Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
b411b363 PR |
26 | #include <linux/module.h> |
27 | #include <linux/drbd.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/fs.h> | |
30 | #include <linux/file.h> | |
31 | #include <linux/slab.h> | |
b411b363 PR |
32 | #include <linux/blkpg.h> |
33 | #include <linux/cpumask.h> | |
34 | #include "drbd_int.h" | |
a3603a6e | 35 | #include "drbd_protocol.h" |
265be2d0 | 36 | #include "drbd_req.h" |
b411b363 PR |
37 | #include "drbd_wrappers.h" |
38 | #include <asm/unaligned.h> | |
b411b363 | 39 | #include <linux/drbd_limits.h> |
87f7be4c | 40 | #include <linux/kthread.h> |
b411b363 | 41 | |
3b98c0c2 LE |
42 | #include <net/genetlink.h> |
43 | ||
44 | /* .doit */ | |
45 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
47 | ||
05a10ec7 AG |
48 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
49 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 50 | |
789c1b62 AG |
51 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
52 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 53 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
54 | |
55 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
56 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 57 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
58 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
59 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 60 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
61 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
62 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 72 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
73 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
74 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
75 | /* .dumpit */ | |
76 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
77 | ||
78 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 79 | #include "drbd_nla.h" |
3b98c0c2 LE |
80 | #include <linux/genl_magic_func.h> |
81 | ||
82 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
83 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
84 | ||
3b98c0c2 LE |
85 | /* Configuration is strictly serialized, because generic netlink message |
86 | * processing is strictly serialized by the genl_lock(). | |
87 | * Which means we can use one static global drbd_config_context struct. | |
88 | */ | |
89 | static struct drbd_config_context { | |
90 | /* assigned from drbd_genlmsghdr */ | |
91 | unsigned int minor; | |
92 | /* assigned from request attributes, if present */ | |
93 | unsigned int volume; | |
94 | #define VOLUME_UNSPECIFIED (-1U) | |
95 | /* pointer into the request skb, | |
96 | * limited lifetime! */ | |
7c3063cc | 97 | char *resource_name; |
089c075d AG |
98 | struct nlattr *my_addr; |
99 | struct nlattr *peer_addr; | |
3b98c0c2 LE |
100 | |
101 | /* reply buffer */ | |
102 | struct sk_buff *reply_skb; | |
103 | /* pointer into reply buffer */ | |
104 | struct drbd_genlmsghdr *reply_dh; | |
105 | /* resolved from attributes, if possible */ | |
b30ab791 | 106 | struct drbd_device *device; |
4bc76048 | 107 | struct drbd_resource *resource; |
bde89a9e | 108 | struct drbd_connection *connection; |
3b98c0c2 LE |
109 | } adm_ctx; |
110 | ||
111 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
112 | { | |
113 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
114 | if (genlmsg_reply(skb, info)) | |
115 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 116 | } |
3b98c0c2 LE |
117 | |
118 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
119 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 120 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
121 | { |
122 | struct sk_buff *skb = adm_ctx.reply_skb; | |
123 | struct nlattr *nla; | |
124 | int err = -EMSGSIZE; | |
125 | ||
126 | if (!info || !info[0]) | |
127 | return 0; | |
128 | ||
129 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
130 | if (!nla) | |
131 | return err; | |
132 | ||
133 | err = nla_put_string(skb, T_info_text, info); | |
134 | if (err) { | |
135 | nla_nest_cancel(skb, nla); | |
136 | return err; | |
137 | } else | |
138 | nla_nest_end(skb, nla); | |
139 | return 0; | |
b411b363 PR |
140 | } |
141 | ||
3b98c0c2 LE |
142 | /* This would be a good candidate for a "pre_doit" hook, |
143 | * and per-family private info->pointers. | |
144 | * But we need to stay compatible with older kernels. | |
145 | * If it returns successfully, adm_ctx members are valid. | |
146 | */ | |
147 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 148 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 149 | #define DRBD_ADM_NEED_CONNECTION 4 |
3b98c0c2 LE |
150 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, |
151 | unsigned flags) | |
152 | { | |
153 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
154 | const u8 cmd = info->genlhdr->cmd; | |
155 | int err; | |
156 | ||
157 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
158 | ||
159 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 160 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
161 | return -EPERM; |
162 | ||
163 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
1e2a2551 AG |
164 | if (!adm_ctx.reply_skb) { |
165 | err = -ENOMEM; | |
3b98c0c2 | 166 | goto fail; |
1e2a2551 | 167 | } |
3b98c0c2 LE |
168 | |
169 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
170 | info, &drbd_genl_family, 0, cmd); | |
171 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
172 | * but anyways */ | |
1e2a2551 AG |
173 | if (!adm_ctx.reply_dh) { |
174 | err = -ENOMEM; | |
3b98c0c2 | 175 | goto fail; |
1e2a2551 | 176 | } |
3b98c0c2 LE |
177 | |
178 | adm_ctx.reply_dh->minor = d_in->minor; | |
179 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
180 | ||
089c075d | 181 | adm_ctx.volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
182 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
183 | struct nlattr *nla; | |
184 | /* parse and validate only */ | |
f399002e | 185 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
186 | if (err) |
187 | goto fail; | |
188 | ||
189 | /* It was present, and valid, | |
190 | * copy it over to the reply skb. */ | |
191 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
192 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
193 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
194 | if (err) | |
195 | goto fail; | |
196 | ||
197 | /* and assign stuff to the global adm_ctx */ | |
198 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
089c075d AG |
199 | if (nla) |
200 | adm_ctx.volume = nla_get_u32(nla); | |
7c3063cc | 201 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 202 | if (nla) |
7c3063cc | 203 | adm_ctx.resource_name = nla_data(nla); |
089c075d AG |
204 | adm_ctx.my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; |
205 | adm_ctx.peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
206 | if ((adm_ctx.my_addr && | |
bde89a9e | 207 | nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.connection->my_addr)) || |
089c075d | 208 | (adm_ctx.peer_addr && |
bde89a9e | 209 | nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.connection->peer_addr))) { |
089c075d AG |
210 | err = -EINVAL; |
211 | goto fail; | |
212 | } | |
213 | } | |
3b98c0c2 LE |
214 | |
215 | adm_ctx.minor = d_in->minor; | |
b30ab791 | 216 | adm_ctx.device = minor_to_device(d_in->minor); |
4bc76048 AG |
217 | if (adm_ctx.resource_name) { |
218 | adm_ctx.resource = drbd_find_resource(adm_ctx.resource_name); | |
4bc76048 | 219 | } |
3b98c0c2 | 220 | |
b30ab791 | 221 | if (!adm_ctx.device && (flags & DRBD_ADM_NEED_MINOR)) { |
3b98c0c2 LE |
222 | drbd_msg_put_info("unknown minor"); |
223 | return ERR_MINOR_INVALID; | |
224 | } | |
4bc76048 | 225 | if (!adm_ctx.resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
44e52cfa | 226 | drbd_msg_put_info("unknown resource"); |
a10f6b8a AG |
227 | if (adm_ctx.resource_name) |
228 | return ERR_RES_NOT_KNOWN; | |
3b98c0c2 LE |
229 | return ERR_INVALID_REQUEST; |
230 | } | |
231 | ||
089c075d | 232 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
3ab706fe | 233 | if (adm_ctx.resource) { |
089c075d AG |
234 | drbd_msg_put_info("no resource name expected"); |
235 | return ERR_INVALID_REQUEST; | |
236 | } | |
b30ab791 | 237 | if (adm_ctx.device) { |
089c075d AG |
238 | drbd_msg_put_info("no minor number expected"); |
239 | return ERR_INVALID_REQUEST; | |
240 | } | |
241 | if (adm_ctx.my_addr && adm_ctx.peer_addr) | |
bde89a9e | 242 | adm_ctx.connection = conn_get_by_addrs(nla_data(adm_ctx.my_addr), |
089c075d AG |
243 | nla_len(adm_ctx.my_addr), |
244 | nla_data(adm_ctx.peer_addr), | |
245 | nla_len(adm_ctx.peer_addr)); | |
bde89a9e | 246 | if (!adm_ctx.connection) { |
089c075d AG |
247 | drbd_msg_put_info("unknown connection"); |
248 | return ERR_INVALID_REQUEST; | |
249 | } | |
250 | } | |
251 | ||
3b98c0c2 | 252 | /* some more paranoia, if the request was over-determined */ |
4bc76048 AG |
253 | if (adm_ctx.device && adm_ctx.resource && |
254 | adm_ctx.device->resource != adm_ctx.resource) { | |
255 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", | |
256 | adm_ctx.minor, adm_ctx.resource->name, | |
d8628a86 | 257 | adm_ctx.device->resource->name); |
44e52cfa | 258 | drbd_msg_put_info("minor exists in different resource"); |
527f4b24 LE |
259 | return ERR_INVALID_REQUEST; |
260 | } | |
b30ab791 | 261 | if (adm_ctx.device && |
3b98c0c2 | 262 | adm_ctx.volume != VOLUME_UNSPECIFIED && |
b30ab791 | 263 | adm_ctx.volume != adm_ctx.device->vnr) { |
3b98c0c2 LE |
264 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
265 | adm_ctx.minor, adm_ctx.volume, | |
77c556f6 | 266 | adm_ctx.device->vnr, |
d8628a86 | 267 | adm_ctx.device->resource->name); |
527f4b24 | 268 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
269 | return ERR_INVALID_REQUEST; |
270 | } | |
0ace9dfa | 271 | |
3b98c0c2 LE |
272 | return NO_ERROR; |
273 | ||
274 | fail: | |
275 | nlmsg_free(adm_ctx.reply_skb); | |
276 | adm_ctx.reply_skb = NULL; | |
1e2a2551 | 277 | return err; |
3b98c0c2 LE |
278 | } |
279 | ||
280 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
281 | { | |
bde89a9e | 282 | if (adm_ctx.connection) { |
05a10ec7 | 283 | kref_put(&adm_ctx.connection->kref, drbd_destroy_connection); |
bde89a9e | 284 | adm_ctx.connection = NULL; |
0ace9dfa | 285 | } |
4bc76048 AG |
286 | if (adm_ctx.resource) { |
287 | kref_put(&adm_ctx.resource->kref, drbd_destroy_resource); | |
288 | adm_ctx.resource = NULL; | |
289 | } | |
0ace9dfa | 290 | |
3b98c0c2 LE |
291 | if (!adm_ctx.reply_skb) |
292 | return -ENOMEM; | |
293 | ||
294 | adm_ctx.reply_dh->ret_code = retcode; | |
3b98c0c2 LE |
295 | drbd_adm_send_reply(adm_ctx.reply_skb, info); |
296 | return 0; | |
297 | } | |
b411b363 | 298 | |
bde89a9e | 299 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 300 | { |
6b75dced | 301 | char *afs; |
b411b363 | 302 | |
089c075d | 303 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 304 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
305 | return; |
306 | ||
bde89a9e | 307 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
308 | case AF_INET6: |
309 | afs = "ipv6"; | |
310 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 311 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 312 | break; |
089c075d AG |
313 | case AF_INET: |
314 | afs = "ipv4"; | |
315 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 316 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 317 | break; |
089c075d AG |
318 | default: |
319 | afs = "ssocks"; | |
320 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 321 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 322 | } |
089c075d | 323 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 324 | } |
b411b363 | 325 | |
b30ab791 | 326 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
327 | { |
328 | char *envp[] = { "HOME=/", | |
329 | "TERM=linux", | |
330 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
331 | (char[20]) { }, /* address family */ |
332 | (char[60]) { }, /* address */ | |
b411b363 | 333 | NULL }; |
6b75dced | 334 | char mb[12]; |
b411b363 | 335 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 336 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 337 | struct sib_info sib; |
b411b363 PR |
338 | int ret; |
339 | ||
bde89a9e AG |
340 | if (current == connection->worker.task) |
341 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 342 | |
b30ab791 | 343 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 344 | setup_khelper_env(connection, envp); |
b411b363 | 345 | |
1090c056 LE |
346 | /* The helper may take some time. |
347 | * write out any unsynced meta data changes now */ | |
b30ab791 | 348 | drbd_md_sync(device); |
1090c056 | 349 | |
d0180171 | 350 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
351 | sib.sib_reason = SIB_HELPER_PRE; |
352 | sib.helper_name = cmd; | |
b30ab791 | 353 | drbd_bcast_event(device, &sib); |
70834d30 | 354 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 355 | if (ret) |
d0180171 | 356 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
357 | usermode_helper, cmd, mb, |
358 | (ret >> 8) & 0xff, ret); | |
359 | else | |
d0180171 | 360 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
361 | usermode_helper, cmd, mb, |
362 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
363 | sib.sib_reason = SIB_HELPER_POST; |
364 | sib.helper_exit_code = ret; | |
b30ab791 | 365 | drbd_bcast_event(device, &sib); |
b411b363 | 366 | |
bde89a9e AG |
367 | if (current == connection->worker.task) |
368 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
369 | |
370 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
371 | ret = 0; | |
372 | ||
373 | return ret; | |
374 | } | |
375 | ||
bde89a9e | 376 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
377 | { |
378 | char *envp[] = { "HOME=/", | |
379 | "TERM=linux", | |
380 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
381 | (char[20]) { }, /* address family */ | |
382 | (char[60]) { }, /* address */ | |
383 | NULL }; | |
77c556f6 AG |
384 | char *resource_name = connection->resource->name; |
385 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
386 | int ret; |
387 | ||
bde89a9e AG |
388 | setup_khelper_env(connection, envp); |
389 | conn_md_sync(connection); | |
6b75dced | 390 | |
1ec861eb | 391 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced PR |
392 | /* TODO: conn_bcast_event() ?? */ |
393 | ||
98683650 | 394 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 395 | if (ret) |
1ec861eb | 396 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 397 | usermode_helper, cmd, resource_name, |
6b75dced PR |
398 | (ret >> 8) & 0xff, ret); |
