconsistent */
462 ex_to_string = "peer is inconsistent or worse";
cb703454
PR		 463 mask.pdsk = D_MASK;
464 val.pdsk = D_INCONSISTENT;
b411b363
PR		 465 break;
466 case 4: /* peer got outdated, or was already outdated */
467 ex_to_string = "peer was fenced";
cb703454
PR		 468 mask.pdsk = D_MASK;
469 val.pdsk = D_OUTDATED;
b411b363
PR		 470 break;
471 case 5: /* peer was down */
cb703454 472 if (conn_highest_disk(tconn) == D_UP_TO_DATE) {
b411b363
PR		 473 /* we will(have) create(d) a new UUID anyways... */
474 ex_to_string = "peer is unreachable, assumed to be dead";
cb703454
PR		 475 mask.pdsk = D_MASK;
476 val.pdsk = D_OUTDATED;
b411b363
PR		 477 } else {
478 ex_to_string = "peer unreachable, doing nothing since disk != UpToDate";
b411b363
PR		 479 }
480 break;
481 case 6: /* Peer is primary, voluntarily outdate myself.
482 * This is useful when an unconnected R_SECONDARY is asked to
483 * become R_PRIMARY, but finds the other peer being active. */
484 ex_to_string = "peer is active";
cb703454
PR		 485 conn_warn(tconn, "Peer is primary, outdating myself.\n");
486 mask.disk = D_MASK;
487 val.disk = D_OUTDATED;
b411b363
PR		 488 break;
489 case 7:
490 if (fp != FP_STONITH)
cb703454 491 conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n");
b411b363 492 ex_to_string = "peer was stonithed";
cb703454
PR		 493 mask.pdsk = D_MASK;
494 val.pdsk = D_OUTDATED;
b411b363
PR		 495 break;
496 default:
497 /* The script is broken ... */
cb703454
PR		 498 conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff);
499 return false; /* Eventually leave IO frozen */
b411b363
PR		 500 }
501
cb703454
PR		 502 conn_info(tconn, "fence-peer helper returned %d (%s)\n",
503 (r>>8) & 0xff, ex_to_string);
fb22c402 504
cb703454 505 out:
fb22c402 506
cb703454
PR		 507 /* Not using
508 conn_request_state(tconn, mask, val, CS_VERBOSE);
509 here, because we might were able to re-establish the connection in the
510 meantime. */
511 spin_lock_irq(&tconn->req_lock);
a1096a6e 512 if (tconn->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &tconn->flags))
cb703454
PR		 513 _conn_request_state(tconn, mask, val, CS_VERBOSE);
514 spin_unlock_irq(&tconn->req_lock);
515
516 return conn_highest_pdsk(tconn) <= D_OUTDATED;
b411b363
PR		 517}
518
87f7be4c
PR		 519static int _try_outdate_peer_async(void *data)
520{
cb703454 521 struct drbd_tconn *tconn = (struct drbd_tconn *)data;
87f7be4c 522
cb703454 523 conn_try_outdate_peer(tconn);
87f7be4c 524
9dc9fbb3 525 kref_put(&tconn->kref, &conn_destroy);
87f7be4c
PR		 526 return 0;
527}
528
cb703454 529void conn_try_outdate_peer_async(struct drbd_tconn *tconn)
87f7be4c
PR		 530{
531 struct task_struct *opa;
532
9dc9fbb3 533 kref_get(&tconn->kref);
cb703454 534 opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h");
9dc9fbb3 535 if (IS_ERR(opa)) {
cb703454 536 conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n");
9dc9fbb3
PR		 537 kref_put(&tconn->kref, &conn_destroy);
538 }
87f7be4c 539}
b411b363 540
bf885f8a
AG		 541enum drbd_state_rv
542drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
b411b363
PR		 543{
544 const int max_tries = 4;
bf885f8a 545 enum drbd_state_rv rv = SS_UNKNOWN_ERROR;
44ed167d 546 struct net_conf *nc;
b411b363
PR		 547 int try = 0;
548 int forced = 0;
549 union drbd_state mask, val;
b411b363
PR		 550
551 if (new_role == R_PRIMARY)
0625ac19 552 request_ping(mdev->tconn); /* Detect a dead peer ASAP */
b411b363 553
8410da8f 554 mutex_lock(mdev->state_mutex);
b411b363
PR		 555
556 mask.i = 0; mask.role = R_MASK;
557 val.i = 0; val.role = new_role;
558
559 while (try++ < max_tries) {
bf885f8a 560 rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE);
b411b363
PR		 561
562 /* in case we first succeeded to outdate,
563 * but now suddenly could establish a connection */
bf885f8a 564 if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) {
b411b363
PR		 565 val.pdsk = 0;
566 mask.pdsk = 0;
567 continue;
568 }
569
bf885f8a 570 if (rv == SS_NO_UP_TO_DATE_DISK && force &&
d10a33c6
PR		 571 (mdev->state.disk < D_UP_TO_DATE &&
572 mdev->state.disk >= D_INCONSISTENT)) {
b411b363
PR		 573 mask.disk = D_MASK;
574 val.disk = D_UP_TO_DATE;
575 forced = 1;
576 continue;
577 }
578
bf885f8a 579 if (rv == SS_NO_UP_TO_DATE_DISK &&
b411b363
PR		 580 mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) {
581 D_ASSERT(mdev->state.pdsk == D_UNKNOWN);
b411b363 582
cb703454 583 if (conn_try_outdate_peer(mdev->tconn)) {
b411b363
PR		 584 val.disk = D_UP_TO_DATE;
585 mask.disk = D_MASK;
586 }
b411b363
PR		 587 continue;
588 }
589
bf885f8a 590 if (rv == SS_NOTHING_TO_DO)
3b98c0c2 591 goto out;
bf885f8a 592 if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) {
cb703454 593 if (!conn_try_outdate_peer(mdev->tconn) && force) {
b411b363 594 dev_warn(DEV, "Forced into split brain situation!\n");
cb703454
PR		 595 mask.pdsk = D_MASK;
596 val.pdsk = D_OUTDATED;
b411b363 597
cb703454 598 }
b411b363
PR		 599 continue;
600 }
bf885f8a 601 if (rv == SS_TWO_PRIMARIES) {
b411b363
PR		 602 /* Maybe the peer is detected as dead very soon...
603 retry at most once more in this case. */
44ed167d
PR		 604 int timeo;
605 rcu_read_lock();
606 nc = rcu_dereference(mdev->tconn->net_conf);
607 timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1;
608 rcu_read_unlock();
609 schedule_timeout_interruptible(timeo);
b411b363
PR		 610 if (try < max_tries)
611 try = max_tries - 1;
612 continue;
613 }
bf885f8a
AG		 614 if (rv < SS_SUCCESS) {
615 rv = _drbd_request_state(mdev, mask, val,
b411b363 616 CS_VERBOSE + CS_WAIT_COMPLETE);
bf885f8a 617 if (rv < SS_SUCCESS)
3b98c0c2 618 goto out;
b411b363
PR		 619 }
620 break;
621 }
622
bf885f8a 623 if (rv < SS_SUCCESS)
3b98c0c2 624 goto out;
b411b363
PR		 625
626 if (forced)
627 dev_warn(DEV, "Forced to consider local data as UpToDate!\n");
628
629 /* Wait until nothing is on the fly :) */
630 wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0);
631
b6dd1a89
LE		 632 /* FIXME also wait for all pending P_BARRIER_ACK? */
633
b411b363 634 if (new_role == R_SECONDARY) {
81e84650 635 set_disk_ro(mdev->vdisk, true);
b411b363
PR		 636 if (get_ldev(mdev)) {
637 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
638 put_ldev(mdev);
639 }
640 } else {
a0095508 641 mutex_lock(&mdev->tconn->conf_update);
91fd4dad 642 nc = mdev->tconn->net_conf;
44ed167d 643 if (nc)
6139f60d 644 nc->discard_my_data = 0; /* without copy; single bit op is atomic */
a0095508 645 mutex_unlock(&mdev->tconn->conf_update);
91fd4dad 646
81e84650 647 set_disk_ro(mdev->vdisk, false);
b411b363
PR		 648 if (get_ldev(mdev)) {
649 if (((mdev->state.conn < C_CONNECTED ||
650 mdev->state.pdsk <= D_FAILED)
651 && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced)
652 drbd_uuid_new_current(mdev);
653
654 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
655 put_ldev(mdev);
656 }
657 }
658
19f843aa
LE		 659 /* writeout of activity log covered areas of the bitmap
660 * to stable storage done in after state change already */
b411b363
PR		 661
662 if (mdev->state.conn >= C_WF_REPORT_PARAMS) {
663 /* if this was forced, we should consider sync */
664 if (forced)
665 drbd_send_uuids(mdev);
43de7c85 666 drbd_send_current_state(mdev);
b411b363
PR		 667 }
668
669 drbd_md_sync(mdev);
670
671 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
3b98c0c2 672out:
8410da8f 673 mutex_unlock(mdev->state_mutex);
bf885f8a 674 return rv;
b411b363
PR		 675}
676
3b98c0c2 677static const char *from_attrs_err_to_txt(int err)
b411b363 678{
3b98c0c2
LE		 679 return err == -ENOMSG ? "required attribute missing" :
680 err == -EOPNOTSUPP ? "unknown mandatory attribute" :
f399002e 681 err == -EEXIST ? "can not change invariant setting" :
3b98c0c2 682 "invalid attribute value";
b411b363
PR		 683}
684
3b98c0c2 685int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info)
b411b363 686{
3b98c0c2
LE		 687 struct set_role_parms parms;
688 int err;
689 enum drbd_ret_code retcode;
b411b363 690
3b98c0c2
LE		 691 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
692 if (!adm_ctx.reply_skb)
693 return retcode;
694 if (retcode != NO_ERROR)
695 goto out;
696
697 memset(&parms, 0, sizeof(parms));
698 if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) {
f399002e 699 err = set_role_parms_from_attrs(&parms, info);
3b98c0c2
LE		 700 if (err) {
701 retcode = ERR_MANDATORY_TAG;
702 drbd_msg_put_info(from_attrs_err_to_txt(err));
703 goto out;
704 }
705 }
706
707 if (info->genlhdr->cmd == DRBD_ADM_PRIMARY)
708 retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate);
709 else
710 retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0);
711out:
712 drbd_adm_finish(info, retcode);
b411b363
PR		 713 return 0;
714}
715
716/* initializes the md.*_offset members, so we are able to find
717 * the on disk meta data */
718static void drbd_md_set_sector_offsets(struct drbd_conf *mdev,
719 struct drbd_backing_dev *bdev)
720{
721 sector_t md_size_sect = 0;
daeda1cc
PR		 722 int meta_dev_idx;
723
724 rcu_read_lock();
725 meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx;
726
727 switch (meta_dev_idx) {
b411b363
PR		 728 default:
729 /* v07 style fixed size indexed meta data */
730 bdev->md.md_size_sect = MD_RESERVED_SECT;
731 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
732 bdev->md.al_offset = MD_AL_OFFSET;
733 bdev->md.bm_offset = MD_BM_OFFSET;
734 break;
735 case DRBD_MD_INDEX_FLEX_EXT:
736 /* just occupy the full device; unit: sectors */
737 bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev);
738 bdev->md.md_offset = 0;
739 bdev->md.al_offset = MD_AL_OFFSET;
740 bdev->md.bm_offset = MD_BM_OFFSET;
741 break;
742 case DRBD_MD_INDEX_INTERNAL:
743 case DRBD_MD_INDEX_FLEX_INT:
744 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
745 /* al size is still fixed */
7ad651b5 746 bdev->md.al_offset = -MD_AL_SECTORS;
b411b363
PR		 747 /* we need (slightly less than) ~ this much bitmap sectors: */
748 md_size_sect = drbd_get_capacity(bdev->backing_bdev);
749 md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT);
750 md_size_sect = BM_SECT_TO_EXT(md_size_sect);
751 md_size_sect = ALIGN(md_size_sect, 8);
752
753 /* plus the "drbd meta data super block",
754 * and the activity log; */
755 md_size_sect += MD_BM_OFFSET;
756
757 bdev->md.md_size_sect = md_size_sect;
758 /* bitmap offset is adjusted by 'super' block size */
759 bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET;
760 break;
761 }
daeda1cc 762 rcu_read_unlock();
b411b363
PR		 763}
764
4b0715f0 765/* input size is expected to be in KB */
b411b363
PR		 766char *ppsize(char *buf, unsigned long long size)
767{
4b0715f0
LE		 768 /* Needs 9 bytes at max including trailing NUL:
769 * -1ULL ==> "16384 EB" */
b411b363
PR		 770 static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
771 int base = 0;
4b0715f0 772 while (size >= 10000 && base < sizeof(units)-1) {
b411b363
PR		 773 /* shift + round */
774 size = (size >> 10) + !!(size & (1<<9));
775 base++;
776 }
4b0715f0 777 sprintf(buf, "%u %cB", (unsigned)size, units[base]);
b411b363
PR		 778
779 return buf;
780}
781
782/* there is still a theoretical deadlock when called from receiver
783 * on an D_INCONSISTENT R_PRIMARY:
784 * remote READ does inc_ap_bio, receiver would need to receive answer
785 * packet from remote to dec_ap_bio again.
786 * receiver receive_sizes(), comes here,
787 * waits for ap_bio_cnt == 0. -> deadlock.
788 * but this cannot happen, actually, because:
789 * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable
790 * (not connected, or bad/no disk on peer):
791 * see drbd_fail_request_early, ap_bio_cnt is zero.
792 * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
793 * peer may not initiate a resize.
794 */
3b98c0c2
LE		 795/* Note these are not to be confused with
796 * drbd_adm_suspend_io/drbd_adm_resume_io,
797 * which are (sub) state changes triggered by admin (drbdsetup),
798 * and can be long lived.
799 * This changes an mdev->flag, is triggered by drbd internals,
800 * and should be short-lived. */
b411b363
PR		 801void drbd_suspend_io(struct drbd_conf *mdev)
802{
803 set_bit(SUSPEND_IO, &mdev->flags);
2aebfabb 804 if (drbd_suspended(mdev))
265be2d0 805 return;
b411b363
PR		 806 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt));
807}
808
809void drbd_resume_io(struct drbd_conf *mdev)
810{
811 clear_bit(SUSPEND_IO, &mdev->flags);
812 wake_up(&mdev->misc_wait);
813}
814
815/**
816 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
817 * @mdev: DRBD device.
818 *
819 * Returns 0 on success, negative return values indicate errors.
820 * You should call drbd_md_sync() after calling this function.
821 */
24c4830c 822enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local)
b411b363
PR		 823{
824 sector_t prev_first_sect, prev_size; /* previous meta location */
ef5e44a6 825 sector_t la_size, u_size;
b411b363
PR		 826 sector_t size;
827 char ppb[10];
828
829 int md_moved, la_size_changed;
830 enum determine_dev_size rv = unchanged;
831
832 /* race:
833 * application request passes inc_ap_bio,
834 * but then cannot get an AL-reference.
