Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
b411b363 PR |
26 | #include <linux/module.h> |
27 | #include <linux/drbd.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/fs.h> | |
30 | #include <linux/file.h> | |
31 | #include <linux/slab.h> | |
b411b363 PR |
32 | #include <linux/blkpg.h> |
33 | #include <linux/cpumask.h> | |
34 | #include "drbd_int.h" | |
265be2d0 | 35 | #include "drbd_req.h" |
b411b363 PR |
36 | #include "drbd_wrappers.h" |
37 | #include <asm/unaligned.h> | |
b411b363 | 38 | #include <linux/drbd_limits.h> |
87f7be4c | 39 | #include <linux/kthread.h> |
b411b363 | 40 | |
3b98c0c2 LE |
41 | #include <net/genetlink.h> |
42 | ||
43 | /* .doit */ | |
44 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
45 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | ||
47 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info); | |
48 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info); | |
49 | ||
50 | int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info); | |
51 | int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 52 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
53 | |
54 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
55 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 56 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
57 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
58 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
62 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 71 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
72 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
73 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
74 | /* .dumpit */ | |
75 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
76 | ||
77 | #include <linux/drbd_genl_api.h> | |
78 | #include <linux/genl_magic_func.h> | |
79 | ||
80 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
81 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
82 | ||
3b98c0c2 LE |
83 | /* Configuration is strictly serialized, because generic netlink message |
84 | * processing is strictly serialized by the genl_lock(). | |
85 | * Which means we can use one static global drbd_config_context struct. | |
86 | */ | |
87 | static struct drbd_config_context { | |
88 | /* assigned from drbd_genlmsghdr */ | |
89 | unsigned int minor; | |
90 | /* assigned from request attributes, if present */ | |
91 | unsigned int volume; | |
92 | #define VOLUME_UNSPECIFIED (-1U) | |
93 | /* pointer into the request skb, | |
94 | * limited lifetime! */ | |
95 | char *conn_name; | |
96 | ||
97 | /* reply buffer */ | |
98 | struct sk_buff *reply_skb; | |
99 | /* pointer into reply buffer */ | |
100 | struct drbd_genlmsghdr *reply_dh; | |
101 | /* resolved from attributes, if possible */ | |
102 | struct drbd_conf *mdev; | |
103 | struct drbd_tconn *tconn; | |
104 | } adm_ctx; | |
105 | ||
106 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
107 | { | |
108 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
109 | if (genlmsg_reply(skb, info)) | |
110 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 111 | } |
3b98c0c2 LE |
112 | |
113 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
114 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 115 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
116 | { |
117 | struct sk_buff *skb = adm_ctx.reply_skb; | |
118 | struct nlattr *nla; | |
119 | int err = -EMSGSIZE; | |
120 | ||
121 | if (!info || !info[0]) | |
122 | return 0; | |
123 | ||
124 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
125 | if (!nla) | |
126 | return err; | |
127 | ||
128 | err = nla_put_string(skb, T_info_text, info); | |
129 | if (err) { | |
130 | nla_nest_cancel(skb, nla); | |
131 | return err; | |
132 | } else | |
133 | nla_nest_end(skb, nla); | |
134 | return 0; | |
b411b363 PR |
135 | } |
136 | ||
3b98c0c2 LE |
137 | /* This would be a good candidate for a "pre_doit" hook, |
138 | * and per-family private info->pointers. | |
139 | * But we need to stay compatible with older kernels. | |
140 | * If it returns successfully, adm_ctx members are valid. | |
141 | */ | |
142 | #define DRBD_ADM_NEED_MINOR 1 | |
143 | #define DRBD_ADM_NEED_CONN 2 | |
144 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, | |
145 | unsigned flags) | |
146 | { | |
147 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
148 | const u8 cmd = info->genlhdr->cmd; | |
149 | int err; | |
150 | ||
151 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
152 | ||
153 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
154 | if (cmd != DRBD_ADM_GET_STATUS | |
155 | && security_netlink_recv(skb, CAP_SYS_ADMIN)) | |
156 | return -EPERM; | |
157 | ||
158 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
159 | if (!adm_ctx.reply_skb) | |
160 | goto fail; | |
161 | ||
162 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
163 | info, &drbd_genl_family, 0, cmd); | |
164 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
165 | * but anyways */ | |
166 | if (!adm_ctx.reply_dh) | |
167 | goto fail; | |
168 | ||
169 | adm_ctx.reply_dh->minor = d_in->minor; | |
170 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
171 | ||
172 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { | |
173 | struct nlattr *nla; | |
174 | /* parse and validate only */ | |
f399002e | 175 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
176 | if (err) |
177 | goto fail; | |
178 | ||
179 | /* It was present, and valid, | |
180 | * copy it over to the reply skb. */ | |
181 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
182 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
183 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
184 | if (err) | |
185 | goto fail; | |
186 | ||
187 | /* and assign stuff to the global adm_ctx */ | |
188 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
189 | adm_ctx.volume = nla ? nla_get_u32(nla) : VOLUME_UNSPECIFIED; | |
190 | nla = nested_attr_tb[__nla_type(T_ctx_conn_name)]; | |
191 | if (nla) | |
192 | adm_ctx.conn_name = nla_data(nla); | |
193 | } else | |
194 | adm_ctx.volume = VOLUME_UNSPECIFIED; | |
195 | ||
196 | adm_ctx.minor = d_in->minor; | |
197 | adm_ctx.mdev = minor_to_mdev(d_in->minor); | |
0ace9dfa | 198 | adm_ctx.tconn = conn_get_by_name(adm_ctx.conn_name); |
3b98c0c2 LE |
199 | |
200 | if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) { | |
201 | drbd_msg_put_info("unknown minor"); | |
202 | return ERR_MINOR_INVALID; | |
203 | } | |
204 | if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_CONN)) { | |
205 | drbd_msg_put_info("unknown connection"); | |
206 | return ERR_INVALID_REQUEST; | |
207 | } | |
208 | ||
209 | /* some more paranoia, if the request was over-determined */ | |
527f4b24 LE |
210 | if (adm_ctx.mdev && adm_ctx.tconn && |
211 | adm_ctx.mdev->tconn != adm_ctx.tconn) { | |
212 | pr_warning("request: minor=%u, conn=%s; but that minor belongs to connection %s\n", | |
213 | adm_ctx.minor, adm_ctx.conn_name, adm_ctx.mdev->tconn->name); | |
214 | drbd_msg_put_info("minor exists in different connection"); | |
215 | return ERR_INVALID_REQUEST; | |
216 | } | |
3b98c0c2 LE |
217 | if (adm_ctx.mdev && |
218 | adm_ctx.volume != VOLUME_UNSPECIFIED && | |
219 | adm_ctx.volume != adm_ctx.mdev->vnr) { | |
220 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", | |
221 | adm_ctx.minor, adm_ctx.volume, | |
222 | adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name); | |
527f4b24 | 223 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
224 | return ERR_INVALID_REQUEST; |
225 | } | |
0ace9dfa | 226 | |
3b98c0c2 LE |
227 | return NO_ERROR; |
228 | ||
229 | fail: | |
230 | nlmsg_free(adm_ctx.reply_skb); | |
231 | adm_ctx.reply_skb = NULL; | |
232 | return -ENOMEM; | |
233 | } | |
234 | ||
235 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
236 | { | |
237 | struct nlattr *nla; | |
238 | const char *conn_name = NULL; | |
239 | ||
0ace9dfa PR |
240 | if (adm_ctx.tconn) { |
241 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
242 | adm_ctx.tconn = NULL; | |
243 | } | |
244 | ||
3b98c0c2 LE |
245 | if (!adm_ctx.reply_skb) |
246 | return -ENOMEM; | |
247 | ||
248 | adm_ctx.reply_dh->ret_code = retcode; | |
249 | ||
250 | nla = info->attrs[DRBD_NLA_CFG_CONTEXT]; | |
251 | if (nla) { | |
252 | nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name)); | |
253 | if (nla) | |
254 | conn_name = nla_data(nla); | |
255 | } | |
256 | ||
257 | drbd_adm_send_reply(adm_ctx.reply_skb, info); | |
258 | return 0; | |
259 | } | |
b411b363 | 260 | |
6b75dced | 261 | static void setup_khelper_env(struct drbd_tconn *tconn, char **envp) |
b411b363 | 262 | { |
6b75dced | 263 | char *afs; |
44ed167d | 264 | struct net_conf *nc; |
b411b363 | 265 | |
44ed167d PR |
266 | rcu_read_lock(); |
267 | nc = rcu_dereference(tconn->net_conf); | |
268 | if (nc) { | |
269 | switch (((struct sockaddr *)nc->peer_addr)->sa_family) { | |
b411b363 PR |
270 | case AF_INET6: |
271 | afs = "ipv6"; | |
6b75dced | 272 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", |
44ed167d | 273 | &((struct sockaddr_in6 *)nc->peer_addr)->sin6_addr); |
b411b363 PR |
274 | break; |
275 | case AF_INET: | |
276 | afs = "ipv4"; | |
6b75dced | 277 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", |
44ed167d | 278 | &((struct sockaddr_in *)nc->peer_addr)->sin_addr); |
b411b363 PR |
279 | break; |
280 | default: | |
281 | afs = "ssocks"; | |
6b75dced | 282 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", |
44ed167d | 283 | &((struct sockaddr_in *)nc->peer_addr)->sin_addr); |
b411b363 | 284 | } |
6b75dced | 285 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
b411b363 | 286 | } |
44ed167d | 287 | rcu_read_unlock(); |
6b75dced PR |
288 | } |
289 | ||
290 | int drbd_khelper(struct drbd_conf *mdev, char *cmd) | |
291 | { | |
292 | char *envp[] = { "HOME=/", | |
293 | "TERM=linux", | |
294 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
295 | (char[20]) { }, /* address family */ | |
296 | (char[60]) { }, /* address */ | |
297 | NULL }; | |
298 | char mb[12]; | |
299 | char *argv[] = {usermode_helper, cmd, mb, NULL }; | |
300 | struct sib_info sib; | |
301 | int ret; | |
302 | ||
303 | snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev)); | |
304 | setup_khelper_env(mdev->tconn, envp); | |
b411b363 | 305 | |
1090c056 LE |
306 | /* The helper may take some time. |
307 | * write out any unsynced meta data changes now */ | |
308 | drbd_md_sync(mdev); | |
309 | ||
b411b363 | 310 | dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
311 | sib.sib_reason = SIB_HELPER_PRE; |
312 | sib.helper_name = cmd; | |
313 | drbd_bcast_event(mdev, &sib); | |
b411b363 PR |
314 | ret = call_usermodehelper(usermode_helper, argv, envp, 1); |
315 | if (ret) | |
316 | dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
317 | usermode_helper, cmd, mb, | |
318 | (ret >> 8) & 0xff, ret); | |
319 | else | |
320 | dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
321 | usermode_helper, cmd, mb, | |
322 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
323 | sib.sib_reason = SIB_HELPER_POST; |
324 | sib.helper_exit_code = ret; | |
325 | drbd_bcast_event(mdev, &sib); | |
b411b363 PR |
326 | |
327 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
328 | ret = 0; | |
329 | ||
330 | return ret; | |
331 | } | |
332 | ||
6b75dced PR |
333 | static void conn_md_sync(struct drbd_tconn *tconn) |
334 | { | |
335 | struct drbd_conf *mdev; | |
e90285e0 | 336 | int vnr; |
6b75dced | 337 | |
d3fcb490 | 338 | down_read(&drbd_cfg_rwsem); |
e90285e0 | 339 | idr_for_each_entry(&tconn->volumes, mdev, vnr) |
6b75dced | 340 | drbd_md_sync(mdev); |
d3fcb490 | 341 | up_read(&drbd_cfg_rwsem); |
6b75dced PR |
342 | } |
343 | ||
344 | int conn_khelper(struct drbd_tconn *tconn, char *cmd) | |
345 | { | |
346 | char *envp[] = { "HOME=/", | |
347 | "TERM=linux", | |
348 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
349 | (char[20]) { }, /* address family */ | |
350 | (char[60]) { }, /* address */ | |
351 | NULL }; | |
352 | char *argv[] = {usermode_helper, cmd, tconn->name, NULL }; | |
353 | int ret; | |
354 | ||
355 | setup_khelper_env(tconn, envp); | |
356 | conn_md_sync(tconn); | |
357 | ||
358 | conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name); | |
359 | /* TODO: conn_bcast_event() ?? */ | |
360 | ||
361 | ret = call_usermodehelper(usermode_helper, argv, envp, 1); | |
362 | if (ret) | |
363 | conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
364 | usermode_helper, cmd, tconn->name, | |
365 | (ret >> 8) & 0xff, ret); | |
366 | else | |
367 | conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
368 | usermode_helper, cmd, tconn->name, | |
369 | (ret >> 8) & 0xff, ret); | |
370 | /* TODO: conn_bcast_event() ?? */ | |
371 | ||
372 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
373 | ret = 0; | |
374 | ||
375 | return ret; | |
376 | } | |
377 | ||
cb703454 | 378 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn) |
b411b363 | 379 | { |
cb703454 PR |
380 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
381 | struct drbd_conf *mdev; | |
382 | int vnr; | |
383 | ||
695d08fa | 384 | rcu_read_lock(); |
cb703454 PR |
385 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
386 | if (get_ldev_if_state(mdev, D_CONSISTENT)) { | |
daeda1cc PR |
387 | fp = max_t(enum drbd_fencing_p, fp, |
388 | rcu_dereference(mdev->ldev->disk_conf)->fencing); | |
cb703454 PR |
389 | put_ldev(mdev); |
390 | } | |
391 | } | |
695d08fa | 392 | rcu_read_unlock(); |
cb703454 PR |
393 | |
394 | return fp; | |
395 | } | |
396 | ||
397 | bool conn_try_outdate_peer(struct drbd_tconn *tconn) | |
398 | { | |
399 | union drbd_state mask = { }; | |
400 | union drbd_state val = { }; | |
401 | enum drbd_fencing_p fp; | |
b411b363 PR |
402 | char *ex_to_string; |
403 | int r; | |
b411b363 | 404 | |
cb703454 PR |
405 | if (tconn->cstate >= C_WF_REPORT_PARAMS) { |
406 | conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n"); | |
407 | return false; | |
408 | } | |
b411b363 | 409 | |
cb703454 PR |
410 | fp = highest_fencing_policy(tconn); |
411 | switch (fp) { | |
412 | case FP_NOT_AVAIL: | |
413 | conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n"); | |
fb22c402 | 414 | goto out; |
cb703454 PR |
415 | case FP_DONT_CARE: |
416 | return true; | |
417 | default: ; | |
b411b363 PR |
418 | } |
419 | ||
cb703454 | 420 | r = conn_khelper(tconn, "fence-peer"); |
b411b363 PR |
421 | |
422 | switch ((r>>8) & 0xff) { | |
423 | case 3: /* peer is inconsistent */ | |
424 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
425 | mask.pdsk = D_MASK; |
426 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
427 | break; |
428 | case 4: /* peer got outdated, or was already outdated */ | |
429 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
430 | mask.pdsk = D_MASK; |
431 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
432 | break; |
433 | case 5: /* peer was down */ | |
cb703454 | 434 | if (conn_highest_disk(tconn) == D_UP_TO_DATE) { |
b411b363 PR |
435 | /* we will(have) create(d) a new UUID anyways... */ |
436 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
437 | mask.pdsk = D_MASK; |