399 | else | |
1ec861eb | 400 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 401 | usermode_helper, cmd, resource_name, |
6b75dced PR |
402 | (ret >> 8) & 0xff, ret); |
403 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 404 | |
b411b363 PR |
405 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
406 | ret = 0; | |
407 | ||
408 | return ret; | |
409 | } | |
410 | ||
bde89a9e | 411 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 412 | { |
cb703454 | 413 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 414 | struct drbd_peer_device *peer_device; |
cb703454 PR |
415 | int vnr; |
416 | ||
695d08fa | 417 | rcu_read_lock(); |
c06ece6b AG |
418 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
419 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 420 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
421 | struct disk_conf *disk_conf = |
422 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
423 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 424 | put_ldev(device); |
cb703454 PR |
425 | } |
426 | } | |
695d08fa | 427 | rcu_read_unlock(); |
cb703454 PR |
428 | |
429 | return fp; | |
430 | } | |
431 | ||
bde89a9e | 432 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 433 | { |
28e448bb | 434 | unsigned int connect_cnt; |
cb703454 PR |
435 | union drbd_state mask = { }; |
436 | union drbd_state val = { }; | |
437 | enum drbd_fencing_p fp; | |
b411b363 PR |
438 | char *ex_to_string; |
439 | int r; | |
b411b363 | 440 | |
bde89a9e | 441 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 442 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
cb703454 PR |
443 | return false; |
444 | } | |
b411b363 | 445 | |
0500813f | 446 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 447 | connect_cnt = connection->connect_cnt; |
0500813f | 448 | spin_unlock_irq(&connection->resource->req_lock); |
28e448bb | 449 | |
bde89a9e | 450 | fp = highest_fencing_policy(connection); |
cb703454 PR |
451 | switch (fp) { |
452 | case FP_NOT_AVAIL: | |
1ec861eb | 453 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 454 | goto out; |
cb703454 PR |
455 | case FP_DONT_CARE: |
456 | return true; | |
457 | default: ; | |
b411b363 PR |
458 | } |
459 | ||
bde89a9e | 460 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
461 | |
462 | switch ((r>>8) & 0xff) { | |
463 | case 3: /* peer is inconsistent */ | |
464 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
465 | mask.pdsk = D_MASK; |
466 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
467 | break; |
468 | case 4: /* peer got outdated, or was already outdated */ | |
469 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
470 | mask.pdsk = D_MASK; |
471 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
472 | break; |
473 | case 5: /* peer was down */ | |
bde89a9e | 474 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
475 | /* we will(have) create(d) a new UUID anyways... */ |
476 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
477 | mask.pdsk = D_MASK; |
478 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
479 | } else { |
480 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
481 | } |
482 | break; | |
483 | case 6: /* Peer is primary, voluntarily outdate myself. | |
484 | * This is useful when an unconnected R_SECONDARY is asked to | |
485 | * become R_PRIMARY, but finds the other peer being active. */ | |
486 | ex_to_string = "peer is active"; | |
1ec861eb | 487 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
488 | mask.disk = D_MASK; |
489 | val.disk = D_OUTDATED; | |
b411b363 PR |
490 | break; |
491 | case 7: | |
492 | if (fp != FP_STONITH) | |
1ec861eb | 493 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 494 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
495 | mask.pdsk = D_MASK; |
496 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
497 | break; |
498 | default: | |
499 | /* The script is broken ... */ | |
1ec861eb | 500 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 501 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
502 | } |
503 | ||
1ec861eb | 504 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 505 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 506 | |
cb703454 | 507 | out: |
fb22c402 | 508 | |
cb703454 | 509 | /* Not using |
bde89a9e | 510 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
511 | here, because we might were able to re-establish the connection in the |
512 | meantime. */ | |
0500813f | 513 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
514 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
515 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
516 | /* In case the connection was established and droped |
517 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 518 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 519 | else |
bde89a9e | 520 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 521 | } |
0500813f | 522 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 | 523 | |
bde89a9e | 524 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
525 | } |
526 | ||
87f7be4c PR |
527 | static int _try_outdate_peer_async(void *data) |
528 | { | |
bde89a9e | 529 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 530 | |
bde89a9e | 531 | conn_try_outdate_peer(connection); |
87f7be4c | 532 | |
05a10ec7 | 533 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
534 | return 0; |
535 | } | |
536 | ||
bde89a9e | 537 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
538 | { |
539 | struct task_struct *opa; | |
540 | ||
bde89a9e AG |
541 | kref_get(&connection->kref); |
542 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); | |
9dc9fbb3 | 543 | if (IS_ERR(opa)) { |
1ec861eb | 544 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 545 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 546 | } |
87f7be4c | 547 | } |
b411b363 | 548 | |
bf885f8a | 549 | enum drbd_state_rv |
b30ab791 | 550 | drbd_set_role(struct drbd_device *device, enum drbd_role new_role, int force) |
b411b363 PR |
551 | { |
552 | const int max_tries = 4; | |
bf885f8a | 553 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 554 | struct net_conf *nc; |
b411b363 PR |
555 | int try = 0; |
556 | int forced = 0; | |
557 | union drbd_state mask, val; | |
b411b363 | 558 | |
b6f85ef9 AG |
559 | if (new_role == R_PRIMARY) { |
560 | struct drbd_connection *connection; | |
561 | ||
562 | /* Detect dead peers as soon as possible. */ | |
563 | ||
564 | rcu_read_lock(); | |
565 | for_each_connection(connection, device->resource) | |
566 | request_ping(connection); | |
567 | rcu_read_unlock(); | |
568 | } | |
b411b363 | 569 | |
b30ab791 | 570 | mutex_lock(device->state_mutex); |
b411b363 PR |
571 | |
572 | mask.i = 0; mask.role = R_MASK; | |
573 | val.i = 0; val.role = new_role; | |
574 | ||
575 | while (try++ < max_tries) { | |
b30ab791 | 576 | rv = _drbd_request_state(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
577 | |
578 | /* in case we first succeeded to outdate, | |
579 | * but now suddenly could establish a connection */ | |
bf885f8a | 580 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
581 | val.pdsk = 0; |
582 | mask.pdsk = 0; | |
583 | continue; | |
584 | } | |
585 | ||
bf885f8a | 586 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
587 | (device->state.disk < D_UP_TO_DATE && |
588 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
589 | mask.disk = D_MASK; |
590 | val.disk = D_UP_TO_DATE; | |
591 | forced = 1; | |
592 | continue; | |
593 | } | |
594 | ||
bf885f8a | 595 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 596 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 597 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 598 | |
a6b32bc3 | 599 | if (conn_try_outdate_peer(first_peer_device(device)->connection)) { |
b411b363 PR |
600 | val.disk = D_UP_TO_DATE; |
601 | mask.disk = D_MASK; | |
602 | } | |
b411b363 PR |
603 | continue; |
604 | } | |
605 | ||
bf885f8a | 606 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 607 | goto out; |
bf885f8a | 608 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
a6b32bc3 | 609 | if (!conn_try_outdate_peer(first_peer_device(device)->connection) && force) { |
d0180171 | 610 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
611 | mask.pdsk = D_MASK; |
612 | val.pdsk = D_OUTDATED; | |
b411b363 | 613 | |
cb703454 | 614 | } |
b411b363 PR |
615 | continue; |
616 | } | |
bf885f8a | 617 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
618 | /* Maybe the peer is detected as dead very soon... |
619 | retry at most once more in this case. */ | |
44ed167d PR |
620 | int timeo; |
621 | rcu_read_lock(); | |
a6b32bc3 | 622 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
623 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
624 | rcu_read_unlock(); | |
625 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
626 | if (try < max_tries) |
627 | try = max_tries - 1; | |
628 | continue; | |
629 | } | |
bf885f8a | 630 | if (rv < SS_SUCCESS) { |
b30ab791 | 631 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 632 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 633 | if (rv < SS_SUCCESS) |
3b98c0c2 | 634 | goto out; |
b411b363 PR |
635 | } |
636 | break; | |
637 | } | |
638 | ||
bf885f8a | 639 | if (rv < SS_SUCCESS) |
3b98c0c2 | 640 | goto out; |
b411b363 PR |
641 | |
642 | if (forced) | |
d0180171 | 643 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
644 | |
645 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 646 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 647 | |
b6dd1a89 LE |
648 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
649 | ||
b411b363 | 650 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
651 | set_disk_ro(device->vdisk, true); |
652 | if (get_ldev(device)) { | |
653 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
654 | put_ldev(device); | |
b411b363 PR |
655 | } |
656 | } else { | |
0500813f | 657 | mutex_lock(&device->resource->conf_update); |
a6b32bc3 | 658 | nc = first_peer_device(device)->connection->net_conf; |
44ed167d | 659 | if (nc) |
6139f60d | 660 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
0500813f | 661 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 662 | |
b30ab791 AG |
663 | set_disk_ro(device->vdisk, false); |
664 | if (get_ldev(device)) { | |
665 | if (((device->state.conn < C_CONNECTED || | |
666 | device->state.pdsk <= D_FAILED) | |
667 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
668 | drbd_uuid_new_current(device); | |
b411b363 | 669 | |
b30ab791 AG |
670 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
671 | put_ldev(device); | |
b411b363 PR |
672 | } |
673 | } | |
674 | ||
19f843aa LE |
675 | /* writeout of activity log covered areas of the bitmap |
676 | * to stable storage done in after state change already */ | |
b411b363 | 677 | |
b30ab791 | 678 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
679 | /* if this was forced, we should consider sync */ |
680 | if (forced) | |
b30ab791 AG |
681 | drbd_send_uuids(device); |
682 | drbd_send_current_state(device); | |
b411b363 PR |
683 | } |
684 | ||
b30ab791 | 685 | drbd_md_sync(device); |
b411b363 | 686 | |
b30ab791 | 687 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 688 | out: |
b30ab791 | 689 | mutex_unlock(device->state_mutex); |
bf885f8a | 690 | return rv; |
b411b363 PR |
691 | } |
692 | ||
3b98c0c2 | 693 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 694 | { |
3b98c0c2 LE |
695 | return err == -ENOMSG ? "required attribute missing" : |
696 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 697 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 698 | "invalid attribute value"; |
ef50a3e3 | 699 | } |
b411b363 | 700 | |
3b98c0c2 | 701 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 702 | { |
3b98c0c2 LE |
703 | struct set_role_parms parms; |
704 | int err; | |
705 | enum drbd_ret_code retcode; | |
b411b363 | 706 | |
3b98c0c2 LE |
707 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
708 | if (!adm_ctx.reply_skb) | |
709 | return retcode; | |
710 | if (retcode != NO_ERROR) | |
711 | goto out; | |
b411b363 | 712 | |
3b98c0c2 LE |
713 | memset(&parms, 0, sizeof(parms)); |
714 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 715 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
716 | if (err) { |
717 | retcode = ERR_MANDATORY_TAG; | |
718 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
719 | goto out; | |
720 | } | |
721 | } | |
b411b363 | 722 | |
3b98c0c2 | 723 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 724 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 725 | else |
b30ab791 | 726 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
3b98c0c2 LE |
727 | out: |
728 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
729 | return 0; |
730 | } | |
731 | ||
ae8bf312 LE |
732 | /* Initializes the md.*_offset members, so we are able to find |
733 | * the on disk meta data. | |
734 | * | |
735 | * We currently have two possible layouts: | |
736 | * external: | |
737 | * |----------- md_size_sect ------------------| | |
738 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
739 | * | al_offset == 8 | | |
740 | * | bm_offset = al_offset + X | | |
741 | * ==> bitmap sectors = md_size_sect - bm_offset | |
742 | * | |
743 | * internal: | |
744 | * |----------- md_size_sect ------------------| | |
745 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
746 | * | al_offset < 0 | | |
747 | * | bm_offset = al_offset - Y | | |
748 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
749 | * | |
750 | * Activity log size used to be fixed 32kB, | |
751 | * but is about to become configurable. | |
752 | */ | |
b30ab791 | 753 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
754 | struct drbd_backing_dev *bdev) |
755 | { | |
756 | sector_t md_size_sect = 0; | |
c04ccaa6 | 757 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 758 | |
3a4d4eb3 LE |
759 | bdev->md.md_offset = drbd_md_ss(bdev); |
760 | ||
68e41a43 | 761 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
762 | default: |
763 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 764 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
765 | bdev->md.al_offset = MD_4kB_SECT; |
766 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
767 | break; |
768 | case DRBD_MD_INDEX_FLEX_EXT: | |
769 | /* just occupy the full device; unit: sectors */ | |
770 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
771 | bdev->md.al_offset = MD_4kB_SECT; |
772 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
773 | break; |
774 | case DRBD_MD_INDEX_INTERNAL: | |
775 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 776 | /* al size is still fixed */ |
