blob: 9d9b93f0885012fcea0ae1f476776d5a01b8431d [file] [log] [blame]
Philipp Reisnerb411b362009-09-25 16:07:19 -07001/*
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
Philipp Reisnerb411b362009-09-25 16:07:19 -070026#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>
Philipp Reisnerb411b362009-09-25 16:07:19 -070032#include <linux/blkpg.h>
33#include <linux/cpumask.h>
34#include "drbd_int.h"
Philipp Reisner265be2d2010-05-31 10:14:17 +020035#include "drbd_req.h"
Philipp Reisnerb411b362009-09-25 16:07:19 -070036#include "drbd_wrappers.h"
37#include <asm/unaligned.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070038#include <linux/drbd_limits.h>
Philipp Reisner87f7be42010-06-11 13:56:33 +020039#include <linux/kthread.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070040
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010041#include <net/genetlink.h>
Philipp Reisnerb411b362009-09-25 16:07:19 -070042
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010043/* .doit */
44// int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info);
45// int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info);
46
47int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info);
48int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info);
49
50int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info);
51int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +010052int drbd_adm_down(struct sk_buff *skb, struct genl_info *info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010053
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);
Lars Ellenbergf3990022011-03-23 14:31:09 +010056int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010057int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info);
58int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info);
Lars Ellenbergf3990022011-03-23 14:31:09 +010059int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010060int 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);
Lars Ellenbergf3990022011-03-23 14:31:09 +010071int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010072int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info);
73int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info);
74/* .dumpit */
75int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb);
76
77#include <linux/drbd_genl_api.h>
78#include <linux/genl_magic_func.h>
79
80/* used blkdev_get_by_path, to claim our meta data device(s) */
Philipp Reisnerb411b362009-09-25 16:07:19 -070081static char *drbd_m_holder = "Hands off! this is DRBD's meta data device.";
82
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010083/* Configuration is strictly serialized, because generic netlink message
84 * processing is strictly serialized by the genl_lock().
85 * Which means we can use one static global drbd_config_context struct.
86 */
87static struct drbd_config_context {
88 /* assigned from drbd_genlmsghdr */
89 unsigned int minor;
90 /* assigned from request attributes, if present */
91 unsigned int volume;
92#define VOLUME_UNSPECIFIED (-1U)
93 /* pointer into the request skb,
94 * limited lifetime! */
95 char *conn_name;
Philipp Reisnerb411b362009-09-25 16:07:19 -070096
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +010097 /* reply buffer */
98 struct sk_buff *reply_skb;
99 /* pointer into reply buffer */
100 struct drbd_genlmsghdr *reply_dh;
101 /* resolved from attributes, if possible */
102 struct drbd_conf *mdev;
103 struct drbd_tconn *tconn;
104} adm_ctx;
105
106static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info)
107{
108 genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb))));
109 if (genlmsg_reply(skb, info))
110 printk(KERN_ERR "drbd: error sending genl reply\n");
Philipp Reisnerb411b362009-09-25 16:07:19 -0700111}
112
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100113/* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only
114 * reason it could fail was no space in skb, and there are 4k available. */
Lars Ellenberg8432b312011-03-08 16:11:16 +0100115int drbd_msg_put_info(const char *info)
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100116{
117 struct sk_buff *skb = adm_ctx.reply_skb;
118 struct nlattr *nla;
119 int err = -EMSGSIZE;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700120
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100121 if (!info || !info[0])
122 return 0;
123
124 nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY);
125 if (!nla)
126 return err;
127
128 err = nla_put_string(skb, T_info_text, info);
129 if (err) {
130 nla_nest_cancel(skb, nla);
131 return err;
132 } else
133 nla_nest_end(skb, nla);
134 return 0;
135}
136
137/* This would be a good candidate for a "pre_doit" hook,
138 * and per-family private info->pointers.
139 * But we need to stay compatible with older kernels.
140 * If it returns successfully, adm_ctx members are valid.
141 */
142#define DRBD_ADM_NEED_MINOR 1
143#define DRBD_ADM_NEED_CONN 2
144static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info,
145 unsigned flags)
146{
147 struct drbd_genlmsghdr *d_in = info->userhdr;
148 const u8 cmd = info->genlhdr->cmd;
149 int err;
150
151 memset(&adm_ctx, 0, sizeof(adm_ctx));
152
153 /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */
154 if (cmd != DRBD_ADM_GET_STATUS
155 && security_netlink_recv(skb, CAP_SYS_ADMIN))
156 return -EPERM;
157
158 adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
159 if (!adm_ctx.reply_skb)
160 goto fail;
161
162 adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb,
163 info, &drbd_genl_family, 0, cmd);
164 /* put of a few bytes into a fresh skb of >= 4k will always succeed.
165 * but anyways */
166 if (!adm_ctx.reply_dh)
167 goto fail;
168
169 adm_ctx.reply_dh->minor = d_in->minor;
170 adm_ctx.reply_dh->ret_code = NO_ERROR;
171
172 if (info->attrs[DRBD_NLA_CFG_CONTEXT]) {
173 struct nlattr *nla;
174 /* parse and validate only */
Lars Ellenbergf3990022011-03-23 14:31:09 +0100175 err = drbd_cfg_context_from_attrs(NULL, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100176 if (err)
177 goto fail;
178
179 /* It was present, and valid,
180 * copy it over to the reply skb. */
181 err = nla_put_nohdr(adm_ctx.reply_skb,
182 info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len,
183 info->attrs[DRBD_NLA_CFG_CONTEXT]);
184 if (err)
185 goto fail;
186
187 /* and assign stuff to the global adm_ctx */
188 nla = nested_attr_tb[__nla_type(T_ctx_volume)];
189 adm_ctx.volume = nla ? nla_get_u32(nla) : VOLUME_UNSPECIFIED;
190 nla = nested_attr_tb[__nla_type(T_ctx_conn_name)];
191 if (nla)
192 adm_ctx.conn_name = nla_data(nla);
193 } else
194 adm_ctx.volume = VOLUME_UNSPECIFIED;
195
196 adm_ctx.minor = d_in->minor;
197 adm_ctx.mdev = minor_to_mdev(d_in->minor);
Philipp Reisner0ace9df2011-04-24 10:53:19 +0200198 adm_ctx.tconn = conn_get_by_name(adm_ctx.conn_name);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100199
200 if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) {
201 drbd_msg_put_info("unknown minor");
202 return ERR_MINOR_INVALID;
203 }
204 if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_CONN)) {
205 drbd_msg_put_info("unknown connection");
206 return ERR_INVALID_REQUEST;
207 }
208
209 /* some more paranoia, if the request was over-determined */
Lars Ellenberg527f4b22011-03-14 13:58:03 +0100210 if (adm_ctx.mdev && adm_ctx.tconn &&
211 adm_ctx.mdev->tconn != adm_ctx.tconn) {
212 pr_warning("request: minor=%u, conn=%s; but that minor belongs to connection %s\n",
213 adm_ctx.minor, adm_ctx.conn_name, adm_ctx.mdev->tconn->name);
214 drbd_msg_put_info("minor exists in different connection");
215 return ERR_INVALID_REQUEST;
216 }
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100217 if (adm_ctx.mdev &&
218 adm_ctx.volume != VOLUME_UNSPECIFIED &&
219 adm_ctx.volume != adm_ctx.mdev->vnr) {
220 pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n",
221 adm_ctx.minor, adm_ctx.volume,
222 adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name);
Lars Ellenberg527f4b22011-03-14 13:58:03 +0100223 drbd_msg_put_info("minor exists as different volume");
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100224 return ERR_INVALID_REQUEST;
225 }
Philipp Reisner0ace9df2011-04-24 10:53:19 +0200226
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100227 return NO_ERROR;
228
229fail:
230 nlmsg_free(adm_ctx.reply_skb);
231 adm_ctx.reply_skb = NULL;
232 return -ENOMEM;
233}
234
235static int drbd_adm_finish(struct genl_info *info, int retcode)
236{
237 struct nlattr *nla;
238 const char *conn_name = NULL;
239
Philipp Reisner0ace9df2011-04-24 10:53:19 +0200240 if (adm_ctx.tconn) {
241 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
242 adm_ctx.tconn = NULL;
243 }
244
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100245 if (!adm_ctx.reply_skb)
246 return -ENOMEM;
247
248 adm_ctx.reply_dh->ret_code = retcode;
249
250 nla = info->attrs[DRBD_NLA_CFG_CONTEXT];
251 if (nla) {
252 nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name));
253 if (nla)
254 conn_name = nla_data(nla);
255 }
256
257 drbd_adm_send_reply(adm_ctx.reply_skb, info);
258 return 0;
259}
Philipp Reisnerb411b362009-09-25 16:07:19 -0700260
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100261static void setup_khelper_env(struct drbd_tconn *tconn, char **envp)
262{
263 char *afs;
Philipp Reisner44ed1672011-04-19 17:10:19 +0200264 struct net_conf *nc;
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100265
Philipp Reisner44ed1672011-04-19 17:10:19 +0200266 rcu_read_lock();
267 nc = rcu_dereference(tconn->net_conf);
268 if (nc) {
269 switch (((struct sockaddr *)nc->peer_addr)->sa_family) {
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100270 case AF_INET6:
271 afs = "ipv6";
272 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6",
Philipp Reisner44ed1672011-04-19 17:10:19 +0200273 &((struct sockaddr_in6 *)nc->peer_addr)->sin6_addr);
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100274 break;
275 case AF_INET:
276 afs = "ipv4";
277 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4",
Philipp Reisner44ed1672011-04-19 17:10:19 +0200278 &((struct sockaddr_in *)nc->peer_addr)->sin_addr);
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100279 break;
280 default:
281 afs = "ssocks";
282 snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4",
Philipp Reisner44ed1672011-04-19 17:10:19 +0200283 &((struct sockaddr_in *)nc->peer_addr)->sin_addr);
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100284 }
285 snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs);
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100286 }
Philipp Reisner44ed1672011-04-19 17:10:19 +0200287 rcu_read_unlock();
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100288}
289
Philipp Reisnerb411b362009-09-25 16:07:19 -0700290int drbd_khelper(struct drbd_conf *mdev, char *cmd)
291{
292 char *envp[] = { "HOME=/",
293 "TERM=linux",
294 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100295 (char[20]) { }, /* address family */
296 (char[60]) { }, /* address */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700297 NULL };
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100298 char mb[12];
Philipp Reisnerb411b362009-09-25 16:07:19 -0700299 char *argv[] = {usermode_helper, cmd, mb, NULL };
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100300 struct sib_info sib;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700301 int ret;
302
303 snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev));
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100304 setup_khelper_env(mdev->tconn, envp);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700305
Lars Ellenberg1090c052010-07-19 17:41:04 +0200306 /* The helper may take some time.
307 * write out any unsynced meta data changes now */
308 drbd_md_sync(mdev);
309
Philipp Reisnerb411b362009-09-25 16:07:19 -0700310 dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100311 sib.sib_reason = SIB_HELPER_PRE;
312 sib.helper_name = cmd;
313 drbd_bcast_event(mdev, &sib);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700314 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
315 if (ret)
316 dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
317 usermode_helper, cmd, mb,
318 (ret >> 8) & 0xff, ret);
319 else
320 dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n",
321 usermode_helper, cmd, mb,
322 (ret >> 8) & 0xff, ret);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100323 sib.sib_reason = SIB_HELPER_POST;
324 sib.helper_exit_code = ret;
325 drbd_bcast_event(mdev, &sib);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700326
327 if (ret < 0) /* Ignore any ERRNOs we got. */
328 ret = 0;
329
330 return ret;
331}
332
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100333static void conn_md_sync(struct drbd_tconn *tconn)
334{
335 struct drbd_conf *mdev;
Philipp Reisnere90285e2011-03-22 12:51:21 +0100336 int vnr;
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100337
Philipp Reisnerd3fcb492011-04-13 14:46:05 -0700338 down_read(&drbd_cfg_rwsem);
Philipp Reisnere90285e2011-03-22 12:51:21 +0100339 idr_for_each_entry(&tconn->volumes, mdev, vnr)
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100340 drbd_md_sync(mdev);
Philipp Reisnerd3fcb492011-04-13 14:46:05 -0700341 up_read(&drbd_cfg_rwsem);
Philipp Reisner6b75dce2011-03-16 17:39:12 +0100342}
343
344int conn_khelper(struct drbd_tconn *tconn, char *cmd)
345{
346 char *envp[] = { "HOME=/",
347 "TERM=linux",
348 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
349 (char[20]) { }, /* address family */
350 (char[60]) { }, /* address */
351 NULL };
352 char *argv[] = {usermode_helper, cmd, tconn->name, NULL };
353 int ret;
354
355 setup_khelper_env(tconn, envp);
356 conn_md_sync(tconn);
357
358 conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name);
359 /* TODO: conn_bcast_event() ?? */
360
361 ret = call_usermodehelper(usermode_helper, argv, envp, 1);
362 if (ret)
363 conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n",
364 usermode_helper, cmd, tconn->name,
365 (ret >> 8) & 0xff, ret);
366 else
367 conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n",
368 usermode_helper, cmd, tconn->name,
369 (ret >> 8) & 0xff, ret);
370 /* TODO: conn_bcast_event() ?? */
371
372 if (ret < 0) /* Ignore any ERRNOs we got. */
373 ret = 0;
374
375 return ret;
376}
377
Philipp Reisnercb703452011-03-24 11:03:07 +0100378static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700379{
Philipp Reisnercb703452011-03-24 11:03:07 +0100380 enum drbd_fencing_p fp = FP_NOT_AVAIL;
381 struct drbd_conf *mdev;
382 int vnr;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700383
Philipp Reisner695d08f2011-04-11 22:53:32 -0700384 rcu_read_lock();
Philipp Reisnercb703452011-03-24 11:03:07 +0100385 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
386 if (get_ldev_if_state(mdev, D_CONSISTENT)) {
387 fp = max_t(enum drbd_fencing_p, fp, mdev->ldev->dc.fencing);
388 put_ldev(mdev);
389 }
Philipp Reisnerb411b362009-09-25 16:07:19 -0700390 }
Philipp Reisner695d08f2011-04-11 22:53:32 -0700391 rcu_read_unlock();
Philipp Reisnerb411b362009-09-25 16:07:19 -0700392
Philipp Reisnercb703452011-03-24 11:03:07 +0100393 return fp;
394}
395
396bool conn_try_outdate_peer(struct drbd_tconn *tconn)
397{
398 union drbd_state mask = { };
399 union drbd_state val = { };
400 enum drbd_fencing_p fp;
401 char *ex_to_string;
402 int r;
403
404 if (tconn->cstate >= C_WF_REPORT_PARAMS) {
405 conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n");
406 return false;
407 }
408
409 fp = highest_fencing_policy(tconn);
410 switch (fp) {
411 case FP_NOT_AVAIL:
412 conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n");
413 goto out;
414 case FP_DONT_CARE:
415 return true;
416 default: ;
417 }
418
419 r = conn_khelper(tconn, "fence-peer");
Philipp Reisnerb411b362009-09-25 16:07:19 -0700420
421 switch ((r>>8) & 0xff) {
422 case 3: /* peer is inconsistent */
423 ex_to_string = "peer is inconsistent or worse";
Philipp Reisnercb703452011-03-24 11:03:07 +0100424 mask.pdsk = D_MASK;
425 val.pdsk = D_INCONSISTENT;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700426 break;
427 case 4: /* peer got outdated, or was already outdated */
428 ex_to_string = "peer was fenced";
Philipp Reisnercb703452011-03-24 11:03:07 +0100429 mask.pdsk = D_MASK;
430 val.pdsk = D_OUTDATED;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700431 break;
432 case 5: /* peer was down */
Philipp Reisnercb703452011-03-24 11:03:07 +0100433 if (conn_highest_disk(tconn) == D_UP_TO_DATE) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700434 /* we will(have) create(d) a new UUID anyways... */
435 ex_to_string = "peer is unreachable, assumed to be dead";
Philipp Reisnercb703452011-03-24 11:03:07 +0100436 mask.pdsk = D_MASK;
437 val.pdsk = D_OUTDATED;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700438 } else {
439 ex_to_string = "peer unreachable, doing nothing since disk != UpToDate";
Philipp Reisnerb411b362009-09-25 16:07:19 -0700440 }
441 break;
442 case 6: /* Peer is primary, voluntarily outdate myself.