835 * this function later may wait on ap_bio_cnt == 0. -> deadlock.
836 *
837 * to avoid that:
838 * Suspend IO right here.
839 * still lock the act_log to not trigger ASSERTs there.
840 */
841 drbd_suspend_io(mdev);
842
843 /* no wait necessary anymore, actually we could assert that */
844 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
845
846 prev_first_sect = drbd_md_first_sector(mdev->ldev);
847 prev_size = mdev->ldev->md.md_size_sect;
848 la_size = mdev->ldev->md.la_size_sect;
849
850 /* TODO: should only be some assert here, not (re)init... */
851 drbd_md_set_sector_offsets(mdev, mdev->ldev);
852
daeda1cc
PR		 853 rcu_read_lock();
854 u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size;
855 rcu_read_unlock();
ef5e44a6 856 size = drbd_new_dev_size(mdev, mdev->ldev, u_size, flags & DDSF_FORCED);
b411b363
PR		 857
858 if (drbd_get_capacity(mdev->this_bdev) != size ||
859 drbd_bm_capacity(mdev) != size) {
860 int err;
02d9a94b 861 err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC));
b411b363
PR		 862 if (unlikely(err)) {
863 /* currently there is only one error: ENOMEM! */
864 size = drbd_bm_capacity(mdev)>>1;
865 if (size == 0) {
866 dev_err(DEV, "OUT OF MEMORY! "
867 "Could not allocate bitmap!\n");
868 } else {
869 dev_err(DEV, "BM resizing failed. "
870 "Leaving size unchanged at size = %lu KB\n",
871 (unsigned long)size);
872 }
873 rv = dev_size_error;
874 }
875 /* racy, see comments above. */
876 drbd_set_my_capacity(mdev, size);
877 mdev->ldev->md.la_size_sect = size;
878 dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1),
879 (unsigned long long)size>>1);
880 }
881 if (rv == dev_size_error)
882 goto out;
883
884 la_size_changed = (la_size != mdev->ldev->md.la_size_sect);
885
886 md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev)
887 || prev_size != mdev->ldev->md.md_size_sect;
888
889 if (la_size_changed || md_moved) {
24dccabb
AG		 890 int err;
891
b411b363
PR		 892 drbd_al_shrink(mdev); /* All extents inactive. */
893 dev_info(DEV, "Writing the whole bitmap, %s\n",
894 la_size_changed && md_moved ? "size changed and md moved" :
895 la_size_changed ? "size changed" : "md moved");
20ceb2b2
LE		 896 /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */
897 err = drbd_bitmap_io(mdev, &drbd_bm_write,
898 "size changed", BM_LOCKED_MASK);
24dccabb
AG		 899 if (err) {
900 rv = dev_size_error;
901 goto out;
902 }
b411b363
PR		 903 drbd_md_mark_dirty(mdev);
904 }
905
906 if (size > la_size)
907 rv = grew;
908 if (size < la_size)
909 rv = shrunk;
910out:
911 lc_unlock(mdev->act_log);
912 wake_up(&mdev->al_wait);
913 drbd_resume_io(mdev);
914
915 return rv;
916}
917
918sector_t
ef5e44a6
PR		 919drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev,
920 sector_t u_size, int assume_peer_has_space)
b411b363
PR		 921{
922 sector_t p_size = mdev->p_size; /* partner's disk size. */
923 sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */
924 sector_t m_size; /* my size */
b411b363
PR		 925 sector_t size = 0;
926
927 m_size = drbd_get_max_capacity(bdev);
928
a393db6f
PR		 929 if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) {
930 dev_warn(DEV, "Resize while not connected was forced by the user!\n");
931 p_size = m_size;
932 }
933
b411b363
PR		 934 if (p_size && m_size) {
935 size = min_t(sector_t, p_size, m_size);
936 } else {
937 if (la_size) {
938 size = la_size;
939 if (m_size && m_size < size)
940 size = m_size;
941 if (p_size && p_size < size)
942 size = p_size;
943 } else {
944 if (m_size)
945 size = m_size;
946 if (p_size)
947 size = p_size;
948 }
949 }
950
951 if (size == 0)
952 dev_err(DEV, "Both nodes diskless!\n");
953
954 if (u_size) {
955 if (u_size > size)
956 dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n",
957 (unsigned long)u_size>>1, (unsigned long)size>>1);
958 else
959 size = u_size;
960 }
961
962 return size;
963}
964
965/**
966 * drbd_check_al_size() - Ensures that the AL is of the right size
967 * @mdev: DRBD device.
968 *
969 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
970 * failed, and 0 on success. You should call drbd_md_sync() after you called
971 * this function.
972 */
f399002e 973static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc)
b411b363
PR		 974{
975 struct lru_cache *n, *t;
976 struct lc_element *e;
977 unsigned int in_use;
978 int i;
979
b411b363 980 if (mdev->act_log &&
f399002e 981 mdev->act_log->nr_elements == dc->al_extents)
b411b363
PR		 982 return 0;
983
984 in_use = 0;
985 t = mdev->act_log;
7ad651b5 986 n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION,
f399002e 987 dc->al_extents, sizeof(struct lc_element), 0);
b411b363
PR		 988
989 if (n == NULL) {
990 dev_err(DEV, "Cannot allocate act_log lru!\n");
991 return -ENOMEM;
992 }
993 spin_lock_irq(&mdev->al_lock);
994 if (t) {
995 for (i = 0; i < t->nr_elements; i++) {
996 e = lc_element_by_index(t, i);
997 if (e->refcnt)
998 dev_err(DEV, "refcnt(%d)==%d\n",
999 e->lc_number, e->refcnt);
1000 in_use += e->refcnt;
1001 }
1002 }
1003 if (!in_use)
1004 mdev->act_log = n;
1005 spin_unlock_irq(&mdev->al_lock);
1006 if (in_use) {
1007 dev_err(DEV, "Activity log still in use!\n");
1008 lc_destroy(n);
1009 return -EBUSY;
1010 } else {
1011 if (t)
1012 lc_destroy(t);
1013 }
1014 drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */
1015 return 0;
1016}
1017
99432fcc 1018static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size)
b411b363
PR		 1019{
1020 struct request_queue * const q = mdev->rq_queue;
99432fcc
PR		 1021 int max_hw_sectors = max_bio_size >> 9;
1022 int max_segments = 0;
1023
1024 if (get_ldev_if_state(mdev, D_ATTACHING)) {
1025 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
1026
1027 max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9);
daeda1cc
PR		 1028 rcu_read_lock();
1029 max_segments = rcu_dereference(mdev->ldev->disk_conf)->max_bio_bvecs;
1030 rcu_read_unlock();
99432fcc
PR		 1031 put_ldev(mdev);
1032 }
b411b363 1033
b411b363 1034 blk_queue_logical_block_size(q, 512);
1816a2b4
LE		 1035 blk_queue_max_hw_sectors(q, max_hw_sectors);
1036 /* This is the workaround for "bio would need to, but cannot, be split" */
1037 blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS);
1038 blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1);
b411b363 1039
99432fcc
PR		 1040 if (get_ldev_if_state(mdev, D_ATTACHING)) {
1041 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
1042
1043 blk_queue_stack_limits(q, b);
1044
1045 if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) {
1046 dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
1047 q->backing_dev_info.ra_pages,
1048 b->backing_dev_info.ra_pages);
1049 q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages;
1050 }
1051 put_ldev(mdev);
1052 }
1053}
1054
1055void drbd_reconsider_max_bio_size(struct drbd_conf *mdev)
1056{
1057 int now, new, local, peer;
1058
1059 now = queue_max_hw_sectors(mdev->rq_queue) << 9;
1060 local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */
1061 peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */
b411b363 1062
99432fcc
PR		 1063 if (get_ldev_if_state(mdev, D_ATTACHING)) {
1064 local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9;
1065 mdev->local_max_bio_size = local;
1066 put_ldev(mdev);
b411b363 1067 }
99432fcc
PR		 1068
1069 /* We may ignore peer limits if the peer is modern enough.
1070 Because new from 8.3.8 onwards the peer can use multiple
1071 BIOs for a single peer_request */
1072 if (mdev->state.conn >= C_CONNECTED) {
31890f4a 1073 if (mdev->tconn->agreed_pro_version < 94)
2ffca4f3
PR		 1074 peer = min_t(int, mdev->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET);
1075 /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */
31890f4a 1076 else if (mdev->tconn->agreed_pro_version == 94)
99432fcc 1077 peer = DRBD_MAX_SIZE_H80_PACKET;
2ffca4f3
PR		 1078 else if (mdev->tconn->agreed_pro_version < 100)
1079 peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */
1080 else
99432fcc
PR		 1081 peer = DRBD_MAX_BIO_SIZE;
1082 }
1083
1084 new = min_t(int, local, peer);
1085
1086 if (mdev->state.role == R_PRIMARY && new < now)
1087 dev_err(DEV, "ASSERT FAILED new < now; (%d < %d)\n", new, now);
1088
1089 if (new != now)
1090 dev_info(DEV, "max BIO size = %u\n", new);
1091
1092 drbd_setup_queue_param(mdev, new);
b411b363
PR		 1093}
1094
a18e9d1e 1095/* Starts the worker thread */
0e29d163 1096static void conn_reconfig_start(struct drbd_tconn *tconn)
b411b363 1097{
0e29d163
PR		 1098 drbd_thread_start(&tconn->worker);
1099 conn_flush_workqueue(tconn);
b411b363
PR		 1100}
1101
a18e9d1e 1102/* if still unconfigured, stops worker again. */
0e29d163 1103static void conn_reconfig_done(struct drbd_tconn *tconn)
b411b363 1104{
992d6e91 1105 bool stop_threads;
0e29d163 1106 spin_lock_irq(&tconn->req_lock);
e0e16653
PR		 1107 stop_threads = conn_all_vols_unconf(tconn) &&
1108 tconn->cstate == C_STANDALONE;
0e29d163 1109 spin_unlock_irq(&tconn->req_lock);
992d6e91
LE		 1110 if (stop_threads) {
1111 /* asender is implicitly stopped by receiver
81fa2e67 1112 * in conn_disconnect() */
992d6e91
LE		 1113 drbd_thread_stop(&tconn->receiver);
1114 drbd_thread_stop(&tconn->worker);
1115 }
b411b363
PR		 1116}
1117
0778286a
PR		 1118/* Make sure IO is suspended before calling this function(). */
1119static void drbd_suspend_al(struct drbd_conf *mdev)
1120{
1121 int s = 0;
1122
61610420 1123 if (!lc_try_lock(mdev->act_log)) {
0778286a
PR		 1124 dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n");
1125 return;
1126 }
1127
61610420 1128 drbd_al_shrink(mdev);
87eeee41 1129 spin_lock_irq(&mdev->tconn->req_lock);
0778286a
PR		 1130 if (mdev->state.conn < C_CONNECTED)
1131 s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags);
87eeee41 1132 spin_unlock_irq(&mdev->tconn->req_lock);
61610420 1133 lc_unlock(mdev->act_log);
0778286a
PR		 1134
1135 if (s)
1136 dev_info(DEV, "Suspended AL updates\n");
1137}
1138
5979e361
LE		 1139
1140static bool should_set_defaults(struct genl_info *info)
1141{
1142 unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags;
1143 return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS);
1144}
1145
d589a21e
PR		 1146static void enforce_disk_conf_limits(struct disk_conf *dc)
1147{
1148 if (dc->al_extents < DRBD_AL_EXTENTS_MIN)
1149 dc->al_extents = DRBD_AL_EXTENTS_MIN;
1150 if (dc->al_extents > DRBD_AL_EXTENTS_MAX)
1151 dc->al_extents = DRBD_AL_EXTENTS_MAX;
1152
1153 if (dc->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX)
1154 dc->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX;
1155}
1156
f399002e
LE		 1157int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info)
1158{
1159 enum drbd_ret_code retcode;
1160 struct drbd_conf *mdev;
daeda1cc 1161 struct disk_conf *new_disk_conf, *old_disk_conf;
813472ce 1162 struct fifo_buffer *old_plan = NULL, *new_plan = NULL;
f399002e 1163 int err, fifo_size;
f399002e
LE		 1164
1165 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1166 if (!adm_ctx.reply_skb)
1167 return retcode;
1168 if (retcode != NO_ERROR)
1169 goto out;
1170
1171 mdev = adm_ctx.mdev;
1172
1173 /* we also need a disk
1174 * to change the options on */
1175 if (!get_ldev(mdev)) {
1176 retcode = ERR_NO_DISK;
1177 goto out;
1178 }
1179
daeda1cc 1180 new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL);
5ecc72c3 1181 if (!new_disk_conf) {
f399002e
LE		 1182 retcode = ERR_NOMEM;
1183 goto fail;
1184 }
1185
daeda1cc
PR		 1186 mutex_lock(&mdev->tconn->conf_update);
1187 old_disk_conf = mdev->ldev->disk_conf;
1188 *new_disk_conf = *old_disk_conf;
5979e361 1189 if (should_set_defaults(info))
b966b5dd 1190 set_disk_conf_defaults(new_disk_conf);
5979e361 1191
5ecc72c3 1192 err = disk_conf_from_attrs_for_change(new_disk_conf, info);
c75b9b10 1193 if (err && err != -ENOMSG) {
f399002e
LE		 1194 retcode = ERR_MANDATORY_TAG;
1195 drbd_msg_put_info(from_attrs_err_to_txt(err));
1196 }
1197
5ecc72c3
LE		 1198 if (!expect(new_disk_conf->resync_rate >= 1))
1199 new_disk_conf->resync_rate = 1;
f399002e 1200
d589a21e 1201 enforce_disk_conf_limits(new_disk_conf);
f399002e 1202
5ecc72c3 1203 fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ;
9958c857 1204 if (fifo_size != mdev->rs_plan_s->size) {
813472ce
PR		 1205 new_plan = fifo_alloc(fifo_size);
1206 if (!new_plan) {
f399002e
LE		 1207 dev_err(DEV, "kmalloc of fifo_buffer failed");
1208 retcode = ERR_NOMEM;
daeda1cc 1209 goto fail_unlock;
f399002e
LE		 1210 }
1211 }
1212
f399002e
LE		 1213 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
1214 drbd_al_shrink(mdev);
5ecc72c3 1215 err = drbd_check_al_size(mdev, new_disk_conf);
