]>
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 | 37 | #include <asm/unaligned.h> |
b411b363 | 38 | #include <linux/drbd_limits.h> |
87f7be4c | 39 | #include <linux/kthread.h> |
b411b363 | 40 | |
3b98c0c2 LE |
41 | #include <net/genetlink.h> |
42 | ||
43 | /* .doit */ | |
44 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
45 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | ||
05a10ec7 AG |
47 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
48 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 49 | |
789c1b62 AG |
50 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
51 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 52 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
53 | |
54 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
55 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 56 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
57 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
58 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
62 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 71 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
72 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
73 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
74 | /* .dumpit */ | |
75 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
76 | ||
77 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 78 | #include "drbd_nla.h" |
3b98c0c2 LE |
79 | #include <linux/genl_magic_func.h> |
80 | ||
81 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
82 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
83 | ||
3b98c0c2 LE |
84 | /* Configuration is strictly serialized, because generic netlink message |
85 | * processing is strictly serialized by the genl_lock(). | |
86 | * Which means we can use one static global drbd_config_context struct. | |
87 | */ | |
88 | static struct drbd_config_context { | |
89 | /* assigned from drbd_genlmsghdr */ | |
90 | unsigned int minor; | |
91 | /* assigned from request attributes, if present */ | |
92 | unsigned int volume; | |
93 | #define VOLUME_UNSPECIFIED (-1U) | |
94 | /* pointer into the request skb, | |
95 | * limited lifetime! */ | |
7c3063cc | 96 | char *resource_name; |
089c075d AG |
97 | struct nlattr *my_addr; |
98 | struct nlattr *peer_addr; | |
3b98c0c2 LE |
99 | |
100 | /* reply buffer */ | |
101 | struct sk_buff *reply_skb; | |
102 | /* pointer into reply buffer */ | |
103 | struct drbd_genlmsghdr *reply_dh; | |
104 | /* resolved from attributes, if possible */ | |
b30ab791 | 105 | struct drbd_device *device; |
4bc76048 | 106 | struct drbd_resource *resource; |
bde89a9e | 107 | struct drbd_connection *connection; |
3b98c0c2 LE |
108 | } adm_ctx; |
109 | ||
110 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
111 | { | |
112 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
113 | if (genlmsg_reply(skb, info)) | |
114 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 115 | } |
3b98c0c2 LE |
116 | |
117 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
118 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 119 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
120 | { |
121 | struct sk_buff *skb = adm_ctx.reply_skb; | |
122 | struct nlattr *nla; | |
123 | int err = -EMSGSIZE; | |
124 | ||
125 | if (!info || !info[0]) | |
126 | return 0; | |
127 | ||
128 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
129 | if (!nla) | |
130 | return err; | |
131 | ||
132 | err = nla_put_string(skb, T_info_text, info); | |
133 | if (err) { | |
134 | nla_nest_cancel(skb, nla); | |
135 | return err; | |
136 | } else | |
137 | nla_nest_end(skb, nla); | |
138 | return 0; | |
b411b363 PR |
139 | } |
140 | ||
3b98c0c2 LE |
141 | /* This would be a good candidate for a "pre_doit" hook, |
142 | * and per-family private info->pointers. | |
143 | * But we need to stay compatible with older kernels. | |
144 | * If it returns successfully, adm_ctx members are valid. | |
145 | */ | |
146 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 147 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 148 | #define DRBD_ADM_NEED_CONNECTION 4 |
3b98c0c2 LE |
149 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, |
150 | unsigned flags) | |
151 | { | |
152 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
153 | const u8 cmd = info->genlhdr->cmd; | |
154 | int err; | |
155 | ||
156 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
157 | ||
158 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 159 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
160 | return -EPERM; |
161 | ||
162 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
1e2a2551 AG |
163 | if (!adm_ctx.reply_skb) { |
164 | err = -ENOMEM; | |
3b98c0c2 | 165 | goto fail; |
1e2a2551 | 166 | } |
3b98c0c2 LE |
167 | |
168 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
169 | info, &drbd_genl_family, 0, cmd); | |
170 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
171 | * but anyways */ | |
1e2a2551 AG |
172 | if (!adm_ctx.reply_dh) { |
173 | err = -ENOMEM; | |
3b98c0c2 | 174 | goto fail; |
1e2a2551 | 175 | } |
3b98c0c2 LE |
176 | |
177 | adm_ctx.reply_dh->minor = d_in->minor; | |
178 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
179 | ||
089c075d | 180 | adm_ctx.volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
181 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
182 | struct nlattr *nla; | |
183 | /* parse and validate only */ | |
f399002e | 184 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
185 | if (err) |
186 | goto fail; | |
187 | ||
188 | /* It was present, and valid, | |
189 | * copy it over to the reply skb. */ | |
190 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
191 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
192 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
193 | if (err) | |
194 | goto fail; | |
195 | ||
196 | /* and assign stuff to the global adm_ctx */ | |
197 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
089c075d AG |
198 | if (nla) |
199 | adm_ctx.volume = nla_get_u32(nla); | |
7c3063cc | 200 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 201 | if (nla) |
7c3063cc | 202 | adm_ctx.resource_name = nla_data(nla); |
089c075d AG |
203 | adm_ctx.my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; |
204 | adm_ctx.peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
205 | if ((adm_ctx.my_addr && | |
bde89a9e | 206 | nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.connection->my_addr)) || |
089c075d | 207 | (adm_ctx.peer_addr && |
bde89a9e | 208 | nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.connection->peer_addr))) { |
089c075d AG |
209 | err = -EINVAL; |
210 | goto fail; | |
211 | } | |
212 | } | |
3b98c0c2 LE |
213 | |
214 | adm_ctx.minor = d_in->minor; | |
b30ab791 | 215 | adm_ctx.device = minor_to_device(d_in->minor); |
4bc76048 AG |
216 | if (adm_ctx.resource_name) { |
217 | adm_ctx.resource = drbd_find_resource(adm_ctx.resource_name); | |
4bc76048 | 218 | } |
3b98c0c2 | 219 | |
b30ab791 | 220 | if (!adm_ctx.device && (flags & DRBD_ADM_NEED_MINOR)) { |
3b98c0c2 LE |
221 | drbd_msg_put_info("unknown minor"); |
222 | return ERR_MINOR_INVALID; | |
223 | } | |
4bc76048 | 224 | if (!adm_ctx.resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
44e52cfa | 225 | drbd_msg_put_info("unknown resource"); |
a10f6b8a AG |
226 | if (adm_ctx.resource_name) |
227 | return ERR_RES_NOT_KNOWN; | |
3b98c0c2 LE |
228 | return ERR_INVALID_REQUEST; |
229 | } | |
230 | ||
089c075d | 231 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
3ab706fe | 232 | if (adm_ctx.resource) { |
089c075d AG |
233 | drbd_msg_put_info("no resource name expected"); |
234 | return ERR_INVALID_REQUEST; | |
235 | } | |
b30ab791 | 236 | if (adm_ctx.device) { |
089c075d AG |
237 | drbd_msg_put_info("no minor number expected"); |
238 | return ERR_INVALID_REQUEST; | |
239 | } | |
240 | if (adm_ctx.my_addr && adm_ctx.peer_addr) | |
bde89a9e | 241 | adm_ctx.connection = conn_get_by_addrs(nla_data(adm_ctx.my_addr), |
089c075d AG |
242 | nla_len(adm_ctx.my_addr), |
243 | nla_data(adm_ctx.peer_addr), | |
244 | nla_len(adm_ctx.peer_addr)); | |
bde89a9e | 245 | if (!adm_ctx.connection) { |
089c075d AG |
246 | drbd_msg_put_info("unknown connection"); |
247 | return ERR_INVALID_REQUEST; | |
248 | } | |
249 | } | |
250 | ||
3b98c0c2 | 251 | /* some more paranoia, if the request was over-determined */ |
4bc76048 AG |
252 | if (adm_ctx.device && adm_ctx.resource && |
253 | adm_ctx.device->resource != adm_ctx.resource) { | |
254 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", | |
255 | adm_ctx.minor, adm_ctx.resource->name, | |
d8628a86 | 256 | adm_ctx.device->resource->name); |
44e52cfa | 257 | drbd_msg_put_info("minor exists in different resource"); |
527f4b24 LE |
258 | return ERR_INVALID_REQUEST; |
259 | } | |
b30ab791 | 260 | if (adm_ctx.device && |
3b98c0c2 | 261 | adm_ctx.volume != VOLUME_UNSPECIFIED && |
b30ab791 | 262 | adm_ctx.volume != adm_ctx.device->vnr) { |
3b98c0c2 LE |
263 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
264 | adm_ctx.minor, adm_ctx.volume, | |
77c556f6 | 265 | adm_ctx.device->vnr, |
d8628a86 | 266 | adm_ctx.device->resource->name); |
527f4b24 | 267 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
268 | return ERR_INVALID_REQUEST; |
269 | } | |
0ace9dfa | 270 | |
3b98c0c2 LE |
271 | return NO_ERROR; |
272 | ||
273 | fail: | |
274 | nlmsg_free(adm_ctx.reply_skb); | |
275 | adm_ctx.reply_skb = NULL; | |
1e2a2551 | 276 | return err; |
3b98c0c2 LE |
277 | } |
278 | ||
279 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
280 | { | |
bde89a9e | 281 | if (adm_ctx.connection) { |
05a10ec7 | 282 | kref_put(&adm_ctx.connection->kref, drbd_destroy_connection); |
bde89a9e | 283 | adm_ctx.connection = NULL; |
0ace9dfa | 284 | } |
4bc76048 AG |
285 | if (adm_ctx.resource) { |
286 | kref_put(&adm_ctx.resource->kref, drbd_destroy_resource); | |
287 | adm_ctx.resource = NULL; | |
288 | } | |
0ace9dfa | 289 | |
3b98c0c2 LE |
290 | if (!adm_ctx.reply_skb) |
291 | return -ENOMEM; | |
292 | ||
293 | adm_ctx.reply_dh->ret_code = retcode; | |
3b98c0c2 LE |
294 | drbd_adm_send_reply(adm_ctx.reply_skb, info); |
295 | return 0; | |
296 | } | |
b411b363 | 297 | |
bde89a9e | 298 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 299 | { |
6b75dced | 300 | char *afs; |
b411b363 | 301 | |
089c075d | 302 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 303 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
304 | return; |
305 | ||
bde89a9e | 306 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
307 | case AF_INET6: |
308 | afs = "ipv6"; | |
309 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 310 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 311 | break; |
089c075d AG |
312 | case AF_INET: |
313 | afs = "ipv4"; | |
314 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 315 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 316 | break; |
089c075d AG |
317 | default: |
318 | afs = "ssocks"; | |
319 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 320 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 321 | } |
089c075d | 322 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 323 | } |
b411b363 | 324 | |
b30ab791 | 325 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
326 | { |
327 | char *envp[] = { "HOME=/", | |
328 | "TERM=linux", | |
329 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
330 | (char[20]) { }, /* address family */ |
331 | (char[60]) { }, /* address */ | |
b411b363 | 332 | NULL }; |
6b75dced | 333 | char mb[12]; |
b411b363 | 334 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 335 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 336 | struct sib_info sib; |
b411b363 PR |
337 | int ret; |
338 | ||
bde89a9e AG |
339 | if (current == connection->worker.task) |
340 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 341 | |
b30ab791 | 342 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 343 | setup_khelper_env(connection, envp); |
b411b363 | 344 | |
1090c056 LE |
345 | /* The helper may take some time. |
346 | * write out any unsynced meta data changes now */ | |
b30ab791 | 347 | drbd_md_sync(device); |
1090c056 | 348 | |
d0180171 | 349 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
350 | sib.sib_reason = SIB_HELPER_PRE; |
351 | sib.helper_name = cmd; | |
b30ab791 | 352 | drbd_bcast_event(device, &sib); |
70834d30 | 353 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 354 | if (ret) |
d0180171 | 355 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
356 | usermode_helper, cmd, mb, |
357 | (ret >> 8) & 0xff, ret); | |
358 | else | |
d0180171 | 359 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
360 | usermode_helper, cmd, mb, |
361 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
362 | sib.sib_reason = SIB_HELPER_POST; |
363 | sib.helper_exit_code = ret; | |
b30ab791 | 364 | drbd_bcast_event(device, &sib); |
b411b363 | 365 | |
bde89a9e AG |
366 | if (current == connection->worker.task) |
367 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
368 | |
369 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
370 | ret = 0; | |
371 | ||
372 | return ret; | |
373 | } | |
374 | ||
bde89a9e | 375 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
376 | { |
377 | char *envp[] = { "HOME=/", | |
378 | "TERM=linux", | |
379 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
380 | (char[20]) { }, /* address family */ | |
381 | (char[60]) { }, /* address */ | |
382 | NULL }; | |
77c556f6 AG |
383 | char *resource_name = connection->resource->name; |
384 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
385 | int ret; |
386 | ||
bde89a9e AG |
387 | setup_khelper_env(connection, envp); |
388 | conn_md_sync(connection); | |
6b75dced | 389 | |
1ec861eb | 390 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced PR |
391 | /* TODO: conn_bcast_event() ?? */ |
392 | ||
98683650 | 393 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 394 | if (ret) |
1ec861eb | 395 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 396 | usermode_helper, cmd, resource_name, |
6b75dced PR |
397 | (ret >> 8) & 0xff, ret); |
398 | else | |
1ec861eb | 399 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 400 | usermode_helper, cmd, resource_name, |
6b75dced PR |
401 | (ret >> 8) & 0xff, ret); |
402 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 403 | |
b411b363 PR |
404 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
405 | ret = 0; | |
406 | ||
407 | return ret; | |
408 | } | |
409 | ||
bde89a9e | 410 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 411 | { |
cb703454 | 412 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 413 | struct drbd_peer_device *peer_device; |
cb703454 PR |
414 | int vnr; |
415 | ||
695d08fa | 416 | rcu_read_lock(); |
c06ece6b AG |
417 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
418 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 419 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
420 | struct disk_conf *disk_conf = |
421 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
422 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 423 | put_ldev(device); |
cb703454 PR |
424 | } |
425 | } | |
695d08fa | 426 | rcu_read_unlock(); |
cb703454 | 427 | |
d7fe69c6 PR |
428 | if (fp == FP_NOT_AVAIL) { |
429 | /* IO Suspending works on the whole resource. | |
430 | Do it only for one device. */ | |
431 | vnr = 0; | |
432 | peer_device = idr_get_next(&connection->peer_devices, &vnr); | |
433 | drbd_change_state(peer_device->device, CS_VERBOSE | CS_HARD, NS(susp_fen, 0)); | |
434 | } | |
435 | ||
cb703454 PR |
436 | return fp; |
437 | } | |
438 | ||
bde89a9e | 439 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 440 | { |
28e448bb | 441 | unsigned int connect_cnt; |
cb703454 PR |
442 | union drbd_state mask = { }; |
443 | union drbd_state val = { }; | |
444 | enum drbd_fencing_p fp; | |
b411b363 PR |
445 | char *ex_to_string; |
446 | int r; | |
b411b363 | 447 | |
bde89a9e | 448 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 449 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
cb703454 PR |
450 | return false; |
451 | } | |
b411b363 | 452 | |
0500813f | 453 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 454 | connect_cnt = connection->connect_cnt; |
0500813f | 455 | spin_unlock_irq(&connection->resource->req_lock); |
28e448bb | 456 | |
bde89a9e | 457 | fp = highest_fencing_policy(connection); |
cb703454 PR |
458 | switch (fp) { |
459 | case FP_NOT_AVAIL: | |
1ec861eb | 460 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 461 | goto out; |
cb703454 PR |
462 | case FP_DONT_CARE: |
463 | return true; | |
464 | default: ; | |
b411b363 PR |
465 | } |
466 | ||
bde89a9e | 467 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
468 | |
469 | switch ((r>>8) & 0xff) { | |
470 | case 3: /* peer is inconsistent */ | |
471 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
472 | mask.pdsk = D_MASK; |
473 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
474 | break; |
475 | case 4: /* peer got outdated, or was already outdated */ | |
476 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
477 | mask.pdsk = D_MASK; |
478 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
479 | break; |
480 | case 5: /* peer was down */ | |
