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" | |
a3603a6e | 35 | #include "drbd_protocol.h" |
265be2d0 | 36 | #include "drbd_req.h" |
b411b363 PR |
37 | #include "drbd_wrappers.h" |
38 | #include <asm/unaligned.h> | |
b411b363 | 39 | #include <linux/drbd_limits.h> |
87f7be4c | 40 | #include <linux/kthread.h> |
b411b363 | 41 | |
3b98c0c2 LE |
42 | #include <net/genetlink.h> |
43 | ||
44 | /* .doit */ | |
45 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
47 | ||
05a10ec7 AG |
48 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
49 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 50 | |
789c1b62 AG |
51 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
52 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 53 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
54 | |
55 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
56 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 57 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
58 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
59 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 60 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
61 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
62 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 72 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
73 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
74 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
75 | /* .dumpit */ | |
76 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
77 | ||
78 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 79 | #include "drbd_nla.h" |
3b98c0c2 LE |
80 | #include <linux/genl_magic_func.h> |
81 | ||
82 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
83 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
84 | ||
3b98c0c2 LE |
85 | /* Configuration is strictly serialized, because generic netlink message |
86 | * processing is strictly serialized by the genl_lock(). | |
87 | * Which means we can use one static global drbd_config_context struct. | |
88 | */ | |
89 | static struct drbd_config_context { | |
90 | /* assigned from drbd_genlmsghdr */ | |
91 | unsigned int minor; | |
92 | /* assigned from request attributes, if present */ | |
93 | unsigned int volume; | |
94 | #define VOLUME_UNSPECIFIED (-1U) | |
95 | /* pointer into the request skb, | |
96 | * limited lifetime! */ | |
7c3063cc | 97 | char *resource_name; |
089c075d AG |
98 | struct nlattr *my_addr; |
99 | struct nlattr *peer_addr; | |
3b98c0c2 LE |
100 | |
101 | /* reply buffer */ | |
102 | struct sk_buff *reply_skb; | |
103 | /* pointer into reply buffer */ | |
104 | struct drbd_genlmsghdr *reply_dh; | |
105 | /* resolved from attributes, if possible */ | |
b30ab791 | 106 | struct drbd_device *device; |
4bc76048 | 107 | struct drbd_resource *resource; |
bde89a9e | 108 | struct drbd_connection *connection; |
3b98c0c2 LE |
109 | } adm_ctx; |
110 | ||
111 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
112 | { | |
113 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
114 | if (genlmsg_reply(skb, info)) | |
115 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 116 | } |
3b98c0c2 LE |
117 | |
118 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
119 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 120 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
121 | { |
122 | struct sk_buff *skb = adm_ctx.reply_skb; | |
123 | struct nlattr *nla; | |
124 | int err = -EMSGSIZE; | |
125 | ||
126 | if (!info || !info[0]) | |
127 | return 0; | |
128 | ||
129 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
130 | if (!nla) | |
131 | return err; | |
132 | ||
133 | err = nla_put_string(skb, T_info_text, info); | |
134 | if (err) { | |
135 | nla_nest_cancel(skb, nla); | |
136 | return err; | |
137 | } else | |
138 | nla_nest_end(skb, nla); | |
139 | return 0; | |
b411b363 PR |
140 | } |
141 | ||
3b98c0c2 LE |
142 | /* This would be a good candidate for a "pre_doit" hook, |
143 | * and per-family private info->pointers. | |
144 | * But we need to stay compatible with older kernels. | |
145 | * If it returns successfully, adm_ctx members are valid. | |
146 | */ | |
147 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 148 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 149 | #define DRBD_ADM_NEED_CONNECTION 4 |
3b98c0c2 LE |
150 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, |
151 | unsigned flags) | |
152 | { | |
153 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
154 | const u8 cmd = info->genlhdr->cmd; | |
155 | int err; | |
156 | ||
157 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
158 | ||
159 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 160 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
161 | return -EPERM; |
162 | ||
163 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
1e2a2551 AG |
164 | if (!adm_ctx.reply_skb) { |
165 | err = -ENOMEM; | |
3b98c0c2 | 166 | goto fail; |
1e2a2551 | 167 | } |
3b98c0c2 LE |
168 | |
169 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
170 | info, &drbd_genl_family, 0, cmd); | |
171 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
172 | * but anyways */ | |
1e2a2551 AG |
173 | if (!adm_ctx.reply_dh) { |
174 | err = -ENOMEM; | |
3b98c0c2 | 175 | goto fail; |
1e2a2551 | 176 | } |
3b98c0c2 LE |
177 | |
178 | adm_ctx.reply_dh->minor = d_in->minor; | |
179 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
180 | ||
089c075d | 181 | adm_ctx.volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
182 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
183 | struct nlattr *nla; | |
184 | /* parse and validate only */ | |
f399002e | 185 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
186 | if (err) |
187 | goto fail; | |
188 | ||
189 | /* It was present, and valid, | |
190 | * copy it over to the reply skb. */ | |
191 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
192 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
193 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
194 | if (err) | |
195 | goto fail; | |
196 | ||
197 | /* and assign stuff to the global adm_ctx */ | |
198 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
089c075d AG |
199 | if (nla) |
200 | adm_ctx.volume = nla_get_u32(nla); | |
7c3063cc | 201 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 202 | if (nla) |
7c3063cc | 203 | adm_ctx.resource_name = nla_data(nla); |
089c075d AG |
204 | adm_ctx.my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; |
205 | adm_ctx.peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
206 | if ((adm_ctx.my_addr && | |
bde89a9e | 207 | nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.connection->my_addr)) || |
089c075d | 208 | (adm_ctx.peer_addr && |
bde89a9e | 209 | nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.connection->peer_addr))) { |
089c075d AG |
210 | err = -EINVAL; |
211 | goto fail; | |
212 | } | |
213 | } | |
3b98c0c2 LE |
214 | |
215 | adm_ctx.minor = d_in->minor; | |
b30ab791 | 216 | adm_ctx.device = minor_to_device(d_in->minor); |
4bc76048 AG |
217 | if (adm_ctx.resource_name) { |
218 | adm_ctx.resource = drbd_find_resource(adm_ctx.resource_name); | |
219 | if (adm_ctx.resource) { | |
220 | adm_ctx.connection = first_connection(adm_ctx.resource); | |
221 | kref_get(&adm_ctx.connection->kref); | |
222 | } | |
223 | } | |
3b98c0c2 | 224 | |
b30ab791 | 225 | if (!adm_ctx.device && (flags & DRBD_ADM_NEED_MINOR)) { |
3b98c0c2 LE |
226 | drbd_msg_put_info("unknown minor"); |
227 | return ERR_MINOR_INVALID; | |
228 | } | |
4bc76048 | 229 | if (!adm_ctx.resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
44e52cfa | 230 | drbd_msg_put_info("unknown resource"); |
a10f6b8a AG |
231 | if (adm_ctx.resource_name) |
232 | return ERR_RES_NOT_KNOWN; | |
3b98c0c2 LE |
233 | return ERR_INVALID_REQUEST; |
234 | } | |
235 | ||
089c075d | 236 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
bde89a9e | 237 | if (adm_ctx.connection && !(flags & DRBD_ADM_NEED_RESOURCE)) { |
089c075d AG |
238 | drbd_msg_put_info("no resource name expected"); |
239 | return ERR_INVALID_REQUEST; | |
240 | } | |
b30ab791 | 241 | if (adm_ctx.device) { |
089c075d AG |
242 | drbd_msg_put_info("no minor number expected"); |
243 | return ERR_INVALID_REQUEST; | |
244 | } | |
245 | if (adm_ctx.my_addr && adm_ctx.peer_addr) | |
bde89a9e | 246 | adm_ctx.connection = conn_get_by_addrs(nla_data(adm_ctx.my_addr), |
089c075d AG |
247 | nla_len(adm_ctx.my_addr), |
248 | nla_data(adm_ctx.peer_addr), | |
249 | nla_len(adm_ctx.peer_addr)); | |
bde89a9e | 250 | if (!adm_ctx.connection) { |
089c075d AG |
251 | drbd_msg_put_info("unknown connection"); |
252 | return ERR_INVALID_REQUEST; | |
253 | } | |
254 | } | |
255 | ||
3b98c0c2 | 256 | /* some more paranoia, if the request was over-determined */ |
4bc76048 AG |
257 | if (adm_ctx.device && adm_ctx.resource && |
258 | adm_ctx.device->resource != adm_ctx.resource) { | |
259 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", | |
260 | adm_ctx.minor, adm_ctx.resource->name, | |
d8628a86 | 261 | adm_ctx.device->resource->name); |
44e52cfa | 262 | drbd_msg_put_info("minor exists in different resource"); |
527f4b24 LE |
263 | return ERR_INVALID_REQUEST; |
264 | } | |
b30ab791 | 265 | if (adm_ctx.device && |
3b98c0c2 | 266 | adm_ctx.volume != VOLUME_UNSPECIFIED && |
b30ab791 | 267 | adm_ctx.volume != adm_ctx.device->vnr) { |
3b98c0c2 LE |
268 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
269 | adm_ctx.minor, adm_ctx.volume, | |
77c556f6 | 270 | adm_ctx.device->vnr, |
d8628a86 | 271 | adm_ctx.device->resource->name); |
527f4b24 | 272 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
273 | return ERR_INVALID_REQUEST; |
274 | } | |
0ace9dfa | 275 | |
3b98c0c2 LE |
276 | return NO_ERROR; |
277 | ||
278 | fail: | |
279 | nlmsg_free(adm_ctx.reply_skb); | |
280 | adm_ctx.reply_skb = NULL; | |
1e2a2551 | 281 | return err; |
3b98c0c2 LE |
282 | } |
283 | ||
284 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
285 | { | |
bde89a9e | 286 | if (adm_ctx.connection) { |
05a10ec7 | 287 | kref_put(&adm_ctx.connection->kref, drbd_destroy_connection); |
bde89a9e | 288 | adm_ctx.connection = NULL; |
0ace9dfa | 289 | } |
4bc76048 AG |
290 | if (adm_ctx.resource) { |
291 | kref_put(&adm_ctx.resource->kref, drbd_destroy_resource); | |
292 | adm_ctx.resource = NULL; | |
293 | } | |
0ace9dfa | 294 | |
3b98c0c2 LE |
295 | if (!adm_ctx.reply_skb) |
296 | return -ENOMEM; | |
297 | ||
298 | adm_ctx.reply_dh->ret_code = retcode; | |
3b98c0c2 LE |
299 | drbd_adm_send_reply(adm_ctx.reply_skb, info); |
300 | return 0; | |
301 | } | |
b411b363 | 302 | |
bde89a9e | 303 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 304 | { |
6b75dced | 305 | char *afs; |
b411b363 | 306 | |
089c075d | 307 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 308 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
309 | return; |
310 | ||
bde89a9e | 311 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
312 | case AF_INET6: |
313 | afs = "ipv6"; | |
314 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 315 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 316 | break; |
089c075d AG |
317 | case AF_INET: |
318 | afs = "ipv4"; | |
319 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 320 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 321 | break; |
089c075d AG |
322 | default: |
323 | afs = "ssocks"; | |
324 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 325 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 326 | } |
089c075d | 327 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 328 | } |
b411b363 | 329 | |
b30ab791 | 330 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
331 | { |
332 | char *envp[] = { "HOME=/", | |
333 | "TERM=linux", | |
334 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
335 | (char[20]) { }, /* address family */ |
336 | (char[60]) { }, /* address */ | |
b411b363 | 337 | NULL }; |
6b75dced | 338 | char mb[12]; |
b411b363 | 339 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 340 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 341 | struct sib_info sib; |
b411b363 PR |
342 | int ret; |
343 | ||
bde89a9e AG |
344 | if (current == connection->worker.task) |
345 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 346 | |
b30ab791 | 347 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 348 | setup_khelper_env(connection, envp); |
b411b363 | 349 | |
1090c056 LE |
350 | /* The helper may take some time. |
351 | * write out any unsynced meta data changes now */ | |
b30ab791 | 352 | drbd_md_sync(device); |
1090c056 | 353 | |
d0180171 | 354 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
355 | sib.sib_reason = SIB_HELPER_PRE; |
356 | sib.helper_name = cmd; | |
b30ab791 | 357 | drbd_bcast_event(device, &sib); |
70834d30 | 358 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 359 | if (ret) |
d0180171 | 360 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
361 | usermode_helper, cmd, mb, |
362 | (ret >> 8) & 0xff, ret); | |
363 | else | |
d0180171 | 364 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
365 | usermode_helper, cmd, mb, |
366 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
367 | sib.sib_reason = SIB_HELPER_POST; |
368 | sib.helper_exit_code = ret; | |
b30ab791 | 369 | drbd_bcast_event(device, &sib); |
b411b363 | 370 | |
bde89a9e AG |
371 | if (current == connection->worker.task) |
372 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
373 | |
374 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
375 | ret = 0; | |
376 | ||
377 | return ret; | |
378 | } | |
379 | ||
bde89a9e | 380 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
381 | { |
382 | char *envp[] = { "HOME=/", | |
383 | "TERM=linux", | |
384 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
385 | (char[20]) { }, /* address family */ | |
386 | (char[60]) { }, /* address */ | |
387 | NULL }; | |
77c556f6 AG |
388 | char *resource_name = connection->resource->name; |
389 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
390 | int ret; |
391 | ||
bde89a9e AG |
392 | setup_khelper_env(connection, envp); |
393 | conn_md_sync(connection); | |
6b75dced | 394 | |
1ec861eb | 395 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced PR |
396 | /* TODO: conn_bcast_event() ?? */ |
397 | ||
98683650 | 398 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 399 | if (ret) |
1ec861eb | 400 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 401 | usermode_helper, cmd, resource_name, |
6b75dced PR |
402 | (ret >> 8) & 0xff, ret); |
403 | else | |
1ec861eb | 404 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 405 | usermode_helper, cmd, resource_name, |
6b75dced PR |
406 | (ret >> 8) & 0xff, ret); |
407 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 408 | |
b411b363 PR |
409 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
410 | ret = 0; | |
411 | ||
412 | return ret; | |
413 | } | |
414 | ||
bde89a9e | 415 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 416 | { |
cb703454 | 417 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 418 | struct drbd_peer_device *peer_device; |
cb703454 PR |
419 | int vnr; |
420 | ||
695d08fa | 421 | rcu_read_lock(); |
c06ece6b AG |
422 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
423 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 424 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
425 | struct disk_conf *disk_conf = |
426 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
427 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 428 | put_ldev(device); |
cb703454 PR |
429 | } |
430 | } | |
695d08fa | 431 | rcu_read_unlock(); |
cb703454 PR |
432 | |
433 | return fp; | |
434 | } | |
435 | ||
bde89a9e | 436 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 437 | { |
28e448bb | 438 | unsigned int connect_cnt; |
cb703454 PR |
439 | union drbd_state mask = { }; |
440 | union drbd_state val = { }; | |
441 | enum drbd_fencing_p fp; | |
b411b363 PR |
442 | char *ex_to_string; |
443 | int r; | |
b411b363 | 444 | |
bde89a9e | 445 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 446 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
cb703454 PR |
447 | return false; |
448 | } | |
b411b363 | 449 | |
bde89a9e AG |
450 | spin_lock_irq(&connection->req_lock); |
451 | connect_cnt = connection->connect_cnt; | |
452 | spin_unlock_irq(&connection->req_lock); | |
28e448bb | 453 | |
bde89a9e | 454 | fp = highest_fencing_policy(connection); |
cb703454 PR |
455 | switch (fp) { |
456 | case FP_NOT_AVAIL: | |
1ec861eb | 457 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 458 | goto out; |
cb703454 PR |
459 | case FP_DONT_CARE: |
460 | return true; | |
461 | default: ; | |
b411b363 PR |
462 | } |
463 | ||
bde89a9e | 464 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
465 | |
466 | switch ((r>>8) & 0xff) { | |
467 | case 3: /* peer is inconsistent */ | |
468 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
469 | mask.pdsk = D_MASK; |
470 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
471 | break; |
472 | case 4: /* peer got outdated, or was already outdated */ | |
473 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
474 | mask.pdsk = D_MASK; |
475 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
476 | break; |
477 | case 5: /* peer was down */ | |