ae8bf312 | 777 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
778 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
779 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
780 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
781 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
782 | md_size_sect = ALIGN(md_size_sect, 8); | |
783 | ||
784 | /* plus the "drbd meta data super block", | |
785 | * and the activity log; */ | |
ae8bf312 | 786 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
787 | |
788 | bdev->md.md_size_sect = md_size_sect; | |
789 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 790 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
791 | break; |
792 | } | |
793 | } | |
794 | ||
4b0715f0 | 795 | /* input size is expected to be in KB */ |
b411b363 PR |
796 | char *ppsize(char *buf, unsigned long long size) |
797 | { | |
4b0715f0 LE |
798 | /* Needs 9 bytes at max including trailing NUL: |
799 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
800 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
801 | int base = 0; | |
4b0715f0 | 802 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
803 | /* shift + round */ |
804 | size = (size >> 10) + !!(size & (1<<9)); | |
805 | base++; | |
806 | } | |
4b0715f0 | 807 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
808 | |
809 | return buf; | |
810 | } | |
811 | ||
812 | /* there is still a theoretical deadlock when called from receiver | |
813 | * on an D_INCONSISTENT R_PRIMARY: | |
814 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
815 | * packet from remote to dec_ap_bio again. | |
816 | * receiver receive_sizes(), comes here, | |
817 | * waits for ap_bio_cnt == 0. -> deadlock. | |
818 | * but this cannot happen, actually, because: | |
819 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
820 | * (not connected, or bad/no disk on peer): | |
821 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
822 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
823 | * peer may not initiate a resize. | |
824 | */ | |
3b98c0c2 LE |
825 | /* Note these are not to be confused with |
826 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
827 | * which are (sub) state changes triggered by admin (drbdsetup), | |
828 | * and can be long lived. | |
b30ab791 | 829 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 830 | * and should be short-lived. */ |
b30ab791 | 831 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 832 | { |
b30ab791 AG |
833 | set_bit(SUSPEND_IO, &device->flags); |
834 | if (drbd_suspended(device)) | |
265be2d0 | 835 | return; |
b30ab791 | 836 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
837 | } |
838 | ||
b30ab791 | 839 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 840 | { |
b30ab791 AG |
841 | clear_bit(SUSPEND_IO, &device->flags); |
842 | wake_up(&device->misc_wait); | |
b411b363 PR |
843 | } |
844 | ||
845 | /** | |
846 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 847 | * @device: DRBD device. |
b411b363 PR |
848 | * |
849 | * Returns 0 on success, negative return values indicate errors. | |
850 | * You should call drbd_md_sync() after calling this function. | |
851 | */ | |
d752b269 | 852 | enum determine_dev_size |
b30ab791 | 853 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
854 | { |
855 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 856 | sector_t la_size_sect, u_size; |
b30ab791 | 857 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
858 | u32 prev_al_stripe_size_4k; |
859 | u32 prev_al_stripes; | |
b411b363 PR |
860 | sector_t size; |
861 | char ppb[10]; | |
d752b269 | 862 | void *buffer; |
b411b363 PR |
863 | |
864 | int md_moved, la_size_changed; | |
e96c9633 | 865 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
866 | |
867 | /* race: | |
868 | * application request passes inc_ap_bio, | |
869 | * but then cannot get an AL-reference. | |
870 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
871 | * | |
872 | * to avoid that: | |
873 | * Suspend IO right here. | |
874 | * still lock the act_log to not trigger ASSERTs there. | |
875 | */ | |
b30ab791 AG |
876 | drbd_suspend_io(device); |
877 | buffer = drbd_md_get_buffer(device); /* Lock meta-data IO */ | |
d752b269 | 878 | if (!buffer) { |
b30ab791 | 879 | drbd_resume_io(device); |
d752b269 PR |
880 | return DS_ERROR; |
881 | } | |
b411b363 PR |
882 | |
883 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 884 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 885 | |
b30ab791 AG |
886 | prev_first_sect = drbd_md_first_sector(device->ldev); |
887 | prev_size = device->ldev->md.md_size_sect; | |
888 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 889 | |
d752b269 PR |
890 | if (rs) { |
891 | /* rs is non NULL if we should change the AL layout only */ | |
892 | ||
893 | prev_al_stripes = md->al_stripes; | |
894 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
895 | ||
896 | md->al_stripes = rs->al_stripes; | |
897 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
898 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
899 | } | |
900 | ||
b30ab791 | 901 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 902 | |
daeda1cc | 903 | rcu_read_lock(); |
b30ab791 | 904 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 905 | rcu_read_unlock(); |
b30ab791 | 906 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 907 | |
d752b269 PR |
908 | if (size < la_size_sect) { |
909 | if (rs && u_size == 0) { | |
910 | /* Remove "rs &&" later. This check should always be active, but | |
911 | right now the receiver expects the permissive behavior */ | |
d0180171 | 912 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
913 | "Use --size=%llus for explicit shrink.\n", |
914 | (unsigned long long)size); | |
915 | rv = DS_ERROR_SHRINK; | |
916 | } | |
917 | if (u_size > size) | |
918 | rv = DS_ERROR_SPACE_MD; | |
919 | if (rv != DS_UNCHANGED) | |
920 | goto err_out; | |
921 | } | |
922 | ||
b30ab791 AG |
923 | if (drbd_get_capacity(device->this_bdev) != size || |
924 | drbd_bm_capacity(device) != size) { | |
b411b363 | 925 | int err; |
b30ab791 | 926 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
927 | if (unlikely(err)) { |
928 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 929 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 930 | if (size == 0) { |
d0180171 | 931 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
932 | "Could not allocate bitmap!\n"); |
933 | } else { | |
d0180171 | 934 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
935 | "Leaving size unchanged at size = %lu KB\n", |
936 | (unsigned long)size); | |
937 | } | |
e96c9633 | 938 | rv = DS_ERROR; |
b411b363 PR |
939 | } |
940 | /* racy, see comments above. */ | |
b30ab791 AG |
941 | drbd_set_my_capacity(device, size); |
942 | device->ldev->md.la_size_sect = size; | |
d0180171 | 943 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
944 | (unsigned long long)size>>1); |
945 | } | |
d752b269 PR |
946 | if (rv <= DS_ERROR) |
947 | goto err_out; | |
b411b363 | 948 | |
b30ab791 | 949 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 950 | |
b30ab791 AG |
951 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
952 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 953 | |
d752b269 PR |
954 | if (la_size_changed || md_moved || rs) { |
955 | u32 prev_flags; | |
24dccabb | 956 | |
b30ab791 | 957 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
958 | |
959 | prev_flags = md->flags; | |
960 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 961 | drbd_md_write(device, buffer); |
d752b269 | 962 | |
d0180171 | 963 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
964 | la_size_changed && md_moved ? "size changed and md moved" : |
965 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 966 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 967 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 968 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 969 | drbd_initialize_al(device, buffer); |
d752b269 PR |
970 | |
971 | md->flags = prev_flags; | |
b30ab791 | 972 | drbd_md_write(device, buffer); |
d752b269 PR |
973 | |
974 | if (rs) | |
d0180171 AG |
975 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
976 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
977 | } |
978 | ||
cccac985 | 979 | if (size > la_size_sect) |
57737adc | 980 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 981 | if (size < la_size_sect) |
e96c9633 | 982 | rv = DS_SHRUNK; |
d752b269 PR |
983 | |
984 | if (0) { | |
985 | err_out: | |
986 | if (rs) { | |
987 | md->al_stripes = prev_al_stripes; | |
988 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
989 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
990 | ||
b30ab791 | 991 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
992 | } |
993 | } | |
b30ab791 AG |
994 | lc_unlock(device->act_log); |
995 | wake_up(&device->al_wait); | |
996 | drbd_md_put_buffer(device); | |
997 | drbd_resume_io(device); | |
b411b363 PR |
998 | |
999 | return rv; | |
1000 | } | |
1001 | ||
1002 | sector_t | |
b30ab791 | 1003 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1004 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1005 | { |
b30ab791 | 1006 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1007 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1008 | sector_t m_size; /* my size */ |
b411b363 PR |
1009 | sector_t size = 0; |
1010 | ||
1011 | m_size = drbd_get_max_capacity(bdev); | |
1012 | ||
b30ab791 | 1013 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1014 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1015 | p_size = m_size; |
1016 | } | |
1017 | ||
b411b363 PR |
1018 | if (p_size && m_size) { |
1019 | size = min_t(sector_t, p_size, m_size); | |
1020 | } else { | |
cccac985 LE |
1021 | if (la_size_sect) { |
1022 | size = la_size_sect; | |
b411b363 PR |
1023 | if (m_size && m_size < size) |
1024 | size = m_size; | |
1025 | if (p_size && p_size < size) | |
1026 | size = p_size; | |
1027 | } else { | |
1028 | if (m_size) | |
1029 | size = m_size; | |
1030 | if (p_size) | |
1031 | size = p_size; | |
1032 | } | |
1033 | } | |
1034 | ||
1035 | if (size == 0) | |
d0180171 | 1036 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1037 | |
1038 | if (u_size) { | |
1039 | if (u_size > size) | |
d0180171 | 1040 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1041 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1042 | else | |
1043 | size = u_size; | |
1044 | } | |
1045 | ||
1046 | return size; | |
1047 | } | |
1048 | ||
1049 | /** | |
1050 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1051 | * @device: DRBD device. |
b411b363 PR |
1052 | * |
1053 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1054 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1055 | * this function. | |
1056 | */ | |
b30ab791 | 1057 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1058 | { |
1059 | struct lru_cache *n, *t; | |
1060 | struct lc_element *e; | |
1061 | unsigned int in_use; | |
1062 | int i; | |
1063 | ||
b30ab791 AG |
1064 | if (device->act_log && |
1065 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1066 | return 0; |
1067 | ||
1068 | in_use = 0; | |
b30ab791 | 1069 | t = device->act_log; |
7ad651b5 | 1070 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1071 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1072 | |
1073 | if (n == NULL) { | |
d0180171 | 1074 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1075 | return -ENOMEM; |
1076 | } | |
b30ab791 | 1077 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1078 | if (t) { |
1079 | for (i = 0; i < t->nr_elements; i++) { | |
1080 | e = lc_element_by_index(t, i); | |
1081 | if (e->refcnt) | |
d0180171 | 1082 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1083 | e->lc_number, e->refcnt); |
1084 | in_use += e->refcnt; | |
1085 | } | |
1086 | } | |
1087 | if (!in_use) | |
b30ab791 AG |
1088 | device->act_log = n; |
1089 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1090 | if (in_use) { |
d0180171 | 1091 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1092 | lc_destroy(n); |
1093 | return -EBUSY; | |
1094 | } else { | |
1095 | if (t) | |
1096 | lc_destroy(t); | |
1097 | } | |
b30ab791 | 1098 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1099 | return 0; |
1100 | } | |
1101 | ||
b30ab791 | 1102 | static void drbd_setup_queue_param(struct drbd_device *device, unsigned int max_bio_size) |
b411b363 | 1103 | { |
b30ab791 | 1104 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1105 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1106 | unsigned int max_segments = 0; | |
99432fcc | 1107 | |
b30ab791 AG |
1108 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1109 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1110 | |
1111 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1112 | rcu_read_lock(); |
b30ab791 | 1113 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1114 | rcu_read_unlock(); |
b30ab791 | 1115 | put_ldev(device); |
99432fcc | 1116 | } |
b411b363 | 1117 | |
b411b363 | 1118 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1119 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1120 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1121 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1122 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1123 | |
b30ab791 AG |
1124 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1125 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1126 | |
1127 | blk_queue_stack_limits(q, b); | |
1128 | ||
1129 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1130 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1131 | q->backing_dev_info.ra_pages, |
1132 | b->backing_dev_info.ra_pages); | |
1133 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1134 | } | |
b30ab791 | 1135 | put_ldev(device); |
99432fcc PR |
1136 | } |
1137 | } | |
1138 | ||
b30ab791 | 1139 | void drbd_reconsider_max_bio_size(struct drbd_device *device) |
99432fcc | 1140 | { |
db141b2f | 1141 | unsigned int now, new, local, peer; |
99432fcc | 1142 | |
b30ab791 AG |
1143 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1144 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1145 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1146 | |
b30ab791 AG |
1147 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1148 | local = queue_max_hw_sectors(device->ldev->backing_bdev->bd_disk->queue) << 9; | |
1149 | device->local_max_bio_size = local; | |
1150 | put_ldev(device); | |
b411b363 | 1151 | } |
db141b2f | 1152 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1153 | |