bde89a9e | 481 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
482 | /* we will(have) create(d) a new UUID anyways... */ |
483 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
484 | mask.pdsk = D_MASK; |
485 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
486 | } else { |
487 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
488 | } |
489 | break; | |
490 | case 6: /* Peer is primary, voluntarily outdate myself. | |
491 | * This is useful when an unconnected R_SECONDARY is asked to | |
492 | * become R_PRIMARY, but finds the other peer being active. */ | |
493 | ex_to_string = "peer is active"; | |
1ec861eb | 494 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
495 | mask.disk = D_MASK; |
496 | val.disk = D_OUTDATED; | |
b411b363 PR |
497 | break; |
498 | case 7: | |
499 | if (fp != FP_STONITH) | |
1ec861eb | 500 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 501 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
502 | mask.pdsk = D_MASK; |
503 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
504 | break; |
505 | default: | |
506 | /* The script is broken ... */ | |
1ec861eb | 507 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 508 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
509 | } |
510 | ||
1ec861eb | 511 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 512 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 513 | |
cb703454 | 514 | out: |
fb22c402 | 515 | |
cb703454 | 516 | /* Not using |
bde89a9e | 517 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
518 | here, because we might were able to re-establish the connection in the |
519 | meantime. */ | |
0500813f | 520 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
521 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
522 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
523 | /* In case the connection was established and droped |
524 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 525 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 526 | else |
bde89a9e | 527 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 528 | } |
0500813f | 529 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 | 530 | |
bde89a9e | 531 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
532 | } |
533 | ||
87f7be4c PR |
534 | static int _try_outdate_peer_async(void *data) |
535 | { | |
bde89a9e | 536 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 537 | |
bde89a9e | 538 | conn_try_outdate_peer(connection); |
87f7be4c | 539 | |
05a10ec7 | 540 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
541 | return 0; |
542 | } | |
543 | ||
bde89a9e | 544 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
545 | { |
546 | struct task_struct *opa; | |
547 | ||
bde89a9e AG |
548 | kref_get(&connection->kref); |
549 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); | |
9dc9fbb3 | 550 | if (IS_ERR(opa)) { |
1ec861eb | 551 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 552 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 553 | } |
87f7be4c | 554 | } |
b411b363 | 555 | |
bf885f8a | 556 | enum drbd_state_rv |
b30ab791 | 557 | drbd_set_role(struct drbd_device *device, enum drbd_role new_role, int force) |
b411b363 PR |
558 | { |
559 | const int max_tries = 4; | |
bf885f8a | 560 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 561 | struct net_conf *nc; |
b411b363 PR |
562 | int try = 0; |
563 | int forced = 0; | |
564 | union drbd_state mask, val; | |
b411b363 | 565 | |
b6f85ef9 AG |
566 | if (new_role == R_PRIMARY) { |
567 | struct drbd_connection *connection; | |
568 | ||
569 | /* Detect dead peers as soon as possible. */ | |
570 | ||
571 | rcu_read_lock(); | |
572 | for_each_connection(connection, device->resource) | |
573 | request_ping(connection); | |
574 | rcu_read_unlock(); | |
575 | } | |
b411b363 | 576 | |
b30ab791 | 577 | mutex_lock(device->state_mutex); |
b411b363 PR |
578 | |
579 | mask.i = 0; mask.role = R_MASK; | |
580 | val.i = 0; val.role = new_role; | |
581 | ||
582 | while (try++ < max_tries) { | |
b30ab791 | 583 | rv = _drbd_request_state(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
584 | |
585 | /* in case we first succeeded to outdate, | |
586 | * but now suddenly could establish a connection */ | |
bf885f8a | 587 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
588 | val.pdsk = 0; |
589 | mask.pdsk = 0; | |
590 | continue; | |
591 | } | |
592 | ||
bf885f8a | 593 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
594 | (device->state.disk < D_UP_TO_DATE && |
595 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
596 | mask.disk = D_MASK; |
597 | val.disk = D_UP_TO_DATE; | |
598 | forced = 1; | |
599 | continue; | |
600 | } | |
601 | ||
bf885f8a | 602 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 603 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 604 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 605 | |
a6b32bc3 | 606 | if (conn_try_outdate_peer(first_peer_device(device)->connection)) { |
b411b363 PR |
607 | val.disk = D_UP_TO_DATE; |
608 | mask.disk = D_MASK; | |
609 | } | |
b411b363 PR |
610 | continue; |
611 | } | |
612 | ||
bf885f8a | 613 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 614 | goto out; |
bf885f8a | 615 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
a6b32bc3 | 616 | if (!conn_try_outdate_peer(first_peer_device(device)->connection) && force) { |
d0180171 | 617 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
618 | mask.pdsk = D_MASK; |
619 | val.pdsk = D_OUTDATED; | |
b411b363 | 620 | |
cb703454 | 621 | } |
b411b363 PR |
622 | continue; |
623 | } | |
bf885f8a | 624 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
625 | /* Maybe the peer is detected as dead very soon... |
626 | retry at most once more in this case. */ | |
44ed167d PR |
627 | int timeo; |
628 | rcu_read_lock(); | |
a6b32bc3 | 629 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
630 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
631 | rcu_read_unlock(); | |
632 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
633 | if (try < max_tries) |
634 | try = max_tries - 1; | |
635 | continue; | |
636 | } | |
bf885f8a | 637 | if (rv < SS_SUCCESS) { |
b30ab791 | 638 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 639 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 640 | if (rv < SS_SUCCESS) |
3b98c0c2 | 641 | goto out; |
b411b363 PR |
642 | } |
643 | break; | |
644 | } | |
645 | ||
bf885f8a | 646 | if (rv < SS_SUCCESS) |
3b98c0c2 | 647 | goto out; |
b411b363 PR |
648 | |
649 | if (forced) | |
d0180171 | 650 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
651 | |
652 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 653 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 654 | |
b6dd1a89 LE |
655 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
656 | ||
b411b363 | 657 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
658 | set_disk_ro(device->vdisk, true); |
659 | if (get_ldev(device)) { | |
660 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
661 | put_ldev(device); | |
b411b363 PR |
662 | } |
663 | } else { | |
0500813f | 664 | mutex_lock(&device->resource->conf_update); |
a6b32bc3 | 665 | nc = first_peer_device(device)->connection->net_conf; |
44ed167d | 666 | if (nc) |
6139f60d | 667 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
0500813f | 668 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 669 | |
b30ab791 AG |
670 | set_disk_ro(device->vdisk, false); |
671 | if (get_ldev(device)) { | |
672 | if (((device->state.conn < C_CONNECTED || | |
673 | device->state.pdsk <= D_FAILED) | |
674 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
675 | drbd_uuid_new_current(device); | |
b411b363 | 676 | |
b30ab791 AG |
677 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
678 | put_ldev(device); | |
b411b363 PR |
679 | } |
680 | } | |
681 | ||
19f843aa LE |
682 | /* writeout of activity log covered areas of the bitmap |
683 | * to stable storage done in after state change already */ | |
b411b363 | 684 | |
b30ab791 | 685 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
686 | /* if this was forced, we should consider sync */ |
687 | if (forced) | |
69a22773 AG |
688 | drbd_send_uuids(first_peer_device(device)); |
689 | drbd_send_current_state(first_peer_device(device)); | |
b411b363 PR |
690 | } |
691 | ||
b30ab791 | 692 | drbd_md_sync(device); |
b411b363 | 693 | |
b30ab791 | 694 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 695 | out: |
b30ab791 | 696 | mutex_unlock(device->state_mutex); |
bf885f8a | 697 | return rv; |
b411b363 PR |
698 | } |
699 | ||
3b98c0c2 | 700 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 701 | { |
3b98c0c2 LE |
702 | return err == -ENOMSG ? "required attribute missing" : |
703 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 704 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 705 | "invalid attribute value"; |
ef50a3e3 | 706 | } |
b411b363 | 707 | |
3b98c0c2 | 708 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 709 | { |
3b98c0c2 LE |
710 | struct set_role_parms parms; |
711 | int err; | |
712 | enum drbd_ret_code retcode; | |
b411b363 | 713 | |
3b98c0c2 LE |
714 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
715 | if (!adm_ctx.reply_skb) | |
716 | return retcode; | |
717 | if (retcode != NO_ERROR) | |
718 | goto out; | |
b411b363 | 719 | |
3b98c0c2 LE |
720 | memset(&parms, 0, sizeof(parms)); |
721 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 722 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
723 | if (err) { |
724 | retcode = ERR_MANDATORY_TAG; | |
725 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
726 | goto out; | |
727 | } | |
728 | } | |
b411b363 | 729 | |
3b98c0c2 | 730 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 731 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 732 | else |
b30ab791 | 733 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
3b98c0c2 LE |
734 | out: |
735 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
736 | return 0; |
737 | } | |
738 | ||
ae8bf312 LE |
739 | /* Initializes the md.*_offset members, so we are able to find |
740 | * the on disk meta data. | |
741 | * | |
742 | * We currently have two possible layouts: | |
743 | * external: | |
744 | * |----------- md_size_sect ------------------| | |
745 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
746 | * | al_offset == 8 | | |
747 | * | bm_offset = al_offset + X | | |
748 | * ==> bitmap sectors = md_size_sect - bm_offset | |
749 | * | |
750 | * internal: | |
751 | * |----------- md_size_sect ------------------| | |
752 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
753 | * | al_offset < 0 | | |
754 | * | bm_offset = al_offset - Y | | |
755 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
756 | * | |
757 | * Activity log size used to be fixed 32kB, | |
758 | * but is about to become configurable. | |
759 | */ | |
b30ab791 | 760 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
761 | struct drbd_backing_dev *bdev) |
762 | { | |
763 | sector_t md_size_sect = 0; | |
c04ccaa6 | 764 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 765 | |
3a4d4eb3 LE |
766 | bdev->md.md_offset = drbd_md_ss(bdev); |
767 | ||
68e41a43 | 768 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
769 | default: |
770 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 771 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
772 | bdev->md.al_offset = MD_4kB_SECT; |
773 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
774 | break; |
775 | case DRBD_MD_INDEX_FLEX_EXT: | |
776 | /* just occupy the full device; unit: sectors */ | |
777 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
778 | bdev->md.al_offset = MD_4kB_SECT; |
779 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
780 | break; |
781 | case DRBD_MD_INDEX_INTERNAL: | |
782 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 783 | /* al size is still fixed */ |
ae8bf312 | 784 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
785 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
786 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
787 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
788 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
789 | md_size_sect = ALIGN(md_size_sect, 8); | |
790 | ||
791 | /* plus the "drbd meta data super block", | |
792 | * and the activity log; */ | |
ae8bf312 | 793 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
794 | |
795 | bdev->md.md_size_sect = md_size_sect; | |
796 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 797 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
798 | break; |
799 | } | |
800 | } | |
801 | ||
4b0715f0 | 802 | /* input size is expected to be in KB */ |
b411b363 PR |
803 | char *ppsize(char *buf, unsigned long long size) |
804 | { | |
4b0715f0 LE |
805 | /* Needs 9 bytes at max including trailing NUL: |
806 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
807 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
808 | int base = 0; | |
4b0715f0 | 809 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
810 | /* shift + round */ |
811 | size = (size >> 10) + !!(size & (1<<9)); | |
812 | base++; | |
813 | } | |
4b0715f0 | 814 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
815 | |
816 | return buf; | |
817 | } | |
818 | ||
819 | /* there is still a theoretical deadlock when called from receiver | |
820 | * on an D_INCONSISTENT R_PRIMARY: | |
821 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
822 | * packet from remote to dec_ap_bio again. | |
823 | * receiver receive_sizes(), comes here, | |
824 | * waits for ap_bio_cnt == 0. -> deadlock. | |
825 | * but this cannot happen, actually, because: | |
826 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
827 | * (not connected, or bad/no disk on peer): | |
828 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
829 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
830 | * peer may not initiate a resize. | |
831 | */ | |
3b98c0c2 LE |
832 | /* Note these are not to be confused with |
833 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
834 | * which are (sub) state changes triggered by admin (drbdsetup), | |
835 | * and can be long lived. | |
b30ab791 | 836 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 837 | * and should be short-lived. */ |
b30ab791 | 838 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 839 | { |
b30ab791 AG |
840 | set_bit(SUSPEND_IO, &device->flags); |
841 | if (drbd_suspended(device)) | |
265be2d0 | 842 | return; |
b30ab791 | 843 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
844 | } |
845 | ||
b30ab791 | 846 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 847 | { |
b30ab791 AG |
848 | clear_bit(SUSPEND_IO, &device->flags); |
849 | wake_up(&device->misc_wait); | |
b411b363 PR |
850 | } |
851 | ||
852 | /** | |
853 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 854 | * @device: DRBD device. |
b411b363 PR |
855 | * |
856 | * Returns 0 on success, negative return values indicate errors. | |
857 | * You should call drbd_md_sync() after calling this function. | |
858 | */ | |
d752b269 | 859 | enum determine_dev_size |
b30ab791 | 860 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
861 | { |
862 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 863 | sector_t la_size_sect, u_size; |
b30ab791 | 864 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
865 | u32 prev_al_stripe_size_4k; |
866 | u32 prev_al_stripes; | |
b411b363 PR |
867 | sector_t size; |
868 | char ppb[10]; | |
d752b269 | 869 | void *buffer; |
b411b363 PR |
870 | |
871 | int md_moved, la_size_changed; | |
e96c9633 | 872 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
873 | |
874 | /* race: | |
875 | * application request passes inc_ap_bio, | |
876 | * but then cannot get an AL-reference. | |
877 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
878 | * | |
879 | * to avoid that: | |
880 | * Suspend IO right here. | |
881 | * still lock the act_log to not trigger ASSERTs there. | |
882 | */ | |
b30ab791 AG |
883 | drbd_suspend_io(device); |
884 | buffer = drbd_md_get_buffer(device); /* Lock meta-data IO */ | |
d752b269 | 885 | if (!buffer) { |
b30ab791 | 886 | drbd_resume_io(device); |
d752b269 PR |
887 | return DS_ERROR; |
888 | } | |
b411b363 PR |
889 | |
890 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 891 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 892 | |
b30ab791 AG |
893 | prev_first_sect = drbd_md_first_sector(device->ldev); |
894 | prev_size = device->ldev->md.md_size_sect; | |
895 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 896 | |
d752b269 PR |
897 | if (rs) { |
898 | /* rs is non NULL if we should change the AL layout only */ | |
899 | ||
900 | prev_al_stripes = md->al_stripes; | |
901 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
902 | ||
903 | md->al_stripes = rs->al_stripes; | |
904 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
905 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
906 | } | |
907 | ||
b30ab791 | 908 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 909 | |
daeda1cc | 910 | rcu_read_lock(); |
b30ab791 | 911 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 912 | rcu_read_unlock(); |
b30ab791 | 913 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 914 | |
d752b269 PR |
915 | if (size < la_size_sect) { |