bde89a9e | 478 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
479 | /* we will(have) create(d) a new UUID anyways... */ |
480 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
481 | mask.pdsk = D_MASK; |
482 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
483 | } else { |
484 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
485 | } |
486 | break; | |
487 | case 6: /* Peer is primary, voluntarily outdate myself. | |
488 | * This is useful when an unconnected R_SECONDARY is asked to | |
489 | * become R_PRIMARY, but finds the other peer being active. */ | |
490 | ex_to_string = "peer is active"; | |
1ec861eb | 491 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
492 | mask.disk = D_MASK; |
493 | val.disk = D_OUTDATED; | |
b411b363 PR |
494 | break; |
495 | case 7: | |
496 | if (fp != FP_STONITH) | |
1ec861eb | 497 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 498 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
499 | mask.pdsk = D_MASK; |
500 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
501 | break; |
502 | default: | |
503 | /* The script is broken ... */ | |
1ec861eb | 504 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 505 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
506 | } |
507 | ||
1ec861eb | 508 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 509 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 510 | |
cb703454 | 511 | out: |
fb22c402 | 512 | |
cb703454 | 513 | /* Not using |
bde89a9e | 514 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
515 | here, because we might were able to re-establish the connection in the |
516 | meantime. */ | |
bde89a9e AG |
517 | spin_lock_irq(&connection->req_lock); |
518 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { | |
519 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
520 | /* In case the connection was established and droped |
521 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 522 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 523 | else |
bde89a9e | 524 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 525 | } |
bde89a9e | 526 | spin_unlock_irq(&connection->req_lock); |
cb703454 | 527 | |
bde89a9e | 528 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
529 | } |
530 | ||
87f7be4c PR |
531 | static int _try_outdate_peer_async(void *data) |
532 | { | |
bde89a9e | 533 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 534 | |
bde89a9e | 535 | conn_try_outdate_peer(connection); |
87f7be4c | 536 | |
05a10ec7 | 537 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
538 | return 0; |
539 | } | |
540 | ||
bde89a9e | 541 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
542 | { |
543 | struct task_struct *opa; | |
544 | ||
bde89a9e AG |
545 | kref_get(&connection->kref); |
546 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); | |
9dc9fbb3 | 547 | if (IS_ERR(opa)) { |
1ec861eb | 548 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 549 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 550 | } |
87f7be4c | 551 | } |
b411b363 | 552 | |
bf885f8a | 553 | enum drbd_state_rv |
b30ab791 | 554 | drbd_set_role(struct drbd_device *device, enum drbd_role new_role, int force) |
b411b363 PR |
555 | { |
556 | const int max_tries = 4; | |
bf885f8a | 557 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 558 | struct net_conf *nc; |
b411b363 PR |
559 | int try = 0; |
560 | int forced = 0; | |
561 | union drbd_state mask, val; | |
b411b363 | 562 | |
b6f85ef9 AG |
563 | if (new_role == R_PRIMARY) { |
564 | struct drbd_connection *connection; | |
565 | ||
566 | /* Detect dead peers as soon as possible. */ | |
567 | ||
568 | rcu_read_lock(); | |
569 | for_each_connection(connection, device->resource) | |
570 | request_ping(connection); | |
571 | rcu_read_unlock(); | |
572 | } | |
b411b363 | 573 | |
b30ab791 | 574 | mutex_lock(device->state_mutex); |
b411b363 PR |
575 | |
576 | mask.i = 0; mask.role = R_MASK; | |
577 | val.i = 0; val.role = new_role; | |
578 | ||
579 | while (try++ < max_tries) { | |
b30ab791 | 580 | rv = _drbd_request_state(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
581 | |
582 | /* in case we first succeeded to outdate, | |
583 | * but now suddenly could establish a connection */ | |
bf885f8a | 584 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
585 | val.pdsk = 0; |
586 | mask.pdsk = 0; | |
587 | continue; | |
588 | } | |
589 | ||
bf885f8a | 590 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
591 | (device->state.disk < D_UP_TO_DATE && |
592 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
593 | mask.disk = D_MASK; |
594 | val.disk = D_UP_TO_DATE; | |
595 | forced = 1; | |
596 | continue; | |
597 | } | |
598 | ||
bf885f8a | 599 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 600 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 601 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 602 | |
a6b32bc3 | 603 | if (conn_try_outdate_peer(first_peer_device(device)->connection)) { |
b411b363 PR |
604 | val.disk = D_UP_TO_DATE; |
605 | mask.disk = D_MASK; | |
606 | } | |
b411b363 PR |
607 | continue; |
608 | } | |
609 | ||
bf885f8a | 610 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 611 | goto out; |
bf885f8a | 612 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
a6b32bc3 | 613 | if (!conn_try_outdate_peer(first_peer_device(device)->connection) && force) { |
d0180171 | 614 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
615 | mask.pdsk = D_MASK; |
616 | val.pdsk = D_OUTDATED; | |
b411b363 | 617 | |
cb703454 | 618 | } |
b411b363 PR |
619 | continue; |
620 | } | |
bf885f8a | 621 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
622 | /* Maybe the peer is detected as dead very soon... |
623 | retry at most once more in this case. */ | |
44ed167d PR |
624 | int timeo; |
625 | rcu_read_lock(); | |
a6b32bc3 | 626 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
627 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
628 | rcu_read_unlock(); | |
629 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
630 | if (try < max_tries) |
631 | try = max_tries - 1; | |
632 | continue; | |
633 | } | |
bf885f8a | 634 | if (rv < SS_SUCCESS) { |
b30ab791 | 635 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 636 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 637 | if (rv < SS_SUCCESS) |
3b98c0c2 | 638 | goto out; |
b411b363 PR |
639 | } |
640 | break; | |
641 | } | |
642 | ||
bf885f8a | 643 | if (rv < SS_SUCCESS) |
3b98c0c2 | 644 | goto out; |
b411b363 PR |
645 | |
646 | if (forced) | |
d0180171 | 647 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
648 | |
649 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 650 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 651 | |
b6dd1a89 LE |
652 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
653 | ||
b411b363 | 654 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
655 | set_disk_ro(device->vdisk, true); |
656 | if (get_ldev(device)) { | |
657 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
658 | put_ldev(device); | |
b411b363 PR |
659 | } |
660 | } else { | |
a6b32bc3 AG |
661 | mutex_lock(&first_peer_device(device)->connection->conf_update); |
662 | nc = first_peer_device(device)->connection->net_conf; | |
44ed167d | 663 | if (nc) |
6139f60d | 664 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
a6b32bc3 | 665 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
91fd4dad | 666 | |
b30ab791 AG |
667 | set_disk_ro(device->vdisk, false); |
668 | if (get_ldev(device)) { | |
669 | if (((device->state.conn < C_CONNECTED || | |
670 | device->state.pdsk <= D_FAILED) | |
671 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
672 | drbd_uuid_new_current(device); | |
b411b363 | 673 | |
b30ab791 AG |
674 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
675 | put_ldev(device); | |
b411b363 PR |
676 | } |
677 | } | |
678 | ||
19f843aa LE |
679 | /* writeout of activity log covered areas of the bitmap |
680 | * to stable storage done in after state change already */ | |
b411b363 | 681 | |
b30ab791 | 682 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
683 | /* if this was forced, we should consider sync */ |
684 | if (forced) | |
b30ab791 AG |
685 | drbd_send_uuids(device); |
686 | drbd_send_current_state(device); | |
b411b363 PR |
687 | } |
688 | ||
b30ab791 | 689 | drbd_md_sync(device); |
b411b363 | 690 | |
b30ab791 | 691 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 692 | out: |
b30ab791 | 693 | mutex_unlock(device->state_mutex); |
bf885f8a | 694 | return rv; |
b411b363 PR |
695 | } |
696 | ||
3b98c0c2 | 697 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 698 | { |
3b98c0c2 LE |
699 | return err == -ENOMSG ? "required attribute missing" : |
700 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 701 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 702 | "invalid attribute value"; |
ef50a3e3 | 703 | } |
b411b363 | 704 | |
3b98c0c2 | 705 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 706 | { |
3b98c0c2 LE |
707 | struct set_role_parms parms; |
708 | int err; | |
709 | enum drbd_ret_code retcode; | |
b411b363 | 710 | |
3b98c0c2 LE |
711 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
712 | if (!adm_ctx.reply_skb) | |
713 | return retcode; | |
714 | if (retcode != NO_ERROR) | |
715 | goto out; | |
b411b363 | 716 | |
3b98c0c2 LE |
717 | memset(&parms, 0, sizeof(parms)); |
718 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 719 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
720 | if (err) { |
721 | retcode = ERR_MANDATORY_TAG; | |
722 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
723 | goto out; | |
724 | } | |
725 | } | |
b411b363 | 726 | |
3b98c0c2 | 727 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 728 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 729 | else |
b30ab791 | 730 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
3b98c0c2 LE |
731 | out: |
732 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
733 | return 0; |
734 | } | |
735 | ||
ae8bf312 LE |
736 | /* Initializes the md.*_offset members, so we are able to find |
737 | * the on disk meta data. | |
738 | * | |
739 | * We currently have two possible layouts: | |
740 | * external: | |
741 | * |----------- md_size_sect ------------------| | |
742 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
743 | * | al_offset == 8 | | |
744 | * | bm_offset = al_offset + X | | |
745 | * ==> bitmap sectors = md_size_sect - bm_offset | |
746 | * | |
747 | * internal: | |
748 | * |----------- md_size_sect ------------------| | |
749 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
750 | * | al_offset < 0 | | |
751 | * | bm_offset = al_offset - Y | | |
752 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
753 | * | |
754 | * Activity log size used to be fixed 32kB, | |
755 | * but is about to become configurable. | |
756 | */ | |
b30ab791 | 757 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
758 | struct drbd_backing_dev *bdev) |
759 | { | |
760 | sector_t md_size_sect = 0; | |
c04ccaa6 | 761 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 762 | |
3a4d4eb3 LE |
763 | bdev->md.md_offset = drbd_md_ss(bdev); |
764 | ||
68e41a43 | 765 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
766 | default: |
767 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 768 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
769 | bdev->md.al_offset = MD_4kB_SECT; |
770 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
771 | break; |
772 | case DRBD_MD_INDEX_FLEX_EXT: | |
773 | /* just occupy the full device; unit: sectors */ | |
774 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
775 | bdev->md.al_offset = MD_4kB_SECT; |
776 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
777 | break; |
778 | case DRBD_MD_INDEX_INTERNAL: | |
779 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 780 | /* al size is still fixed */ |
ae8bf312 | 781 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
782 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
783 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
784 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
785 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
786 | md_size_sect = ALIGN(md_size_sect, 8); | |
787 | ||
788 | /* plus the "drbd meta data super block", | |
789 | * and the activity log; */ | |
ae8bf312 | 790 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
791 | |
792 | bdev->md.md_size_sect = md_size_sect; | |
793 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 794 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
795 | break; |
796 | } | |
797 | } | |
798 | ||
4b0715f0 | 799 | /* input size is expected to be in KB */ |
b411b363 PR |
800 | char *ppsize(char *buf, unsigned long long size) |
801 | { | |
4b0715f0 LE |
802 | /* Needs 9 bytes at max including trailing NUL: |
803 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
804 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
805 | int base = 0; | |
4b0715f0 | 806 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
807 | /* shift + round */ |
808 | size = (size >> 10) + !!(size & (1<<9)); | |
809 | base++; | |
810 | } | |
4b0715f0 | 811 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
812 | |
813 | return buf; | |
814 | } | |
815 | ||
816 | /* there is still a theoretical deadlock when called from receiver | |
817 | * on an D_INCONSISTENT R_PRIMARY: | |
818 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
819 | * packet from remote to dec_ap_bio again. | |
820 | * receiver receive_sizes(), comes here, | |
821 | * waits for ap_bio_cnt == 0. -> deadlock. | |
822 | * but this cannot happen, actually, because: | |
823 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
824 | * (not connected, or bad/no disk on peer): | |
825 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
826 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
827 | * peer may not initiate a resize. | |
828 | */ | |
3b98c0c2 LE |
829 | /* Note these are not to be confused with |
830 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
831 | * which are (sub) state changes triggered by admin (drbdsetup), | |
832 | * and can be long lived. | |
b30ab791 | 833 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 834 | * and should be short-lived. */ |
b30ab791 | 835 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 836 | { |
b30ab791 AG |
837 | set_bit(SUSPEND_IO, &device->flags); |
838 | if (drbd_suspended(device)) | |
265be2d0 | 839 | return; |
b30ab791 | 840 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
841 | } |
842 | ||
b30ab791 | 843 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 844 | { |
b30ab791 AG |
845 | clear_bit(SUSPEND_IO, &device->flags); |
846 | wake_up(&device->misc_wait); | |
b411b363 PR |
847 | } |
848 | ||
849 | /** | |
850 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 851 | * @device: DRBD device. |
b411b363 PR |
852 | * |
853 | * Returns 0 on success, negative return values indicate errors. | |
854 | * You should call drbd_md_sync() after calling this function. | |
855 | */ | |
d752b269 | 856 | enum determine_dev_size |
b30ab791 | 857 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
858 | { |
859 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 860 | sector_t la_size_sect, u_size; |
b30ab791 | 861 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
862 | u32 prev_al_stripe_size_4k; |
863 | u32 prev_al_stripes; | |
b411b363 PR |
864 | sector_t size; |
865 | char ppb[10]; | |
d752b269 | 866 | void *buffer; |
b411b363 PR |
867 | |
868 | int md_moved, la_size_changed; | |
e96c9633 | 869 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
870 | |
871 | /* race: | |
872 | * application request passes inc_ap_bio, | |
873 | * but then cannot get an AL-reference. | |
874 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
875 | * | |
876 | * to avoid that: | |
877 | * Suspend IO right here. | |
878 | * still lock the act_log to not trigger ASSERTs there. | |
879 | */ | |
b30ab791 AG |
880 | drbd_suspend_io(device); |
881 | buffer = drbd_md_get_buffer(device); /* Lock meta-data IO */ | |
d752b269 | 882 | if (!buffer) { |
b30ab791 | 883 | drbd_resume_io(device); |
d752b269 PR |
884 | return DS_ERROR; |
885 | } | |
b411b363 PR |
886 | |
887 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 888 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 889 | |
b30ab791 AG |
890 | prev_first_sect = drbd_md_first_sector(device->ldev); |
891 | prev_size = device->ldev->md.md_size_sect; | |
892 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 893 | |
d752b269 PR |
894 | if (rs) { |
895 | /* rs is non NULL if we should change the AL layout only */ | |
896 | ||
897 | prev_al_stripes = md->al_stripes; | |
898 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
899 | ||
900 | md->al_stripes = rs->al_stripes; | |
901 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
902 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
903 | } | |
904 | ||
b30ab791 | 905 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 906 | |
daeda1cc | 907 | rcu_read_lock(); |
b30ab791 | 908 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 909 | rcu_read_unlock(); |
b30ab791 | 910 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 911 | |
d752b269 PR |
912 | if (size < la_size_sect) { |