1154 | /* We may ignore peer limits if the peer is modern enough. | |
1155 | Because new from 8.3.8 onwards the peer can use multiple | |
1156 | BIOs for a single peer_request */ | |
b30ab791 | 1157 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1158 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1159 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1160 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1161 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1162 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1163 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1164 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1165 | else | |
99432fcc PR |
1166 | peer = DRBD_MAX_BIO_SIZE; |
1167 | } | |
1168 | ||
db141b2f | 1169 | new = min(local, peer); |
99432fcc | 1170 | |
b30ab791 | 1171 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1172 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1173 | |
1174 | if (new != now) | |
d0180171 | 1175 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1176 | |
b30ab791 | 1177 | drbd_setup_queue_param(device, new); |
b411b363 PR |
1178 | } |
1179 | ||
a18e9d1e | 1180 | /* Starts the worker thread */ |
bde89a9e | 1181 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1182 | { |
bde89a9e AG |
1183 | drbd_thread_start(&connection->worker); |
1184 | conn_flush_workqueue(connection); | |
b411b363 PR |
1185 | } |
1186 | ||
a18e9d1e | 1187 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1188 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1189 | { |
992d6e91 | 1190 | bool stop_threads; |
0500813f | 1191 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1192 | stop_threads = conn_all_vols_unconf(connection) && |
1193 | connection->cstate == C_STANDALONE; | |
0500813f | 1194 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1195 | if (stop_threads) { |
1196 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1197 | * in conn_disconnect() */ |
bde89a9e AG |
1198 | drbd_thread_stop(&connection->receiver); |
1199 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1200 | } |
b411b363 PR |
1201 | } |
1202 | ||
0778286a | 1203 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1204 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1205 | { |
1206 | int s = 0; | |
1207 | ||
b30ab791 | 1208 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1209 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1210 | return; |
1211 | } | |
1212 | ||
b30ab791 | 1213 | drbd_al_shrink(device); |
0500813f | 1214 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1215 | if (device->state.conn < C_CONNECTED) |
1216 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1217 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1218 | lc_unlock(device->act_log); |
0778286a PR |
1219 | |
1220 | if (s) | |
d0180171 | 1221 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1222 | } |
1223 | ||
5979e361 LE |
1224 | |
1225 | static bool should_set_defaults(struct genl_info *info) | |
1226 | { | |
1227 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1228 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1229 | } | |
1230 | ||
5bbcf5e6 | 1231 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1232 | { |
5bbcf5e6 LE |
1233 | /* This is limited by 16 bit "slot" numbers, |
1234 | * and by available on-disk context storage. | |
1235 | * | |
1236 | * Also (u16)~0 is special (denotes a "free" extent). | |
1237 | * | |
1238 | * One transaction occupies one 4kB on-disk block, | |
1239 | * we have n such blocks in the on disk ring buffer, | |
1240 | * the "current" transaction may fail (n-1), | |
1241 | * and there is 919 slot numbers context information per transaction. | |
1242 | * | |
1243 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1244 | * so cap there first. | |
1245 | */ | |
1246 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1247 | const unsigned int sufficient_on_disk = | |
1248 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1249 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1250 | |
5bbcf5e6 LE |
1251 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1252 | ||
1253 | if (al_size_4k > sufficient_on_disk) | |
1254 | return max_al_nr; | |
1255 | ||
1256 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1257 | } |
1258 | ||
f399002e LE |
1259 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1260 | { | |
1261 | enum drbd_ret_code retcode; | |
b30ab791 | 1262 | struct drbd_device *device; |
daeda1cc | 1263 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1264 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1265 | int err, fifo_size; |
f399002e LE |
1266 | |
1267 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1268 | if (!adm_ctx.reply_skb) | |
1269 | return retcode; | |
1270 | if (retcode != NO_ERROR) | |
1271 | goto out; | |
1272 | ||
b30ab791 | 1273 | device = adm_ctx.device; |
f399002e LE |
1274 | |
1275 | /* we also need a disk | |
1276 | * to change the options on */ | |
b30ab791 | 1277 | if (!get_ldev(device)) { |
f399002e LE |
1278 | retcode = ERR_NO_DISK; |
1279 | goto out; | |
1280 | } | |
1281 | ||
daeda1cc | 1282 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1283 | if (!new_disk_conf) { |
f399002e LE |
1284 | retcode = ERR_NOMEM; |
1285 | goto fail; | |
1286 | } | |
1287 | ||
0500813f | 1288 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1289 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1290 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1291 | if (should_set_defaults(info)) |
b966b5dd | 1292 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1293 | |
5ecc72c3 | 1294 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1295 | if (err && err != -ENOMSG) { |
f399002e LE |
1296 | retcode = ERR_MANDATORY_TAG; |
1297 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
8e229434 | 1298 | goto fail_unlock; |
f399002e LE |
1299 | } |
1300 | ||
5ecc72c3 LE |
1301 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1302 | new_disk_conf->resync_rate = 1; | |
f399002e | 1303 | |
5bbcf5e6 LE |
1304 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1305 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1306 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1307 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1308 | |
1309 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1310 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1311 | |
5ecc72c3 | 1312 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1313 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1314 | new_plan = fifo_alloc(fifo_size); |
1315 | if (!new_plan) { | |
d0180171 | 1316 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1317 | retcode = ERR_NOMEM; |
daeda1cc | 1318 | goto fail_unlock; |
f399002e LE |
1319 | } |
1320 | } | |
1321 | ||
b30ab791 AG |
1322 | drbd_suspend_io(device); |
1323 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1324 | drbd_al_shrink(device); | |
1325 | err = drbd_check_al_size(device, new_disk_conf); | |
1326 | lc_unlock(device->act_log); | |
1327 | wake_up(&device->al_wait); | |
1328 | drbd_resume_io(device); | |
f399002e LE |
1329 | |
1330 | if (err) { | |
1331 | retcode = ERR_NOMEM; | |
daeda1cc | 1332 | goto fail_unlock; |
f399002e LE |
1333 | } |
1334 | ||
dc97b708 | 1335 | write_lock_irq(&global_state_lock); |
b30ab791 | 1336 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1337 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1338 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1339 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1340 | } |
1341 | write_unlock_irq(&global_state_lock); | |
f399002e | 1342 | |
daeda1cc PR |
1343 | if (retcode != NO_ERROR) |
1344 | goto fail_unlock; | |
f399002e | 1345 | |
813472ce | 1346 | if (new_plan) { |
b30ab791 AG |
1347 | old_plan = device->rs_plan_s; |
1348 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1349 | } |
9958c857 | 1350 | |
0500813f | 1351 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1352 | |
9a51ab1c | 1353 | if (new_disk_conf->al_updates) |
b30ab791 | 1354 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1355 | else |
b30ab791 | 1356 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1357 | |
691631c0 | 1358 | if (new_disk_conf->md_flushes) |
b30ab791 | 1359 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1360 | else |
b30ab791 | 1361 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1362 | |
a6b32bc3 | 1363 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
27eb13e9 | 1364 | |
b30ab791 | 1365 | drbd_md_sync(device); |
f399002e | 1366 | |
b30ab791 AG |
1367 | if (device->state.conn >= C_CONNECTED) |
1368 | drbd_send_sync_param(device); | |
f399002e | 1369 | |
daeda1cc PR |
1370 | synchronize_rcu(); |
1371 | kfree(old_disk_conf); | |
813472ce | 1372 | kfree(old_plan); |
b30ab791 | 1373 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1374 | goto success; |
1375 | ||
1376 | fail_unlock: | |
0500813f | 1377 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1378 | fail: |
5ecc72c3 | 1379 | kfree(new_disk_conf); |
813472ce | 1380 | kfree(new_plan); |
daeda1cc | 1381 | success: |
b30ab791 | 1382 | put_ldev(device); |
f399002e LE |
1383 | out: |
1384 | drbd_adm_finish(info, retcode); | |
1385 | return 0; | |
1386 | } | |
1387 | ||
3b98c0c2 | 1388 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1389 | { |
b30ab791 | 1390 | struct drbd_device *device; |
3b98c0c2 | 1391 | int err; |
116676ca | 1392 | enum drbd_ret_code retcode; |
b411b363 PR |
1393 | enum determine_dev_size dd; |
1394 | sector_t max_possible_sectors; | |
1395 | sector_t min_md_device_sectors; | |
1396 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1397 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1398 | struct block_device *bdev; |
b411b363 | 1399 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1400 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1401 | union drbd_state ns, os; |
f2024e7c | 1402 | enum drbd_state_rv rv; |
44ed167d | 1403 | struct net_conf *nc; |
b411b363 | 1404 | |
3b98c0c2 LE |
1405 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1406 | if (!adm_ctx.reply_skb) | |
1407 | return retcode; | |
1408 | if (retcode != NO_ERROR) | |
40cbf085 | 1409 | goto finish; |
b411b363 | 1410 | |
b30ab791 | 1411 | device = adm_ctx.device; |
a6b32bc3 | 1412 | conn_reconfig_start(first_peer_device(device)->connection); |
b411b363 PR |
1413 | |
1414 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1415 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1416 | retcode = ERR_DISK_CONFIGURED; |
1417 | goto fail; | |
1418 | } | |
82f59cc6 LE |
1419 | /* It may just now have detached because of IO error. Make sure |
1420 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1421 | * e.g. if someone calls attach from the on-io-error handler, | |
1422 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
b30ab791 | 1423 | wait_event(device->misc_wait, !atomic_read(&device->local_cnt)); |
b411b363 | 1424 | |
383606e0 | 1425 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1426 | clear_bit(FORCE_DETACH, &device->flags); |
1427 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1428 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1429 | |
0029d624 | 1430 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1431 | device->rs_total = 0; |
1432 | device->rs_failed = 0; | |
1433 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1434 | |
3b98c0c2 | 1435 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1436 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1437 | if (!nbc) { | |
1438 | retcode = ERR_NOMEM; | |
1439 | goto fail; | |
1440 | } | |
9f2247bb PR |
1441 | spin_lock_init(&nbc->md.uuid_lock); |
1442 | ||
daeda1cc PR |
1443 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1444 | if (!new_disk_conf) { | |
1445 | retcode = ERR_NOMEM; | |
b411b363 PR |
1446 | goto fail; |
1447 | } | |
daeda1cc | 1448 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1449 | |
daeda1cc PR |
1450 | set_disk_conf_defaults(new_disk_conf); |
1451 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1452 | if (err) { |
b411b363 | 1453 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1454 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1455 | goto fail; |
1456 | } | |
1457 | ||
5bbcf5e6 LE |
1458 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1459 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1460 | |
9958c857 PR |
1461 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1462 | if (!new_plan) { | |
1463 | retcode = ERR_NOMEM; | |
1464 | goto fail; | |
1465 | } | |
1466 | ||
daeda1cc | 1467 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1468 | retcode = ERR_MD_IDX_INVALID; |
1469 | goto fail; | |
1470 | } | |
1471 | ||
a3f8f7dc | 1472 | write_lock_irq(&global_state_lock); |
b30ab791 | 1473 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1474 | write_unlock_irq(&global_state_lock); |
1475 | if (retcode != NO_ERROR) | |
1476 | goto fail; | |
1477 | ||
44ed167d | 1478 | rcu_read_lock(); |
a6b32bc3 | 1479 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d | 1480 | if (nc) { |
daeda1cc | 1481 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1482 | rcu_read_unlock(); |
47ff2d0a PR |
1483 | retcode = ERR_STONITH_AND_PROT_A; |
1484 | goto fail; | |
1485 | } | |
1486 | } | |
44ed167d | 1487 | rcu_read_unlock(); |
47ff2d0a | 1488 | |
daeda1cc | 1489 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1490 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1491 | if (IS_ERR(bdev)) { |
d0180171 | 1492 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1493 | PTR_ERR(bdev)); |
b411b363 PR |
1494 | retcode = ERR_OPEN_DISK; |
1495 | goto fail; | |
1496 | } | |
e525fd89 TH |
1497 | nbc->backing_bdev = bdev; |
1498 | ||
1499 | /* | |
1500 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1501 | * drbd sharing one meta device. TODO in that case, paranoia | |
1502 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1503 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1504 | * should check it for you already; but if you don't, or | |
1505 | * someone fooled it, we need to double check here) | |
1506 | */ | |
daeda1cc | 1507 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1508 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1509 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1510 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1511 | if (IS_ERR(bdev)) { |
d0180171 | 1512 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1513 | PTR_ERR(bdev)); |
b411b363 PR |
1514 | retcode = ERR_OPEN_MD_DISK; |
1515 | goto fail; | |
1516 | } | |
e525fd89 | 1517 | nbc->md_bdev = bdev; |