443 * This is useful when an unconnected R_SECONDARY is asked to
444 * become R_PRIMARY, but finds the other peer being active. */
445 ex_to_string = "peer is active";
Philipp Reisnercb703452011-03-24 11:03:07 +0100446 conn_warn(tconn, "Peer is primary, outdating myself.\n");
447 mask.disk = D_MASK;
448 val.disk = D_OUTDATED;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700449 break;
450 case 7:
451 if (fp != FP_STONITH)
Philipp Reisnercb703452011-03-24 11:03:07 +0100452 conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n");
Philipp Reisnerb411b362009-09-25 16:07:19 -0700453 ex_to_string = "peer was stonithed";
Philipp Reisnercb703452011-03-24 11:03:07 +0100454 mask.pdsk = D_MASK;
455 val.pdsk = D_OUTDATED;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700456 break;
457 default:
458 /* The script is broken ... */
Philipp Reisnercb703452011-03-24 11:03:07 +0100459 conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff);
460 return false; /* Eventually leave IO frozen */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700461 }
462
Philipp Reisnercb703452011-03-24 11:03:07 +0100463 conn_info(tconn, "fence-peer helper returned %d (%s)\n",
464 (r>>8) & 0xff, ex_to_string);
Philipp Reisnerfb22c402010-09-08 23:20:21 +0200465
Philipp Reisnercb703452011-03-24 11:03:07 +0100466 out:
Philipp Reisnerfb22c402010-09-08 23:20:21 +0200467
Philipp Reisnercb703452011-03-24 11:03:07 +0100468 /* Not using
469 conn_request_state(tconn, mask, val, CS_VERBOSE);
470 here, because we might were able to re-establish the connection in the
471 meantime. */
472 spin_lock_irq(&tconn->req_lock);
473 if (tconn->cstate < C_WF_REPORT_PARAMS)
474 _conn_request_state(tconn, mask, val, CS_VERBOSE);
475 spin_unlock_irq(&tconn->req_lock);
476
477 return conn_highest_pdsk(tconn) <= D_OUTDATED;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700478}
479
Philipp Reisner87f7be42010-06-11 13:56:33 +0200480static int _try_outdate_peer_async(void *data)
481{
Philipp Reisnercb703452011-03-24 11:03:07 +0100482 struct drbd_tconn *tconn = (struct drbd_tconn *)data;
Philipp Reisner87f7be42010-06-11 13:56:33 +0200483
Philipp Reisnercb703452011-03-24 11:03:07 +0100484 conn_try_outdate_peer(tconn);
Philipp Reisner87f7be42010-06-11 13:56:33 +0200485
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +0200486 kref_put(&tconn->kref, &conn_destroy);
Philipp Reisner87f7be42010-06-11 13:56:33 +0200487 return 0;
488}
489
Philipp Reisnercb703452011-03-24 11:03:07 +0100490void conn_try_outdate_peer_async(struct drbd_tconn *tconn)
Philipp Reisner87f7be42010-06-11 13:56:33 +0200491{
492 struct task_struct *opa;
493
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +0200494 kref_get(&tconn->kref);
Philipp Reisnercb703452011-03-24 11:03:07 +0100495 opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h");
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +0200496 if (IS_ERR(opa)) {
Philipp Reisnercb703452011-03-24 11:03:07 +0100497 conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n");
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +0200498 kref_put(&tconn->kref, &conn_destroy);
499 }
Philipp Reisner87f7be42010-06-11 13:56:33 +0200500}
Philipp Reisnerb411b362009-09-25 16:07:19 -0700501
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100502enum drbd_state_rv
503drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700504{
505 const int max_tries = 4;
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100506 enum drbd_state_rv rv = SS_UNKNOWN_ERROR;
Philipp Reisner44ed1672011-04-19 17:10:19 +0200507 struct net_conf *nc;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700508 int try = 0;
509 int forced = 0;
510 union drbd_state mask, val;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700511
512 if (new_role == R_PRIMARY)
Philipp Reisner0625ac12011-02-07 14:49:19 +0100513 request_ping(mdev->tconn); /* Detect a dead peer ASAP */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700514
Philipp Reisner8410da82011-02-11 20:11:10 +0100515 mutex_lock(mdev->state_mutex);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700516
517 mask.i = 0; mask.role = R_MASK;
518 val.i = 0; val.role = new_role;
519
520 while (try++ < max_tries) {
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100521 rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700522
523 /* in case we first succeeded to outdate,
524 * but now suddenly could establish a connection */
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100525 if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700526 val.pdsk = 0;
527 mask.pdsk = 0;
528 continue;
529 }
530
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100531 if (rv == SS_NO_UP_TO_DATE_DISK && force &&
Philipp Reisnerd10a33c2010-03-04 15:11:39 +0100532 (mdev->state.disk < D_UP_TO_DATE &&
533 mdev->state.disk >= D_INCONSISTENT)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700534 mask.disk = D_MASK;
535 val.disk = D_UP_TO_DATE;
536 forced = 1;
537 continue;
538 }
539
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100540 if (rv == SS_NO_UP_TO_DATE_DISK &&
Philipp Reisnerb411b362009-09-25 16:07:19 -0700541 mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) {
542 D_ASSERT(mdev->state.pdsk == D_UNKNOWN);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700543
Philipp Reisnercb703452011-03-24 11:03:07 +0100544 if (conn_try_outdate_peer(mdev->tconn)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700545 val.disk = D_UP_TO_DATE;
546 mask.disk = D_MASK;
547 }
Philipp Reisnerb411b362009-09-25 16:07:19 -0700548 continue;
549 }
550
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100551 if (rv == SS_NOTHING_TO_DO)
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100552 goto out;
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100553 if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) {
Philipp Reisnercb703452011-03-24 11:03:07 +0100554 if (!conn_try_outdate_peer(mdev->tconn) && force) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700555 dev_warn(DEV, "Forced into split brain situation!\n");
Philipp Reisnercb703452011-03-24 11:03:07 +0100556 mask.pdsk = D_MASK;
557 val.pdsk = D_OUTDATED;
558
Philipp Reisnerb411b362009-09-25 16:07:19 -0700559 }
Philipp Reisnerb411b362009-09-25 16:07:19 -0700560 continue;
561 }
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100562 if (rv == SS_TWO_PRIMARIES) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700563 /* Maybe the peer is detected as dead very soon...
564 retry at most once more in this case. */
Philipp Reisner44ed1672011-04-19 17:10:19 +0200565 int timeo;
566 rcu_read_lock();
567 nc = rcu_dereference(mdev->tconn->net_conf);
568 timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1;
569 rcu_read_unlock();
570 schedule_timeout_interruptible(timeo);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700571 if (try < max_tries)
572 try = max_tries - 1;
573 continue;
574 }
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100575 if (rv < SS_SUCCESS) {
576 rv = _drbd_request_state(mdev, mask, val,
Philipp Reisnerb411b362009-09-25 16:07:19 -0700577 CS_VERBOSE + CS_WAIT_COMPLETE);
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100578 if (rv < SS_SUCCESS)
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100579 goto out;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700580 }
581 break;
582 }
583
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100584 if (rv < SS_SUCCESS)
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100585 goto out;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700586
587 if (forced)
588 dev_warn(DEV, "Forced to consider local data as UpToDate!\n");
589
590 /* Wait until nothing is on the fly :) */
591 wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0);
592
593 if (new_role == R_SECONDARY) {
Andreas Gruenbacher81e84652010-12-09 15:03:57 +0100594 set_disk_ro(mdev->vdisk, true);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700595 if (get_ldev(mdev)) {
596 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
597 put_ldev(mdev);
598 }
599 } else {
Philipp Reisner91fd4da2011-04-20 17:47:29 +0200600 mutex_lock(&mdev->tconn->net_conf_update);
601 nc = mdev->tconn->net_conf;
Philipp Reisner44ed1672011-04-19 17:10:19 +0200602 if (nc)
Philipp Reisner91fd4da2011-04-20 17:47:29 +0200603 nc->want_lose = 0; /* without copy; single bit op is atomic */
604 mutex_unlock(&mdev->tconn->net_conf_update);
605
Andreas Gruenbacher81e84652010-12-09 15:03:57 +0100606 set_disk_ro(mdev->vdisk, false);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700607 if (get_ldev(mdev)) {
608 if (((mdev->state.conn < C_CONNECTED ||
609 mdev->state.pdsk <= D_FAILED)
610 && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced)
611 drbd_uuid_new_current(mdev);
612
613 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
614 put_ldev(mdev);
615 }
616 }
617
Lars Ellenberg19f843a2010-12-15 08:59:11 +0100618 /* writeout of activity log covered areas of the bitmap
619 * to stable storage done in after state change already */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700620
621 if (mdev->state.conn >= C_WF_REPORT_PARAMS) {
622 /* if this was forced, we should consider sync */
623 if (forced)
624 drbd_send_uuids(mdev);
625 drbd_send_state(mdev);
626 }
627
628 drbd_md_sync(mdev);
629
630 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100631out:
Philipp Reisner8410da82011-02-11 20:11:10 +0100632 mutex_unlock(mdev->state_mutex);
Andreas Gruenbacherbf885f82010-12-08 00:39:32 +0100633 return rv;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700634}
635
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100636static const char *from_attrs_err_to_txt(int err)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700637{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100638 return err == -ENOMSG ? "required attribute missing" :
639 err == -EOPNOTSUPP ? "unknown mandatory attribute" :
Lars Ellenbergf3990022011-03-23 14:31:09 +0100640 err == -EEXIST ? "can not change invariant setting" :
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100641 "invalid attribute value";
Philipp Reisnerb411b362009-09-25 16:07:19 -0700642}
643
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100644int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700645{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100646 struct set_role_parms parms;
647 int err;
648 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700649
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100650 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
651 if (!adm_ctx.reply_skb)
652 return retcode;
653 if (retcode != NO_ERROR)
654 goto out;
655
656 memset(&parms, 0, sizeof(parms));
657 if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) {
Lars Ellenbergf3990022011-03-23 14:31:09 +0100658 err = set_role_parms_from_attrs(&parms, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100659 if (err) {
660 retcode = ERR_MANDATORY_TAG;
661 drbd_msg_put_info(from_attrs_err_to_txt(err));
662 goto out;
663 }
664 }
665
666 if (info->genlhdr->cmd == DRBD_ADM_PRIMARY)
667 retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate);
668 else
669 retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0);
670out:
671 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700672 return 0;
673}
674
675/* initializes the md.*_offset members, so we are able to find
676 * the on disk meta data */
677static void drbd_md_set_sector_offsets(struct drbd_conf *mdev,
678 struct drbd_backing_dev *bdev)
679{
680 sector_t md_size_sect = 0;
681 switch (bdev->dc.meta_dev_idx) {
682 default:
683 /* v07 style fixed size indexed meta data */
684 bdev->md.md_size_sect = MD_RESERVED_SECT;
685 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
686 bdev->md.al_offset = MD_AL_OFFSET;
687 bdev->md.bm_offset = MD_BM_OFFSET;
688 break;
689 case DRBD_MD_INDEX_FLEX_EXT:
690 /* just occupy the full device; unit: sectors */
691 bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev);
692 bdev->md.md_offset = 0;
693 bdev->md.al_offset = MD_AL_OFFSET;
694 bdev->md.bm_offset = MD_BM_OFFSET;
695 break;
696 case DRBD_MD_INDEX_INTERNAL:
697 case DRBD_MD_INDEX_FLEX_INT:
698 bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
699 /* al size is still fixed */
Lars Ellenberg7ad651b2011-02-21 13:21:03 +0100700 bdev->md.al_offset = -MD_AL_SECTORS;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700701 /* we need (slightly less than) ~ this much bitmap sectors: */
702 md_size_sect = drbd_get_capacity(bdev->backing_bdev);
703 md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT);
704 md_size_sect = BM_SECT_TO_EXT(md_size_sect);
705 md_size_sect = ALIGN(md_size_sect, 8);
706
707 /* plus the "drbd meta data super block",
708 * and the activity log; */
709 md_size_sect += MD_BM_OFFSET;
710
711 bdev->md.md_size_sect = md_size_sect;
712 /* bitmap offset is adjusted by 'super' block size */
713 bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET;
714 break;
715 }
716}
717
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100718/* input size is expected to be in KB */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700719char *ppsize(char *buf, unsigned long long size)
720{
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100721 /* Needs 9 bytes at max including trailing NUL:
722 * -1ULL ==> "16384 EB" */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700723 static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' };
724 int base = 0;
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100725 while (size >= 10000 && base < sizeof(units)-1) {
Philipp Reisnerb411b362009-09-25 16:07:19 -0700726 /* shift + round */
727 size = (size >> 10) + !!(size & (1<<9));
728 base++;
729 }
Lars Ellenberg4b0715f2010-12-14 15:13:04 +0100730 sprintf(buf, "%u %cB", (unsigned)size, units[base]);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700731
732 return buf;
733}
734
735/* there is still a theoretical deadlock when called from receiver
736 * on an D_INCONSISTENT R_PRIMARY:
737 * remote READ does inc_ap_bio, receiver would need to receive answer
738 * packet from remote to dec_ap_bio again.
739 * receiver receive_sizes(), comes here,
740 * waits for ap_bio_cnt == 0. -> deadlock.
741 * but this cannot happen, actually, because:
742 * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable
743 * (not connected, or bad/no disk on peer):
744 * see drbd_fail_request_early, ap_bio_cnt is zero.
745 * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET:
746 * peer may not initiate a resize.
747 */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +0100748/* Note these are not to be confused with
749 * drbd_adm_suspend_io/drbd_adm_resume_io,
750 * which are (sub) state changes triggered by admin (drbdsetup),
751 * and can be long lived.
752 * This changes an mdev->flag, is triggered by drbd internals,
753 * and should be short-lived. */
Philipp Reisnerb411b362009-09-25 16:07:19 -0700754void drbd_suspend_io(struct drbd_conf *mdev)
755{
756 set_bit(SUSPEND_IO, &mdev->flags);
Philipp Reisner2aebfab2011-03-28 16:48:11 +0200757 if (drbd_suspended(mdev))
Philipp Reisner265be2d2010-05-31 10:14:17 +0200758 return;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700759 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt));
760}
761
762void drbd_resume_io(struct drbd_conf *mdev)
763{
764 clear_bit(SUSPEND_IO, &mdev->flags);
765 wake_up(&mdev->misc_wait);
766}
767
768/**
769 * drbd_determine_dev_size() - Sets the right device size obeying all constraints
770 * @mdev: DRBD device.
771 *
772 * Returns 0 on success, negative return values indicate errors.
773 * You should call drbd_md_sync() after calling this function.
774 */
Bart Van Assche24c48302011-05-21 18:32:29 +0200775enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700776{
777 sector_t prev_first_sect, prev_size; /* previous meta location */
778 sector_t la_size;
779 sector_t size;
780 char ppb[10];
781
782 int md_moved, la_size_changed;
783 enum determine_dev_size rv = unchanged;
784
785 /* race:
786 * application request passes inc_ap_bio,
787 * but then cannot get an AL-reference.
788 * this function later may wait on ap_bio_cnt == 0. -> deadlock.
789 *
790 * to avoid that:
791 * Suspend IO right here.
792 * still lock the act_log to not trigger ASSERTs there.
793 */
794 drbd_suspend_io(mdev);
795
796 /* no wait necessary anymore, actually we could assert that */
797 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
798
799 prev_first_sect = drbd_md_first_sector(mdev->ldev);
800 prev_size = mdev->ldev->md.md_size_sect;
801 la_size = mdev->ldev->md.la_size_sect;
802
803 /* TODO: should only be some assert here, not (re)init... */
804 drbd_md_set_sector_offsets(mdev, mdev->ldev);
805
Philipp Reisnerd8450302010-03-24 15:51:26 +0100806 size = drbd_new_dev_size(mdev, mdev->ldev, flags & DDSF_FORCED);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700807
808 if (drbd_get_capacity(mdev->this_bdev) != size ||
809 drbd_bm_capacity(mdev) != size) {
810 int err;
Philipp Reisner02d9a942010-03-24 16:23:03 +0100811 err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC));
Philipp Reisnerb411b362009-09-25 16:07:19 -0700812 if (unlikely(err)) {
813 /* currently there is only one error: ENOMEM! */
814 size = drbd_bm_capacity(mdev)>>1;
815 if (size == 0) {
816 dev_err(DEV, "OUT OF MEMORY! "
817 "Could not allocate bitmap!\n");
818 } else {
819 dev_err(DEV, "BM resizing failed. "
820 "Leaving size unchanged at size = %lu KB\n",
821 (unsigned long)size);
822 }
823 rv = dev_size_error;
824 }
825 /* racy, see comments above. */
826 drbd_set_my_capacity(mdev, size);
827 mdev->ldev->md.la_size_sect = size;
828 dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1),
829 (unsigned long long)size>>1);
830 }
831 if (rv == dev_size_error)
832 goto out;
833
834 la_size_changed = (la_size != mdev->ldev->md.la_size_sect);
835
836 md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev)
837 || prev_size != mdev->ldev->md.md_size_sect;
838
839 if (la_size_changed || md_moved) {
Andreas Gruenbacher24dccab2010-12-12 17:45:41 +0100840 int err;
841
Philipp Reisnerb411b362009-09-25 16:07:19 -0700842 drbd_al_shrink(mdev); /* All extents inactive. */
843 dev_info(DEV, "Writing the whole bitmap, %s\n",
844 la_size_changed && md_moved ? "size changed and md moved" :
845 la_size_changed ? "size changed" : "md moved");
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +0100846 /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */
847 err = drbd_bitmap_io(mdev, &drbd_bm_write,
848 "size changed", BM_LOCKED_MASK);
Andreas Gruenbacher24dccab2010-12-12 17:45:41 +0100849 if (err) {
850 rv = dev_size_error;
851 goto out;
852 }
Philipp Reisnerb411b362009-09-25 16:07:19 -0700853 drbd_md_mark_dirty(mdev);
854 }
855
856 if (size > la_size)
857 rv = grew;
858 if (size < la_size)
859 rv = shrunk;
860out:
861 lc_unlock(mdev->act_log);
862 wake_up(&mdev->al_wait);
863 drbd_resume_io(mdev);
864
865 return rv;
866}
867
868sector_t
Philipp Reisnera393db62009-12-22 13:35:52 +0100869drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, int assume_peer_has_space)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700870{
871 sector_t p_size = mdev->p_size; /* partner's disk size. */
872 sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */
873 sector_t m_size; /* my size */
874 sector_t u_size = bdev->dc.disk_size; /* size requested by user. */
875 sector_t size = 0;
876
877 m_size = drbd_get_max_capacity(bdev);
878
Philipp Reisnera393db62009-12-22 13:35:52 +0100879 if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) {
880 dev_warn(DEV, "Resize while not connected was forced by the user!\n");
881 p_size = m_size;
882 }
883
Philipp Reisnerb411b362009-09-25 16:07:19 -0700884 if (p_size && m_size) {
885 size = min_t(sector_t, p_size, m_size);
886 } else {
887 if (la_size) {
888 size = la_size;
889 if (m_size && m_size < size)
890 size = m_size;
891 if (p_size && p_size < size)
892 size = p_size;
893 } else {
894 if (m_size)
895 size = m_size;
896 if (p_size)
897 size = p_size;
898 }
899 }
900
901 if (size == 0)
902 dev_err(DEV, "Both nodes diskless!\n");
903
904 if (u_size) {
905 if (u_size > size)
906 dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n",
907 (unsigned long)u_size>>1, (unsigned long)size>>1);
908 else
909 size = u_size;
910 }
911
912 return size;
913}
914
915/**
916 * drbd_check_al_size() - Ensures that the AL is of the right size
917 * @mdev: DRBD device.