913 | if (rs && u_size == 0) { | |
914 | /* Remove "rs &&" later. This check should always be active, but | |
915 | right now the receiver expects the permissive behavior */ | |
d0180171 | 916 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
917 | "Use --size=%llus for explicit shrink.\n", |
918 | (unsigned long long)size); | |
919 | rv = DS_ERROR_SHRINK; | |
920 | } | |
921 | if (u_size > size) | |
922 | rv = DS_ERROR_SPACE_MD; | |
923 | if (rv != DS_UNCHANGED) | |
924 | goto err_out; | |
925 | } | |
926 | ||
b30ab791 AG |
927 | if (drbd_get_capacity(device->this_bdev) != size || |
928 | drbd_bm_capacity(device) != size) { | |
b411b363 | 929 | int err; |
b30ab791 | 930 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
931 | if (unlikely(err)) { |
932 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 933 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 934 | if (size == 0) { |
d0180171 | 935 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
936 | "Could not allocate bitmap!\n"); |
937 | } else { | |
d0180171 | 938 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
939 | "Leaving size unchanged at size = %lu KB\n", |
940 | (unsigned long)size); | |
941 | } | |
e96c9633 | 942 | rv = DS_ERROR; |
b411b363 PR |
943 | } |
944 | /* racy, see comments above. */ | |
b30ab791 AG |
945 | drbd_set_my_capacity(device, size); |
946 | device->ldev->md.la_size_sect = size; | |
d0180171 | 947 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
948 | (unsigned long long)size>>1); |
949 | } | |
d752b269 PR |
950 | if (rv <= DS_ERROR) |
951 | goto err_out; | |
b411b363 | 952 | |
b30ab791 | 953 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 954 | |
b30ab791 AG |
955 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
956 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 957 | |
d752b269 PR |
958 | if (la_size_changed || md_moved || rs) { |
959 | u32 prev_flags; | |
24dccabb | 960 | |
b30ab791 | 961 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
962 | |
963 | prev_flags = md->flags; | |
964 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 965 | drbd_md_write(device, buffer); |
d752b269 | 966 | |
d0180171 | 967 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
968 | la_size_changed && md_moved ? "size changed and md moved" : |
969 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 970 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 971 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 972 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 973 | drbd_initialize_al(device, buffer); |
d752b269 PR |
974 | |
975 | md->flags = prev_flags; | |
b30ab791 | 976 | drbd_md_write(device, buffer); |
d752b269 PR |
977 | |
978 | if (rs) | |
d0180171 AG |
979 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
980 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
981 | } |
982 | ||
cccac985 | 983 | if (size > la_size_sect) |
57737adc | 984 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 985 | if (size < la_size_sect) |
e96c9633 | 986 | rv = DS_SHRUNK; |
d752b269 PR |
987 | |
988 | if (0) { | |
989 | err_out: | |
990 | if (rs) { | |
991 | md->al_stripes = prev_al_stripes; | |
992 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
993 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
994 | ||
b30ab791 | 995 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
996 | } |
997 | } | |
b30ab791 AG |
998 | lc_unlock(device->act_log); |
999 | wake_up(&device->al_wait); | |
1000 | drbd_md_put_buffer(device); | |
1001 | drbd_resume_io(device); | |
b411b363 PR |
1002 | |
1003 | return rv; | |
1004 | } | |
1005 | ||
1006 | sector_t | |
b30ab791 | 1007 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1008 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1009 | { |
b30ab791 | 1010 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1011 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1012 | sector_t m_size; /* my size */ |
b411b363 PR |
1013 | sector_t size = 0; |
1014 | ||
1015 | m_size = drbd_get_max_capacity(bdev); | |
1016 | ||
b30ab791 | 1017 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1018 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1019 | p_size = m_size; |
1020 | } | |
1021 | ||
b411b363 PR |
1022 | if (p_size && m_size) { |
1023 | size = min_t(sector_t, p_size, m_size); | |
1024 | } else { | |
cccac985 LE |
1025 | if (la_size_sect) { |
1026 | size = la_size_sect; | |
b411b363 PR |
1027 | if (m_size && m_size < size) |
1028 | size = m_size; | |
1029 | if (p_size && p_size < size) | |
1030 | size = p_size; | |
1031 | } else { | |
1032 | if (m_size) | |
1033 | size = m_size; | |
1034 | if (p_size) | |
1035 | size = p_size; | |
1036 | } | |
1037 | } | |
1038 | ||
1039 | if (size == 0) | |
d0180171 | 1040 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1041 | |
1042 | if (u_size) { | |
1043 | if (u_size > size) | |
d0180171 | 1044 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1045 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1046 | else | |
1047 | size = u_size; | |
1048 | } | |
1049 | ||
1050 | return size; | |
1051 | } | |
1052 | ||
1053 | /** | |
1054 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1055 | * @device: DRBD device. |
b411b363 PR |
1056 | * |
1057 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1058 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1059 | * this function. | |
1060 | */ | |
b30ab791 | 1061 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1062 | { |
1063 | struct lru_cache *n, *t; | |
1064 | struct lc_element *e; | |
1065 | unsigned int in_use; | |
1066 | int i; | |
1067 | ||
b30ab791 AG |
1068 | if (device->act_log && |
1069 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1070 | return 0; |
1071 | ||
1072 | in_use = 0; | |
b30ab791 | 1073 | t = device->act_log; |
7ad651b5 | 1074 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1075 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1076 | |
1077 | if (n == NULL) { | |
d0180171 | 1078 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1079 | return -ENOMEM; |
1080 | } | |
b30ab791 | 1081 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1082 | if (t) { |
1083 | for (i = 0; i < t->nr_elements; i++) { | |
1084 | e = lc_element_by_index(t, i); | |
1085 | if (e->refcnt) | |
d0180171 | 1086 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1087 | e->lc_number, e->refcnt); |
1088 | in_use += e->refcnt; | |
1089 | } | |
1090 | } | |
1091 | if (!in_use) | |
b30ab791 AG |
1092 | device->act_log = n; |
1093 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1094 | if (in_use) { |
d0180171 | 1095 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1096 | lc_destroy(n); |
1097 | return -EBUSY; | |
1098 | } else { | |
1099 | if (t) | |
1100 | lc_destroy(t); | |
1101 | } | |
b30ab791 | 1102 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1103 | return 0; |
1104 | } | |
1105 | ||
b30ab791 | 1106 | static void drbd_setup_queue_param(struct drbd_device *device, unsigned int max_bio_size) |
b411b363 | 1107 | { |
b30ab791 | 1108 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1109 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1110 | unsigned int max_segments = 0; | |
99432fcc | 1111 | |
b30ab791 AG |
1112 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1113 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1114 | |
1115 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1116 | rcu_read_lock(); |
b30ab791 | 1117 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1118 | rcu_read_unlock(); |
b30ab791 | 1119 | put_ldev(device); |
99432fcc | 1120 | } |
b411b363 | 1121 | |
b411b363 | 1122 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1123 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1124 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1125 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1126 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1127 | |
b30ab791 AG |
1128 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1129 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1130 | |
1131 | blk_queue_stack_limits(q, b); | |
1132 | ||
1133 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1134 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1135 | q->backing_dev_info.ra_pages, |
1136 | b->backing_dev_info.ra_pages); | |
1137 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1138 | } | |
b30ab791 | 1139 | put_ldev(device); |
99432fcc PR |
1140 | } |
1141 | } | |
1142 | ||
b30ab791 | 1143 | void drbd_reconsider_max_bio_size(struct drbd_device *device) |
99432fcc | 1144 | { |
db141b2f | 1145 | unsigned int now, new, local, peer; |
99432fcc | 1146 | |
b30ab791 AG |
1147 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1148 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1149 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1150 | |
b30ab791 AG |
1151 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1152 | local = queue_max_hw_sectors(device->ldev->backing_bdev->bd_disk->queue) << 9; | |
1153 | device->local_max_bio_size = local; | |
1154 | put_ldev(device); | |
b411b363 | 1155 | } |
db141b2f | 1156 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1157 | |
1158 | /* We may ignore peer limits if the peer is modern enough. | |
1159 | Because new from 8.3.8 onwards the peer can use multiple | |
1160 | BIOs for a single peer_request */ | |
b30ab791 | 1161 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1162 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1163 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1164 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1165 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1166 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1167 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1168 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1169 | else | |
99432fcc PR |
1170 | peer = DRBD_MAX_BIO_SIZE; |
1171 | } | |
1172 | ||
db141b2f | 1173 | new = min(local, peer); |
99432fcc | 1174 | |
b30ab791 | 1175 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1176 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1177 | |
1178 | if (new != now) | |
d0180171 | 1179 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1180 | |
b30ab791 | 1181 | drbd_setup_queue_param(device, new); |
b411b363 PR |
1182 | } |
1183 | ||
a18e9d1e | 1184 | /* Starts the worker thread */ |
bde89a9e | 1185 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1186 | { |
bde89a9e AG |
1187 | drbd_thread_start(&connection->worker); |
1188 | conn_flush_workqueue(connection); | |
b411b363 PR |
1189 | } |
1190 | ||
a18e9d1e | 1191 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1192 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1193 | { |
992d6e91 | 1194 | bool stop_threads; |
bde89a9e AG |
1195 | spin_lock_irq(&connection->req_lock); |
1196 | stop_threads = conn_all_vols_unconf(connection) && | |
1197 | connection->cstate == C_STANDALONE; | |
1198 | spin_unlock_irq(&connection->req_lock); | |
992d6e91 LE |
1199 | if (stop_threads) { |
1200 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1201 | * in conn_disconnect() */ |
bde89a9e AG |
1202 | drbd_thread_stop(&connection->receiver); |
1203 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1204 | } |
b411b363 PR |
1205 | } |
1206 | ||
0778286a | 1207 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1208 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1209 | { |
1210 | int s = 0; | |
1211 | ||
b30ab791 | 1212 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1213 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1214 | return; |
1215 | } | |
1216 | ||
b30ab791 | 1217 | drbd_al_shrink(device); |
a6b32bc3 | 1218 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 AG |
1219 | if (device->state.conn < C_CONNECTED) |
1220 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
a6b32bc3 | 1221 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 1222 | lc_unlock(device->act_log); |
0778286a PR |
1223 | |
1224 | if (s) | |
d0180171 | 1225 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1226 | } |
1227 | ||
5979e361 LE |
1228 | |
1229 | static bool should_set_defaults(struct genl_info *info) | |
1230 | { | |
1231 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1232 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1233 | } | |
1234 | ||
5bbcf5e6 | 1235 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1236 | { |
5bbcf5e6 LE |
1237 | /* This is limited by 16 bit "slot" numbers, |
1238 | * and by available on-disk context storage. | |
1239 | * | |
1240 | * Also (u16)~0 is special (denotes a "free" extent). | |
1241 | * | |
1242 | * One transaction occupies one 4kB on-disk block, | |
1243 | * we have n such blocks in the on disk ring buffer, | |
1244 | * the "current" transaction may fail (n-1), | |
1245 | * and there is 919 slot numbers context information per transaction. | |
1246 | * | |
1247 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1248 | * so cap there first. | |
1249 | */ | |
1250 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1251 | const unsigned int sufficient_on_disk = | |
1252 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1253 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1254 | |
5bbcf5e6 LE |
1255 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1256 | ||
1257 | if (al_size_4k > sufficient_on_disk) | |
1258 | return max_al_nr; | |
1259 | ||
1260 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1261 | } |
1262 | ||
f399002e LE |
1263 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1264 | { | |
1265 | enum drbd_ret_code retcode; | |
b30ab791 | 1266 | struct drbd_device *device; |
daeda1cc | 1267 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1268 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1269 | int err, fifo_size; |
f399002e LE |
1270 | |
1271 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1272 | if (!adm_ctx.reply_skb) | |
1273 | return retcode; | |
1274 | if (retcode != NO_ERROR) | |
1275 | goto out; | |
1276 | ||
b30ab791 | 1277 | device = adm_ctx.device; |
f399002e LE |
1278 | |
1279 | /* we also need a disk | |
1280 | * to change the options on */ | |
b30ab791 | 1281 | if (!get_ldev(device)) { |
f399002e LE |
1282 | retcode = ERR_NO_DISK; |
1283 | goto out; | |
1284 | } | |
1285 | ||
daeda1cc | 1286 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1287 | if (!new_disk_conf) { |
f399002e LE |
1288 | retcode = ERR_NOMEM; |
1289 | goto fail; | |
1290 | } | |
1291 | ||
a6b32bc3 | 1292 | mutex_lock(&first_peer_device(device)->connection->conf_update); |
b30ab791 | 1293 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1294 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1295 | if (should_set_defaults(info)) |
b966b5dd | 1296 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1297 | |
5ecc72c3 | 1298 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1299 | if (err && err != -ENOMSG) { |
f399002e LE |
1300 | retcode = ERR_MANDATORY_TAG; |
1301 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
8e229434 | 1302 | goto fail_unlock; |
f399002e LE |
1303 | } |
1304 | ||
5ecc72c3 LE |
1305 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1306 | new_disk_conf->resync_rate = 1; | |
f399002e | 1307 | |
5bbcf5e6 LE |
1308 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1309 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1310 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1311 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1312 | |
1313 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1314 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1315 | |
5ecc72c3 | 1316 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1317 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1318 | new_plan = fifo_alloc(fifo_size); |
1319 | if (!new_plan) { | |
d0180171 | 1320 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1321 | retcode = ERR_NOMEM; |
daeda1cc | 1322 | goto fail_unlock; |
f399002e LE |
1323 | } |
1324 | } | |
1325 | ||
b30ab791 AG |
1326 | drbd_suspend_io(device); |
1327 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1328 | drbd_al_shrink(device); | |
1329 | err = drbd_check_al_size(device, new_disk_conf); | |
1330 | lc_unlock(device->act_log); | |
1331 | wake_up(&device->al_wait); | |
1332 | drbd_resume_io(device); | |
f399002e LE |
1333 | |
1334 | if (err) { | |
1335 | retcode = ERR_NOMEM; | |
daeda1cc | 1336 | goto fail_unlock; |
f399002e LE |
1337 | } |
1338 | ||
dc97b708 | 1339 | write_lock_irq(&global_state_lock); |
b30ab791 | 1340 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1341 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1342 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1343 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1344 | } |
1345 | write_unlock_irq(&global_state_lock); | |
f399002e | 1346 | |
daeda1cc PR |
1347 | if (retcode != NO_ERROR) |
1348 | goto fail_unlock; | |
f399002e | 1349 | |
813472ce | 1350 | if (new_plan) { |
b30ab791 AG |
1351 | old_plan = device->rs_plan_s; |
1352 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1353 | } |
9958c857 | 1354 | |
a6b32bc3 | 1355 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
27eb13e9 | 1356 | |
9a51ab1c | 1357 | if (new_disk_conf->al_updates) |