b411b363 | 1518 | |
e525fd89 | 1519 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1520 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1521 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1522 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1523 | goto fail; |
1524 | } | |
1525 | ||
1526 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1527 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1528 | offsetof(struct bm_extent, lce)); |
1529 | if (!resync_lru) { | |
1530 | retcode = ERR_NOMEM; | |
e525fd89 | 1531 | goto fail; |
b411b363 PR |
1532 | } |
1533 | ||
c04ccaa6 LE |
1534 | /* Read our meta data super block early. |
1535 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1536 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1537 | if (retcode != NO_ERROR) |
1538 | goto fail; | |
b411b363 | 1539 | |
5bbcf5e6 LE |
1540 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1541 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1542 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1543 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1544 | ||
daeda1cc | 1545 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1546 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1547 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1548 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1549 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1550 | goto fail; |
b411b363 PR |
1551 | } |
1552 | ||
daeda1cc | 1553 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1554 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1555 | /* at least one MB, otherwise it does not make sense */ | |
1556 | min_md_device_sectors = (2<<10); | |
1557 | } else { | |
1558 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1559 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1560 | } |
1561 | ||
b411b363 | 1562 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1563 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1564 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1565 | "at least %llu sectors needed for this meta-disk type\n", |
1566 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1567 | goto fail; |
b411b363 PR |
1568 | } |
1569 | ||
1570 | /* Make sure the new disk is big enough | |
1571 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1572 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1573 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1574 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1575 | goto fail; |
b411b363 PR |
1576 | } |
1577 | ||
1578 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1579 | ||
1352994b | 1580 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1581 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1582 | "to currently maximum possible %llu sectors <==\n", |
1583 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1584 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1585 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1586 | "meta data may help <<==\n"); |
1587 | } | |
1588 | ||
b30ab791 | 1589 | drbd_suspend_io(device); |
b411b363 | 1590 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1591 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1592 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1593 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1594 | * As barriers are counted per resource, | |
1595 | * we'd need to suspend io on all devices of a resource. | |
1596 | */ | |
b30ab791 | 1597 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1598 | /* and for any other previously queued work */ |
b30ab791 | 1599 | drbd_flush_workqueue(device); |
b411b363 | 1600 | |
b30ab791 | 1601 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1602 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1603 | drbd_resume_io(device); |
f2024e7c | 1604 | if (rv < SS_SUCCESS) |
e525fd89 | 1605 | goto fail; |
b411b363 | 1606 | |
b30ab791 | 1607 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1608 | goto force_diskless; |
1609 | ||
b30ab791 AG |
1610 | if (!device->bitmap) { |
1611 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1612 | retcode = ERR_NOMEM; |
1613 | goto force_diskless_dec; | |
1614 | } | |
1615 | } | |
1616 | ||
b30ab791 AG |
1617 | if (device->state.conn < C_CONNECTED && |
1618 | device->state.role == R_PRIMARY && | |
1619 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
d0180171 | 1620 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1621 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1622 | retcode = ERR_DATA_NOT_CURRENT; |
1623 | goto force_diskless_dec; | |
1624 | } | |
1625 | ||
1626 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1627 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1628 | retcode = ERR_NOMEM; |
1629 | goto force_diskless_dec; | |
1630 | } | |
1631 | ||
1632 | /* Prevent shrinking of consistent devices ! */ | |
1633 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1634 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1635 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1636 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1637 | goto force_diskless_dec; |
1638 | } | |
1639 | ||
b411b363 PR |
1640 | /* Reset the "barriers don't work" bits here, then force meta data to |
1641 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1642 | if (new_disk_conf->md_flushes) |
b30ab791 | 1643 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1644 | else |
b30ab791 | 1645 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1646 | |
1647 | /* Point of no return reached. | |
1648 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1649 | * now device takes over responsibility, and the state engine should |
b411b363 | 1650 | * clean it up somewhere. */ |
0b0ba1ef | 1651 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1652 | device->ldev = nbc; |
1653 | device->resync = resync_lru; | |
1654 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1655 | nbc = NULL; |
1656 | resync_lru = NULL; | |
daeda1cc | 1657 | new_disk_conf = NULL; |
9958c857 | 1658 | new_plan = NULL; |
b411b363 | 1659 | |
a6b32bc3 | 1660 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
b411b363 | 1661 | |
b30ab791 AG |
1662 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1663 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1664 | else |
b30ab791 | 1665 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1666 | |
b30ab791 | 1667 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1668 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1669 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1670 | |
b30ab791 AG |
1671 | device->send_cnt = 0; |
1672 | device->recv_cnt = 0; | |
1673 | device->read_cnt = 0; | |
1674 | device->writ_cnt = 0; | |
b411b363 | 1675 | |
b30ab791 | 1676 | drbd_reconsider_max_bio_size(device); |
b411b363 PR |
1677 | |
1678 | /* If I am currently not R_PRIMARY, | |
1679 | * but meta data primary indicator is set, | |
1680 | * I just now recover from a hard crash, | |
1681 | * and have been R_PRIMARY before that crash. | |
1682 | * | |
1683 | * Now, if I had no connection before that crash | |
1684 | * (have been degraded R_PRIMARY), chances are that | |
1685 | * I won't find my peer now either. | |
1686 | * | |
1687 | * In that case, and _only_ in that case, | |
1688 | * we use the degr-wfc-timeout instead of the default, | |
1689 | * so we can automatically recover from a crash of a | |
1690 | * degraded but active "cluster" after a certain timeout. | |
1691 | */ | |
b30ab791 AG |
1692 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1693 | if (device->state.role != R_PRIMARY && | |
1694 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1695 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1696 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1697 | |
b30ab791 | 1698 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1699 | if (dd <= DS_ERROR) { |
b411b363 PR |
1700 | retcode = ERR_NOMEM_BITMAP; |
1701 | goto force_diskless_dec; | |
e96c9633 | 1702 | } else if (dd == DS_GREW) |
b30ab791 | 1703 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1704 | |
b30ab791 AG |
1705 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1706 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1707 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1708 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1709 | "(aka FullSync)\n"); |
b30ab791 | 1710 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1711 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1712 | retcode = ERR_IO_MD_DISK; |
1713 | goto force_diskless_dec; | |
1714 | } | |
1715 | } else { | |
b30ab791 | 1716 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1717 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1718 | retcode = ERR_IO_MD_DISK; |
1719 | goto force_diskless_dec; | |
1720 | } | |
b411b363 PR |
1721 | } |
1722 | ||
b30ab791 AG |
1723 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1724 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1725 | |
0500813f | 1726 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1727 | os = drbd_read_state(device); |
78bae59b | 1728 | ns = os; |
b411b363 PR |
1729 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1730 | otherwise investigate MDF_WasUpToDate... | |
1731 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1732 | otherwise into D_CONSISTENT state. | |
1733 | */ | |
b30ab791 AG |
1734 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1735 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1736 | ns.disk = D_CONSISTENT; |
1737 | else | |
1738 | ns.disk = D_OUTDATED; | |
1739 | } else { | |
1740 | ns.disk = D_INCONSISTENT; | |
1741 | } | |
1742 | ||
b30ab791 | 1743 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1744 | ns.pdsk = D_OUTDATED; |
1745 | ||
daeda1cc PR |
1746 | rcu_read_lock(); |
1747 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1748 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1749 | ns.disk = D_UP_TO_DATE; |
1750 | ||
1751 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1752 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1753 | this point, because drbd_request_state() modifies these | |
1754 | flags. */ | |
1755 | ||
b30ab791 AG |
1756 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1757 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1758 | else |
b30ab791 | 1759 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1760 | |
1761 | rcu_read_unlock(); | |
1762 | ||
b411b363 PR |
1763 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1764 | state after the negotiation phase. */ | |
b30ab791 AG |
1765 | if (device->state.conn == C_CONNECTED) { |
1766 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1767 | ns.i = os.i; |
1768 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1769 | |
1770 | /* We expect to receive up-to-date UUIDs soon. | |
1771 | To avoid a race in receive_state, free p_uuid while | |
1772 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1773 | kfree(device->p_uuid); |
1774 | device->p_uuid = NULL; | |
b411b363 PR |
1775 | } |
1776 | ||
b30ab791 | 1777 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1778 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1779 | |
1780 | if (rv < SS_SUCCESS) | |
1781 | goto force_diskless_dec; | |
1782 | ||
b30ab791 | 1783 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1784 | |
b30ab791 AG |
1785 | if (device->state.role == R_PRIMARY) |
1786 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1787 | else |
b30ab791 | 1788 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1789 | |
b30ab791 AG |
1790 | drbd_md_mark_dirty(device); |
1791 | drbd_md_sync(device); | |
b411b363 | 1792 | |
b30ab791 AG |
1793 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1794 | put_ldev(device); | |
a6b32bc3 | 1795 | conn_reconfig_done(first_peer_device(device)->connection); |
3b98c0c2 | 1796 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1797 | return 0; |
1798 | ||
1799 | force_diskless_dec: | |
b30ab791 | 1800 | put_ldev(device); |
b411b363 | 1801 | force_diskless: |
b30ab791 AG |
1802 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1803 | drbd_md_sync(device); | |
b411b363 | 1804 | fail: |
a6b32bc3 | 1805 | conn_reconfig_done(first_peer_device(device)->connection); |
b411b363 | 1806 | if (nbc) { |
e525fd89 TH |
1807 | if (nbc->backing_bdev) |
1808 | blkdev_put(nbc->backing_bdev, | |
1809 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1810 | if (nbc->md_bdev) | |
1811 | blkdev_put(nbc->md_bdev, | |
1812 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1813 | kfree(nbc); |
1814 | } | |
daeda1cc | 1815 | kfree(new_disk_conf); |
b411b363 | 1816 | lc_destroy(resync_lru); |
9958c857 | 1817 | kfree(new_plan); |
b411b363 | 1818 | |
40cbf085 | 1819 | finish: |
3b98c0c2 | 1820 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1821 | return 0; |
1822 | } | |
1823 | ||
b30ab791 | 1824 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1825 | { |
19f83c76 | 1826 | enum drbd_state_rv retcode; |
9a0d9d03 | 1827 | int ret; |
02ee8f95 | 1828 | |
cdfda633 | 1829 | if (force) { |
b30ab791 AG |
1830 | set_bit(FORCE_DETACH, &device->flags); |
1831 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1832 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1833 | goto out; |
1834 | } | |
1835 | ||
b30ab791 AG |
1836 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
1837 | drbd_md_get_buffer(device); /* make sure there is no in-flight meta-data IO */ | |
1838 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
1839 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1840 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1841 | ret = wait_event_interruptible(device->misc_wait, |
1842 | device->state.disk != D_FAILED); | |
1843 | drbd_resume_io(device); | |
9b2f61ae | 1844 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1845 | retcode = SS_NOTHING_TO_DO; |
1846 | if (ret) | |
1847 | retcode = ERR_INTR; | |
02ee8f95 | 1848 | out: |
85f75dd7 | 1849 | return retcode; |
b411b363 PR |
1850 | } |
1851 | ||
82f59cc6 LE |
1852 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1853 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1854 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1855 | * internal references as well. | |
1856 | * Only then we have finally detached. */ | |
3b98c0c2 | 1857 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1858 | { |
116676ca | 1859 | enum drbd_ret_code retcode; |
cdfda633 PR |
1860 | struct detach_parms parms = { }; |
1861 | int err; | |
b411b363 | 1862 | |
3b98c0c2 LE |
1863 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1864 | if (!adm_ctx.reply_skb) | |