918 *
919 * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation
920 * failed, and 0 on success. You should call drbd_md_sync() after you called
921 * this function.
922 */
Lars Ellenbergf3990022011-03-23 14:31:09 +0100923static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700924{
925 struct lru_cache *n, *t;
926 struct lc_element *e;
927 unsigned int in_use;
928 int i;
929
Lars Ellenbergf3990022011-03-23 14:31:09 +0100930 if (!expect(dc->al_extents >= DRBD_AL_EXTENTS_MIN))
931 dc->al_extents = DRBD_AL_EXTENTS_MIN;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700932
933 if (mdev->act_log &&
Lars Ellenbergf3990022011-03-23 14:31:09 +0100934 mdev->act_log->nr_elements == dc->al_extents)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700935 return 0;
936
937 in_use = 0;
938 t = mdev->act_log;
Lars Ellenberg7ad651b2011-02-21 13:21:03 +0100939 n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION,
Lars Ellenbergf3990022011-03-23 14:31:09 +0100940 dc->al_extents, sizeof(struct lc_element), 0);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700941
942 if (n == NULL) {
943 dev_err(DEV, "Cannot allocate act_log lru!\n");
944 return -ENOMEM;
945 }
946 spin_lock_irq(&mdev->al_lock);
947 if (t) {
948 for (i = 0; i < t->nr_elements; i++) {
949 e = lc_element_by_index(t, i);
950 if (e->refcnt)
951 dev_err(DEV, "refcnt(%d)==%d\n",
952 e->lc_number, e->refcnt);
953 in_use += e->refcnt;
954 }
955 }
956 if (!in_use)
957 mdev->act_log = n;
958 spin_unlock_irq(&mdev->al_lock);
959 if (in_use) {
960 dev_err(DEV, "Activity log still in use!\n");
961 lc_destroy(n);
962 return -EBUSY;
963 } else {
964 if (t)
965 lc_destroy(t);
966 }
967 drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */
968 return 0;
969}
970
Philipp Reisner99432fc2011-05-20 16:39:13 +0200971static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size)
Philipp Reisnerb411b362009-09-25 16:07:19 -0700972{
973 struct request_queue * const q = mdev->rq_queue;
Philipp Reisner99432fc2011-05-20 16:39:13 +0200974 int max_hw_sectors = max_bio_size >> 9;
975 int max_segments = 0;
976
977 if (get_ldev_if_state(mdev, D_ATTACHING)) {
978 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
979
980 max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9);
981 max_segments = mdev->ldev->dc.max_bio_bvecs;
982 put_ldev(mdev);
983 }
Philipp Reisnerb411b362009-09-25 16:07:19 -0700984
Philipp Reisnerb411b362009-09-25 16:07:19 -0700985 blk_queue_logical_block_size(q, 512);
Lars Ellenberg1816a2b2010-11-11 15:19:07 +0100986 blk_queue_max_hw_sectors(q, max_hw_sectors);
987 /* This is the workaround for "bio would need to, but cannot, be split" */
988 blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS);
989 blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1);
Philipp Reisnerb411b362009-09-25 16:07:19 -0700990
Philipp Reisner99432fc2011-05-20 16:39:13 +0200991 if (get_ldev_if_state(mdev, D_ATTACHING)) {
992 struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue;
Philipp Reisnerb411b362009-09-25 16:07:19 -0700993
Philipp Reisner99432fc2011-05-20 16:39:13 +0200994 blk_queue_stack_limits(q, b);
995
996 if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) {
997 dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n",
998 q->backing_dev_info.ra_pages,
999 b->backing_dev_info.ra_pages);
1000 q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages;
1001 }
1002 put_ldev(mdev);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001003 }
1004}
1005
Philipp Reisner99432fc2011-05-20 16:39:13 +02001006void drbd_reconsider_max_bio_size(struct drbd_conf *mdev)
1007{
1008 int now, new, local, peer;
1009
1010 now = queue_max_hw_sectors(mdev->rq_queue) << 9;
1011 local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */
1012 peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */
1013
1014 if (get_ldev_if_state(mdev, D_ATTACHING)) {
1015 local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9;
1016 mdev->local_max_bio_size = local;
1017 put_ldev(mdev);
1018 }
1019
1020 /* We may ignore peer limits if the peer is modern enough.
1021 Because new from 8.3.8 onwards the peer can use multiple
1022 BIOs for a single peer_request */
1023 if (mdev->state.conn >= C_CONNECTED) {
Philipp Reisner31890f42011-01-19 14:12:51 +01001024 if (mdev->tconn->agreed_pro_version < 94)
Philipp Reisner99432fc2011-05-20 16:39:13 +02001025 peer = mdev->peer_max_bio_size;
Philipp Reisner31890f42011-01-19 14:12:51 +01001026 else if (mdev->tconn->agreed_pro_version == 94)
Philipp Reisner99432fc2011-05-20 16:39:13 +02001027 peer = DRBD_MAX_SIZE_H80_PACKET;
1028 else /* drbd 8.3.8 onwards */
1029 peer = DRBD_MAX_BIO_SIZE;
1030 }
1031
1032 new = min_t(int, local, peer);
1033
1034 if (mdev->state.role == R_PRIMARY && new < now)
1035 dev_err(DEV, "ASSERT FAILED new < now; (%d < %d)\n", new, now);
1036
1037 if (new != now)
1038 dev_info(DEV, "max BIO size = %u\n", new);
1039
1040 drbd_setup_queue_param(mdev, new);
1041}
1042
Philipp Reisnera18e9d12011-04-24 11:09:55 +02001043/* Starts the worker thread */
Philipp Reisner0e29d162011-02-18 14:23:11 +01001044static void conn_reconfig_start(struct drbd_tconn *tconn)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001045{
Philipp Reisner0e29d162011-02-18 14:23:11 +01001046 drbd_thread_start(&tconn->worker);
1047 conn_flush_workqueue(tconn);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001048}
1049
Philipp Reisnera18e9d12011-04-24 11:09:55 +02001050/* if still unconfigured, stops worker again. */
Philipp Reisner0e29d162011-02-18 14:23:11 +01001051static void conn_reconfig_done(struct drbd_tconn *tconn)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001052{
Philipp Reisner0e29d162011-02-18 14:23:11 +01001053 spin_lock_irq(&tconn->req_lock);
Philipp Reisnera18e9d12011-04-24 11:09:55 +02001054 if (conn_all_vols_unconf(tconn))
Philipp Reisner0e29d162011-02-18 14:23:11 +01001055 drbd_thread_stop_nowait(&tconn->worker);
Philipp Reisner0e29d162011-02-18 14:23:11 +01001056 spin_unlock_irq(&tconn->req_lock);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001057}
1058
Philipp Reisner07782862010-08-31 12:00:50 +02001059/* Make sure IO is suspended before calling this function(). */
1060static void drbd_suspend_al(struct drbd_conf *mdev)
1061{
1062 int s = 0;
1063
Lars Ellenberg61610422011-02-21 13:20:54 +01001064 if (!lc_try_lock(mdev->act_log)) {
Philipp Reisner07782862010-08-31 12:00:50 +02001065 dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n");
1066 return;
1067 }
1068
Lars Ellenberg61610422011-02-21 13:20:54 +01001069 drbd_al_shrink(mdev);
Philipp Reisner87eeee42011-01-19 14:16:30 +01001070 spin_lock_irq(&mdev->tconn->req_lock);
Philipp Reisner07782862010-08-31 12:00:50 +02001071 if (mdev->state.conn < C_CONNECTED)
1072 s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags);
Philipp Reisner87eeee42011-01-19 14:16:30 +01001073 spin_unlock_irq(&mdev->tconn->req_lock);
Lars Ellenberg61610422011-02-21 13:20:54 +01001074 lc_unlock(mdev->act_log);
Philipp Reisner07782862010-08-31 12:00:50 +02001075
1076 if (s)
1077 dev_info(DEV, "Suspended AL updates\n");
1078}
1079
Lars Ellenberg5979e362011-04-27 21:09:55 +02001080
1081static bool should_set_defaults(struct genl_info *info)
1082{
1083 unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags;
1084 return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS);
1085}
1086
1087/* Maybe we should we generate these functions
1088 * from the drbd_genl.h magic as well?
1089 * That way we would not "accidentally forget" to add defaults here. */
1090
1091#define RESET_ARRAY_FIELD(field) do { \
1092 memset(field, 0, sizeof(field)); \
1093 field ## _len = 0; \
1094} while (0)
1095void drbd_set_res_opts_default(struct res_opts *r)
1096{
1097 RESET_ARRAY_FIELD(r->cpu_mask);
1098 r->on_no_data = DRBD_ON_NO_DATA_DEF;
1099}
1100
1101static void drbd_set_net_conf_defaults(struct net_conf *nc)
1102{
1103 /* Do NOT (re)set those fields marked as GENLA_F_INVARIANT
1104 * in drbd_genl.h, they can only be change with disconnect/reconnect */
1105 RESET_ARRAY_FIELD(nc->shared_secret);
1106
1107 RESET_ARRAY_FIELD(nc->cram_hmac_alg);
1108 RESET_ARRAY_FIELD(nc->integrity_alg);
1109 RESET_ARRAY_FIELD(nc->verify_alg);
1110 RESET_ARRAY_FIELD(nc->csums_alg);
1111#undef RESET_ARRAY_FIELD
1112
1113 nc->wire_protocol = DRBD_PROTOCOL_DEF;
1114 nc->try_connect_int = DRBD_CONNECT_INT_DEF;
1115 nc->timeout = DRBD_TIMEOUT_DEF;
1116 nc->ping_int = DRBD_PING_INT_DEF;
1117 nc->ping_timeo = DRBD_PING_TIMEO_DEF;
1118 nc->sndbuf_size = DRBD_SNDBUF_SIZE_DEF;
1119 nc->rcvbuf_size = DRBD_RCVBUF_SIZE_DEF;
1120 nc->ko_count = DRBD_KO_COUNT_DEF;
1121 nc->max_buffers = DRBD_MAX_BUFFERS_DEF;
1122 nc->max_epoch_size = DRBD_MAX_EPOCH_SIZE_DEF;
1123 nc->unplug_watermark = DRBD_UNPLUG_WATERMARK_DEF;
1124 nc->after_sb_0p = DRBD_AFTER_SB_0P_DEF;
1125 nc->after_sb_1p = DRBD_AFTER_SB_1P_DEF;
1126 nc->after_sb_2p = DRBD_AFTER_SB_2P_DEF;
1127 nc->rr_conflict = DRBD_RR_CONFLICT_DEF;
1128 nc->on_congestion = DRBD_ON_CONGESTION_DEF;
1129 nc->cong_fill = DRBD_CONG_FILL_DEF;
1130 nc->cong_extents = DRBD_CONG_EXTENTS_DEF;
1131 nc->two_primaries = 0;
1132 nc->no_cork = 0;
1133 nc->always_asbp = 0;
1134 nc->use_rle = 0;
1135}
1136
1137static void drbd_set_disk_conf_defaults(struct disk_conf *dc)
1138{
1139 /* Do NOT (re)set those fields marked as GENLA_F_INVARIANT
1140 * in drbd_genl.h, they can only be change with detach/reattach */
1141 dc->on_io_error = DRBD_ON_IO_ERROR_DEF;
1142 dc->fencing = DRBD_FENCING_DEF;
1143 dc->resync_rate = DRBD_RATE_DEF;
1144 dc->resync_after = DRBD_AFTER_DEF;
1145 dc->al_extents = DRBD_AL_EXTENTS_DEF;
1146 dc->c_plan_ahead = DRBD_C_PLAN_AHEAD_DEF;
1147 dc->c_delay_target = DRBD_C_DELAY_TARGET_DEF;
1148 dc->c_fill_target = DRBD_C_FILL_TARGET_DEF;
1149 dc->c_max_rate = DRBD_C_MAX_RATE_DEF;
1150 dc->c_min_rate = DRBD_C_MIN_RATE_DEF;
1151 dc->no_disk_barrier = 0;
1152 dc->no_disk_flush = 0;
1153 dc->no_disk_drain = 0;
1154 dc->no_md_flush = 0;
1155}
1156
1157
Lars Ellenbergf3990022011-03-23 14:31:09 +01001158int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info)
1159{
1160 enum drbd_ret_code retcode;
1161 struct drbd_conf *mdev;
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001162 struct disk_conf *new_disk_conf;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001163 int err, fifo_size;
1164 int *rs_plan_s = NULL;
1165
1166 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1167 if (!adm_ctx.reply_skb)
1168 return retcode;
1169 if (retcode != NO_ERROR)
1170 goto out;
1171
1172 mdev = adm_ctx.mdev;
1173
1174 /* we also need a disk
1175 * to change the options on */
1176 if (!get_ldev(mdev)) {
1177 retcode = ERR_NO_DISK;
1178 goto out;
1179 }
1180
1181/* FIXME freeze IO, cluster wide.
1182 *
1183 * We should make sure no-one uses
1184 * some half-updated struct when we
1185 * assign it later. */
1186
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001187 new_disk_conf = kmalloc(sizeof(*new_disk_conf), GFP_KERNEL);
1188 if (!new_disk_conf) {
Lars Ellenbergf3990022011-03-23 14:31:09 +01001189 retcode = ERR_NOMEM;
1190 goto fail;
1191 }
1192
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001193 memcpy(new_disk_conf, &mdev->ldev->dc, sizeof(*new_disk_conf));
Lars Ellenberg5979e362011-04-27 21:09:55 +02001194 if (should_set_defaults(info))
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001195 drbd_set_disk_conf_defaults(new_disk_conf);
Lars Ellenberg5979e362011-04-27 21:09:55 +02001196
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001197 err = disk_conf_from_attrs_for_change(new_disk_conf, info);
Lars Ellenbergf3990022011-03-23 14:31:09 +01001198 if (err) {
1199 retcode = ERR_MANDATORY_TAG;
1200 drbd_msg_put_info(from_attrs_err_to_txt(err));
1201 }
1202
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001203 if (!expect(new_disk_conf->resync_rate >= 1))
1204 new_disk_conf->resync_rate = 1;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001205
1206 /* clip to allowed range */
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001207 if (!expect(new_disk_conf->al_extents >= DRBD_AL_EXTENTS_MIN))
1208 new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN;
1209 if (!expect(new_disk_conf->al_extents <= DRBD_AL_EXTENTS_MAX))
1210 new_disk_conf->al_extents = DRBD_AL_EXTENTS_MAX;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001211
1212 /* most sanity checks done, try to assign the new sync-after
1213 * dependency. need to hold the global lock in there,
1214 * to avoid a race in the dependency loop check. */
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001215 retcode = drbd_alter_sa(mdev, new_disk_conf->resync_after);
Lars Ellenbergf3990022011-03-23 14:31:09 +01001216 if (retcode != NO_ERROR)
1217 goto fail;
1218
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001219 fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001220 if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) {
1221 rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL);
1222 if (!rs_plan_s) {
1223 dev_err(DEV, "kmalloc of fifo_buffer failed");
1224 retcode = ERR_NOMEM;
1225 goto fail;
1226 }
1227 }
1228
1229 if (fifo_size != mdev->rs_plan_s.size) {
1230 kfree(mdev->rs_plan_s.values);
1231 mdev->rs_plan_s.values = rs_plan_s;
1232 mdev->rs_plan_s.size = fifo_size;
1233 mdev->rs_planed = 0;
1234 rs_plan_s = NULL;
1235 }
1236
1237 wait_event(mdev->al_wait, lc_try_lock(mdev->act_log));
1238 drbd_al_shrink(mdev);
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001239 err = drbd_check_al_size(mdev, new_disk_conf);
Lars Ellenbergf3990022011-03-23 14:31:09 +01001240 lc_unlock(mdev->act_log);
1241 wake_up(&mdev->al_wait);
1242
1243 if (err) {
1244 retcode = ERR_NOMEM;
1245 goto fail;
1246 }
1247
1248 /* FIXME
1249 * To avoid someone looking at a half-updated struct, we probably
1250 * should have a rw-semaphor on net_conf and disk_conf.