916 | if (rs && u_size == 0) { | |
917 | /* Remove "rs &&" later. This check should always be active, but | |
918 | right now the receiver expects the permissive behavior */ | |
d0180171 | 919 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
920 | "Use --size=%llus for explicit shrink.\n", |
921 | (unsigned long long)size); | |
922 | rv = DS_ERROR_SHRINK; | |
923 | } | |
924 | if (u_size > size) | |
925 | rv = DS_ERROR_SPACE_MD; | |
926 | if (rv != DS_UNCHANGED) | |
927 | goto err_out; | |
928 | } | |
929 | ||
b30ab791 AG |
930 | if (drbd_get_capacity(device->this_bdev) != size || |
931 | drbd_bm_capacity(device) != size) { | |
b411b363 | 932 | int err; |
b30ab791 | 933 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
934 | if (unlikely(err)) { |
935 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 936 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 937 | if (size == 0) { |
d0180171 | 938 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
939 | "Could not allocate bitmap!\n"); |
940 | } else { | |
d0180171 | 941 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
942 | "Leaving size unchanged at size = %lu KB\n", |
943 | (unsigned long)size); | |
944 | } | |
e96c9633 | 945 | rv = DS_ERROR; |
b411b363 PR |
946 | } |
947 | /* racy, see comments above. */ | |
b30ab791 AG |
948 | drbd_set_my_capacity(device, size); |
949 | device->ldev->md.la_size_sect = size; | |
d0180171 | 950 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
951 | (unsigned long long)size>>1); |
952 | } | |
d752b269 PR |
953 | if (rv <= DS_ERROR) |
954 | goto err_out; | |
b411b363 | 955 | |
b30ab791 | 956 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 957 | |
b30ab791 AG |
958 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
959 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 960 | |
d752b269 PR |
961 | if (la_size_changed || md_moved || rs) { |
962 | u32 prev_flags; | |
24dccabb | 963 | |
b30ab791 | 964 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
965 | |
966 | prev_flags = md->flags; | |
967 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 968 | drbd_md_write(device, buffer); |
d752b269 | 969 | |
d0180171 | 970 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
971 | la_size_changed && md_moved ? "size changed and md moved" : |
972 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 973 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 974 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 975 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 976 | drbd_initialize_al(device, buffer); |
d752b269 PR |
977 | |
978 | md->flags = prev_flags; | |
b30ab791 | 979 | drbd_md_write(device, buffer); |
d752b269 PR |
980 | |
981 | if (rs) | |
d0180171 AG |
982 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
983 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
984 | } |
985 | ||
cccac985 | 986 | if (size > la_size_sect) |
57737adc | 987 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 988 | if (size < la_size_sect) |
e96c9633 | 989 | rv = DS_SHRUNK; |
d752b269 PR |
990 | |
991 | if (0) { | |
992 | err_out: | |
993 | if (rs) { | |
994 | md->al_stripes = prev_al_stripes; | |
995 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
996 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
997 | ||
b30ab791 | 998 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
999 | } |
1000 | } | |
b30ab791 AG |
1001 | lc_unlock(device->act_log); |
1002 | wake_up(&device->al_wait); | |
1003 | drbd_md_put_buffer(device); | |
1004 | drbd_resume_io(device); | |
b411b363 PR |
1005 | |
1006 | return rv; | |
1007 | } | |
1008 | ||
1009 | sector_t | |
b30ab791 | 1010 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1011 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1012 | { |
b30ab791 | 1013 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1014 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1015 | sector_t m_size; /* my size */ |
b411b363 PR |
1016 | sector_t size = 0; |
1017 | ||
1018 | m_size = drbd_get_max_capacity(bdev); | |
1019 | ||
b30ab791 | 1020 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1021 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1022 | p_size = m_size; |
1023 | } | |
1024 | ||
b411b363 PR |
1025 | if (p_size && m_size) { |
1026 | size = min_t(sector_t, p_size, m_size); | |
1027 | } else { | |
cccac985 LE |
1028 | if (la_size_sect) { |
1029 | size = la_size_sect; | |
b411b363 PR |
1030 | if (m_size && m_size < size) |
1031 | size = m_size; | |
1032 | if (p_size && p_size < size) | |
1033 | size = p_size; | |
1034 | } else { | |
1035 | if (m_size) | |
1036 | size = m_size; | |
1037 | if (p_size) | |
1038 | size = p_size; | |
1039 | } | |
1040 | } | |
1041 | ||
1042 | if (size == 0) | |
d0180171 | 1043 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1044 | |
1045 | if (u_size) { | |
1046 | if (u_size > size) | |
d0180171 | 1047 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1048 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1049 | else | |
1050 | size = u_size; | |
1051 | } | |
1052 | ||
1053 | return size; | |
1054 | } | |
1055 | ||
1056 | /** | |
1057 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1058 | * @device: DRBD device. |
b411b363 PR |
1059 | * |
1060 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1061 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1062 | * this function. | |
1063 | */ | |
b30ab791 | 1064 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1065 | { |
1066 | struct lru_cache *n, *t; | |
1067 | struct lc_element *e; | |
1068 | unsigned int in_use; | |
1069 | int i; | |
1070 | ||
b30ab791 AG |
1071 | if (device->act_log && |
1072 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1073 | return 0; |
1074 | ||
1075 | in_use = 0; | |
b30ab791 | 1076 | t = device->act_log; |
7ad651b5 | 1077 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1078 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1079 | |
1080 | if (n == NULL) { | |
d0180171 | 1081 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1082 | return -ENOMEM; |
1083 | } | |
b30ab791 | 1084 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1085 | if (t) { |
1086 | for (i = 0; i < t->nr_elements; i++) { | |
1087 | e = lc_element_by_index(t, i); | |
1088 | if (e->refcnt) | |
d0180171 | 1089 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1090 | e->lc_number, e->refcnt); |
1091 | in_use += e->refcnt; | |
1092 | } | |
1093 | } | |
1094 | if (!in_use) | |
b30ab791 AG |
1095 | device->act_log = n; |
1096 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1097 | if (in_use) { |
d0180171 | 1098 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1099 | lc_destroy(n); |
1100 | return -EBUSY; | |
1101 | } else { | |
1102 | if (t) | |
1103 | lc_destroy(t); | |
1104 | } | |
b30ab791 | 1105 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1106 | return 0; |
1107 | } | |
1108 | ||
b30ab791 | 1109 | static void drbd_setup_queue_param(struct drbd_device *device, unsigned int max_bio_size) |
b411b363 | 1110 | { |
b30ab791 | 1111 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1112 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1113 | unsigned int max_segments = 0; | |
99432fcc | 1114 | |
b30ab791 AG |
1115 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1116 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1117 | |
1118 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1119 | rcu_read_lock(); |
b30ab791 | 1120 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1121 | rcu_read_unlock(); |
b30ab791 | 1122 | put_ldev(device); |
99432fcc | 1123 | } |
b411b363 | 1124 | |
b411b363 | 1125 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1126 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1127 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1128 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1129 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1130 | |
b30ab791 AG |
1131 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1132 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1133 | |
1134 | blk_queue_stack_limits(q, b); | |
1135 | ||
1136 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1137 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1138 | q->backing_dev_info.ra_pages, |
1139 | b->backing_dev_info.ra_pages); | |
1140 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1141 | } | |
b30ab791 | 1142 | put_ldev(device); |
99432fcc PR |
1143 | } |
1144 | } | |
1145 | ||
b30ab791 | 1146 | void drbd_reconsider_max_bio_size(struct drbd_device *device) |
99432fcc | 1147 | { |
db141b2f | 1148 | unsigned int now, new, local, peer; |
99432fcc | 1149 | |
b30ab791 AG |
1150 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1151 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1152 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1153 | |
b30ab791 AG |
1154 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1155 | local = queue_max_hw_sectors(device->ldev->backing_bdev->bd_disk->queue) << 9; | |
1156 | device->local_max_bio_size = local; | |
1157 | put_ldev(device); | |
b411b363 | 1158 | } |
db141b2f | 1159 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1160 | |
1161 | /* We may ignore peer limits if the peer is modern enough. | |
1162 | Because new from 8.3.8 onwards the peer can use multiple | |
1163 | BIOs for a single peer_request */ | |
b30ab791 | 1164 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1165 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1166 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1167 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1168 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1169 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1170 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1171 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1172 | else | |
99432fcc PR |
1173 | peer = DRBD_MAX_BIO_SIZE; |
1174 | } | |
1175 | ||
db141b2f | 1176 | new = min(local, peer); |
99432fcc | 1177 | |
b30ab791 | 1178 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1179 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1180 | |
1181 | if (new != now) | |
d0180171 | 1182 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1183 | |
b30ab791 | 1184 | drbd_setup_queue_param(device, new); |
b411b363 PR |
1185 | } |
1186 | ||
a18e9d1e | 1187 | /* Starts the worker thread */ |
bde89a9e | 1188 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1189 | { |
bde89a9e | 1190 | drbd_thread_start(&connection->worker); |
b5043c5e | 1191 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1192 | } |
1193 | ||
a18e9d1e | 1194 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1195 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1196 | { |
992d6e91 | 1197 | bool stop_threads; |
0500813f | 1198 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1199 | stop_threads = conn_all_vols_unconf(connection) && |
1200 | connection->cstate == C_STANDALONE; | |
0500813f | 1201 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1202 | if (stop_threads) { |
1203 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1204 | * in conn_disconnect() */ |
bde89a9e AG |
1205 | drbd_thread_stop(&connection->receiver); |
1206 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1207 | } |
b411b363 PR |
1208 | } |
1209 | ||
0778286a | 1210 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1211 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1212 | { |
1213 | int s = 0; | |
1214 | ||
b30ab791 | 1215 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1216 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1217 | return; |
1218 | } | |
1219 | ||
b30ab791 | 1220 | drbd_al_shrink(device); |
0500813f | 1221 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1222 | if (device->state.conn < C_CONNECTED) |
1223 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1224 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1225 | lc_unlock(device->act_log); |
0778286a PR |
1226 | |
1227 | if (s) | |
d0180171 | 1228 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1229 | } |
1230 | ||
5979e361 LE |
1231 | |
1232 | static bool should_set_defaults(struct genl_info *info) | |
1233 | { | |
1234 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1235 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1236 | } | |
1237 | ||
5bbcf5e6 | 1238 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1239 | { |
5bbcf5e6 LE |
1240 | /* This is limited by 16 bit "slot" numbers, |
1241 | * and by available on-disk context storage. | |
1242 | * | |
1243 | * Also (u16)~0 is special (denotes a "free" extent). | |
1244 | * | |
1245 | * One transaction occupies one 4kB on-disk block, | |
1246 | * we have n such blocks in the on disk ring buffer, | |
1247 | * the "current" transaction may fail (n-1), | |
1248 | * and there is 919 slot numbers context information per transaction. | |
1249 | * | |
1250 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1251 | * so cap there first. | |
1252 | */ | |
1253 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1254 | const unsigned int sufficient_on_disk = | |
1255 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1256 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1257 | |
5bbcf5e6 LE |
1258 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1259 | ||
1260 | if (al_size_4k > sufficient_on_disk) | |
1261 | return max_al_nr; | |
1262 | ||
1263 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1264 | } |
1265 | ||
f399002e LE |
1266 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1267 | { | |
1268 | enum drbd_ret_code retcode; | |
b30ab791 | 1269 | struct drbd_device *device; |
daeda1cc | 1270 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1271 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1272 | int err, fifo_size; |
f399002e LE |
1273 | |
1274 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1275 | if (!adm_ctx.reply_skb) | |
1276 | return retcode; | |
1277 | if (retcode != NO_ERROR) | |
1278 | goto out; | |
1279 | ||
b30ab791 | 1280 | device = adm_ctx.device; |
f399002e LE |
1281 | |
1282 | /* we also need a disk | |
1283 | * to change the options on */ | |
b30ab791 | 1284 | if (!get_ldev(device)) { |
f399002e LE |
1285 | retcode = ERR_NO_DISK; |
1286 | goto out; | |
1287 | } | |
1288 | ||
daeda1cc | 1289 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1290 | if (!new_disk_conf) { |
f399002e LE |
1291 | retcode = ERR_NOMEM; |
1292 | goto fail; | |
1293 | } | |
1294 | ||
0500813f | 1295 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1296 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1297 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1298 | if (should_set_defaults(info)) |
b966b5dd | 1299 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1300 | |
5ecc72c3 | 1301 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1302 | if (err && err != -ENOMSG) { |
f399002e LE |
1303 | retcode = ERR_MANDATORY_TAG; |
1304 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
8e229434 | 1305 | goto fail_unlock; |
f399002e LE |
1306 | } |
1307 | ||
5ecc72c3 LE |
1308 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1309 | new_disk_conf->resync_rate = 1; | |
f399002e | 1310 | |
5bbcf5e6 LE |
1311 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1312 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1313 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1314 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1315 | |
1316 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1317 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1318 | |
5ecc72c3 | 1319 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1320 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1321 | new_plan = fifo_alloc(fifo_size); |
1322 | if (!new_plan) { | |
d0180171 | 1323 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1324 | retcode = ERR_NOMEM; |
daeda1cc | 1325 | goto fail_unlock; |
f399002e LE |
1326 | } |
1327 | } | |
1328 | ||
b30ab791 AG |
1329 | drbd_suspend_io(device); |
1330 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1331 | drbd_al_shrink(device); | |
1332 | err = drbd_check_al_size(device, new_disk_conf); | |
1333 | lc_unlock(device->act_log); | |
1334 | wake_up(&device->al_wait); | |
1335 | drbd_resume_io(device); | |
f399002e LE |
1336 | |
1337 | if (err) { | |
1338 | retcode = ERR_NOMEM; | |
daeda1cc | 1339 | goto fail_unlock; |
f399002e LE |
1340 | } |
1341 | ||
dc97b708 | 1342 | write_lock_irq(&global_state_lock); |
b30ab791 | 1343 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1344 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1345 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1346 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1347 | } |
1348 | write_unlock_irq(&global_state_lock); | |
f399002e | 1349 | |
daeda1cc PR |
1350 | if (retcode != NO_ERROR) |
1351 | goto fail_unlock; | |
f399002e | 1352 | |
813472ce | 1353 | if (new_plan) { |
b30ab791 AG |
1354 | old_plan = device->rs_plan_s; |
1355 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1356 | } |
9958c857 | 1357 | |
0500813f | 1358 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1359 | |
9a51ab1c | 1360 | if (new_disk_conf->al_updates) |
b30ab791 | 1361 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1362 | else |