1865 | return retcode; | |
1866 | if (retcode != NO_ERROR) | |
1867 | goto out; | |
b411b363 | 1868 | |
cdfda633 PR |
1869 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1870 | err = detach_parms_from_attrs(&parms, info); | |
1871 | if (err) { | |
1872 | retcode = ERR_MANDATORY_TAG; | |
1873 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1874 | goto out; | |
1875 | } | |
b411b363 PR |
1876 | } |
1877 | ||
b30ab791 | 1878 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
3b98c0c2 LE |
1879 | out: |
1880 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1881 | return 0; |
1882 | } | |
b411b363 | 1883 | |
bde89a9e | 1884 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1885 | { |
c06ece6b | 1886 | struct drbd_peer_device *peer_device; |
695d08fa | 1887 | bool rv = false; |
f399002e LE |
1888 | int vnr; |
1889 | ||
695d08fa | 1890 | rcu_read_lock(); |
c06ece6b AG |
1891 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1892 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1893 | if (device->state.conn == C_SYNC_SOURCE || |
1894 | device->state.conn == C_SYNC_TARGET || | |
1895 | device->state.conn == C_PAUSED_SYNC_S || | |
1896 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1897 | rv = true; |
1898 | break; | |
1899 | } | |
b411b363 | 1900 | } |
695d08fa | 1901 | rcu_read_unlock(); |
b411b363 | 1902 | |
695d08fa | 1903 | return rv; |
f399002e | 1904 | } |
47ff2d0a | 1905 | |
bde89a9e | 1906 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1907 | { |
c06ece6b | 1908 | struct drbd_peer_device *peer_device; |
695d08fa | 1909 | bool rv = false; |
f399002e LE |
1910 | int vnr; |
1911 | ||
695d08fa | 1912 | rcu_read_lock(); |
c06ece6b AG |
1913 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1914 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1915 | if (device->state.conn == C_VERIFY_S || |
1916 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1917 | rv = true; |
1918 | break; | |
47ff2d0a PR |
1919 | } |
1920 | } | |
695d08fa | 1921 | rcu_read_unlock(); |
b411b363 | 1922 | |
695d08fa | 1923 | return rv; |
f399002e | 1924 | } |
422028b1 | 1925 | |
cd64397c | 1926 | static enum drbd_ret_code |
270eb5c9 | 1927 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 1928 | { |
c06ece6b | 1929 | struct drbd_peer_device *peer_device; |
cd64397c | 1930 | int i; |
b411b363 | 1931 | |
270eb5c9 AG |
1932 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
1933 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 1934 | return ERR_NEED_APV_100; |
b411b363 | 1935 | |
270eb5c9 | 1936 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
1937 | return ERR_NEED_APV_100; |
1938 | ||
270eb5c9 | 1939 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 1940 | return ERR_NEED_APV_100; |
b411b363 PR |
1941 | } |
1942 | ||
270eb5c9 | 1943 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
1944 | conn_highest_role(connection) == R_PRIMARY && |
1945 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 1946 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 1947 | |
270eb5c9 AG |
1948 | if (new_net_conf->two_primaries && |
1949 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
1950 | return ERR_NOT_PROTO_C; |
1951 | ||
c06ece6b AG |
1952 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
1953 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1954 | if (get_ldev(device)) { |
1955 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
1956 | put_ldev(device); | |
270eb5c9 | 1957 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1958 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 1959 | } |
270eb5c9 | 1960 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 1961 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
1962 | } |
1963 | ||
270eb5c9 | 1964 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 1965 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 1966 | |
cd64397c PR |
1967 | return NO_ERROR; |
1968 | } | |
b411b363 | 1969 | |
44ed167d | 1970 | static enum drbd_ret_code |
270eb5c9 | 1971 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
1972 | { |
1973 | static enum drbd_ret_code rv; | |
c06ece6b | 1974 | struct drbd_peer_device *peer_device; |
44ed167d | 1975 | int i; |
b411b363 | 1976 | |
44ed167d | 1977 | rcu_read_lock(); |
270eb5c9 | 1978 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 1979 | rcu_read_unlock(); |
b411b363 | 1980 | |
bde89a9e | 1981 | /* connection->volumes protected by genl_lock() here */ |
c06ece6b AG |
1982 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
1983 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1984 | if (!device->bitmap) { |
1985 | if (drbd_bm_init(device)) | |
44ed167d | 1986 | return ERR_NOMEM; |
b411b363 PR |
1987 | } |
1988 | } | |
1989 | ||
44ed167d PR |
1990 | return rv; |
1991 | } | |
b411b363 | 1992 | |
0fd0ea06 PR |
1993 | struct crypto { |
1994 | struct crypto_hash *verify_tfm; | |
1995 | struct crypto_hash *csums_tfm; | |
1996 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 1997 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 1998 | }; |
b411b363 | 1999 | |
0fd0ea06 | 2000 | static int |
4b6ad6d4 | 2001 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2002 | { |
2003 | if (!tfm_name[0]) | |
2004 | return NO_ERROR; | |
b411b363 | 2005 | |
0fd0ea06 PR |
2006 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2007 | if (IS_ERR(*tfm)) { | |
2008 | *tfm = NULL; | |
2009 | return err_alg; | |
b411b363 | 2010 | } |
b411b363 | 2011 | |
0fd0ea06 PR |
2012 | return NO_ERROR; |
2013 | } | |
b411b363 | 2014 | |
0fd0ea06 | 2015 | static enum drbd_ret_code |
270eb5c9 | 2016 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2017 | { |
2018 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2019 | enum drbd_ret_code rv; | |
0fd0ea06 | 2020 | |
270eb5c9 | 2021 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2022 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2023 | if (rv != NO_ERROR) |
2024 | return rv; | |
270eb5c9 | 2025 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2026 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2027 | if (rv != NO_ERROR) |
2028 | return rv; | |
270eb5c9 | 2029 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2030 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2031 | if (rv != NO_ERROR) |
2032 | return rv; | |
270eb5c9 | 2033 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2034 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2035 | new_net_conf->cram_hmac_alg); |
b411b363 | 2036 | |
4b6ad6d4 AG |
2037 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2038 | ERR_AUTH_ALG); | |
b411b363 PR |
2039 | } |
2040 | ||
0fd0ea06 PR |
2041 | return rv; |
2042 | } | |
b411b363 | 2043 | |
0fd0ea06 PR |
2044 | static void free_crypto(struct crypto *crypto) |
2045 | { | |
0fd0ea06 | 2046 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2047 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2048 | crypto_free_hash(crypto->csums_tfm); |
2049 | crypto_free_hash(crypto->verify_tfm); | |
2050 | } | |
b411b363 | 2051 | |
f399002e LE |
2052 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2053 | { | |
2054 | enum drbd_ret_code retcode; | |
bde89a9e | 2055 | struct drbd_connection *connection; |
270eb5c9 | 2056 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2057 | int err; |
2058 | int ovr; /* online verify running */ | |
2059 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2060 | struct crypto crypto = { }; |
b411b363 | 2061 | |
089c075d | 2062 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2063 | if (!adm_ctx.reply_skb) |
2064 | return retcode; | |
2065 | if (retcode != NO_ERROR) | |
2066 | goto out; | |
b411b363 | 2067 | |
bde89a9e | 2068 | connection = adm_ctx.connection; |
b411b363 | 2069 | |
270eb5c9 AG |
2070 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2071 | if (!new_net_conf) { | |
f399002e LE |
2072 | retcode = ERR_NOMEM; |
2073 | goto out; | |
2074 | } | |
b411b363 | 2075 | |
bde89a9e | 2076 | conn_reconfig_start(connection); |
b411b363 | 2077 | |
bde89a9e | 2078 | mutex_lock(&connection->data.mutex); |
0500813f | 2079 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2080 | old_net_conf = connection->net_conf; |
2561b9c1 | 2081 | |
270eb5c9 | 2082 | if (!old_net_conf) { |
f399002e LE |
2083 | drbd_msg_put_info("net conf missing, try connect"); |
2084 | retcode = ERR_INVALID_REQUEST; | |
2561b9c1 PR |
2085 | goto fail; |
2086 | } | |
2087 | ||
270eb5c9 | 2088 | *new_net_conf = *old_net_conf; |
5979e361 | 2089 | if (should_set_defaults(info)) |
270eb5c9 | 2090 | set_net_conf_defaults(new_net_conf); |
f399002e | 2091 | |
270eb5c9 | 2092 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2093 | if (err && err != -ENOMSG) { |
f399002e LE |
2094 | retcode = ERR_MANDATORY_TAG; |
2095 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2096 | goto fail; | |
2561b9c1 | 2097 | } |
b411b363 | 2098 | |
270eb5c9 | 2099 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2100 | if (retcode != NO_ERROR) |
2101 | goto fail; | |
b411b363 | 2102 | |
f399002e | 2103 | /* re-sync running */ |
bde89a9e | 2104 | rsr = conn_resync_running(connection); |
270eb5c9 | 2105 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2106 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2107 | goto fail; |
b411b363 PR |
2108 | } |
2109 | ||
f399002e | 2110 | /* online verify running */ |
bde89a9e | 2111 | ovr = conn_ov_running(connection); |
270eb5c9 | 2112 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2113 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2114 | goto fail; |
f399002e | 2115 | } |
b411b363 | 2116 | |
270eb5c9 | 2117 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2118 | if (retcode != NO_ERROR) |
b411b363 | 2119 | goto fail; |
f399002e | 2120 | |
270eb5c9 | 2121 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2122 | |
2123 | if (!rsr) { | |
bde89a9e AG |
2124 | crypto_free_hash(connection->csums_tfm); |
2125 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2126 | crypto.csums_tfm = NULL; |
f399002e LE |
2127 | } |
2128 | if (!ovr) { | |
bde89a9e AG |
2129 | crypto_free_hash(connection->verify_tfm); |
2130 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2131 | crypto.verify_tfm = NULL; |
b411b363 PR |
2132 | } |
2133 | ||
bde89a9e AG |
2134 | crypto_free_hash(connection->integrity_tfm); |
2135 | connection->integrity_tfm = crypto.integrity_tfm; | |
2136 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2137 | /* Do this without trying to take connection->data.mutex again. */ | |
2138 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2139 | |
bde89a9e AG |
2140 | crypto_free_hash(connection->cram_hmac_tfm); |
2141 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2142 | |
0500813f | 2143 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2144 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2145 | synchronize_rcu(); |
270eb5c9 | 2146 | kfree(old_net_conf); |
91fd4dad | 2147 | |
bde89a9e AG |
2148 | if (connection->cstate >= C_WF_REPORT_PARAMS) |
2149 | drbd_send_sync_param(minor_to_device(conn_lowest_minor(connection))); | |
f399002e | 2150 | |
91fd4dad PR |
2151 | goto done; |
2152 | ||
b411b363 | 2153 | fail: |
0500813f | 2154 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2155 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2156 | free_crypto(&crypto); |
270eb5c9 | 2157 | kfree(new_net_conf); |
91fd4dad | 2158 | done: |
bde89a9e | 2159 | conn_reconfig_done(connection); |
f399002e LE |
2160 | out: |
2161 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2162 | return 0; |
2163 | } | |
2164 | ||
3b98c0c2 | 2165 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2166 | { |
c06ece6b | 2167 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2168 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2169 | struct crypto crypto = { }; |
77c556f6 | 2170 | struct drbd_resource *resource; |
bde89a9e | 2171 | struct drbd_connection *connection; |
3b98c0c2 LE |
2172 | enum drbd_ret_code retcode; |
2173 | int i; | |
2174 | int err; | |
b411b363 | 2175 | |
44e52cfa | 2176 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2177 | |
3b98c0c2 LE |
2178 | if (!adm_ctx.reply_skb) |
2179 | return retcode; | |
2180 | if (retcode != NO_ERROR) | |
2181 | goto out; | |
089c075d AG |
2182 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
2183 | drbd_msg_put_info("connection endpoint(s) missing"); | |
2184 | retcode = ERR_INVALID_REQUEST; | |
2185 | goto out; | |
2186 | } | |
b411b363 | 2187 | |
089c075d AG |
2188 | /* No need for _rcu here. All reconfiguration is |
2189 | * strictly serialized on genl_lock(). We are protected against | |
2190 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2191 | for_each_resource(resource, &drbd_resources) { |
2192 | for_each_connection(connection, resource) { | |
2193 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2194 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2195 | connection->my_addr_len)) { | |
2196 | retcode = ERR_LOCAL_ADDR; | |
2197 | goto out; | |
2198 | } | |
b411b363 | 2199 | |
77c556f6 AG |
2200 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2201 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2202 | connection->peer_addr_len)) { | |
2203 | retcode = ERR_PEER_ADDR; | |
2204 | goto out; | |
2205 | } | |
089c075d | 2206 | } |
b411b363 PR |
2207 | } |
2208 | ||
3ab706fe | 2209 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2210 | conn_reconfig_start(connection); |
b411b363 | 2211 | |
bde89a9e | 2212 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2213 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2214 | goto fail; |
2215 | } | |
2216 | ||
a209b4ae | 2217 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2218 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2219 | if (!new_net_conf) { | |
b411b363 | 2220 | retcode = ERR_NOMEM; |
b411b363 PR |
2221 | goto fail; |
2222 | } | |
2223 | ||
270eb5c9 | 2224 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2225 | |
270eb5c9 | 2226 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2227 | if (err && err != -ENOMSG) { |
b411b363 | 2228 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2229 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2230 | goto fail; |
2231 | } | |
2232 | ||
270eb5c9 | 2233 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2234 | if (retcode != NO_ERROR) |
422028b1 | 2235 | goto fail; |
b411b363 | 2236 | |
270eb5c9 | 2237 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2238 | if (retcode != NO_ERROR) |
2239 | goto fail; | |
b411b363 | 2240 | |