1251 */
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001252 mdev->ldev->dc = *new_disk_conf;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001253
1254 drbd_md_sync(mdev);
1255
1256
1257 if (mdev->state.conn >= C_CONNECTED)
1258 drbd_send_sync_param(mdev);
1259
1260 fail:
1261 put_ldev(mdev);
Lars Ellenberg5ecc72c2011-04-27 21:14:57 +02001262 kfree(new_disk_conf);
Lars Ellenbergf3990022011-03-23 14:31:09 +01001263 kfree(rs_plan_s);
1264 out:
1265 drbd_adm_finish(info, retcode);
1266 return 0;
1267}
1268
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001269int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001270{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001271 struct drbd_conf *mdev;
1272 int err;
Andreas Gruenbacher116676c2010-12-08 13:33:11 +01001273 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001274 enum determine_dev_size dd;
1275 sector_t max_possible_sectors;
1276 sector_t min_md_device_sectors;
1277 struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */
Tejun Heoe525fd82010-11-13 11:55:17 +01001278 struct block_device *bdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001279 struct lru_cache *resync_lru = NULL;
1280 union drbd_state ns, os;
Andreas Gruenbacherf2024e72010-12-10 13:44:05 +01001281 enum drbd_state_rv rv;
Philipp Reisner44ed1672011-04-19 17:10:19 +02001282 struct net_conf *nc;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001283 int cp_discovered = 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001284
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001285 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1286 if (!adm_ctx.reply_skb)
1287 return retcode;
1288 if (retcode != NO_ERROR)
Lars Ellenberg40cbf082011-03-16 16:52:10 +01001289 goto finish;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001290
1291 mdev = adm_ctx.mdev;
Philipp Reisner0e29d162011-02-18 14:23:11 +01001292 conn_reconfig_start(mdev->tconn);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001293
1294 /* if you want to reconfigure, please tear down first */
1295 if (mdev->state.disk > D_DISKLESS) {
1296 retcode = ERR_DISK_CONFIGURED;
1297 goto fail;
1298 }
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02001299 /* It may just now have detached because of IO error. Make sure
1300 * drbd_ldev_destroy is done already, we may end up here very fast,
1301 * e.g. if someone calls attach from the on-io-error handler,
1302 * to realize a "hot spare" feature (not that I'd recommend that) */
1303 wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001304
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001305 /* allocation not in the IO path, drbdsetup context */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001306 nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL);
1307 if (!nbc) {
1308 retcode = ERR_NOMEM;
1309 goto fail;
1310 }
1311
Lars Ellenberg5979e362011-04-27 21:09:55 +02001312 drbd_set_disk_conf_defaults(&nbc->dc);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001313
Lars Ellenbergf3990022011-03-23 14:31:09 +01001314 err = disk_conf_from_attrs(&nbc->dc, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001315 if (err) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001316 retcode = ERR_MANDATORY_TAG;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001317 drbd_msg_put_info(from_attrs_err_to_txt(err));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001318 goto fail;
1319 }
1320
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001321 if ((int)nbc->dc.meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001322 retcode = ERR_MD_IDX_INVALID;
1323 goto fail;
1324 }
1325
Philipp Reisner44ed1672011-04-19 17:10:19 +02001326 rcu_read_lock();
1327 nc = rcu_dereference(mdev->tconn->net_conf);
1328 if (nc) {
1329 if (nbc->dc.fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) {
1330 rcu_read_unlock();
Philipp Reisner47ff2d02010-06-18 13:56:57 +02001331 retcode = ERR_STONITH_AND_PROT_A;
1332 goto fail;
1333 }
1334 }
Philipp Reisner44ed1672011-04-19 17:10:19 +02001335 rcu_read_unlock();
Philipp Reisner47ff2d02010-06-18 13:56:57 +02001336
Tejun Heod4d77622010-11-13 11:55:18 +01001337 bdev = blkdev_get_by_path(nbc->dc.backing_dev,
1338 FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev);
Tejun Heoe525fd82010-11-13 11:55:17 +01001339 if (IS_ERR(bdev)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001340 dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.backing_dev,
Tejun Heoe525fd82010-11-13 11:55:17 +01001341 PTR_ERR(bdev));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001342 retcode = ERR_OPEN_DISK;
1343 goto fail;
1344 }
Tejun Heoe525fd82010-11-13 11:55:17 +01001345 nbc->backing_bdev = bdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001346
Tejun Heoe525fd82010-11-13 11:55:17 +01001347 /*
1348 * meta_dev_idx >= 0: external fixed size, possibly multiple
1349 * drbd sharing one meta device. TODO in that case, paranoia
1350 * check that [md_bdev, meta_dev_idx] is not yet used by some
1351 * other drbd minor! (if you use drbd.conf + drbdadm, that
1352 * should check it for you already; but if you don't, or
1353 * someone fooled it, we need to double check here)
1354 */
Tejun Heod4d77622010-11-13 11:55:18 +01001355 bdev = blkdev_get_by_path(nbc->dc.meta_dev,
1356 FMODE_READ | FMODE_WRITE | FMODE_EXCL,
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001357 ((int)nbc->dc.meta_dev_idx < 0) ?
Tejun Heod4d77622010-11-13 11:55:18 +01001358 (void *)mdev : (void *)drbd_m_holder);
Tejun Heoe525fd82010-11-13 11:55:17 +01001359 if (IS_ERR(bdev)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001360 dev_err(DEV, "open(\"%s\") failed with %ld\n", nbc->dc.meta_dev,
Tejun Heoe525fd82010-11-13 11:55:17 +01001361 PTR_ERR(bdev));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001362 retcode = ERR_OPEN_MD_DISK;
1363 goto fail;
1364 }
Tejun Heoe525fd82010-11-13 11:55:17 +01001365 nbc->md_bdev = bdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001366
Tejun Heoe525fd82010-11-13 11:55:17 +01001367 if ((nbc->backing_bdev == nbc->md_bdev) !=
1368 (nbc->dc.meta_dev_idx == DRBD_MD_INDEX_INTERNAL ||
1369 nbc->dc.meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) {
1370 retcode = ERR_MD_IDX_INVALID;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001371 goto fail;
1372 }
1373
1374 resync_lru = lc_create("resync", drbd_bm_ext_cache,
Lars Ellenberg46a15bc2011-02-21 13:21:01 +01001375 1, 61, sizeof(struct bm_extent),
Philipp Reisnerb411b362009-09-25 16:07:19 -07001376 offsetof(struct bm_extent, lce));
1377 if (!resync_lru) {
1378 retcode = ERR_NOMEM;
Tejun Heoe525fd82010-11-13 11:55:17 +01001379 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001380 }
1381
1382 /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */
1383 drbd_md_set_sector_offsets(mdev, nbc);
1384
1385 if (drbd_get_max_capacity(nbc) < nbc->dc.disk_size) {
1386 dev_err(DEV, "max capacity %llu smaller than disk size %llu\n",
1387 (unsigned long long) drbd_get_max_capacity(nbc),
1388 (unsigned long long) nbc->dc.disk_size);
1389 retcode = ERR_DISK_TO_SMALL;
Tejun Heoe525fd82010-11-13 11:55:17 +01001390 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001391 }
1392
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001393 if ((int)nbc->dc.meta_dev_idx < 0) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001394 max_possible_sectors = DRBD_MAX_SECTORS_FLEX;
1395 /* at least one MB, otherwise it does not make sense */
1396 min_md_device_sectors = (2<<10);
1397 } else {
1398 max_possible_sectors = DRBD_MAX_SECTORS;
1399 min_md_device_sectors = MD_RESERVED_SECT * (nbc->dc.meta_dev_idx + 1);
1400 }
1401
Philipp Reisnerb411b362009-09-25 16:07:19 -07001402 if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) {
1403 retcode = ERR_MD_DISK_TO_SMALL;
1404 dev_warn(DEV, "refusing attach: md-device too small, "
1405 "at least %llu sectors needed for this meta-disk type\n",
1406 (unsigned long long) min_md_device_sectors);
Tejun Heoe525fd82010-11-13 11:55:17 +01001407 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001408 }
1409
1410 /* Make sure the new disk is big enough
1411 * (we may currently be R_PRIMARY with no local disk...) */
1412 if (drbd_get_max_capacity(nbc) <
1413 drbd_get_capacity(mdev->this_bdev)) {
1414 retcode = ERR_DISK_TO_SMALL;
Tejun Heoe525fd82010-11-13 11:55:17 +01001415 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001416 }
1417
1418 nbc->known_size = drbd_get_capacity(nbc->backing_bdev);
1419
Lars Ellenberg13529942009-10-12 19:07:49 +02001420 if (nbc->known_size > max_possible_sectors) {
1421 dev_warn(DEV, "==> truncating very big lower level device "
1422 "to currently maximum possible %llu sectors <==\n",
1423 (unsigned long long) max_possible_sectors);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001424 if ((int)nbc->dc.meta_dev_idx >= 0)
Lars Ellenberg13529942009-10-12 19:07:49 +02001425 dev_warn(DEV, "==>> using internal or flexible "
1426 "meta data may help <<==\n");
1427 }
1428
Philipp Reisnerb411b362009-09-25 16:07:19 -07001429 drbd_suspend_io(mdev);
1430 /* also wait for the last barrier ack. */
Philipp Reisner2aebfab2011-03-28 16:48:11 +02001431 wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001432 /* and for any other previously queued work */
Philipp Reisnera21e9292011-02-08 15:08:49 +01001433 drbd_flush_workqueue(mdev);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001434
Andreas Gruenbacherf2024e72010-12-10 13:44:05 +01001435 rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE);
1436 retcode = rv; /* FIXME: Type mismatch. */
Philipp Reisnerb411b362009-09-25 16:07:19 -07001437 drbd_resume_io(mdev);
Andreas Gruenbacherf2024e72010-12-10 13:44:05 +01001438 if (rv < SS_SUCCESS)
Tejun Heoe525fd82010-11-13 11:55:17 +01001439 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001440
1441 if (!get_ldev_if_state(mdev, D_ATTACHING))
1442 goto force_diskless;
1443
1444 drbd_md_set_sector_offsets(mdev, nbc);
1445
1446 if (!mdev->bitmap) {
1447 if (drbd_bm_init(mdev)) {
1448 retcode = ERR_NOMEM;
1449 goto force_diskless_dec;
1450 }
1451 }
1452
1453 retcode = drbd_md_read(mdev, nbc);
1454 if (retcode != NO_ERROR)
1455 goto force_diskless_dec;
1456
1457 if (mdev->state.conn < C_CONNECTED &&
1458 mdev->state.role == R_PRIMARY &&
1459 (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) {
1460 dev_err(DEV, "Can only attach to data with current UUID=%016llX\n",
1461 (unsigned long long)mdev->ed_uuid);
1462 retcode = ERR_DATA_NOT_CURRENT;
1463 goto force_diskless_dec;
1464 }
1465
1466 /* Since we are diskless, fix the activity log first... */
Lars Ellenbergf3990022011-03-23 14:31:09 +01001467 if (drbd_check_al_size(mdev, &nbc->dc)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001468 retcode = ERR_NOMEM;
1469 goto force_diskless_dec;
1470 }
1471
1472 /* Prevent shrinking of consistent devices ! */
1473 if (drbd_md_test_flag(nbc, MDF_CONSISTENT) &&
Philipp Reisnera393db62009-12-22 13:35:52 +01001474 drbd_new_dev_size(mdev, nbc, 0) < nbc->md.la_size_sect) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001475 dev_warn(DEV, "refusing to truncate a consistent device\n");
1476 retcode = ERR_DISK_TO_SMALL;
1477 goto force_diskless_dec;
1478 }
1479
1480 if (!drbd_al_read_log(mdev, nbc)) {
1481 retcode = ERR_IO_MD_DISK;
1482 goto force_diskless_dec;
1483 }
1484
Philipp Reisnerb411b362009-09-25 16:07:19 -07001485 /* Reset the "barriers don't work" bits here, then force meta data to
1486 * be written, to ensure we determine if barriers are supported. */
1487 if (nbc->dc.no_md_flush)
Philipp Reisnera8a4e512010-08-25 10:21:04 +02001488 set_bit(MD_NO_FUA, &mdev->flags);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001489 else
Philipp Reisnera8a4e512010-08-25 10:21:04 +02001490 clear_bit(MD_NO_FUA, &mdev->flags);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001491
1492 /* Point of no return reached.
1493 * Devices and memory are no longer released by error cleanup below.
1494 * now mdev takes over responsibility, and the state engine should
1495 * clean it up somewhere. */
1496 D_ASSERT(mdev->ldev == NULL);
1497 mdev->ldev = nbc;
1498 mdev->resync = resync_lru;
1499 nbc = NULL;
1500 resync_lru = NULL;
1501
Philipp Reisner2451fc32010-08-24 13:43:11 +02001502 mdev->write_ordering = WO_bdev_flush;
1503 drbd_bump_write_ordering(mdev, WO_bdev_flush);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001504
1505 if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY))
1506 set_bit(CRASHED_PRIMARY, &mdev->flags);
1507 else
1508 clear_bit(CRASHED_PRIMARY, &mdev->flags);
1509
Philipp Reisner894c6a92010-06-18 16:03:20 +02001510 if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
Philipp Reisnerda9fbc22011-03-29 10:52:01 +02001511 !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001512 set_bit(CRASHED_PRIMARY, &mdev->flags);
1513 cp_discovered = 1;
1514 }
1515
1516 mdev->send_cnt = 0;
1517 mdev->recv_cnt = 0;
1518 mdev->read_cnt = 0;
1519 mdev->writ_cnt = 0;
1520
Philipp Reisner99432fc2011-05-20 16:39:13 +02001521 drbd_reconsider_max_bio_size(mdev);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001522
1523 /* If I am currently not R_PRIMARY,
1524 * but meta data primary indicator is set,
1525 * I just now recover from a hard crash,
1526 * and have been R_PRIMARY before that crash.
1527 *
1528 * Now, if I had no connection before that crash
1529 * (have been degraded R_PRIMARY), chances are that
1530 * I won't find my peer now either.
1531 *
1532 * In that case, and _only_ in that case,
1533 * we use the degr-wfc-timeout instead of the default,
1534 * so we can automatically recover from a crash of a
1535 * degraded but active "cluster" after a certain timeout.
1536 */
1537 clear_bit(USE_DEGR_WFC_T, &mdev->flags);
1538 if (mdev->state.role != R_PRIMARY &&
1539 drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) &&
1540 !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND))
1541 set_bit(USE_DEGR_WFC_T, &mdev->flags);
1542
Bart Van Assche24c48302011-05-21 18:32:29 +02001543 dd = drbd_determine_dev_size(mdev, 0);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001544 if (dd == dev_size_error) {
1545 retcode = ERR_NOMEM_BITMAP;
1546 goto force_diskless_dec;
1547 } else if (dd == grew)
1548 set_bit(RESYNC_AFTER_NEG, &mdev->flags);
1549
1550 if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) {
1551 dev_info(DEV, "Assuming that all blocks are out of sync "
1552 "(aka FullSync)\n");
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +01001553 if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write,
1554 "set_n_write from attaching", BM_LOCKED_MASK)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001555 retcode = ERR_IO_MD_DISK;
1556 goto force_diskless_dec;
1557 }
1558 } else {
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +01001559 if (drbd_bitmap_io(mdev, &drbd_bm_read,
Andreas Gruenbacher22ab6a32010-12-13 01:44:11 +01001560 "read from attaching", BM_LOCKED_MASK)) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07001561 retcode = ERR_IO_MD_DISK;
1562 goto force_diskless_dec;
1563 }
1564 }
1565
1566 if (cp_discovered) {
1567 drbd_al_apply_to_bm(mdev);
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +01001568 if (drbd_bitmap_io(mdev, &drbd_bm_write,
1569 "crashed primary apply AL", BM_LOCKED_MASK)) {
Lars Ellenberg19f843a2010-12-15 08:59:11 +01001570 retcode = ERR_IO_MD_DISK;
1571 goto force_diskless_dec;
1572 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07001573 }
1574
Philipp Reisner07782862010-08-31 12:00:50 +02001575 if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev))
1576 drbd_suspend_al(mdev); /* IO is still suspended here... */
1577
Philipp Reisner87eeee42011-01-19 14:16:30 +01001578 spin_lock_irq(&mdev->tconn->req_lock);
Philipp Reisner78bae592011-03-28 15:40:12 +02001579 os = drbd_read_state(mdev);
1580 ns = os;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001581 /* If MDF_CONSISTENT is not set go into inconsistent state,
1582 otherwise investigate MDF_WasUpToDate...
1583 If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state,
1584 otherwise into D_CONSISTENT state.
1585 */
1586 if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) {
1587 if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE))
1588 ns.disk = D_CONSISTENT;
1589 else
1590 ns.disk = D_OUTDATED;
1591 } else {
1592 ns.disk = D_INCONSISTENT;
1593 }
1594
1595 if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED))
1596 ns.pdsk = D_OUTDATED;
1597
1598 if ( ns.disk == D_CONSISTENT &&
1599 (ns.pdsk == D_OUTDATED || mdev->ldev->dc.fencing == FP_DONT_CARE))
1600 ns.disk = D_UP_TO_DATE;
1601
1602 /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND,
1603 MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before
1604 this point, because drbd_request_state() modifies these
1605 flags. */
1606
1607 /* In case we are C_CONNECTED postpone any decision on the new disk
1608 state after the negotiation phase. */
1609 if (mdev->state.conn == C_CONNECTED) {
1610 mdev->new_state_tmp.i = ns.i;
1611 ns.i = os.i;
1612 ns.disk = D_NEGOTIATING;
Philipp Reisnerdc66c742010-06-02 14:31:29 +02001613
1614 /* We expect to receive up-to-date UUIDs soon.