b30ab791 | 1358 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1359 | else |
b30ab791 | 1360 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1361 | |
691631c0 | 1362 | if (new_disk_conf->md_flushes) |
b30ab791 | 1363 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1364 | else |
b30ab791 | 1365 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1366 | |
a6b32bc3 | 1367 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
27eb13e9 | 1368 | |
b30ab791 | 1369 | drbd_md_sync(device); |
f399002e | 1370 | |
b30ab791 AG |
1371 | if (device->state.conn >= C_CONNECTED) |
1372 | drbd_send_sync_param(device); | |
f399002e | 1373 | |
daeda1cc PR |
1374 | synchronize_rcu(); |
1375 | kfree(old_disk_conf); | |
813472ce | 1376 | kfree(old_plan); |
b30ab791 | 1377 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1378 | goto success; |
1379 | ||
1380 | fail_unlock: | |
a6b32bc3 | 1381 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
f399002e | 1382 | fail: |
5ecc72c3 | 1383 | kfree(new_disk_conf); |
813472ce | 1384 | kfree(new_plan); |
daeda1cc | 1385 | success: |
b30ab791 | 1386 | put_ldev(device); |
f399002e LE |
1387 | out: |
1388 | drbd_adm_finish(info, retcode); | |
1389 | return 0; | |
1390 | } | |
1391 | ||
3b98c0c2 | 1392 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1393 | { |
b30ab791 | 1394 | struct drbd_device *device; |
3b98c0c2 | 1395 | int err; |
116676ca | 1396 | enum drbd_ret_code retcode; |
b411b363 PR |
1397 | enum determine_dev_size dd; |
1398 | sector_t max_possible_sectors; | |
1399 | sector_t min_md_device_sectors; | |
1400 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1401 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1402 | struct block_device *bdev; |
b411b363 | 1403 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1404 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1405 | union drbd_state ns, os; |
f2024e7c | 1406 | enum drbd_state_rv rv; |
44ed167d | 1407 | struct net_conf *nc; |
b411b363 | 1408 | |
3b98c0c2 LE |
1409 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1410 | if (!adm_ctx.reply_skb) | |
1411 | return retcode; | |
1412 | if (retcode != NO_ERROR) | |
40cbf085 | 1413 | goto finish; |
b411b363 | 1414 | |
b30ab791 | 1415 | device = adm_ctx.device; |
a6b32bc3 | 1416 | conn_reconfig_start(first_peer_device(device)->connection); |
b411b363 PR |
1417 | |
1418 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1419 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1420 | retcode = ERR_DISK_CONFIGURED; |
1421 | goto fail; | |
1422 | } | |
82f59cc6 LE |
1423 | /* It may just now have detached because of IO error. Make sure |
1424 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1425 | * e.g. if someone calls attach from the on-io-error handler, | |
1426 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
b30ab791 | 1427 | wait_event(device->misc_wait, !atomic_read(&device->local_cnt)); |
b411b363 | 1428 | |
383606e0 | 1429 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1430 | clear_bit(FORCE_DETACH, &device->flags); |
1431 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1432 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1433 | |
0029d624 | 1434 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1435 | device->rs_total = 0; |
1436 | device->rs_failed = 0; | |
1437 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1438 | |
3b98c0c2 | 1439 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1440 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1441 | if (!nbc) { | |
1442 | retcode = ERR_NOMEM; | |
1443 | goto fail; | |
1444 | } | |
9f2247bb PR |
1445 | spin_lock_init(&nbc->md.uuid_lock); |
1446 | ||
daeda1cc PR |
1447 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1448 | if (!new_disk_conf) { | |
1449 | retcode = ERR_NOMEM; | |
b411b363 PR |
1450 | goto fail; |
1451 | } | |
daeda1cc | 1452 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1453 | |
daeda1cc PR |
1454 | set_disk_conf_defaults(new_disk_conf); |
1455 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1456 | if (err) { |
b411b363 | 1457 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1458 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1459 | goto fail; |
1460 | } | |
1461 | ||
5bbcf5e6 LE |
1462 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1463 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1464 | |
9958c857 PR |
1465 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1466 | if (!new_plan) { | |
1467 | retcode = ERR_NOMEM; | |
1468 | goto fail; | |
1469 | } | |
1470 | ||
daeda1cc | 1471 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1472 | retcode = ERR_MD_IDX_INVALID; |
1473 | goto fail; | |
1474 | } | |
1475 | ||
a3f8f7dc | 1476 | write_lock_irq(&global_state_lock); |
b30ab791 | 1477 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1478 | write_unlock_irq(&global_state_lock); |
1479 | if (retcode != NO_ERROR) | |
1480 | goto fail; | |
1481 | ||
44ed167d | 1482 | rcu_read_lock(); |
a6b32bc3 | 1483 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d | 1484 | if (nc) { |
daeda1cc | 1485 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1486 | rcu_read_unlock(); |
47ff2d0a PR |
1487 | retcode = ERR_STONITH_AND_PROT_A; |
1488 | goto fail; | |
1489 | } | |
1490 | } | |
44ed167d | 1491 | rcu_read_unlock(); |
47ff2d0a | 1492 | |
daeda1cc | 1493 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1494 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1495 | if (IS_ERR(bdev)) { |
d0180171 | 1496 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1497 | PTR_ERR(bdev)); |
b411b363 PR |
1498 | retcode = ERR_OPEN_DISK; |
1499 | goto fail; | |
1500 | } | |
e525fd89 TH |
1501 | nbc->backing_bdev = bdev; |
1502 | ||
1503 | /* | |
1504 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1505 | * drbd sharing one meta device. TODO in that case, paranoia | |
1506 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1507 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1508 | * should check it for you already; but if you don't, or | |
1509 | * someone fooled it, we need to double check here) | |
1510 | */ | |
daeda1cc | 1511 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1512 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1513 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1514 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1515 | if (IS_ERR(bdev)) { |
d0180171 | 1516 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1517 | PTR_ERR(bdev)); |
b411b363 PR |
1518 | retcode = ERR_OPEN_MD_DISK; |
1519 | goto fail; | |
1520 | } | |
e525fd89 | 1521 | nbc->md_bdev = bdev; |
b411b363 | 1522 | |
e525fd89 | 1523 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1524 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1525 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1526 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1527 | goto fail; |
1528 | } | |
1529 | ||
1530 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1531 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1532 | offsetof(struct bm_extent, lce)); |
1533 | if (!resync_lru) { | |
1534 | retcode = ERR_NOMEM; | |
e525fd89 | 1535 | goto fail; |
b411b363 PR |
1536 | } |
1537 | ||
c04ccaa6 LE |
1538 | /* Read our meta data super block early. |
1539 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1540 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1541 | if (retcode != NO_ERROR) |
1542 | goto fail; | |
b411b363 | 1543 | |
5bbcf5e6 LE |
1544 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1545 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1546 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1547 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1548 | ||
daeda1cc | 1549 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1550 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1551 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1552 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1553 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1554 | goto fail; |
b411b363 PR |
1555 | } |
1556 | ||
daeda1cc | 1557 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1558 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1559 | /* at least one MB, otherwise it does not make sense */ | |
1560 | min_md_device_sectors = (2<<10); | |
1561 | } else { | |
1562 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1563 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1564 | } |
1565 | ||
b411b363 | 1566 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1567 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1568 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1569 | "at least %llu sectors needed for this meta-disk type\n", |
1570 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1571 | goto fail; |
b411b363 PR |
1572 | } |
1573 | ||
1574 | /* Make sure the new disk is big enough | |
1575 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1576 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1577 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1578 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1579 | goto fail; |
b411b363 PR |
1580 | } |
1581 | ||
1582 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1583 | ||
1352994b | 1584 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1585 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1586 | "to currently maximum possible %llu sectors <==\n", |
1587 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1588 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1589 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1590 | "meta data may help <<==\n"); |
1591 | } | |
1592 | ||
b30ab791 | 1593 | drbd_suspend_io(device); |
b411b363 | 1594 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1595 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1596 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1597 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1598 | * As barriers are counted per resource, | |
1599 | * we'd need to suspend io on all devices of a resource. | |
1600 | */ | |
b30ab791 | 1601 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1602 | /* and for any other previously queued work */ |
b30ab791 | 1603 | drbd_flush_workqueue(device); |
b411b363 | 1604 | |
b30ab791 | 1605 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1606 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1607 | drbd_resume_io(device); |
f2024e7c | 1608 | if (rv < SS_SUCCESS) |
e525fd89 | 1609 | goto fail; |
b411b363 | 1610 | |
b30ab791 | 1611 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1612 | goto force_diskless; |
1613 | ||
b30ab791 AG |
1614 | if (!device->bitmap) { |
1615 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1616 | retcode = ERR_NOMEM; |
1617 | goto force_diskless_dec; | |
1618 | } | |
1619 | } | |
1620 | ||
b30ab791 AG |
1621 | if (device->state.conn < C_CONNECTED && |
1622 | device->state.role == R_PRIMARY && | |
1623 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
d0180171 | 1624 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1625 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1626 | retcode = ERR_DATA_NOT_CURRENT; |
1627 | goto force_diskless_dec; | |
1628 | } | |
1629 | ||
1630 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1631 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1632 | retcode = ERR_NOMEM; |
1633 | goto force_diskless_dec; | |
1634 | } | |
1635 | ||
1636 | /* Prevent shrinking of consistent devices ! */ | |
1637 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1638 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1639 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1640 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1641 | goto force_diskless_dec; |
1642 | } | |
1643 | ||
b411b363 PR |
1644 | /* Reset the "barriers don't work" bits here, then force meta data to |
1645 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1646 | if (new_disk_conf->md_flushes) |
b30ab791 | 1647 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1648 | else |
b30ab791 | 1649 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1650 | |
1651 | /* Point of no return reached. | |
1652 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1653 | * now device takes over responsibility, and the state engine should |
b411b363 | 1654 | * clean it up somewhere. */ |
0b0ba1ef | 1655 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1656 | device->ldev = nbc; |
1657 | device->resync = resync_lru; | |
1658 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1659 | nbc = NULL; |
1660 | resync_lru = NULL; | |
daeda1cc | 1661 | new_disk_conf = NULL; |
9958c857 | 1662 | new_plan = NULL; |
b411b363 | 1663 | |
a6b32bc3 | 1664 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
b411b363 | 1665 | |
b30ab791 AG |
1666 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1667 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1668 | else |
b30ab791 | 1669 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1670 | |
b30ab791 | 1671 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
a6b32bc3 AG |
1672 | !(device->state.role == R_PRIMARY && |
1673 | first_peer_device(device)->connection->susp_nod)) | |
b30ab791 | 1674 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1675 | |
b30ab791 AG |
1676 | device->send_cnt = 0; |
1677 | device->recv_cnt = 0; | |
1678 | device->read_cnt = 0; | |
1679 | device->writ_cnt = 0; | |
b411b363 | 1680 | |
b30ab791 | 1681 | drbd_reconsider_max_bio_size(device); |
b411b363 PR |
1682 | |
1683 | /* If I am currently not R_PRIMARY, | |
1684 | * but meta data primary indicator is set, | |
1685 | * I just now recover from a hard crash, | |
1686 | * and have been R_PRIMARY before that crash. | |
1687 | * | |
1688 | * Now, if I had no connection before that crash | |
1689 | * (have been degraded R_PRIMARY), chances are that | |
1690 | * I won't find my peer now either. | |
1691 | * | |
1692 | * In that case, and _only_ in that case, | |
1693 | * we use the degr-wfc-timeout instead of the default, | |
1694 | * so we can automatically recover from a crash of a | |
1695 | * degraded but active "cluster" after a certain timeout. | |
1696 | */ | |
b30ab791 AG |
1697 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1698 | if (device->state.role != R_PRIMARY && | |
1699 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1700 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1701 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1702 | |
b30ab791 | 1703 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1704 | if (dd <= DS_ERROR) { |
b411b363 PR |
1705 | retcode = ERR_NOMEM_BITMAP; |
1706 | goto force_diskless_dec; | |
e96c9633 | 1707 | } else if (dd == DS_GREW) |
b30ab791 | 1708 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1709 | |
b30ab791 AG |
1710 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1711 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1712 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1713 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1714 | "(aka FullSync)\n"); |
b30ab791 | 1715 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1716 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1717 | retcode = ERR_IO_MD_DISK; |
1718 | goto force_diskless_dec; | |
1719 | } | |
1720 | } else { | |
b30ab791 | 1721 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1722 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1723 | retcode = ERR_IO_MD_DISK; |
1724 | goto force_diskless_dec; | |
1725 | } | |
b411b363 PR |
1726 | } |
1727 | ||
b30ab791 AG |
1728 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1729 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1730 | |
a6b32bc3 | 1731 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 1732 | os = drbd_read_state(device); |
78bae59b | 1733 | ns = os; |
b411b363 PR |
1734 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1735 | otherwise investigate MDF_WasUpToDate... | |
1736 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1737 | otherwise into D_CONSISTENT state. | |
1738 | */ | |
b30ab791 AG |
1739 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1740 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1741 | ns.disk = D_CONSISTENT; |
1742 | else | |
1743 | ns.disk = D_OUTDATED; | |
1744 | } else { | |
1745 | ns.disk = D_INCONSISTENT; | |
1746 | } | |
1747 | ||
b30ab791 | 1748 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1749 | ns.pdsk = D_OUTDATED; |
1750 | ||
daeda1cc PR |
1751 | rcu_read_lock(); |
1752 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1753 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1754 | ns.disk = D_UP_TO_DATE; |
1755 | ||
1756 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1757 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1758 | this point, because drbd_request_state() modifies these | |
1759 | flags. */ | |
1760 | ||
b30ab791 AG |
1761 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1762 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1763 | else |
b30ab791 | 1764 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1765 | |
1766 | rcu_read_unlock(); | |
1767 | ||
b411b363 PR |
1768 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1769 | state after the negotiation phase. */ | |
b30ab791 AG |