f399002e
LE		 1216 lc_unlock(mdev->act_log);
1217 wake_up(&mdev->al_wait);
1218
1219 if (err) {
1220 retcode = ERR_NOMEM;
daeda1cc 1221 goto fail_unlock;
f399002e
LE		 1222 }
1223
dc97b708 1224 write_lock_irq(&global_state_lock);
95f8efd0 1225 retcode = drbd_resync_after_valid(mdev, new_disk_conf->resync_after);
dc97b708 1226 if (retcode == NO_ERROR) {
daeda1cc 1227 rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf);
95f8efd0 1228 drbd_resync_after_changed(mdev);
dc97b708
PR		 1229 }
1230 write_unlock_irq(&global_state_lock);
f399002e 1231
daeda1cc
PR		 1232 if (retcode != NO_ERROR)
1233 goto fail_unlock;
f399002e 1234
813472ce
PR		 1235 if (new_plan) {
1236 old_plan = mdev->rs_plan_s;
1237 rcu_assign_pointer(mdev->rs_plan_s, new_plan);
9958c857 1238 }
9958c857 1239
c141ebda 1240 mutex_unlock(&mdev->tconn->conf_update);
27eb13e9 1241
9a51ab1c
PR		 1242 if (new_disk_conf->al_updates)
1243 mdev->ldev->md.flags &= MDF_AL_DISABLED;
1244 else
1245 mdev->ldev->md.flags |= MDF_AL_DISABLED;
1246
27eb13e9
PR		 1247 drbd_bump_write_ordering(mdev->tconn, WO_bdev_flush);
1248
daeda1cc 1249 drbd_md_sync(mdev);
f399002e
LE		 1250
1251 if (mdev->state.conn >= C_CONNECTED)
1252 drbd_send_sync_param(mdev);
1253
daeda1cc
PR		 1254 synchronize_rcu();
1255 kfree(old_disk_conf);
813472ce 1256 kfree(old_plan);
cdfda633 1257 mod_timer(&mdev->request_timer, jiffies + HZ);
daeda1cc
PR		 1258 goto success;
1259
1260fail_unlock:
1261 mutex_unlock(&mdev->tconn->conf_update);
f399002e 1262 fail:
5ecc72c3 1263 kfree(new_disk_conf);
813472ce 1264 kfree(new_plan);
daeda1cc
PR		 1265success:
1266 put_ldev(mdev);
f399002e
LE		 1267 out:
1268 drbd_adm_finish(info, retcode);
1269 return 0;
1270}
1271
3b98c0c2 1272int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info)
b411b363 1273{
3b98c0c2
LE		 1274 struct drbd_conf *mdev;
1275 int err;
116676ca 1276 enum drbd_ret_code retcode;
b411b363
PR		 1277 enum determine_dev_size dd;
1278 sector_t max_possible_sectors;
1279 sector_t min_md_device_sectors;
1280 struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */
daeda1cc 1281 struct disk_conf *new_disk_conf = NULL;
e525fd89 1282 struct block_device *bdev;
b411b363 1283 struct lru_cache *resync_lru = NULL;
9958c857 1284 struct fifo_buffer *new_plan = NULL;
b411b363 1285 union drbd_state ns, os;
f2024e7c 1286 enum drbd_state_rv rv;
44ed167d 1287 struct net_conf *nc;
b411b363 1288
3b98c0c2
LE		 1289 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1290 if (!adm_ctx.reply_skb)
1291 return retcode;
1292 if (retcode != NO_ERROR)
40cbf085 1293 goto finish;
3b98c0c2
LE		 1294
1295 mdev = adm_ctx.mdev;
0e29d163 1296 conn_reconfig_start(mdev->tconn);
b411b363
PR		 1297
1298 /* if you want to reconfigure, please tear down first */
1299 if (mdev->state.disk > D_DISKLESS) {
1300 retcode = ERR_DISK_CONFIGURED;
1301 goto fail;
1302 }
82f59cc6
LE		 1303 /* It may just now have detached because of IO error. Make sure
1304 * drbd_ldev_destroy is done already, we may end up here very fast,
1305 * e.g. if someone calls attach from the on-io-error handler,
1306 * to realize a "hot spare" feature (not that I'd recommend that) */
1307 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
b411b363 1308
0c849666
LE		 1309 /* make sure there is no leftover from previous force-detach attempts */
1310 clear_bit(FORCE_DETACH, &mdev->flags);
1311
a324896b
LE		 1312 /* and no leftover from previously aborted resync or verify, either */
1313 mdev->rs_total = 0;
1314 mdev->rs_failed = 0;
1315 atomic_set(&mdev->rs_pending_cnt, 0);
1316
3b98c0c2 1317 /* allocation not in the IO path, drbdsetup context */
b411b363
PR		 1318 nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL);
1319 if (!nbc) {
1320 retcode = ERR_NOMEM;
1321 goto fail;
1322 }
daeda1cc
PR		 1323 new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL);
1324 if (!new_disk_conf) {
1325 retcode = ERR_NOMEM;
1326 goto fail;
1327 }
1328 nbc->disk_conf = new_disk_conf;
b411b363 1329
daeda1cc
PR		 1330 set_disk_conf_defaults(new_disk_conf);
1331 err = disk_conf_from_attrs(new_disk_conf, info);
3b98c0c2 1332 if (err) {
b411b363 1333 retcode = ERR_MANDATORY_TAG;
3b98c0c2 1334 drbd_msg_put_info(from_attrs_err_to_txt(err));
b411b363
PR		 1335 goto fail;
1336 }
1337
d589a21e
PR		 1338 enforce_disk_conf_limits(new_disk_conf);
1339
9958c857
PR		 1340 new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ);
1341 if (!new_plan) {
1342 retcode = ERR_NOMEM;
1343 goto fail;
1344 }
1345
daeda1cc 1346 if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) {
b411b363
PR		 1347 retcode = ERR_MD_IDX_INVALID;
1348 goto fail;
1349 }
1350
44ed167d
PR		 1351 rcu_read_lock();
1352 nc = rcu_dereference(mdev->tconn->net_conf);
1353 if (nc) {
daeda1cc 1354 if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) {
44ed167d 1355 rcu_read_unlock();
47ff2d0a
PR		 1356 retcode = ERR_STONITH_AND_PROT_A;
1357 goto fail;
1358 }
1359 }
44ed167d 1360 rcu_read_unlock();
47ff2d0a 1361
daeda1cc 1362 bdev = blkdev_get_by_path(new_disk_conf->backing_dev,
d4d77629 1363 FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev);
e525fd89 1364 if (IS_ERR(bdev)) {
daeda1cc 1365 dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev,
e525fd89 1366 PTR_ERR(bdev));
b411b363
PR		 1367 retcode = ERR_OPEN_DISK;
1368 goto fail;
1369 }
e525fd89
TH		 1370 nbc->backing_bdev = bdev;
1371
1372 /*
1373 * meta_dev_idx >= 0: external fixed size, possibly multiple
1374 * drbd sharing one meta device. TODO in that case, paranoia
1375 * check that [md_bdev, meta_dev_idx] is not yet used by some
1376 * other drbd minor! (if you use drbd.conf + drbdadm, that
1377 * should check it for you already; but if you don't, or
1378 * someone fooled it, we need to double check here)
1379 */
daeda1cc 1380 bdev = blkdev_get_by_path(new_disk_conf->meta_dev,
d4d77629 1381 FMODE_READ | FMODE_WRITE | FMODE_EXCL,
daeda1cc 1382 (new_disk_conf->meta_dev_idx < 0) ?
d4d77629 1383 (void *)mdev : (void *)drbd_m_holder);
e525fd89 1384 if (IS_ERR(bdev)) {
daeda1cc 1385 dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev,
e525fd89 1386 PTR_ERR(bdev));
b411b363
PR		 1387 retcode = ERR_OPEN_MD_DISK;
1388 goto fail;
1389 }
e525fd89 1390 nbc->md_bdev = bdev;
b411b363 1391
e525fd89 1392 if ((nbc->backing_bdev == nbc->md_bdev) !=
daeda1cc
PR		 1393 (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1394 new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) {
e525fd89 1395 retcode = ERR_MD_IDX_INVALID;
b411b363
PR		 1396 goto fail;
1397 }
1398
1399 resync_lru = lc_create("resync", drbd_bm_ext_cache,
46a15bc3 1400 1, 61, sizeof(struct bm_extent),
b411b363
PR		 1401 offsetof(struct bm_extent, lce));
1402 if (!resync_lru) {
1403 retcode = ERR_NOMEM;
e525fd89 1404 goto fail;
b411b363
PR		 1405 }
1406
1407 /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */
1408 drbd_md_set_sector_offsets(mdev, nbc);
1409
daeda1cc 1410 if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) {
b411b363
PR		 1411 dev_err(DEV, "max capacity %llu smaller than disk size %llu\n",
1412 (unsigned long long) drbd_get_max_capacity(nbc),
daeda1cc 1413 (unsigned long long) new_disk_conf->disk_size);
67b58bf7 1414 retcode = ERR_DISK_TOO_SMALL;
e525fd89 1415 goto fail;
b411b363
PR		 1416 }
1417
daeda1cc 1418 if (new_disk_conf->meta_dev_idx < 0) {
b411b363
PR		 1419 max_possible_sectors = DRBD_MAX_SECTORS_FLEX;
1420 /* at least one MB, otherwise it does not make sense */
1421 min_md_device_sectors = (2<<10);
1422 } else {
1423 max_possible_sectors = DRBD_MAX_SECTORS;
daeda1cc 1424 min_md_device_sectors = MD_RESERVED_SECT * (new_disk_conf->meta_dev_idx + 1);
b411b363
PR		 1425 }
1426
b411b363 1427 if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) {
67b58bf7 1428 retcode = ERR_MD_DISK_TOO_SMALL;
b411b363
PR		 1429 dev_warn(DEV, "refusing attach: md-device too small, "
1430 "at least %llu sectors needed for this meta-disk type\n",
1431 (unsigned long long) min_md_device_sectors);
e525fd89 1432 goto fail;
b411b363
PR		 1433 }
1434
1435 /* Make sure the new disk is big enough
1436 * (we may currently be R_PRIMARY with no local disk...) */
1437 if (drbd_get_max_capacity(nbc) <
1438 drbd_get_capacity(mdev->this_bdev)) {
67b58bf7 1439 retcode = ERR_DISK_TOO_SMALL;
e525fd89 1440 goto fail;
b411b363
PR		 1441 }
1442
1443 nbc->known_size = drbd_get_capacity(nbc->backing_bdev);
1444
1352994b
LE		 1445 if (nbc->known_size > max_possible_sectors) {
1446 dev_warn(DEV, "==> truncating very big lower level device "
1447 "to currently maximum possible %llu sectors <==\n",
1448 (unsigned long long) max_possible_sectors);
daeda1cc 1449 if (new_disk_conf->meta_dev_idx >= 0)
1352994b
LE		 1450 dev_warn(DEV, "==>> using internal or flexible "
1451 "meta data may help <<==\n");
1452 }
1453
b411b363
PR		 1454 drbd_suspend_io(mdev);
1455 /* also wait for the last barrier ack. */
b6dd1a89
LE		 1456 /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171
1457 * We need a way to either ignore barrier acks for barriers sent before a device
1458 * was attached, or a way to wait for all pending barrier acks to come in.
1459 * As barriers are counted per resource,
1460 * we'd need to suspend io on all devices of a resource.
1461 */
2aebfabb 1462 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev));
b411b363 1463 /* and for any other previously queued work */
a21e9298 1464 drbd_flush_workqueue(mdev);
b411b363 1465
f2024e7c
AG		 1466 rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE);
1467 retcode = rv; /* FIXME: Type mismatch. */
b411b363 1468 drbd_resume_io(mdev);
f2024e7c 1469 if (rv < SS_SUCCESS)
e525fd89 1470 goto fail;
b411b363
PR		 1471
1472 if (!get_ldev_if_state(mdev, D_ATTACHING))
1473 goto force_diskless;
1474
1475 drbd_md_set_sector_offsets(mdev, nbc);
1476
1477 if (!mdev->bitmap) {
1478 if (drbd_bm_init(mdev)) {
1479 retcode = ERR_NOMEM;
1480 goto force_diskless_dec;
1481 }
1482 }
1483
1484 retcode = drbd_md_read(mdev, nbc);
1485 if (retcode != NO_ERROR)
1486 goto force_diskless_dec;
1487
1488 if (mdev->state.conn < C_CONNECTED &&
1489 mdev->state.role == R_PRIMARY &&
1490 (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) {
1491 dev_err(DEV, "Can only attach to data with current UUID=%016llX\n",
1492 (unsigned long long)mdev->ed_uuid);
1493 retcode = ERR_DATA_NOT_CURRENT;
1494 goto force_diskless_dec;
1495 }
1496
1497 /* Since we are diskless, fix the activity log first... */
daeda1cc 1498 if (drbd_check_al_size(mdev, new_disk_conf)) {
b411b363
PR		 1499 retcode = ERR_NOMEM;
1500 goto force_diskless_dec;
1501 }
1502
1503 /* Prevent shrinking of consistent devices ! */
1504 if (drbd_md_test_flag(nbc, MDF_CONSISTENT) &&
daeda1cc 1505 drbd_new_dev_size(mdev, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) {
b411b363 1506 dev_warn(DEV, "refusing to truncate a consistent device\n");
67b58bf7 1507 retcode = ERR_DISK_TOO_SMALL;
b411b363
PR		 1508 goto force_diskless_dec;
1509 }
1510
b411b363
PR		 1511 /* Reset the "barriers don't work" bits here, then force meta data to
1512 * be written, to ensure we determine if barriers are supported. */
e544046a 1513 if (new_disk_conf->md_flushes)
a8a4e51e 1514 clear_bit(MD_NO_FUA, &mdev->flags);
e544046a
AG		 1515 else
1516 set_bit(MD_NO_FUA, &mdev->flags);
b411b363
PR		 1517
1518 /* Point of no return reached.
1519 * Devices and memory are no longer released by error cleanup below.
1520 * now mdev takes over responsibility, and the state engine should
1521 * clean it up somewhere. */
1522 D_ASSERT(mdev->ldev == NULL);
1523 mdev->ldev = nbc;
1524 mdev->resync = resync_lru;
9958c857 1525 mdev->rs_plan_s = new_plan;
b411b363
PR		 1526 nbc = NULL;
1527 resync_lru = NULL;
daeda1cc 1528 new_disk_conf = NULL;
9958c857 1529 new_plan = NULL;
b411b363 1530
4b0007c0 1531 drbd_bump_write_ordering(mdev->tconn, WO_bdev_flush);
b411b363
PR		 1532
1533 if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY))
1534 set_bit(CRASHED_PRIMARY, &mdev->flags);
1535 else
1536 clear_bit(CRASHED_PRIMARY, &mdev->flags);
1537
894c6a94 1538 if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
d5d7ebd4 1539 !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod))
b411b363 1540 set_bit(CRASHED_PRIMARY, &mdev->flags);
b411b363
PR		 1541
1542 mdev->send_cnt = 0;
1543 mdev->recv_cnt = 0;
1544 mdev->read_cnt = 0;
1545 mdev->writ_cnt = 0;
1546
99432fcc 1547 drbd_reconsider_max_bio_size(mdev);
b411b363
PR		 1548
1549 /* If I am currently not R_PRIMARY,
1550 * but meta data primary indicator is set,
1551 * I just now recover from a hard crash,
1552 * and have been R_PRIMARY before that crash.