1615 To avoid a race in receive_state, free p_uuid while
1616 holding req_lock. I.e. atomic with the state change */
1617 kfree(mdev->p_uuid);
1618 mdev->p_uuid = NULL;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001619 }
1620
1621 rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL);
Philipp Reisner87eeee42011-01-19 14:16:30 +01001622 spin_unlock_irq(&mdev->tconn->req_lock);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001623
1624 if (rv < SS_SUCCESS)
1625 goto force_diskless_dec;
1626
1627 if (mdev->state.role == R_PRIMARY)
1628 mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1;
1629 else
1630 mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1;
1631
1632 drbd_md_mark_dirty(mdev);
1633 drbd_md_sync(mdev);
1634
1635 kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE);
1636 put_ldev(mdev);
Philipp Reisner0e29d162011-02-18 14:23:11 +01001637 conn_reconfig_done(mdev->tconn);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001638 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001639 return 0;
1640
1641 force_diskless_dec:
1642 put_ldev(mdev);
1643 force_diskless:
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02001644 drbd_force_state(mdev, NS(disk, D_FAILED));
Philipp Reisnerb411b362009-09-25 16:07:19 -07001645 drbd_md_sync(mdev);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001646 fail:
Lars Ellenberg40cbf082011-03-16 16:52:10 +01001647 conn_reconfig_done(mdev->tconn);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001648 if (nbc) {
Tejun Heoe525fd82010-11-13 11:55:17 +01001649 if (nbc->backing_bdev)
1650 blkdev_put(nbc->backing_bdev,
1651 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
1652 if (nbc->md_bdev)
1653 blkdev_put(nbc->md_bdev,
1654 FMODE_READ | FMODE_WRITE | FMODE_EXCL);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001655 kfree(nbc);
1656 }
1657 lc_destroy(resync_lru);
1658
Lars Ellenberg40cbf082011-03-16 16:52:10 +01001659 finish:
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001660 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001661 return 0;
1662}
1663
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01001664static int adm_detach(struct drbd_conf *mdev)
1665{
Philipp Reisner19f83c72011-03-29 14:21:03 +02001666 enum drbd_state_rv retcode;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01001667 drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */
1668 retcode = drbd_request_state(mdev, NS(disk, D_DISKLESS));
1669 wait_event(mdev->misc_wait,
1670 mdev->state.disk != D_DISKLESS ||
1671 !atomic_read(&mdev->local_cnt));
1672 drbd_resume_io(mdev);
1673 return retcode;
1674}
1675
Lars Ellenberg82f59cc2010-10-16 12:13:47 +02001676/* Detaching the disk is a process in multiple stages. First we need to lock
1677 * out application IO, in-flight IO, IO stuck in drbd_al_begin_io.
1678 * Then we transition to D_DISKLESS, and wait for put_ldev() to return all
1679 * internal references as well.
1680 * Only then we have finally detached. */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001681int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001682{
Lars Ellenberg9a0d9d02011-05-02 11:51:31 +02001683 enum drbd_ret_code retcode;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001684
1685 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
1686 if (!adm_ctx.reply_skb)
1687 return retcode;
1688 if (retcode != NO_ERROR)
1689 goto out;
1690
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01001691 retcode = adm_detach(adm_ctx.mdev);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001692out:
1693 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07001694 return 0;
1695}
1696
Lars Ellenbergf3990022011-03-23 14:31:09 +01001697static bool conn_resync_running(struct drbd_tconn *tconn)
1698{
1699 struct drbd_conf *mdev;
Philipp Reisner695d08f2011-04-11 22:53:32 -07001700 bool rv = false;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001701 int vnr;
1702
Philipp Reisner695d08f2011-04-11 22:53:32 -07001703 rcu_read_lock();
Lars Ellenbergf3990022011-03-23 14:31:09 +01001704 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1705 if (mdev->state.conn == C_SYNC_SOURCE ||
1706 mdev->state.conn == C_SYNC_TARGET ||
1707 mdev->state.conn == C_PAUSED_SYNC_S ||
Philipp Reisner695d08f2011-04-11 22:53:32 -07001708 mdev->state.conn == C_PAUSED_SYNC_T) {
1709 rv = true;
1710 break;
1711 }
Lars Ellenbergf3990022011-03-23 14:31:09 +01001712 }
Philipp Reisner695d08f2011-04-11 22:53:32 -07001713 rcu_read_unlock();
1714
1715 return rv;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001716}
1717
1718static bool conn_ov_running(struct drbd_tconn *tconn)
1719{
1720 struct drbd_conf *mdev;
Philipp Reisner695d08f2011-04-11 22:53:32 -07001721 bool rv = false;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001722 int vnr;
1723
Philipp Reisner695d08f2011-04-11 22:53:32 -07001724 rcu_read_lock();
Lars Ellenbergf3990022011-03-23 14:31:09 +01001725 idr_for_each_entry(&tconn->volumes, mdev, vnr) {
1726 if (mdev->state.conn == C_VERIFY_S ||
Philipp Reisner695d08f2011-04-11 22:53:32 -07001727 mdev->state.conn == C_VERIFY_T) {
1728 rv = true;
1729 break;
1730 }
Lars Ellenbergf3990022011-03-23 14:31:09 +01001731 }
Philipp Reisner695d08f2011-04-11 22:53:32 -07001732 rcu_read_unlock();
1733
1734 return rv;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001735}
1736
Philipp Reisnercd643972011-04-13 18:00:59 -07001737static enum drbd_ret_code
Philipp Reisner44ed1672011-04-19 17:10:19 +02001738_check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf)
Philipp Reisnercd643972011-04-13 18:00:59 -07001739{
1740 struct drbd_conf *mdev;
1741 int i;
1742
Philipp Reisner44ed1672011-04-19 17:10:19 +02001743 if (old_conf && tconn->agreed_pro_version < 100 &&
Philipp Reisnerb032b6f2011-04-13 18:16:10 -07001744 tconn->cstate == C_WF_REPORT_PARAMS &&
Philipp Reisner44ed1672011-04-19 17:10:19 +02001745 new_conf->wire_protocol != old_conf->wire_protocol)
Philipp Reisnerb032b6f2011-04-13 18:16:10 -07001746 return ERR_NEED_APV_100;
1747
Philipp Reisnercd643972011-04-13 18:00:59 -07001748 if (new_conf->two_primaries &&
1749 (new_conf->wire_protocol != DRBD_PROT_C))
1750 return ERR_NOT_PROTO_C;
1751
Philipp Reisnercd643972011-04-13 18:00:59 -07001752 idr_for_each_entry(&tconn->volumes, mdev, i) {
1753 if (get_ldev(mdev)) {
1754 enum drbd_fencing_p fp = mdev->ldev->dc.fencing;
1755 put_ldev(mdev);
Philipp Reisner44ed1672011-04-19 17:10:19 +02001756 if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH)
Philipp Reisnercd643972011-04-13 18:00:59 -07001757 return ERR_STONITH_AND_PROT_A;
Philipp Reisnercd643972011-04-13 18:00:59 -07001758 }
Philipp Reisner44ed1672011-04-19 17:10:19 +02001759 if (mdev->state.role == R_PRIMARY && new_conf->want_lose)
Philipp Reisnercd643972011-04-13 18:00:59 -07001760 return ERR_DISCARD;
Philipp Reisnercd643972011-04-13 18:00:59 -07001761 }
Philipp Reisnercd643972011-04-13 18:00:59 -07001762
1763 if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A)
1764 return ERR_CONG_NOT_PROTO_A;
1765
1766 return NO_ERROR;
1767}
1768
Philipp Reisner44ed1672011-04-19 17:10:19 +02001769static enum drbd_ret_code
1770check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf)
1771{
1772 static enum drbd_ret_code rv;
1773 struct drbd_conf *mdev;
1774 int i;
1775
1776 rcu_read_lock();
1777 rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf);
1778 rcu_read_unlock();
1779
1780 /* tconn->volumes protected by genl_lock() here */
1781 idr_for_each_entry(&tconn->volumes, mdev, i) {
1782 if (!mdev->bitmap) {
1783 if(drbd_bm_init(mdev))
1784 return ERR_NOMEM;
1785 }
1786 }
1787
1788 return rv;
1789}
1790
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001791struct crypto {
1792 struct crypto_hash *verify_tfm;
1793 struct crypto_hash *csums_tfm;
1794 struct crypto_hash *cram_hmac_tfm;
Andreas Gruenbacher8d412fc2011-04-27 20:59:18 +02001795 struct crypto_hash *integrity_tfm;
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001796 void *int_dig_in;
1797 void *int_dig_vv;
1798};
1799
1800static int
Andreas Gruenbacher4b6ad6d2011-04-29 10:20:08 +02001801alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg)
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001802{
1803 if (!tfm_name[0])
1804 return NO_ERROR;
1805
1806 *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC);
1807 if (IS_ERR(*tfm)) {
1808 *tfm = NULL;
1809 return err_alg;
1810 }
1811
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001812 return NO_ERROR;
1813}
1814
1815static enum drbd_ret_code
1816alloc_crypto(struct crypto *crypto, struct net_conf *new_conf)
1817{
1818 char hmac_name[CRYPTO_MAX_ALG_NAME];
1819 enum drbd_ret_code rv;
1820 int hash_size;
1821
Andreas Gruenbacher4b6ad6d2011-04-29 10:20:08 +02001822 rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg,
1823 ERR_CSUMS_ALG);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001824 if (rv != NO_ERROR)
1825 return rv;
Andreas Gruenbacher4b6ad6d2011-04-29 10:20:08 +02001826 rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg,
1827 ERR_VERIFY_ALG);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001828 if (rv != NO_ERROR)
1829 return rv;
Andreas Gruenbacher4b6ad6d2011-04-29 10:20:08 +02001830 rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg,
1831 ERR_INTEGRITY_ALG);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001832 if (rv != NO_ERROR)
1833 return rv;
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001834 if (new_conf->cram_hmac_alg[0] != 0) {
1835 snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)",
1836 new_conf->cram_hmac_alg);
1837
Andreas Gruenbacher4b6ad6d2011-04-29 10:20:08 +02001838 rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name,
1839 ERR_AUTH_ALG);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001840 }
Andreas Gruenbacher8d412fc2011-04-27 20:59:18 +02001841 if (crypto->integrity_tfm) {
1842 hash_size = crypto_hash_digestsize(crypto->integrity_tfm);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001843 crypto->int_dig_in = kmalloc(hash_size, GFP_KERNEL);
1844 if (!crypto->int_dig_in)
1845 return ERR_NOMEM;
1846 crypto->int_dig_vv = kmalloc(hash_size, GFP_KERNEL);
1847 if (!crypto->int_dig_vv)
1848 return ERR_NOMEM;
1849 }
1850
1851 return rv;
1852}
1853
1854static void free_crypto(struct crypto *crypto)
1855{
1856 kfree(crypto->int_dig_in);
1857 kfree(crypto->int_dig_vv);
1858 crypto_free_hash(crypto->cram_hmac_tfm);
Andreas Gruenbacher8d412fc2011-04-27 20:59:18 +02001859 crypto_free_hash(crypto->integrity_tfm);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001860 crypto_free_hash(crypto->csums_tfm);
1861 crypto_free_hash(crypto->verify_tfm);
1862}
1863
Lars Ellenbergf3990022011-03-23 14:31:09 +01001864int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info)
1865{
1866 enum drbd_ret_code retcode;
1867 struct drbd_tconn *tconn;
Philipp Reisner44ed1672011-04-19 17:10:19 +02001868 struct net_conf *old_conf, *new_conf = NULL;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001869 int err;
1870 int ovr; /* online verify running */
1871 int rsr; /* re-sync running */
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001872 struct crypto crypto = { };
Andreas Gruenbacher88104ca2011-04-28 21:47:21 +02001873 bool change_integrity_alg;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001874
1875 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
1876 if (!adm_ctx.reply_skb)
1877 return retcode;
1878 if (retcode != NO_ERROR)
1879 goto out;
1880
1881 tconn = adm_ctx.tconn;
1882
1883 new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL);
1884 if (!new_conf) {
1885 retcode = ERR_NOMEM;
1886 goto out;
1887 }
1888
Lars Ellenbergf3990022011-03-23 14:31:09 +01001889 conn_reconfig_start(tconn);
1890
Andreas Gruenbacher88104ca2011-04-28 21:47:21 +02001891 mutex_lock(&tconn->data.mutex);
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001892 mutex_lock(&tconn->net_conf_update);
1893 old_conf = tconn->net_conf;
Philipp Reisner44ed1672011-04-19 17:10:19 +02001894
1895 if (!old_conf) {
1896 drbd_msg_put_info("net conf missing, try connect");
1897 retcode = ERR_INVALID_REQUEST;
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001898 goto fail;
Philipp Reisner44ed1672011-04-19 17:10:19 +02001899 }
1900
1901 *new_conf = *old_conf;
Lars Ellenberg5979e362011-04-27 21:09:55 +02001902 if (should_set_defaults(info))
1903 drbd_set_net_conf_defaults(new_conf);
Philipp Reisner44ed1672011-04-19 17:10:19 +02001904
Lars Ellenbergf3990022011-03-23 14:31:09 +01001905 err = net_conf_from_attrs_for_change(new_conf, info);
1906 if (err) {
1907 retcode = ERR_MANDATORY_TAG;
1908 drbd_msg_put_info(from_attrs_err_to_txt(err));
1909 goto fail;
1910 }
1911
Philipp Reisnercd643972011-04-13 18:00:59 -07001912 retcode = check_net_options(tconn, new_conf);
1913 if (retcode != NO_ERROR)
1914 goto fail;
1915
Lars Ellenbergf3990022011-03-23 14:31:09 +01001916 /* re-sync running */
1917 rsr = conn_resync_running(tconn);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001918 if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) {
Lars Ellenbergf3990022011-03-23 14:31:09 +01001919 retcode = ERR_CSUMS_RESYNC_RUNNING;
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001920 goto fail;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001921 }
1922
Lars Ellenbergf3990022011-03-23 14:31:09 +01001923 /* online verify running */
1924 ovr = conn_ov_running(tconn);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001925 if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) {
1926 retcode = ERR_VERIFY_RUNNING;
1927 goto fail;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001928 }
1929
Andreas Gruenbacher88104ca2011-04-28 21:47:21 +02001930 change_integrity_alg = strcmp(old_conf->integrity_alg,
1931 new_conf->integrity_alg);
1932
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001933 retcode = alloc_crypto(&crypto, new_conf);
1934 if (retcode != NO_ERROR)
1935 goto fail;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001936
Philipp Reisner44ed1672011-04-19 17:10:19 +02001937 rcu_assign_pointer(tconn->net_conf, new_conf);
Lars Ellenbergf3990022011-03-23 14:31:09 +01001938
1939 if (!rsr) {
1940 crypto_free_hash(tconn->csums_tfm);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001941 tconn->csums_tfm = crypto.csums_tfm;
1942 crypto.csums_tfm = NULL;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001943 }
1944 if (!ovr) {
1945 crypto_free_hash(tconn->verify_tfm);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001946 tconn->verify_tfm = crypto.verify_tfm;
1947 crypto.verify_tfm = NULL;
Lars Ellenbergf3990022011-03-23 14:31:09 +01001948 }
1949
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001950 kfree(tconn->int_dig_in);
1951 tconn->int_dig_in = crypto.int_dig_in;
1952 kfree(tconn->int_dig_vv);
1953 tconn->int_dig_vv = crypto.int_dig_vv;
Andreas Gruenbacher8d412fc2011-04-27 20:59:18 +02001954 crypto_free_hash(tconn->integrity_tfm);
1955 tconn->integrity_tfm = crypto.integrity_tfm;
Andreas Gruenbacher88104ca2011-04-28 21:47:21 +02001956 if (change_integrity_alg) {
1957 /* Do this without trying to take tconn->data.mutex again. */
1958 if (__drbd_send_protocol(tconn))
1959 goto fail;
1960 }
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001961
1962 /* FIXME Changing cram_hmac while the connection is established is useless */
1963 crypto_free_hash(tconn->cram_hmac_tfm);
1964 tconn->cram_hmac_tfm = crypto.cram_hmac_tfm;
1965
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001966 mutex_unlock(&tconn->net_conf_update);
Andreas Gruenbacher88104ca2011-04-28 21:47:21 +02001967 mutex_unlock(&tconn->data.mutex);
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001968 synchronize_rcu();
1969 kfree(old_conf);
1970
Lars Ellenbergf3990022011-03-23 14:31:09 +01001971 if (tconn->cstate >= C_WF_REPORT_PARAMS)
1972 drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn)));
1973
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001974 goto done;
1975
Lars Ellenbergf3990022011-03-23 14:31:09 +01001976 fail:
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001977 mutex_unlock(&tconn->net_conf_update);
Andreas Gruenbacher88104ca2011-04-28 21:47:21 +02001978 mutex_unlock(&tconn->data.mutex);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001979 free_crypto(&crypto);
Lars Ellenbergf3990022011-03-23 14:31:09 +01001980 kfree(new_conf);
Philipp Reisner91fd4da2011-04-20 17:47:29 +02001981 done:
Lars Ellenbergf3990022011-03-23 14:31:09 +01001982 conn_reconfig_done(tconn);
1983 out:
1984 drbd_adm_finish(info, retcode);
1985 return 0;
1986}
1987
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001988int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07001989{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001990 struct drbd_conf *mdev;
Philipp Reisner44ed1672011-04-19 17:10:19 +02001991 struct net_conf *old_conf, *new_conf = NULL;
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02001992 struct crypto crypto = { };
Philipp Reisner80883192011-02-18 14:56:45 +01001993 struct drbd_tconn *oconn;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001994 struct drbd_tconn *tconn;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001995 struct sockaddr *new_my_addr, *new_peer_addr, *taken_addr;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01001996 enum drbd_ret_code retcode;
1997 int i;
1998 int err;
Philipp Reisnerb411b362009-09-25 16:07:19 -07001999
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002000 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
2001 if (!adm_ctx.reply_skb)
2002 return retcode;
2003 if (retcode != NO_ERROR)
2004 goto out;
2005
2006 tconn = adm_ctx.tconn;
Philipp Reisner80883192011-02-18 14:56:45 +01002007 conn_reconfig_start(tconn);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002008