438 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
439 | } else { |
440 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
441 | } |
442 | break; | |
443 | case 6: /* Peer is primary, voluntarily outdate myself. | |
444 | * This is useful when an unconnected R_SECONDARY is asked to | |
445 | * become R_PRIMARY, but finds the other peer being active. */ | |
446 | ex_to_string = "peer is active"; | |
cb703454 PR |
447 | conn_warn(tconn, "Peer is primary, outdating myself.\n"); |
448 | mask.disk = D_MASK; | |
449 | val.disk = D_OUTDATED; | |
b411b363 PR |
450 | break; |
451 | case 7: | |
452 | if (fp != FP_STONITH) | |
cb703454 | 453 | conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 454 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
455 | mask.pdsk = D_MASK; |
456 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
457 | break; |
458 | default: | |
459 | /* The script is broken ... */ | |
cb703454 PR |
460 | conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
461 | return false; /* Eventually leave IO frozen */ | |
b411b363 PR |
462 | } |
463 | ||
cb703454 PR |
464 | conn_info(tconn, "fence-peer helper returned %d (%s)\n", |
465 | (r>>8) & 0xff, ex_to_string); | |
fb22c402 | 466 | |
cb703454 | 467 | out: |
fb22c402 | 468 | |
cb703454 PR |
469 | /* Not using |
470 | conn_request_state(tconn, mask, val, CS_VERBOSE); | |
471 | here, because we might were able to re-establish the connection in the | |
472 | meantime. */ | |
473 | spin_lock_irq(&tconn->req_lock); | |
474 | if (tconn->cstate < C_WF_REPORT_PARAMS) | |
475 | _conn_request_state(tconn, mask, val, CS_VERBOSE); | |
476 | spin_unlock_irq(&tconn->req_lock); | |
477 | ||
478 | return conn_highest_pdsk(tconn) <= D_OUTDATED; | |
b411b363 PR |
479 | } |
480 | ||
87f7be4c PR |
481 | static int _try_outdate_peer_async(void *data) |
482 | { | |
cb703454 | 483 | struct drbd_tconn *tconn = (struct drbd_tconn *)data; |
87f7be4c | 484 | |
cb703454 | 485 | conn_try_outdate_peer(tconn); |
87f7be4c | 486 | |
9dc9fbb3 | 487 | kref_put(&tconn->kref, &conn_destroy); |
87f7be4c PR |
488 | return 0; |
489 | } | |
490 | ||
cb703454 | 491 | void conn_try_outdate_peer_async(struct drbd_tconn *tconn) |
87f7be4c PR |
492 | { |
493 | struct task_struct *opa; | |
494 | ||
9dc9fbb3 | 495 | kref_get(&tconn->kref); |
cb703454 | 496 | opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h"); |
9dc9fbb3 | 497 | if (IS_ERR(opa)) { |
cb703454 | 498 | conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n"); |
9dc9fbb3 PR |
499 | kref_put(&tconn->kref, &conn_destroy); |
500 | } | |
87f7be4c | 501 | } |
b411b363 | 502 | |
bf885f8a AG |
503 | enum drbd_state_rv |
504 | drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) | |
b411b363 PR |
505 | { |
506 | const int max_tries = 4; | |
bf885f8a | 507 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 508 | struct net_conf *nc; |
b411b363 PR |
509 | int try = 0; |
510 | int forced = 0; | |
511 | union drbd_state mask, val; | |
b411b363 PR |
512 | |
513 | if (new_role == R_PRIMARY) | |
0625ac19 | 514 | request_ping(mdev->tconn); /* Detect a dead peer ASAP */ |
b411b363 | 515 | |
8410da8f | 516 | mutex_lock(mdev->state_mutex); |
b411b363 PR |
517 | |
518 | mask.i = 0; mask.role = R_MASK; | |
519 | val.i = 0; val.role = new_role; | |
520 | ||
521 | while (try++ < max_tries) { | |
bf885f8a | 522 | rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
523 | |
524 | /* in case we first succeeded to outdate, | |
525 | * but now suddenly could establish a connection */ | |
bf885f8a | 526 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
527 | val.pdsk = 0; |
528 | mask.pdsk = 0; | |
529 | continue; | |
530 | } | |
531 | ||
bf885f8a | 532 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
d10a33c6 PR |
533 | (mdev->state.disk < D_UP_TO_DATE && |
534 | mdev->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
535 | mask.disk = D_MASK; |
536 | val.disk = D_UP_TO_DATE; | |
537 | forced = 1; | |
538 | continue; | |
539 | } | |
540 | ||
bf885f8a | 541 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b411b363 PR |
542 | mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
543 | D_ASSERT(mdev->state.pdsk == D_UNKNOWN); | |
b411b363 | 544 | |
cb703454 | 545 | if (conn_try_outdate_peer(mdev->tconn)) { |
b411b363 PR |
546 | val.disk = D_UP_TO_DATE; |
547 | mask.disk = D_MASK; | |
548 | } | |
b411b363 PR |
549 | continue; |
550 | } | |
551 | ||
bf885f8a | 552 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 553 | goto out; |
bf885f8a | 554 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
cb703454 | 555 | if (!conn_try_outdate_peer(mdev->tconn) && force) { |
b411b363 | 556 | dev_warn(DEV, "Forced into split brain situation!\n"); |
cb703454 PR |
557 | mask.pdsk = D_MASK; |
558 | val.pdsk = D_OUTDATED; | |
b411b363 | 559 | |
cb703454 | 560 | } |
b411b363 PR |
561 | continue; |
562 | } | |
bf885f8a | 563 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
564 | /* Maybe the peer is detected as dead very soon... |
565 | retry at most once more in this case. */ | |
44ed167d PR |
566 | int timeo; |
567 | rcu_read_lock(); | |
568 | nc = rcu_dereference(mdev->tconn->net_conf); | |
569 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; | |
570 | rcu_read_unlock(); | |
571 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
572 | if (try < max_tries) |
573 | try = max_tries - 1; | |
574 | continue; | |
575 | } | |
bf885f8a AG |
576 | if (rv < SS_SUCCESS) { |
577 | rv = _drbd_request_state(mdev, mask, val, | |
b411b363 | 578 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 579 | if (rv < SS_SUCCESS) |
3b98c0c2 | 580 | goto out; |
b411b363 PR |
581 | } |
582 | break; | |
583 | } | |
584 | ||
bf885f8a | 585 | if (rv < SS_SUCCESS) |
3b98c0c2 | 586 | goto out; |
b411b363 PR |
587 | |
588 | if (forced) | |
589 | dev_warn(DEV, "Forced to consider local data as UpToDate!\n"); | |
590 | ||
591 | /* Wait until nothing is on the fly :) */ | |
592 | wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0); | |
593 | ||
594 | if (new_role == R_SECONDARY) { | |
81e84650 | 595 | set_disk_ro(mdev->vdisk, true); |
b411b363 PR |
596 | if (get_ldev(mdev)) { |
597 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
598 | put_ldev(mdev); | |
599 | } | |
600 | } else { | |
a0095508 | 601 | mutex_lock(&mdev->tconn->conf_update); |
91fd4dad | 602 | nc = mdev->tconn->net_conf; |
44ed167d | 603 | if (nc) |
91fd4dad | 604 | nc->want_lose = 0; /* without copy; single bit op is atomic */ |
a0095508 | 605 | mutex_unlock(&mdev->tconn->conf_update); |
91fd4dad | 606 | |
81e84650 | 607 | set_disk_ro(mdev->vdisk, false); |
b411b363 PR |
608 | if (get_ldev(mdev)) { |
609 | if (((mdev->state.conn < C_CONNECTED || | |
610 | mdev->state.pdsk <= D_FAILED) | |
611 | && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
612 | drbd_uuid_new_current(mdev); | |
613 | ||
614 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
615 | put_ldev(mdev); | |
616 | } | |
617 | } | |
618 | ||
19f843aa LE |
619 | /* writeout of activity log covered areas of the bitmap |
620 | * to stable storage done in after state change already */ | |
b411b363 PR |
621 | |
622 | if (mdev->state.conn >= C_WF_REPORT_PARAMS) { | |
623 | /* if this was forced, we should consider sync */ | |
624 | if (forced) | |
625 | drbd_send_uuids(mdev); | |
626 | drbd_send_state(mdev); | |
627 | } | |
628 | ||
629 | drbd_md_sync(mdev); | |
630 | ||
631 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
3b98c0c2 | 632 | out: |
8410da8f | 633 | mutex_unlock(mdev->state_mutex); |
bf885f8a | 634 | return rv; |
b411b363 PR |
635 | } |
636 | ||
3b98c0c2 | 637 | static const char *from_attrs_err_to_txt(int err) |
b411b363 | 638 | { |
3b98c0c2 LE |
639 | return err == -ENOMSG ? "required attribute missing" : |
640 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 641 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 642 | "invalid attribute value"; |
b411b363 PR |
643 | } |
644 | ||
3b98c0c2 | 645 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 646 | { |
3b98c0c2 LE |
647 | struct set_role_parms parms; |
648 | int err; | |
649 | enum drbd_ret_code retcode; | |
b411b363 | 650 | |
3b98c0c2 LE |
651 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
652 | if (!adm_ctx.reply_skb) | |
653 | return retcode; | |
654 | if (retcode != NO_ERROR) | |
655 | goto out; | |
656 | ||
657 | memset(&parms, 0, sizeof(parms)); | |
658 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 659 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
660 | if (err) { |
661 | retcode = ERR_MANDATORY_TAG; | |
662 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
663 | goto out; | |
664 | } | |
665 | } | |
666 | ||
667 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) | |
668 | retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate); | |
669 | else | |
670 | retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0); | |
671 | out: | |
672 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
673 | return 0; |
674 | } | |
675 | ||
676 | /* initializes the md.*_offset members, so we are able to find | |
677 | * the on disk meta data */ | |
678 | static void drbd_md_set_sector_offsets(struct drbd_conf *mdev, | |
679 | struct drbd_backing_dev *bdev) | |
680 | { | |
681 | sector_t md_size_sect = 0; | |
daeda1cc PR |
682 | int meta_dev_idx; |
683 | ||
684 | rcu_read_lock(); | |
685 | meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx; | |
686 | ||
687 | switch (meta_dev_idx) { | |
b411b363 PR |
688 | default: |
689 | /* v07 style fixed size indexed meta data */ | |
690 | bdev->md.md_size_sect = MD_RESERVED_SECT; | |
691 | bdev->md.md_offset = drbd_md_ss__(mdev, bdev); | |
692 | bdev->md.al_offset = MD_AL_OFFSET; | |
693 | bdev->md.bm_offset = MD_BM_OFFSET; | |
694 | break; | |
695 | case DRBD_MD_INDEX_FLEX_EXT: | |
696 | /* just occupy the full device; unit: sectors */ | |
697 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
698 | bdev->md.md_offset = 0; | |
699 | bdev->md.al_offset = MD_AL_OFFSET; | |
700 | bdev->md.bm_offset = MD_BM_OFFSET; | |
701 | break; | |
702 | case DRBD_MD_INDEX_INTERNAL: | |
703 | case DRBD_MD_INDEX_FLEX_INT: | |
704 | bdev->md.md_offset = drbd_md_ss__(mdev, bdev); | |
705 | /* al size is still fixed */ | |
7ad651b5 | 706 | bdev->md.al_offset = -MD_AL_SECTORS; |
b411b363 PR |
707 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
708 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
709 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
710 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
711 | md_size_sect = ALIGN(md_size_sect, 8); | |
712 | ||
713 | /* plus the "drbd meta data super block", | |
714 | * and the activity log; */ | |
715 | md_size_sect += MD_BM_OFFSET; | |
716 | ||
717 | bdev->md.md_size_sect = md_size_sect; | |
718 | /* bitmap offset is adjusted by 'super' block size */ | |
719 | bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET; | |
720 | break; | |
721 | } | |
daeda1cc | 722 | rcu_read_unlock(); |
b411b363 PR |
723 | } |
724 | ||
4b0715f0 | 725 | /* input size is expected to be in KB */ |
b411b363 PR |
726 | char *ppsize(char *buf, unsigned long long size) |
727 | { | |
4b0715f0 LE |
728 | /* Needs 9 bytes at max including trailing NUL: |
729 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
730 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
731 | int base = 0; | |
4b0715f0 | 732 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
733 | /* shift + round */ |
734 | size = (size >> 10) + !!(size & (1<<9)); | |
735 | base++; | |
736 | } | |
4b0715f0 | 737 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
738 | |
739 | return buf; | |
740 | } | |
741 | ||
742 | /* there is still a theoretical deadlock when called from receiver | |
743 | * on an D_INCONSISTENT R_PRIMARY: | |
744 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
745 | * packet from remote to dec_ap_bio again. | |
746 | * receiver receive_sizes(), comes here, | |
747 | * waits for ap_bio_cnt == 0. -> deadlock. | |
748 | * but this cannot happen, actually, because: | |
749 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
750 | * (not connected, or bad/no disk on peer): | |
751 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
752 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
753 | * peer may not initiate a resize. | |
754 | */ | |
3b98c0c2 LE |
755 | /* Note these are not to be confused with |
756 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
757 | * which are (sub) state changes triggered by admin (drbdsetup), | |
758 | * and can be long lived. | |
759 | * This changes an mdev->flag, is triggered by drbd internals, | |
760 | * and should be short-lived. */ | |
b411b363 PR |
761 | void drbd_suspend_io(struct drbd_conf *mdev) |
762 | { | |
763 | set_bit(SUSPEND_IO, &mdev->flags); | |
2aebfabb | 764 | if (drbd_suspended(mdev)) |
265be2d0 | 765 | return; |
b411b363 PR |
766 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); |
767 | } | |
768 | ||
769 | void drbd_resume_io(struct drbd_conf *mdev) | |
770 | { | |
771 | clear_bit(SUSPEND_IO, &mdev->flags); | |
772 | wake_up(&mdev->misc_wait); | |
773 | } | |
774 | ||
775 | /** | |
776 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
777 | * @mdev: DRBD device. | |
778 | * | |
779 | * Returns 0 on success, negative return values indicate errors. | |
780 | * You should call drbd_md_sync() after calling this function. | |
781 | */ | |
24c4830c | 782 | enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local) |
b411b363 PR |
783 | { |
784 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
ef5e44a6 | 785 | sector_t la_size, u_size; |
b411b363 PR |
786 | sector_t size; |
787 | char ppb[10]; | |
788 | ||
789 | int md_moved, la_size_changed; | |
790 | enum determine_dev_size rv = unchanged; | |
791 | ||
792 | /* race: | |
793 | * application request passes inc_ap_bio, | |
794 | * but then cannot get an AL-reference. | |
795 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
796 | * | |
797 | * to avoid that: | |
798 | * Suspend IO right here. | |
799 | * still lock the act_log to not trigger ASSERTs there. | |
800 | */ | |
801 | drbd_suspend_io(mdev); | |
802 | ||
803 | /* no wait necessary anymore, actually we could assert that */ | |
804 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); | |
805 | ||
806 | prev_first_sect = drbd_md_first_sector(mdev->ldev); | |
807 | prev_size = mdev->ldev->md.md_size_sect; | |
808 | la_size = mdev->ldev->md.la_size_sect; | |
809 | ||
810 | /* TODO: should only be some assert here, not (re)init... */ | |
811 | drbd_md_set_sector_offsets(mdev, mdev->ldev); | |
812 | ||
daeda1cc PR |
813 | rcu_read_lock(); |
814 | u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size; | |
815 | rcu_read_unlock(); | |
ef5e44a6 | 816 | size = drbd_new_dev_size(mdev, mdev->ldev, u_size, flags & DDSF_FORCED); |
b411b363 PR |
817 | |
818 | if (drbd_get_capacity(mdev->this_bdev) != size || | |
819 | drbd_bm_capacity(mdev) != size) { | |
820 | int err; | |
02d9a94b | 821 | err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
822 | if (unlikely(err)) { |
823 | /* currently there is only one error: ENOMEM! */ | |