1553 *
1554 * Now, if I had no connection before that crash
1555 * (have been degraded R_PRIMARY), chances are that
1556 * I won't find my peer now either.
1557 *
1558 * In that case, and _only_ in that case,
1559 * we use the degr-wfc-timeout instead of the default,
1560 * so we can automatically recover from a crash of a
1561 * degraded but active "cluster" after a certain timeout.
1562 */
1563 clear_bit(USE_DEGR_WFC_T, &mdev->flags);
1564 if (mdev->state.role != R_PRIMARY &&
1565 drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
1566 !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND))
1567 set_bit(USE_DEGR_WFC_T, &mdev->flags);
1568
24c4830c 1569 dd = drbd_determine_dev_size(mdev, 0);
b411b363
PR		 1570 if (dd == dev_size_error) {
1571 retcode = ERR_NOMEM_BITMAP;
1572 goto force_diskless_dec;
1573 } else if (dd == grew)
1574 set_bit(RESYNC_AFTER_NEG, &mdev->flags);
1575
9a51ab1c
PR		 1576 if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC) ||
1577 (test_bit(CRASHED_PRIMARY, &mdev->flags) &&
1578 drbd_md_test_flag(mdev->ldev, MDF_AL_DISABLED))) {
b411b363
PR		 1579 dev_info(DEV, "Assuming that all blocks are out of sync "
1580 "(aka FullSync)\n");
20ceb2b2
LE		 1581 if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write,
1582 "set_n_write from attaching", BM_LOCKED_MASK)) {
b411b363
PR		 1583 retcode = ERR_IO_MD_DISK;
1584 goto force_diskless_dec;
1585 }
1586 } else {
20ceb2b2 1587 if (drbd_bitmap_io(mdev, &drbd_bm_read,
22ab6a30 1588 "read from attaching", BM_LOCKED_MASK)) {
b411b363
PR		 1589 retcode = ERR_IO_MD_DISK;
1590 goto force_diskless_dec;
1591 }
1592 }
1593
0778286a
PR		 1594 if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev))
1595 drbd_suspend_al(mdev); /* IO is still suspended here... */
1596
87eeee41 1597 spin_lock_irq(&mdev->tconn->req_lock);
78bae59b
PR		 1598 os = drbd_read_state(mdev);
1599 ns = os;
b411b363
PR		 1600 /* If MDF_CONSISTENT is not set go into inconsistent state,
1601 otherwise investigate MDF_WasUpToDate...
1602 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1603 otherwise into D_CONSISTENT state.
1604 */
1605 if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) {
1606 if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE))
1607 ns.disk = D_CONSISTENT;
1608 else
1609 ns.disk = D_OUTDATED;
1610 } else {
1611 ns.disk = D_INCONSISTENT;
1612 }
1613
1614 if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED))
1615 ns.pdsk = D_OUTDATED;
1616
daeda1cc
PR		 1617 rcu_read_lock();
1618 if (ns.disk == D_CONSISTENT &&
1619 (ns.pdsk == D_OUTDATED || rcu_dereference(mdev->ldev->disk_conf)->fencing == FP_DONT_CARE))
b411b363
PR		 1620 ns.disk = D_UP_TO_DATE;
1621
1622 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1623 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1624 this point, because drbd_request_state() modifies these
1625 flags. */
1626
9a51ab1c
PR		 1627 if (rcu_dereference(mdev->ldev->disk_conf)->al_updates)
1628 mdev->ldev->md.flags &= MDF_AL_DISABLED;
1629 else
1630 mdev->ldev->md.flags |= MDF_AL_DISABLED;
1631
1632 rcu_read_unlock();
1633
b411b363
PR		 1634 /* In case we are C_CONNECTED postpone any decision on the new disk
1635 state after the negotiation phase. */
1636 if (mdev->state.conn == C_CONNECTED) {
1637 mdev->new_state_tmp.i = ns.i;
1638 ns.i = os.i;
1639 ns.disk = D_NEGOTIATING;
dc66c74d
PR		 1640
1641 /* We expect to receive up-to-date UUIDs soon.
1642 To avoid a race in receive_state, free p_uuid while
1643 holding req_lock. I.e. atomic with the state change */
1644 kfree(mdev->p_uuid);
1645 mdev->p_uuid = NULL;
b411b363
PR		 1646 }
1647
1648 rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
87eeee41 1649 spin_unlock_irq(&mdev->tconn->req_lock);
b411b363
PR		 1650
1651 if (rv < SS_SUCCESS)
1652 goto force_diskless_dec;
1653
cdfda633
PR		 1654 mod_timer(&mdev->request_timer, jiffies + HZ);
1655
b411b363
PR		 1656 if (mdev->state.role == R_PRIMARY)
1657 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
1658 else
1659 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
1660
1661 drbd_md_mark_dirty(mdev);
1662 drbd_md_sync(mdev);
1663
1664 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
1665 put_ldev(mdev);
0e29d163 1666 conn_reconfig_done(mdev->tconn);
3b98c0c2 1667 drbd_adm_finish(info, retcode);
b411b363
PR		 1668 return 0;
1669
1670 force_diskless_dec:
1671 put_ldev(mdev);
1672 force_diskless:
9510b241 1673 drbd_force_state(mdev, NS(disk, D_DISKLESS));
b411b363 1674 drbd_md_sync(mdev);
b411b363 1675 fail:
40cbf085 1676 conn_reconfig_done(mdev->tconn);
b411b363 1677 if (nbc) {
e525fd89
TH		 1678 if (nbc->backing_bdev)
1679 blkdev_put(nbc->backing_bdev,
1680 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
1681 if (nbc->md_bdev)
1682 blkdev_put(nbc->md_bdev,
1683 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
b411b363
PR		 1684 kfree(nbc);
1685 }
daeda1cc 1686 kfree(new_disk_conf);
b411b363 1687 lc_destroy(resync_lru);
9958c857 1688 kfree(new_plan);
b411b363 1689
40cbf085 1690 finish:
3b98c0c2 1691 drbd_adm_finish(info, retcode);
b411b363
PR		 1692 return 0;
1693}
1694
cdfda633 1695static int adm_detach(struct drbd_conf *mdev, int force)
85f75dd7 1696{
19f83c76 1697 enum drbd_state_rv retcode;
009ba89d 1698 int ret;
cdfda633
PR		 1699
1700 if (force) {
0c849666 1701 set_bit(FORCE_DETACH, &mdev->flags);
cdfda633
PR		 1702 drbd_force_state(mdev, NS(disk, D_FAILED));
1703 retcode = SS_SUCCESS;
1704 goto out;
1705 }
1706
85f75dd7 1707 drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */
0cfac5dd 1708 drbd_md_get_buffer(mdev); /* make sure there is no in-flight meta-data IO */
009ba89d 1709 retcode = drbd_request_state(mdev, NS(disk, D_FAILED));
0cfac5dd 1710 drbd_md_put_buffer(mdev);
009ba89d
LE		 1711 /* D_FAILED will transition to DISKLESS. */
1712 ret = wait_event_interruptible(mdev->misc_wait,
1713 mdev->state.disk != D_FAILED);
85f75dd7 1714 drbd_resume_io(mdev);
009ba89d
LE		 1715 if ((int)retcode == (int)SS_IS_DISKLESS)
1716 retcode = SS_NOTHING_TO_DO;
1717 if (ret)
1718 retcode = ERR_INTR;
cdfda633 1719out:
85f75dd7
LE		 1720 return retcode;
1721}
1722
82f59cc6
LE		 1723/* Detaching the disk is a process in multiple stages. First we need to lock
1724 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1725 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1726 * internal references as well.
1727 * Only then we have finally detached. */
3b98c0c2 1728int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info)
b411b363 1729{
9a0d9d03 1730 enum drbd_ret_code retcode;
cdfda633
PR		 1731 struct detach_parms parms = { };
1732 int err;
3b98c0c2
LE		 1733
1734 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1735 if (!adm_ctx.reply_skb)
1736 return retcode;
1737 if (retcode != NO_ERROR)
1738 goto out;
1739
cdfda633
PR		 1740 if (info->attrs[DRBD_NLA_DETACH_PARMS]) {
1741 err = detach_parms_from_attrs(&parms, info);
1742 if (err) {
1743 retcode = ERR_MANDATORY_TAG;
1744 drbd_msg_put_info(from_attrs_err_to_txt(err));
1745 goto out;
1746 }
1747 }
1748
1749 retcode = adm_detach(adm_ctx.mdev, parms.force_detach);
3b98c0c2
LE		 1750out:
1751 drbd_adm_finish(info, retcode);
b411b363
PR		 1752 return 0;
1753}
1754
f399002e
LE		 1755static bool conn_resync_running(struct drbd_tconn *tconn)
1756{
1757 struct drbd_conf *mdev;
695d08fa 1758 bool rv = false;
f399002e
LE		 1759 int vnr;
1760
695d08fa 1761 rcu_read_lock();
f399002e
LE		 1762 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1763 if (mdev->state.conn == C_SYNC_SOURCE ||
1764 mdev->state.conn == C_SYNC_TARGET ||
1765 mdev->state.conn == C_PAUSED_SYNC_S ||
695d08fa
PR		 1766 mdev->state.conn == C_PAUSED_SYNC_T) {
1767 rv = true;
1768 break;
1769 }
f399002e 1770 }
695d08fa
PR		 1771 rcu_read_unlock();
1772
1773 return rv;
f399002e
LE		 1774}
1775
1776static bool conn_ov_running(struct drbd_tconn *tconn)
1777{
1778 struct drbd_conf *mdev;
695d08fa 1779 bool rv = false;
f399002e
LE		 1780 int vnr;
1781
695d08fa 1782 rcu_read_lock();
f399002e
LE		 1783 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1784 if (mdev->state.conn == C_VERIFY_S ||
695d08fa
PR		 1785 mdev->state.conn == C_VERIFY_T) {
1786 rv = true;
1787 break;
1788 }
f399002e 1789 }
695d08fa
PR		 1790 rcu_read_unlock();
1791
1792 return rv;
f399002e
LE		 1793}
1794
cd64397c 1795static enum drbd_ret_code
44ed167d 1796_check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf)
cd64397c
PR		 1797{
1798 struct drbd_conf *mdev;
1799 int i;
1800
dcb20d1a
PR		 1801 if (old_conf && tconn->cstate == C_WF_REPORT_PARAMS && tconn->agreed_pro_version < 100) {
1802 if (new_conf->wire_protocol != old_conf->wire_protocol)
1803 return ERR_NEED_APV_100;
1804
1805 if (new_conf->two_primaries != old_conf->two_primaries)
1806 return ERR_NEED_APV_100;
1807
1808 if (!new_conf->integrity_alg != !old_conf->integrity_alg)
1809 return ERR_NEED_APV_100;
1810
1811 if (strcmp(new_conf->integrity_alg, old_conf->integrity_alg))
1812 return ERR_NEED_APV_100;
1813 }
1814
1815 if (!new_conf->two_primaries &&
1816 conn_highest_role(tconn) == R_PRIMARY &&
1817 conn_highest_peer(tconn) == R_PRIMARY)
1818 return ERR_NEED_ALLOW_TWO_PRI;
b032b6fa 1819
cd64397c
PR		 1820 if (new_conf->two_primaries &&
1821 (new_conf->wire_protocol != DRBD_PROT_C))
1822 return ERR_NOT_PROTO_C;
1823
cd64397c
PR		 1824 idr_for_each_entry(&tconn->volumes, mdev, i) {
1825 if (get_ldev(mdev)) {
daeda1cc 1826 enum drbd_fencing_p fp = rcu_dereference(mdev->ldev->disk_conf)->fencing;
cd64397c 1827 put_ldev(mdev);
44ed167d 1828 if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH)
cd64397c 1829 return ERR_STONITH_AND_PROT_A;
cd64397c 1830 }
6139f60d 1831 if (mdev->state.role == R_PRIMARY && new_conf->discard_my_data)
cd64397c 1832 return ERR_DISCARD;
cd64397c 1833 }
cd64397c
PR		 1834
1835 if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A)
1836 return ERR_CONG_NOT_PROTO_A;
1837
1838 return NO_ERROR;
1839}
1840
44ed167d
PR		 1841static enum drbd_ret_code
1842check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf)
1843{
1844 static enum drbd_ret_code rv;
1845 struct drbd_conf *mdev;
1846 int i;
1847
1848 rcu_read_lock();
1849 rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf);
1850 rcu_read_unlock();
1851
1852 /* tconn->volumes protected by genl_lock() here */
1853 idr_for_each_entry(&tconn->volumes, mdev, i) {
1854 if (!mdev->bitmap) {
1855 if(drbd_bm_init(mdev))
1856 return ERR_NOMEM;
1857 }
1858 }
1859
1860 return rv;
1861}
1862
0fd0ea06
PR		 1863struct crypto {
1864 struct crypto_hash *verify_tfm;
1865 struct crypto_hash *csums_tfm;
1866 struct crypto_hash *cram_hmac_tfm;
8d412fc6 1867 struct crypto_hash *integrity_tfm;
0fd0ea06
PR		 1868};
1869
1870static int
4b6ad6d4 1871alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg)
0fd0ea06
PR		 1872{
1873 if (!tfm_name[0])
1874 return NO_ERROR;
1875
1876 *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC);
1877 if (IS_ERR(*tfm)) {
1878 *tfm = NULL;