1770 | if (device->state.conn == C_CONNECTED) { |
1771 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1772 | ns.i = os.i; |
1773 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1774 | |
1775 | /* We expect to receive up-to-date UUIDs soon. | |
1776 | To avoid a race in receive_state, free p_uuid while | |
1777 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1778 | kfree(device->p_uuid); |
1779 | device->p_uuid = NULL; | |
b411b363 PR |
1780 | } |
1781 | ||
b30ab791 | 1782 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
a6b32bc3 | 1783 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
b411b363 PR |
1784 | |
1785 | if (rv < SS_SUCCESS) | |
1786 | goto force_diskless_dec; | |
1787 | ||
b30ab791 | 1788 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1789 | |
b30ab791 AG |
1790 | if (device->state.role == R_PRIMARY) |
1791 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1792 | else |
b30ab791 | 1793 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1794 | |
b30ab791 AG |
1795 | drbd_md_mark_dirty(device); |
1796 | drbd_md_sync(device); | |
b411b363 | 1797 | |
b30ab791 AG |
1798 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1799 | put_ldev(device); | |
a6b32bc3 | 1800 | conn_reconfig_done(first_peer_device(device)->connection); |
3b98c0c2 | 1801 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1802 | return 0; |
1803 | ||
1804 | force_diskless_dec: | |
b30ab791 | 1805 | put_ldev(device); |
b411b363 | 1806 | force_diskless: |
b30ab791 AG |
1807 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1808 | drbd_md_sync(device); | |
b411b363 | 1809 | fail: |
a6b32bc3 | 1810 | conn_reconfig_done(first_peer_device(device)->connection); |
b411b363 | 1811 | if (nbc) { |
e525fd89 TH |
1812 | if (nbc->backing_bdev) |
1813 | blkdev_put(nbc->backing_bdev, | |
1814 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1815 | if (nbc->md_bdev) | |
1816 | blkdev_put(nbc->md_bdev, | |
1817 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1818 | kfree(nbc); |
1819 | } | |
daeda1cc | 1820 | kfree(new_disk_conf); |
b411b363 | 1821 | lc_destroy(resync_lru); |
9958c857 | 1822 | kfree(new_plan); |
b411b363 | 1823 | |
40cbf085 | 1824 | finish: |
3b98c0c2 | 1825 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1826 | return 0; |
1827 | } | |
1828 | ||
b30ab791 | 1829 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1830 | { |
19f83c76 | 1831 | enum drbd_state_rv retcode; |
9a0d9d03 | 1832 | int ret; |
02ee8f95 | 1833 | |
cdfda633 | 1834 | if (force) { |
b30ab791 AG |
1835 | set_bit(FORCE_DETACH, &device->flags); |
1836 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1837 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1838 | goto out; |
1839 | } | |
1840 | ||
b30ab791 AG |
1841 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
1842 | drbd_md_get_buffer(device); /* make sure there is no in-flight meta-data IO */ | |
1843 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
1844 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1845 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1846 | ret = wait_event_interruptible(device->misc_wait, |
1847 | device->state.disk != D_FAILED); | |
1848 | drbd_resume_io(device); | |
9b2f61ae | 1849 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1850 | retcode = SS_NOTHING_TO_DO; |
1851 | if (ret) | |
1852 | retcode = ERR_INTR; | |
02ee8f95 | 1853 | out: |
85f75dd7 | 1854 | return retcode; |
b411b363 PR |
1855 | } |
1856 | ||
82f59cc6 LE |
1857 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1858 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1859 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1860 | * internal references as well. | |
1861 | * Only then we have finally detached. */ | |
3b98c0c2 | 1862 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1863 | { |
116676ca | 1864 | enum drbd_ret_code retcode; |
cdfda633 PR |
1865 | struct detach_parms parms = { }; |
1866 | int err; | |
b411b363 | 1867 | |
3b98c0c2 LE |
1868 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1869 | if (!adm_ctx.reply_skb) | |
1870 | return retcode; | |
1871 | if (retcode != NO_ERROR) | |
1872 | goto out; | |
b411b363 | 1873 | |
cdfda633 PR |
1874 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1875 | err = detach_parms_from_attrs(&parms, info); | |
1876 | if (err) { | |
1877 | retcode = ERR_MANDATORY_TAG; | |
1878 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1879 | goto out; | |
1880 | } | |
b411b363 PR |
1881 | } |
1882 | ||
b30ab791 | 1883 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
3b98c0c2 LE |
1884 | out: |
1885 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1886 | return 0; |
1887 | } | |
b411b363 | 1888 | |
bde89a9e | 1889 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1890 | { |
c06ece6b | 1891 | struct drbd_peer_device *peer_device; |
695d08fa | 1892 | bool rv = false; |
f399002e LE |
1893 | int vnr; |
1894 | ||
695d08fa | 1895 | rcu_read_lock(); |
c06ece6b AG |
1896 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1897 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1898 | if (device->state.conn == C_SYNC_SOURCE || |
1899 | device->state.conn == C_SYNC_TARGET || | |
1900 | device->state.conn == C_PAUSED_SYNC_S || | |
1901 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1902 | rv = true; |
1903 | break; | |
1904 | } | |
b411b363 | 1905 | } |
695d08fa | 1906 | rcu_read_unlock(); |
b411b363 | 1907 | |
695d08fa | 1908 | return rv; |
f399002e | 1909 | } |
47ff2d0a | 1910 | |
bde89a9e | 1911 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1912 | { |
c06ece6b | 1913 | struct drbd_peer_device *peer_device; |
695d08fa | 1914 | bool rv = false; |
f399002e LE |
1915 | int vnr; |
1916 | ||
695d08fa | 1917 | rcu_read_lock(); |
c06ece6b AG |
1918 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1919 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1920 | if (device->state.conn == C_VERIFY_S || |
1921 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1922 | rv = true; |
1923 | break; | |
47ff2d0a PR |
1924 | } |
1925 | } | |
695d08fa | 1926 | rcu_read_unlock(); |
b411b363 | 1927 | |
695d08fa | 1928 | return rv; |
f399002e | 1929 | } |
422028b1 | 1930 | |
cd64397c | 1931 | static enum drbd_ret_code |
270eb5c9 | 1932 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 1933 | { |
c06ece6b | 1934 | struct drbd_peer_device *peer_device; |
cd64397c | 1935 | int i; |
b411b363 | 1936 | |
270eb5c9 AG |
1937 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
1938 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 1939 | return ERR_NEED_APV_100; |
b411b363 | 1940 | |
270eb5c9 | 1941 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
1942 | return ERR_NEED_APV_100; |
1943 | ||
270eb5c9 | 1944 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 1945 | return ERR_NEED_APV_100; |
b411b363 PR |
1946 | } |
1947 | ||
270eb5c9 | 1948 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
1949 | conn_highest_role(connection) == R_PRIMARY && |
1950 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 1951 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 1952 | |
270eb5c9 AG |
1953 | if (new_net_conf->two_primaries && |
1954 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
1955 | return ERR_NOT_PROTO_C; |
1956 | ||
c06ece6b AG |
1957 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
1958 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1959 | if (get_ldev(device)) { |
1960 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
1961 | put_ldev(device); | |
270eb5c9 | 1962 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1963 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 1964 | } |
270eb5c9 | 1965 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 1966 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
1967 | } |
1968 | ||
270eb5c9 | 1969 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 1970 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 1971 | |
cd64397c PR |
1972 | return NO_ERROR; |
1973 | } | |
b411b363 | 1974 | |
44ed167d | 1975 | static enum drbd_ret_code |
270eb5c9 | 1976 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
1977 | { |
1978 | static enum drbd_ret_code rv; | |
c06ece6b | 1979 | struct drbd_peer_device *peer_device; |
44ed167d | 1980 | int i; |
b411b363 | 1981 | |
44ed167d | 1982 | rcu_read_lock(); |
270eb5c9 | 1983 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 1984 | rcu_read_unlock(); |
b411b363 | 1985 | |
bde89a9e | 1986 | /* connection->volumes protected by genl_lock() here */ |
c06ece6b AG |
1987 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
1988 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1989 | if (!device->bitmap) { |
1990 | if (drbd_bm_init(device)) | |
44ed167d | 1991 | return ERR_NOMEM; |
b411b363 PR |
1992 | } |
1993 | } | |
1994 | ||
44ed167d PR |
1995 | return rv; |
1996 | } | |
b411b363 | 1997 | |
0fd0ea06 PR |
1998 | struct crypto { |
1999 | struct crypto_hash *verify_tfm; | |
2000 | struct crypto_hash *csums_tfm; | |
2001 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2002 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2003 | }; |
b411b363 | 2004 | |
0fd0ea06 | 2005 | static int |
4b6ad6d4 | 2006 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2007 | { |
2008 | if (!tfm_name[0]) | |
2009 | return NO_ERROR; | |
b411b363 | 2010 | |
0fd0ea06 PR |
2011 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2012 | if (IS_ERR(*tfm)) { | |
2013 | *tfm = NULL; | |
2014 | return err_alg; | |
b411b363 | 2015 | } |
b411b363 | 2016 | |
0fd0ea06 PR |
2017 | return NO_ERROR; |
2018 | } | |
b411b363 | 2019 | |
0fd0ea06 | 2020 | static enum drbd_ret_code |
270eb5c9 | 2021 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2022 | { |
2023 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2024 | enum drbd_ret_code rv; | |
0fd0ea06 | 2025 | |
270eb5c9 | 2026 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2027 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2028 | if (rv != NO_ERROR) |
2029 | return rv; | |
270eb5c9 | 2030 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2031 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2032 | if (rv != NO_ERROR) |
2033 | return rv; | |
270eb5c9 | 2034 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2035 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2036 | if (rv != NO_ERROR) |
2037 | return rv; | |
270eb5c9 | 2038 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2039 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2040 | new_net_conf->cram_hmac_alg); |
b411b363 | 2041 | |
4b6ad6d4 AG |
2042 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2043 | ERR_AUTH_ALG); | |
b411b363 PR |
2044 | } |
2045 | ||
0fd0ea06 PR |
2046 | return rv; |
2047 | } | |
b411b363 | 2048 | |
0fd0ea06 PR |
2049 | static void free_crypto(struct crypto *crypto) |
2050 | { | |
0fd0ea06 | 2051 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2052 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2053 | crypto_free_hash(crypto->csums_tfm); |
2054 | crypto_free_hash(crypto->verify_tfm); | |
2055 | } | |
b411b363 | 2056 | |
f399002e LE |
2057 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2058 | { | |
2059 | enum drbd_ret_code retcode; | |
bde89a9e | 2060 | struct drbd_connection *connection; |
270eb5c9 | 2061 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2062 | int err; |
2063 | int ovr; /* online verify running */ | |
2064 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2065 | struct crypto crypto = { }; |
b411b363 | 2066 | |
089c075d | 2067 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2068 | if (!adm_ctx.reply_skb) |
2069 | return retcode; | |
2070 | if (retcode != NO_ERROR) | |
2071 | goto out; | |
b411b363 | 2072 | |
bde89a9e | 2073 | connection = adm_ctx.connection; |
b411b363 | 2074 | |
270eb5c9 AG |
2075 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2076 | if (!new_net_conf) { | |
f399002e LE |
2077 | retcode = ERR_NOMEM; |
2078 | goto out; | |
2079 | } | |
b411b363 | 2080 | |
bde89a9e | 2081 | conn_reconfig_start(connection); |
b411b363 | 2082 | |
bde89a9e AG |
2083 | mutex_lock(&connection->data.mutex); |
2084 | mutex_lock(&connection->conf_update); | |
270eb5c9 | 2085 | old_net_conf = connection->net_conf; |
2561b9c1 | 2086 | |
270eb5c9 | 2087 | if (!old_net_conf) { |
f399002e LE |
2088 | drbd_msg_put_info("net conf missing, try connect"); |
2089 | retcode = ERR_INVALID_REQUEST; | |
2561b9c1 PR |
2090 | goto fail; |
2091 | } | |
2092 | ||
270eb5c9 | 2093 | *new_net_conf = *old_net_conf; |
5979e361 | 2094 | if (should_set_defaults(info)) |
270eb5c9 | 2095 | set_net_conf_defaults(new_net_conf); |
f399002e | 2096 | |
270eb5c9 | 2097 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2098 | if (err && err != -ENOMSG) { |
f399002e LE |
2099 | retcode = ERR_MANDATORY_TAG; |
2100 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2101 | goto fail; | |
2561b9c1 | 2102 | } |
b411b363 | 2103 | |
270eb5c9 | 2104 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2105 | if (retcode != NO_ERROR) |
2106 | goto fail; | |
b411b363 | 2107 | |
f399002e | 2108 | /* re-sync running */ |
bde89a9e | 2109 | rsr = conn_resync_running(connection); |
270eb5c9 | 2110 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2111 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2112 | goto fail; |
b411b363 PR |
2113 | } |
2114 | ||
f399002e | 2115 | /* online verify running */ |
bde89a9e | 2116 | ovr = conn_ov_running(connection); |
270eb5c9 | 2117 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2118 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2119 | goto fail; |
f399002e | 2120 | } |
b411b363 | 2121 | |
270eb5c9 | 2122 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2123 | if (retcode != NO_ERROR) |
b411b363 | 2124 | goto fail; |
f399002e | 2125 | |
270eb5c9 | 2126 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2127 | |
2128 | if (!rsr) { | |
bde89a9e AG |
2129 | crypto_free_hash(connection->csums_tfm); |
2130 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2131 | crypto.csums_tfm = NULL; |
f399002e LE |
2132 | } |
2133 | if (!ovr) { | |
bde89a9e AG |
2134 | crypto_free_hash(connection->verify_tfm); |
2135 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2136 | crypto.verify_tfm = NULL; |
b411b363 PR |
2137 | } |
2138 | ||
bde89a9e AG |
2139 | crypto_free_hash(connection->integrity_tfm); |
2140 | connection->integrity_tfm = crypto.integrity_tfm; | |
2141 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2142 | /* Do this without trying to take connection->data.mutex again. */ | |
2143 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2144 | |
bde89a9e AG |
2145 | crypto_free_hash(connection->cram_hmac_tfm); |
2146 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2147 | |
bde89a9e AG |
2148 | mutex_unlock(&connection->conf_update); |
2149 | mutex_unlock(&connection->data.mutex); | |
91fd4dad | 2150 | synchronize_rcu(); |
270eb5c9 | 2151 | kfree(old_net_conf); |
91fd4dad | 2152 | |
bde89a9e AG |
2153 | if (connection->cstate >= C_WF_REPORT_PARAMS) |
2154 | drbd_send_sync_param(minor_to_device(conn_lowest_minor(connection))); | |
f399002e | 2155 | |
91fd4dad PR |
2156 | goto done; |
2157 | ||
b411b363 | 2158 | fail: |
bde89a9e AG |
2159 | mutex_unlock(&connection->conf_update); |
2160 | mutex_unlock(&connection->data.mutex); | |
0fd0ea06 | 2161 | free_crypto(&crypto); |
270eb5c9 | 2162 | kfree(new_net_conf); |
91fd4dad | 2163 | done: |
bde89a9e | 2164 | conn_reconfig_done(connection); |
f399002e LE |
2165 | out: |
2166 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2167 | return 0; |
2168 | } | |
2169 | ||
3b98c0c2 | 2170 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2171 | { |
c06ece6b | 2172 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2173 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2174 | struct crypto crypto = { }; |
77c556f6 | 2175 | struct drbd_resource *resource; |
bde89a9e | 2176 | struct drbd_connection *connection; |
3b98c0c2 LE |
2177 | enum drbd_ret_code retcode; |
2178 | int i; | |
2179 | int err; | |
b411b363 | 2180 | |
44e52cfa | 2181 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2182 | |
3b98c0c2 LE |
2183 | if (!adm_ctx.reply_skb) |
2184 | return retcode; | |
2185 | if (retcode != NO_ERROR) | |
2186 | goto out; | |
089c075d AG |
2187 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
2188 | drbd_msg_put_info("connection endpoint(s) missing"); | |
2189 | retcode = ERR_INVALID_REQUEST; | |
2190 | goto out; | |
2191 | } | |
b411b363 | 2192 | |
089c075d AG |
2193 | /* No need for _rcu here. All reconfiguration is |
2194 | * strictly serialized on genl_lock(). We are protected against | |
2195 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2196 | for_each_resource(resource, &drbd_resources) { |
2197 | for_each_connection(connection, resource) { | |
2198 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2199 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2200 | connection->my_addr_len)) { | |
2201 | retcode = ERR_LOCAL_ADDR; | |
2202 | goto out; | |
2203 | } | |
b411b363 | 2204 | |
77c556f6 AG |
2205 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2206 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2207 | connection->peer_addr_len)) { | |