824 | size = drbd_bm_capacity(mdev)>>1; | |
825 | if (size == 0) { | |
826 | dev_err(DEV, "OUT OF MEMORY! " | |
827 | "Could not allocate bitmap!\n"); | |
828 | } else { | |
829 | dev_err(DEV, "BM resizing failed. " | |
830 | "Leaving size unchanged at size = %lu KB\n", | |
831 | (unsigned long)size); | |
832 | } | |
833 | rv = dev_size_error; | |
834 | } | |
835 | /* racy, see comments above. */ | |
836 | drbd_set_my_capacity(mdev, size); | |
837 | mdev->ldev->md.la_size_sect = size; | |
838 | dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), | |
839 | (unsigned long long)size>>1); | |
840 | } | |
841 | if (rv == dev_size_error) | |
842 | goto out; | |
843 | ||
844 | la_size_changed = (la_size != mdev->ldev->md.la_size_sect); | |
845 | ||
846 | md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev) | |
847 | || prev_size != mdev->ldev->md.md_size_sect; | |
848 | ||
849 | if (la_size_changed || md_moved) { | |
24dccabb AG |
850 | int err; |
851 | ||
b411b363 PR |
852 | drbd_al_shrink(mdev); /* All extents inactive. */ |
853 | dev_info(DEV, "Writing the whole bitmap, %s\n", | |
854 | la_size_changed && md_moved ? "size changed and md moved" : | |
855 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 LE |
856 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
857 | err = drbd_bitmap_io(mdev, &drbd_bm_write, | |
858 | "size changed", BM_LOCKED_MASK); | |
24dccabb AG |
859 | if (err) { |
860 | rv = dev_size_error; | |
861 | goto out; | |
862 | } | |
b411b363 PR |
863 | drbd_md_mark_dirty(mdev); |
864 | } | |
865 | ||
866 | if (size > la_size) | |
867 | rv = grew; | |
868 | if (size < la_size) | |
869 | rv = shrunk; | |
870 | out: | |
871 | lc_unlock(mdev->act_log); | |
872 | wake_up(&mdev->al_wait); | |
873 | drbd_resume_io(mdev); | |
874 | ||
875 | return rv; | |
876 | } | |
877 | ||
878 | sector_t | |
ef5e44a6 PR |
879 | drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, |
880 | sector_t u_size, int assume_peer_has_space) | |
b411b363 PR |
881 | { |
882 | sector_t p_size = mdev->p_size; /* partner's disk size. */ | |
883 | sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */ | |
884 | sector_t m_size; /* my size */ | |
b411b363 PR |
885 | sector_t size = 0; |
886 | ||
887 | m_size = drbd_get_max_capacity(bdev); | |
888 | ||
a393db6f PR |
889 | if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) { |
890 | dev_warn(DEV, "Resize while not connected was forced by the user!\n"); | |
891 | p_size = m_size; | |
892 | } | |
893 | ||
b411b363 PR |
894 | if (p_size && m_size) { |
895 | size = min_t(sector_t, p_size, m_size); | |
896 | } else { | |
897 | if (la_size) { | |
898 | size = la_size; | |
899 | if (m_size && m_size < size) | |
900 | size = m_size; | |
901 | if (p_size && p_size < size) | |
902 | size = p_size; | |
903 | } else { | |
904 | if (m_size) | |
905 | size = m_size; | |
906 | if (p_size) | |
907 | size = p_size; | |
908 | } | |
909 | } | |
910 | ||
911 | if (size == 0) | |
912 | dev_err(DEV, "Both nodes diskless!\n"); | |
913 | ||
914 | if (u_size) { | |
915 | if (u_size > size) | |
916 | dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n", | |
917 | (unsigned long)u_size>>1, (unsigned long)size>>1); | |
918 | else | |
919 | size = u_size; | |
920 | } | |
921 | ||
922 | return size; | |
923 | } | |
924 | ||
925 | /** | |
926 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
927 | * @mdev: DRBD device. | |
928 | * | |
929 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
930 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
931 | * this function. | |
932 | */ | |
f399002e | 933 | static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc) |
b411b363 PR |
934 | { |
935 | struct lru_cache *n, *t; | |
936 | struct lc_element *e; | |
937 | unsigned int in_use; | |
938 | int i; | |
939 | ||
f399002e LE |
940 | if (!expect(dc->al_extents >= DRBD_AL_EXTENTS_MIN)) |
941 | dc->al_extents = DRBD_AL_EXTENTS_MIN; | |
b411b363 PR |
942 | |
943 | if (mdev->act_log && | |
f399002e | 944 | mdev->act_log->nr_elements == dc->al_extents) |
b411b363 PR |
945 | return 0; |
946 | ||
947 | in_use = 0; | |
948 | t = mdev->act_log; | |
7ad651b5 | 949 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 950 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
951 | |
952 | if (n == NULL) { | |
953 | dev_err(DEV, "Cannot allocate act_log lru!\n"); | |
954 | return -ENOMEM; | |
955 | } | |
956 | spin_lock_irq(&mdev->al_lock); | |
957 | if (t) { | |
958 | for (i = 0; i < t->nr_elements; i++) { | |
959 | e = lc_element_by_index(t, i); | |
960 | if (e->refcnt) | |
961 | dev_err(DEV, "refcnt(%d)==%d\n", | |
962 | e->lc_number, e->refcnt); | |
963 | in_use += e->refcnt; | |
964 | } | |
965 | } | |
966 | if (!in_use) | |
967 | mdev->act_log = n; | |
968 | spin_unlock_irq(&mdev->al_lock); | |
969 | if (in_use) { | |
970 | dev_err(DEV, "Activity log still in use!\n"); | |
971 | lc_destroy(n); | |
972 | return -EBUSY; | |
973 | } else { | |
974 | if (t) | |
975 | lc_destroy(t); | |
976 | } | |
977 | drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */ | |
978 | return 0; | |
979 | } | |
980 | ||
99432fcc | 981 | static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size) |
b411b363 PR |
982 | { |
983 | struct request_queue * const q = mdev->rq_queue; | |
99432fcc PR |
984 | int max_hw_sectors = max_bio_size >> 9; |
985 | int max_segments = 0; | |
986 | ||
987 | if (get_ldev_if_state(mdev, D_ATTACHING)) { | |
988 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
989 | ||
990 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc PR |
991 | rcu_read_lock(); |
992 | max_segments = rcu_dereference(mdev->ldev->disk_conf)->max_bio_bvecs; | |
993 | rcu_read_unlock(); | |
99432fcc PR |
994 | put_ldev(mdev); |
995 | } | |
b411b363 | 996 | |
b411b363 | 997 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
998 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
999 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1000 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1001 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1002 | |
99432fcc PR |
1003 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1004 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
1005 | ||
1006 | blk_queue_stack_limits(q, b); | |
1007 | ||
1008 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
1009 | dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", | |
1010 | q->backing_dev_info.ra_pages, | |
1011 | b->backing_dev_info.ra_pages); | |
1012 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1013 | } | |
1014 | put_ldev(mdev); | |
1015 | } | |
1016 | } | |
1017 | ||
1018 | void drbd_reconsider_max_bio_size(struct drbd_conf *mdev) | |
1019 | { | |
1020 | int now, new, local, peer; | |
1021 | ||
1022 | now = queue_max_hw_sectors(mdev->rq_queue) << 9; | |
1023 | local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1024 | peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1025 | |
99432fcc PR |
1026 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1027 | local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9; | |
1028 | mdev->local_max_bio_size = local; | |
1029 | put_ldev(mdev); | |
b411b363 | 1030 | } |
99432fcc PR |
1031 | |
1032 | /* We may ignore peer limits if the peer is modern enough. | |
1033 | Because new from 8.3.8 onwards the peer can use multiple | |
1034 | BIOs for a single peer_request */ | |
1035 | if (mdev->state.conn >= C_CONNECTED) { | |
31890f4a | 1036 | if (mdev->tconn->agreed_pro_version < 94) |
99432fcc | 1037 | peer = mdev->peer_max_bio_size; |
31890f4a | 1038 | else if (mdev->tconn->agreed_pro_version == 94) |
99432fcc PR |
1039 | peer = DRBD_MAX_SIZE_H80_PACKET; |
1040 | else /* drbd 8.3.8 onwards */ | |
1041 | peer = DRBD_MAX_BIO_SIZE; | |
1042 | } | |
1043 | ||
1044 | new = min_t(int, local, peer); | |
1045 | ||
1046 | if (mdev->state.role == R_PRIMARY && new < now) | |
1047 | dev_err(DEV, "ASSERT FAILED new < now; (%d < %d)\n", new, now); | |
1048 | ||
1049 | if (new != now) | |
1050 | dev_info(DEV, "max BIO size = %u\n", new); | |
1051 | ||
1052 | drbd_setup_queue_param(mdev, new); | |
b411b363 PR |
1053 | } |
1054 | ||
a18e9d1e | 1055 | /* Starts the worker thread */ |
0e29d163 | 1056 | static void conn_reconfig_start(struct drbd_tconn *tconn) |
b411b363 | 1057 | { |
0e29d163 PR |
1058 | drbd_thread_start(&tconn->worker); |
1059 | conn_flush_workqueue(tconn); | |
b411b363 PR |
1060 | } |
1061 | ||
a18e9d1e | 1062 | /* if still unconfigured, stops worker again. */ |
0e29d163 | 1063 | static void conn_reconfig_done(struct drbd_tconn *tconn) |
b411b363 | 1064 | { |
992d6e91 | 1065 | bool stop_threads; |
0e29d163 | 1066 | spin_lock_irq(&tconn->req_lock); |
992d6e91 | 1067 | stop_threads = conn_all_vols_unconf(tconn); |
0e29d163 | 1068 | spin_unlock_irq(&tconn->req_lock); |
992d6e91 LE |
1069 | if (stop_threads) { |
1070 | /* asender is implicitly stopped by receiver | |
1071 | * in drbd_disconnect() */ | |
1072 | drbd_thread_stop(&tconn->receiver); | |
1073 | drbd_thread_stop(&tconn->worker); | |
1074 | } | |
b411b363 PR |
1075 | } |
1076 | ||
0778286a PR |
1077 | /* Make sure IO is suspended before calling this function(). */ |
1078 | static void drbd_suspend_al(struct drbd_conf *mdev) | |
1079 | { | |
1080 | int s = 0; | |
1081 | ||
61610420 | 1082 | if (!lc_try_lock(mdev->act_log)) { |
0778286a PR |
1083 | dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n"); |
1084 | return; | |
1085 | } | |
1086 | ||
61610420 | 1087 | drbd_al_shrink(mdev); |
87eeee41 | 1088 | spin_lock_irq(&mdev->tconn->req_lock); |
0778286a PR |
1089 | if (mdev->state.conn < C_CONNECTED) |
1090 | s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags); | |
87eeee41 | 1091 | spin_unlock_irq(&mdev->tconn->req_lock); |
61610420 | 1092 | lc_unlock(mdev->act_log); |
0778286a PR |
1093 | |
1094 | if (s) | |
1095 | dev_info(DEV, "Suspended AL updates\n"); | |
1096 | } | |
1097 | ||
5979e361 LE |
1098 | |
1099 | static bool should_set_defaults(struct genl_info *info) | |
1100 | { | |
1101 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1102 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1103 | } | |
1104 | ||
f399002e LE |
1105 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1106 | { | |
1107 | enum drbd_ret_code retcode; | |
1108 | struct drbd_conf *mdev; | |
daeda1cc | 1109 | struct disk_conf *new_disk_conf, *old_disk_conf; |
f399002e LE |
1110 | int err, fifo_size; |
1111 | int *rs_plan_s = NULL; | |
1112 | ||
1113 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1114 | if (!adm_ctx.reply_skb) | |
1115 | return retcode; | |
1116 | if (retcode != NO_ERROR) | |
1117 | goto out; | |
1118 | ||
1119 | mdev = adm_ctx.mdev; | |
1120 | ||
1121 | /* we also need a disk | |
1122 | * to change the options on */ | |
1123 | if (!get_ldev(mdev)) { | |
1124 | retcode = ERR_NO_DISK; | |
1125 | goto out; | |
1126 | } | |
1127 | ||
daeda1cc | 1128 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1129 | if (!new_disk_conf) { |
f399002e LE |
1130 | retcode = ERR_NOMEM; |
1131 | goto fail; | |
1132 | } | |
1133 | ||
daeda1cc PR |
1134 | mutex_lock(&mdev->tconn->conf_update); |
1135 | old_disk_conf = mdev->ldev->disk_conf; | |
1136 | *new_disk_conf = *old_disk_conf; | |
5979e361 | 1137 | if (should_set_defaults(info)) |
b966b5dd | 1138 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1139 | |
5ecc72c3 | 1140 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
f399002e LE |
1141 | if (err) { |
1142 | retcode = ERR_MANDATORY_TAG; | |
1143 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1144 | } | |
1145 | ||
5ecc72c3 LE |
1146 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1147 | new_disk_conf->resync_rate = 1; | |
f399002e LE |
1148 | |
1149 | /* clip to allowed range */ | |
5ecc72c3 LE |
1150 | if (!expect(new_disk_conf->al_extents >= DRBD_AL_EXTENTS_MIN)) |
1151 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1152 | if (!expect(new_disk_conf->al_extents <= DRBD_AL_EXTENTS_MAX)) | |
1153 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MAX; | |
f399002e | 1154 | |
5ecc72c3 | 1155 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
f399002e LE |
1156 | if (fifo_size != mdev->rs_plan_s.size && fifo_size > 0) { |
1157 | rs_plan_s = kzalloc(sizeof(int) * fifo_size, GFP_KERNEL); | |
1158 | if (!rs_plan_s) { | |
1159 | dev_err(DEV, "kmalloc of fifo_buffer failed"); | |
1160 | retcode = ERR_NOMEM; | |
daeda1cc | 1161 | goto fail_unlock; |
f399002e LE |
1162 | } |
1163 | } | |
1164 | ||
1165 | if (fifo_size != mdev->rs_plan_s.size) { | |
1166 | kfree(mdev->rs_plan_s.values); | |
1167 | mdev->rs_plan_s.values = rs_plan_s; | |
1168 | mdev->rs_plan_s.size = fifo_size; | |
1169 | mdev->rs_planed = 0; | |
1170 | rs_plan_s = NULL; | |
1171 | } | |
1172 | ||
1173 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); | |
1174 | drbd_al_shrink(mdev); | |
5ecc72c3 | 1175 | err = drbd_check_al_size(mdev, new_disk_conf); |
f399002e LE |
1176 | lc_unlock(mdev->act_log); |
1177 | wake_up(&mdev->al_wait); | |
1178 | ||
1179 | if (err) { | |
1180 | retcode = ERR_NOMEM; | |
daeda1cc | 1181 | goto fail_unlock; |
f399002e LE |
1182 | } |
1183 | ||
dc97b708 PR |
1184 | write_lock_irq(&global_state_lock); |
1185 | retcode = drbd_sync_after_valid(mdev, new_disk_conf->resync_after); | |
1186 | if (retcode == NO_ERROR) { | |
daeda1cc | 1187 | rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf); |
dc97b708 PR |
1188 | drbd_sync_after_changed(mdev); |
1189 | } | |
1190 | write_unlock_irq(&global_state_lock); | |
f399002e | 1191 | |
daeda1cc PR |
1192 | if (retcode != NO_ERROR) |
1193 | goto fail_unlock; | |
f399002e | 1194 | |
daeda1cc | 1195 | drbd_md_sync(mdev); |
f399002e LE |
1196 | |
1197 | if (mdev->state.conn >= C_CONNECTED) | |
1198 | drbd_send_sync_param(mdev); | |
1199 | ||
daeda1cc PR |
1200 | mutex_unlock(&mdev->tconn->conf_update); |
1201 | synchronize_rcu(); | |
1202 | kfree(old_disk_conf); | |
1203 | goto success; | |
1204 | ||
1205 | fail_unlock: | |
1206 | mutex_unlock(&mdev->tconn->conf_update); | |
f399002e | 1207 | fail: |
5ecc72c3 | 1208 | kfree(new_disk_conf); |
f399002e | 1209 | kfree(rs_plan_s); |
daeda1cc PR |
1210 | success: |
1211 | put_ldev(mdev); | |
f399002e LE |
1212 | out: |
1213 | drbd_adm_finish(info, retcode); | |
1214 | return 0; | |
1215 | } | |
1216 | ||
3b98c0c2 | 1217 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1218 | { |
3b98c0c2 LE |
1219 | struct drbd_conf *mdev; |
1220 | int err; | |
116676ca | 1221 | enum drbd_ret_code retcode; |
b411b363 PR |
1222 | enum determine_dev_size dd; |
1223 | sector_t max_possible_sectors; | |
1224 | sector_t min_md_device_sectors; | |
1225 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1226 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1227 | struct block_device *bdev; |
b411b363 PR |
1228 | struct lru_cache *resync_lru = NULL; |
1229 | union drbd_state ns, os; | |
f2024e7c | 1230 | enum drbd_state_rv rv; |
44ed167d | 1231 | struct net_conf *nc; |
b411b363 | 1232 | int cp_discovered = 0; |
b411b363 | 1233 | |
3b98c0c2 LE |
1234 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1235 | if (!adm_ctx.reply_skb) | |