Philipp Reisner80883192011-02-18 14:56:45 +01002009 if (tconn->cstate > C_STANDALONE) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07002010 retcode = ERR_NET_CONFIGURED;
2011 goto fail;
2012 }
2013
2014 /* allocation not in the IO path, cqueue thread context */
Lars Ellenberg5979e362011-04-27 21:09:55 +02002015 new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002016 if (!new_conf) {
2017 retcode = ERR_NOMEM;
2018 goto fail;
2019 }
2020
Lars Ellenberg5979e362011-04-27 21:09:55 +02002021 drbd_set_net_conf_defaults(new_conf);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002022
Lars Ellenbergf3990022011-03-23 14:31:09 +01002023 err = net_conf_from_attrs(new_conf, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002024 if (err) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07002025 retcode = ERR_MANDATORY_TAG;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002026 drbd_msg_put_info(from_attrs_err_to_txt(err));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002027 goto fail;
2028 }
2029
Philipp Reisnercd643972011-04-13 18:00:59 -07002030 retcode = check_net_options(tconn, new_conf);
2031 if (retcode != NO_ERROR)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002032 goto fail;
Philipp Reisner422028b2010-10-27 11:12:07 +02002033
Philipp Reisnerb411b362009-09-25 16:07:19 -07002034 retcode = NO_ERROR;
2035
2036 new_my_addr = (struct sockaddr *)&new_conf->my_addr;
2037 new_peer_addr = (struct sockaddr *)&new_conf->peer_addr;
Lars Ellenberg543cc102011-03-10 22:18:18 +01002038
Philipp Reisneref356262011-04-13 14:21:29 -07002039 /* No need to take drbd_cfg_rwsem here. All reconfiguration is
Lars Ellenberg543cc102011-03-10 22:18:18 +01002040 * strictly serialized on genl_lock(). We are protected against
2041 * concurrent reconfiguration/addition/deletion */
Philipp Reisner80883192011-02-18 14:56:45 +01002042 list_for_each_entry(oconn, &drbd_tconns, all_tconn) {
Philipp Reisner44ed1672011-04-19 17:10:19 +02002043 struct net_conf *nc;
Philipp Reisner80883192011-02-18 14:56:45 +01002044 if (oconn == tconn)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002045 continue;
Philipp Reisner44ed1672011-04-19 17:10:19 +02002046
2047 rcu_read_lock();
2048 nc = rcu_dereference(oconn->net_conf);
2049 if (nc) {
2050 taken_addr = (struct sockaddr *)&nc->my_addr;
2051 if (new_conf->my_addr_len == nc->my_addr_len &&
Philipp Reisnerb411b362009-09-25 16:07:19 -07002052 !memcmp(new_my_addr, taken_addr, new_conf->my_addr_len))
2053 retcode = ERR_LOCAL_ADDR;
2054
Philipp Reisner44ed1672011-04-19 17:10:19 +02002055 taken_addr = (struct sockaddr *)&nc->peer_addr;
2056 if (new_conf->peer_addr_len == nc->peer_addr_len &&
Philipp Reisnerb411b362009-09-25 16:07:19 -07002057 !memcmp(new_peer_addr, taken_addr, new_conf->peer_addr_len))
2058 retcode = ERR_PEER_ADDR;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002059 }
Philipp Reisner44ed1672011-04-19 17:10:19 +02002060 rcu_read_unlock();
2061 if (retcode != NO_ERROR)
2062 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002063 }
2064
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02002065 retcode = alloc_crypto(&crypto, new_conf);
2066 if (retcode != NO_ERROR)
2067 goto fail;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002068
Philipp Reisnerb411b362009-09-25 16:07:19 -07002069 ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0;
2070
Philipp Reisner80883192011-02-18 14:56:45 +01002071 conn_flush_workqueue(tconn);
Philipp Reisner91fd4da2011-04-20 17:47:29 +02002072
2073 mutex_lock(&tconn->net_conf_update);
2074 old_conf = tconn->net_conf;
2075 if (old_conf) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07002076 retcode = ERR_NET_CONFIGURED;
Philipp Reisner91fd4da2011-04-20 17:47:29 +02002077 mutex_unlock(&tconn->net_conf_update);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002078 goto fail;
2079 }
Philipp Reisner44ed1672011-04-19 17:10:19 +02002080 rcu_assign_pointer(tconn->net_conf, new_conf);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002081
Philipp Reisner91fd4da2011-04-20 17:47:29 +02002082 conn_free_crypto(tconn);
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02002083 tconn->int_dig_in = crypto.int_dig_in;
2084 tconn->int_dig_vv = crypto.int_dig_vv;
2085 tconn->cram_hmac_tfm = crypto.cram_hmac_tfm;
Andreas Gruenbacher8d412fc2011-04-27 20:59:18 +02002086 tconn->integrity_tfm = crypto.integrity_tfm;
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02002087 tconn->csums_tfm = crypto.csums_tfm;
2088 tconn->verify_tfm = crypto.verify_tfm;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002089
Philipp Reisner91fd4da2011-04-20 17:47:29 +02002090 mutex_unlock(&tconn->net_conf_update);
2091
Philipp Reisner695d08f2011-04-11 22:53:32 -07002092 rcu_read_lock();
Philipp Reisner80883192011-02-18 14:56:45 +01002093 idr_for_each_entry(&tconn->volumes, mdev, i) {
2094 mdev->send_cnt = 0;
2095 mdev->recv_cnt = 0;
Philipp Reisner80883192011-02-18 14:56:45 +01002096 }
Philipp Reisner695d08f2011-04-11 22:53:32 -07002097 rcu_read_unlock();
Lars Ellenberg5ee743e2011-04-26 16:22:25 +02002098
2099 retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE);
2100
Philipp Reisner80883192011-02-18 14:56:45 +01002101 conn_reconfig_done(tconn);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002102 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002103 return 0;
2104
2105fail:
Philipp Reisner0fd0ea02011-04-27 11:27:47 +02002106 free_crypto(&crypto);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002107 kfree(new_conf);
2108
Philipp Reisner80883192011-02-18 14:56:45 +01002109 conn_reconfig_done(tconn);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002110out:
2111 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002112 return 0;
2113}
2114
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002115static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force)
2116{
2117 enum drbd_state_rv rv;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002118
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02002119 rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING),
2120 force ? CS_HARD : 0);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002121
2122 switch (rv) {
2123 case SS_NOTHING_TO_DO:
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02002124 break;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002125 case SS_ALREADY_STANDALONE:
2126 return SS_SUCCESS;
2127 case SS_PRIMARY_NOP:
2128 /* Our state checking code wants to see the peer outdated. */
2129 rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING,
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02002130 pdsk, D_OUTDATED), CS_VERBOSE);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002131 break;
2132 case SS_CW_FAILED_BY_PEER:
2133 /* The peer probably wants to see us outdated. */
2134 rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING,
2135 disk, D_OUTDATED), 0);
2136 if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) {
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02002137 rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING),
2138 CS_HARD);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002139 }
2140 break;
2141 default:;
2142 /* no special handling necessary */
2143 }
2144
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02002145 if (rv >= SS_SUCCESS) {
2146 enum drbd_state_rv rv2;
2147 /* No one else can reconfigure the network while I am here.
2148 * The state handling only uses drbd_thread_stop_nowait(),
2149 * we want to really wait here until the receiver is no more.
2150 */
2151 drbd_thread_stop(&adm_ctx.tconn->receiver);
2152
2153 /* Race breaker. This additional state change request may be
2154 * necessary, if this was a forced disconnect during a receiver
2155 * restart. We may have "killed" the receiver thread just
2156 * after drbdd_init() returned. Typically, we should be
2157 * C_STANDALONE already, now, and this becomes a no-op.
2158 */
2159 rv2 = conn_request_state(tconn, NS(conn, C_STANDALONE),
2160 CS_VERBOSE | CS_HARD);
2161 if (rv2 < SS_SUCCESS)
2162 conn_err(tconn,
2163 "unexpected rv2=%d in conn_try_disconnect()\n",
2164 rv2);
2165 }
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002166 return rv;
2167}
2168
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002169int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002170{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002171 struct disconnect_parms parms;
2172 struct drbd_tconn *tconn;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002173 enum drbd_state_rv rv;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002174 enum drbd_ret_code retcode;
2175 int err;
Philipp Reisner2561b9c2010-12-03 15:22:48 +01002176
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002177 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
2178 if (!adm_ctx.reply_skb)
2179 return retcode;
2180 if (retcode != NO_ERROR)
Philipp Reisner2561b9c2010-12-03 15:22:48 +01002181 goto fail;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002182
2183 tconn = adm_ctx.tconn;
2184 memset(&parms, 0, sizeof(parms));
2185 if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) {
Lars Ellenbergf3990022011-03-23 14:31:09 +01002186 err = disconnect_parms_from_attrs(&parms, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002187 if (err) {
2188 retcode = ERR_MANDATORY_TAG;
2189 drbd_msg_put_info(from_attrs_err_to_txt(err));
2190 goto fail;
2191 }
Philipp Reisner2561b9c2010-12-03 15:22:48 +01002192 }
2193
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01002194 rv = conn_try_disconnect(tconn, parms.force_disconnect);
2195 if (rv < SS_SUCCESS)
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02002196 retcode = rv; /* FIXME: Type mismatch. */
2197 else
2198 retcode = NO_ERROR;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002199 fail:
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002200 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002201 return 0;
2202}
2203
2204void resync_after_online_grow(struct drbd_conf *mdev)
2205{
2206 int iass; /* I am sync source */
2207
2208 dev_info(DEV, "Resync of new storage after online grow\n");
2209 if (mdev->state.role != mdev->state.peer)
2210 iass = (mdev->state.role == R_PRIMARY);
2211 else
Philipp Reisner25703f82011-02-07 14:35:25 +01002212 iass = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002213
2214 if (iass)
2215 drbd_start_resync(mdev, C_SYNC_SOURCE);
2216 else
2217 _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE);
2218}
2219
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002220int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002221{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002222 struct resize_parms rs;
2223 struct drbd_conf *mdev;
2224 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002225 enum determine_dev_size dd;
Philipp Reisner6495d2c2010-03-24 16:07:04 +01002226 enum dds_flags ddsf;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002227 int err;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002228
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002229 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2230 if (!adm_ctx.reply_skb)
2231 return retcode;
2232 if (retcode != NO_ERROR)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002233 goto fail;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002234
2235 memset(&rs, 0, sizeof(struct resize_parms));
2236 if (info->attrs[DRBD_NLA_RESIZE_PARMS]) {
Lars Ellenbergf3990022011-03-23 14:31:09 +01002237 err = resize_parms_from_attrs(&rs, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002238 if (err) {
2239 retcode = ERR_MANDATORY_TAG;
2240 drbd_msg_put_info(from_attrs_err_to_txt(err));
2241 goto fail;
2242 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07002243 }
2244
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002245 mdev = adm_ctx.mdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002246 if (mdev->state.conn > C_CONNECTED) {
2247 retcode = ERR_RESIZE_RESYNC;
2248 goto fail;
2249 }
2250
2251 if (mdev->state.role == R_SECONDARY &&
2252 mdev->state.peer == R_SECONDARY) {
2253 retcode = ERR_NO_PRIMARY;
2254 goto fail;
2255 }
2256
2257 if (!get_ldev(mdev)) {
2258 retcode = ERR_NO_DISK;
2259 goto fail;
2260 }
2261
Philipp Reisner31890f42011-01-19 14:12:51 +01002262 if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) {
Philipp Reisner6495d2c2010-03-24 16:07:04 +01002263 retcode = ERR_NEED_APV_93;
2264 goto fail;
2265 }
2266
Philipp Reisner087c2492010-03-26 13:49:56 +01002267 if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev))
Philipp Reisnerb411b362009-09-25 16:07:19 -07002268 mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002269
2270 mdev->ldev->dc.disk_size = (sector_t)rs.resize_size;
Philipp Reisner6495d2c2010-03-24 16:07:04 +01002271 ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0);
Bart Van Assche24c48302011-05-21 18:32:29 +02002272 dd = drbd_determine_dev_size(mdev, ddsf);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002273 drbd_md_sync(mdev);
2274 put_ldev(mdev);
2275 if (dd == dev_size_error) {
2276 retcode = ERR_NOMEM_BITMAP;
2277 goto fail;
2278 }
2279
Philipp Reisner087c2492010-03-26 13:49:56 +01002280 if (mdev->state.conn == C_CONNECTED) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07002281 if (dd == grew)
2282 set_bit(RESIZE_PENDING, &mdev->flags);
2283
2284 drbd_send_uuids(mdev);
Philipp Reisner6495d2c2010-03-24 16:07:04 +01002285 drbd_send_sizes(mdev, 1, ddsf);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002286 }
2287
2288 fail:
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002289 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002290 return 0;
2291}
2292
Lars Ellenbergf3990022011-03-23 14:31:09 +01002293int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002294{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002295 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002296 cpumask_var_t new_cpu_mask;
Lars Ellenbergf3990022011-03-23 14:31:09 +01002297 struct drbd_tconn *tconn;
Philipp Reisner778f2712010-07-06 11:14:00 +02002298 int *rs_plan_s = NULL;
Lars Ellenbergb57a1e22011-04-27 21:17:33 +02002299 struct res_opts res_opts;
Lars Ellenbergf3990022011-03-23 14:31:09 +01002300 int err;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002301
Lars Ellenbergf3990022011-03-23 14:31:09 +01002302 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002303 if (!adm_ctx.reply_skb)
2304 return retcode;
2305 if (retcode != NO_ERROR)
2306 goto fail;
Lars Ellenbergf3990022011-03-23 14:31:09 +01002307 tconn = adm_ctx.tconn;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002308
Philipp Reisnerb411b362009-09-25 16:07:19 -07002309 if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) {
2310 retcode = ERR_NOMEM;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002311 drbd_msg_put_info("unable to allocate cpumask");
Philipp Reisnerb411b362009-09-25 16:07:19 -07002312 goto fail;
2313 }
2314
Lars Ellenbergb57a1e22011-04-27 21:17:33 +02002315 res_opts = tconn->res_opts;
Lars Ellenberg5979e362011-04-27 21:09:55 +02002316 if (should_set_defaults(info))
Lars Ellenbergb57a1e22011-04-27 21:17:33 +02002317 drbd_set_res_opts_default(&res_opts);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002318
Lars Ellenbergb57a1e22011-04-27 21:17:33 +02002319 err = res_opts_from_attrs(&res_opts, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002320 if (err) {
Philipp Reisnerb411b362009-09-25 16:07:19 -07002321 retcode = ERR_MANDATORY_TAG;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002322 drbd_msg_put_info(from_attrs_err_to_txt(err));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002323 goto fail;
2324 }
2325
Philipp Reisnerb411b362009-09-25 16:07:19 -07002326 /* silently ignore cpu mask on UP kernel */
Lars Ellenbergb57a1e22011-04-27 21:17:33 +02002327 if (nr_cpu_ids > 1 && res_opts.cpu_mask[0] != 0) {
2328 err = __bitmap_parse(res_opts.cpu_mask, 32, 0,
Philipp Reisnerb411b362009-09-25 16:07:19 -07002329 cpumask_bits(new_cpu_mask), nr_cpu_ids);
2330 if (err) {
Lars Ellenbergf3990022011-03-23 14:31:09 +01002331 conn_warn(tconn, "__bitmap_parse() failed with %d\n", err);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002332 retcode = ERR_CPU_MASK_PARSE;
2333 goto fail;
2334 }
2335 }
2336
Lars Ellenberg7ad651b2011-02-21 13:21:03 +01002337
Lars Ellenbergb57a1e22011-04-27 21:17:33 +02002338 tconn->res_opts = res_opts;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002339
Lars Ellenbergf3990022011-03-23 14:31:09 +01002340 if (!cpumask_equal(tconn->cpu_mask, new_cpu_mask)) {
2341 cpumask_copy(tconn->cpu_mask, new_cpu_mask);
2342 drbd_calc_cpu_mask(tconn);
2343 tconn->receiver.reset_cpu_mask = 1;
2344 tconn->asender.reset_cpu_mask = 1;
2345 tconn->worker.reset_cpu_mask = 1;
Philipp Reisner778f2712010-07-06 11:14:00 +02002346 }
2347
Philipp Reisnerb411b362009-09-25 16:07:19 -07002348fail:
Philipp Reisner778f2712010-07-06 11:14:00 +02002349 kfree(rs_plan_s);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002350 free_cpumask_var(new_cpu_mask);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002351
2352 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002353 return 0;
2354}
2355
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002356int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002357{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002358 struct drbd_conf *mdev;
2359 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2360
2361 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2362 if (!adm_ctx.reply_skb)
2363 return retcode;
2364 if (retcode != NO_ERROR)
2365 goto out;
2366
2367 mdev = adm_ctx.mdev;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002368
Lars Ellenberg194bfb32011-01-18 10:38:01 +01002369 /* If there is still bitmap IO pending, probably because of a previous