b30ab791 | 1363 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1364 | |
691631c0 | 1365 | if (new_disk_conf->md_flushes) |
b30ab791 | 1366 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1367 | else |
b30ab791 | 1368 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1369 | |
a6b32bc3 | 1370 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
27eb13e9 | 1371 | |
b30ab791 | 1372 | drbd_md_sync(device); |
f399002e | 1373 | |
69a22773 AG |
1374 | if (device->state.conn >= C_CONNECTED) { |
1375 | struct drbd_peer_device *peer_device; | |
1376 | ||
1377 | for_each_peer_device(peer_device, device) | |
1378 | drbd_send_sync_param(peer_device); | |
1379 | } | |
f399002e | 1380 | |
daeda1cc PR |
1381 | synchronize_rcu(); |
1382 | kfree(old_disk_conf); | |
813472ce | 1383 | kfree(old_plan); |
b30ab791 | 1384 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1385 | goto success; |
1386 | ||
1387 | fail_unlock: | |
0500813f | 1388 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1389 | fail: |
5ecc72c3 | 1390 | kfree(new_disk_conf); |
813472ce | 1391 | kfree(new_plan); |
daeda1cc | 1392 | success: |
b30ab791 | 1393 | put_ldev(device); |
f399002e LE |
1394 | out: |
1395 | drbd_adm_finish(info, retcode); | |
1396 | return 0; | |
1397 | } | |
1398 | ||
3b98c0c2 | 1399 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1400 | { |
b30ab791 | 1401 | struct drbd_device *device; |
3b98c0c2 | 1402 | int err; |
116676ca | 1403 | enum drbd_ret_code retcode; |
b411b363 PR |
1404 | enum determine_dev_size dd; |
1405 | sector_t max_possible_sectors; | |
1406 | sector_t min_md_device_sectors; | |
1407 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1408 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1409 | struct block_device *bdev; |
b411b363 | 1410 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1411 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1412 | union drbd_state ns, os; |
f2024e7c | 1413 | enum drbd_state_rv rv; |
44ed167d | 1414 | struct net_conf *nc; |
b411b363 | 1415 | |
3b98c0c2 LE |
1416 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1417 | if (!adm_ctx.reply_skb) | |
1418 | return retcode; | |
1419 | if (retcode != NO_ERROR) | |
40cbf085 | 1420 | goto finish; |
b411b363 | 1421 | |
b30ab791 | 1422 | device = adm_ctx.device; |
a6b32bc3 | 1423 | conn_reconfig_start(first_peer_device(device)->connection); |
b411b363 PR |
1424 | |
1425 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1426 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1427 | retcode = ERR_DISK_CONFIGURED; |
1428 | goto fail; | |
1429 | } | |
82f59cc6 LE |
1430 | /* It may just now have detached because of IO error. Make sure |
1431 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1432 | * e.g. if someone calls attach from the on-io-error handler, | |
1433 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
b30ab791 | 1434 | wait_event(device->misc_wait, !atomic_read(&device->local_cnt)); |
b411b363 | 1435 | |
383606e0 | 1436 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1437 | clear_bit(FORCE_DETACH, &device->flags); |
1438 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1439 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1440 | |
0029d624 | 1441 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1442 | device->rs_total = 0; |
1443 | device->rs_failed = 0; | |
1444 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1445 | |
3b98c0c2 | 1446 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1447 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1448 | if (!nbc) { | |
1449 | retcode = ERR_NOMEM; | |
1450 | goto fail; | |
1451 | } | |
9f2247bb PR |
1452 | spin_lock_init(&nbc->md.uuid_lock); |
1453 | ||
daeda1cc PR |
1454 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1455 | if (!new_disk_conf) { | |
1456 | retcode = ERR_NOMEM; | |
b411b363 PR |
1457 | goto fail; |
1458 | } | |
daeda1cc | 1459 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1460 | |
daeda1cc PR |
1461 | set_disk_conf_defaults(new_disk_conf); |
1462 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1463 | if (err) { |
b411b363 | 1464 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1465 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1466 | goto fail; |
1467 | } | |
1468 | ||
5bbcf5e6 LE |
1469 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1470 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1471 | |
9958c857 PR |
1472 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1473 | if (!new_plan) { | |
1474 | retcode = ERR_NOMEM; | |
1475 | goto fail; | |
1476 | } | |
1477 | ||
daeda1cc | 1478 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1479 | retcode = ERR_MD_IDX_INVALID; |
1480 | goto fail; | |
1481 | } | |
1482 | ||
a3f8f7dc | 1483 | write_lock_irq(&global_state_lock); |
b30ab791 | 1484 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1485 | write_unlock_irq(&global_state_lock); |
1486 | if (retcode != NO_ERROR) | |
1487 | goto fail; | |
1488 | ||
44ed167d | 1489 | rcu_read_lock(); |
a6b32bc3 | 1490 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d | 1491 | if (nc) { |
daeda1cc | 1492 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1493 | rcu_read_unlock(); |
47ff2d0a PR |
1494 | retcode = ERR_STONITH_AND_PROT_A; |
1495 | goto fail; | |
1496 | } | |
1497 | } | |
44ed167d | 1498 | rcu_read_unlock(); |
47ff2d0a | 1499 | |
daeda1cc | 1500 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1501 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1502 | if (IS_ERR(bdev)) { |
d0180171 | 1503 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1504 | PTR_ERR(bdev)); |
b411b363 PR |
1505 | retcode = ERR_OPEN_DISK; |
1506 | goto fail; | |
1507 | } | |
e525fd89 TH |
1508 | nbc->backing_bdev = bdev; |
1509 | ||
1510 | /* | |
1511 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1512 | * drbd sharing one meta device. TODO in that case, paranoia | |
1513 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1514 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1515 | * should check it for you already; but if you don't, or | |
1516 | * someone fooled it, we need to double check here) | |
1517 | */ | |
daeda1cc | 1518 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1519 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1520 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1521 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1522 | if (IS_ERR(bdev)) { |
d0180171 | 1523 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1524 | PTR_ERR(bdev)); |
b411b363 PR |
1525 | retcode = ERR_OPEN_MD_DISK; |
1526 | goto fail; | |
1527 | } | |
e525fd89 | 1528 | nbc->md_bdev = bdev; |
b411b363 | 1529 | |
e525fd89 | 1530 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1531 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1532 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1533 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1534 | goto fail; |
1535 | } | |
1536 | ||
1537 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1538 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1539 | offsetof(struct bm_extent, lce)); |
1540 | if (!resync_lru) { | |
1541 | retcode = ERR_NOMEM; | |
e525fd89 | 1542 | goto fail; |
b411b363 PR |
1543 | } |
1544 | ||
c04ccaa6 LE |
1545 | /* Read our meta data super block early. |
1546 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1547 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1548 | if (retcode != NO_ERROR) |
1549 | goto fail; | |
b411b363 | 1550 | |
5bbcf5e6 LE |
1551 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1552 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1553 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1554 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1555 | ||
daeda1cc | 1556 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1557 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1558 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1559 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1560 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1561 | goto fail; |
b411b363 PR |
1562 | } |
1563 | ||
daeda1cc | 1564 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1565 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1566 | /* at least one MB, otherwise it does not make sense */ | |
1567 | min_md_device_sectors = (2<<10); | |
1568 | } else { | |
1569 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1570 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1571 | } |
1572 | ||
b411b363 | 1573 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1574 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1575 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1576 | "at least %llu sectors needed for this meta-disk type\n", |
1577 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1578 | goto fail; |
b411b363 PR |
1579 | } |
1580 | ||
1581 | /* Make sure the new disk is big enough | |
1582 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1583 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1584 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1585 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1586 | goto fail; |
b411b363 PR |
1587 | } |
1588 | ||
1589 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1590 | ||
1352994b | 1591 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1592 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1593 | "to currently maximum possible %llu sectors <==\n", |
1594 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1595 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1596 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1597 | "meta data may help <<==\n"); |
1598 | } | |
1599 | ||
b30ab791 | 1600 | drbd_suspend_io(device); |
b411b363 | 1601 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1602 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1603 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1604 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1605 | * As barriers are counted per resource, | |
1606 | * we'd need to suspend io on all devices of a resource. | |
1607 | */ | |
b30ab791 | 1608 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1609 | /* and for any other previously queued work */ |
b5043c5e | 1610 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
b411b363 | 1611 | |
b30ab791 | 1612 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1613 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1614 | drbd_resume_io(device); |
f2024e7c | 1615 | if (rv < SS_SUCCESS) |
e525fd89 | 1616 | goto fail; |
b411b363 | 1617 | |
b30ab791 | 1618 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1619 | goto force_diskless; |
1620 | ||
b30ab791 AG |
1621 | if (!device->bitmap) { |
1622 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1623 | retcode = ERR_NOMEM; |
1624 | goto force_diskless_dec; | |
1625 | } | |
1626 | } | |
1627 | ||
b30ab791 AG |
1628 | if (device->state.conn < C_CONNECTED && |
1629 | device->state.role == R_PRIMARY && | |
1630 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
d0180171 | 1631 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1632 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1633 | retcode = ERR_DATA_NOT_CURRENT; |
1634 | goto force_diskless_dec; | |
1635 | } | |
1636 | ||
1637 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1638 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1639 | retcode = ERR_NOMEM; |
1640 | goto force_diskless_dec; | |
1641 | } | |
1642 | ||
1643 | /* Prevent shrinking of consistent devices ! */ | |
1644 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1645 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1646 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1647 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1648 | goto force_diskless_dec; |
1649 | } | |
1650 | ||
b411b363 PR |
1651 | /* Reset the "barriers don't work" bits here, then force meta data to |
1652 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1653 | if (new_disk_conf->md_flushes) |
b30ab791 | 1654 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1655 | else |
b30ab791 | 1656 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1657 | |
1658 | /* Point of no return reached. | |
1659 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1660 | * now device takes over responsibility, and the state engine should |
b411b363 | 1661 | * clean it up somewhere. */ |
0b0ba1ef | 1662 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1663 | device->ldev = nbc; |
1664 | device->resync = resync_lru; | |
1665 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1666 | nbc = NULL; |
1667 | resync_lru = NULL; | |
daeda1cc | 1668 | new_disk_conf = NULL; |
9958c857 | 1669 | new_plan = NULL; |
b411b363 | 1670 | |
a6b32bc3 | 1671 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
b411b363 | 1672 | |
b30ab791 AG |
1673 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1674 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1675 | else |
b30ab791 | 1676 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1677 | |
b30ab791 | 1678 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1679 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1680 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1681 | |
b30ab791 AG |
1682 | device->send_cnt = 0; |
1683 | device->recv_cnt = 0; | |
1684 | device->read_cnt = 0; | |
1685 | device->writ_cnt = 0; | |
b411b363 | 1686 | |
b30ab791 | 1687 | drbd_reconsider_max_bio_size(device); |
b411b363 PR |
1688 | |
1689 | /* If I am currently not R_PRIMARY, | |
1690 | * but meta data primary indicator is set, | |
1691 | * I just now recover from a hard crash, | |
1692 | * and have been R_PRIMARY before that crash. | |
1693 | * | |
1694 | * Now, if I had no connection before that crash | |
1695 | * (have been degraded R_PRIMARY), chances are that | |
1696 | * I won't find my peer now either. | |
1697 | * | |
1698 | * In that case, and _only_ in that case, | |
1699 | * we use the degr-wfc-timeout instead of the default, | |
1700 | * so we can automatically recover from a crash of a | |
1701 | * degraded but active "cluster" after a certain timeout. | |
1702 | */ | |
b30ab791 AG |
1703 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1704 | if (device->state.role != R_PRIMARY && | |
1705 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1706 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1707 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1708 | |
b30ab791 | 1709 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1710 | if (dd <= DS_ERROR) { |
b411b363 PR |
1711 | retcode = ERR_NOMEM_BITMAP; |
1712 | goto force_diskless_dec; | |
e96c9633 | 1713 | } else if (dd == DS_GREW) |
b30ab791 | 1714 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1715 | |
b30ab791 AG |
1716 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1717 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1718 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1719 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1720 | "(aka FullSync)\n"); |
b30ab791 | 1721 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1722 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1723 | retcode = ERR_IO_MD_DISK; |
1724 | goto force_diskless_dec; | |
1725 | } | |
1726 | } else { | |
b30ab791 | 1727 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1728 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1729 | retcode = ERR_IO_MD_DISK; |
1730 | goto force_diskless_dec; | |
1731 | } | |
b411b363 PR |
1732 | } |
1733 | ||
b30ab791 AG |
1734 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1735 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1736 | |
0500813f | 1737 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1738 | os = drbd_read_state(device); |
78bae59b | 1739 | ns = os; |
b411b363 PR |
1740 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1741 | otherwise investigate MDF_WasUpToDate... | |
1742 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1743 | otherwise into D_CONSISTENT state. | |
1744 | */ | |
b30ab791 AG |
1745 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1746 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1747 | ns.disk = D_CONSISTENT; |
1748 | else | |
1749 | ns.disk = D_OUTDATED; | |
1750 | } else { | |
1751 | ns.disk = D_INCONSISTENT; | |
1752 | } | |
1753 | ||
b30ab791 | 1754 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1755 | ns.pdsk = D_OUTDATED; |
1756 | ||
daeda1cc PR |
1757 | rcu_read_lock(); |
1758 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1759 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1760 | ns.disk = D_UP_TO_DATE; |
1761 | ||
1762 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1763 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1764 | this point, because drbd_request_state() modifies these | |
1765 | flags. */ | |
1766 | ||
b30ab791 AG |
1767 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1768 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1769 | else |
b30ab791 | 1770 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1771 | |
1772 | rcu_read_unlock(); | |
1773 | ||
b411b363 PR |
1774 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1775 | state after the negotiation phase. */ | |