1236 | return retcode; | |
1237 | if (retcode != NO_ERROR) | |
40cbf085 | 1238 | goto finish; |
3b98c0c2 LE |
1239 | |
1240 | mdev = adm_ctx.mdev; | |
0e29d163 | 1241 | conn_reconfig_start(mdev->tconn); |
b411b363 PR |
1242 | |
1243 | /* if you want to reconfigure, please tear down first */ | |
1244 | if (mdev->state.disk > D_DISKLESS) { | |
1245 | retcode = ERR_DISK_CONFIGURED; | |
1246 | goto fail; | |
1247 | } | |
82f59cc6 LE |
1248 | /* It may just now have detached because of IO error. Make sure |
1249 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1250 | * e.g. if someone calls attach from the on-io-error handler, | |
1251 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
1252 | wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); | |
b411b363 | 1253 | |
3b98c0c2 | 1254 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1255 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1256 | if (!nbc) { | |
1257 | retcode = ERR_NOMEM; | |
1258 | goto fail; | |
1259 | } | |
daeda1cc PR |
1260 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1261 | if (!new_disk_conf) { | |
1262 | retcode = ERR_NOMEM; | |
1263 | goto fail; | |
1264 | } | |
1265 | nbc->disk_conf = new_disk_conf; | |
b411b363 | 1266 | |
daeda1cc PR |
1267 | set_disk_conf_defaults(new_disk_conf); |
1268 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1269 | if (err) { |
b411b363 | 1270 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1271 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1272 | goto fail; |
1273 | } | |
1274 | ||
daeda1cc | 1275 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1276 | retcode = ERR_MD_IDX_INVALID; |
1277 | goto fail; | |
1278 | } | |
1279 | ||
44ed167d PR |
1280 | rcu_read_lock(); |
1281 | nc = rcu_dereference(mdev->tconn->net_conf); | |
1282 | if (nc) { | |
daeda1cc | 1283 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1284 | rcu_read_unlock(); |
47ff2d0a PR |
1285 | retcode = ERR_STONITH_AND_PROT_A; |
1286 | goto fail; | |
1287 | } | |
1288 | } | |
44ed167d | 1289 | rcu_read_unlock(); |
47ff2d0a | 1290 | |
daeda1cc | 1291 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
d4d77629 | 1292 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev); |
e525fd89 | 1293 | if (IS_ERR(bdev)) { |
daeda1cc | 1294 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1295 | PTR_ERR(bdev)); |
b411b363 PR |
1296 | retcode = ERR_OPEN_DISK; |
1297 | goto fail; | |
1298 | } | |
e525fd89 TH |
1299 | nbc->backing_bdev = bdev; |
1300 | ||
1301 | /* | |
1302 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1303 | * drbd sharing one meta device. TODO in that case, paranoia | |
1304 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1305 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1306 | * should check it for you already; but if you don't, or | |
1307 | * someone fooled it, we need to double check here) | |
1308 | */ | |
daeda1cc | 1309 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1310 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1311 | (new_disk_conf->meta_dev_idx < 0) ? |
d4d77629 | 1312 | (void *)mdev : (void *)drbd_m_holder); |
e525fd89 | 1313 | if (IS_ERR(bdev)) { |
daeda1cc | 1314 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1315 | PTR_ERR(bdev)); |
b411b363 PR |
1316 | retcode = ERR_OPEN_MD_DISK; |
1317 | goto fail; | |
1318 | } | |
e525fd89 | 1319 | nbc->md_bdev = bdev; |
b411b363 | 1320 | |
e525fd89 | 1321 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1322 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1323 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1324 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1325 | goto fail; |
1326 | } | |
1327 | ||
1328 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1329 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1330 | offsetof(struct bm_extent, lce)); |
1331 | if (!resync_lru) { | |
1332 | retcode = ERR_NOMEM; | |
e525fd89 | 1333 | goto fail; |
b411b363 PR |
1334 | } |
1335 | ||
1336 | /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */ | |
1337 | drbd_md_set_sector_offsets(mdev, nbc); | |
1338 | ||
daeda1cc | 1339 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
b411b363 PR |
1340 | dev_err(DEV, "max capacity %llu smaller than disk size %llu\n", |
1341 | (unsigned long long) drbd_get_max_capacity(nbc), | |
daeda1cc | 1342 | (unsigned long long) new_disk_conf->disk_size); |
b411b363 | 1343 | retcode = ERR_DISK_TO_SMALL; |
e525fd89 | 1344 | goto fail; |
b411b363 PR |
1345 | } |
1346 | ||
daeda1cc | 1347 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1348 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1349 | /* at least one MB, otherwise it does not make sense */ | |
1350 | min_md_device_sectors = (2<<10); | |
1351 | } else { | |
1352 | max_possible_sectors = DRBD_MAX_SECTORS; | |
daeda1cc | 1353 | min_md_device_sectors = MD_RESERVED_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1354 | } |
1355 | ||
b411b363 PR |
1356 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
1357 | retcode = ERR_MD_DISK_TO_SMALL; | |
1358 | dev_warn(DEV, "refusing attach: md-device too small, " | |
1359 | "at least %llu sectors needed for this meta-disk type\n", | |
1360 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1361 | goto fail; |
b411b363 PR |
1362 | } |
1363 | ||
1364 | /* Make sure the new disk is big enough | |
1365 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1366 | if (drbd_get_max_capacity(nbc) < | |
1367 | drbd_get_capacity(mdev->this_bdev)) { | |
1368 | retcode = ERR_DISK_TO_SMALL; | |
e525fd89 | 1369 | goto fail; |
b411b363 PR |
1370 | } |
1371 | ||
1372 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1373 | ||
1352994b LE |
1374 | if (nbc->known_size > max_possible_sectors) { |
1375 | dev_warn(DEV, "==> truncating very big lower level device " | |
1376 | "to currently maximum possible %llu sectors <==\n", | |
1377 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1378 | if (new_disk_conf->meta_dev_idx >= 0) |
1352994b LE |
1379 | dev_warn(DEV, "==>> using internal or flexible " |
1380 | "meta data may help <<==\n"); | |
1381 | } | |
1382 | ||
b411b363 PR |
1383 | drbd_suspend_io(mdev); |
1384 | /* also wait for the last barrier ack. */ | |
2aebfabb | 1385 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev)); |
b411b363 | 1386 | /* and for any other previously queued work */ |
a21e9298 | 1387 | drbd_flush_workqueue(mdev); |
b411b363 | 1388 | |
f2024e7c AG |
1389 | rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE); |
1390 | retcode = rv; /* FIXME: Type mismatch. */ | |
b411b363 | 1391 | drbd_resume_io(mdev); |
f2024e7c | 1392 | if (rv < SS_SUCCESS) |
e525fd89 | 1393 | goto fail; |
b411b363 PR |
1394 | |
1395 | if (!get_ldev_if_state(mdev, D_ATTACHING)) | |
1396 | goto force_diskless; | |
1397 | ||
1398 | drbd_md_set_sector_offsets(mdev, nbc); | |
1399 | ||
1400 | if (!mdev->bitmap) { | |
1401 | if (drbd_bm_init(mdev)) { | |
1402 | retcode = ERR_NOMEM; | |
1403 | goto force_diskless_dec; | |
1404 | } | |
1405 | } | |
1406 | ||
1407 | retcode = drbd_md_read(mdev, nbc); | |
1408 | if (retcode != NO_ERROR) | |
1409 | goto force_diskless_dec; | |
1410 | ||
1411 | if (mdev->state.conn < C_CONNECTED && | |
1412 | mdev->state.role == R_PRIMARY && | |
1413 | (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
1414 | dev_err(DEV, "Can only attach to data with current UUID=%016llX\n", | |
1415 | (unsigned long long)mdev->ed_uuid); | |
1416 | retcode = ERR_DATA_NOT_CURRENT; | |
1417 | goto force_diskless_dec; | |
1418 | } | |
1419 | ||
1420 | /* Since we are diskless, fix the activity log first... */ | |
daeda1cc | 1421 | if (drbd_check_al_size(mdev, new_disk_conf)) { |
b411b363 PR |
1422 | retcode = ERR_NOMEM; |
1423 | goto force_diskless_dec; | |
1424 | } | |
1425 | ||
1426 | /* Prevent shrinking of consistent devices ! */ | |
1427 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
daeda1cc | 1428 | drbd_new_dev_size(mdev, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
b411b363 PR |
1429 | dev_warn(DEV, "refusing to truncate a consistent device\n"); |
1430 | retcode = ERR_DISK_TO_SMALL; | |
1431 | goto force_diskless_dec; | |
1432 | } | |
1433 | ||
1434 | if (!drbd_al_read_log(mdev, nbc)) { | |
1435 | retcode = ERR_IO_MD_DISK; | |
1436 | goto force_diskless_dec; | |
1437 | } | |
1438 | ||
b411b363 PR |
1439 | /* Reset the "barriers don't work" bits here, then force meta data to |
1440 | * be written, to ensure we determine if barriers are supported. */ | |
daeda1cc | 1441 | if (new_disk_conf->no_md_flush) |
a8a4e51e | 1442 | set_bit(MD_NO_FUA, &mdev->flags); |
b411b363 | 1443 | else |
a8a4e51e | 1444 | clear_bit(MD_NO_FUA, &mdev->flags); |
b411b363 | 1445 | |
daeda1cc PR |
1446 | /* FIXME Missing stuff: rs_plan_s, clip al range */ |
1447 | ||
b411b363 PR |
1448 | /* Point of no return reached. |
1449 | * Devices and memory are no longer released by error cleanup below. | |
1450 | * now mdev takes over responsibility, and the state engine should | |
1451 | * clean it up somewhere. */ | |
1452 | D_ASSERT(mdev->ldev == NULL); | |
1453 | mdev->ldev = nbc; | |
1454 | mdev->resync = resync_lru; | |
1455 | nbc = NULL; | |
1456 | resync_lru = NULL; | |
daeda1cc | 1457 | new_disk_conf = NULL; |
b411b363 | 1458 | |
2451fc3b PR |
1459 | mdev->write_ordering = WO_bdev_flush; |
1460 | drbd_bump_write_ordering(mdev, WO_bdev_flush); | |
b411b363 PR |
1461 | |
1462 | if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY)) | |
1463 | set_bit(CRASHED_PRIMARY, &mdev->flags); | |
1464 | else | |
1465 | clear_bit(CRASHED_PRIMARY, &mdev->flags); | |
1466 | ||
894c6a94 | 1467 | if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && |
da9fbc27 | 1468 | !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod)) { |
b411b363 PR |
1469 | set_bit(CRASHED_PRIMARY, &mdev->flags); |
1470 | cp_discovered = 1; | |
1471 | } | |
1472 | ||
1473 | mdev->send_cnt = 0; | |
1474 | mdev->recv_cnt = 0; | |
1475 | mdev->read_cnt = 0; | |
1476 | mdev->writ_cnt = 0; | |
1477 | ||
99432fcc | 1478 | drbd_reconsider_max_bio_size(mdev); |
b411b363 PR |
1479 | |
1480 | /* If I am currently not R_PRIMARY, | |
1481 | * but meta data primary indicator is set, | |
1482 | * I just now recover from a hard crash, | |
1483 | * and have been R_PRIMARY before that crash. | |
1484 | * | |
1485 | * Now, if I had no connection before that crash | |
1486 | * (have been degraded R_PRIMARY), chances are that | |
1487 | * I won't find my peer now either. | |
1488 | * | |
1489 | * In that case, and _only_ in that case, | |
1490 | * we use the degr-wfc-timeout instead of the default, | |
1491 | * so we can automatically recover from a crash of a | |
1492 | * degraded but active "cluster" after a certain timeout. | |
1493 | */ | |
1494 | clear_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1495 | if (mdev->state.role != R_PRIMARY && | |
1496 | drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && | |
1497 | !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND)) | |
1498 | set_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1499 | ||
24c4830c | 1500 | dd = drbd_determine_dev_size(mdev, 0); |
b411b363 PR |
1501 | if (dd == dev_size_error) { |
1502 | retcode = ERR_NOMEM_BITMAP; | |
1503 | goto force_diskless_dec; | |
1504 | } else if (dd == grew) | |
1505 | set_bit(RESYNC_AFTER_NEG, &mdev->flags); | |
1506 | ||
1507 | if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) { | |
1508 | dev_info(DEV, "Assuming that all blocks are out of sync " | |
1509 | "(aka FullSync)\n"); | |
20ceb2b2 LE |
1510 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, |
1511 | "set_n_write from attaching", BM_LOCKED_MASK)) { | |
b411b363 PR |
1512 | retcode = ERR_IO_MD_DISK; |
1513 | goto force_diskless_dec; | |
1514 | } | |
1515 | } else { | |
20ceb2b2 | 1516 | if (drbd_bitmap_io(mdev, &drbd_bm_read, |
22ab6a30 | 1517 | "read from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1518 | retcode = ERR_IO_MD_DISK; |
1519 | goto force_diskless_dec; | |
1520 | } | |
1521 | } | |
1522 | ||
1523 | if (cp_discovered) { | |
1524 | drbd_al_apply_to_bm(mdev); | |
20ceb2b2 LE |
1525 | if (drbd_bitmap_io(mdev, &drbd_bm_write, |
1526 | "crashed primary apply AL", BM_LOCKED_MASK)) { | |
19f843aa LE |
1527 | retcode = ERR_IO_MD_DISK; |
1528 | goto force_diskless_dec; | |
1529 | } | |
b411b363 PR |
1530 | } |
1531 | ||
0778286a PR |
1532 | if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev)) |
1533 | drbd_suspend_al(mdev); /* IO is still suspended here... */ | |
1534 | ||
87eeee41 | 1535 | spin_lock_irq(&mdev->tconn->req_lock); |
78bae59b PR |
1536 | os = drbd_read_state(mdev); |
1537 | ns = os; | |
b411b363 PR |
1538 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1539 | otherwise investigate MDF_WasUpToDate... | |
1540 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1541 | otherwise into D_CONSISTENT state. | |
1542 | */ | |
1543 | if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) { | |
1544 | if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE)) | |
1545 | ns.disk = D_CONSISTENT; | |
1546 | else | |
1547 | ns.disk = D_OUTDATED; | |
1548 | } else { | |
1549 | ns.disk = D_INCONSISTENT; | |
1550 | } | |
1551 | ||
1552 | if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED)) | |
1553 | ns.pdsk = D_OUTDATED; | |
1554 | ||
daeda1cc PR |
1555 | rcu_read_lock(); |
1556 | if (ns.disk == D_CONSISTENT && | |
1557 | (ns.pdsk == D_OUTDATED || rcu_dereference(mdev->ldev->disk_conf)->fencing == FP_DONT_CARE)) | |
b411b363 | 1558 | ns.disk = D_UP_TO_DATE; |
daeda1cc | 1559 | rcu_read_unlock(); |
b411b363 PR |
1560 | |
1561 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1562 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1563 | this point, because drbd_request_state() modifies these | |
1564 | flags. */ | |
1565 | ||
1566 | /* In case we are C_CONNECTED postpone any decision on the new disk | |
1567 | state after the negotiation phase. */ | |
1568 | if (mdev->state.conn == C_CONNECTED) { | |
1569 | mdev->new_state_tmp.i = ns.i; | |
1570 | ns.i = os.i; | |
1571 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1572 | |
1573 | /* We expect to receive up-to-date UUIDs soon. | |
1574 | To avoid a race in receive_state, free p_uuid while | |
1575 | holding req_lock. I.e. atomic with the state change */ | |
1576 | kfree(mdev->p_uuid); | |
1577 | mdev->p_uuid = NULL; | |
b411b363 PR |
1578 | } |
1579 | ||
1580 | rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL); | |
87eeee41 | 1581 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
1582 | |
1583 | if (rv < SS_SUCCESS) | |
1584 | goto force_diskless_dec; | |
1585 | ||
1586 | if (mdev->state.role == R_PRIMARY) | |
1587 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
1588 | else | |
1589 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
1590 | ||
1591 | drbd_md_mark_dirty(mdev); | |
1592 | drbd_md_sync(mdev); | |
1593 | ||
1594 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
1595 | put_ldev(mdev); | |
0e29d163 | 1596 | conn_reconfig_done(mdev->tconn); |
3b98c0c2 | 1597 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1598 | return 0; |
1599 | ||