2370 * resync just being finished, wait for it before requesting a new resync. */
2371 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
2372
Philipp Reisnerb411b362009-09-25 16:07:19 -07002373 retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED);
2374
2375 if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION)
2376 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
2377
2378 while (retcode == SS_NEED_CONNECTION) {
Philipp Reisner87eeee42011-01-19 14:16:30 +01002379 spin_lock_irq(&mdev->tconn->req_lock);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002380 if (mdev->state.conn < C_CONNECTED)
2381 retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL);
Philipp Reisner87eeee42011-01-19 14:16:30 +01002382 spin_unlock_irq(&mdev->tconn->req_lock);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002383
2384 if (retcode != SS_NEED_CONNECTION)
2385 break;
2386
2387 retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T));
2388 }
2389
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002390out:
2391 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002392 return 0;
2393}
2394
Philipp Reisner07782862010-08-31 12:00:50 +02002395static int drbd_bmio_set_susp_al(struct drbd_conf *mdev)
2396{
2397 int rv;
2398
2399 rv = drbd_bmio_set_n_write(mdev);
2400 drbd_suspend_al(mdev);
2401 return rv;
2402}
2403
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002404static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info,
2405 union drbd_state mask, union drbd_state val)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002406{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002407 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002408
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002409 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2410 if (!adm_ctx.reply_skb)
2411 return retcode;
2412 if (retcode != NO_ERROR)
2413 goto out;
Lars Ellenberg194bfb32011-01-18 10:38:01 +01002414
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002415 retcode = drbd_request_state(adm_ctx.mdev, mask, val);
2416out:
2417 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002418 return 0;
2419}
2420
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002421int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002422{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002423 return drbd_adm_simple_request_state(skb, info, NS(conn, C_STARTING_SYNC_S));
2424}
Philipp Reisnerb411b362009-09-25 16:07:19 -07002425
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002426int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info)
2427{
2428 enum drbd_ret_code retcode;
2429
2430 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2431 if (!adm_ctx.reply_skb)
2432 return retcode;
2433 if (retcode != NO_ERROR)
2434 goto out;
2435
2436 if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002437 retcode = ERR_PAUSE_IS_SET;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002438out:
2439 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002440 return 0;
2441}
2442
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002443int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002444{
Philipp Reisnerda9fbc22011-03-29 10:52:01 +02002445 union drbd_dev_state s;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002446 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002447
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002448 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2449 if (!adm_ctx.reply_skb)
2450 return retcode;
2451 if (retcode != NO_ERROR)
2452 goto out;
2453
2454 if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) {
2455 s = adm_ctx.mdev->state;
Philipp Reisnercd88d032011-01-20 11:46:41 +01002456 if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) {
2457 retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP :
2458 s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR;
2459 } else {
2460 retcode = ERR_PAUSE_IS_CLEAR;
2461 }
2462 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07002463
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002464out:
2465 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002466 return 0;
2467}
2468
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002469int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002470{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002471 return drbd_adm_simple_request_state(skb, info, NS(susp, 1));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002472}
2473
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002474int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002475{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002476 struct drbd_conf *mdev;
2477 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
2478
2479 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2480 if (!adm_ctx.reply_skb)
2481 return retcode;
2482 if (retcode != NO_ERROR)
2483 goto out;
2484
2485 mdev = adm_ctx.mdev;
Philipp Reisner43a51822010-06-11 11:26:34 +02002486 if (test_bit(NEW_CUR_UUID, &mdev->flags)) {
2487 drbd_uuid_new_current(mdev);
2488 clear_bit(NEW_CUR_UUID, &mdev->flags);
Philipp Reisner43a51822010-06-11 11:26:34 +02002489 }
Philipp Reisner265be2d2010-05-31 10:14:17 +02002490 drbd_suspend_io(mdev);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002491 retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0));
2492 if (retcode == SS_SUCCESS) {
Philipp Reisner265be2d2010-05-31 10:14:17 +02002493 if (mdev->state.conn < C_CONNECTED)
Philipp Reisner2f5cdd02011-02-21 14:29:27 +01002494 tl_clear(mdev->tconn);
Philipp Reisner265be2d2010-05-31 10:14:17 +02002495 if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED)
Philipp Reisner2f5cdd02011-02-21 14:29:27 +01002496 tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO);
Philipp Reisner265be2d2010-05-31 10:14:17 +02002497 }
2498 drbd_resume_io(mdev);
2499
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002500out:
2501 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002502 return 0;
2503}
2504
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002505int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002506{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002507 return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002508}
2509
Lars Ellenberg543cc102011-03-10 22:18:18 +01002510int nla_put_drbd_cfg_context(struct sk_buff *skb, const char *conn_name, unsigned vnr)
2511{
2512 struct nlattr *nla;
2513 nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT);
2514 if (!nla)
2515 goto nla_put_failure;
2516 if (vnr != VOLUME_UNSPECIFIED)
2517 NLA_PUT_U32(skb, T_ctx_volume, vnr);
2518 NLA_PUT_STRING(skb, T_ctx_conn_name, conn_name);
2519 nla_nest_end(skb, nla);
2520 return 0;
2521
2522nla_put_failure:
2523 if (nla)
2524 nla_nest_cancel(skb, nla);
2525 return -EMSGSIZE;
2526}
2527
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002528int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev,
2529 const struct sib_info *sib)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002530{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002531 struct state_info *si = NULL; /* for sizeof(si->member); */
Philipp Reisner44ed1672011-04-19 17:10:19 +02002532 struct net_conf *nc;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002533 struct nlattr *nla;
2534 int got_ldev;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002535 int err = 0;
2536 int exclude_sensitive;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002537
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002538 /* If sib != NULL, this is drbd_bcast_event, which anyone can listen
2539 * to. So we better exclude_sensitive information.
2540 *
2541 * If sib == NULL, this is drbd_adm_get_status, executed synchronously
2542 * in the context of the requesting user process. Exclude sensitive
2543 * information, unless current has superuser.
2544 *
2545 * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and
2546 * relies on the current implementation of netlink_dump(), which
2547 * executes the dump callback successively from netlink_recvmsg(),
2548 * always in the context of the receiving process */
2549 exclude_sensitive = sib || !capable(CAP_SYS_ADMIN);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002550
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002551 got_ldev = get_ldev(mdev);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002552
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002553 /* We need to add connection name and volume number information still.
2554 * Minor number is in drbd_genlmsghdr. */
Lars Ellenberg543cc102011-03-10 22:18:18 +01002555 if (nla_put_drbd_cfg_context(skb, mdev->tconn->name, mdev->vnr))
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002556 goto nla_put_failure;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002557
Lars Ellenbergf3990022011-03-23 14:31:09 +01002558 if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive))
2559 goto nla_put_failure;
2560
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002561 if (got_ldev)
2562 if (disk_conf_to_skb(skb, &mdev->ldev->dc, exclude_sensitive))
2563 goto nla_put_failure;
Philipp Reisner44ed1672011-04-19 17:10:19 +02002564
2565 rcu_read_lock();
2566 nc = rcu_dereference(mdev->tconn->net_conf);
2567 if (nc)
2568 err = net_conf_to_skb(skb, nc, exclude_sensitive);
2569 rcu_read_unlock();
2570 if (err)
2571 goto nla_put_failure;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002572
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002573 nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO);
2574 if (!nla)
2575 goto nla_put_failure;
2576 NLA_PUT_U32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY);
2577 NLA_PUT_U32(skb, T_current_state, mdev->state.i);
2578 NLA_PUT_U64(skb, T_ed_uuid, mdev->ed_uuid);
2579 NLA_PUT_U64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev));
Philipp Reisnerb411b362009-09-25 16:07:19 -07002580
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002581 if (got_ldev) {
2582 NLA_PUT_U32(skb, T_disk_flags, mdev->ldev->md.flags);
2583 NLA_PUT(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid);
2584 NLA_PUT_U64(skb, T_bits_total, drbd_bm_bits(mdev));
2585 NLA_PUT_U64(skb, T_bits_oos, drbd_bm_total_weight(mdev));
2586 if (C_SYNC_SOURCE <= mdev->state.conn &&
2587 C_PAUSED_SYNC_T >= mdev->state.conn) {
2588 NLA_PUT_U64(skb, T_bits_rs_total, mdev->rs_total);
2589 NLA_PUT_U64(skb, T_bits_rs_failed, mdev->rs_failed);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002590 }
2591 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07002592
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002593 if (sib) {
2594 switch(sib->sib_reason) {
2595 case SIB_SYNC_PROGRESS:
2596 case SIB_GET_STATUS_REPLY:
2597 break;
2598 case SIB_STATE_CHANGE:
2599 NLA_PUT_U32(skb, T_prev_state, sib->os.i);
2600 NLA_PUT_U32(skb, T_new_state, sib->ns.i);
2601 break;
2602 case SIB_HELPER_POST:
2603 NLA_PUT_U32(skb,
2604 T_helper_exit_code, sib->helper_exit_code);
2605 /* fall through */
2606 case SIB_HELPER_PRE:
2607 NLA_PUT_STRING(skb, T_helper, sib->helper_name);
2608 break;
2609 }
2610 }
2611 nla_nest_end(skb, nla);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002612
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002613 if (0)
2614nla_put_failure:
2615 err = -EMSGSIZE;
2616 if (got_ldev)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002617 put_ldev(mdev);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002618 return err;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002619}
2620
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002621int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002622{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002623 enum drbd_ret_code retcode;
2624 int err;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002625
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002626 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2627 if (!adm_ctx.reply_skb)
2628 return retcode;
2629 if (retcode != NO_ERROR)
2630 goto out;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002631
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002632 err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL);
2633 if (err) {
2634 nlmsg_free(adm_ctx.reply_skb);
2635 return err;
2636 }
2637out:
2638 drbd_adm_finish(info, retcode);
2639 return 0;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002640}
2641
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002642int get_one_status(struct sk_buff *skb, struct netlink_callback *cb)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002643{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002644 struct drbd_conf *mdev;
2645 struct drbd_genlmsghdr *dh;
Lars Ellenberg543cc102011-03-10 22:18:18 +01002646 struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0];
2647 struct drbd_tconn *tconn = NULL;
2648 struct drbd_tconn *tmp;
2649 unsigned volume = cb->args[1];
Philipp Reisnerb411b362009-09-25 16:07:19 -07002650
Lars Ellenberg543cc102011-03-10 22:18:18 +01002651 /* Open coded, deferred, iteration:
2652 * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) {
2653 * idr_for_each_entry(&tconn->volumes, mdev, i) {
2654 * ...
2655 * }
2656 * }
2657 * where tconn is cb->args[0];
2658 * and i is cb->args[1];
2659 *
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002660 * cb->args[2] indicates if we shall loop over all resources,
2661 * or just dump all volumes of a single resource.
2662 *
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002663 * This may miss entries inserted after this dump started,
2664 * or entries deleted before they are reached.
Lars Ellenberg543cc102011-03-10 22:18:18 +01002665 *
2666 * We need to make sure the mdev won't disappear while
2667 * we are looking at it, and revalidate our iterators
2668 * on each iteration.
2669 */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002670
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +02002671 /* synchronize with conn_create()/conn_destroy() */
Philipp Reisneref356262011-04-13 14:21:29 -07002672 down_read(&drbd_cfg_rwsem);
Lars Ellenberg543cc102011-03-10 22:18:18 +01002673 /* revalidate iterator position */
2674 list_for_each_entry(tmp, &drbd_tconns, all_tconn) {
2675 if (pos == NULL) {
2676 /* first iteration */
2677 pos = tmp;
2678 tconn = pos;
2679 break;
2680 }
2681 if (tmp == pos) {
2682 tconn = pos;
2683 break;
2684 }
2685 }
2686 if (tconn) {
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002687next_tconn:
Lars Ellenberg543cc102011-03-10 22:18:18 +01002688 mdev = idr_get_next(&tconn->volumes, &volume);
2689 if (!mdev) {
2690 /* No more volumes to dump on this tconn.
2691 * Advance tconn iterator. */
2692 pos = list_entry(tconn->all_tconn.next,
2693 struct drbd_tconn, all_tconn);
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002694 /* Did we dump any volume on this tconn yet? */
Lars Ellenberg543cc102011-03-10 22:18:18 +01002695 if (volume != 0) {
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002696 /* If we reached the end of the list,
2697 * or only a single resource dump was requested,
2698 * we are done. */
2699 if (&pos->all_tconn == &drbd_tconns || cb->args[2])
2700 goto out;
Lars Ellenberg543cc102011-03-10 22:18:18 +01002701 volume = 0;
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002702 tconn = pos;
Lars Ellenberg543cc102011-03-10 22:18:18 +01002703 goto next_tconn;
2704 }
2705 }
2706
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002707 dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid,
2708 cb->nlh->nlmsg_seq, &drbd_genl_family,
2709 NLM_F_MULTI, DRBD_ADM_GET_STATUS);
2710 if (!dh)
Lars Ellenberg543cc102011-03-10 22:18:18 +01002711 goto out;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002712
Lars Ellenberg543cc102011-03-10 22:18:18 +01002713 if (!mdev) {
2714 /* this is a tconn without a single volume */
2715 dh->minor = -1U;
2716 dh->ret_code = NO_ERROR;
2717 if (nla_put_drbd_cfg_context(skb, tconn->name, VOLUME_UNSPECIFIED))
2718 genlmsg_cancel(skb, dh);
2719 else
2720 genlmsg_end(skb, dh);
2721 goto out;
2722 }
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002723
Lars Ellenberg543cc102011-03-10 22:18:18 +01002724 D_ASSERT(mdev->vnr == volume);
2725 D_ASSERT(mdev->tconn == tconn);
2726
2727 dh->minor = mdev_to_minor(mdev);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002728 dh->ret_code = NO_ERROR;
2729
2730 if (nla_put_status_info(skb, mdev, NULL)) {
2731 genlmsg_cancel(skb, dh);
Lars Ellenberg543cc102011-03-10 22:18:18 +01002732 goto out;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002733 }
2734 genlmsg_end(skb, dh);
2735 }
2736
Lars Ellenberg543cc102011-03-10 22:18:18 +01002737out:
Philipp Reisneref356262011-04-13 14:21:29 -07002738 up_read(&drbd_cfg_rwsem);
Lars Ellenberg543cc102011-03-10 22:18:18 +01002739 /* where to start the next iteration */
2740 cb->args[0] = (long)pos;
2741 cb->args[1] = (pos == tconn) ? volume + 1 : 0;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002742
Lars Ellenberg543cc102011-03-10 22:18:18 +01002743 /* No more tconns/volumes/minors found results in an empty skb.
2744 * Which will terminate the dump. */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002745 return skb->len;
2746}
2747
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002748/*
2749 * Request status of all resources, or of all volumes within a single resource.
2750 *
2751 * This is a dump, as the answer may not fit in a single reply skb otherwise.
2752 * Which means we cannot use the family->attrbuf or other such members, because
2753 * dump is NOT protected by the genl_lock(). During dump, we only have access
2754 * to the incoming skb, and need to opencode "parsing" of the nlattr payload.
2755 *
2756 * Once things are setup properly, we call into get_one_status().