270eb5c9 | 2241 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2242 | |
bde89a9e | 2243 | conn_flush_workqueue(connection); |
b411b363 | 2244 | |
0500813f | 2245 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2246 | old_net_conf = connection->net_conf; |
2247 | if (old_net_conf) { | |
b411b363 | 2248 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2249 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2250 | goto fail; |
2251 | } | |
270eb5c9 | 2252 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2253 | |
bde89a9e AG |
2254 | conn_free_crypto(connection); |
2255 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2256 | connection->integrity_tfm = crypto.integrity_tfm; | |
2257 | connection->csums_tfm = crypto.csums_tfm; | |
2258 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2259 | |
bde89a9e AG |
2260 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2261 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2262 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2263 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2264 | |
0500813f | 2265 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2266 | |
695d08fa | 2267 | rcu_read_lock(); |
c06ece6b AG |
2268 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2269 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2270 | device->send_cnt = 0; |
2271 | device->recv_cnt = 0; | |
b411b363 | 2272 | } |
695d08fa | 2273 | rcu_read_unlock(); |
b411b363 | 2274 | |
bde89a9e | 2275 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2276 | |
bde89a9e | 2277 | conn_reconfig_done(connection); |
3b98c0c2 | 2278 | drbd_adm_finish(info, retcode); |
b411b363 | 2279 | return 0; |
b411b363 | 2280 | |
b411b363 | 2281 | fail: |
0fd0ea06 | 2282 | free_crypto(&crypto); |
270eb5c9 | 2283 | kfree(new_net_conf); |
b411b363 | 2284 | |
bde89a9e | 2285 | conn_reconfig_done(connection); |
3b98c0c2 LE |
2286 | out: |
2287 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2288 | return 0; |
2289 | } | |
2290 | ||
bde89a9e | 2291 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2292 | { |
2293 | enum drbd_state_rv rv; | |
85f75dd7 | 2294 | |
bde89a9e | 2295 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2296 | force ? CS_HARD : 0); |
85f75dd7 LE |
2297 | |
2298 | switch (rv) { | |
2299 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2300 | break; |
85f75dd7 LE |
2301 | case SS_ALREADY_STANDALONE: |
2302 | return SS_SUCCESS; | |
2303 | case SS_PRIMARY_NOP: | |
2304 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2305 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2306 | |
2307 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2308 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2309 | |
85f75dd7 LE |
2310 | break; |
2311 | case SS_CW_FAILED_BY_PEER: | |
2312 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2313 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2314 | disk, D_OUTDATED), 0); |
2315 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2316 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2317 | CS_HARD); |
b411b363 | 2318 | } |
85f75dd7 LE |
2319 | break; |
2320 | default:; | |
2321 | /* no special handling necessary */ | |
2322 | } | |
2323 | ||
f3dfa40a LE |
2324 | if (rv >= SS_SUCCESS) { |
2325 | enum drbd_state_rv rv2; | |
2326 | /* No one else can reconfigure the network while I am here. | |
2327 | * The state handling only uses drbd_thread_stop_nowait(), | |
2328 | * we want to really wait here until the receiver is no more. | |
2329 | */ | |
9693da23 | 2330 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2331 | |
2332 | /* Race breaker. This additional state change request may be | |
2333 | * necessary, if this was a forced disconnect during a receiver | |
2334 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2335 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2336 | * C_STANDALONE already, now, and this becomes a no-op. |
2337 | */ | |
bde89a9e | 2338 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2339 | CS_VERBOSE | CS_HARD); |
2340 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2341 | drbd_err(connection, |
f3dfa40a LE |
2342 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2343 | rv2); | |
b411b363 | 2344 | } |
85f75dd7 LE |
2345 | return rv; |
2346 | } | |
b411b363 | 2347 | |
3b98c0c2 | 2348 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2349 | { |
3b98c0c2 | 2350 | struct disconnect_parms parms; |
bde89a9e | 2351 | struct drbd_connection *connection; |
85f75dd7 | 2352 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2353 | enum drbd_ret_code retcode; |
2354 | int err; | |
2561b9c1 | 2355 | |
089c075d | 2356 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2357 | if (!adm_ctx.reply_skb) |
2358 | return retcode; | |
2359 | if (retcode != NO_ERROR) | |
2561b9c1 | 2360 | goto fail; |
b411b363 | 2361 | |
bde89a9e | 2362 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2363 | memset(&parms, 0, sizeof(parms)); |
2364 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2365 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2366 | if (err) { |
2367 | retcode = ERR_MANDATORY_TAG; | |
2368 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
b411b363 PR |
2369 | goto fail; |
2370 | } | |
2371 | } | |
2372 | ||
bde89a9e | 2373 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2374 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2375 | retcode = rv; /* FIXME: Type mismatch. */ |
2376 | else | |
2377 | retcode = NO_ERROR; | |
b411b363 | 2378 | fail: |
3b98c0c2 | 2379 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2380 | return 0; |
2381 | } | |
2382 | ||
b30ab791 | 2383 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2384 | { |
2385 | int iass; /* I am sync source */ | |
2386 | ||
d0180171 | 2387 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2388 | if (device->state.role != device->state.peer) |
2389 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2390 | else |
a6b32bc3 | 2391 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2392 | |
2393 | if (iass) | |
b30ab791 | 2394 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2395 | else |
b30ab791 | 2396 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2397 | } |
2398 | ||
3b98c0c2 | 2399 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2400 | { |
daeda1cc | 2401 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2402 | struct resize_parms rs; |
b30ab791 | 2403 | struct drbd_device *device; |
3b98c0c2 | 2404 | enum drbd_ret_code retcode; |
b411b363 | 2405 | enum determine_dev_size dd; |
d752b269 | 2406 | bool change_al_layout = false; |
6495d2c6 | 2407 | enum dds_flags ddsf; |
daeda1cc | 2408 | sector_t u_size; |
3b98c0c2 | 2409 | int err; |
b411b363 | 2410 | |
3b98c0c2 LE |
2411 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2412 | if (!adm_ctx.reply_skb) | |
2413 | return retcode; | |
2414 | if (retcode != NO_ERROR) | |
b411b363 | 2415 | goto fail; |
3b98c0c2 | 2416 | |
b30ab791 AG |
2417 | device = adm_ctx.device; |
2418 | if (!get_ldev(device)) { | |
d752b269 PR |
2419 | retcode = ERR_NO_DISK; |
2420 | goto fail; | |
2421 | } | |
2422 | ||
3b98c0c2 | 2423 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2424 | rs.al_stripes = device->ldev->md.al_stripes; |
2425 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2426 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2427 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2428 | if (err) { |
3b98c0c2 LE |
2429 | retcode = ERR_MANDATORY_TAG; |
2430 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
d752b269 | 2431 | goto fail_ldev; |
b411b363 PR |
2432 | } |
2433 | } | |
2434 | ||
b30ab791 | 2435 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2436 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2437 | goto fail_ldev; |
b411b363 | 2438 | } |
b411b363 | 2439 | |
b30ab791 AG |
2440 | if (device->state.role == R_SECONDARY && |
2441 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2442 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2443 | goto fail_ldev; |
b411b363 | 2444 | } |
b411b363 | 2445 | |
a6b32bc3 | 2446 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2447 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2448 | goto fail_ldev; |
6495d2c6 PR |
2449 | } |
2450 | ||
daeda1cc | 2451 | rcu_read_lock(); |
b30ab791 | 2452 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2453 | rcu_read_unlock(); |
2454 | if (u_size != (sector_t)rs.resize_size) { | |
2455 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2456 | if (!new_disk_conf) { | |
778f271d | 2457 | retcode = ERR_NOMEM; |
9bcd2521 | 2458 | goto fail_ldev; |
778f271d PR |
2459 | } |
2460 | } | |
2461 | ||
b30ab791 AG |
2462 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2463 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2464 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2465 | ||
2466 | if (al_size_k > (16 * 1024 * 1024)) { | |
2467 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2468 | goto fail_ldev; | |
2469 | } | |
2470 | ||
2471 | if (al_size_k < MD_32kB_SECT/2) { | |
2472 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2473 | goto fail_ldev; | |
2474 | } | |
2475 | ||
b30ab791 | 2476 | if (device->state.conn != C_CONNECTED) { |
d752b269 PR |
2477 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2478 | goto fail_ldev; | |
2479 | } | |
2480 | ||
2481 | change_al_layout = true; | |
2482 | } | |
2483 | ||
b30ab791 AG |
2484 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2485 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2486 | |
daeda1cc | 2487 | if (new_disk_conf) { |
0500813f | 2488 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2489 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2490 | *new_disk_conf = *old_disk_conf; |
2491 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2492 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2493 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2494 | synchronize_rcu(); |
2495 | kfree(old_disk_conf); | |
b411b363 PR |
2496 | } |
2497 | ||
6495d2c6 | 2498 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2499 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2500 | drbd_md_sync(device); | |
2501 | put_ldev(device); | |
e96c9633 | 2502 | if (dd == DS_ERROR) { |
b411b363 PR |
2503 | retcode = ERR_NOMEM_BITMAP; |
2504 | goto fail; | |
d752b269 PR |
2505 | } else if (dd == DS_ERROR_SPACE_MD) { |
2506 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2507 | goto fail; | |
2508 | } else if (dd == DS_ERROR_SHRINK) { | |
2509 | retcode = ERR_IMPLICIT_SHRINK; | |
2510 | goto fail; | |
b411b363 | 2511 | } |
778f271d | 2512 | |
b30ab791 | 2513 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2514 | if (dd == DS_GREW) |
b30ab791 | 2515 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2516 | |
b30ab791 AG |
2517 | drbd_send_uuids(device); |
2518 | drbd_send_sizes(device, 1, ddsf); | |
778f271d PR |
2519 | } |
2520 | ||
b411b363 | 2521 | fail: |
3b98c0c2 | 2522 | drbd_adm_finish(info, retcode); |
b411b363 | 2523 | return 0; |
b411b363 | 2524 | |
9bcd2521 | 2525 | fail_ldev: |
b30ab791 | 2526 | put_ldev(device); |
9bcd2521 | 2527 | goto fail; |
b411b363 | 2528 | } |
b411b363 | 2529 | |
f399002e | 2530 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2531 | { |
3b98c0c2 | 2532 | enum drbd_ret_code retcode; |
b57a1e27 | 2533 | struct res_opts res_opts; |
f399002e | 2534 | int err; |
b411b363 | 2535 | |
44e52cfa | 2536 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2537 | if (!adm_ctx.reply_skb) |
2538 | return retcode; | |
2539 | if (retcode != NO_ERROR) | |
2540 | goto fail; | |
b411b363 | 2541 | |
eb6bea67 | 2542 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2543 | if (should_set_defaults(info)) |
b966b5dd | 2544 | set_res_opts_defaults(&res_opts); |
b411b363 | 2545 | |
b57a1e27 | 2546 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2547 | if (err && err != -ENOMSG) { |
b411b363 | 2548 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2549 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2550 | goto fail; |
2551 | } | |
2552 | ||
eb6bea67 | 2553 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2554 | if (err) { |
2555 | retcode = ERR_INVALID_REQUEST; | |
2556 | if (err == -ENOMEM) | |
2557 | retcode = ERR_NOMEM; | |
b411b363 PR |
2558 | } |
2559 | ||
b411b363 | 2560 | fail: |
3b98c0c2 | 2561 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2562 | return 0; |
2563 | } | |
2564 | ||
3b98c0c2 | 2565 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2566 | { |
b30ab791 | 2567 | struct drbd_device *device; |
3b98c0c2 LE |
2568 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2569 | ||
2570 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2571 | if (!adm_ctx.reply_skb) | |
2572 | return retcode; | |
2573 | if (retcode != NO_ERROR) | |
2574 | goto out; | |
2575 | ||
b30ab791 | 2576 | device = adm_ctx.device; |
b411b363 | 2577 | |
194bfb32 | 2578 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2579 | * resync just being finished, wait for it before requesting a new resync. |
2580 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2581 | drbd_suspend_io(device); |
2582 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
2583 | drbd_flush_workqueue(device); | |
194bfb32 | 2584 | |
0b2dafcd PR |
2585 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2586 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2587 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2588 | */ |
b30ab791 AG |
2589 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2590 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2591 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2592 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2593 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2594 | retcode = ERR_IO_MD_DISK; | |
2595 | } | |
2596 | } else | |
b30ab791 AG |
2597 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2598 | drbd_resume_io(device); | |
b411b363 | 2599 | |
3b98c0c2 LE |
2600 | out: |
2601 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2602 | return 0; |
2603 | } | |
2604 | ||
3b98c0c2 LE |
2605 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2606 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2607 | { |
3b98c0c2 | 2608 | enum drbd_ret_code retcode; |
194bfb32 | 2609 | |
3b98c0c2 LE |
2610 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2611 | if (!adm_ctx.reply_skb) | |
2612 | return retcode; | |
2613 | if (retcode != NO_ERROR) | |
2614 | goto out; | |
b411b363 | 2615 | |
b30ab791 | 2616 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
3b98c0c2 LE |
2617 | out: |