b30ab791 AG |
1776 | if (device->state.conn == C_CONNECTED) { |
1777 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1778 | ns.i = os.i; |
1779 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1780 | |
1781 | /* We expect to receive up-to-date UUIDs soon. | |
1782 | To avoid a race in receive_state, free p_uuid while | |
1783 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1784 | kfree(device->p_uuid); |
1785 | device->p_uuid = NULL; | |
b411b363 PR |
1786 | } |
1787 | ||
b30ab791 | 1788 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1789 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1790 | |
1791 | if (rv < SS_SUCCESS) | |
1792 | goto force_diskless_dec; | |
1793 | ||
b30ab791 | 1794 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1795 | |
b30ab791 AG |
1796 | if (device->state.role == R_PRIMARY) |
1797 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1798 | else |
b30ab791 | 1799 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1800 | |
b30ab791 AG |
1801 | drbd_md_mark_dirty(device); |
1802 | drbd_md_sync(device); | |
b411b363 | 1803 | |
b30ab791 AG |
1804 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1805 | put_ldev(device); | |
a6b32bc3 | 1806 | conn_reconfig_done(first_peer_device(device)->connection); |
3b98c0c2 | 1807 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1808 | return 0; |
1809 | ||
1810 | force_diskless_dec: | |
b30ab791 | 1811 | put_ldev(device); |
b411b363 | 1812 | force_diskless: |
b30ab791 AG |
1813 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1814 | drbd_md_sync(device); | |
b411b363 | 1815 | fail: |
a6b32bc3 | 1816 | conn_reconfig_done(first_peer_device(device)->connection); |
b411b363 | 1817 | if (nbc) { |
e525fd89 TH |
1818 | if (nbc->backing_bdev) |
1819 | blkdev_put(nbc->backing_bdev, | |
1820 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1821 | if (nbc->md_bdev) | |
1822 | blkdev_put(nbc->md_bdev, | |
1823 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1824 | kfree(nbc); |
1825 | } | |
daeda1cc | 1826 | kfree(new_disk_conf); |
b411b363 | 1827 | lc_destroy(resync_lru); |
9958c857 | 1828 | kfree(new_plan); |
b411b363 | 1829 | |
40cbf085 | 1830 | finish: |
3b98c0c2 | 1831 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1832 | return 0; |
1833 | } | |
1834 | ||
b30ab791 | 1835 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1836 | { |
19f83c76 | 1837 | enum drbd_state_rv retcode; |
9a0d9d03 | 1838 | int ret; |
02ee8f95 | 1839 | |
cdfda633 | 1840 | if (force) { |
b30ab791 AG |
1841 | set_bit(FORCE_DETACH, &device->flags); |
1842 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1843 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1844 | goto out; |
1845 | } | |
1846 | ||
b30ab791 AG |
1847 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
1848 | drbd_md_get_buffer(device); /* make sure there is no in-flight meta-data IO */ | |
1849 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
1850 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1851 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1852 | ret = wait_event_interruptible(device->misc_wait, |
1853 | device->state.disk != D_FAILED); | |
1854 | drbd_resume_io(device); | |
9b2f61ae | 1855 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1856 | retcode = SS_NOTHING_TO_DO; |
1857 | if (ret) | |
1858 | retcode = ERR_INTR; | |
02ee8f95 | 1859 | out: |
85f75dd7 | 1860 | return retcode; |
b411b363 PR |
1861 | } |
1862 | ||
82f59cc6 LE |
1863 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1864 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1865 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1866 | * internal references as well. | |
1867 | * Only then we have finally detached. */ | |
3b98c0c2 | 1868 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1869 | { |
116676ca | 1870 | enum drbd_ret_code retcode; |
cdfda633 PR |
1871 | struct detach_parms parms = { }; |
1872 | int err; | |
b411b363 | 1873 | |
3b98c0c2 LE |
1874 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1875 | if (!adm_ctx.reply_skb) | |
1876 | return retcode; | |
1877 | if (retcode != NO_ERROR) | |
1878 | goto out; | |
b411b363 | 1879 | |
cdfda633 PR |
1880 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1881 | err = detach_parms_from_attrs(&parms, info); | |
1882 | if (err) { | |
1883 | retcode = ERR_MANDATORY_TAG; | |
1884 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1885 | goto out; | |
1886 | } | |
b411b363 PR |
1887 | } |
1888 | ||
b30ab791 | 1889 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
3b98c0c2 LE |
1890 | out: |
1891 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1892 | return 0; |
1893 | } | |
b411b363 | 1894 | |
bde89a9e | 1895 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1896 | { |
c06ece6b | 1897 | struct drbd_peer_device *peer_device; |
695d08fa | 1898 | bool rv = false; |
f399002e LE |
1899 | int vnr; |
1900 | ||
695d08fa | 1901 | rcu_read_lock(); |
c06ece6b AG |
1902 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1903 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1904 | if (device->state.conn == C_SYNC_SOURCE || |
1905 | device->state.conn == C_SYNC_TARGET || | |
1906 | device->state.conn == C_PAUSED_SYNC_S || | |
1907 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1908 | rv = true; |
1909 | break; | |
1910 | } | |
b411b363 | 1911 | } |
695d08fa | 1912 | rcu_read_unlock(); |
b411b363 | 1913 | |
695d08fa | 1914 | return rv; |
f399002e | 1915 | } |
47ff2d0a | 1916 | |
bde89a9e | 1917 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1918 | { |
c06ece6b | 1919 | struct drbd_peer_device *peer_device; |
695d08fa | 1920 | bool rv = false; |
f399002e LE |
1921 | int vnr; |
1922 | ||
695d08fa | 1923 | rcu_read_lock(); |
c06ece6b AG |
1924 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1925 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1926 | if (device->state.conn == C_VERIFY_S || |
1927 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1928 | rv = true; |
1929 | break; | |
47ff2d0a PR |
1930 | } |
1931 | } | |
695d08fa | 1932 | rcu_read_unlock(); |
b411b363 | 1933 | |
695d08fa | 1934 | return rv; |
f399002e | 1935 | } |
422028b1 | 1936 | |
cd64397c | 1937 | static enum drbd_ret_code |
270eb5c9 | 1938 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 1939 | { |
c06ece6b | 1940 | struct drbd_peer_device *peer_device; |
cd64397c | 1941 | int i; |
b411b363 | 1942 | |
270eb5c9 AG |
1943 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
1944 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 1945 | return ERR_NEED_APV_100; |
b411b363 | 1946 | |
270eb5c9 | 1947 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
1948 | return ERR_NEED_APV_100; |
1949 | ||
270eb5c9 | 1950 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 1951 | return ERR_NEED_APV_100; |
b411b363 PR |
1952 | } |
1953 | ||
270eb5c9 | 1954 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
1955 | conn_highest_role(connection) == R_PRIMARY && |
1956 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 1957 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 1958 | |
270eb5c9 AG |
1959 | if (new_net_conf->two_primaries && |
1960 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
1961 | return ERR_NOT_PROTO_C; |
1962 | ||
c06ece6b AG |
1963 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
1964 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1965 | if (get_ldev(device)) { |
1966 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
1967 | put_ldev(device); | |
270eb5c9 | 1968 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1969 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 1970 | } |
270eb5c9 | 1971 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 1972 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
1973 | } |
1974 | ||
270eb5c9 | 1975 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 1976 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 1977 | |
cd64397c PR |
1978 | return NO_ERROR; |
1979 | } | |
b411b363 | 1980 | |
44ed167d | 1981 | static enum drbd_ret_code |
270eb5c9 | 1982 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
1983 | { |
1984 | static enum drbd_ret_code rv; | |
c06ece6b | 1985 | struct drbd_peer_device *peer_device; |
44ed167d | 1986 | int i; |
b411b363 | 1987 | |
44ed167d | 1988 | rcu_read_lock(); |
270eb5c9 | 1989 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 1990 | rcu_read_unlock(); |
b411b363 | 1991 | |
bde89a9e | 1992 | /* connection->volumes protected by genl_lock() here */ |
c06ece6b AG |
1993 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
1994 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1995 | if (!device->bitmap) { |
1996 | if (drbd_bm_init(device)) | |
44ed167d | 1997 | return ERR_NOMEM; |
b411b363 PR |
1998 | } |
1999 | } | |
2000 | ||
44ed167d PR |
2001 | return rv; |
2002 | } | |
b411b363 | 2003 | |
0fd0ea06 PR |
2004 | struct crypto { |
2005 | struct crypto_hash *verify_tfm; | |
2006 | struct crypto_hash *csums_tfm; | |
2007 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2008 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2009 | }; |
b411b363 | 2010 | |
0fd0ea06 | 2011 | static int |
4b6ad6d4 | 2012 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2013 | { |
2014 | if (!tfm_name[0]) | |
2015 | return NO_ERROR; | |
b411b363 | 2016 | |
0fd0ea06 PR |
2017 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2018 | if (IS_ERR(*tfm)) { | |
2019 | *tfm = NULL; | |
2020 | return err_alg; | |
b411b363 | 2021 | } |
b411b363 | 2022 | |
0fd0ea06 PR |
2023 | return NO_ERROR; |
2024 | } | |
b411b363 | 2025 | |
0fd0ea06 | 2026 | static enum drbd_ret_code |
270eb5c9 | 2027 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2028 | { |
2029 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2030 | enum drbd_ret_code rv; | |
0fd0ea06 | 2031 | |
270eb5c9 | 2032 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2033 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2034 | if (rv != NO_ERROR) |
2035 | return rv; | |
270eb5c9 | 2036 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2037 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2038 | if (rv != NO_ERROR) |
2039 | return rv; | |
270eb5c9 | 2040 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2041 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2042 | if (rv != NO_ERROR) |
2043 | return rv; | |
270eb5c9 | 2044 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2045 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2046 | new_net_conf->cram_hmac_alg); |
b411b363 | 2047 | |
4b6ad6d4 AG |
2048 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2049 | ERR_AUTH_ALG); | |
b411b363 PR |
2050 | } |
2051 | ||
0fd0ea06 PR |
2052 | return rv; |
2053 | } | |
b411b363 | 2054 | |
0fd0ea06 PR |
2055 | static void free_crypto(struct crypto *crypto) |
2056 | { | |
0fd0ea06 | 2057 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2058 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2059 | crypto_free_hash(crypto->csums_tfm); |
2060 | crypto_free_hash(crypto->verify_tfm); | |
2061 | } | |
b411b363 | 2062 | |
f399002e LE |
2063 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2064 | { | |
2065 | enum drbd_ret_code retcode; | |
bde89a9e | 2066 | struct drbd_connection *connection; |
270eb5c9 | 2067 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2068 | int err; |
2069 | int ovr; /* online verify running */ | |
2070 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2071 | struct crypto crypto = { }; |
b411b363 | 2072 | |
089c075d | 2073 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2074 | if (!adm_ctx.reply_skb) |
2075 | return retcode; | |
2076 | if (retcode != NO_ERROR) | |
2077 | goto out; | |
b411b363 | 2078 | |
bde89a9e | 2079 | connection = adm_ctx.connection; |
b411b363 | 2080 | |
270eb5c9 AG |
2081 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2082 | if (!new_net_conf) { | |
f399002e LE |
2083 | retcode = ERR_NOMEM; |
2084 | goto out; | |
2085 | } | |
b411b363 | 2086 | |
bde89a9e | 2087 | conn_reconfig_start(connection); |
b411b363 | 2088 | |
bde89a9e | 2089 | mutex_lock(&connection->data.mutex); |
0500813f | 2090 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2091 | old_net_conf = connection->net_conf; |
2561b9c1 | 2092 | |
270eb5c9 | 2093 | if (!old_net_conf) { |
f399002e LE |
2094 | drbd_msg_put_info("net conf missing, try connect"); |
2095 | retcode = ERR_INVALID_REQUEST; | |
2561b9c1 PR |
2096 | goto fail; |
2097 | } | |
2098 | ||
270eb5c9 | 2099 | *new_net_conf = *old_net_conf; |
5979e361 | 2100 | if (should_set_defaults(info)) |
270eb5c9 | 2101 | set_net_conf_defaults(new_net_conf); |
f399002e | 2102 | |
270eb5c9 | 2103 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2104 | if (err && err != -ENOMSG) { |
f399002e LE |
2105 | retcode = ERR_MANDATORY_TAG; |
2106 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2107 | goto fail; | |
2561b9c1 | 2108 | } |
b411b363 | 2109 | |
270eb5c9 | 2110 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2111 | if (retcode != NO_ERROR) |
2112 | goto fail; | |
b411b363 | 2113 | |
f399002e | 2114 | /* re-sync running */ |
bde89a9e | 2115 | rsr = conn_resync_running(connection); |
270eb5c9 | 2116 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2117 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2118 | goto fail; |
b411b363 PR |
2119 | } |
2120 | ||
f399002e | 2121 | /* online verify running */ |
bde89a9e | 2122 | ovr = conn_ov_running(connection); |
270eb5c9 | 2123 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2124 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2125 | goto fail; |
f399002e | 2126 | } |
b411b363 | 2127 | |
270eb5c9 | 2128 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2129 | if (retcode != NO_ERROR) |
b411b363 | 2130 | goto fail; |
f399002e | 2131 | |
270eb5c9 | 2132 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2133 | |
2134 | if (!rsr) { | |
bde89a9e AG |
2135 | crypto_free_hash(connection->csums_tfm); |
2136 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2137 | crypto.csums_tfm = NULL; |
f399002e LE |
2138 | } |
2139 | if (!ovr) { | |
bde89a9e AG |
2140 | crypto_free_hash(connection->verify_tfm); |
2141 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2142 | crypto.verify_tfm = NULL; |
b411b363 PR |
2143 | } |
2144 | ||
bde89a9e AG |
2145 | crypto_free_hash(connection->integrity_tfm); |
2146 | connection->integrity_tfm = crypto.integrity_tfm; | |
2147 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2148 | /* Do this without trying to take connection->data.mutex again. */ | |
2149 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2150 | |
bde89a9e AG |
2151 | crypto_free_hash(connection->cram_hmac_tfm); |
2152 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2153 | |
0500813f | 2154 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2155 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2156 | synchronize_rcu(); |
270eb5c9 | 2157 | kfree(old_net_conf); |
91fd4dad | 2158 | |
69a22773 AG |
2159 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2160 | struct drbd_peer_device *peer_device; | |
2161 | int vnr; | |
2162 | ||
2163 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2164 | drbd_send_sync_param(peer_device); | |
2165 | } | |
f399002e | 2166 | |
91fd4dad PR |
2167 | goto done; |
2168 | ||
b411b363 | 2169 | fail: |
0500813f | 2170 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2171 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2172 | free_crypto(&crypto); |
270eb5c9 | 2173 | kfree(new_net_conf); |
91fd4dad | 2174 | done: |
bde89a9e | 2175 | conn_reconfig_done(connection); |
f399002e LE |
2176 | out: |
2177 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2178 | return 0; |
2179 | } | |
2180 | ||
3b98c0c2 | 2181 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2182 | { |
c06ece6b | 2183 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2184 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2185 | struct crypto crypto = { }; |
77c556f6 | 2186 | struct drbd_resource *resource; |
bde89a9e | 2187 | struct drbd_connection *connection; |
3b98c0c2 LE |
2188 | enum drbd_ret_code retcode; |
2189 | int i; | |
2190 | int err; | |
b411b363 | 2191 | |
44e52cfa | 2192 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2193 | |
3b98c0c2 LE |
2194 | if (!adm_ctx.reply_skb) |
2195 | return retcode; | |
2196 | if (retcode != NO_ERROR) | |
2197 | goto out; | |
089c075d AG |
2198 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
2199 | drbd_msg_put_info("connection endpoint(s) missing"); | |
2200 | retcode = ERR_INVALID_REQUEST; | |
2201 | goto out; | |
2202 | } | |
b411b363 | 2203 | |
089c075d AG |
2204 | /* No need for _rcu here. All reconfiguration is |
2205 | * strictly serialized on genl_lock(). We are protected against | |
2206 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2207 | for_each_resource(resource, &drbd_resources) { |
2208 | for_each_connection(connection, resource) { | |
2209 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2210 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2211 | connection->my_addr_len)) { | |
2212 | retcode = ERR_LOCAL_ADDR; | |
2213 | goto out; | |
2214 | } | |
b411b363 | 2215 | |
77c556f6 AG |