2208 | retcode = ERR_PEER_ADDR; | |
2209 | goto out; | |
2210 | } | |
089c075d | 2211 | } |
b411b363 PR |
2212 | } |
2213 | ||
bde89a9e AG |
2214 | connection = adm_ctx.connection; |
2215 | conn_reconfig_start(connection); | |
b411b363 | 2216 | |
bde89a9e | 2217 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2218 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2219 | goto fail; |
2220 | } | |
2221 | ||
a209b4ae | 2222 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2223 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2224 | if (!new_net_conf) { | |
b411b363 | 2225 | retcode = ERR_NOMEM; |
b411b363 PR |
2226 | goto fail; |
2227 | } | |
2228 | ||
270eb5c9 | 2229 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2230 | |
270eb5c9 | 2231 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2232 | if (err && err != -ENOMSG) { |
b411b363 | 2233 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2234 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2235 | goto fail; |
2236 | } | |
2237 | ||
270eb5c9 | 2238 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2239 | if (retcode != NO_ERROR) |
422028b1 | 2240 | goto fail; |
b411b363 | 2241 | |
270eb5c9 | 2242 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2243 | if (retcode != NO_ERROR) |
2244 | goto fail; | |
b411b363 | 2245 | |
270eb5c9 | 2246 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2247 | |
bde89a9e | 2248 | conn_flush_workqueue(connection); |
b411b363 | 2249 | |
bde89a9e | 2250 | mutex_lock(&connection->conf_update); |
270eb5c9 AG |
2251 | old_net_conf = connection->net_conf; |
2252 | if (old_net_conf) { | |
b411b363 | 2253 | retcode = ERR_NET_CONFIGURED; |
bde89a9e | 2254 | mutex_unlock(&connection->conf_update); |
b411b363 PR |
2255 | goto fail; |
2256 | } | |
270eb5c9 | 2257 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2258 | |
bde89a9e AG |
2259 | conn_free_crypto(connection); |
2260 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2261 | connection->integrity_tfm = crypto.integrity_tfm; | |
2262 | connection->csums_tfm = crypto.csums_tfm; | |
2263 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2264 | |
bde89a9e AG |
2265 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2266 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2267 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2268 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2269 | |
bde89a9e | 2270 | mutex_unlock(&connection->conf_update); |
b411b363 | 2271 | |
695d08fa | 2272 | rcu_read_lock(); |
c06ece6b AG |
2273 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2274 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2275 | device->send_cnt = 0; |
2276 | device->recv_cnt = 0; | |
b411b363 | 2277 | } |
695d08fa | 2278 | rcu_read_unlock(); |
b411b363 | 2279 | |
bde89a9e | 2280 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2281 | |
bde89a9e | 2282 | conn_reconfig_done(connection); |
3b98c0c2 | 2283 | drbd_adm_finish(info, retcode); |
b411b363 | 2284 | return 0; |
b411b363 | 2285 | |
b411b363 | 2286 | fail: |
0fd0ea06 | 2287 | free_crypto(&crypto); |
270eb5c9 | 2288 | kfree(new_net_conf); |
b411b363 | 2289 | |
bde89a9e | 2290 | conn_reconfig_done(connection); |
3b98c0c2 LE |
2291 | out: |
2292 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2293 | return 0; |
2294 | } | |
2295 | ||
bde89a9e | 2296 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2297 | { |
2298 | enum drbd_state_rv rv; | |
85f75dd7 | 2299 | |
bde89a9e | 2300 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2301 | force ? CS_HARD : 0); |
85f75dd7 LE |
2302 | |
2303 | switch (rv) { | |
2304 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2305 | break; |
85f75dd7 LE |
2306 | case SS_ALREADY_STANDALONE: |
2307 | return SS_SUCCESS; | |
2308 | case SS_PRIMARY_NOP: | |
2309 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2310 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2311 | |
2312 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2313 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2314 | |
85f75dd7 LE |
2315 | break; |
2316 | case SS_CW_FAILED_BY_PEER: | |
2317 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2318 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2319 | disk, D_OUTDATED), 0); |
2320 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2321 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2322 | CS_HARD); |
b411b363 | 2323 | } |
85f75dd7 LE |
2324 | break; |
2325 | default:; | |
2326 | /* no special handling necessary */ | |
2327 | } | |
2328 | ||
f3dfa40a LE |
2329 | if (rv >= SS_SUCCESS) { |
2330 | enum drbd_state_rv rv2; | |
2331 | /* No one else can reconfigure the network while I am here. | |
2332 | * The state handling only uses drbd_thread_stop_nowait(), | |
2333 | * we want to really wait here until the receiver is no more. | |
2334 | */ | |
9693da23 | 2335 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2336 | |
2337 | /* Race breaker. This additional state change request may be | |
2338 | * necessary, if this was a forced disconnect during a receiver | |
2339 | * restart. We may have "killed" the receiver thread just | |
2340 | * after drbdd_init() returned. Typically, we should be | |
2341 | * C_STANDALONE already, now, and this becomes a no-op. | |
2342 | */ | |
bde89a9e | 2343 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2344 | CS_VERBOSE | CS_HARD); |
2345 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2346 | drbd_err(connection, |
f3dfa40a LE |
2347 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2348 | rv2); | |
b411b363 | 2349 | } |
85f75dd7 LE |
2350 | return rv; |
2351 | } | |
b411b363 | 2352 | |
3b98c0c2 | 2353 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2354 | { |
3b98c0c2 | 2355 | struct disconnect_parms parms; |
bde89a9e | 2356 | struct drbd_connection *connection; |
85f75dd7 | 2357 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2358 | enum drbd_ret_code retcode; |
2359 | int err; | |
2561b9c1 | 2360 | |
089c075d | 2361 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2362 | if (!adm_ctx.reply_skb) |
2363 | return retcode; | |
2364 | if (retcode != NO_ERROR) | |
2561b9c1 | 2365 | goto fail; |
b411b363 | 2366 | |
bde89a9e | 2367 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2368 | memset(&parms, 0, sizeof(parms)); |
2369 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2370 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2371 | if (err) { |
2372 | retcode = ERR_MANDATORY_TAG; | |
2373 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
b411b363 PR |
2374 | goto fail; |
2375 | } | |
2376 | } | |
2377 | ||
bde89a9e | 2378 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2379 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2380 | retcode = rv; /* FIXME: Type mismatch. */ |
2381 | else | |
2382 | retcode = NO_ERROR; | |
b411b363 | 2383 | fail: |
3b98c0c2 | 2384 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2385 | return 0; |
2386 | } | |
2387 | ||
b30ab791 | 2388 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2389 | { |
2390 | int iass; /* I am sync source */ | |
2391 | ||
d0180171 | 2392 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2393 | if (device->state.role != device->state.peer) |
2394 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2395 | else |
a6b32bc3 | 2396 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2397 | |
2398 | if (iass) | |
b30ab791 | 2399 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2400 | else |
b30ab791 | 2401 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2402 | } |
2403 | ||
3b98c0c2 | 2404 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2405 | { |
daeda1cc | 2406 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2407 | struct resize_parms rs; |
b30ab791 | 2408 | struct drbd_device *device; |
3b98c0c2 | 2409 | enum drbd_ret_code retcode; |
b411b363 | 2410 | enum determine_dev_size dd; |
d752b269 | 2411 | bool change_al_layout = false; |
6495d2c6 | 2412 | enum dds_flags ddsf; |
daeda1cc | 2413 | sector_t u_size; |
3b98c0c2 | 2414 | int err; |
b411b363 | 2415 | |
3b98c0c2 LE |
2416 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2417 | if (!adm_ctx.reply_skb) | |
2418 | return retcode; | |
2419 | if (retcode != NO_ERROR) | |
b411b363 | 2420 | goto fail; |
3b98c0c2 | 2421 | |
b30ab791 AG |
2422 | device = adm_ctx.device; |
2423 | if (!get_ldev(device)) { | |
d752b269 PR |
2424 | retcode = ERR_NO_DISK; |
2425 | goto fail; | |
2426 | } | |
2427 | ||
3b98c0c2 | 2428 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2429 | rs.al_stripes = device->ldev->md.al_stripes; |
2430 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2431 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2432 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2433 | if (err) { |
3b98c0c2 LE |
2434 | retcode = ERR_MANDATORY_TAG; |
2435 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
d752b269 | 2436 | goto fail_ldev; |
b411b363 PR |
2437 | } |
2438 | } | |
2439 | ||
b30ab791 | 2440 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2441 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2442 | goto fail_ldev; |
b411b363 | 2443 | } |
b411b363 | 2444 | |
b30ab791 AG |
2445 | if (device->state.role == R_SECONDARY && |
2446 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2447 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2448 | goto fail_ldev; |
b411b363 | 2449 | } |
b411b363 | 2450 | |
a6b32bc3 | 2451 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2452 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2453 | goto fail_ldev; |
6495d2c6 PR |
2454 | } |
2455 | ||
daeda1cc | 2456 | rcu_read_lock(); |
b30ab791 | 2457 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2458 | rcu_read_unlock(); |
2459 | if (u_size != (sector_t)rs.resize_size) { | |
2460 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2461 | if (!new_disk_conf) { | |
778f271d | 2462 | retcode = ERR_NOMEM; |
9bcd2521 | 2463 | goto fail_ldev; |
778f271d PR |
2464 | } |
2465 | } | |
2466 | ||
b30ab791 AG |
2467 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2468 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2469 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2470 | ||
2471 | if (al_size_k > (16 * 1024 * 1024)) { | |
2472 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2473 | goto fail_ldev; | |
2474 | } | |
2475 | ||
2476 | if (al_size_k < MD_32kB_SECT/2) { | |
2477 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2478 | goto fail_ldev; | |
2479 | } | |
2480 | ||
b30ab791 | 2481 | if (device->state.conn != C_CONNECTED) { |
d752b269 PR |
2482 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2483 | goto fail_ldev; | |
2484 | } | |
2485 | ||
2486 | change_al_layout = true; | |
2487 | } | |
2488 | ||
b30ab791 AG |
2489 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2490 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2491 | |
daeda1cc | 2492 | if (new_disk_conf) { |
a6b32bc3 | 2493 | mutex_lock(&first_peer_device(device)->connection->conf_update); |
b30ab791 | 2494 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2495 | *new_disk_conf = *old_disk_conf; |
2496 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2497 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
a6b32bc3 | 2498 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
daeda1cc PR |
2499 | synchronize_rcu(); |
2500 | kfree(old_disk_conf); | |
b411b363 PR |
2501 | } |
2502 | ||
6495d2c6 | 2503 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2504 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2505 | drbd_md_sync(device); | |
2506 | put_ldev(device); | |
e96c9633 | 2507 | if (dd == DS_ERROR) { |
b411b363 PR |
2508 | retcode = ERR_NOMEM_BITMAP; |
2509 | goto fail; | |
d752b269 PR |
2510 | } else if (dd == DS_ERROR_SPACE_MD) { |
2511 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2512 | goto fail; | |
2513 | } else if (dd == DS_ERROR_SHRINK) { | |
2514 | retcode = ERR_IMPLICIT_SHRINK; | |
2515 | goto fail; | |
b411b363 | 2516 | } |
778f271d | 2517 | |
b30ab791 | 2518 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2519 | if (dd == DS_GREW) |
b30ab791 | 2520 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2521 | |
b30ab791 AG |
2522 | drbd_send_uuids(device); |
2523 | drbd_send_sizes(device, 1, ddsf); | |
778f271d PR |
2524 | } |
2525 | ||
b411b363 | 2526 | fail: |
3b98c0c2 | 2527 | drbd_adm_finish(info, retcode); |
b411b363 | 2528 | return 0; |
b411b363 | 2529 | |
9bcd2521 | 2530 | fail_ldev: |
b30ab791 | 2531 | put_ldev(device); |
9bcd2521 | 2532 | goto fail; |
b411b363 | 2533 | } |
b411b363 | 2534 | |
f399002e | 2535 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2536 | { |
3b98c0c2 | 2537 | enum drbd_ret_code retcode; |
b57a1e27 | 2538 | struct res_opts res_opts; |
f399002e | 2539 | int err; |
b411b363 | 2540 | |
44e52cfa | 2541 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2542 | if (!adm_ctx.reply_skb) |
2543 | return retcode; | |
2544 | if (retcode != NO_ERROR) | |
2545 | goto fail; | |
b411b363 | 2546 | |
eb6bea67 | 2547 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2548 | if (should_set_defaults(info)) |
b966b5dd | 2549 | set_res_opts_defaults(&res_opts); |
b411b363 | 2550 | |
b57a1e27 | 2551 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2552 | if (err && err != -ENOMSG) { |
b411b363 | 2553 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2554 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2555 | goto fail; |
2556 | } | |
2557 | ||
eb6bea67 | 2558 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2559 | if (err) { |
2560 | retcode = ERR_INVALID_REQUEST; | |
2561 | if (err == -ENOMEM) | |
2562 | retcode = ERR_NOMEM; | |
b411b363 PR |
2563 | } |
2564 | ||
b411b363 | 2565 | fail: |
3b98c0c2 | 2566 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2567 | return 0; |
2568 | } | |
2569 | ||
3b98c0c2 | 2570 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2571 | { |
b30ab791 | 2572 | struct drbd_device *device; |
3b98c0c2 LE |
2573 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2574 | ||
2575 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2576 | if (!adm_ctx.reply_skb) | |
2577 | return retcode; | |
2578 | if (retcode != NO_ERROR) | |
2579 | goto out; | |
2580 | ||
b30ab791 | 2581 | device = adm_ctx.device; |
b411b363 | 2582 | |
194bfb32 | 2583 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2584 | * resync just being finished, wait for it before requesting a new resync. |
2585 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2586 | drbd_suspend_io(device); |
2587 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
2588 | drbd_flush_workqueue(device); | |
194bfb32 | 2589 | |
0b2dafcd PR |
2590 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2591 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2592 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2593 | */ |
b30ab791 AG |
2594 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2595 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2596 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2597 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2598 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2599 | retcode = ERR_IO_MD_DISK; | |
2600 | } | |
2601 | } else | |
b30ab791 AG |
2602 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2603 | drbd_resume_io(device); | |
b411b363 | 2604 | |
3b98c0c2 LE |
2605 | out: |
2606 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2607 | return 0; |
2608 | } | |
2609 | ||
3b98c0c2 LE |
2610 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2611 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2612 | { |
3b98c0c2 | 2613 | enum drbd_ret_code retcode; |
194bfb32 | 2614 | |
3b98c0c2 LE |
2615 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2616 | if (!adm_ctx.reply_skb) | |
2617 | return retcode; | |
2618 | if (retcode != NO_ERROR) | |
2619 | goto out; | |
b411b363 | 2620 | |
b30ab791 | 2621 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
3b98c0c2 LE |
2622 | out: |
2623 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2624 | return 0; |
2625 | } | |
2626 | ||
b30ab791 | 2627 | static int drbd_bmio_set_susp_al(struct drbd_device *device) |
0778286a PR |
2628 | { |
2629 | int rv; | |
2630 | ||
b30ab791 AG |
2631 | rv = drbd_bmio_set_n_write(device); |
2632 | drbd_suspend_al(device); | |
0778286a PR |
2633 | return rv; |
2634 | } | |
2635 | ||
3b98c0c2 | 2636 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2637 | { |
25b0d6c8 | 2638 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2639 | struct drbd_device *device; |
25b0d6c8 PR |
2640 | |
2641 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2642 | if (!adm_ctx.reply_skb) | |
2643 | return retcode; | |
2644 | if (retcode != NO_ERROR) | |
2645 | goto out; | |
2646 | ||
b30ab791 | 2647 | device = adm_ctx.device; |
b411b363 | 2648 | |
194bfb32 | 2649 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2650 | * resync just being finished, wait for it before requesting a new resync. |
2651 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2652 | drbd_suspend_io(device); |
2653 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
2654 | drbd_flush_workqueue(device); | |
194bfb32 | 2655 | |
0b2dafcd PR |