1600 | force_diskless_dec: | |
1601 | put_ldev(mdev); | |
1602 | force_diskless: | |
82f59cc6 | 1603 | drbd_force_state(mdev, NS(disk, D_FAILED)); |
b411b363 | 1604 | drbd_md_sync(mdev); |
b411b363 | 1605 | fail: |
40cbf085 | 1606 | conn_reconfig_done(mdev->tconn); |
b411b363 | 1607 | if (nbc) { |
e525fd89 TH |
1608 | if (nbc->backing_bdev) |
1609 | blkdev_put(nbc->backing_bdev, | |
1610 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1611 | if (nbc->md_bdev) | |
1612 | blkdev_put(nbc->md_bdev, | |
1613 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1614 | kfree(nbc); |
1615 | } | |
daeda1cc | 1616 | kfree(new_disk_conf); |
b411b363 PR |
1617 | lc_destroy(resync_lru); |
1618 | ||
40cbf085 | 1619 | finish: |
3b98c0c2 | 1620 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1621 | return 0; |
1622 | } | |
1623 | ||
85f75dd7 LE |
1624 | static int adm_detach(struct drbd_conf *mdev) |
1625 | { | |
19f83c76 | 1626 | enum drbd_state_rv retcode; |
009ba89d | 1627 | int ret; |
85f75dd7 | 1628 | drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */ |
009ba89d LE |
1629 | retcode = drbd_request_state(mdev, NS(disk, D_FAILED)); |
1630 | /* D_FAILED will transition to DISKLESS. */ | |
1631 | ret = wait_event_interruptible(mdev->misc_wait, | |
1632 | mdev->state.disk != D_FAILED); | |
85f75dd7 | 1633 | drbd_resume_io(mdev); |
009ba89d LE |
1634 | if ((int)retcode == (int)SS_IS_DISKLESS) |
1635 | retcode = SS_NOTHING_TO_DO; | |
1636 | if (ret) | |
1637 | retcode = ERR_INTR; | |
85f75dd7 LE |
1638 | return retcode; |
1639 | } | |
1640 | ||
82f59cc6 LE |
1641 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1642 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1643 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1644 | * internal references as well. | |
1645 | * Only then we have finally detached. */ | |
3b98c0c2 | 1646 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1647 | { |
9a0d9d03 | 1648 | enum drbd_ret_code retcode; |
3b98c0c2 LE |
1649 | |
1650 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1651 | if (!adm_ctx.reply_skb) | |
1652 | return retcode; | |
1653 | if (retcode != NO_ERROR) | |
1654 | goto out; | |
1655 | ||
85f75dd7 | 1656 | retcode = adm_detach(adm_ctx.mdev); |
3b98c0c2 LE |
1657 | out: |
1658 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1659 | return 0; |
1660 | } | |
1661 | ||
f399002e LE |
1662 | static bool conn_resync_running(struct drbd_tconn *tconn) |
1663 | { | |
1664 | struct drbd_conf *mdev; | |
695d08fa | 1665 | bool rv = false; |
f399002e LE |
1666 | int vnr; |
1667 | ||
695d08fa | 1668 | rcu_read_lock(); |
f399002e LE |
1669 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1670 | if (mdev->state.conn == C_SYNC_SOURCE || | |
1671 | mdev->state.conn == C_SYNC_TARGET || | |
1672 | mdev->state.conn == C_PAUSED_SYNC_S || | |
695d08fa PR |
1673 | mdev->state.conn == C_PAUSED_SYNC_T) { |
1674 | rv = true; | |
1675 | break; | |
1676 | } | |
f399002e | 1677 | } |
695d08fa PR |
1678 | rcu_read_unlock(); |
1679 | ||
1680 | return rv; | |
f399002e LE |
1681 | } |
1682 | ||
1683 | static bool conn_ov_running(struct drbd_tconn *tconn) | |
1684 | { | |
1685 | struct drbd_conf *mdev; | |
695d08fa | 1686 | bool rv = false; |
f399002e LE |
1687 | int vnr; |
1688 | ||
695d08fa | 1689 | rcu_read_lock(); |
f399002e LE |
1690 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1691 | if (mdev->state.conn == C_VERIFY_S || | |
695d08fa PR |
1692 | mdev->state.conn == C_VERIFY_T) { |
1693 | rv = true; | |
1694 | break; | |
1695 | } | |
f399002e | 1696 | } |
695d08fa PR |
1697 | rcu_read_unlock(); |
1698 | ||
1699 | return rv; | |
f399002e LE |
1700 | } |
1701 | ||
cd64397c | 1702 | static enum drbd_ret_code |
44ed167d | 1703 | _check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf) |
cd64397c PR |
1704 | { |
1705 | struct drbd_conf *mdev; | |
1706 | int i; | |
1707 | ||
44ed167d | 1708 | if (old_conf && tconn->agreed_pro_version < 100 && |
b032b6fa | 1709 | tconn->cstate == C_WF_REPORT_PARAMS && |
44ed167d | 1710 | new_conf->wire_protocol != old_conf->wire_protocol) |
b032b6fa PR |
1711 | return ERR_NEED_APV_100; |
1712 | ||
cd64397c PR |
1713 | if (new_conf->two_primaries && |
1714 | (new_conf->wire_protocol != DRBD_PROT_C)) | |
1715 | return ERR_NOT_PROTO_C; | |
1716 | ||
cd64397c PR |
1717 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
1718 | if (get_ldev(mdev)) { | |
daeda1cc | 1719 | enum drbd_fencing_p fp = rcu_dereference(mdev->ldev->disk_conf)->fencing; |
cd64397c | 1720 | put_ldev(mdev); |
44ed167d | 1721 | if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1722 | return ERR_STONITH_AND_PROT_A; |
cd64397c | 1723 | } |
44ed167d | 1724 | if (mdev->state.role == R_PRIMARY && new_conf->want_lose) |
cd64397c | 1725 | return ERR_DISCARD; |
cd64397c | 1726 | } |
cd64397c PR |
1727 | |
1728 | if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A) | |
1729 | return ERR_CONG_NOT_PROTO_A; | |
1730 | ||
1731 | return NO_ERROR; | |
1732 | } | |
1733 | ||
44ed167d PR |
1734 | static enum drbd_ret_code |
1735 | check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf) | |
1736 | { | |
1737 | static enum drbd_ret_code rv; | |
1738 | struct drbd_conf *mdev; | |
1739 | int i; | |
1740 | ||
1741 | rcu_read_lock(); | |
1742 | rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf); | |
1743 | rcu_read_unlock(); | |
1744 | ||
1745 | /* tconn->volumes protected by genl_lock() here */ | |
1746 | idr_for_each_entry(&tconn->volumes, mdev, i) { | |
1747 | if (!mdev->bitmap) { | |
1748 | if(drbd_bm_init(mdev)) | |
1749 | return ERR_NOMEM; | |
1750 | } | |
1751 | } | |
1752 | ||
1753 | return rv; | |
1754 | } | |
1755 | ||
0fd0ea06 PR |
1756 | struct crypto { |
1757 | struct crypto_hash *verify_tfm; | |
1758 | struct crypto_hash *csums_tfm; | |
1759 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 1760 | struct crypto_hash *integrity_tfm; |
0fd0ea06 PR |
1761 | void *int_dig_in; |
1762 | void *int_dig_vv; | |
1763 | }; | |
1764 | ||
1765 | static int | |
4b6ad6d4 | 1766 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
1767 | { |
1768 | if (!tfm_name[0]) | |
1769 | return NO_ERROR; | |
1770 | ||
1771 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); | |
1772 | if (IS_ERR(*tfm)) { | |
1773 | *tfm = NULL; | |
1774 | return err_alg; | |
1775 | } | |
1776 | ||
0fd0ea06 PR |
1777 | return NO_ERROR; |
1778 | } | |
1779 | ||
1780 | static enum drbd_ret_code | |
1781 | alloc_crypto(struct crypto *crypto, struct net_conf *new_conf) | |
1782 | { | |
1783 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
1784 | enum drbd_ret_code rv; | |
1785 | int hash_size; | |
1786 | ||
4b6ad6d4 AG |
1787 | rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg, |
1788 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
1789 | if (rv != NO_ERROR) |
1790 | return rv; | |
4b6ad6d4 AG |
1791 | rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg, |
1792 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
1793 | if (rv != NO_ERROR) |
1794 | return rv; | |
4b6ad6d4 AG |
1795 | rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg, |
1796 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
1797 | if (rv != NO_ERROR) |
1798 | return rv; | |
0fd0ea06 PR |
1799 | if (new_conf->cram_hmac_alg[0] != 0) { |
1800 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", | |
1801 | new_conf->cram_hmac_alg); | |
1802 | ||
4b6ad6d4 AG |
1803 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
1804 | ERR_AUTH_ALG); | |
0fd0ea06 | 1805 | } |
8d412fc6 AG |
1806 | if (crypto->integrity_tfm) { |
1807 | hash_size = crypto_hash_digestsize(crypto->integrity_tfm); | |
0fd0ea06 PR |
1808 | crypto->int_dig_in = kmalloc(hash_size, GFP_KERNEL); |
1809 | if (!crypto->int_dig_in) | |
1810 | return ERR_NOMEM; | |
1811 | crypto->int_dig_vv = kmalloc(hash_size, GFP_KERNEL); | |
1812 | if (!crypto->int_dig_vv) | |
1813 | return ERR_NOMEM; | |
1814 | } | |
1815 | ||
1816 | return rv; | |
1817 | } | |
1818 | ||
1819 | static void free_crypto(struct crypto *crypto) | |
1820 | { | |
1821 | kfree(crypto->int_dig_in); | |
1822 | kfree(crypto->int_dig_vv); | |
1823 | crypto_free_hash(crypto->cram_hmac_tfm); | |
8d412fc6 | 1824 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
1825 | crypto_free_hash(crypto->csums_tfm); |
1826 | crypto_free_hash(crypto->verify_tfm); | |
1827 | } | |
1828 | ||
f399002e LE |
1829 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
1830 | { | |
1831 | enum drbd_ret_code retcode; | |
1832 | struct drbd_tconn *tconn; | |
44ed167d | 1833 | struct net_conf *old_conf, *new_conf = NULL; |
f399002e LE |
1834 | int err; |
1835 | int ovr; /* online verify running */ | |
1836 | int rsr; /* re-sync running */ | |
0fd0ea06 | 1837 | struct crypto crypto = { }; |
88104ca4 | 1838 | bool change_integrity_alg; |
f399002e LE |
1839 | |
1840 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); | |
1841 | if (!adm_ctx.reply_skb) | |
1842 | return retcode; | |
1843 | if (retcode != NO_ERROR) | |
1844 | goto out; | |
1845 | ||
1846 | tconn = adm_ctx.tconn; | |
1847 | ||
1848 | new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); | |
1849 | if (!new_conf) { | |
1850 | retcode = ERR_NOMEM; | |
1851 | goto out; | |
1852 | } | |
1853 | ||
44ed167d PR |
1854 | conn_reconfig_start(tconn); |
1855 | ||
88104ca4 | 1856 | mutex_lock(&tconn->data.mutex); |
a0095508 | 1857 | mutex_lock(&tconn->conf_update); |
91fd4dad | 1858 | old_conf = tconn->net_conf; |
44ed167d PR |
1859 | |
1860 | if (!old_conf) { | |
f399002e LE |
1861 | drbd_msg_put_info("net conf missing, try connect"); |
1862 | retcode = ERR_INVALID_REQUEST; | |
91fd4dad | 1863 | goto fail; |
f399002e LE |
1864 | } |
1865 | ||
44ed167d | 1866 | *new_conf = *old_conf; |
5979e361 | 1867 | if (should_set_defaults(info)) |
b966b5dd | 1868 | set_net_conf_defaults(new_conf); |
f399002e | 1869 | |
f399002e LE |
1870 | err = net_conf_from_attrs_for_change(new_conf, info); |
1871 | if (err) { | |
1872 | retcode = ERR_MANDATORY_TAG; | |
1873 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1874 | goto fail; | |
1875 | } | |
1876 | ||
cd64397c PR |
1877 | retcode = check_net_options(tconn, new_conf); |
1878 | if (retcode != NO_ERROR) | |
1879 | goto fail; | |
1880 | ||
f399002e LE |
1881 | /* re-sync running */ |
1882 | rsr = conn_resync_running(tconn); | |
0fd0ea06 | 1883 | if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) { |
f399002e | 1884 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 1885 | goto fail; |
f399002e LE |
1886 | } |
1887 | ||
f399002e LE |
1888 | /* online verify running */ |
1889 | ovr = conn_ov_running(tconn); | |
0fd0ea06 PR |
1890 | if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) { |
1891 | retcode = ERR_VERIFY_RUNNING; | |
1892 | goto fail; | |
f399002e LE |
1893 | } |
1894 | ||
88104ca4 AG |
1895 | change_integrity_alg = strcmp(old_conf->integrity_alg, |
1896 | new_conf->integrity_alg); | |
1897 | ||
0fd0ea06 PR |
1898 | retcode = alloc_crypto(&crypto, new_conf); |
1899 | if (retcode != NO_ERROR) | |
1900 | goto fail; | |
f399002e | 1901 | |
44ed167d | 1902 | rcu_assign_pointer(tconn->net_conf, new_conf); |
f399002e LE |
1903 | |
1904 | if (!rsr) { | |
1905 | crypto_free_hash(tconn->csums_tfm); | |
0fd0ea06 PR |
1906 | tconn->csums_tfm = crypto.csums_tfm; |
1907 | crypto.csums_tfm = NULL; | |
f399002e LE |
1908 | } |
1909 | if (!ovr) { | |
1910 | crypto_free_hash(tconn->verify_tfm); | |
0fd0ea06 PR |
1911 | tconn->verify_tfm = crypto.verify_tfm; |
1912 | crypto.verify_tfm = NULL; | |
f399002e LE |
1913 | } |
1914 | ||
0fd0ea06 PR |
1915 | kfree(tconn->int_dig_in); |
1916 | tconn->int_dig_in = crypto.int_dig_in; | |
1917 | kfree(tconn->int_dig_vv); | |
1918 | tconn->int_dig_vv = crypto.int_dig_vv; | |
8d412fc6 AG |
1919 | crypto_free_hash(tconn->integrity_tfm); |
1920 | tconn->integrity_tfm = crypto.integrity_tfm; | |
88104ca4 AG |
1921 | if (change_integrity_alg) { |
1922 | /* Do this without trying to take tconn->data.mutex again. */ | |
1923 | if (__drbd_send_protocol(tconn)) | |
1924 | goto fail; | |
1925 | } | |
0fd0ea06 PR |
1926 | |
1927 | /* FIXME Changing cram_hmac while the connection is established is useless */ | |
1928 | crypto_free_hash(tconn->cram_hmac_tfm); | |
1929 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
1930 | ||
a0095508 | 1931 | mutex_unlock(&tconn->conf_update); |
88104ca4 | 1932 | mutex_unlock(&tconn->data.mutex); |
91fd4dad PR |
1933 | synchronize_rcu(); |
1934 | kfree(old_conf); | |
1935 | ||
f399002e LE |
1936 | if (tconn->cstate >= C_WF_REPORT_PARAMS) |
1937 | drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn))); | |
1938 | ||
91fd4dad PR |
1939 | goto done; |
1940 | ||
f399002e | 1941 | fail: |
a0095508 | 1942 | mutex_unlock(&tconn->conf_update); |
88104ca4 | 1943 | mutex_unlock(&tconn->data.mutex); |
0fd0ea06 | 1944 | free_crypto(&crypto); |
f399002e | 1945 | kfree(new_conf); |
91fd4dad | 1946 | done: |
f399002e LE |
1947 | conn_reconfig_done(tconn); |
1948 | out: | |
1949 | drbd_adm_finish(info, retcode); | |
1950 | return 0; | |
1951 | } | |
1952 | ||
3b98c0c2 | 1953 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1954 | { |
3b98c0c2 | 1955 | struct drbd_conf *mdev; |
44ed167d | 1956 | struct net_conf *old_conf, *new_conf = NULL; |
0fd0ea06 | 1957 | struct crypto crypto = { }; |
80883197 | 1958 | struct drbd_tconn *oconn; |
3b98c0c2 | 1959 | struct drbd_tconn *tconn; |
b411b363 | 1960 | struct sockaddr *new_my_addr, *new_peer_addr, *taken_addr; |
3b98c0c2 LE |
1961 | enum drbd_ret_code retcode; |
1962 | int i; | |
1963 | int err; | |
b411b363 | 1964 | |
3b98c0c2 LE |
1965 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
1966 | if (!adm_ctx.reply_skb) | |
1967 | return retcode; | |
1968 | if (retcode != NO_ERROR) | |
1969 | goto out; | |
1970 | ||
1971 | tconn = adm_ctx.tconn; | |
80883197 | 1972 | conn_reconfig_start(tconn); |
b411b363 | 1973 | |
80883197 | 1974 | if (tconn->cstate > C_STANDALONE) { |
b411b363 PR |
1975 | retcode = ERR_NET_CONFIGURED; |
1976 | goto fail; | |
1977 | } | |
1978 | ||
1979 | /* allocation not in the IO path, cqueue thread context */ | |
5979e361 | 1980 | new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL); |
b411b363 PR |
1981 | if (!new_conf) { |
1982 | retcode = ERR_NOMEM; | |
1983 | goto fail; | |
1984 | } | |
1985 | ||
b966b5dd | 1986 | set_net_conf_defaults(new_conf); |
f399002e LE |
1987 | |
1988 | err = net_conf_from_attrs(new_conf, info); | |
3b98c0c2 | 1989 | if (err) { |
b411b363 | 1990 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1991 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1992 | goto fail; |
1993 | } | |
1994 | ||
cd64397c PR |
1995 | retcode = check_net_options(tconn, new_conf); |
1996 | if (retcode != NO_ERROR) | |
422028b1 | 1997 | goto fail; |
422028b1 | 1998 | |
b411b363 PR |
1999 | retcode = NO_ERROR; |
2000 | ||
2001 | new_my_addr = (struct sockaddr *)&new_conf->my_addr; | |
2002 | new_peer_addr = (struct sockaddr *)&new_conf->peer_addr; | |
543cc10b | 2003 | |
ef356262 | 2004 | /* No need to take drbd_cfg_rwsem here. All reconfiguration is |
543cc10b LE |
2005 | * strictly serialized on genl_lock(). We are protected against |