2216 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2217 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2218 | connection->peer_addr_len)) { | |
2219 | retcode = ERR_PEER_ADDR; | |
2220 | goto out; | |
2221 | } | |
089c075d | 2222 | } |
b411b363 PR |
2223 | } |
2224 | ||
3ab706fe | 2225 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2226 | conn_reconfig_start(connection); |
b411b363 | 2227 | |
bde89a9e | 2228 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2229 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2230 | goto fail; |
2231 | } | |
2232 | ||
a209b4ae | 2233 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2234 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2235 | if (!new_net_conf) { | |
b411b363 | 2236 | retcode = ERR_NOMEM; |
b411b363 PR |
2237 | goto fail; |
2238 | } | |
2239 | ||
270eb5c9 | 2240 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2241 | |
270eb5c9 | 2242 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2243 | if (err && err != -ENOMSG) { |
b411b363 | 2244 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2245 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2246 | goto fail; |
2247 | } | |
2248 | ||
270eb5c9 | 2249 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2250 | if (retcode != NO_ERROR) |
422028b1 | 2251 | goto fail; |
b411b363 | 2252 | |
270eb5c9 | 2253 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2254 | if (retcode != NO_ERROR) |
2255 | goto fail; | |
b411b363 | 2256 | |
270eb5c9 | 2257 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2258 | |
b5043c5e | 2259 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2260 | |
0500813f | 2261 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2262 | old_net_conf = connection->net_conf; |
2263 | if (old_net_conf) { | |
b411b363 | 2264 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2265 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2266 | goto fail; |
2267 | } | |
270eb5c9 | 2268 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2269 | |
bde89a9e AG |
2270 | conn_free_crypto(connection); |
2271 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2272 | connection->integrity_tfm = crypto.integrity_tfm; | |
2273 | connection->csums_tfm = crypto.csums_tfm; | |
2274 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2275 | |
bde89a9e AG |
2276 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2277 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2278 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2279 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2280 | |
0500813f | 2281 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2282 | |
695d08fa | 2283 | rcu_read_lock(); |
c06ece6b AG |
2284 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2285 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2286 | device->send_cnt = 0; |
2287 | device->recv_cnt = 0; | |
b411b363 | 2288 | } |
695d08fa | 2289 | rcu_read_unlock(); |
b411b363 | 2290 | |
bde89a9e | 2291 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2292 | |
bde89a9e | 2293 | conn_reconfig_done(connection); |
3b98c0c2 | 2294 | drbd_adm_finish(info, retcode); |
b411b363 | 2295 | return 0; |
b411b363 | 2296 | |
b411b363 | 2297 | fail: |
0fd0ea06 | 2298 | free_crypto(&crypto); |
270eb5c9 | 2299 | kfree(new_net_conf); |
b411b363 | 2300 | |
bde89a9e | 2301 | conn_reconfig_done(connection); |
3b98c0c2 LE |
2302 | out: |
2303 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2304 | return 0; |
2305 | } | |
2306 | ||
bde89a9e | 2307 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2308 | { |
2309 | enum drbd_state_rv rv; | |
85f75dd7 | 2310 | |
bde89a9e | 2311 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2312 | force ? CS_HARD : 0); |
85f75dd7 LE |
2313 | |
2314 | switch (rv) { | |
2315 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2316 | break; |
85f75dd7 LE |
2317 | case SS_ALREADY_STANDALONE: |
2318 | return SS_SUCCESS; | |
2319 | case SS_PRIMARY_NOP: | |
2320 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2321 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2322 | |
2323 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2324 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2325 | |
85f75dd7 LE |
2326 | break; |
2327 | case SS_CW_FAILED_BY_PEER: | |
2328 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2329 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2330 | disk, D_OUTDATED), 0); |
2331 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2332 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2333 | CS_HARD); |
b411b363 | 2334 | } |
85f75dd7 LE |
2335 | break; |
2336 | default:; | |
2337 | /* no special handling necessary */ | |
2338 | } | |
2339 | ||
f3dfa40a LE |
2340 | if (rv >= SS_SUCCESS) { |
2341 | enum drbd_state_rv rv2; | |
2342 | /* No one else can reconfigure the network while I am here. | |
2343 | * The state handling only uses drbd_thread_stop_nowait(), | |
2344 | * we want to really wait here until the receiver is no more. | |
2345 | */ | |
9693da23 | 2346 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2347 | |
2348 | /* Race breaker. This additional state change request may be | |
2349 | * necessary, if this was a forced disconnect during a receiver | |
2350 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2351 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2352 | * C_STANDALONE already, now, and this becomes a no-op. |
2353 | */ | |
bde89a9e | 2354 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2355 | CS_VERBOSE | CS_HARD); |
2356 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2357 | drbd_err(connection, |
f3dfa40a LE |
2358 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2359 | rv2); | |
b411b363 | 2360 | } |
85f75dd7 LE |
2361 | return rv; |
2362 | } | |
b411b363 | 2363 | |
3b98c0c2 | 2364 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2365 | { |
3b98c0c2 | 2366 | struct disconnect_parms parms; |
bde89a9e | 2367 | struct drbd_connection *connection; |
85f75dd7 | 2368 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2369 | enum drbd_ret_code retcode; |
2370 | int err; | |
2561b9c1 | 2371 | |
089c075d | 2372 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2373 | if (!adm_ctx.reply_skb) |
2374 | return retcode; | |
2375 | if (retcode != NO_ERROR) | |
2561b9c1 | 2376 | goto fail; |
b411b363 | 2377 | |
bde89a9e | 2378 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2379 | memset(&parms, 0, sizeof(parms)); |
2380 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2381 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2382 | if (err) { |
2383 | retcode = ERR_MANDATORY_TAG; | |
2384 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
b411b363 PR |
2385 | goto fail; |
2386 | } | |
2387 | } | |
2388 | ||
bde89a9e | 2389 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2390 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2391 | retcode = rv; /* FIXME: Type mismatch. */ |
2392 | else | |
2393 | retcode = NO_ERROR; | |
b411b363 | 2394 | fail: |
3b98c0c2 | 2395 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2396 | return 0; |
2397 | } | |
2398 | ||
b30ab791 | 2399 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2400 | { |
2401 | int iass; /* I am sync source */ | |
2402 | ||
d0180171 | 2403 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2404 | if (device->state.role != device->state.peer) |
2405 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2406 | else |
a6b32bc3 | 2407 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2408 | |
2409 | if (iass) | |
b30ab791 | 2410 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2411 | else |
b30ab791 | 2412 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2413 | } |
2414 | ||
3b98c0c2 | 2415 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2416 | { |
daeda1cc | 2417 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2418 | struct resize_parms rs; |
b30ab791 | 2419 | struct drbd_device *device; |
3b98c0c2 | 2420 | enum drbd_ret_code retcode; |
b411b363 | 2421 | enum determine_dev_size dd; |
d752b269 | 2422 | bool change_al_layout = false; |
6495d2c6 | 2423 | enum dds_flags ddsf; |
daeda1cc | 2424 | sector_t u_size; |
3b98c0c2 | 2425 | int err; |
b411b363 | 2426 | |
3b98c0c2 LE |
2427 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2428 | if (!adm_ctx.reply_skb) | |
2429 | return retcode; | |
2430 | if (retcode != NO_ERROR) | |
b411b363 | 2431 | goto fail; |
3b98c0c2 | 2432 | |
b30ab791 AG |
2433 | device = adm_ctx.device; |
2434 | if (!get_ldev(device)) { | |
d752b269 PR |
2435 | retcode = ERR_NO_DISK; |
2436 | goto fail; | |
2437 | } | |
2438 | ||
3b98c0c2 | 2439 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2440 | rs.al_stripes = device->ldev->md.al_stripes; |
2441 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2442 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2443 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2444 | if (err) { |
3b98c0c2 LE |
2445 | retcode = ERR_MANDATORY_TAG; |
2446 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
d752b269 | 2447 | goto fail_ldev; |
b411b363 PR |
2448 | } |
2449 | } | |
2450 | ||
b30ab791 | 2451 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2452 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2453 | goto fail_ldev; |
b411b363 | 2454 | } |
b411b363 | 2455 | |
b30ab791 AG |
2456 | if (device->state.role == R_SECONDARY && |
2457 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2458 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2459 | goto fail_ldev; |
b411b363 | 2460 | } |
b411b363 | 2461 | |
a6b32bc3 | 2462 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2463 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2464 | goto fail_ldev; |
6495d2c6 PR |
2465 | } |
2466 | ||
daeda1cc | 2467 | rcu_read_lock(); |
b30ab791 | 2468 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2469 | rcu_read_unlock(); |
2470 | if (u_size != (sector_t)rs.resize_size) { | |
2471 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2472 | if (!new_disk_conf) { | |
778f271d | 2473 | retcode = ERR_NOMEM; |
9bcd2521 | 2474 | goto fail_ldev; |
778f271d PR |
2475 | } |
2476 | } | |
2477 | ||
b30ab791 AG |
2478 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2479 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2480 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2481 | ||
2482 | if (al_size_k > (16 * 1024 * 1024)) { | |
2483 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2484 | goto fail_ldev; | |
2485 | } | |
2486 | ||
2487 | if (al_size_k < MD_32kB_SECT/2) { | |
2488 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2489 | goto fail_ldev; | |
2490 | } | |
2491 | ||
b30ab791 | 2492 | if (device->state.conn != C_CONNECTED) { |
d752b269 PR |
2493 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2494 | goto fail_ldev; | |
2495 | } | |
2496 | ||
2497 | change_al_layout = true; | |
2498 | } | |
2499 | ||
b30ab791 AG |
2500 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2501 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2502 | |
daeda1cc | 2503 | if (new_disk_conf) { |
0500813f | 2504 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2505 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2506 | *new_disk_conf = *old_disk_conf; |
2507 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2508 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2509 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2510 | synchronize_rcu(); |
2511 | kfree(old_disk_conf); | |
b411b363 PR |
2512 | } |
2513 | ||
6495d2c6 | 2514 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2515 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2516 | drbd_md_sync(device); | |
2517 | put_ldev(device); | |
e96c9633 | 2518 | if (dd == DS_ERROR) { |
b411b363 PR |
2519 | retcode = ERR_NOMEM_BITMAP; |
2520 | goto fail; | |
d752b269 PR |
2521 | } else if (dd == DS_ERROR_SPACE_MD) { |
2522 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2523 | goto fail; | |
2524 | } else if (dd == DS_ERROR_SHRINK) { | |
2525 | retcode = ERR_IMPLICIT_SHRINK; | |
2526 | goto fail; | |
b411b363 | 2527 | } |
778f271d | 2528 | |
b30ab791 | 2529 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2530 | if (dd == DS_GREW) |
b30ab791 | 2531 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2532 | |
69a22773 AG |
2533 | drbd_send_uuids(first_peer_device(device)); |
2534 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2535 | } |
2536 | ||
b411b363 | 2537 | fail: |
3b98c0c2 | 2538 | drbd_adm_finish(info, retcode); |
b411b363 | 2539 | return 0; |
b411b363 | 2540 | |
9bcd2521 | 2541 | fail_ldev: |
b30ab791 | 2542 | put_ldev(device); |
9bcd2521 | 2543 | goto fail; |
b411b363 | 2544 | } |
b411b363 | 2545 | |
f399002e | 2546 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2547 | { |
3b98c0c2 | 2548 | enum drbd_ret_code retcode; |
b57a1e27 | 2549 | struct res_opts res_opts; |
f399002e | 2550 | int err; |
b411b363 | 2551 | |
44e52cfa | 2552 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2553 | if (!adm_ctx.reply_skb) |
2554 | return retcode; | |
2555 | if (retcode != NO_ERROR) | |
2556 | goto fail; | |
b411b363 | 2557 | |
eb6bea67 | 2558 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2559 | if (should_set_defaults(info)) |
b966b5dd | 2560 | set_res_opts_defaults(&res_opts); |
b411b363 | 2561 | |
b57a1e27 | 2562 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2563 | if (err && err != -ENOMSG) { |
b411b363 | 2564 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2565 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2566 | goto fail; |
2567 | } | |
2568 | ||
eb6bea67 | 2569 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2570 | if (err) { |
2571 | retcode = ERR_INVALID_REQUEST; | |
2572 | if (err == -ENOMEM) | |
2573 | retcode = ERR_NOMEM; | |
b411b363 PR |
2574 | } |
2575 | ||
b411b363 | 2576 | fail: |
3b98c0c2 | 2577 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2578 | return 0; |
2579 | } | |
2580 | ||
3b98c0c2 | 2581 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2582 | { |
b30ab791 | 2583 | struct drbd_device *device; |
3b98c0c2 LE |
2584 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2585 | ||
2586 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2587 | if (!adm_ctx.reply_skb) | |
2588 | return retcode; | |
2589 | if (retcode != NO_ERROR) | |
2590 | goto out; | |
2591 | ||
b30ab791 | 2592 | device = adm_ctx.device; |
b411b363 | 2593 | |
194bfb32 | 2594 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2595 | * resync just being finished, wait for it before requesting a new resync. |
2596 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2597 | drbd_suspend_io(device); |
2598 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2599 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2600 | |
0b2dafcd PR |
2601 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2602 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2603 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2604 | */ |
b30ab791 AG |
2605 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2606 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2607 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2608 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2609 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2610 | retcode = ERR_IO_MD_DISK; | |
2611 | } | |
2612 | } else | |
b30ab791 AG |
2613 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2614 | drbd_resume_io(device); | |
b411b363 | 2615 | |
3b98c0c2 LE |
2616 | out: |
2617 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2618 | return 0; |
2619 | } | |
2620 | ||
3b98c0c2 LE |
2621 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2622 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2623 | { |
3b98c0c2 | 2624 | enum drbd_ret_code retcode; |
194bfb32 | 2625 | |
3b98c0c2 LE |
2626 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2627 | if (!adm_ctx.reply_skb) | |
2628 | return retcode; | |
2629 | if (retcode != NO_ERROR) | |
2630 | goto out; | |
b411b363 | 2631 | |
b30ab791 | 2632 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
3b98c0c2 LE |
2633 | out: |
2634 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2635 | return 0; |
2636 | } | |
2637 | ||
b30ab791 | 2638 | static int drbd_bmio_set_susp_al(struct drbd_device *device) |
0778286a PR |
2639 | { |
2640 | int rv; | |
2641 | ||
b30ab791 AG |
2642 | rv = drbd_bmio_set_n_write(device); |
2643 | drbd_suspend_al(device); | |
0778286a PR |
2644 | return rv; |
2645 | } | |
2646 | ||
3b98c0c2 | 2647 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2648 | { |
25b0d6c8 | 2649 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2650 | struct drbd_device *device; |
25b0d6c8 PR |
2651 | |
2652 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2653 | if (!adm_ctx.reply_skb) | |
2654 | return retcode; | |
2655 | if (retcode != NO_ERROR) | |
2656 | goto out; | |
2657 | ||
b30ab791 | 2658 | device = adm_ctx.device; |
b411b363 | 2659 | |
194bfb32 | 2660 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2661 | * resync just being finished, wait for it before requesting a new resync. |
2662 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2663 | drbd_suspend_io(device); |
2664 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2665 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2666 | |