1879 return err_alg;
1880 }
1881
0fd0ea06
PR		 1882 return NO_ERROR;
1883}
1884
1885static enum drbd_ret_code
1886alloc_crypto(struct crypto *crypto, struct net_conf *new_conf)
1887{
1888 char hmac_name[CRYPTO_MAX_ALG_NAME];
1889 enum drbd_ret_code rv;
0fd0ea06 1890
4b6ad6d4
AG		 1891 rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg,
1892 ERR_CSUMS_ALG);
0fd0ea06
PR		 1893 if (rv != NO_ERROR)
1894 return rv;
4b6ad6d4
AG		 1895 rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg,
1896 ERR_VERIFY_ALG);
0fd0ea06
PR		 1897 if (rv != NO_ERROR)
1898 return rv;
4b6ad6d4
AG		 1899 rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg,
1900 ERR_INTEGRITY_ALG);
0fd0ea06
PR		 1901 if (rv != NO_ERROR)
1902 return rv;
0fd0ea06
PR		 1903 if (new_conf->cram_hmac_alg[0] != 0) {
1904 snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)",
1905 new_conf->cram_hmac_alg);
1906
4b6ad6d4
AG		 1907 rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name,
1908 ERR_AUTH_ALG);
0fd0ea06 1909 }
0fd0ea06
PR		 1910
1911 return rv;
1912}
1913
1914static void free_crypto(struct crypto *crypto)
1915{
0fd0ea06 1916 crypto_free_hash(crypto->cram_hmac_tfm);
8d412fc6 1917 crypto_free_hash(crypto->integrity_tfm);
0fd0ea06
PR		 1918 crypto_free_hash(crypto->csums_tfm);
1919 crypto_free_hash(crypto->verify_tfm);
1920}
1921
f399002e
LE		 1922int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info)
1923{
1924 enum drbd_ret_code retcode;
1925 struct drbd_tconn *tconn;
44ed167d 1926 struct net_conf *old_conf, *new_conf = NULL;
f399002e
LE		 1927 int err;
1928 int ovr; /* online verify running */
1929 int rsr; /* re-sync running */
0fd0ea06 1930 struct crypto crypto = { };
f399002e 1931
089c075d 1932 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION);
f399002e
LE		 1933 if (!adm_ctx.reply_skb)
1934 return retcode;
1935 if (retcode != NO_ERROR)
1936 goto out;
1937
1938 tconn = adm_ctx.tconn;
1939
1940 new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
1941 if (!new_conf) {
1942 retcode = ERR_NOMEM;
1943 goto out;
1944 }
1945
44ed167d
PR		 1946 conn_reconfig_start(tconn);
1947
88104ca4 1948 mutex_lock(&tconn->data.mutex);
a0095508 1949 mutex_lock(&tconn->conf_update);
91fd4dad 1950 old_conf = tconn->net_conf;
44ed167d
PR		 1951
1952 if (!old_conf) {
f399002e
LE		 1953 drbd_msg_put_info("net conf missing, try connect");
1954 retcode = ERR_INVALID_REQUEST;
91fd4dad 1955 goto fail;
f399002e
LE		 1956 }
1957
44ed167d 1958 *new_conf = *old_conf;
5979e361 1959 if (should_set_defaults(info))
b966b5dd 1960 set_net_conf_defaults(new_conf);
f399002e 1961
f399002e 1962 err = net_conf_from_attrs_for_change(new_conf, info);
c75b9b10 1963 if (err && err != -ENOMSG) {
f399002e
LE		 1964 retcode = ERR_MANDATORY_TAG;
1965 drbd_msg_put_info(from_attrs_err_to_txt(err));
1966 goto fail;
1967 }
1968
cd64397c
PR		 1969 retcode = check_net_options(tconn, new_conf);
1970 if (retcode != NO_ERROR)
1971 goto fail;
1972
f399002e
LE		 1973 /* re-sync running */
1974 rsr = conn_resync_running(tconn);
0fd0ea06 1975 if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) {
f399002e 1976 retcode = ERR_CSUMS_RESYNC_RUNNING;
91fd4dad 1977 goto fail;
f399002e
LE		 1978 }
1979
f399002e
LE		 1980 /* online verify running */
1981 ovr = conn_ov_running(tconn);
0fd0ea06
PR		 1982 if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) {
1983 retcode = ERR_VERIFY_RUNNING;
1984 goto fail;
f399002e
LE		 1985 }
1986
0fd0ea06
PR		 1987 retcode = alloc_crypto(&crypto, new_conf);
1988 if (retcode != NO_ERROR)
1989 goto fail;
f399002e 1990
44ed167d 1991 rcu_assign_pointer(tconn->net_conf, new_conf);
f399002e
LE		 1992
1993 if (!rsr) {
1994 crypto_free_hash(tconn->csums_tfm);
0fd0ea06
PR		 1995 tconn->csums_tfm = crypto.csums_tfm;
1996 crypto.csums_tfm = NULL;
f399002e
LE		 1997 }
1998 if (!ovr) {
1999 crypto_free_hash(tconn->verify_tfm);
0fd0ea06
PR		 2000 tconn->verify_tfm = crypto.verify_tfm;
2001 crypto.verify_tfm = NULL;
f399002e
LE		 2002 }
2003
8d412fc6
AG		 2004 crypto_free_hash(tconn->integrity_tfm);
2005 tconn->integrity_tfm = crypto.integrity_tfm;
d659f2aa 2006 if (tconn->cstate >= C_WF_REPORT_PARAMS && tconn->agreed_pro_version >= 100)
88104ca4 2007 /* Do this without trying to take tconn->data.mutex again. */
d659f2aa 2008 __drbd_send_protocol(tconn, P_PROTOCOL_UPDATE);
0fd0ea06 2009
0fd0ea06
PR		 2010 crypto_free_hash(tconn->cram_hmac_tfm);
2011 tconn->cram_hmac_tfm = crypto.cram_hmac_tfm;
2012
a0095508 2013 mutex_unlock(&tconn->conf_update);
88104ca4 2014 mutex_unlock(&tconn->data.mutex);
91fd4dad
PR		 2015 synchronize_rcu();
2016 kfree(old_conf);
2017
f399002e
LE		 2018 if (tconn->cstate >= C_WF_REPORT_PARAMS)
2019 drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn)));
2020
91fd4dad
PR		 2021 goto done;
2022
f399002e 2023 fail:
a0095508 2024 mutex_unlock(&tconn->conf_update);
88104ca4 2025 mutex_unlock(&tconn->data.mutex);
0fd0ea06 2026 free_crypto(&crypto);
f399002e 2027 kfree(new_conf);
91fd4dad 2028 done:
f399002e
LE		 2029 conn_reconfig_done(tconn);
2030 out:
2031 drbd_adm_finish(info, retcode);
2032 return 0;
2033}
2034
3b98c0c2 2035int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info)
b411b363 2036{
3b98c0c2 2037 struct drbd_conf *mdev;
44ed167d 2038 struct net_conf *old_conf, *new_conf = NULL;
0fd0ea06 2039 struct crypto crypto = { };
3b98c0c2 2040 struct drbd_tconn *tconn;
3b98c0c2
LE		 2041 enum drbd_ret_code retcode;
2042 int i;
2043 int err;
b411b363 2044
44e52cfa 2045 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE);
089c075d 2046
3b98c0c2
LE		 2047 if (!adm_ctx.reply_skb)
2048 return retcode;
2049 if (retcode != NO_ERROR)
2050 goto out;
089c075d
AG		 2051 if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) {
2052 drbd_msg_put_info("connection endpoint(s) missing");
2053 retcode = ERR_INVALID_REQUEST;
2054 goto out;
2055 }
2056
2057 /* No need for _rcu here. All reconfiguration is
2058 * strictly serialized on genl_lock(). We are protected against
2059 * concurrent reconfiguration/addition/deletion */
2060 list_for_each_entry(tconn, &drbd_tconns, all_tconn) {
2061 if (nla_len(adm_ctx.my_addr) == tconn->my_addr_len &&
2062 !memcmp(nla_data(adm_ctx.my_addr), &tconn->my_addr, tconn->my_addr_len)) {
2063 retcode = ERR_LOCAL_ADDR;
2064 goto out;
2065 }
2066
2067 if (nla_len(adm_ctx.peer_addr) == tconn->peer_addr_len &&
2068 !memcmp(nla_data(adm_ctx.peer_addr), &tconn->peer_addr, tconn->peer_addr_len)) {
2069 retcode = ERR_PEER_ADDR;
2070 goto out;
2071 }
2072 }
3b98c0c2
LE		 2073
2074 tconn = adm_ctx.tconn;
80883197 2075 conn_reconfig_start(tconn);
b411b363 2076
80883197 2077 if (tconn->cstate > C_STANDALONE) {
b411b363
PR		 2078 retcode = ERR_NET_CONFIGURED;
2079 goto fail;
2080 }
2081
a209b4ae 2082 /* allocation not in the IO path, drbdsetup / netlink process context */
5979e361 2083 new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL);
b411b363
PR		 2084 if (!new_conf) {
2085 retcode = ERR_NOMEM;
2086 goto fail;
2087 }
2088
b966b5dd 2089 set_net_conf_defaults(new_conf);
f399002e
LE		 2090
2091 err = net_conf_from_attrs(new_conf, info);
25e40932 2092 if (err && err != -ENOMSG) {
b411b363 2093 retcode = ERR_MANDATORY_TAG;
3b98c0c2 2094 drbd_msg_put_info(from_attrs_err_to_txt(err));
b411b363
PR		 2095 goto fail;
2096 }
2097
cd64397c
PR		 2098 retcode = check_net_options(tconn, new_conf);
2099 if (retcode != NO_ERROR)
422028b1 2100 goto fail;
422028b1 2101
0fd0ea06
PR		 2102 retcode = alloc_crypto(&crypto, new_conf);
2103 if (retcode != NO_ERROR)
2104 goto fail;
b411b363 2105
b411b363
PR		 2106 ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0;
2107
80883197 2108 conn_flush_workqueue(tconn);
91fd4dad 2109
a0095508 2110 mutex_lock(&tconn->conf_update);
91fd4dad
PR		 2111 old_conf = tconn->net_conf;
2112 if (old_conf) {
b411b363 2113 retcode = ERR_NET_CONFIGURED;
a0095508 2114 mutex_unlock(&tconn->conf_update);
b411b363
PR		 2115 goto fail;
2116 }
44ed167d 2117 rcu_assign_pointer(tconn->net_conf, new_conf);
b411b363 2118
91fd4dad 2119 conn_free_crypto(tconn);
0fd0ea06 2120 tconn->cram_hmac_tfm = crypto.cram_hmac_tfm;
8d412fc6 2121 tconn->integrity_tfm = crypto.integrity_tfm;
0fd0ea06
PR		 2122 tconn->csums_tfm = crypto.csums_tfm;
2123 tconn->verify_tfm = crypto.verify_tfm;
b411b363 2124
089c075d
AG		 2125 tconn->my_addr_len = nla_len(adm_ctx.my_addr);
2126 memcpy(&tconn->my_addr, nla_data(adm_ctx.my_addr), tconn->my_addr_len);
2127 tconn->peer_addr_len = nla_len(adm_ctx.peer_addr);
2128 memcpy(&tconn->peer_addr, nla_data(adm_ctx.peer_addr), tconn->peer_addr_len);
2129
a0095508 2130 mutex_unlock(&tconn->conf_update);
91fd4dad 2131
695d08fa 2132 rcu_read_lock();
80883197
PR		 2133 idr_for_each_entry(&tconn->volumes, mdev, i) {
2134 mdev->send_cnt = 0;
2135 mdev->recv_cnt = 0;
80883197 2136 }
695d08fa 2137 rcu_read_unlock();
5ee743e9
LE		 2138
2139 retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE);
2140
80883197 2141 conn_reconfig_done(tconn);
3b98c0c2 2142 drbd_adm_finish(info, retcode);
b411b363
PR		 2143 return 0;
2144
2145fail:
0fd0ea06 2146 free_crypto(&crypto);
b411b363
PR		 2147 kfree(new_conf);
2148
80883197 2149 conn_reconfig_done(tconn);
3b98c0c2
LE		 2150out:
2151 drbd_adm_finish(info, retcode);
b411b363
PR		 2152 return 0;
2153}
2154
85f75dd7
LE		 2155static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force)
2156{
2157 enum drbd_state_rv rv;
85f75dd7 2158
f3dfa40a
LE		 2159 rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING),
2160 force ? CS_HARD : 0);
85f75dd7
LE		 2161
2162 switch (rv) {
2163 case SS_NOTHING_TO_DO:
f3dfa40a 2164 break;
85f75dd7
LE		 2165 case SS_ALREADY_STANDALONE:
2166 return SS_SUCCESS;
2167 case SS_PRIMARY_NOP:
2168 /* Our state checking code wants to see the peer outdated. */
2169 rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING,
f3dfa40a 2170 pdsk, D_OUTDATED), CS_VERBOSE);
85f75dd7
LE		 2171 break;
2172 case SS_CW_FAILED_BY_PEER:
2173 /* The peer probably wants to see us outdated. */
2174 rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING,
2175 disk, D_OUTDATED), 0);
2176 if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) {
f3dfa40a
LE		 2177 rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING),
2178 CS_HARD);
85f75dd7
LE		 2179 }
2180 break;
2181 default:;
2182 /* no special handling necessary */
2183 }
2184
f3dfa40a
LE		 2185 if (rv >= SS_SUCCESS) {
2186 enum drbd_state_rv rv2;
2187 /* No one else can reconfigure the network while I am here.
2188 * The state handling only uses drbd_thread_stop_nowait(),
2189 * we want to really wait here until the receiver is no more.
2190 */
2191 drbd_thread_stop(&adm_ctx.tconn->receiver);
2192
2193 /* Race breaker. This additional state change request may be
2194 * necessary, if this was a forced disconnect during a receiver
2195 * restart. We may have "killed" the receiver thread just
2196 * after drbdd_init() returned. Typically, we should be
2197 * C_STANDALONE already, now, and this becomes a no-op.