2006 | * concurrent reconfiguration/addition/deletion */ | |
80883197 | 2007 | list_for_each_entry(oconn, &drbd_tconns, all_tconn) { |
44ed167d | 2008 | struct net_conf *nc; |
80883197 | 2009 | if (oconn == tconn) |
b411b363 | 2010 | continue; |
44ed167d PR |
2011 | |
2012 | rcu_read_lock(); | |
2013 | nc = rcu_dereference(oconn->net_conf); | |
2014 | if (nc) { | |
2015 | taken_addr = (struct sockaddr *)&nc->my_addr; | |
2016 | if (new_conf->my_addr_len == nc->my_addr_len && | |
b411b363 PR |
2017 | !memcmp(new_my_addr, taken_addr, new_conf->my_addr_len)) |
2018 | retcode = ERR_LOCAL_ADDR; | |
2019 | ||
44ed167d PR |
2020 | taken_addr = (struct sockaddr *)&nc->peer_addr; |
2021 | if (new_conf->peer_addr_len == nc->peer_addr_len && | |
b411b363 PR |
2022 | !memcmp(new_peer_addr, taken_addr, new_conf->peer_addr_len)) |
2023 | retcode = ERR_PEER_ADDR; | |
b411b363 | 2024 | } |
44ed167d PR |
2025 | rcu_read_unlock(); |
2026 | if (retcode != NO_ERROR) | |
2027 | goto fail; | |
b411b363 PR |
2028 | } |
2029 | ||
0fd0ea06 PR |
2030 | retcode = alloc_crypto(&crypto, new_conf); |
2031 | if (retcode != NO_ERROR) | |
2032 | goto fail; | |
b411b363 | 2033 | |
b411b363 PR |
2034 | ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
2035 | ||
80883197 | 2036 | conn_flush_workqueue(tconn); |
91fd4dad | 2037 | |
a0095508 | 2038 | mutex_lock(&tconn->conf_update); |
91fd4dad PR |
2039 | old_conf = tconn->net_conf; |
2040 | if (old_conf) { | |
b411b363 | 2041 | retcode = ERR_NET_CONFIGURED; |
a0095508 | 2042 | mutex_unlock(&tconn->conf_update); |
b411b363 PR |
2043 | goto fail; |
2044 | } | |
44ed167d | 2045 | rcu_assign_pointer(tconn->net_conf, new_conf); |
b411b363 | 2046 | |
91fd4dad | 2047 | conn_free_crypto(tconn); |
0fd0ea06 PR |
2048 | tconn->int_dig_in = crypto.int_dig_in; |
2049 | tconn->int_dig_vv = crypto.int_dig_vv; | |
2050 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
8d412fc6 | 2051 | tconn->integrity_tfm = crypto.integrity_tfm; |
0fd0ea06 PR |
2052 | tconn->csums_tfm = crypto.csums_tfm; |
2053 | tconn->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2054 | |
a0095508 | 2055 | mutex_unlock(&tconn->conf_update); |
91fd4dad | 2056 | |
695d08fa | 2057 | rcu_read_lock(); |
80883197 PR |
2058 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
2059 | mdev->send_cnt = 0; | |
2060 | mdev->recv_cnt = 0; | |
80883197 | 2061 | } |
695d08fa | 2062 | rcu_read_unlock(); |
5ee743e9 LE |
2063 | |
2064 | retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE); | |
2065 | ||
80883197 | 2066 | conn_reconfig_done(tconn); |
3b98c0c2 | 2067 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2068 | return 0; |
2069 | ||
2070 | fail: | |
0fd0ea06 | 2071 | free_crypto(&crypto); |
b411b363 PR |
2072 | kfree(new_conf); |
2073 | ||
80883197 | 2074 | conn_reconfig_done(tconn); |
3b98c0c2 LE |
2075 | out: |
2076 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2077 | return 0; |
2078 | } | |
2079 | ||
85f75dd7 LE |
2080 | static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force) |
2081 | { | |
2082 | enum drbd_state_rv rv; | |
85f75dd7 | 2083 | |
f3dfa40a LE |
2084 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2085 | force ? CS_HARD : 0); | |
85f75dd7 LE |
2086 | |
2087 | switch (rv) { | |
2088 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2089 | break; |
85f75dd7 LE |
2090 | case SS_ALREADY_STANDALONE: |
2091 | return SS_SUCCESS; | |
2092 | case SS_PRIMARY_NOP: | |
2093 | /* Our state checking code wants to see the peer outdated. */ | |
2094 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
f3dfa40a | 2095 | pdsk, D_OUTDATED), CS_VERBOSE); |
85f75dd7 LE |
2096 | break; |
2097 | case SS_CW_FAILED_BY_PEER: | |
2098 | /* The peer probably wants to see us outdated. */ | |
2099 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
2100 | disk, D_OUTDATED), 0); | |
2101 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
f3dfa40a LE |
2102 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2103 | CS_HARD); | |
85f75dd7 LE |
2104 | } |
2105 | break; | |
2106 | default:; | |
2107 | /* no special handling necessary */ | |
2108 | } | |
2109 | ||
f3dfa40a LE |
2110 | if (rv >= SS_SUCCESS) { |
2111 | enum drbd_state_rv rv2; | |
2112 | /* No one else can reconfigure the network while I am here. | |
2113 | * The state handling only uses drbd_thread_stop_nowait(), | |
2114 | * we want to really wait here until the receiver is no more. | |
2115 | */ | |
2116 | drbd_thread_stop(&adm_ctx.tconn->receiver); | |
2117 | ||
2118 | /* Race breaker. This additional state change request may be | |
2119 | * necessary, if this was a forced disconnect during a receiver | |
2120 | * restart. We may have "killed" the receiver thread just | |
2121 | * after drbdd_init() returned. Typically, we should be | |
2122 | * C_STANDALONE already, now, and this becomes a no-op. | |
2123 | */ | |
2124 | rv2 = conn_request_state(tconn, NS(conn, C_STANDALONE), | |
2125 | CS_VERBOSE | CS_HARD); | |
2126 | if (rv2 < SS_SUCCESS) | |
2127 | conn_err(tconn, | |
2128 | "unexpected rv2=%d in conn_try_disconnect()\n", | |
2129 | rv2); | |
2130 | } | |
85f75dd7 LE |
2131 | return rv; |
2132 | } | |
2133 | ||
3b98c0c2 | 2134 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2135 | { |
3b98c0c2 LE |
2136 | struct disconnect_parms parms; |
2137 | struct drbd_tconn *tconn; | |
85f75dd7 | 2138 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2139 | enum drbd_ret_code retcode; |
2140 | int err; | |
2561b9c1 | 2141 | |
3b98c0c2 LE |
2142 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
2143 | if (!adm_ctx.reply_skb) | |
2144 | return retcode; | |
2145 | if (retcode != NO_ERROR) | |
2561b9c1 | 2146 | goto fail; |
3b98c0c2 LE |
2147 | |
2148 | tconn = adm_ctx.tconn; | |
2149 | memset(&parms, 0, sizeof(parms)); | |
2150 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2151 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2152 | if (err) { |
2153 | retcode = ERR_MANDATORY_TAG; | |
2154 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2155 | goto fail; | |
2156 | } | |
2561b9c1 PR |
2157 | } |
2158 | ||
85f75dd7 LE |
2159 | rv = conn_try_disconnect(tconn, parms.force_disconnect); |
2160 | if (rv < SS_SUCCESS) | |
f3dfa40a LE |
2161 | retcode = rv; /* FIXME: Type mismatch. */ |
2162 | else | |
2163 | retcode = NO_ERROR; | |
b411b363 | 2164 | fail: |
3b98c0c2 | 2165 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2166 | return 0; |
2167 | } | |
2168 | ||
2169 | void resync_after_online_grow(struct drbd_conf *mdev) | |
2170 | { | |
2171 | int iass; /* I am sync source */ | |
2172 | ||
2173 | dev_info(DEV, "Resync of new storage after online grow\n"); | |
2174 | if (mdev->state.role != mdev->state.peer) | |
2175 | iass = (mdev->state.role == R_PRIMARY); | |
2176 | else | |
25703f83 | 2177 | iass = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags); |
b411b363 PR |
2178 | |
2179 | if (iass) | |
2180 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
2181 | else | |
2182 | _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); | |
2183 | } | |
2184 | ||
3b98c0c2 | 2185 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2186 | { |
daeda1cc | 2187 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 LE |
2188 | struct resize_parms rs; |
2189 | struct drbd_conf *mdev; | |
2190 | enum drbd_ret_code retcode; | |
b411b363 | 2191 | enum determine_dev_size dd; |
6495d2c6 | 2192 | enum dds_flags ddsf; |
daeda1cc | 2193 | sector_t u_size; |
3b98c0c2 | 2194 | int err; |
b411b363 | 2195 | |
3b98c0c2 LE |
2196 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2197 | if (!adm_ctx.reply_skb) | |
2198 | return retcode; | |
2199 | if (retcode != NO_ERROR) | |
b411b363 | 2200 | goto fail; |
3b98c0c2 LE |
2201 | |
2202 | memset(&rs, 0, sizeof(struct resize_parms)); | |
2203 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { | |
f399002e | 2204 | err = resize_parms_from_attrs(&rs, info); |
3b98c0c2 LE |
2205 | if (err) { |
2206 | retcode = ERR_MANDATORY_TAG; | |
2207 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2208 | goto fail; | |
2209 | } | |
b411b363 PR |
2210 | } |
2211 | ||
3b98c0c2 | 2212 | mdev = adm_ctx.mdev; |
b411b363 PR |
2213 | if (mdev->state.conn > C_CONNECTED) { |
2214 | retcode = ERR_RESIZE_RESYNC; | |
2215 | goto fail; | |
2216 | } | |
2217 | ||
2218 | if (mdev->state.role == R_SECONDARY && | |
2219 | mdev->state.peer == R_SECONDARY) { | |
2220 | retcode = ERR_NO_PRIMARY; | |
2221 | goto fail; | |
2222 | } | |
2223 | ||
2224 | if (!get_ldev(mdev)) { | |
2225 | retcode = ERR_NO_DISK; | |
2226 | goto fail; | |
2227 | } | |
2228 | ||
31890f4a | 2229 | if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) { |
6495d2c6 PR |
2230 | retcode = ERR_NEED_APV_93; |
2231 | goto fail; | |
2232 | } | |
2233 | ||
daeda1cc PR |
2234 | rcu_read_lock(); |
2235 | u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size; | |
2236 | rcu_read_unlock(); | |
2237 | if (u_size != (sector_t)rs.resize_size) { | |
2238 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2239 | if (!new_disk_conf) { | |
2240 | retcode = ERR_NOMEM; | |
2241 | goto fail; | |
2242 | } | |
2243 | } | |
2244 | ||
087c2492 | 2245 | if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) |
b411b363 | 2246 | mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); |
b411b363 | 2247 | |
daeda1cc PR |
2248 | if (new_disk_conf) { |
2249 | mutex_lock(&mdev->tconn->conf_update); | |
2250 | old_disk_conf = mdev->ldev->disk_conf; | |
2251 | *new_disk_conf = *old_disk_conf; | |
2252 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
2253 | rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf); | |
2254 | mutex_unlock(&mdev->tconn->conf_update); | |
2255 | synchronize_rcu(); | |
2256 | kfree(old_disk_conf); | |
2257 | } | |
2258 | ||
6495d2c6 | 2259 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
24c4830c | 2260 | dd = drbd_determine_dev_size(mdev, ddsf); |
b411b363 PR |
2261 | drbd_md_sync(mdev); |
2262 | put_ldev(mdev); | |
2263 | if (dd == dev_size_error) { | |
2264 | retcode = ERR_NOMEM_BITMAP; | |
2265 | goto fail; | |
2266 | } | |
2267 | ||
087c2492 | 2268 | if (mdev->state.conn == C_CONNECTED) { |
b411b363 PR |
2269 | if (dd == grew) |
2270 | set_bit(RESIZE_PENDING, &mdev->flags); | |
2271 | ||
2272 | drbd_send_uuids(mdev); | |
6495d2c6 | 2273 | drbd_send_sizes(mdev, 1, ddsf); |
b411b363 PR |
2274 | } |
2275 | ||
2276 | fail: | |
3b98c0c2 | 2277 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2278 | return 0; |
2279 | } | |
2280 | ||
b966b5dd AG |
2281 | void drbd_set_res_opts_defaults(struct res_opts *r) |
2282 | { | |
2283 | return set_res_opts_defaults(r); | |
2284 | } | |
2285 | ||
f399002e | 2286 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2287 | { |
3b98c0c2 | 2288 | enum drbd_ret_code retcode; |
b411b363 | 2289 | cpumask_var_t new_cpu_mask; |
f399002e | 2290 | struct drbd_tconn *tconn; |
b57a1e27 | 2291 | struct res_opts res_opts; |
f399002e | 2292 | int err; |
b411b363 | 2293 | |
f399002e | 2294 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
3b98c0c2 LE |
2295 | if (!adm_ctx.reply_skb) |
2296 | return retcode; | |
2297 | if (retcode != NO_ERROR) | |
2298 | goto fail; | |
f399002e | 2299 | tconn = adm_ctx.tconn; |
3b98c0c2 | 2300 | |
b411b363 PR |
2301 | if (!zalloc_cpumask_var(&new_cpu_mask, GFP_KERNEL)) { |
2302 | retcode = ERR_NOMEM; | |
3b98c0c2 | 2303 | drbd_msg_put_info("unable to allocate cpumask"); |
b411b363 PR |
2304 | goto fail; |
2305 | } | |
2306 | ||
b57a1e27 | 2307 | res_opts = tconn->res_opts; |
5979e361 | 2308 | if (should_set_defaults(info)) |
b966b5dd | 2309 | set_res_opts_defaults(&res_opts); |
b411b363 | 2310 | |
b57a1e27 | 2311 | err = res_opts_from_attrs(&res_opts, info); |
3b98c0c2 | 2312 | if (err) { |
b411b363 | 2313 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2314 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2315 | goto fail; |
2316 | } | |
2317 | ||
b411b363 | 2318 | /* silently ignore cpu mask on UP kernel */ |
b57a1e27 LE |
2319 | if (nr_cpu_ids > 1 && res_opts.cpu_mask[0] != 0) { |
2320 | err = __bitmap_parse(res_opts.cpu_mask, 32, 0, | |
b411b363 PR |
2321 | cpumask_bits(new_cpu_mask), nr_cpu_ids); |
2322 | if (err) { | |
f399002e | 2323 | conn_warn(tconn, "__bitmap_parse() failed with %d\n", err); |
b411b363 PR |
2324 | retcode = ERR_CPU_MASK_PARSE; |
2325 | goto fail; | |
2326 | } | |
2327 | } | |
2328 | ||
b411b363 | 2329 | |
b57a1e27 | 2330 | tconn->res_opts = res_opts; |
b411b363 | 2331 | |
f399002e LE |
2332 | if (!cpumask_equal(tconn->cpu_mask, new_cpu_mask)) { |
2333 | cpumask_copy(tconn->cpu_mask, new_cpu_mask); | |
2334 | drbd_calc_cpu_mask(tconn); | |
2335 | tconn->receiver.reset_cpu_mask = 1; | |
2336 | tconn->asender.reset_cpu_mask = 1; | |
2337 | tconn->worker.reset_cpu_mask = 1; | |
b411b363 PR |
2338 | } |
2339 | ||
b411b363 PR |
2340 | fail: |
2341 | free_cpumask_var(new_cpu_mask); | |
3b98c0c2 LE |
2342 | |
2343 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2344 | return 0; |
2345 | } | |
2346 | ||
3b98c0c2 | 2347 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2348 | { |
3b98c0c2 LE |
2349 | struct drbd_conf *mdev; |
2350 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2351 | ||
2352 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2353 | if (!adm_ctx.reply_skb) | |
2354 | return retcode; | |
2355 | if (retcode != NO_ERROR) | |
2356 | goto out; | |
2357 | ||
2358 | mdev = adm_ctx.mdev; | |
b411b363 | 2359 | |
194bfb32 LE |
2360 | /* If there is still bitmap IO pending, probably because of a previous |
2361 | * resync just being finished, wait for it before requesting a new resync. */ | |
2362 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
2363 | ||
b411b363 PR |
2364 | retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED); |
2365 | ||
2366 | if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION) | |
2367 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
2368 | ||
2369 | while (retcode == SS_NEED_CONNECTION) { | |
87eeee41 | 2370 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2371 | if (mdev->state.conn < C_CONNECTED) |
2372 | retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL); | |
87eeee41 | 2373 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2374 | |
2375 | if (retcode != SS_NEED_CONNECTION) | |
2376 | break; | |
2377 | ||
2378 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
2379 | } | |
2380 | ||
3b98c0c2 LE |
2381 | out: |
2382 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2383 | return 0; |
2384 | } | |
2385 | ||
0778286a PR |
2386 | static int drbd_bmio_set_susp_al(struct drbd_conf *mdev) |
2387 | { | |
2388 | int rv; | |
2389 | ||
2390 | rv = drbd_bmio_set_n_write(mdev); | |
2391 | drbd_suspend_al(mdev); | |
2392 | return rv; | |
2393 | } | |
2394 | ||
3b98c0c2 LE |
2395 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2396 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2397 | { |
3b98c0c2 | 2398 | enum drbd_ret_code retcode; |
194bfb32 | 2399 | |
3b98c0c2 LE |
2400 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2401 | if (!adm_ctx.reply_skb) | |
2402 | return retcode; | |
2403 | if (retcode != NO_ERROR) | |