0b2dafcd PR |
2667 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2668 | * in the bitmap. Otherwise, try to start a resync handshake | |
2669 | * as sync source for full sync. | |
2670 | */ | |
b30ab791 | 2671 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2672 | /* The peer will get a resync upon connect anyways. Just make that |
2673 | into a full resync. */ | |
b30ab791 | 2674 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2675 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2676 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2677 | "set_n_write from invalidate_peer", |
2678 | BM_LOCKED_SET_ALLOWED)) | |
2679 | retcode = ERR_IO_MD_DISK; | |
2680 | } | |
2681 | } else | |
b30ab791 AG |
2682 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2683 | drbd_resume_io(device); | |
b411b363 | 2684 | |
25b0d6c8 PR |
2685 | out: |
2686 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2687 | return 0; |
2688 | } | |
2689 | ||
3b98c0c2 | 2690 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2691 | { |
3b98c0c2 | 2692 | enum drbd_ret_code retcode; |
b411b363 | 2693 | |
3b98c0c2 LE |
2694 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2695 | if (!adm_ctx.reply_skb) | |
2696 | return retcode; | |
2697 | if (retcode != NO_ERROR) | |
2698 | goto out; | |
b411b363 | 2699 | |
b30ab791 | 2700 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 LE |
2701 | retcode = ERR_PAUSE_IS_SET; |
2702 | out: | |
2703 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2704 | return 0; |
2705 | } | |
2706 | ||
3b98c0c2 | 2707 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2708 | { |
da9fbc27 | 2709 | union drbd_dev_state s; |
3b98c0c2 LE |
2710 | enum drbd_ret_code retcode; |
2711 | ||
2712 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2713 | if (!adm_ctx.reply_skb) | |
2714 | return retcode; | |
2715 | if (retcode != NO_ERROR) | |
2716 | goto out; | |
b411b363 | 2717 | |
b30ab791 AG |
2718 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2719 | s = adm_ctx.device->state; | |
cd88d030 PR |
2720 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2721 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2722 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2723 | } else { | |
2724 | retcode = ERR_PAUSE_IS_CLEAR; | |
2725 | } | |
2726 | } | |
b411b363 | 2727 | |
3b98c0c2 LE |
2728 | out: |
2729 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2730 | return 0; |
2731 | } | |
2732 | ||
3b98c0c2 | 2733 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2734 | { |
3b98c0c2 | 2735 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2736 | } |
2737 | ||
3b98c0c2 | 2738 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2739 | { |
b30ab791 | 2740 | struct drbd_device *device; |
3b98c0c2 LE |
2741 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2742 | ||
2743 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2744 | if (!adm_ctx.reply_skb) | |
2745 | return retcode; | |
2746 | if (retcode != NO_ERROR) | |
2747 | goto out; | |
2748 | ||
b30ab791 AG |
2749 | device = adm_ctx.device; |
2750 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2751 | drbd_uuid_new_current(device); | |
2752 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2753 | } |
b30ab791 AG |
2754 | drbd_suspend_io(device); |
2755 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2756 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2757 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2758 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2759 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2760 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2761 | } |
b30ab791 | 2762 | drbd_resume_io(device); |
265be2d0 | 2763 | |
3b98c0c2 LE |
2764 | out: |
2765 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2766 | return 0; |
2767 | } | |
2768 | ||
3b98c0c2 | 2769 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2770 | { |
3b98c0c2 | 2771 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2772 | } |
2773 | ||
251b8f8e AG |
2774 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2775 | struct drbd_resource *resource, | |
2776 | struct drbd_connection *connection, | |
2777 | struct drbd_device *device) | |
b411b363 | 2778 | { |
543cc10b LE |
2779 | struct nlattr *nla; |
2780 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2781 | if (!nla) | |
2782 | goto nla_put_failure; | |
251b8f8e AG |
2783 | if (device && |
2784 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2785 | goto nla_put_failure; |
f597f6b8 | 2786 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 2787 | goto nla_put_failure; |
251b8f8e AG |
2788 | if (connection) { |
2789 | if (connection->my_addr_len && | |
2790 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2791 | goto nla_put_failure; | |
2792 | if (connection->peer_addr_len && | |
2793 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2794 | goto nla_put_failure; | |
2795 | } | |
543cc10b LE |
2796 | nla_nest_end(skb, nla); |
2797 | return 0; | |
b411b363 | 2798 | |
543cc10b LE |
2799 | nla_put_failure: |
2800 | if (nla) | |
2801 | nla_nest_cancel(skb, nla); | |
2802 | return -EMSGSIZE; | |
2803 | } | |
b411b363 | 2804 | |
251b8f8e AG |
2805 | /* |
2806 | * Return the connection of @resource if @resource has exactly one connection. | |
2807 | */ | |
2808 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2809 | { | |
2810 | struct list_head *connections = &resource->connections; | |
2811 | ||
2812 | if (list_empty(connections) || connections->next->next != connections) | |
2813 | return NULL; | |
2814 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2815 | } | |
2816 | ||
2817 | int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, | |
3b98c0c2 | 2818 | const struct sib_info *sib) |
b411b363 | 2819 | { |
251b8f8e | 2820 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2821 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2822 | struct nlattr *nla; | |
2823 | int got_ldev; | |
3b98c0c2 LE |
2824 | int err = 0; |
2825 | int exclude_sensitive; | |
2826 | ||
2827 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2828 | * to. So we better exclude_sensitive information. | |
2829 | * | |
2830 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2831 | * in the context of the requesting user process. Exclude sensitive | |
2832 | * information, unless current has superuser. | |
2833 | * | |
2834 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2835 | * relies on the current implementation of netlink_dump(), which | |
2836 | * executes the dump callback successively from netlink_recvmsg(), | |
2837 | * always in the context of the receiving process */ | |
2838 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2839 | ||
b30ab791 | 2840 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2841 | |
2842 | /* We need to add connection name and volume number information still. | |
2843 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 2844 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 2845 | goto nla_put_failure; |
3b98c0c2 | 2846 | |
eb6bea67 | 2847 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
2848 | goto nla_put_failure; |
2849 | ||
daeda1cc | 2850 | rcu_read_lock(); |
f9eb7bf4 AG |
2851 | if (got_ldev) { |
2852 | struct disk_conf *disk_conf; | |
44ed167d | 2853 | |
b30ab791 | 2854 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2855 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2856 | } | |
2857 | if (!err) { | |
2858 | struct net_conf *nc; | |
2859 | ||
a6b32bc3 | 2860 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2861 | if (nc) |
2862 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2863 | } | |
44ed167d PR |
2864 | rcu_read_unlock(); |
2865 | if (err) | |
2866 | goto nla_put_failure; | |
3b98c0c2 | 2867 | |
3b98c0c2 LE |
2868 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2869 | if (!nla) | |
2870 | goto nla_put_failure; | |
26ec9287 | 2871 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2872 | nla_put_u32(skb, T_current_state, device->state.i) || |
2873 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2874 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2875 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2876 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2877 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2878 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2879 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2880 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2881 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2882 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2883 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2884 | goto nla_put_failure; |
3b98c0c2 LE |
2885 | |
2886 | if (got_ldev) { | |
39a1aa7f | 2887 | int err; |
b411b363 | 2888 | |
b30ab791 AG |
2889 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2890 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2891 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2892 | |
2893 | if (err) | |
2894 | goto nla_put_failure; | |
2895 | ||
b30ab791 AG |
2896 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
2897 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
2898 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 2899 | goto nla_put_failure; |
b30ab791 AG |
2900 | if (C_SYNC_SOURCE <= device->state.conn && |
2901 | C_PAUSED_SYNC_T >= device->state.conn) { | |
2902 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
2903 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 2904 | goto nla_put_failure; |
3b98c0c2 | 2905 | } |
b411b363 | 2906 | } |
b411b363 | 2907 | |
3b98c0c2 LE |
2908 | if (sib) { |
2909 | switch(sib->sib_reason) { | |
2910 | case SIB_SYNC_PROGRESS: | |
2911 | case SIB_GET_STATUS_REPLY: | |
2912 | break; | |
2913 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2914 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2915 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2916 | goto nla_put_failure; | |
3b98c0c2 LE |
2917 | break; |
2918 | case SIB_HELPER_POST: | |
26ec9287 AG |
2919 | if (nla_put_u32(skb, T_helper_exit_code, |
2920 | sib->helper_exit_code)) | |
2921 | goto nla_put_failure; | |
3b98c0c2 LE |
2922 | /* fall through */ |
2923 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2924 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2925 | goto nla_put_failure; | |
3b98c0c2 LE |
2926 | break; |
2927 | } | |
b411b363 | 2928 | } |
3b98c0c2 | 2929 | nla_nest_end(skb, nla); |
b411b363 | 2930 | |
3b98c0c2 LE |
2931 | if (0) |
2932 | nla_put_failure: | |
2933 | err = -EMSGSIZE; | |
2934 | if (got_ldev) | |
b30ab791 | 2935 | put_ldev(device); |
3b98c0c2 | 2936 | return err; |
b411b363 PR |
2937 | } |
2938 | ||
3b98c0c2 | 2939 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2940 | { |
3b98c0c2 LE |
2941 | enum drbd_ret_code retcode; |
2942 | int err; | |
b411b363 | 2943 | |
3b98c0c2 LE |
2944 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2945 | if (!adm_ctx.reply_skb) | |
2946 | return retcode; | |
2947 | if (retcode != NO_ERROR) | |
2948 | goto out; | |
b411b363 | 2949 | |
b30ab791 | 2950 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
2951 | if (err) { |
2952 | nlmsg_free(adm_ctx.reply_skb); | |
2953 | return err; | |
b411b363 | 2954 | } |
3b98c0c2 LE |
2955 | out: |
2956 | drbd_adm_finish(info, retcode); | |
2957 | return 0; | |
b411b363 PR |
2958 | } |
2959 | ||
4b7a530f | 2960 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2961 | { |
b30ab791 | 2962 | struct drbd_device *device; |
3b98c0c2 | 2963 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
2964 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
2965 | struct drbd_resource *resource = NULL; | |
77c556f6 | 2966 | struct drbd_resource *tmp; |
543cc10b LE |
2967 | unsigned volume = cb->args[1]; |
2968 | ||
2969 | /* Open coded, deferred, iteration: | |
77c556f6 | 2970 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
2971 | * connection = "first connection of resource or undefined"; |
2972 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
2973 | * ... |
2974 | * } | |
2975 | * } | |
77c556f6 | 2976 | * where resource is cb->args[0]; |
543cc10b LE |
2977 | * and i is cb->args[1]; |
2978 | * | |
71932efc LE |
2979 | * cb->args[2] indicates if we shall loop over all resources, |
2980 | * or just dump all volumes of a single resource. | |
2981 | * | |
3b98c0c2 LE |
2982 | * This may miss entries inserted after this dump started, |
2983 | * or entries deleted before they are reached. | |
543cc10b | 2984 | * |
b30ab791 | 2985 | * We need to make sure the device won't disappear while |
543cc10b LE |
2986 | * we are looking at it, and revalidate our iterators |
2987 | * on each iteration. | |
2988 | */ | |
b411b363 | 2989 | |
05a10ec7 | 2990 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 2991 | rcu_read_lock(); |
543cc10b | 2992 | /* revalidate iterator position */ |
77c556f6 | 2993 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
2994 | if (pos == NULL) { |
2995 | /* first iteration */ | |
2996 | pos = tmp; | |
77c556f6 | 2997 | resource = pos; |
543cc10b LE |
2998 | break; |
2999 | } | |
3000 | if (tmp == pos) { | |
77c556f6 | 3001 | resource = pos; |
543cc10b LE |
3002 | break; |
3003 | } | |
b411b363 | 3004 | } |
77c556f6 AG |
3005 | if (resource) { |
3006 | next_resource: | |
251b8f8e AG |
3007 | device = idr_get_next(&resource->devices, &volume); |
3008 | if (!device) { | |
77c556f6 AG |
3009 | /* No more volumes to dump on this resource. |
3010 | * Advance resource iterator. */ | |
3011 | pos = list_entry_rcu(resource->resources.next, | |
3012 | struct drbd_resource, resources); | |
3013 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3014 | if (volume != 0) { |
71932efc LE |
3015 | /* If we reached the end of the list, |
3016 | * or only a single resource dump was requested, | |
3017 | * we are done. */ | |
77c556f6 | 3018 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3019 | goto out; |
543cc10b | 3020 | volume = 0; |
77c556f6 AG |
3021 | resource = pos; |
3022 | goto next_resource; | |
543cc10b LE |
3023 | } |
3024 | } | |
3025 | ||
98683650 | 3026 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3027 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3028 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3029 | if (!dh) | |
543cc10b LE |
3030 | goto out; |
3031 | ||
251b8f8e | 3032 | if (!device) { |
bde89a9e | 3033 | /* This is a connection without a single volume. |
367d675d LE |
3034 | * Suprisingly enough, it may have a network |
3035 | * configuration. */ | |
251b8f8e AG |
3036 | struct drbd_connection *connection; |
3037 | ||
543cc10b LE |
3038 | dh->minor = -1U; |
3039 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3040 | connection = the_only_connection(resource); |
3041 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3042 | goto cancel; |
251b8f8e AG |
3043 | if (connection) { |
3044 | struct net_conf *nc; | |
3045 | ||
3046 | nc = rcu_dereference(connection->net_conf); | |
3047 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3048 | goto cancel; | |
3049 | } | |
367d675d | 3050 | goto done; |
543cc10b | 3051 | } |
b411b363 | 3052 | |
0b0ba1ef | 3053 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3054 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3055 | |
b30ab791 | 3056 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3057 | dh->ret_code = NO_ERROR; |
3058 | ||
b30ab791 | 3059 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3060 | cancel: |
3b98c0c2 | 3061 | genlmsg_cancel(skb, dh); |
543cc10b | 3062 | goto out; |
3b98c0c2 | 3063 | } |
367d675d | 3064 | done: |
3b98c0c2 | 3065 | genlmsg_end(skb, dh); |
bde89a9e | 3066 | } |
b411b363 | 3067 | |
543cc10b | 3068 | out: |
c141ebda | 3069 | rcu_read_unlock(); |
543cc10b | 3070 | /* where to start the next iteration */ |
bde89a9e | 3071 | cb->args[0] = (long)pos; |
77c556f6 | 3072 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3073 | |
77c556f6 | 3074 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3075 | * Which will terminate the dump. */ |
3b98c0c2 | 3076 | return skb->len; |
b411b363 PR |
3077 | } |
3078 | ||
71932efc LE |
3079 | /* |
3080 | * Request status of all resources, or of all volumes within a single resource. | |
3081 | * | |
3082 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3083 | * Which means we cannot use the family->attrbuf or other such members, because | |
3084 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3085 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3086 | * | |
3087 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3088 | */ |
71932efc | 3089 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3090 | { |
71932efc LE |
3091 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3092 | struct nlattr *nla; | |
7c3063cc | 3093 | const char *resource_name; |
4bc76048 | 3094 | struct drbd_resource *resource; |
7c3063cc | 3095 | int maxtype; |
71932efc LE |
3096 | |
3097 | /* Is this a followup call? */ | |
3098 | if (cb->args[0]) { | |
3099 | /* ... of a single resource dump, | |
3100 | * and the resource iterator has been advanced already? */ | |
3101 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3102 | return 0; /* DONE. */ | |
3103 | goto dump; | |