2757 */
2758int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb)
2759{
2760 const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ;
2761 struct nlattr *nla;
2762 const char *conn_name;
2763 struct drbd_tconn *tconn;
2764
2765 /* Is this a followup call? */
2766 if (cb->args[0]) {
2767 /* ... of a single resource dump,
2768 * and the resource iterator has been advanced already? */
2769 if (cb->args[2] && cb->args[2] != cb->args[0])
2770 return 0; /* DONE. */
2771 goto dump;
2772 }
2773
2774 /* First call (from netlink_dump_start). We need to figure out
2775 * which resource(s) the user wants us to dump. */
2776 nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen),
2777 nlmsg_attrlen(cb->nlh, hdrlen),
2778 DRBD_NLA_CFG_CONTEXT);
2779
2780 /* No explicit context given. Dump all. */
2781 if (!nla)
2782 goto dump;
2783 nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name));
2784 /* context given, but no name present? */
2785 if (!nla)
2786 return -EINVAL;
2787 conn_name = nla_data(nla);
Philipp Reisner0ace9df2011-04-24 10:53:19 +02002788 tconn = conn_get_by_name(conn_name);
2789
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002790 if (!tconn)
2791 return -ENODEV;
2792
Philipp Reisner0ace9df2011-04-24 10:53:19 +02002793 kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */
2794
Lars Ellenberg71932ef2011-04-18 09:43:25 +02002795 /* prime iterators, and set "filter" mode mark:
2796 * only dump this tconn. */
2797 cb->args[0] = (long)tconn;
2798 /* cb->args[1] = 0; passed in this way. */
2799 cb->args[2] = (long)tconn;
2800
2801dump:
2802 return get_one_status(skb, cb);
2803}
2804
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002805int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info)
2806{
2807 enum drbd_ret_code retcode;
2808 struct timeout_parms tp;
2809 int err;
2810
2811 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2812 if (!adm_ctx.reply_skb)
2813 return retcode;
2814 if (retcode != NO_ERROR)
2815 goto out;
2816
2817 tp.timeout_type =
2818 adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED :
2819 test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED :
2820 UT_DEFAULT;
2821
2822 err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp);
2823 if (err) {
2824 nlmsg_free(adm_ctx.reply_skb);
2825 return err;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002826 }
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002827out:
2828 drbd_adm_finish(info, retcode);
2829 return 0;
2830}
Lars Ellenberg873b0d52011-01-21 22:53:48 +01002831
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002832int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info)
2833{
2834 struct drbd_conf *mdev;
2835 enum drbd_ret_code retcode;
2836
2837 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2838 if (!adm_ctx.reply_skb)
2839 return retcode;
2840 if (retcode != NO_ERROR)
2841 goto out;
2842
2843 mdev = adm_ctx.mdev;
2844 if (info->attrs[DRBD_NLA_START_OV_PARMS]) {
2845 /* resume from last known position, if possible */
2846 struct start_ov_parms parms =
2847 { .ov_start_sector = mdev->ov_start_sector };
Lars Ellenbergf3990022011-03-23 14:31:09 +01002848 int err = start_ov_parms_from_attrs(&parms, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002849 if (err) {
2850 retcode = ERR_MANDATORY_TAG;
2851 drbd_msg_put_info(from_attrs_err_to_txt(err));
2852 goto out;
2853 }
2854 /* w_make_ov_request expects position to be aligned */
2855 mdev->ov_start_sector = parms.ov_start_sector & ~BM_SECT_PER_BIT;
2856 }
Lars Ellenberg873b0d52011-01-21 22:53:48 +01002857 /* If there is still bitmap IO pending, e.g. previous resync or verify
2858 * just being finished, wait for it before requesting a new resync. */
2859 wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags));
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002860 retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S));
2861out:
2862 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002863 return 0;
2864}
2865
2866
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002867int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info)
Philipp Reisnerb411b362009-09-25 16:07:19 -07002868{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002869 struct drbd_conf *mdev;
2870 enum drbd_ret_code retcode;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002871 int skip_initial_sync = 0;
2872 int err;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002873 struct new_c_uuid_parms args;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002874
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002875 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
2876 if (!adm_ctx.reply_skb)
2877 return retcode;
2878 if (retcode != NO_ERROR)
2879 goto out_nolock;
Philipp Reisnerb411b362009-09-25 16:07:19 -07002880
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002881 mdev = adm_ctx.mdev;
2882 memset(&args, 0, sizeof(args));
2883 if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) {
Lars Ellenbergf3990022011-03-23 14:31:09 +01002884 err = new_c_uuid_parms_from_attrs(&args, info);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002885 if (err) {
2886 retcode = ERR_MANDATORY_TAG;
2887 drbd_msg_put_info(from_attrs_err_to_txt(err));
2888 goto out_nolock;
2889 }
Philipp Reisnerb411b362009-09-25 16:07:19 -07002890 }
2891
Philipp Reisner8410da82011-02-11 20:11:10 +01002892 mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */
Philipp Reisnerb411b362009-09-25 16:07:19 -07002893
2894 if (!get_ldev(mdev)) {
2895 retcode = ERR_NO_DISK;
2896 goto out;
2897 }
2898
2899 /* this is "skip initial sync", assume to be clean */
Philipp Reisner31890f42011-01-19 14:12:51 +01002900 if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 &&
Philipp Reisnerb411b362009-09-25 16:07:19 -07002901 mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) {
2902 dev_info(DEV, "Preparing to skip initial sync\n");
2903 skip_initial_sync = 1;
2904 } else if (mdev->state.conn != C_STANDALONE) {
2905 retcode = ERR_CONNECTED;
2906 goto out_dec;
2907 }
2908
2909 drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */
2910 drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */
2911
2912 if (args.clear_bm) {
Lars Ellenberg20ceb2b2011-01-21 10:56:44 +01002913 err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write,
2914 "clear_n_write from new_c_uuid", BM_LOCKED_MASK);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002915 if (err) {
2916 dev_err(DEV, "Writing bitmap failed with %d\n",err);
2917 retcode = ERR_IO_MD_DISK;
2918 }
2919 if (skip_initial_sync) {
2920 drbd_send_uuids_skip_initial_sync(mdev);
2921 _drbd_uuid_set(mdev, UI_BITMAP, 0);
Lars Ellenberg62b0da32011-01-20 13:25:21 +01002922 drbd_print_uuids(mdev, "cleared bitmap UUID");
Philipp Reisner87eeee42011-01-19 14:16:30 +01002923 spin_lock_irq(&mdev->tconn->req_lock);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002924 _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE),
2925 CS_VERBOSE, NULL);
Philipp Reisner87eeee42011-01-19 14:16:30 +01002926 spin_unlock_irq(&mdev->tconn->req_lock);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002927 }
2928 }
2929
2930 drbd_md_sync(mdev);
2931out_dec:
2932 put_ldev(mdev);
2933out:
Philipp Reisner8410da82011-02-11 20:11:10 +01002934 mutex_unlock(mdev->state_mutex);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002935out_nolock:
2936 drbd_adm_finish(info, retcode);
Philipp Reisnerb411b362009-09-25 16:07:19 -07002937 return 0;
2938}
2939
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002940static enum drbd_ret_code
2941drbd_check_conn_name(const char *name)
Philipp Reisner774b3052011-02-22 02:07:03 -05002942{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002943 if (!name || !name[0]) {
2944 drbd_msg_put_info("connection name missing");
2945 return ERR_MANDATORY_TAG;
2946 }
2947 /* if we want to use these in sysfs/configfs/debugfs some day,
2948 * we must not allow slashes */
2949 if (strchr(name, '/')) {
2950 drbd_msg_put_info("invalid connection name");
2951 return ERR_INVALID_REQUEST;
2952 }
2953 return NO_ERROR;
2954}
Philipp Reisner774b3052011-02-22 02:07:03 -05002955
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002956int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info)
2957{
2958 enum drbd_ret_code retcode;
2959
2960 retcode = drbd_adm_prepare(skb, info, 0);
2961 if (!adm_ctx.reply_skb)
2962 return retcode;
2963 if (retcode != NO_ERROR)
2964 goto out;
2965
2966 retcode = drbd_check_conn_name(adm_ctx.conn_name);
2967 if (retcode != NO_ERROR)
2968 goto out;
2969
2970 if (adm_ctx.tconn) {
Lars Ellenberg38f19612011-03-14 13:22:35 +01002971 if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) {
2972 retcode = ERR_INVALID_REQUEST;
2973 drbd_msg_put_info("connection exists");
2974 }
2975 /* else: still NO_ERROR */
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002976 goto out;
Philipp Reisner774b3052011-02-22 02:07:03 -05002977 }
2978
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +02002979 if (!conn_create(adm_ctx.conn_name))
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002980 retcode = ERR_NOMEM;
2981out:
2982 drbd_adm_finish(info, retcode);
Philipp Reisner774b3052011-02-22 02:07:03 -05002983 return 0;
2984}
2985
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002986int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info)
Philipp Reisner774b3052011-02-22 02:07:03 -05002987{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002988 struct drbd_genlmsghdr *dh = info->userhdr;
2989 enum drbd_ret_code retcode;
Philipp Reisner774b3052011-02-22 02:07:03 -05002990
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002991 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
2992 if (!adm_ctx.reply_skb)
2993 return retcode;
2994 if (retcode != NO_ERROR)
2995 goto out;
Philipp Reisner774b3052011-02-22 02:07:03 -05002996
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01002997 /* FIXME drop minor_count parameter, limit to MINORMASK */
2998 if (dh->minor >= minor_count) {
2999 drbd_msg_put_info("requested minor out of range");
3000 retcode = ERR_INVALID_REQUEST;
3001 goto out;
3002 }
Andreas Gruenbacher0c8e36d2011-03-30 16:00:17 +02003003 if (adm_ctx.volume > DRBD_VOLUME_MAX) {
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003004 drbd_msg_put_info("requested volume id out of range");
3005 retcode = ERR_INVALID_REQUEST;
3006 goto out;
Philipp Reisner774b3052011-02-22 02:07:03 -05003007 }
3008
Lars Ellenberg38f19612011-03-14 13:22:35 +01003009 /* drbd_adm_prepare made sure already
3010 * that mdev->tconn and mdev->vnr match the request. */
3011 if (adm_ctx.mdev) {
3012 if (info->nlhdr->nlmsg_flags & NLM_F_EXCL)
3013 retcode = ERR_MINOR_EXISTS;
3014 /* else: still NO_ERROR */
3015 goto out;
3016 }
3017
Philipp Reisnerd3fcb492011-04-13 14:46:05 -07003018 down_write(&drbd_cfg_rwsem);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003019 retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume);
Philipp Reisnerd3fcb492011-04-13 14:46:05 -07003020 up_write(&drbd_cfg_rwsem);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003021out:
3022 drbd_adm_finish(info, retcode);
Philipp Reisner774b3052011-02-22 02:07:03 -05003023 return 0;
3024}
3025
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003026static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev)
3027{
3028 if (mdev->state.disk == D_DISKLESS &&
3029 /* no need to be mdev->state.conn == C_STANDALONE &&
3030 * we may want to delete a minor from a live replication group.
3031 */
3032 mdev->state.role == R_SECONDARY) {
Philipp Reisnerff370e52011-04-11 21:10:11 -07003033 drbd_delete_device(mdev);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003034 return NO_ERROR;
3035 } else
3036 return ERR_MINOR_CONFIGURED;
3037}
3038
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003039int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info)
Philipp Reisner774b3052011-02-22 02:07:03 -05003040{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003041 enum drbd_ret_code retcode;
3042
3043 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR);
3044 if (!adm_ctx.reply_skb)
3045 return retcode;
3046 if (retcode != NO_ERROR)
3047 goto out;
3048
Philipp Reisneref356262011-04-13 14:21:29 -07003049 down_write(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003050 retcode = adm_delete_minor(adm_ctx.mdev);
Philipp Reisneref356262011-04-13 14:21:29 -07003051 up_write(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003052out:
3053 drbd_adm_finish(info, retcode);
3054 return 0;
3055}
3056
3057int drbd_adm_down(struct sk_buff *skb, struct genl_info *info)
3058{
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003059 int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003060 struct drbd_conf *mdev;
3061 unsigned i;
3062
3063 retcode = drbd_adm_prepare(skb, info, 0);
3064 if (!adm_ctx.reply_skb)
3065 return retcode;
3066 if (retcode != NO_ERROR)
3067 goto out;
3068
3069 if (!adm_ctx.tconn) {
3070 retcode = ERR_CONN_NOT_KNOWN;
3071 goto out;
3072 }
3073
Philipp Reisneref356262011-04-13 14:21:29 -07003074 down_read(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003075 /* demote */
3076 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3077 retcode = drbd_set_role(mdev, R_SECONDARY, 0);
3078 if (retcode < SS_SUCCESS) {
3079 drbd_msg_put_info("failed to demote");
3080 goto out_unlock;
3081 }
3082 }
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003083 up_read(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003084
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003085 /* disconnect; may stop the receiver;
3086 * must not hold the drbd_cfg_rwsem */
3087 retcode = conn_try_disconnect(adm_ctx.tconn, 0);
3088 if (retcode < SS_SUCCESS) {
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003089 drbd_msg_put_info("failed to disconnect");
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003090 goto out;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003091 }
3092
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003093 down_read(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003094 /* detach */
3095 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003096 retcode = adm_detach(mdev);
3097 if (retcode < SS_SUCCESS) {
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003098 drbd_msg_put_info("failed to detach");
3099 goto out_unlock;
3100 }
3101 }
Philipp Reisneref356262011-04-13 14:21:29 -07003102 up_read(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003103
Lars Ellenbergf3dfa402011-05-02 10:45:05 +02003104 /* If we reach this, all volumes (of this tconn) are Secondary,
3105 * Disconnected, Diskless, aka Unconfigured. Make sure all threads have
3106 * actually stopped, state handling only does drbd_thread_stop_nowait().
3107 * This needs to be done without holding drbd_cfg_rwsem. */
3108 drbd_thread_stop(&adm_ctx.tconn->worker);
3109
3110 /* Now, nothing can fail anymore */
3111
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003112 /* delete volumes */
Philipp Reisneref356262011-04-13 14:21:29 -07003113 down_write(&drbd_cfg_rwsem);
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003114 idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) {
3115 retcode = adm_delete_minor(mdev);
3116 if (retcode != NO_ERROR) {
3117 /* "can not happen" */
3118 drbd_msg_put_info("failed to delete volume");
Philipp Reisneref356262011-04-13 14:21:29 -07003119 up_write(&drbd_cfg_rwsem);
3120 goto out;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003121 }
3122 }
3123
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003124 /* delete connection */
3125 if (conn_lowest_minor(adm_ctx.tconn) < 0) {
Lars Ellenbergd9cc6e22011-04-27 10:25:28 +02003126 drbd_thread_stop(&adm_ctx.tconn->worker);
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +02003127 list_del(&adm_ctx.tconn->all_tconn);
3128 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
3129
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003130 retcode = NO_ERROR;
3131 } else {
3132 /* "can not happen" */
3133 retcode = ERR_CONN_IN_USE;
3134 drbd_msg_put_info("failed to delete connection");
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003135 }
Philipp Reisneref356262011-04-13 14:21:29 -07003136
3137 up_write(&drbd_cfg_rwsem);
3138 goto out;
Lars Ellenberg85f75dd72011-03-15 16:26:37 +01003139out_unlock:
Philipp Reisneref356262011-04-13 14:21:29 -07003140 up_read(&drbd_cfg_rwsem);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003141out:
3142 drbd_adm_finish(info, retcode);
Philipp Reisner774b3052011-02-22 02:07:03 -05003143 return 0;
3144}
3145
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003146int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info)
Philipp Reisner774b3052011-02-22 02:07:03 -05003147{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003148 enum drbd_ret_code retcode;
3149
3150 retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN);
3151 if (!adm_ctx.reply_skb)
3152 return retcode;
3153 if (retcode != NO_ERROR)
3154 goto out;
3155
Philipp Reisneref356262011-04-13 14:21:29 -07003156 down_write(&drbd_cfg_rwsem);
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003157 if (conn_lowest_minor(adm_ctx.tconn) < 0) {
Philipp Reisner9dc9fbb2011-04-22 15:23:32 +02003158 list_del(&adm_ctx.tconn->all_tconn);
3159 kref_put(&adm_ctx.tconn->kref, &conn_destroy);
3160
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003161 retcode = NO_ERROR;
Philipp Reisner774b3052011-02-22 02:07:03 -05003162 } else {
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003163 retcode = ERR_CONN_IN_USE;
Philipp Reisner774b3052011-02-22 02:07:03 -05003164 }
Philipp Reisneref356262011-04-13 14:21:29 -07003165 up_write(&drbd_cfg_rwsem);
Philipp Reisner774b3052011-02-22 02:07:03 -05003166
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003167out:
3168 drbd_adm_finish(info, retcode);
Philipp Reisner774b3052011-02-22 02:07:03 -05003169 return 0;
3170}
3171
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003172void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib)
Philipp Reisnerb411b362009-09-25 16:07:19 -07003173{
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003174 static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */
3175 struct sk_buff *msg;
3176 struct drbd_genlmsghdr *d_out;
3177 unsigned seq;
3178 int err = -ENOMEM;
Philipp Reisnerb411b362009-09-25 16:07:19 -07003179
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003180 seq = atomic_inc_return(&drbd_genl_seq);
3181 msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO);
3182 if (!msg)
3183 goto failed;
Philipp Reisnerb411b362009-09-25 16:07:19 -07003184
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003185 err = -EMSGSIZE;
3186 d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT);
3187 if (!d_out) /* cannot happen, but anyways. */
3188 goto nla_put_failure;
3189 d_out->minor = mdev_to_minor(mdev);
3190 d_out->ret_code = 0;
Philipp Reisner9f5180e2009-10-06 09:30:14 +02003191
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003192 if (nla_put_status_info(msg, mdev, sib))
3193 goto nla_put_failure;
3194 genlmsg_end(msg, d_out);
3195 err = drbd_genl_multicast_events(msg, 0);
3196 /* msg has been consumed or freed in netlink_broadcast() */
3197 if (err && err != -ESRCH)
3198 goto failed;
Philipp Reisnerb411b362009-09-25 16:07:19 -07003199
Philipp Reisnerb411b362009-09-25 16:07:19 -07003200 return;
Lars Ellenberg3b98c0c2011-03-07 12:49:34 +01003201
3202nla_put_failure:
3203 nlmsg_free(msg);
3204failed:
3205 dev_err(DEV, "Error %d while broadcasting event. "
3206 "Event seq:%u sib_reason:%u\n",
3207 err, seq, sib->sib_reason);
Philipp Reisnerb411b362009-09-25 16:07:19 -07003208}