2404 | goto out; | |
b411b363 | 2405 | |
3b98c0c2 LE |
2406 | retcode = drbd_request_state(adm_ctx.mdev, mask, val); |
2407 | out: | |
2408 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2409 | return 0; |
2410 | } | |
2411 | ||
3b98c0c2 | 2412 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2413 | { |
3b98c0c2 LE |
2414 | return drbd_adm_simple_request_state(skb, info, NS(conn, C_STARTING_SYNC_S)); |
2415 | } | |
b411b363 | 2416 | |
3b98c0c2 LE |
2417 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
2418 | { | |
2419 | enum drbd_ret_code retcode; | |
2420 | ||
2421 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2422 | if (!adm_ctx.reply_skb) | |
2423 | return retcode; | |
2424 | if (retcode != NO_ERROR) | |
2425 | goto out; | |
b411b363 | 2426 | |
3b98c0c2 LE |
2427 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
2428 | retcode = ERR_PAUSE_IS_SET; | |
2429 | out: | |
2430 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2431 | return 0; |
2432 | } | |
2433 | ||
3b98c0c2 | 2434 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2435 | { |
da9fbc27 | 2436 | union drbd_dev_state s; |
3b98c0c2 LE |
2437 | enum drbd_ret_code retcode; |
2438 | ||
2439 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2440 | if (!adm_ctx.reply_skb) | |
2441 | return retcode; | |
2442 | if (retcode != NO_ERROR) | |
2443 | goto out; | |
b411b363 | 2444 | |
3b98c0c2 LE |
2445 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2446 | s = adm_ctx.mdev->state; | |
cd88d030 PR |
2447 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2448 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2449 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2450 | } else { | |
2451 | retcode = ERR_PAUSE_IS_CLEAR; | |
2452 | } | |
2453 | } | |
b411b363 | 2454 | |
3b98c0c2 LE |
2455 | out: |
2456 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2457 | return 0; |
2458 | } | |
2459 | ||
3b98c0c2 | 2460 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2461 | { |
3b98c0c2 | 2462 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2463 | } |
2464 | ||
3b98c0c2 | 2465 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2466 | { |
3b98c0c2 LE |
2467 | struct drbd_conf *mdev; |
2468 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2469 | ||
2470 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2471 | if (!adm_ctx.reply_skb) | |
2472 | return retcode; | |
2473 | if (retcode != NO_ERROR) | |
2474 | goto out; | |
2475 | ||
2476 | mdev = adm_ctx.mdev; | |
43a5182c PR |
2477 | if (test_bit(NEW_CUR_UUID, &mdev->flags)) { |
2478 | drbd_uuid_new_current(mdev); | |
2479 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
43a5182c | 2480 | } |
265be2d0 | 2481 | drbd_suspend_io(mdev); |
3b98c0c2 LE |
2482 | retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); |
2483 | if (retcode == SS_SUCCESS) { | |
265be2d0 | 2484 | if (mdev->state.conn < C_CONNECTED) |
2f5cdd0b | 2485 | tl_clear(mdev->tconn); |
265be2d0 | 2486 | if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED) |
2f5cdd0b | 2487 | tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO); |
265be2d0 PR |
2488 | } |
2489 | drbd_resume_io(mdev); | |
2490 | ||
3b98c0c2 LE |
2491 | out: |
2492 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2493 | return 0; |
2494 | } | |
2495 | ||
3b98c0c2 | 2496 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2497 | { |
3b98c0c2 | 2498 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2499 | } |
2500 | ||
543cc10b LE |
2501 | int nla_put_drbd_cfg_context(struct sk_buff *skb, const char *conn_name, unsigned vnr) |
2502 | { | |
2503 | struct nlattr *nla; | |
2504 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2505 | if (!nla) | |
2506 | goto nla_put_failure; | |
2507 | if (vnr != VOLUME_UNSPECIFIED) | |
2508 | NLA_PUT_U32(skb, T_ctx_volume, vnr); | |
2509 | NLA_PUT_STRING(skb, T_ctx_conn_name, conn_name); | |
2510 | nla_nest_end(skb, nla); | |
2511 | return 0; | |
2512 | ||
2513 | nla_put_failure: | |
2514 | if (nla) | |
2515 | nla_nest_cancel(skb, nla); | |
2516 | return -EMSGSIZE; | |
2517 | } | |
2518 | ||
3b98c0c2 LE |
2519 | int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev, |
2520 | const struct sib_info *sib) | |
b411b363 | 2521 | { |
3b98c0c2 | 2522 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
44ed167d | 2523 | struct net_conf *nc; |
3b98c0c2 LE |
2524 | struct nlattr *nla; |
2525 | int got_ldev; | |
3b98c0c2 LE |
2526 | int err = 0; |
2527 | int exclude_sensitive; | |
2528 | ||
2529 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2530 | * to. So we better exclude_sensitive information. | |
2531 | * | |
2532 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2533 | * in the context of the requesting user process. Exclude sensitive | |
2534 | * information, unless current has superuser. | |
2535 | * | |
2536 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2537 | * relies on the current implementation of netlink_dump(), which | |
2538 | * executes the dump callback successively from netlink_recvmsg(), | |
2539 | * always in the context of the receiving process */ | |
2540 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2541 | ||
2542 | got_ldev = get_ldev(mdev); | |
3b98c0c2 LE |
2543 | |
2544 | /* We need to add connection name and volume number information still. | |
2545 | * Minor number is in drbd_genlmsghdr. */ | |
543cc10b | 2546 | if (nla_put_drbd_cfg_context(skb, mdev->tconn->name, mdev->vnr)) |
3b98c0c2 | 2547 | goto nla_put_failure; |
3b98c0c2 | 2548 | |
f399002e LE |
2549 | if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive)) |
2550 | goto nla_put_failure; | |
2551 | ||
daeda1cc | 2552 | rcu_read_lock(); |
3b98c0c2 | 2553 | if (got_ldev) |
daeda1cc | 2554 | if (disk_conf_to_skb(skb, rcu_dereference(mdev->ldev->disk_conf), exclude_sensitive)) |
3b98c0c2 | 2555 | goto nla_put_failure; |
44ed167d | 2556 | |
44ed167d PR |
2557 | nc = rcu_dereference(mdev->tconn->net_conf); |
2558 | if (nc) | |
2559 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2560 | rcu_read_unlock(); | |
2561 | if (err) | |
2562 | goto nla_put_failure; | |
3b98c0c2 | 2563 | |
3b98c0c2 LE |
2564 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2565 | if (!nla) | |
2566 | goto nla_put_failure; | |
2567 | NLA_PUT_U32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY); | |
2568 | NLA_PUT_U32(skb, T_current_state, mdev->state.i); | |
2569 | NLA_PUT_U64(skb, T_ed_uuid, mdev->ed_uuid); | |
2570 | NLA_PUT_U64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev)); | |
2571 | ||
2572 | if (got_ldev) { | |
2573 | NLA_PUT_U32(skb, T_disk_flags, mdev->ldev->md.flags); | |
2574 | NLA_PUT(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid); | |
2575 | NLA_PUT_U64(skb, T_bits_total, drbd_bm_bits(mdev)); | |
2576 | NLA_PUT_U64(skb, T_bits_oos, drbd_bm_total_weight(mdev)); | |
2577 | if (C_SYNC_SOURCE <= mdev->state.conn && | |
2578 | C_PAUSED_SYNC_T >= mdev->state.conn) { | |
2579 | NLA_PUT_U64(skb, T_bits_rs_total, mdev->rs_total); | |
2580 | NLA_PUT_U64(skb, T_bits_rs_failed, mdev->rs_failed); | |
2581 | } | |
b411b363 PR |
2582 | } |
2583 | ||
3b98c0c2 LE |
2584 | if (sib) { |
2585 | switch(sib->sib_reason) { | |
2586 | case SIB_SYNC_PROGRESS: | |
2587 | case SIB_GET_STATUS_REPLY: | |
2588 | break; | |
2589 | case SIB_STATE_CHANGE: | |
2590 | NLA_PUT_U32(skb, T_prev_state, sib->os.i); | |
2591 | NLA_PUT_U32(skb, T_new_state, sib->ns.i); | |
2592 | break; | |
2593 | case SIB_HELPER_POST: | |
2594 | NLA_PUT_U32(skb, | |
2595 | T_helper_exit_code, sib->helper_exit_code); | |
2596 | /* fall through */ | |
2597 | case SIB_HELPER_PRE: | |
2598 | NLA_PUT_STRING(skb, T_helper, sib->helper_name); | |
2599 | break; | |
2600 | } | |
b411b363 | 2601 | } |
3b98c0c2 | 2602 | nla_nest_end(skb, nla); |
b411b363 | 2603 | |
3b98c0c2 LE |
2604 | if (0) |
2605 | nla_put_failure: | |
2606 | err = -EMSGSIZE; | |
2607 | if (got_ldev) | |
2608 | put_ldev(mdev); | |
3b98c0c2 | 2609 | return err; |
b411b363 PR |
2610 | } |
2611 | ||
3b98c0c2 | 2612 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2613 | { |
3b98c0c2 LE |
2614 | enum drbd_ret_code retcode; |
2615 | int err; | |
b411b363 | 2616 | |
3b98c0c2 LE |
2617 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2618 | if (!adm_ctx.reply_skb) | |
2619 | return retcode; | |
2620 | if (retcode != NO_ERROR) | |
2621 | goto out; | |
b411b363 | 2622 | |
3b98c0c2 LE |
2623 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL); |
2624 | if (err) { | |
2625 | nlmsg_free(adm_ctx.reply_skb); | |
2626 | return err; | |
b411b363 | 2627 | } |
3b98c0c2 LE |
2628 | out: |
2629 | drbd_adm_finish(info, retcode); | |
2630 | return 0; | |
b411b363 PR |
2631 | } |
2632 | ||
71932efc | 2633 | int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2634 | { |
3b98c0c2 LE |
2635 | struct drbd_conf *mdev; |
2636 | struct drbd_genlmsghdr *dh; | |
543cc10b LE |
2637 | struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0]; |
2638 | struct drbd_tconn *tconn = NULL; | |
2639 | struct drbd_tconn *tmp; | |
2640 | unsigned volume = cb->args[1]; | |
2641 | ||
2642 | /* Open coded, deferred, iteration: | |
2643 | * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) { | |
2644 | * idr_for_each_entry(&tconn->volumes, mdev, i) { | |
2645 | * ... | |
2646 | * } | |
2647 | * } | |
2648 | * where tconn is cb->args[0]; | |
2649 | * and i is cb->args[1]; | |
2650 | * | |
71932efc LE |
2651 | * cb->args[2] indicates if we shall loop over all resources, |
2652 | * or just dump all volumes of a single resource. | |
2653 | * | |
3b98c0c2 LE |
2654 | * This may miss entries inserted after this dump started, |
2655 | * or entries deleted before they are reached. | |
543cc10b LE |
2656 | * |
2657 | * We need to make sure the mdev won't disappear while | |
2658 | * we are looking at it, and revalidate our iterators | |
2659 | * on each iteration. | |
2660 | */ | |
3b98c0c2 | 2661 | |
9dc9fbb3 | 2662 | /* synchronize with conn_create()/conn_destroy() */ |
ef356262 | 2663 | down_read(&drbd_cfg_rwsem); |
543cc10b LE |
2664 | /* revalidate iterator position */ |
2665 | list_for_each_entry(tmp, &drbd_tconns, all_tconn) { | |
2666 | if (pos == NULL) { | |
2667 | /* first iteration */ | |
2668 | pos = tmp; | |
2669 | tconn = pos; | |
2670 | break; | |
2671 | } | |
2672 | if (tmp == pos) { | |
2673 | tconn = pos; | |
2674 | break; | |
2675 | } | |
2676 | } | |
2677 | if (tconn) { | |
71932efc | 2678 | next_tconn: |
543cc10b LE |
2679 | mdev = idr_get_next(&tconn->volumes, &volume); |
2680 | if (!mdev) { | |
2681 | /* No more volumes to dump on this tconn. | |
2682 | * Advance tconn iterator. */ | |
2683 | pos = list_entry(tconn->all_tconn.next, | |
2684 | struct drbd_tconn, all_tconn); | |
71932efc | 2685 | /* Did we dump any volume on this tconn yet? */ |
543cc10b | 2686 | if (volume != 0) { |
71932efc LE |
2687 | /* If we reached the end of the list, |
2688 | * or only a single resource dump was requested, | |
2689 | * we are done. */ | |
2690 | if (&pos->all_tconn == &drbd_tconns || cb->args[2]) | |
2691 | goto out; | |
543cc10b | 2692 | volume = 0; |
71932efc | 2693 | tconn = pos; |
543cc10b LE |
2694 | goto next_tconn; |
2695 | } | |
2696 | } | |
2697 | ||
3b98c0c2 LE |
2698 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, |
2699 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
2700 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
2701 | if (!dh) | |
543cc10b LE |
2702 | goto out; |
2703 | ||
2704 | if (!mdev) { | |
2705 | /* this is a tconn without a single volume */ | |
2706 | dh->minor = -1U; | |
2707 | dh->ret_code = NO_ERROR; | |
2708 | if (nla_put_drbd_cfg_context(skb, tconn->name, VOLUME_UNSPECIFIED)) | |
2709 | genlmsg_cancel(skb, dh); | |
2710 | else | |
2711 | genlmsg_end(skb, dh); | |
2712 | goto out; | |
2713 | } | |
3b98c0c2 | 2714 | |
543cc10b LE |
2715 | D_ASSERT(mdev->vnr == volume); |
2716 | D_ASSERT(mdev->tconn == tconn); | |
3b98c0c2 | 2717 | |
543cc10b | 2718 | dh->minor = mdev_to_minor(mdev); |
3b98c0c2 LE |
2719 | dh->ret_code = NO_ERROR; |
2720 | ||
2721 | if (nla_put_status_info(skb, mdev, NULL)) { | |
2722 | genlmsg_cancel(skb, dh); | |
543cc10b | 2723 | goto out; |
3b98c0c2 LE |
2724 | } |
2725 | genlmsg_end(skb, dh); | |
2726 | } | |
b411b363 | 2727 | |
543cc10b | 2728 | out: |
ef356262 | 2729 | up_read(&drbd_cfg_rwsem); |
543cc10b LE |
2730 | /* where to start the next iteration */ |
2731 | cb->args[0] = (long)pos; | |
2732 | cb->args[1] = (pos == tconn) ? volume + 1 : 0; | |
b411b363 | 2733 | |
543cc10b LE |
2734 | /* No more tconns/volumes/minors found results in an empty skb. |
2735 | * Which will terminate the dump. */ | |
3b98c0c2 | 2736 | return skb->len; |
b411b363 PR |
2737 | } |
2738 | ||
71932efc LE |
2739 | /* |
2740 | * Request status of all resources, or of all volumes within a single resource. | |
2741 | * | |
2742 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
2743 | * Which means we cannot use the family->attrbuf or other such members, because | |
2744 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
2745 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
2746 | * | |
2747 | * Once things are setup properly, we call into get_one_status(). | |
2748 | */ | |
2749 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) | |
2750 | { | |
2751 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; | |
2752 | struct nlattr *nla; | |
2753 | const char *conn_name; | |
2754 | struct drbd_tconn *tconn; | |
2755 | ||
2756 | /* Is this a followup call? */ | |
2757 | if (cb->args[0]) { | |
2758 | /* ... of a single resource dump, | |
2759 | * and the resource iterator has been advanced already? */ | |
2760 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
2761 | return 0; /* DONE. */ | |
2762 | goto dump; | |
2763 | } | |
2764 | ||
2765 | /* First call (from netlink_dump_start). We need to figure out | |
2766 | * which resource(s) the user wants us to dump. */ | |
2767 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
2768 | nlmsg_attrlen(cb->nlh, hdrlen), | |
2769 | DRBD_NLA_CFG_CONTEXT); | |
2770 | ||
2771 | /* No explicit context given. Dump all. */ | |
2772 | if (!nla) | |
2773 | goto dump; | |
2774 | nla = nla_find_nested(nla, __nla_type(T_ctx_conn_name)); | |
2775 | /* context given, but no name present? */ | |
2776 | if (!nla) | |
2777 | return -EINVAL; | |
2778 | conn_name = nla_data(nla); | |
0ace9dfa PR |
2779 | tconn = conn_get_by_name(conn_name); |
2780 | ||
71932efc LE |
2781 | if (!tconn) |
2782 | return -ENODEV; | |
2783 | ||
0ace9dfa PR |
2784 | kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */ |
2785 | ||
71932efc LE |
2786 | /* prime iterators, and set "filter" mode mark: |
2787 | * only dump this tconn. */ | |
2788 | cb->args[0] = (long)tconn; | |
2789 | /* cb->args[1] = 0; passed in this way. */ | |
2790 | cb->args[2] = (long)tconn; | |
2791 | ||
2792 | dump: | |
2793 | return get_one_status(skb, cb); | |
2794 | } | |
2795 | ||
3b98c0c2 | 2796 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2797 | { |
3b98c0c2 LE |
2798 | enum drbd_ret_code retcode; |
2799 | struct timeout_parms tp; | |
2800 | int err; | |
b411b363 | 2801 | |
3b98c0c2 LE |