3104 | } | |
3105 | ||
3106 | /* First call (from netlink_dump_start). We need to figure out | |
3107 | * which resource(s) the user wants us to dump. */ | |
3108 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3109 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3110 | DRBD_NLA_CFG_CONTEXT); | |
3111 | ||
3112 | /* No explicit context given. Dump all. */ | |
3113 | if (!nla) | |
3114 | goto dump; | |
7c3063cc AG |
3115 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3116 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3117 | if (IS_ERR(nla)) | |
3118 | return PTR_ERR(nla); | |
71932efc LE |
3119 | /* context given, but no name present? */ |
3120 | if (!nla) | |
3121 | return -EINVAL; | |
7c3063cc | 3122 | resource_name = nla_data(nla); |
4bc76048 AG |
3123 | if (!*resource_name) |
3124 | return -ENODEV; | |
3125 | resource = drbd_find_resource(resource_name); | |
3126 | if (!resource) | |
71932efc LE |
3127 | return -ENODEV; |
3128 | ||
4bc76048 | 3129 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3130 | |
71932efc | 3131 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3132 | * only dump this connection. */ |
4bc76048 | 3133 | cb->args[0] = (long)resource; |
71932efc | 3134 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3135 | cb->args[2] = (long)resource; |
71932efc LE |
3136 | |
3137 | dump: | |
3138 | return get_one_status(skb, cb); | |
3139 | } | |
b411b363 | 3140 | |
3b98c0c2 | 3141 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3142 | { |
3b98c0c2 LE |
3143 | enum drbd_ret_code retcode; |
3144 | struct timeout_parms tp; | |
3145 | int err; | |
b411b363 | 3146 | |
3b98c0c2 LE |
3147 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3148 | if (!adm_ctx.reply_skb) | |
3149 | return retcode; | |
3150 | if (retcode != NO_ERROR) | |
3151 | goto out; | |
b411b363 | 3152 | |
3b98c0c2 | 3153 | tp.timeout_type = |
b30ab791 AG |
3154 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3155 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3156 | UT_DEFAULT; |
b411b363 | 3157 | |
3b98c0c2 LE |
3158 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3159 | if (err) { | |
3160 | nlmsg_free(adm_ctx.reply_skb); | |
3161 | return err; | |
3162 | } | |
3163 | out: | |
3164 | drbd_adm_finish(info, retcode); | |
3165 | return 0; | |
b411b363 PR |
3166 | } |
3167 | ||
3b98c0c2 | 3168 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3169 | { |
b30ab791 | 3170 | struct drbd_device *device; |
3b98c0c2 | 3171 | enum drbd_ret_code retcode; |
58ffa580 | 3172 | struct start_ov_parms parms; |
b411b363 | 3173 | |
3b98c0c2 LE |
3174 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3175 | if (!adm_ctx.reply_skb) | |
3176 | return retcode; | |
3177 | if (retcode != NO_ERROR) | |
3178 | goto out; | |
873b0d5f | 3179 | |
b30ab791 | 3180 | device = adm_ctx.device; |
58ffa580 LE |
3181 | |
3182 | /* resume from last known position, if possible */ | |
b30ab791 | 3183 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3184 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3185 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3186 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3187 | if (err) { |
3188 | retcode = ERR_MANDATORY_TAG; | |
3189 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3190 | goto out; | |
3191 | } | |
b411b363 | 3192 | } |
58ffa580 | 3193 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3194 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3195 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3196 | |
3197 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3198 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3199 | drbd_suspend_io(device); |
3200 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3201 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3202 | drbd_resume_io(device); | |
3b98c0c2 LE |
3203 | out: |
3204 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3205 | return 0; |
3206 | } | |
3207 | ||
3208 | ||
3b98c0c2 | 3209 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3210 | { |
b30ab791 | 3211 | struct drbd_device *device; |
3b98c0c2 | 3212 | enum drbd_ret_code retcode; |
b411b363 PR |
3213 | int skip_initial_sync = 0; |
3214 | int err; | |
3b98c0c2 | 3215 | struct new_c_uuid_parms args; |
b411b363 | 3216 | |
3b98c0c2 LE |
3217 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3218 | if (!adm_ctx.reply_skb) | |
3219 | return retcode; | |
3220 | if (retcode != NO_ERROR) | |
3221 | goto out_nolock; | |
b411b363 | 3222 | |
b30ab791 | 3223 | device = adm_ctx.device; |
3b98c0c2 LE |
3224 | memset(&args, 0, sizeof(args)); |
3225 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3226 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3227 | if (err) { |
3228 | retcode = ERR_MANDATORY_TAG; | |
3229 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3230 | goto out_nolock; | |
3231 | } | |
b411b363 PR |
3232 | } |
3233 | ||
b30ab791 | 3234 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3235 | |
b30ab791 | 3236 | if (!get_ldev(device)) { |
b411b363 PR |
3237 | retcode = ERR_NO_DISK; |
3238 | goto out; | |
3239 | } | |
3240 | ||
3241 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3242 | if (device->state.conn == C_CONNECTED && |
3243 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3244 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3245 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3246 | skip_initial_sync = 1; |
b30ab791 | 3247 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3248 | retcode = ERR_CONNECTED; |
3249 | goto out_dec; | |
3250 | } | |
3251 | ||
b30ab791 AG |
3252 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3253 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3254 | |
3255 | if (args.clear_bm) { | |
b30ab791 | 3256 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3257 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3258 | if (err) { |
d0180171 | 3259 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3260 | retcode = ERR_IO_MD_DISK; |
3261 | } | |
3262 | if (skip_initial_sync) { | |
69a22773 | 3263 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
3264 | _drbd_uuid_set(device, UI_BITMAP, 0); |
3265 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3266 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3267 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3268 | CS_VERBOSE, NULL); |
0500813f | 3269 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3270 | } |
3271 | } | |
3272 | ||
b30ab791 | 3273 | drbd_md_sync(device); |
b411b363 | 3274 | out_dec: |
b30ab791 | 3275 | put_ldev(device); |
b411b363 | 3276 | out: |
b30ab791 | 3277 | mutex_unlock(device->state_mutex); |
3b98c0c2 LE |
3278 | out_nolock: |
3279 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3280 | return 0; |
3281 | } | |
3282 | ||
3b98c0c2 | 3283 | static enum drbd_ret_code |
7c3063cc | 3284 | drbd_check_resource_name(const char *name) |
b411b363 | 3285 | { |
3b98c0c2 | 3286 | if (!name || !name[0]) { |
7c3063cc | 3287 | drbd_msg_put_info("resource name missing"); |
3b98c0c2 | 3288 | return ERR_MANDATORY_TAG; |
b411b363 | 3289 | } |
3b98c0c2 LE |
3290 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3291 | * we must not allow slashes */ | |
3292 | if (strchr(name, '/')) { | |
7c3063cc | 3293 | drbd_msg_put_info("invalid resource name"); |
3b98c0c2 | 3294 | return ERR_INVALID_REQUEST; |
b411b363 | 3295 | } |
3b98c0c2 | 3296 | return NO_ERROR; |
774b3055 | 3297 | } |
b411b363 | 3298 | |
789c1b62 | 3299 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3300 | { |
3b98c0c2 | 3301 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3302 | struct res_opts res_opts; |
3303 | int err; | |
b411b363 | 3304 | |
3b98c0c2 LE |
3305 | retcode = drbd_adm_prepare(skb, info, 0); |
3306 | if (!adm_ctx.reply_skb) | |
3307 | return retcode; | |
3308 | if (retcode != NO_ERROR) | |
3309 | goto out; | |
b411b363 | 3310 | |
afbbfa88 AG |
3311 | set_res_opts_defaults(&res_opts); |
3312 | err = res_opts_from_attrs(&res_opts, info); | |
3313 | if (err && err != -ENOMSG) { | |
3314 | retcode = ERR_MANDATORY_TAG; | |
3315 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3316 | goto out; | |
b411b363 PR |
3317 | } |
3318 | ||
7c3063cc | 3319 | retcode = drbd_check_resource_name(adm_ctx.resource_name); |
3b98c0c2 LE |
3320 | if (retcode != NO_ERROR) |
3321 | goto out; | |
b411b363 | 3322 | |
5c661042 | 3323 | if (adm_ctx.resource) { |
38f19616 LE |
3324 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3325 | retcode = ERR_INVALID_REQUEST; | |
789c1b62 | 3326 | drbd_msg_put_info("resource exists"); |
38f19616 LE |
3327 | } |
3328 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3329 | goto out; |
b411b363 | 3330 | } |
b411b363 | 3331 | |
afbbfa88 | 3332 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3333 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
3334 | out: |
3335 | drbd_adm_finish(info, retcode); | |
3336 | return 0; | |
b411b363 PR |
3337 | } |
3338 | ||
05a10ec7 | 3339 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3340 | { |
3b98c0c2 LE |
3341 | struct drbd_genlmsghdr *dh = info->userhdr; |
3342 | enum drbd_ret_code retcode; | |
b411b363 | 3343 | |
44e52cfa | 3344 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3345 | if (!adm_ctx.reply_skb) |
3346 | return retcode; | |
3347 | if (retcode != NO_ERROR) | |
3348 | goto out; | |
b411b363 | 3349 | |
f2257a56 | 3350 | if (dh->minor > MINORMASK) { |
3b98c0c2 LE |
3351 | drbd_msg_put_info("requested minor out of range"); |
3352 | retcode = ERR_INVALID_REQUEST; | |
3353 | goto out; | |
b411b363 | 3354 | } |
0c8e36d9 | 3355 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
3356 | drbd_msg_put_info("requested volume id out of range"); |
3357 | retcode = ERR_INVALID_REQUEST; | |
3358 | goto out; | |
b411b363 | 3359 | } |
b411b363 | 3360 | |
38f19616 | 3361 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3362 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3363 | if (adm_ctx.device) { |
38f19616 LE |
3364 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
3365 | retcode = ERR_MINOR_EXISTS; | |
3366 | /* else: still NO_ERROR */ | |
3367 | goto out; | |
b411b363 | 3368 | } |
38f19616 | 3369 | |
59515a2e | 3370 | retcode = drbd_create_device(adm_ctx.resource, dh->minor, adm_ctx.volume); |
3b98c0c2 LE |
3371 | out: |
3372 | drbd_adm_finish(info, retcode); | |
3373 | return 0; | |
b411b363 PR |
3374 | } |
3375 | ||
05a10ec7 | 3376 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3377 | { |
b30ab791 AG |
3378 | if (device->state.disk == D_DISKLESS && |
3379 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3380 | * we may want to delete a minor from a live replication group. |
3381 | */ | |
b30ab791 AG |
3382 | device->state.role == R_SECONDARY) { |
3383 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3384 | CS_VERBOSE + CS_WAIT_COMPLETE); |
f82795d6 | 3385 | drbd_delete_device(device); |
85f75dd7 LE |
3386 | return NO_ERROR; |
3387 | } else | |
3388 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3389 | } |
3390 | ||
05a10ec7 | 3391 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3392 | { |
3b98c0c2 | 3393 | enum drbd_ret_code retcode; |
b411b363 | 3394 | |
3b98c0c2 LE |
3395 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3396 | if (!adm_ctx.reply_skb) | |
3397 | return retcode; | |
3398 | if (retcode != NO_ERROR) | |
3399 | goto out; | |
b411b363 | 3400 | |
05a10ec7 | 3401 | retcode = adm_del_minor(adm_ctx.device); |
85f75dd7 LE |
3402 | out: |
3403 | drbd_adm_finish(info, retcode); | |
3404 | return 0; | |
b411b363 PR |
3405 | } |
3406 | ||
85f75dd7 | 3407 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3408 | { |
b6f85ef9 AG |
3409 | struct drbd_resource *resource; |
3410 | struct drbd_connection *connection; | |
3411 | struct drbd_device *device; | |
f3dfa40a | 3412 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3413 | unsigned i; |
b411b363 | 3414 | |
a10f6b8a | 3415 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3416 | if (!adm_ctx.reply_skb) |
3417 | return retcode; | |
3418 | if (retcode != NO_ERROR) | |
3419 | goto out; | |
b411b363 | 3420 | |
b6f85ef9 | 3421 | resource = adm_ctx.resource; |
85f75dd7 | 3422 | /* demote */ |
b6f85ef9 AG |
3423 | for_each_connection(connection, resource) { |
3424 | struct drbd_peer_device *peer_device; | |
3425 | ||
3426 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3427 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3428 | if (retcode < SS_SUCCESS) { | |
3429 | drbd_msg_put_info("failed to demote"); | |
3430 | goto out; | |
3431 | } | |
3432 | } | |
3433 | ||
3434 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3435 | if (retcode < SS_SUCCESS) { |
b6f85ef9 | 3436 | drbd_msg_put_info("failed to disconnect"); |
c141ebda | 3437 | goto out; |
85f75dd7 | 3438 | } |
b411b363 | 3439 | } |
b411b363 | 3440 | |
85f75dd7 | 3441 | /* detach */ |
b6f85ef9 AG |
3442 | idr_for_each_entry(&resource->devices, device, i) { |
3443 | retcode = adm_detach(device, 0); | |
27012382 | 3444 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
85f75dd7 | 3445 | drbd_msg_put_info("failed to detach"); |
c141ebda | 3446 | goto out; |
85f75dd7 LE |
3447 | } |
3448 | } | |
b411b363 | 3449 | |
bde89a9e | 3450 | /* If we reach this, all volumes (of this connection) are Secondary, |
f3dfa40a | 3451 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have |
c141ebda | 3452 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
b6f85ef9 AG |
3453 | for_each_connection(connection, resource) |
3454 | drbd_thread_stop(&connection->worker); | |
b411b363 | 3455 | |
f3dfa40a | 3456 | /* Now, nothing can fail anymore */ |
b411b363 | 3457 | |
85f75dd7 | 3458 | /* delete volumes */ |
b6f85ef9 AG |
3459 | idr_for_each_entry(&resource->devices, device, i) { |
3460 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3461 | if (retcode != NO_ERROR) { |
3462 | /* "can not happen" */ | |
3463 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 | 3464 | goto out; |
85f75dd7 LE |
3465 | } |
3466 | } | |
b411b363 | 3467 | |
b6f85ef9 AG |
3468 | list_del_rcu(&resource->resources); |
3469 | synchronize_rcu(); | |
3470 | drbd_free_resource(resource); | |
3471 | retcode = NO_ERROR; | |
b411b363 | 3472 | |
3b98c0c2 LE |
3473 | out: |
3474 | drbd_adm_finish(info, retcode); | |
3475 | return 0; | |
b411b363 PR |
3476 | } |
3477 | ||
789c1b62 | 3478 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3479 | { |
77c556f6 AG |
3480 | struct drbd_resource *resource; |
3481 | struct drbd_connection *connection; | |
3b98c0c2 | 3482 | enum drbd_ret_code retcode; |
b411b363 | 3483 | |
44e52cfa | 3484 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3485 | if (!adm_ctx.reply_skb) |
3486 | return retcode; | |
3487 | if (retcode != NO_ERROR) | |
3488 | goto out; | |
b411b363 | 3489 | |
77c556f6 AG |
3490 | resource = adm_ctx.resource; |
3491 | for_each_connection(connection, resource) { | |
3492 | if (connection->cstate > C_STANDALONE) { | |
3493 | retcode = ERR_NET_CONFIGURED; | |
3494 | goto out; | |
3495 | } | |
3496 | } | |
3497 | if (!idr_is_empty(&resource->devices)) { | |
789c1b62 | 3498 | retcode = ERR_RES_IN_USE; |
77c556f6 | 3499 | goto out; |
b411b363 PR |
3500 | } |
3501 | ||
77c556f6 AG |
3502 | list_del_rcu(&resource->resources); |
3503 | for_each_connection(connection, resource) | |
3504 | drbd_thread_stop(&connection->worker); | |
3505 | synchronize_rcu(); | |
3506 | drbd_free_resource(resource); | |
3507 | retcode = NO_ERROR; | |
3b98c0c2 LE |
3508 | out: |
3509 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3510 | return 0; |
3511 | } | |
3512 | ||
b30ab791 | 3513 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3514 | { |
3b98c0c2 LE |
3515 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3516 | struct sk_buff *msg; | |
3517 | struct drbd_genlmsghdr *d_out; | |
3518 | unsigned seq; | |
3519 | int err = -ENOMEM; | |
3520 | ||
ef86b779 | 3521 | if (sib->sib_reason == SIB_SYNC_PROGRESS) { |
b30ab791 AG |
3522 | if (time_after(jiffies, device->rs_last_bcast + HZ)) |
3523 | device->rs_last_bcast = jiffies; | |
ef86b779 PR |
3524 | else |
3525 | return; | |
3526 | } | |
b411b363 | 3527 | |
3b98c0c2 LE |
3528 | seq = atomic_inc_return(&drbd_genl_seq); |
3529 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3530 | if (!msg) | |
3531 | goto failed; | |
3532 | ||
3533 | err = -EMSGSIZE; | |
3534 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3535 | if (!d_out) /* cannot happen, but anyways. */ | |
3536 | goto nla_put_failure; | |
b30ab791 | 3537 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3538 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3539 | |
b30ab791 | 3540 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3541 | goto nla_put_failure; |
3542 | genlmsg_end(msg, d_out); | |
3543 | err = drbd_genl_multicast_events(msg, 0); | |
3544 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3545 | if (err && err != -ESRCH) | |
3546 | goto failed; | |
b411b363 | 3547 | |
3b98c0c2 | 3548 | return; |
b411b363 | 3549 | |
3b98c0c2 LE |
3550 | nla_put_failure: |
3551 | nlmsg_free(msg); | |
3552 | failed: | |
d0180171 | 3553 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3554 | "Event seq:%u sib_reason:%u\n", |
3555 | err, seq, sib->sib_reason); | |
b411b363 | 3556 | } |