2198 */
2199 rv2 = conn_request_state(tconn, NS(conn, C_STANDALONE),
2200 CS_VERBOSE | CS_HARD);
2201 if (rv2 < SS_SUCCESS)
2202 conn_err(tconn,
2203 "unexpected rv2=%d in conn_try_disconnect()\n",
2204 rv2);
2205 }
85f75dd7
LE		 2206 return rv;
2207}
2208
3b98c0c2 2209int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info)
b411b363 2210{
3b98c0c2
LE		 2211 struct disconnect_parms parms;
2212 struct drbd_tconn *tconn;
85f75dd7 2213 enum drbd_state_rv rv;
3b98c0c2
LE		 2214 enum drbd_ret_code retcode;
2215 int err;
2561b9c1 2216
089c075d 2217 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION);
3b98c0c2
LE		 2218 if (!adm_ctx.reply_skb)
2219 return retcode;
2220 if (retcode != NO_ERROR)
2561b9c1 2221 goto fail;
3b98c0c2
LE		 2222
2223 tconn = adm_ctx.tconn;
2224 memset(&parms, 0, sizeof(parms));
2225 if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) {
f399002e 2226 err = disconnect_parms_from_attrs(&parms, info);
3b98c0c2
LE		 2227 if (err) {
2228 retcode = ERR_MANDATORY_TAG;
2229 drbd_msg_put_info(from_attrs_err_to_txt(err));
2230 goto fail;
2231 }
2561b9c1
PR		 2232 }
2233
85f75dd7
LE		 2234 rv = conn_try_disconnect(tconn, parms.force_disconnect);
2235 if (rv < SS_SUCCESS)
f3dfa40a
LE		 2236 retcode = rv; /* FIXME: Type mismatch. */
2237 else
2238 retcode = NO_ERROR;
b411b363 2239 fail:
3b98c0c2 2240 drbd_adm_finish(info, retcode);
b411b363
PR		 2241 return 0;
2242}
2243
2244void resync_after_online_grow(struct drbd_conf *mdev)
2245{
2246 int iass; /* I am sync source */
2247
2248 dev_info(DEV, "Resync of new storage after online grow\n");
2249 if (mdev->state.role != mdev->state.peer)
2250 iass = (mdev->state.role == R_PRIMARY);
2251 else
427c0434 2252 iass = test_bit(RESOLVE_CONFLICTS, &mdev->tconn->flags);
b411b363
PR		 2253
2254 if (iass)
2255 drbd_start_resync(mdev, C_SYNC_SOURCE);
2256 else
2257 _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE);
2258}
2259
3b98c0c2 2260int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
b411b363 2261{
daeda1cc 2262 struct disk_conf *old_disk_conf, *new_disk_conf = NULL;
3b98c0c2
LE		 2263 struct resize_parms rs;
2264 struct drbd_conf *mdev;
2265 enum drbd_ret_code retcode;
b411b363 2266 enum determine_dev_size dd;
6495d2c6 2267 enum dds_flags ddsf;
daeda1cc 2268 sector_t u_size;
3b98c0c2 2269 int err;
b411b363 2270
3b98c0c2
LE		 2271 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2272 if (!adm_ctx.reply_skb)
2273 return retcode;
2274 if (retcode != NO_ERROR)
b411b363 2275 goto fail;
3b98c0c2
LE		 2276
2277 memset(&rs, 0, sizeof(struct resize_parms));
2278 if (info->attrs[DRBD_NLA_RESIZE_PARMS]) {
f399002e 2279 err = resize_parms_from_attrs(&rs, info);
3b98c0c2
LE		 2280 if (err) {
2281 retcode = ERR_MANDATORY_TAG;
2282 drbd_msg_put_info(from_attrs_err_to_txt(err));
2283 goto fail;
2284 }
b411b363
PR		 2285 }
2286
3b98c0c2 2287 mdev = adm_ctx.mdev;
b411b363
PR		 2288 if (mdev->state.conn > C_CONNECTED) {
2289 retcode = ERR_RESIZE_RESYNC;
2290 goto fail;
2291 }
2292
2293 if (mdev->state.role == R_SECONDARY &&
2294 mdev->state.peer == R_SECONDARY) {
2295 retcode = ERR_NO_PRIMARY;
2296 goto fail;
2297 }
2298
2299 if (!get_ldev(mdev)) {
2300 retcode = ERR_NO_DISK;
2301 goto fail;
2302 }
2303
31890f4a 2304 if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) {
6495d2c6 2305 retcode = ERR_NEED_APV_93;
9bcd2521 2306 goto fail_ldev;
6495d2c6
PR		 2307 }
2308
daeda1cc
PR		 2309 rcu_read_lock();
2310 u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size;
2311 rcu_read_unlock();
2312 if (u_size != (sector_t)rs.resize_size) {
2313 new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL);
2314 if (!new_disk_conf) {
2315 retcode = ERR_NOMEM;
9bcd2521 2316 goto fail_ldev;
daeda1cc
PR		 2317 }
2318 }
2319
087c2492 2320 if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev))
b411b363 2321 mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
b411b363 2322
daeda1cc
PR		 2323 if (new_disk_conf) {
2324 mutex_lock(&mdev->tconn->conf_update);
2325 old_disk_conf = mdev->ldev->disk_conf;
2326 *new_disk_conf = *old_disk_conf;
2327 new_disk_conf->disk_size = (sector_t)rs.resize_size;
2328 rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf);
2329 mutex_unlock(&mdev->tconn->conf_update);
2330 synchronize_rcu();
2331 kfree(old_disk_conf);
2332 }
2333
6495d2c6 2334 ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0);
24c4830c 2335 dd = drbd_determine_dev_size(mdev, ddsf);
b411b363
PR		 2336 drbd_md_sync(mdev);
2337 put_ldev(mdev);
2338 if (dd == dev_size_error) {
2339 retcode = ERR_NOMEM_BITMAP;
2340 goto fail;
2341 }
2342
087c2492 2343 if (mdev->state.conn == C_CONNECTED) {
b411b363
PR		 2344 if (dd == grew)
2345 set_bit(RESIZE_PENDING, &mdev->flags);
2346
2347 drbd_send_uuids(mdev);
6495d2c6 2348 drbd_send_sizes(mdev, 1, ddsf);
b411b363
PR		 2349 }
2350
2351 fail:
3b98c0c2 2352 drbd_adm_finish(info, retcode);
b411b363 2353 return 0;
9bcd2521
PR		 2354
2355 fail_ldev:
2356 put_ldev(mdev);
2357 goto fail;
b411b363
PR		 2358}
2359
f399002e 2360int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info)
b411b363 2361{
3b98c0c2 2362 enum drbd_ret_code retcode;
f399002e 2363 struct drbd_tconn *tconn;
b57a1e27 2364 struct res_opts res_opts;
f399002e 2365 int err;
b411b363 2366
44e52cfa 2367 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE);
3b98c0c2
LE		 2368 if (!adm_ctx.reply_skb)
2369 return retcode;
2370 if (retcode != NO_ERROR)
2371 goto fail;
f399002e 2372 tconn = adm_ctx.tconn;
3b98c0c2 2373
b57a1e27 2374 res_opts = tconn->res_opts;
5979e361 2375 if (should_set_defaults(info))
b966b5dd 2376 set_res_opts_defaults(&res_opts);
b411b363 2377
b57a1e27 2378 err = res_opts_from_attrs(&res_opts, info);
c75b9b10 2379 if (err && err != -ENOMSG) {
b411b363 2380 retcode = ERR_MANDATORY_TAG;
3b98c0c2 2381 drbd_msg_put_info(from_attrs_err_to_txt(err));
b411b363
PR		 2382 goto fail;
2383 }
2384
afbbfa88
AG		 2385 err = set_resource_options(tconn, &res_opts);
2386 if (err) {
2387 retcode = ERR_INVALID_REQUEST;
2388 if (err == -ENOMEM)
2389 retcode = ERR_NOMEM;
b411b363
PR		 2390 }
2391
b411b363 2392fail:
3b98c0c2 2393 drbd_adm_finish(info, retcode);
b411b363
PR		 2394 return 0;
2395}
2396
3b98c0c2 2397int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info)
b411b363 2398{
3b98c0c2
LE		 2399 struct drbd_conf *mdev;
2400 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2401
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;
2407
2408 mdev = adm_ctx.mdev;
b411b363 2409
194bfb32 2410 /* If there is still bitmap IO pending, probably because of a previous
970fbde1
LE		 2411 * resync just being finished, wait for it before requesting a new resync.
2412 * Also wait for it's after_state_ch(). */
5016b82a 2413 drbd_suspend_io(mdev);
194bfb32 2414 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
970fbde1 2415 drbd_flush_workqueue(mdev);
194bfb32 2416
b411b363
PR		 2417 retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED);
2418
2419 if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION)
2420 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
2421
2422 while (retcode == SS_NEED_CONNECTION) {
87eeee41 2423 spin_lock_irq(&mdev->tconn->req_lock);
b411b363
PR		 2424 if (mdev->state.conn < C_CONNECTED)
2425 retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL);
87eeee41 2426 spin_unlock_irq(&mdev->tconn->req_lock);
b411b363
PR		 2427
2428 if (retcode != SS_NEED_CONNECTION)
2429 break;
2430
2431 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
2432 }
5016b82a 2433 drbd_resume_io(mdev);
b411b363 2434
3b98c0c2
LE		 2435out:
2436 drbd_adm_finish(info, retcode);
b411b363
PR		 2437 return 0;
2438}
2439
3b98c0c2
LE		 2440static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info,
2441 union drbd_state mask, union drbd_state val)
b411b363 2442{
3b98c0c2 2443 enum drbd_ret_code retcode;
194bfb32 2444
3b98c0c2
LE		 2445 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2446 if (!adm_ctx.reply_skb)
2447 return retcode;
2448 if (retcode != NO_ERROR)
2449 goto out;
b411b363 2450
3b98c0c2
LE		 2451 retcode = drbd_request_state(adm_ctx.mdev, mask, val);
2452out:
2453 drbd_adm_finish(info, retcode);
b411b363
PR		 2454 return 0;
2455}
2456
25b0d6c8
PR		 2457static int drbd_bmio_set_susp_al(struct drbd_conf *mdev)
2458{
2459 int rv;
2460
2461 rv = drbd_bmio_set_n_write(mdev);
2462 drbd_suspend_al(mdev);
2463 return rv;
2464}
2465
3b98c0c2 2466int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info)
b411b363 2467{
25b0d6c8
PR		 2468 int retcode; /* drbd_ret_code, drbd_state_rv */
2469 struct drbd_conf *mdev;
2470
2471 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2472 if (!adm_ctx.reply_skb)
2473 return retcode;
2474 if (retcode != NO_ERROR)
2475 goto out;
2476
2477 mdev = adm_ctx.mdev;
2478
5016b82a 2479 /* If there is still bitmap IO pending, probably because of a previous
970fbde1
LE		 2480 * resync just being finished, wait for it before requesting a new resync.
2481 * Also wait for it's after_state_ch(). */
5016b82a
LE		 2482 drbd_suspend_io(mdev);
2483 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
970fbde1 2484 drbd_flush_workqueue(mdev);
5016b82a 2485
25b0d6c8
PR		 2486 retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED);
2487 if (retcode < SS_SUCCESS) {
2488 if (retcode == SS_NEED_CONNECTION && mdev->state.role == R_PRIMARY) {
2489 /* The peer will get a resync upon connect anyways.
2490 * Just make that into a full resync. */
2491 retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT));
2492 if (retcode >= SS_SUCCESS) {
2493 if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al,
2494 "set_n_write from invalidate_peer",
2495 BM_LOCKED_SET_ALLOWED))
2496 retcode = ERR_IO_MD_DISK;
2497 }
2498 } else
2499 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S));
2500 }
5016b82a 2501 drbd_resume_io(mdev);
25b0d6c8
PR		 2502
2503out:
2504 drbd_adm_finish(info, retcode);
2505 return 0;
3b98c0c2 2506}
b411b363 2507
3b98c0c2
LE		 2508int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info)
2509{
2510 enum drbd_ret_code retcode;
2511
2512 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2513 if (!adm_ctx.reply_skb)
2514 return retcode;
2515 if (retcode != NO_ERROR)
2516 goto out;
b411b363 2517
3b98c0c2
LE		 2518 if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO)
2519 retcode = ERR_PAUSE_IS_SET;
2520out:
2521 drbd_adm_finish(info, retcode);
b411b363
PR		 2522 return 0;
2523}
2524
3b98c0c2 2525int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info)
b411b363 2526{
da9fbc27 2527 union drbd_dev_state s;
3b98c0c2
LE		 2528 enum drbd_ret_code retcode;
2529
2530 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2531 if (!adm_ctx.reply_skb)
2532 return retcode;
2533 if (retcode != NO_ERROR)
2534 goto out;
b411b363 2535
3b98c0c2
LE		 2536 if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) {
2537 s = adm_ctx.mdev->state;
cd88d030
PR		 2538 if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) {
2539 retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP :
2540 s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR;
2541 } else {
2542 retcode = ERR_PAUSE_IS_CLEAR;
2543 }
2544 }
b411b363 2545
3b98c0c2
LE		 2546out:
2547 drbd_adm_finish(info, retcode);
b411b363
PR		 2548 return 0;
2549}
2550
3b98c0c2 2551int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info)
b411b363 2552{
3b98c0c2 2553 return drbd_adm_simple_request_state(skb, info, NS(susp, 1));
b411b363
PR		 2554}
2555
3b98c0c2 2556int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info)
b411b363 2557{
3b98c0c2
LE		 2558 struct drbd_conf *mdev;
2559 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2560
2561 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2562 if (!adm_ctx.reply_skb)
2563 return retcode;
2564 if (retcode != NO_ERROR)
2565 goto out;
2566
2567 mdev = adm_ctx.mdev;
43a5182c
PR		 2568 if (test_bit(NEW_CUR_UUID, &mdev->flags)) {
2569 drbd_uuid_new_current(mdev);
2570 clear_bit(NEW_CUR_UUID, &mdev->flags);
43a5182c 2571 }
265be2d0 2572 drbd_suspend_io(mdev);
3b98c0c2
LE		 2573 retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0));
2574 if (retcode == SS_SUCCESS) {
265be2d0 2575 if (mdev->state.conn < C_CONNECTED)
2f5cdd0b 2576 tl_clear(mdev->tconn);
265be2d0 2577 if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED)
2f5cdd0b 2578 tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO);
265be2d0
PR		 2579 }
2580 drbd_resume_io(mdev);
2581
3b98c0c2
LE		 2582out:
2583 drbd_adm_finish(info, retcode);
b411b363
PR		 2584 return 0;
2585}
2586
3b98c0c2 2587int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info)
b411b363 2588{
3b98c0c2 2589 return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED));
b411b363
PR		 2590}
2591
089c075d 2592int nla_put_drbd_cfg_context(struct sk_buff *skb, struct drbd_tconn *tconn, unsigned vnr)
543cc10b
LE		 2593{
2594 struct nlattr *nla;
2595 nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT);
2596 if (!nla)
2597 goto nla_put_failure;
26ec9287
AG		 2598 if (vnr != VOLUME_UNSPECIFIED &&
2599 nla_put_u32(skb, T_ctx_volume, vnr))
2600 goto nla_put_failure;
2601 if (nla_put_string(skb, T_ctx_resource_name, tconn->name))
2602 goto nla_put_failure;
2603 if (tconn->my_addr_len &&
2604 nla_put(skb, T_ctx_my_addr, tconn->my_addr_len, &tconn->my_addr))
2605 goto nla_put_failure;
2606 if (tconn->peer_addr_len &&
2607 nla_put(skb, T_ctx_peer_addr, tconn->peer_addr_len, &tconn->peer_addr))
2608 goto nla_put_failure;
543cc10b
LE		 2609 nla_nest_end(skb, nla);
2610 return 0;
2611
2612nla_put_failure:
2613 if (nla)
2614 nla_nest_cancel(skb, nla);
2615 return -EMSGSIZE;
2616}
2617
3b98c0c2
LE		 2618int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev,
2619 const struct sib_info *sib)
b411b363 2620{
3b98c0c2 2621 struct state_info *si = NULL; /* for sizeof(si->member); */
44ed167d 2622 struct net_conf *nc;
3b98c0c2
LE		 2623 struct nlattr *nla;
2624 int got_ldev;
3b98c0c2
LE		 2625 int err = 0;
2626 int exclude_sensitive;
2627
2628 /* If sib != NULL, this is drbd_bcast_event, which anyone can listen
2629 * to. So we better exclude_sensitive information.
2630 *
2631 * If sib == NULL, this is drbd_adm_get_status, executed synchronously
2632 * in the context of the requesting user process. Exclude sensitive
2633 * information, unless current has superuser.