2656 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2657 | * in the bitmap. Otherwise, try to start a resync handshake | |
2658 | * as sync source for full sync. | |
2659 | */ | |
b30ab791 | 2660 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2661 | /* The peer will get a resync upon connect anyways. Just make that |
2662 | into a full resync. */ | |
b30ab791 | 2663 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2664 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2665 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2666 | "set_n_write from invalidate_peer", |
2667 | BM_LOCKED_SET_ALLOWED)) | |
2668 | retcode = ERR_IO_MD_DISK; | |
2669 | } | |
2670 | } else | |
b30ab791 AG |
2671 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2672 | drbd_resume_io(device); | |
b411b363 | 2673 | |
25b0d6c8 PR |
2674 | out: |
2675 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2676 | return 0; |
2677 | } | |
2678 | ||
3b98c0c2 | 2679 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2680 | { |
3b98c0c2 | 2681 | enum drbd_ret_code retcode; |
b411b363 | 2682 | |
3b98c0c2 LE |
2683 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2684 | if (!adm_ctx.reply_skb) | |
2685 | return retcode; | |
2686 | if (retcode != NO_ERROR) | |
2687 | goto out; | |
b411b363 | 2688 | |
b30ab791 | 2689 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 LE |
2690 | retcode = ERR_PAUSE_IS_SET; |
2691 | out: | |
2692 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2693 | return 0; |
2694 | } | |
2695 | ||
3b98c0c2 | 2696 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2697 | { |
da9fbc27 | 2698 | union drbd_dev_state s; |
3b98c0c2 LE |
2699 | enum drbd_ret_code retcode; |
2700 | ||
2701 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2702 | if (!adm_ctx.reply_skb) | |
2703 | return retcode; | |
2704 | if (retcode != NO_ERROR) | |
2705 | goto out; | |
b411b363 | 2706 | |
b30ab791 AG |
2707 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2708 | s = adm_ctx.device->state; | |
cd88d030 PR |
2709 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2710 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2711 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2712 | } else { | |
2713 | retcode = ERR_PAUSE_IS_CLEAR; | |
2714 | } | |
2715 | } | |
b411b363 | 2716 | |
3b98c0c2 LE |
2717 | out: |
2718 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2719 | return 0; |
2720 | } | |
2721 | ||
3b98c0c2 | 2722 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2723 | { |
3b98c0c2 | 2724 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2725 | } |
2726 | ||
3b98c0c2 | 2727 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2728 | { |
b30ab791 | 2729 | struct drbd_device *device; |
3b98c0c2 LE |
2730 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2731 | ||
2732 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2733 | if (!adm_ctx.reply_skb) | |
2734 | return retcode; | |
2735 | if (retcode != NO_ERROR) | |
2736 | goto out; | |
2737 | ||
b30ab791 AG |
2738 | device = adm_ctx.device; |
2739 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2740 | drbd_uuid_new_current(device); | |
2741 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2742 | } |
b30ab791 AG |
2743 | drbd_suspend_io(device); |
2744 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2745 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2746 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2747 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2748 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2749 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2750 | } |
b30ab791 | 2751 | drbd_resume_io(device); |
265be2d0 | 2752 | |
3b98c0c2 LE |
2753 | out: |
2754 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2755 | return 0; |
2756 | } | |
2757 | ||
3b98c0c2 | 2758 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2759 | { |
3b98c0c2 | 2760 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2761 | } |
2762 | ||
251b8f8e AG |
2763 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2764 | struct drbd_resource *resource, | |
2765 | struct drbd_connection *connection, | |
2766 | struct drbd_device *device) | |
b411b363 | 2767 | { |
543cc10b LE |
2768 | struct nlattr *nla; |
2769 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2770 | if (!nla) | |
2771 | goto nla_put_failure; | |
251b8f8e AG |
2772 | if (device && |
2773 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2774 | goto nla_put_failure; |
77c556f6 | 2775 | if (nla_put_string(skb, T_ctx_resource_name, connection->resource->name)) |
26ec9287 | 2776 | goto nla_put_failure; |
251b8f8e AG |
2777 | if (connection) { |
2778 | if (connection->my_addr_len && | |
2779 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2780 | goto nla_put_failure; | |
2781 | if (connection->peer_addr_len && | |
2782 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2783 | goto nla_put_failure; | |
2784 | } | |
543cc10b LE |
2785 | nla_nest_end(skb, nla); |
2786 | return 0; | |
b411b363 | 2787 | |
543cc10b LE |
2788 | nla_put_failure: |
2789 | if (nla) | |
2790 | nla_nest_cancel(skb, nla); | |
2791 | return -EMSGSIZE; | |
2792 | } | |
b411b363 | 2793 | |
251b8f8e AG |
2794 | /* |
2795 | * Return the connection of @resource if @resource has exactly one connection. | |
2796 | */ | |
2797 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2798 | { | |
2799 | struct list_head *connections = &resource->connections; | |
2800 | ||
2801 | if (list_empty(connections) || connections->next->next != connections) | |
2802 | return NULL; | |
2803 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2804 | } | |
2805 | ||
2806 | int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, | |
3b98c0c2 | 2807 | const struct sib_info *sib) |
b411b363 | 2808 | { |
251b8f8e | 2809 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2810 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2811 | struct nlattr *nla; | |
2812 | int got_ldev; | |
3b98c0c2 LE |
2813 | int err = 0; |
2814 | int exclude_sensitive; | |
2815 | ||
2816 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2817 | * to. So we better exclude_sensitive information. | |
2818 | * | |
2819 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2820 | * in the context of the requesting user process. Exclude sensitive | |
2821 | * information, unless current has superuser. | |
2822 | * | |
2823 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2824 | * relies on the current implementation of netlink_dump(), which | |
2825 | * executes the dump callback successively from netlink_recvmsg(), | |
2826 | * always in the context of the receiving process */ | |
2827 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2828 | ||
b30ab791 | 2829 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2830 | |
2831 | /* We need to add connection name and volume number information still. | |
2832 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 2833 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 2834 | goto nla_put_failure; |
3b98c0c2 | 2835 | |
eb6bea67 | 2836 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
2837 | goto nla_put_failure; |
2838 | ||
daeda1cc | 2839 | rcu_read_lock(); |
f9eb7bf4 AG |
2840 | if (got_ldev) { |
2841 | struct disk_conf *disk_conf; | |
44ed167d | 2842 | |
b30ab791 | 2843 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2844 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2845 | } | |
2846 | if (!err) { | |
2847 | struct net_conf *nc; | |
2848 | ||
a6b32bc3 | 2849 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2850 | if (nc) |
2851 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2852 | } | |
44ed167d PR |
2853 | rcu_read_unlock(); |
2854 | if (err) | |
2855 | goto nla_put_failure; | |
3b98c0c2 | 2856 | |
3b98c0c2 LE |
2857 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2858 | if (!nla) | |
2859 | goto nla_put_failure; | |
26ec9287 | 2860 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2861 | nla_put_u32(skb, T_current_state, device->state.i) || |
2862 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2863 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2864 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2865 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2866 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2867 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2868 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2869 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2870 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2871 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2872 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2873 | goto nla_put_failure; |
3b98c0c2 LE |
2874 | |
2875 | if (got_ldev) { | |
39a1aa7f | 2876 | int err; |
b411b363 | 2877 | |
b30ab791 AG |
2878 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2879 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2880 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2881 | |
2882 | if (err) | |
2883 | goto nla_put_failure; | |
2884 | ||
b30ab791 AG |
2885 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
2886 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
2887 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 2888 | goto nla_put_failure; |
b30ab791 AG |
2889 | if (C_SYNC_SOURCE <= device->state.conn && |
2890 | C_PAUSED_SYNC_T >= device->state.conn) { | |
2891 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
2892 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 2893 | goto nla_put_failure; |
3b98c0c2 | 2894 | } |
b411b363 | 2895 | } |
b411b363 | 2896 | |
3b98c0c2 LE |
2897 | if (sib) { |
2898 | switch(sib->sib_reason) { | |
2899 | case SIB_SYNC_PROGRESS: | |
2900 | case SIB_GET_STATUS_REPLY: | |
2901 | break; | |
2902 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2903 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2904 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2905 | goto nla_put_failure; | |
3b98c0c2 LE |
2906 | break; |
2907 | case SIB_HELPER_POST: | |
26ec9287 AG |
2908 | if (nla_put_u32(skb, T_helper_exit_code, |
2909 | sib->helper_exit_code)) | |
2910 | goto nla_put_failure; | |
3b98c0c2 LE |
2911 | /* fall through */ |
2912 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2913 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2914 | goto nla_put_failure; | |
3b98c0c2 LE |
2915 | break; |
2916 | } | |
b411b363 | 2917 | } |
3b98c0c2 | 2918 | nla_nest_end(skb, nla); |
b411b363 | 2919 | |
3b98c0c2 LE |
2920 | if (0) |
2921 | nla_put_failure: | |
2922 | err = -EMSGSIZE; | |
2923 | if (got_ldev) | |
b30ab791 | 2924 | put_ldev(device); |
3b98c0c2 | 2925 | return err; |
b411b363 PR |
2926 | } |
2927 | ||
3b98c0c2 | 2928 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2929 | { |
3b98c0c2 LE |
2930 | enum drbd_ret_code retcode; |
2931 | int err; | |
b411b363 | 2932 | |
3b98c0c2 LE |
2933 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2934 | if (!adm_ctx.reply_skb) | |
2935 | return retcode; | |
2936 | if (retcode != NO_ERROR) | |
2937 | goto out; | |
b411b363 | 2938 | |
b30ab791 | 2939 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
2940 | if (err) { |
2941 | nlmsg_free(adm_ctx.reply_skb); | |
2942 | return err; | |
b411b363 | 2943 | } |
3b98c0c2 LE |
2944 | out: |
2945 | drbd_adm_finish(info, retcode); | |
2946 | return 0; | |
b411b363 PR |
2947 | } |
2948 | ||
4b7a530f | 2949 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2950 | { |
b30ab791 | 2951 | struct drbd_device *device; |
3b98c0c2 | 2952 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
2953 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
2954 | struct drbd_resource *resource = NULL; | |
77c556f6 | 2955 | struct drbd_resource *tmp; |
543cc10b LE |
2956 | unsigned volume = cb->args[1]; |
2957 | ||
2958 | /* Open coded, deferred, iteration: | |
77c556f6 | 2959 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
2960 | * connection = "first connection of resource or undefined"; |
2961 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
2962 | * ... |
2963 | * } | |
2964 | * } | |
77c556f6 | 2965 | * where resource is cb->args[0]; |
543cc10b LE |
2966 | * and i is cb->args[1]; |
2967 | * | |
71932efc LE |
2968 | * cb->args[2] indicates if we shall loop over all resources, |
2969 | * or just dump all volumes of a single resource. | |
2970 | * | |
3b98c0c2 LE |
2971 | * This may miss entries inserted after this dump started, |
2972 | * or entries deleted before they are reached. | |
543cc10b | 2973 | * |
b30ab791 | 2974 | * We need to make sure the device won't disappear while |
543cc10b LE |
2975 | * we are looking at it, and revalidate our iterators |
2976 | * on each iteration. | |
2977 | */ | |
b411b363 | 2978 | |
05a10ec7 | 2979 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 2980 | rcu_read_lock(); |
543cc10b | 2981 | /* revalidate iterator position */ |
77c556f6 | 2982 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
2983 | if (pos == NULL) { |
2984 | /* first iteration */ | |
2985 | pos = tmp; | |
77c556f6 | 2986 | resource = pos; |
543cc10b LE |
2987 | break; |
2988 | } | |
2989 | if (tmp == pos) { | |
77c556f6 | 2990 | resource = pos; |
543cc10b LE |
2991 | break; |
2992 | } | |
b411b363 | 2993 | } |
77c556f6 AG |
2994 | if (resource) { |
2995 | next_resource: | |
251b8f8e AG |
2996 | device = idr_get_next(&resource->devices, &volume); |
2997 | if (!device) { | |
77c556f6 AG |
2998 | /* No more volumes to dump on this resource. |
2999 | * Advance resource iterator. */ | |
3000 | pos = list_entry_rcu(resource->resources.next, | |
3001 | struct drbd_resource, resources); | |
3002 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3003 | if (volume != 0) { |
71932efc LE |
3004 | /* If we reached the end of the list, |
3005 | * or only a single resource dump was requested, | |
3006 | * we are done. */ | |
77c556f6 | 3007 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3008 | goto out; |
543cc10b | 3009 | volume = 0; |
77c556f6 AG |
3010 | resource = pos; |
3011 | goto next_resource; | |
543cc10b LE |
3012 | } |
3013 | } | |
3014 | ||
98683650 | 3015 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3016 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3017 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3018 | if (!dh) | |
543cc10b LE |
3019 | goto out; |
3020 | ||
251b8f8e | 3021 | if (!device) { |
bde89a9e | 3022 | /* This is a connection without a single volume. |
367d675d LE |
3023 | * Suprisingly enough, it may have a network |
3024 | * configuration. */ | |
251b8f8e AG |
3025 | struct drbd_connection *connection; |
3026 | ||
543cc10b LE |
3027 | dh->minor = -1U; |
3028 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3029 | connection = the_only_connection(resource); |
3030 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3031 | goto cancel; |
251b8f8e AG |
3032 | if (connection) { |
3033 | struct net_conf *nc; | |
3034 | ||
3035 | nc = rcu_dereference(connection->net_conf); | |
3036 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3037 | goto cancel; | |
3038 | } | |
367d675d | 3039 | goto done; |
543cc10b | 3040 | } |
b411b363 | 3041 | |
0b0ba1ef | 3042 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3043 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3044 | |
b30ab791 | 3045 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3046 | dh->ret_code = NO_ERROR; |
3047 | ||
b30ab791 | 3048 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3049 | cancel: |
3b98c0c2 | 3050 | genlmsg_cancel(skb, dh); |
543cc10b | 3051 | goto out; |
3b98c0c2 | 3052 | } |
367d675d | 3053 | done: |
3b98c0c2 | 3054 | genlmsg_end(skb, dh); |
bde89a9e | 3055 | } |
b411b363 | 3056 | |
543cc10b | 3057 | out: |
c141ebda | 3058 | rcu_read_unlock(); |
543cc10b | 3059 | /* where to start the next iteration */ |
bde89a9e | 3060 | cb->args[0] = (long)pos; |
77c556f6 | 3061 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3062 | |
77c556f6 | 3063 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3064 | * Which will terminate the dump. */ |
3b98c0c2 | 3065 | return skb->len; |
b411b363 PR |
3066 | } |
3067 | ||
71932efc LE |
3068 | /* |
3069 | * Request status of all resources, or of all volumes within a single resource. | |
3070 | * | |
3071 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3072 | * Which means we cannot use the family->attrbuf or other such members, because | |
3073 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3074 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3075 | * | |
3076 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3077 | */ |
71932efc | 3078 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3079 | { |
71932efc LE |
3080 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3081 | struct nlattr *nla; | |
7c3063cc | 3082 | const char *resource_name; |
4bc76048 | 3083 | struct drbd_resource *resource; |
7c3063cc | 3084 | int maxtype; |
71932efc LE |
3085 | |
3086 | /* Is this a followup call? */ | |
3087 | if (cb->args[0]) { | |
3088 | /* ... of a single resource dump, | |
3089 | * and the resource iterator has been advanced already? */ | |
3090 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3091 | return 0; /* DONE. */ | |
3092 | goto dump; | |
3093 | } | |
3094 | ||