2802 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2803 | if (!adm_ctx.reply_skb) | |
2804 | return retcode; | |
2805 | if (retcode != NO_ERROR) | |
2806 | goto out; | |
b411b363 | 2807 | |
3b98c0c2 LE |
2808 | tp.timeout_type = |
2809 | adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : | |
2810 | test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED : | |
2811 | UT_DEFAULT; | |
b411b363 | 2812 | |
3b98c0c2 LE |
2813 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
2814 | if (err) { | |
2815 | nlmsg_free(adm_ctx.reply_skb); | |
2816 | return err; | |
2817 | } | |
2818 | out: | |
2819 | drbd_adm_finish(info, retcode); | |
2820 | return 0; | |
b411b363 PR |
2821 | } |
2822 | ||
3b98c0c2 | 2823 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2824 | { |
3b98c0c2 LE |
2825 | struct drbd_conf *mdev; |
2826 | enum drbd_ret_code retcode; | |
b411b363 | 2827 | |
3b98c0c2 LE |
2828 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2829 | if (!adm_ctx.reply_skb) | |
2830 | return retcode; | |
2831 | if (retcode != NO_ERROR) | |
2832 | goto out; | |
873b0d5f | 2833 | |
3b98c0c2 LE |
2834 | mdev = adm_ctx.mdev; |
2835 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { | |
2836 | /* resume from last known position, if possible */ | |
2837 | struct start_ov_parms parms = | |
2838 | { .ov_start_sector = mdev->ov_start_sector }; | |
f399002e | 2839 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2840 | if (err) { |
2841 | retcode = ERR_MANDATORY_TAG; | |
2842 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2843 | goto out; | |
2844 | } | |
2845 | /* w_make_ov_request expects position to be aligned */ | |
2846 | mdev->ov_start_sector = parms.ov_start_sector & ~BM_SECT_PER_BIT; | |
2847 | } | |
873b0d5f LE |
2848 | /* If there is still bitmap IO pending, e.g. previous resync or verify |
2849 | * just being finished, wait for it before requesting a new resync. */ | |
2850 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
3b98c0c2 LE |
2851 | retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S)); |
2852 | out: | |
2853 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2854 | return 0; |
2855 | } | |
2856 | ||
2857 | ||
3b98c0c2 | 2858 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2859 | { |
3b98c0c2 LE |
2860 | struct drbd_conf *mdev; |
2861 | enum drbd_ret_code retcode; | |
b411b363 PR |
2862 | int skip_initial_sync = 0; |
2863 | int err; | |
3b98c0c2 | 2864 | struct new_c_uuid_parms args; |
b411b363 | 2865 | |
3b98c0c2 LE |
2866 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2867 | if (!adm_ctx.reply_skb) | |
2868 | return retcode; | |
2869 | if (retcode != NO_ERROR) | |
2870 | goto out_nolock; | |
b411b363 | 2871 | |
3b98c0c2 LE |
2872 | mdev = adm_ctx.mdev; |
2873 | memset(&args, 0, sizeof(args)); | |
2874 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 2875 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
2876 | if (err) { |
2877 | retcode = ERR_MANDATORY_TAG; | |
2878 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2879 | goto out_nolock; | |
2880 | } | |
b411b363 PR |
2881 | } |
2882 | ||
8410da8f | 2883 | mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 PR |
2884 | |
2885 | if (!get_ldev(mdev)) { | |
2886 | retcode = ERR_NO_DISK; | |
2887 | goto out; | |
2888 | } | |
2889 | ||
2890 | /* this is "skip initial sync", assume to be clean */ | |
31890f4a | 2891 | if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 && |
b411b363 PR |
2892 | mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
2893 | dev_info(DEV, "Preparing to skip initial sync\n"); | |
2894 | skip_initial_sync = 1; | |
2895 | } else if (mdev->state.conn != C_STANDALONE) { | |
2896 | retcode = ERR_CONNECTED; | |
2897 | goto out_dec; | |
2898 | } | |
2899 | ||
2900 | drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ | |
2901 | drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */ | |
2902 | ||
2903 | if (args.clear_bm) { | |
20ceb2b2 LE |
2904 | err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, |
2905 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); | |
b411b363 PR |
2906 | if (err) { |
2907 | dev_err(DEV, "Writing bitmap failed with %d\n",err); | |
2908 | retcode = ERR_IO_MD_DISK; | |
2909 | } | |
2910 | if (skip_initial_sync) { | |
2911 | drbd_send_uuids_skip_initial_sync(mdev); | |
2912 | _drbd_uuid_set(mdev, UI_BITMAP, 0); | |
62b0da3a | 2913 | drbd_print_uuids(mdev, "cleared bitmap UUID"); |
87eeee41 | 2914 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2915 | _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
2916 | CS_VERBOSE, NULL); | |
87eeee41 | 2917 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2918 | } |
2919 | } | |
2920 | ||
2921 | drbd_md_sync(mdev); | |
2922 | out_dec: | |
2923 | put_ldev(mdev); | |
2924 | out: | |
8410da8f | 2925 | mutex_unlock(mdev->state_mutex); |
3b98c0c2 LE |
2926 | out_nolock: |
2927 | drbd_adm_finish(info, retcode); | |
774b3055 PR |
2928 | return 0; |
2929 | } | |
2930 | ||
3b98c0c2 LE |
2931 | static enum drbd_ret_code |
2932 | drbd_check_conn_name(const char *name) | |
774b3055 | 2933 | { |
3b98c0c2 LE |
2934 | if (!name || !name[0]) { |
2935 | drbd_msg_put_info("connection name missing"); | |
2936 | return ERR_MANDATORY_TAG; | |
774b3055 | 2937 | } |
3b98c0c2 LE |
2938 | /* if we want to use these in sysfs/configfs/debugfs some day, |
2939 | * we must not allow slashes */ | |
2940 | if (strchr(name, '/')) { | |
2941 | drbd_msg_put_info("invalid connection name"); | |
2942 | return ERR_INVALID_REQUEST; | |
774b3055 | 2943 | } |
3b98c0c2 | 2944 | return NO_ERROR; |
774b3055 PR |
2945 | } |
2946 | ||
3b98c0c2 | 2947 | int drbd_adm_create_connection(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2948 | { |
3b98c0c2 | 2949 | enum drbd_ret_code retcode; |
9f5180e5 | 2950 | |
3b98c0c2 LE |
2951 | retcode = drbd_adm_prepare(skb, info, 0); |
2952 | if (!adm_ctx.reply_skb) | |
2953 | return retcode; | |
2954 | if (retcode != NO_ERROR) | |
2955 | goto out; | |
b411b363 | 2956 | |
3b98c0c2 LE |
2957 | retcode = drbd_check_conn_name(adm_ctx.conn_name); |
2958 | if (retcode != NO_ERROR) | |
2959 | goto out; | |
b411b363 | 2960 | |
3b98c0c2 | 2961 | if (adm_ctx.tconn) { |
38f19616 LE |
2962 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
2963 | retcode = ERR_INVALID_REQUEST; | |
2964 | drbd_msg_put_info("connection exists"); | |
2965 | } | |
2966 | /* else: still NO_ERROR */ | |
3b98c0c2 | 2967 | goto out; |
b411b363 PR |
2968 | } |
2969 | ||
9dc9fbb3 | 2970 | if (!conn_create(adm_ctx.conn_name)) |
b411b363 | 2971 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
2972 | out: |
2973 | drbd_adm_finish(info, retcode); | |
2974 | return 0; | |
b411b363 PR |
2975 | } |
2976 | ||
3b98c0c2 | 2977 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2978 | { |
3b98c0c2 LE |
2979 | struct drbd_genlmsghdr *dh = info->userhdr; |
2980 | enum drbd_ret_code retcode; | |
b411b363 | 2981 | |
3b98c0c2 LE |
2982 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
2983 | if (!adm_ctx.reply_skb) | |
2984 | return retcode; | |
2985 | if (retcode != NO_ERROR) | |
2986 | goto out; | |
b411b363 | 2987 | |
3b98c0c2 LE |
2988 | /* FIXME drop minor_count parameter, limit to MINORMASK */ |
2989 | if (dh->minor >= minor_count) { | |
2990 | drbd_msg_put_info("requested minor out of range"); | |
2991 | retcode = ERR_INVALID_REQUEST; | |
2992 | goto out; | |
b411b363 | 2993 | } |
0c8e36d9 | 2994 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
2995 | drbd_msg_put_info("requested volume id out of range"); |
2996 | retcode = ERR_INVALID_REQUEST; | |
2997 | goto out; | |
b411b363 | 2998 | } |
b411b363 | 2999 | |
38f19616 LE |
3000 | /* drbd_adm_prepare made sure already |
3001 | * that mdev->tconn and mdev->vnr match the request. */ | |
3002 | if (adm_ctx.mdev) { | |
3003 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) | |
3004 | retcode = ERR_MINOR_EXISTS; | |
3005 | /* else: still NO_ERROR */ | |
3006 | goto out; | |
3007 | } | |
3008 | ||
d3fcb490 | 3009 | down_write(&drbd_cfg_rwsem); |
3b98c0c2 | 3010 | retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume); |
d3fcb490 | 3011 | up_write(&drbd_cfg_rwsem); |
3b98c0c2 LE |
3012 | out: |
3013 | drbd_adm_finish(info, retcode); | |
3014 | return 0; | |
b411b363 PR |
3015 | } |
3016 | ||
85f75dd7 LE |
3017 | static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev) |
3018 | { | |
3019 | if (mdev->state.disk == D_DISKLESS && | |
3020 | /* no need to be mdev->state.conn == C_STANDALONE && | |
3021 | * we may want to delete a minor from a live replication group. | |
3022 | */ | |
3023 | mdev->state.role == R_SECONDARY) { | |
ff370e5a | 3024 | drbd_delete_device(mdev); |
85f75dd7 LE |
3025 | return NO_ERROR; |
3026 | } else | |
3027 | return ERR_MINOR_CONFIGURED; | |
3028 | } | |
3029 | ||
3b98c0c2 | 3030 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3031 | { |
3b98c0c2 | 3032 | enum drbd_ret_code retcode; |
b411b363 | 3033 | |
3b98c0c2 LE |
3034 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3035 | if (!adm_ctx.reply_skb) | |
3036 | return retcode; | |
3037 | if (retcode != NO_ERROR) | |
3038 | goto out; | |
b411b363 | 3039 | |
ef356262 | 3040 | down_write(&drbd_cfg_rwsem); |
85f75dd7 | 3041 | retcode = adm_delete_minor(adm_ctx.mdev); |
ef356262 | 3042 | up_write(&drbd_cfg_rwsem); |
85f75dd7 LE |
3043 | out: |
3044 | drbd_adm_finish(info, retcode); | |
3045 | return 0; | |
3046 | } | |
3047 | ||
3048 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) | |
3049 | { | |
f3dfa40a | 3050 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 LE |
3051 | struct drbd_conf *mdev; |
3052 | unsigned i; | |
3053 | ||
3054 | retcode = drbd_adm_prepare(skb, info, 0); | |
3055 | if (!adm_ctx.reply_skb) | |
3056 | return retcode; | |
3057 | if (retcode != NO_ERROR) | |
3058 | goto out; | |
3059 | ||
3060 | if (!adm_ctx.tconn) { | |
3061 | retcode = ERR_CONN_NOT_KNOWN; | |
3062 | goto out; | |
3063 | } | |
3064 | ||
ef356262 | 3065 | down_read(&drbd_cfg_rwsem); |
85f75dd7 LE |
3066 | /* demote */ |
3067 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
3068 | retcode = drbd_set_role(mdev, R_SECONDARY, 0); | |
3069 | if (retcode < SS_SUCCESS) { | |
3070 | drbd_msg_put_info("failed to demote"); | |
3071 | goto out_unlock; | |
3072 | } | |
3073 | } | |
f3dfa40a | 3074 | up_read(&drbd_cfg_rwsem); |
85f75dd7 | 3075 | |
f3dfa40a LE |
3076 | /* disconnect; may stop the receiver; |
3077 | * must not hold the drbd_cfg_rwsem */ | |
3078 | retcode = conn_try_disconnect(adm_ctx.tconn, 0); | |
3079 | if (retcode < SS_SUCCESS) { | |
85f75dd7 | 3080 | drbd_msg_put_info("failed to disconnect"); |
f3dfa40a | 3081 | goto out; |
85f75dd7 LE |
3082 | } |
3083 | ||
f3dfa40a | 3084 | down_read(&drbd_cfg_rwsem); |
85f75dd7 LE |
3085 | /* detach */ |
3086 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
f3dfa40a LE |
3087 | retcode = adm_detach(mdev); |
3088 | if (retcode < SS_SUCCESS) { | |
85f75dd7 LE |
3089 | drbd_msg_put_info("failed to detach"); |
3090 | goto out_unlock; | |
3091 | } | |
3092 | } | |
ef356262 | 3093 | up_read(&drbd_cfg_rwsem); |
85f75dd7 | 3094 | |
f3dfa40a LE |
3095 | /* If we reach this, all volumes (of this tconn) are Secondary, |
3096 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have | |
3097 | * actually stopped, state handling only does drbd_thread_stop_nowait(). | |
3098 | * This needs to be done without holding drbd_cfg_rwsem. */ | |
3099 | drbd_thread_stop(&adm_ctx.tconn->worker); | |
3100 | ||
3101 | /* Now, nothing can fail anymore */ | |
3102 | ||
85f75dd7 | 3103 | /* delete volumes */ |
ef356262 | 3104 | down_write(&drbd_cfg_rwsem); |
85f75dd7 LE |
3105 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { |
3106 | retcode = adm_delete_minor(mdev); | |
3107 | if (retcode != NO_ERROR) { | |
3108 | /* "can not happen" */ | |
3109 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 PR |
3110 | up_write(&drbd_cfg_rwsem); |
3111 | goto out; | |
85f75dd7 LE |
3112 | } |
3113 | } | |
3114 | ||
85f75dd7 LE |
3115 | /* delete connection */ |
3116 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { | |
9dc9fbb3 PR |
3117 | list_del(&adm_ctx.tconn->all_tconn); |
3118 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
3119 | ||
85f75dd7 LE |
3120 | retcode = NO_ERROR; |
3121 | } else { | |
3122 | /* "can not happen" */ | |
3123 | retcode = ERR_CONN_IN_USE; | |
3124 | drbd_msg_put_info("failed to delete connection"); | |
85f75dd7 | 3125 | } |
ef356262 PR |
3126 | up_write(&drbd_cfg_rwsem); |
3127 | goto out; | |
85f75dd7 | 3128 | out_unlock: |
ef356262 | 3129 | up_read(&drbd_cfg_rwsem); |
3b98c0c2 LE |
3130 | out: |
3131 | drbd_adm_finish(info, retcode); | |
3132 | return 0; | |
b411b363 PR |
3133 | } |
3134 | ||
3b98c0c2 | 3135 | int drbd_adm_delete_connection(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3136 | { |
3b98c0c2 | 3137 | enum drbd_ret_code retcode; |
b411b363 | 3138 | |
3b98c0c2 LE |
3139 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONN); |
3140 | if (!adm_ctx.reply_skb) | |
3141 | return retcode; | |
3142 | if (retcode != NO_ERROR) | |
3143 | goto out; | |
3144 | ||
ef356262 | 3145 | down_write(&drbd_cfg_rwsem); |
3b98c0c2 | 3146 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { |
9dc9fbb3 PR |
3147 | list_del(&adm_ctx.tconn->all_tconn); |
3148 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
3149 | ||
3b98c0c2 LE |
3150 | retcode = NO_ERROR; |
3151 | } else { | |
3152 | retcode = ERR_CONN_IN_USE; | |
b411b363 | 3153 | } |
ef356262 | 3154 | up_write(&drbd_cfg_rwsem); |
b411b363 | 3155 | |
992d6e91 LE |
3156 | if (retcode == NO_ERROR) |
3157 | drbd_thread_stop(&adm_ctx.tconn->worker); | |
3b98c0c2 LE |
3158 | out: |
3159 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3160 | return 0; |
3161 | } | |
3162 | ||
3b98c0c2 | 3163 | void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib) |
b411b363 | 3164 | { |
3b98c0c2 LE |
3165 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3166 | struct sk_buff *msg; | |
3167 | struct drbd_genlmsghdr *d_out; | |
3168 | unsigned seq; | |
3169 | int err = -ENOMEM; | |
3170 | ||
3171 | seq = atomic_inc_return(&drbd_genl_seq); | |
3172 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3173 | if (!msg) | |
3174 | goto failed; | |
3175 | ||
3176 | err = -EMSGSIZE; | |
3177 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3178 | if (!d_out) /* cannot happen, but anyways. */ | |
3179 | goto nla_put_failure; | |
3180 | d_out->minor = mdev_to_minor(mdev); | |
3181 | d_out->ret_code = 0; | |
3182 | ||
3183 | if (nla_put_status_info(msg, mdev, sib)) | |
3184 | goto nla_put_failure; | |
3185 | genlmsg_end(msg, d_out); | |
3186 | err = drbd_genl_multicast_events(msg, 0); | |
3187 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3188 | if (err && err != -ESRCH) | |
3189 | goto failed; | |
b411b363 | 3190 | |
3b98c0c2 | 3191 | return; |
b411b363 | 3192 | |
3b98c0c2 LE |
3193 | nla_put_failure: |
3194 | nlmsg_free(msg); | |
3195 | failed: | |
3196 | dev_err(DEV, "Error %d while broadcasting event. " | |
3197 | "Event seq:%u sib_reason:%u\n", | |
3198 | err, seq, sib->sib_reason); | |
b411b363 | 3199 | } |