2634 *
2635 * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and
2636 * relies on the current implementation of netlink_dump(), which
2637 * executes the dump callback successively from netlink_recvmsg(),
2638 * always in the context of the receiving process */
2639 exclude_sensitive = sib || !capable(CAP_SYS_ADMIN);
2640
2641 got_ldev = get_ldev(mdev);
3b98c0c2
LE		 2642
2643 /* We need to add connection name and volume number information still.
2644 * Minor number is in drbd_genlmsghdr. */
089c075d 2645 if (nla_put_drbd_cfg_context(skb, mdev->tconn, mdev->vnr))
3b98c0c2 2646 goto nla_put_failure;
3b98c0c2 2647
f399002e
LE		 2648 if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive))
2649 goto nla_put_failure;
2650
daeda1cc 2651 rcu_read_lock();
3b98c0c2 2652 if (got_ldev)
daeda1cc 2653 if (disk_conf_to_skb(skb, rcu_dereference(mdev->ldev->disk_conf), exclude_sensitive))
3b98c0c2 2654 goto nla_put_failure;
44ed167d 2655
44ed167d
PR		 2656 nc = rcu_dereference(mdev->tconn->net_conf);
2657 if (nc)
2658 err = net_conf_to_skb(skb, nc, exclude_sensitive);
2659 rcu_read_unlock();
2660 if (err)
2661 goto nla_put_failure;
3b98c0c2 2662
3b98c0c2
LE		 2663 nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO);
2664 if (!nla)
2665 goto nla_put_failure;
26ec9287
AG		 2666 if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) ||
2667 nla_put_u32(skb, T_current_state, mdev->state.i) ||
2668 nla_put_u64(skb, T_ed_uuid, mdev->ed_uuid) ||
3174f8c5
PM		 2669 nla_put_u64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev)) ||
2670 nla_put_u64(skb, T_send_cnt, mdev->send_cnt) ||
2671 nla_put_u64(skb, T_recv_cnt, mdev->recv_cnt) ||
2672 nla_put_u64(skb, T_read_cnt, mdev->read_cnt) ||
2673 nla_put_u64(skb, T_writ_cnt, mdev->writ_cnt) ||
2674 nla_put_u64(skb, T_al_writ_cnt, mdev->al_writ_cnt) ||
2675 nla_put_u64(skb, T_bm_writ_cnt, mdev->bm_writ_cnt) ||
2676 nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&mdev->ap_bio_cnt)) ||
2677 nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&mdev->ap_pending_cnt)) ||
2678 nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&mdev->rs_pending_cnt)))
26ec9287 2679 goto nla_put_failure;
3b98c0c2
LE		 2680
2681 if (got_ldev) {
26ec9287
AG		 2682 if (nla_put_u32(skb, T_disk_flags, mdev->ldev->md.flags) ||
2683 nla_put(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid) ||
2684 nla_put_u64(skb, T_bits_total, drbd_bm_bits(mdev)) ||
2685 nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(mdev)))
2686 goto nla_put_failure;
3b98c0c2
LE		 2687 if (C_SYNC_SOURCE <= mdev->state.conn &&
2688 C_PAUSED_SYNC_T >= mdev->state.conn) {
26ec9287
AG		 2689 if (nla_put_u64(skb, T_bits_rs_total, mdev->rs_total) ||
2690 nla_put_u64(skb, T_bits_rs_failed, mdev->rs_failed))
2691 goto nla_put_failure;
3b98c0c2 2692 }
b411b363
PR		 2693 }
2694
3b98c0c2
LE		 2695 if (sib) {
2696 switch(sib->sib_reason) {
2697 case SIB_SYNC_PROGRESS:
2698 case SIB_GET_STATUS_REPLY:
2699 break;
2700 case SIB_STATE_CHANGE:
26ec9287
AG		 2701 if (nla_put_u32(skb, T_prev_state, sib->os.i) ||
2702 nla_put_u32(skb, T_new_state, sib->ns.i))
2703 goto nla_put_failure;
3b98c0c2
LE		 2704 break;
2705 case SIB_HELPER_POST:
26ec9287
AG		 2706 if (nla_put_u32(skb, T_helper_exit_code,
2707 sib->helper_exit_code))
2708 goto nla_put_failure;
3b98c0c2
LE		 2709 /* fall through */
2710 case SIB_HELPER_PRE:
26ec9287
AG		 2711 if (nla_put_string(skb, T_helper, sib->helper_name))
2712 goto nla_put_failure;
3b98c0c2
LE		 2713 break;
2714 }
b411b363 2715 }
3b98c0c2 2716 nla_nest_end(skb, nla);
b411b363 2717
3b98c0c2
LE		 2718 if (0)
2719nla_put_failure:
2720 err = -EMSGSIZE;
2721 if (got_ldev)
2722 put_ldev(mdev);
3b98c0c2 2723 return err;
b411b363
PR		 2724}
2725
3b98c0c2 2726int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info)
b411b363 2727{
3b98c0c2
LE		 2728 enum drbd_ret_code retcode;
2729 int err;
b411b363 2730
3b98c0c2
LE		 2731 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2732 if (!adm_ctx.reply_skb)
2733 return retcode;
2734 if (retcode != NO_ERROR)
2735 goto out;
b411b363 2736
3b98c0c2
LE		 2737 err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL);
2738 if (err) {
2739 nlmsg_free(adm_ctx.reply_skb);
2740 return err;
b411b363 2741 }
3b98c0c2
LE		 2742out:
2743 drbd_adm_finish(info, retcode);
2744 return 0;
b411b363
PR		 2745}
2746
71932efc 2747int get_one_status(struct sk_buff *skb, struct netlink_callback *cb)
b411b363 2748{
3b98c0c2
LE		 2749 struct drbd_conf *mdev;
2750 struct drbd_genlmsghdr *dh;
543cc10b
LE		 2751 struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0];
2752 struct drbd_tconn *tconn = NULL;
2753 struct drbd_tconn *tmp;
2754 unsigned volume = cb->args[1];
2755
2756 /* Open coded, deferred, iteration:
2757 * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) {
2758 * idr_for_each_entry(&tconn->volumes, mdev, i) {
2759 * ...
2760 * }
2761 * }
2762 * where tconn is cb->args[0];
2763 * and i is cb->args[1];
2764 *
71932efc
LE		 2765 * cb->args[2] indicates if we shall loop over all resources,
2766 * or just dump all volumes of a single resource.
2767 *
3b98c0c2
LE		 2768 * This may miss entries inserted after this dump started,
2769 * or entries deleted before they are reached.
543cc10b
LE		 2770 *
2771 * We need to make sure the mdev won't disappear while
2772 * we are looking at it, and revalidate our iterators
2773 * on each iteration.
2774 */
3b98c0c2 2775
9dc9fbb3 2776 /* synchronize with conn_create()/conn_destroy() */
c141ebda 2777 rcu_read_lock();
543cc10b 2778 /* revalidate iterator position */
ec0bddbc 2779 list_for_each_entry_rcu(tmp, &drbd_tconns, all_tconn) {
543cc10b
LE		 2780 if (pos == NULL) {
2781 /* first iteration */
2782 pos = tmp;
2783 tconn = pos;
2784 break;
2785 }
2786 if (tmp == pos) {
2787 tconn = pos;
2788 break;
2789 }
2790 }
2791 if (tconn) {
71932efc 2792next_tconn:
543cc10b
LE		 2793 mdev = idr_get_next(&tconn->volumes, &volume);
2794 if (!mdev) {
2795 /* No more volumes to dump on this tconn.
2796 * Advance tconn iterator. */
ec0bddbc
PR		 2797 pos = list_entry_rcu(tconn->all_tconn.next,
2798 struct drbd_tconn, all_tconn);
71932efc 2799 /* Did we dump any volume on this tconn yet? */
543cc10b 2800 if (volume != 0) {
71932efc
LE		 2801 /* If we reached the end of the list,
2802 * or only a single resource dump was requested,
2803 * we are done. */
2804 if (&pos->all_tconn == &drbd_tconns || cb->args[2])
2805 goto out;
543cc10b 2806 volume = 0;
71932efc 2807 tconn = pos;
543cc10b
LE		 2808 goto next_tconn;
2809 }
2810 }
2811
3b98c0c2
LE		 2812 dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid,
2813 cb->nlh->nlmsg_seq, &drbd_genl_family,
2814 NLM_F_MULTI, DRBD_ADM_GET_STATUS);
2815 if (!dh)
543cc10b
LE		 2816 goto out;
2817
2818 if (!mdev) {
367d675d
LE		 2819 /* This is a tconn without a single volume.
2820 * Suprisingly enough, it may have a network
2821 * configuration. */
2822 struct net_conf *nc;
543cc10b
LE		 2823 dh->minor = -1U;
2824 dh->ret_code = NO_ERROR;
089c075d 2825 if (nla_put_drbd_cfg_context(skb, tconn, VOLUME_UNSPECIFIED))
367d675d
LE		 2826 goto cancel;
2827 nc = rcu_dereference(tconn->net_conf);
2828 if (nc && net_conf_to_skb(skb, nc, 1) != 0)
2829 goto cancel;
2830 goto done;
543cc10b 2831 }
3b98c0c2 2832
543cc10b
LE		 2833 D_ASSERT(mdev->vnr == volume);
2834 D_ASSERT(mdev->tconn == tconn);
3b98c0c2 2835
543cc10b 2836 dh->minor = mdev_to_minor(mdev);
3b98c0c2
LE		 2837 dh->ret_code = NO_ERROR;
2838
2839 if (nla_put_status_info(skb, mdev, NULL)) {
367d675d 2840cancel:
3b98c0c2 2841 genlmsg_cancel(skb, dh);
543cc10b 2842 goto out;
3b98c0c2 2843 }
367d675d 2844done:
3b98c0c2
LE		 2845 genlmsg_end(skb, dh);
2846 }
b411b363 2847
543cc10b 2848out:
c141ebda 2849 rcu_read_unlock();
543cc10b
LE		 2850 /* where to start the next iteration */
2851 cb->args[0] = (long)pos;
2852 cb->args[1] = (pos == tconn) ? volume + 1 : 0;
b411b363 2853
543cc10b
LE		 2854 /* No more tconns/volumes/minors found results in an empty skb.
2855 * Which will terminate the dump. */
3b98c0c2 2856 return skb->len;
b411b363
PR		 2857}
2858
71932efc
LE		 2859/*
2860 * Request status of all resources, or of all volumes within a single resource.
2861 *
2862 * This is a dump, as the answer may not fit in a single reply skb otherwise.
2863 * Which means we cannot use the family->attrbuf or other such members, because
2864 * dump is NOT protected by the genl_lock(). During dump, we only have access
2865 * to the incoming skb, and need to opencode "parsing" of the nlattr payload.
2866 *
2867 * Once things are setup properly, we call into get_one_status().