3095 | /* First call (from netlink_dump_start). We need to figure out | |
3096 | * which resource(s) the user wants us to dump. */ | |
3097 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3098 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3099 | DRBD_NLA_CFG_CONTEXT); | |
3100 | ||
3101 | /* No explicit context given. Dump all. */ | |
3102 | if (!nla) | |
3103 | goto dump; | |
7c3063cc AG |
3104 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3105 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3106 | if (IS_ERR(nla)) | |
3107 | return PTR_ERR(nla); | |
71932efc LE |
3108 | /* context given, but no name present? */ |
3109 | if (!nla) | |
3110 | return -EINVAL; | |
7c3063cc | 3111 | resource_name = nla_data(nla); |
4bc76048 AG |
3112 | if (!*resource_name) |
3113 | return -ENODEV; | |
3114 | resource = drbd_find_resource(resource_name); | |
3115 | if (!resource) | |
71932efc LE |
3116 | return -ENODEV; |
3117 | ||
4bc76048 | 3118 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3119 | |
71932efc | 3120 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3121 | * only dump this connection. */ |
4bc76048 | 3122 | cb->args[0] = (long)resource; |
71932efc | 3123 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3124 | cb->args[2] = (long)resource; |
71932efc LE |
3125 | |
3126 | dump: | |
3127 | return get_one_status(skb, cb); | |
3128 | } | |
b411b363 | 3129 | |
3b98c0c2 | 3130 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3131 | { |
3b98c0c2 LE |
3132 | enum drbd_ret_code retcode; |
3133 | struct timeout_parms tp; | |
3134 | int err; | |
b411b363 | 3135 | |
3b98c0c2 LE |
3136 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3137 | if (!adm_ctx.reply_skb) | |
3138 | return retcode; | |
3139 | if (retcode != NO_ERROR) | |
3140 | goto out; | |
b411b363 | 3141 | |
3b98c0c2 | 3142 | tp.timeout_type = |
b30ab791 AG |
3143 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3144 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3145 | UT_DEFAULT; |
b411b363 | 3146 | |
3b98c0c2 LE |
3147 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3148 | if (err) { | |
3149 | nlmsg_free(adm_ctx.reply_skb); | |
3150 | return err; | |
3151 | } | |
3152 | out: | |
3153 | drbd_adm_finish(info, retcode); | |
3154 | return 0; | |
b411b363 PR |
3155 | } |
3156 | ||
3b98c0c2 | 3157 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3158 | { |
b30ab791 | 3159 | struct drbd_device *device; |
3b98c0c2 | 3160 | enum drbd_ret_code retcode; |
58ffa580 | 3161 | struct start_ov_parms parms; |
b411b363 | 3162 | |
3b98c0c2 LE |
3163 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3164 | if (!adm_ctx.reply_skb) | |
3165 | return retcode; | |
3166 | if (retcode != NO_ERROR) | |
3167 | goto out; | |
873b0d5f | 3168 | |
b30ab791 | 3169 | device = adm_ctx.device; |
58ffa580 LE |
3170 | |
3171 | /* resume from last known position, if possible */ | |
b30ab791 | 3172 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3173 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3174 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3175 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3176 | if (err) { |
3177 | retcode = ERR_MANDATORY_TAG; | |
3178 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3179 | goto out; | |
3180 | } | |
b411b363 | 3181 | } |
58ffa580 | 3182 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3183 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3184 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3185 | |
3186 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3187 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3188 | drbd_suspend_io(device); |
3189 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3190 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3191 | drbd_resume_io(device); | |
3b98c0c2 LE |
3192 | out: |
3193 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3194 | return 0; |
3195 | } | |
3196 | ||
3197 | ||
3b98c0c2 | 3198 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3199 | { |
b30ab791 | 3200 | struct drbd_device *device; |
3b98c0c2 | 3201 | enum drbd_ret_code retcode; |
b411b363 PR |
3202 | int skip_initial_sync = 0; |
3203 | int err; | |
3b98c0c2 | 3204 | struct new_c_uuid_parms args; |
b411b363 | 3205 | |
3b98c0c2 LE |
3206 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3207 | if (!adm_ctx.reply_skb) | |
3208 | return retcode; | |
3209 | if (retcode != NO_ERROR) | |
3210 | goto out_nolock; | |
b411b363 | 3211 | |
b30ab791 | 3212 | device = adm_ctx.device; |
3b98c0c2 LE |
3213 | memset(&args, 0, sizeof(args)); |
3214 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3215 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3216 | if (err) { |
3217 | retcode = ERR_MANDATORY_TAG; | |
3218 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3219 | goto out_nolock; | |
3220 | } | |
b411b363 PR |
3221 | } |
3222 | ||
b30ab791 | 3223 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3224 | |
b30ab791 | 3225 | if (!get_ldev(device)) { |
b411b363 PR |
3226 | retcode = ERR_NO_DISK; |
3227 | goto out; | |
3228 | } | |
3229 | ||
3230 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3231 | if (device->state.conn == C_CONNECTED && |
3232 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3233 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3234 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3235 | skip_initial_sync = 1; |
b30ab791 | 3236 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3237 | retcode = ERR_CONNECTED; |
3238 | goto out_dec; | |
3239 | } | |
3240 | ||
b30ab791 AG |
3241 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3242 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3243 | |
3244 | if (args.clear_bm) { | |
b30ab791 | 3245 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3246 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3247 | if (err) { |
d0180171 | 3248 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3249 | retcode = ERR_IO_MD_DISK; |
3250 | } | |
3251 | if (skip_initial_sync) { | |
b30ab791 AG |
3252 | drbd_send_uuids_skip_initial_sync(device); |
3253 | _drbd_uuid_set(device, UI_BITMAP, 0); | |
3254 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
a6b32bc3 | 3255 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 3256 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3257 | CS_VERBOSE, NULL); |
a6b32bc3 | 3258 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
b411b363 PR |
3259 | } |
3260 | } | |
3261 | ||
b30ab791 | 3262 | drbd_md_sync(device); |
b411b363 | 3263 | out_dec: |
b30ab791 | 3264 | put_ldev(device); |
b411b363 | 3265 | out: |
b30ab791 | 3266 | mutex_unlock(device->state_mutex); |
3b98c0c2 LE |
3267 | out_nolock: |
3268 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3269 | return 0; |
3270 | } | |
3271 | ||
3b98c0c2 | 3272 | static enum drbd_ret_code |
7c3063cc | 3273 | drbd_check_resource_name(const char *name) |
b411b363 | 3274 | { |
3b98c0c2 | 3275 | if (!name || !name[0]) { |
7c3063cc | 3276 | drbd_msg_put_info("resource name missing"); |
3b98c0c2 | 3277 | return ERR_MANDATORY_TAG; |
b411b363 | 3278 | } |
3b98c0c2 LE |
3279 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3280 | * we must not allow slashes */ | |
3281 | if (strchr(name, '/')) { | |
7c3063cc | 3282 | drbd_msg_put_info("invalid resource name"); |
3b98c0c2 | 3283 | return ERR_INVALID_REQUEST; |
b411b363 | 3284 | } |
3b98c0c2 | 3285 | return NO_ERROR; |
774b3055 | 3286 | } |
b411b363 | 3287 | |
789c1b62 | 3288 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3289 | { |
3b98c0c2 | 3290 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3291 | struct res_opts res_opts; |
3292 | int err; | |
b411b363 | 3293 | |
3b98c0c2 LE |
3294 | retcode = drbd_adm_prepare(skb, info, 0); |
3295 | if (!adm_ctx.reply_skb) | |
3296 | return retcode; | |
3297 | if (retcode != NO_ERROR) | |
3298 | goto out; | |
b411b363 | 3299 | |
afbbfa88 AG |
3300 | set_res_opts_defaults(&res_opts); |
3301 | err = res_opts_from_attrs(&res_opts, info); | |
3302 | if (err && err != -ENOMSG) { | |
3303 | retcode = ERR_MANDATORY_TAG; | |
3304 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3305 | goto out; | |
b411b363 PR |
3306 | } |
3307 | ||
7c3063cc | 3308 | retcode = drbd_check_resource_name(adm_ctx.resource_name); |
3b98c0c2 LE |
3309 | if (retcode != NO_ERROR) |
3310 | goto out; | |
b411b363 | 3311 | |
5c661042 | 3312 | if (adm_ctx.resource) { |
38f19616 LE |
3313 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3314 | retcode = ERR_INVALID_REQUEST; | |
789c1b62 | 3315 | drbd_msg_put_info("resource exists"); |
38f19616 LE |
3316 | } |
3317 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3318 | goto out; |
b411b363 | 3319 | } |
b411b363 | 3320 | |
afbbfa88 | 3321 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3322 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
3323 | out: |
3324 | drbd_adm_finish(info, retcode); | |
3325 | return 0; | |
b411b363 PR |
3326 | } |
3327 | ||
05a10ec7 | 3328 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3329 | { |
3b98c0c2 LE |
3330 | struct drbd_genlmsghdr *dh = info->userhdr; |
3331 | enum drbd_ret_code retcode; | |
b411b363 | 3332 | |
44e52cfa | 3333 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3334 | if (!adm_ctx.reply_skb) |
3335 | return retcode; | |
3336 | if (retcode != NO_ERROR) | |
3337 | goto out; | |
b411b363 | 3338 | |
f2257a56 | 3339 | if (dh->minor > MINORMASK) { |
3b98c0c2 LE |
3340 | drbd_msg_put_info("requested minor out of range"); |
3341 | retcode = ERR_INVALID_REQUEST; | |
3342 | goto out; | |
b411b363 | 3343 | } |
0c8e36d9 | 3344 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
3345 | drbd_msg_put_info("requested volume id out of range"); |
3346 | retcode = ERR_INVALID_REQUEST; | |
3347 | goto out; | |
b411b363 | 3348 | } |
b411b363 | 3349 | |
38f19616 | 3350 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3351 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3352 | if (adm_ctx.device) { |
38f19616 LE |
3353 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
3354 | retcode = ERR_MINOR_EXISTS; | |
3355 | /* else: still NO_ERROR */ | |
3356 | goto out; | |
b411b363 | 3357 | } |
38f19616 | 3358 | |
59515a2e | 3359 | retcode = drbd_create_device(adm_ctx.resource, dh->minor, adm_ctx.volume); |
3b98c0c2 LE |
3360 | out: |
3361 | drbd_adm_finish(info, retcode); | |
3362 | return 0; | |
b411b363 PR |
3363 | } |
3364 | ||
05a10ec7 | 3365 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3366 | { |
b30ab791 AG |
3367 | if (device->state.disk == D_DISKLESS && |
3368 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3369 | * we may want to delete a minor from a live replication group. |
3370 | */ | |
b30ab791 AG |
3371 | device->state.role == R_SECONDARY) { |
3372 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3373 | CS_VERBOSE + CS_WAIT_COMPLETE); |
f82795d6 | 3374 | drbd_delete_device(device); |
85f75dd7 LE |
3375 | return NO_ERROR; |
3376 | } else | |
3377 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3378 | } |
3379 | ||
05a10ec7 | 3380 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3381 | { |
3b98c0c2 | 3382 | enum drbd_ret_code retcode; |
b411b363 | 3383 | |
3b98c0c2 LE |
3384 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3385 | if (!adm_ctx.reply_skb) | |
3386 | return retcode; | |
3387 | if (retcode != NO_ERROR) | |
3388 | goto out; | |
b411b363 | 3389 | |
05a10ec7 | 3390 | retcode = adm_del_minor(adm_ctx.device); |
85f75dd7 LE |
3391 | out: |
3392 | drbd_adm_finish(info, retcode); | |
3393 | return 0; | |
b411b363 PR |
3394 | } |
3395 | ||
85f75dd7 | 3396 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3397 | { |
b6f85ef9 AG |
3398 | struct drbd_resource *resource; |
3399 | struct drbd_connection *connection; | |
3400 | struct drbd_device *device; | |
f3dfa40a | 3401 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3402 | unsigned i; |
b411b363 | 3403 | |
a10f6b8a | 3404 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3405 | if (!adm_ctx.reply_skb) |
3406 | return retcode; | |
3407 | if (retcode != NO_ERROR) | |
3408 | goto out; | |
b411b363 | 3409 | |
b6f85ef9 | 3410 | resource = adm_ctx.resource; |
85f75dd7 | 3411 | /* demote */ |
b6f85ef9 AG |
3412 | for_each_connection(connection, resource) { |
3413 | struct drbd_peer_device *peer_device; | |
3414 | ||
3415 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3416 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3417 | if (retcode < SS_SUCCESS) { | |
3418 | drbd_msg_put_info("failed to demote"); | |
3419 | goto out; | |
3420 | } | |
3421 | } | |
3422 | ||
3423 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3424 | if (retcode < SS_SUCCESS) { |
b6f85ef9 | 3425 | drbd_msg_put_info("failed to disconnect"); |
c141ebda | 3426 | goto out; |
85f75dd7 | 3427 | } |
b411b363 | 3428 | } |
b411b363 | 3429 | |
85f75dd7 | 3430 | /* detach */ |
b6f85ef9 AG |
3431 | idr_for_each_entry(&resource->devices, device, i) { |
3432 | retcode = adm_detach(device, 0); | |
27012382 | 3433 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
85f75dd7 | 3434 | drbd_msg_put_info("failed to detach"); |
c141ebda | 3435 | goto out; |
85f75dd7 LE |
3436 | } |
3437 | } | |
b411b363 | 3438 | |
bde89a9e | 3439 | /* If we reach this, all volumes (of this connection) are Secondary, |
f3dfa40a | 3440 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have |
c141ebda | 3441 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
b6f85ef9 AG |
3442 | for_each_connection(connection, resource) |
3443 | drbd_thread_stop(&connection->worker); | |
b411b363 | 3444 | |
f3dfa40a | 3445 | /* Now, nothing can fail anymore */ |
b411b363 | 3446 | |
85f75dd7 | 3447 | /* delete volumes */ |
b6f85ef9 AG |
3448 | idr_for_each_entry(&resource->devices, device, i) { |
3449 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3450 | if (retcode != NO_ERROR) { |
3451 | /* "can not happen" */ | |
3452 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 | 3453 | goto out; |
85f75dd7 LE |
3454 | } |
3455 | } | |
b411b363 | 3456 | |
b6f85ef9 AG |
3457 | list_del_rcu(&resource->resources); |
3458 | synchronize_rcu(); | |
3459 | drbd_free_resource(resource); | |
3460 | retcode = NO_ERROR; | |
b411b363 | 3461 | |
3b98c0c2 LE |
3462 | out: |
3463 | drbd_adm_finish(info, retcode); | |
3464 | return 0; | |
b411b363 PR |
3465 | } |
3466 | ||
789c1b62 | 3467 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3468 | { |
77c556f6 AG |
3469 | struct drbd_resource *resource; |
3470 | struct drbd_connection *connection; | |
3b98c0c2 | 3471 | enum drbd_ret_code retcode; |
b411b363 | 3472 | |
44e52cfa | 3473 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3474 | if (!adm_ctx.reply_skb) |
3475 | return retcode; | |
3476 | if (retcode != NO_ERROR) | |
3477 | goto out; | |
b411b363 | 3478 | |
77c556f6 AG |
3479 | resource = adm_ctx.resource; |
3480 | for_each_connection(connection, resource) { | |
3481 | if (connection->cstate > C_STANDALONE) { | |
3482 | retcode = ERR_NET_CONFIGURED; | |
3483 | goto out; | |
3484 | } | |
3485 | } | |
3486 | if (!idr_is_empty(&resource->devices)) { | |
789c1b62 | 3487 | retcode = ERR_RES_IN_USE; |
77c556f6 | 3488 | goto out; |
b411b363 PR |
3489 | } |
3490 | ||
77c556f6 AG |
3491 | list_del_rcu(&resource->resources); |
3492 | for_each_connection(connection, resource) | |
3493 | drbd_thread_stop(&connection->worker); | |
3494 | synchronize_rcu(); | |
3495 | drbd_free_resource(resource); | |
3496 | retcode = NO_ERROR; | |
3b98c0c2 LE |
3497 | out: |
3498 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3499 | return 0; |
3500 | } | |
3501 | ||
b30ab791 | 3502 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3503 | { |
3b98c0c2 LE |
3504 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3505 | struct sk_buff *msg; | |
3506 | struct drbd_genlmsghdr *d_out; | |
3507 | unsigned seq; | |
3508 | int err = -ENOMEM; | |
3509 | ||
ef86b779 | 3510 | if (sib->sib_reason == SIB_SYNC_PROGRESS) { |
b30ab791 AG |
3511 | if (time_after(jiffies, device->rs_last_bcast + HZ)) |
3512 | device->rs_last_bcast = jiffies; | |
ef86b779 PR |
3513 | else |
3514 | return; | |
3515 | } | |
b411b363 | 3516 | |
3b98c0c2 LE |
3517 | seq = atomic_inc_return(&drbd_genl_seq); |
3518 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3519 | if (!msg) | |
3520 | goto failed; | |
3521 | ||
3522 | err = -EMSGSIZE; | |
3523 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3524 | if (!d_out) /* cannot happen, but anyways. */ | |
3525 | goto nla_put_failure; | |
b30ab791 | 3526 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3527 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3528 | |
b30ab791 | 3529 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3530 | goto nla_put_failure; |
3531 | genlmsg_end(msg, d_out); | |
3532 | err = drbd_genl_multicast_events(msg, 0); | |
3533 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3534 | if (err && err != -ESRCH) | |
3535 | goto failed; | |
b411b363 | 3536 | |
3b98c0c2 | 3537 | return; |
b411b363 | 3538 | |
3b98c0c2 LE |
3539 | nla_put_failure: |
3540 | nlmsg_free(msg); | |
3541 | failed: | |
d0180171 | 3542 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3543 | "Event seq:%u sib_reason:%u\n", |
3544 | err, seq, sib->sib_reason); | |
b411b363 | 3545 | } |