2868 */
2869int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb)
2870{
2871 const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ;
2872 struct nlattr *nla;
7c3063cc 2873 const char *resource_name;
71932efc 2874 struct drbd_tconn *tconn;
7c3063cc 2875 int maxtype;
71932efc
LE		 2876
2877 /* Is this a followup call? */
2878 if (cb->args[0]) {
2879 /* ... of a single resource dump,
2880 * and the resource iterator has been advanced already? */
2881 if (cb->args[2] && cb->args[2] != cb->args[0])
2882 return 0; /* DONE. */
2883 goto dump;
2884 }
2885
2886 /* First call (from netlink_dump_start). We need to figure out
2887 * which resource(s) the user wants us to dump. */
2888 nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen),
2889 nlmsg_attrlen(cb->nlh, hdrlen),
2890 DRBD_NLA_CFG_CONTEXT);
2891
2892 /* No explicit context given. Dump all. */
2893 if (!nla)
2894 goto dump;
7c3063cc
AG		 2895 maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1;
2896 nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name));
2897 if (IS_ERR(nla))
2898 return PTR_ERR(nla);
71932efc
LE		 2899 /* context given, but no name present? */
2900 if (!nla)
2901 return -EINVAL;
7c3063cc
AG		 2902 resource_name = nla_data(nla);
2903 tconn = conn_get_by_name(resource_name);
0ace9dfa 2904
71932efc
LE		 2905 if (!tconn)
2906 return -ENODEV;
2907
0ace9dfa
PR		 2908 kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */
2909
71932efc
LE		 2910 /* prime iterators, and set "filter" mode mark:
2911 * only dump this tconn. */
2912 cb->args[0] = (long)tconn;
2913 /* cb->args[1] = 0; passed in this way. */
2914 cb->args[2] = (long)tconn;
2915
2916dump:
2917 return get_one_status(skb, cb);
2918}
2919
3b98c0c2 2920int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info)
b411b363 2921{
3b98c0c2
LE		 2922 enum drbd_ret_code retcode;
2923 struct timeout_parms tp;
2924 int err;
b411b363 2925
3b98c0c2
LE		 2926 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2927 if (!adm_ctx.reply_skb)
2928 return retcode;
2929 if (retcode != NO_ERROR)
2930 goto out;
b411b363 2931
3b98c0c2
LE		 2932 tp.timeout_type =
2933 adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED :
2934 test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED :
2935 UT_DEFAULT;
b411b363 2936
3b98c0c2
LE		 2937 err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp);
2938 if (err) {
2939 nlmsg_free(adm_ctx.reply_skb);
2940 return err;
2941 }
2942out:
2943 drbd_adm_finish(info, retcode);
2944 return 0;
b411b363
PR		 2945}
2946
3b98c0c2 2947int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info)
b411b363 2948{
3b98c0c2
LE		 2949 struct drbd_conf *mdev;
2950 enum drbd_ret_code retcode;
58ffa580 2951 struct start_ov_parms parms;
b411b363 2952
3b98c0c2
LE		 2953 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2954 if (!adm_ctx.reply_skb)
2955 return retcode;
2956 if (retcode != NO_ERROR)
2957 goto out;
873b0d5f 2958
3b98c0c2 2959 mdev = adm_ctx.mdev;
58ffa580
LE		 2960
2961 /* resume from last known position, if possible */
2962 parms.ov_start_sector = mdev->ov_start_sector;
2963 parms.ov_stop_sector = ULLONG_MAX;
3b98c0c2 2964 if (info->attrs[DRBD_NLA_START_OV_PARMS]) {
f399002e 2965 int err = start_ov_parms_from_attrs(&parms, info);
3b98c0c2
LE		 2966 if (err) {
2967 retcode = ERR_MANDATORY_TAG;
2968 drbd_msg_put_info(from_attrs_err_to_txt(err));
2969 goto out;
2970 }
3b98c0c2 2971 }
58ffa580
LE		 2972 /* w_make_ov_request expects position to be aligned */
2973 mdev->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1);
2974 mdev->ov_stop_sector = parms.ov_stop_sector;
2975
873b0d5f
LE		 2976 /* If there is still bitmap IO pending, e.g. previous resync or verify
2977 * just being finished, wait for it before requesting a new resync. */
5016b82a 2978 drbd_suspend_io(mdev);
873b0d5f 2979 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
3b98c0c2 2980 retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
5016b82a 2981 drbd_resume_io(mdev);
3b98c0c2
LE		 2982out:
2983 drbd_adm_finish(info, retcode);
b411b363
PR		 2984 return 0;
2985}
2986
2987
3b98c0c2 2988int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info)
b411b363 2989{
3b98c0c2
LE		 2990 struct drbd_conf *mdev;
2991 enum drbd_ret_code retcode;
b411b363
PR		 2992 int skip_initial_sync = 0;
2993 int err;
3b98c0c2 2994 struct new_c_uuid_parms args;
b411b363 2995
3b98c0c2
LE		 2996 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2997 if (!adm_ctx.reply_skb)
2998 return retcode;
2999 if (retcode != NO_ERROR)
3000 goto out_nolock;
b411b363 3001
3b98c0c2
LE		 3002 mdev = adm_ctx.mdev;
3003 memset(&args, 0, sizeof(args));
3004 if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) {
f399002e 3005 err = new_c_uuid_parms_from_attrs(&args, info);
3b98c0c2
LE		 3006 if (err) {
3007 retcode = ERR_MANDATORY_TAG;
3008 drbd_msg_put_info(from_attrs_err_to_txt(err));
3009 goto out_nolock;
3010 }
b411b363
PR		 3011 }
3012
8410da8f 3013 mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */
b411b363
PR		 3014
3015 if (!get_ldev(mdev)) {
3016 retcode = ERR_NO_DISK;
3017 goto out;
3018 }
3019
3020 /* this is "skip initial sync", assume to be clean */
31890f4a 3021 if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 &&
b411b363
PR		 3022 mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) {
3023 dev_info(DEV, "Preparing to skip initial sync\n");
3024 skip_initial_sync = 1;
3025 } else if (mdev->state.conn != C_STANDALONE) {
3026 retcode = ERR_CONNECTED;
3027 goto out_dec;
3028 }
3029
3030 drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */
3031 drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */
3032
3033 if (args.clear_bm) {
20ceb2b2
LE		 3034 err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write,
3035 "clear_n_write from new_c_uuid", BM_LOCKED_MASK);
b411b363
PR		 3036 if (err) {
3037 dev_err(DEV, "Writing bitmap failed with %d\n",err);
3038 retcode = ERR_IO_MD_DISK;
3039 }
3040 if (skip_initial_sync) {
3041 drbd_send_uuids_skip_initial_sync(mdev);
3042 _drbd_uuid_set(mdev, UI_BITMAP, 0);
62b0da3a 3043 drbd_print_uuids(mdev, "cleared bitmap UUID");
87eeee41 3044 spin_lock_irq(&mdev->tconn->req_lock);
b411b363
PR		 3045 _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE),
3046 CS_VERBOSE, NULL);
87eeee41 3047 spin_unlock_irq(&mdev->tconn->req_lock);
b411b363
PR		 3048 }
3049 }
3050
3051 drbd_md_sync(mdev);
3052out_dec:
3053 put_ldev(mdev);
3054out:
8410da8f 3055 mutex_unlock(mdev->state_mutex);
3b98c0c2
LE		 3056out_nolock:
3057 drbd_adm_finish(info, retcode);
774b3055
PR		 3058 return 0;
3059}
3060
3b98c0c2 3061static enum drbd_ret_code
7c3063cc 3062drbd_check_resource_name(const char *name)
774b3055 3063{
3b98c0c2 3064 if (!name || !name[0]) {
7c3063cc 3065 drbd_msg_put_info("resource name missing");
3b98c0c2 3066 return ERR_MANDATORY_TAG;
774b3055 3067 }
3b98c0c2
LE		 3068 /* if we want to use these in sysfs/configfs/debugfs some day,
3069 * we must not allow slashes */
3070 if (strchr(name, '/')) {
7c3063cc 3071 drbd_msg_put_info("invalid resource name");
3b98c0c2 3072 return ERR_INVALID_REQUEST;
774b3055 3073 }
3b98c0c2 3074 return NO_ERROR;
774b3055
PR		 3075}
3076
789c1b62 3077int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info)
b411b363 3078{
3b98c0c2 3079 enum drbd_ret_code retcode;
afbbfa88
AG		 3080 struct res_opts res_opts;
3081 int err;
9f5180e5 3082
3b98c0c2
LE		 3083 retcode = drbd_adm_prepare(skb, info, 0);
3084 if (!adm_ctx.reply_skb)
3085 return retcode;
3086 if (retcode != NO_ERROR)
3087 goto out;
b411b363 3088
afbbfa88
AG		 3089 set_res_opts_defaults(&res_opts);
3090 err = res_opts_from_attrs(&res_opts, info);
3091 if (err && err != -ENOMSG) {
3092 retcode = ERR_MANDATORY_TAG;
3093 drbd_msg_put_info(from_attrs_err_to_txt(err));
3094 goto out;
3095 }
3096
7c3063cc 3097 retcode = drbd_check_resource_name(adm_ctx.resource_name);
3b98c0c2
LE		 3098 if (retcode != NO_ERROR)
3099 goto out;
b411b363 3100
3b98c0c2 3101 if (adm_ctx.tconn) {
38f19616
LE		 3102 if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) {
3103 retcode = ERR_INVALID_REQUEST;
789c1b62 3104 drbd_msg_put_info("resource exists");
38f19616
LE		 3105 }
3106 /* else: still NO_ERROR */
3b98c0c2 3107 goto out;
b411b363
PR		 3108 }
3109
afbbfa88 3110 if (!conn_create(adm_ctx.resource_name, &res_opts))
b411b363 3111 retcode = ERR_NOMEM;
3b98c0c2
LE		 3112out:
3113 drbd_adm_finish(info, retcode);
3114 return 0;
b411b363
PR		 3115}
3116
3b98c0c2 3117int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info)
b411b363 3118{
3b98c0c2
LE		 3119 struct drbd_genlmsghdr *dh = info->userhdr;
3120 enum drbd_ret_code retcode;
b411b363 3121
44e52cfa 3122 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE);
3b98c0c2
LE		 3123 if (!adm_ctx.reply_skb)
3124 return retcode;
3125 if (retcode != NO_ERROR)
3126 goto out;
b411b363 3127
f2257a56 3128 if (dh->minor > MINORMASK) {
3b98c0c2
LE		 3129 drbd_msg_put_info("requested minor out of range");
3130 retcode = ERR_INVALID_REQUEST;
3131 goto out;
b411b363 3132 }
0c8e36d9 3133 if (adm_ctx.volume > DRBD_VOLUME_MAX) {
3b98c0c2
LE		 3134 drbd_msg_put_info("requested volume id out of range");
3135 retcode = ERR_INVALID_REQUEST;
3136 goto out;
b411b363 3137 }
b411b363 3138
38f19616
LE		 3139 /* drbd_adm_prepare made sure already
3140 * that mdev->tconn and mdev->vnr match the request. */
3141 if (adm_ctx.mdev) {
3142 if (info->nlhdr->nlmsg_flags & NLM_F_EXCL)
3143 retcode = ERR_MINOR_EXISTS;
3144 /* else: still NO_ERROR */
3145 goto out;
3146 }
3147
3b98c0c2
LE		 3148 retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume);
3149out:
3150 drbd_adm_finish(info, retcode);
3151 return 0;
b411b363
PR		 3152}
3153
85f75dd7
LE		 3154static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev)
3155{
3156 if (mdev->state.disk == D_DISKLESS &&
3157 /* no need to be mdev->state.conn == C_STANDALONE &&
3158 * we may want to delete a minor from a live replication group.
3159 */
3160 mdev->state.role == R_SECONDARY) {
369bea63
PR		 3161 _drbd_request_state(mdev, NS(conn, C_WF_REPORT_PARAMS),
3162 CS_VERBOSE + CS_WAIT_COMPLETE);
81fa2e67
PR		 3163 idr_remove(&mdev->tconn->volumes, mdev->vnr);
3164 idr_remove(&minors, mdev_to_minor(mdev));
3165 del_gendisk(mdev->vdisk);
3166 synchronize_rcu();
3167 kref_put(&mdev->kref, &drbd_minor_destroy);
85f75dd7
LE		 3168 return NO_ERROR;
3169 } else
3170 return ERR_MINOR_CONFIGURED;
3171}
3172
3b98c0c2 3173int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info)
b411b363 3174{
3b98c0c2 3175 enum drbd_ret_code retcode;
b411b363 3176
3b98c0c2
LE		 3177 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
3178 if (!adm_ctx.reply_skb)
3179 return retcode;
3180 if (retcode != NO_ERROR)
3181 goto out;
b411b363 3182
85f75dd7 3183 retcode = adm_delete_minor(adm_ctx.mdev);
85f75dd7
LE		 3184out:
3185 drbd_adm_finish(info, retcode);
3186 return 0;
3187}
3188
3189int drbd_adm_down(struct sk_buff *skb, struct genl_info *info)
3190{
f3dfa40a 3191 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
85f75dd7
LE		 3192 struct drbd_conf *mdev;
3193 unsigned i;
3194
3195 retcode = drbd_adm_prepare(skb, info, 0);
3196 if (!adm_ctx.reply_skb)
3197 return retcode;
3198 if (retcode != NO_ERROR)
3199 goto out;
3200
3201 if (!adm_ctx.tconn) {
789c1b62 3202 retcode = ERR_RES_NOT_KNOWN;
85f75dd7
LE		 3203 goto out;
3204 }
3205
85f75dd7
LE		 3206 /* demote */
3207 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3208 retcode = drbd_set_role(mdev, R_SECONDARY, 0);
3209 if (retcode < SS_SUCCESS) {
3210 drbd_msg_put_info("failed to demote");
c141ebda 3211 goto out;
85f75dd7
LE		 3212 }
3213 }
3214
f3dfa40a
LE		 3215 retcode = conn_try_disconnect(adm_ctx.tconn, 0);
3216 if (retcode < SS_SUCCESS) {
85f75dd7 3217 drbd_msg_put_info("failed to disconnect");
f3dfa40a 3218 goto out;
85f75dd7
LE		 3219 }
3220
3221 /* detach */
3222 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
cdfda633 3223 retcode = adm_detach(mdev, 0);
27012382 3224 if (retcode < SS_SUCCESS || retcode > NO_ERROR) {
85f75dd7 3225 drbd_msg_put_info("failed to detach");
c141ebda 3226 goto out;
85f75dd7
LE		 3227 }
3228 }
3229
f3dfa40a
LE		 3230 /* If we reach this, all volumes (of this tconn) are Secondary,
3231 * Disconnected, Diskless, aka Unconfigured. Make sure all threads have
c141ebda 3232 * actually stopped, state handling only does drbd_thread_stop_nowait(). */
f3dfa40a
LE		 3233 drbd_thread_stop(&adm_ctx.tconn->worker);
3234
3235 /* Now, nothing can fail anymore */
3236
85f75dd7
LE		 3237 /* delete volumes */
3238 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3239 retcode = adm_delete_minor(mdev);
3240 if (retcode != NO_ERROR) {
3241 /* "can not happen" */
3242 drbd_msg_put_info("failed to delete volume");
ef356262 3243 goto out;
85f75dd7
LE		 3244 }
3245 }
3246
85f75dd7
LE		 3247 /* delete connection */
3248 if (conn_lowest_minor(adm_ctx.tconn) < 0) {
ec0bddbc
PR		 3249 list_del_rcu(&adm_ctx.tconn->all_tconn);
3250 synchronize_rcu();
9dc9fbb3
PR		 3251 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
3252
85f75dd7
LE		 3253 retcode = NO_ERROR;
3254 } else {
3255 /* "can not happen" */
789c1b62 3256 retcode = ERR_RES_IN_USE;
85f75dd7 3257 drbd_msg_put_info("failed to delete connection");
85f75dd7 3258 }
ef356262 3259 goto out;
3b98c0c2
LE		 3260out:
3261 drbd_adm_finish(info, retcode);
3262 return 0;
b411b363
PR		 3263}
3264
789c1b62 3265int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info)
b411b363 3266{
3b98c0c2 3267 enum drbd_ret_code retcode;
b411b363 3268
44e52cfa 3269 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE);
3b98c0c2
LE		 3270 if (!adm_ctx.reply_skb)
3271 return retcode;
3272 if (retcode != NO_ERROR)
3273 goto out;
3274
3275 if (conn_lowest_minor(adm_ctx.tconn) < 0) {
ec0bddbc
PR		 3276 list_del_rcu(&adm_ctx.tconn->all_tconn);
3277 synchronize_rcu();
9dc9fbb3
PR		 3278 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
3279
3b98c0c2
LE		 3280 retcode = NO_ERROR;
3281 } else {
789c1b62 3282 retcode = ERR_RES_IN_USE;
b411b363
PR		 3283 }
3284
992d6e91
LE		 3285 if (retcode == NO_ERROR)
3286 drbd_thread_stop(&adm_ctx.tconn->worker);
3b98c0c2
LE		 3287out:
3288 drbd_adm_finish(info, retcode);
b411b363
PR		 3289 return 0;
3290}
3291
3b98c0c2 3292void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib)
b411b363 3293{
3b98c0c2
LE		 3294 static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */
3295 struct sk_buff *msg;
3296 struct drbd_genlmsghdr *d_out;
3297 unsigned seq;
3298 int err = -ENOMEM;
3299
3300 seq = atomic_inc_return(&drbd_genl_seq);
3301 msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO);
3302 if (!msg)
3303 goto failed;
3304
3305 err = -EMSGSIZE;
3306 d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT);
3307 if (!d_out) /* cannot happen, but anyways. */
3308 goto nla_put_failure;
3309 d_out->minor = mdev_to_minor(mdev);
6f9b5f84 3310 d_out->ret_code = NO_ERROR;
3b98c0c2
LE		 3311
3312 if (nla_put_status_info(msg, mdev, sib))
3313 goto nla_put_failure;
3314 genlmsg_end(msg, d_out);
3315 err = drbd_genl_multicast_events(msg, 0);
3316 /* msg has been consumed or freed in netlink_broadcast() */
3317 if (err && err != -ESRCH)
3318 goto failed;
b411b363 3319
3b98c0c2 3320 return;
b411b363 3321
3b98c0c2
LE		 3322nla_put_failure:
3323 nlmsg_free(msg);
3324failed:
3325 dev_err(DEV, "Error %d while broadcasting event. "
3326 "Event seq:%u sib_reason:%u\n",
3327 err, seq, sib->sib_reason);
b411b363 3328}
Linux kernel - Deliverables
This page took 0.36175 seconds and 5 git commands to generate.