2618 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2619 | return 0; |
2620 | } | |
2621 | ||
b30ab791 | 2622 | static int drbd_bmio_set_susp_al(struct drbd_device *device) |
0778286a PR |
2623 | { |
2624 | int rv; | |
2625 | ||
b30ab791 AG |
2626 | rv = drbd_bmio_set_n_write(device); |
2627 | drbd_suspend_al(device); | |
0778286a PR |
2628 | return rv; |
2629 | } | |
2630 | ||
3b98c0c2 | 2631 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2632 | { |
25b0d6c8 | 2633 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2634 | struct drbd_device *device; |
25b0d6c8 PR |
2635 | |
2636 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2637 | if (!adm_ctx.reply_skb) | |
2638 | return retcode; | |
2639 | if (retcode != NO_ERROR) | |
2640 | goto out; | |
2641 | ||
b30ab791 | 2642 | device = adm_ctx.device; |
b411b363 | 2643 | |
194bfb32 | 2644 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2645 | * resync just being finished, wait for it before requesting a new resync. |
2646 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2647 | drbd_suspend_io(device); |
2648 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
2649 | drbd_flush_workqueue(device); | |
194bfb32 | 2650 | |
0b2dafcd PR |
2651 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2652 | * in the bitmap. Otherwise, try to start a resync handshake | |
2653 | * as sync source for full sync. | |
2654 | */ | |
b30ab791 | 2655 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2656 | /* The peer will get a resync upon connect anyways. Just make that |
2657 | into a full resync. */ | |
b30ab791 | 2658 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2659 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2660 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2661 | "set_n_write from invalidate_peer", |
2662 | BM_LOCKED_SET_ALLOWED)) | |
2663 | retcode = ERR_IO_MD_DISK; | |
2664 | } | |
2665 | } else | |
b30ab791 AG |
2666 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2667 | drbd_resume_io(device); | |
b411b363 | 2668 | |
25b0d6c8 PR |
2669 | out: |
2670 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2671 | return 0; |
2672 | } | |
2673 | ||
3b98c0c2 | 2674 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2675 | { |
3b98c0c2 | 2676 | enum drbd_ret_code retcode; |
b411b363 | 2677 | |
3b98c0c2 LE |
2678 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2679 | if (!adm_ctx.reply_skb) | |
2680 | return retcode; | |
2681 | if (retcode != NO_ERROR) | |
2682 | goto out; | |
b411b363 | 2683 | |
b30ab791 | 2684 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 LE |
2685 | retcode = ERR_PAUSE_IS_SET; |
2686 | out: | |
2687 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2688 | return 0; |
2689 | } | |
2690 | ||
3b98c0c2 | 2691 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2692 | { |
da9fbc27 | 2693 | union drbd_dev_state s; |
3b98c0c2 LE |
2694 | enum drbd_ret_code retcode; |
2695 | ||
2696 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2697 | if (!adm_ctx.reply_skb) | |
2698 | return retcode; | |
2699 | if (retcode != NO_ERROR) | |
2700 | goto out; | |
b411b363 | 2701 | |
b30ab791 AG |
2702 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2703 | s = adm_ctx.device->state; | |
cd88d030 PR |
2704 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2705 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2706 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2707 | } else { | |
2708 | retcode = ERR_PAUSE_IS_CLEAR; | |
2709 | } | |
2710 | } | |
b411b363 | 2711 | |
3b98c0c2 LE |
2712 | out: |
2713 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2714 | return 0; |
2715 | } | |
2716 | ||
3b98c0c2 | 2717 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2718 | { |
3b98c0c2 | 2719 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2720 | } |
2721 | ||
3b98c0c2 | 2722 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2723 | { |
b30ab791 | 2724 | struct drbd_device *device; |
3b98c0c2 LE |
2725 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2726 | ||
2727 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2728 | if (!adm_ctx.reply_skb) | |
2729 | return retcode; | |
2730 | if (retcode != NO_ERROR) | |
2731 | goto out; | |
2732 | ||
b30ab791 AG |
2733 | device = adm_ctx.device; |
2734 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2735 | drbd_uuid_new_current(device); | |
2736 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2737 | } |
b30ab791 AG |
2738 | drbd_suspend_io(device); |
2739 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2740 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2741 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2742 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2743 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2744 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2745 | } |
b30ab791 | 2746 | drbd_resume_io(device); |
265be2d0 | 2747 | |
3b98c0c2 LE |
2748 | out: |
2749 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2750 | return 0; |
2751 | } | |
2752 | ||
3b98c0c2 | 2753 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2754 | { |
3b98c0c2 | 2755 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2756 | } |
2757 | ||
251b8f8e AG |
2758 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2759 | struct drbd_resource *resource, | |
2760 | struct drbd_connection *connection, | |
2761 | struct drbd_device *device) | |
b411b363 | 2762 | { |
543cc10b LE |
2763 | struct nlattr *nla; |
2764 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2765 | if (!nla) | |
2766 | goto nla_put_failure; | |
251b8f8e AG |
2767 | if (device && |
2768 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2769 | goto nla_put_failure; |
77c556f6 | 2770 | if (nla_put_string(skb, T_ctx_resource_name, connection->resource->name)) |
26ec9287 | 2771 | goto nla_put_failure; |
251b8f8e AG |
2772 | if (connection) { |
2773 | if (connection->my_addr_len && | |
2774 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2775 | goto nla_put_failure; | |
2776 | if (connection->peer_addr_len && | |
2777 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2778 | goto nla_put_failure; | |
2779 | } | |
543cc10b LE |
2780 | nla_nest_end(skb, nla); |
2781 | return 0; | |
b411b363 | 2782 | |
543cc10b LE |
2783 | nla_put_failure: |
2784 | if (nla) | |
2785 | nla_nest_cancel(skb, nla); | |
2786 | return -EMSGSIZE; | |
2787 | } | |
b411b363 | 2788 | |
251b8f8e AG |
2789 | /* |
2790 | * Return the connection of @resource if @resource has exactly one connection. | |
2791 | */ | |
2792 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2793 | { | |
2794 | struct list_head *connections = &resource->connections; | |
2795 | ||
2796 | if (list_empty(connections) || connections->next->next != connections) | |
2797 | return NULL; | |
2798 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2799 | } | |
2800 | ||
2801 | int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, | |
3b98c0c2 | 2802 | const struct sib_info *sib) |
b411b363 | 2803 | { |
251b8f8e | 2804 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2805 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2806 | struct nlattr *nla; | |
2807 | int got_ldev; | |
3b98c0c2 LE |
2808 | int err = 0; |
2809 | int exclude_sensitive; | |
2810 | ||
2811 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2812 | * to. So we better exclude_sensitive information. | |
2813 | * | |
2814 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2815 | * in the context of the requesting user process. Exclude sensitive | |
2816 | * information, unless current has superuser. | |
2817 | * | |
2818 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2819 | * relies on the current implementation of netlink_dump(), which | |
2820 | * executes the dump callback successively from netlink_recvmsg(), | |
2821 | * always in the context of the receiving process */ | |
2822 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2823 | ||
b30ab791 | 2824 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2825 | |
2826 | /* We need to add connection name and volume number information still. | |
2827 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 2828 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 2829 | goto nla_put_failure; |
3b98c0c2 | 2830 | |
eb6bea67 | 2831 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
2832 | goto nla_put_failure; |
2833 | ||
daeda1cc | 2834 | rcu_read_lock(); |
f9eb7bf4 AG |
2835 | if (got_ldev) { |
2836 | struct disk_conf *disk_conf; | |
44ed167d | 2837 | |
b30ab791 | 2838 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2839 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2840 | } | |
2841 | if (!err) { | |
2842 | struct net_conf *nc; | |
2843 | ||
a6b32bc3 | 2844 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2845 | if (nc) |
2846 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2847 | } | |
44ed167d PR |
2848 | rcu_read_unlock(); |
2849 | if (err) | |
2850 | goto nla_put_failure; | |
3b98c0c2 | 2851 | |
3b98c0c2 LE |
2852 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2853 | if (!nla) | |
2854 | goto nla_put_failure; | |
26ec9287 | 2855 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2856 | nla_put_u32(skb, T_current_state, device->state.i) || |
2857 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2858 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2859 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2860 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2861 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2862 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2863 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2864 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2865 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2866 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2867 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2868 | goto nla_put_failure; |
3b98c0c2 LE |
2869 | |
2870 | if (got_ldev) { | |
39a1aa7f | 2871 | int err; |
b411b363 | 2872 | |
b30ab791 AG |
2873 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2874 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2875 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2876 | |
2877 | if (err) | |
2878 | goto nla_put_failure; | |
2879 | ||
b30ab791 AG |
2880 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
2881 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
2882 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 2883 | goto nla_put_failure; |
b30ab791 AG |
2884 | if (C_SYNC_SOURCE <= device->state.conn && |
2885 | C_PAUSED_SYNC_T >= device->state.conn) { | |
2886 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
2887 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 2888 | goto nla_put_failure; |
3b98c0c2 | 2889 | } |
b411b363 | 2890 | } |
b411b363 | 2891 | |
3b98c0c2 LE |
2892 | if (sib) { |
2893 | switch(sib->sib_reason) { | |
2894 | case SIB_SYNC_PROGRESS: | |
2895 | case SIB_GET_STATUS_REPLY: | |
2896 | break; | |
2897 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2898 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2899 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2900 | goto nla_put_failure; | |
3b98c0c2 LE |
2901 | break; |
2902 | case SIB_HELPER_POST: | |
26ec9287 AG |
2903 | if (nla_put_u32(skb, T_helper_exit_code, |
2904 | sib->helper_exit_code)) | |
2905 | goto nla_put_failure; | |
3b98c0c2 LE |
2906 | /* fall through */ |
2907 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2908 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2909 | goto nla_put_failure; | |
3b98c0c2 LE |
2910 | break; |
2911 | } | |
b411b363 | 2912 | } |
3b98c0c2 | 2913 | nla_nest_end(skb, nla); |
b411b363 | 2914 | |
3b98c0c2 LE |
2915 | if (0) |
2916 | nla_put_failure: | |
2917 | err = -EMSGSIZE; | |
2918 | if (got_ldev) | |
b30ab791 | 2919 | put_ldev(device); |
3b98c0c2 | 2920 | return err; |
b411b363 PR |
2921 | } |
2922 | ||
3b98c0c2 | 2923 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2924 | { |
3b98c0c2 LE |
2925 | enum drbd_ret_code retcode; |
2926 | int err; | |
b411b363 | 2927 | |
3b98c0c2 LE |
2928 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2929 | if (!adm_ctx.reply_skb) | |
2930 | return retcode; | |
2931 | if (retcode != NO_ERROR) | |
2932 | goto out; | |
b411b363 | 2933 | |
b30ab791 | 2934 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
2935 | if (err) { |
2936 | nlmsg_free(adm_ctx.reply_skb); | |
2937 | return err; | |
b411b363 | 2938 | } |
3b98c0c2 LE |
2939 | out: |
2940 | drbd_adm_finish(info, retcode); | |
2941 | return 0; | |
b411b363 PR |
2942 | } |
2943 | ||
4b7a530f | 2944 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2945 | { |
b30ab791 | 2946 | struct drbd_device *device; |
3b98c0c2 | 2947 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
2948 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
2949 | struct drbd_resource *resource = NULL; | |
77c556f6 | 2950 | struct drbd_resource *tmp; |
543cc10b LE |
2951 | unsigned volume = cb->args[1]; |
2952 | ||
2953 | /* Open coded, deferred, iteration: | |
77c556f6 | 2954 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
2955 | * connection = "first connection of resource or undefined"; |
2956 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
2957 | * ... |
2958 | * } | |
2959 | * } | |
77c556f6 | 2960 | * where resource is cb->args[0]; |
543cc10b LE |
2961 | * and i is cb->args[1]; |
2962 | * | |
71932efc LE |
2963 | * cb->args[2] indicates if we shall loop over all resources, |
2964 | * or just dump all volumes of a single resource. | |
2965 | * | |
3b98c0c2 LE |
2966 | * This may miss entries inserted after this dump started, |
2967 | * or entries deleted before they are reached. | |
543cc10b | 2968 | * |
b30ab791 | 2969 | * We need to make sure the device won't disappear while |
543cc10b LE |
2970 | * we are looking at it, and revalidate our iterators |
2971 | * on each iteration. | |
2972 | */ | |
b411b363 | 2973 | |
05a10ec7 | 2974 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 2975 | rcu_read_lock(); |
543cc10b | 2976 | /* revalidate iterator position */ |
77c556f6 | 2977 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
2978 | if (pos == NULL) { |
2979 | /* first iteration */ | |
2980 | pos = tmp; | |
77c556f6 | 2981 | resource = pos; |
543cc10b LE |
2982 | break; |
2983 | } | |
2984 | if (tmp == pos) { | |
77c556f6 | 2985 | resource = pos; |
543cc10b LE |
2986 | break; |
2987 | } | |
b411b363 | 2988 | } |
77c556f6 AG |
2989 | if (resource) { |
2990 | next_resource: | |
251b8f8e AG |
2991 | device = idr_get_next(&resource->devices, &volume); |
2992 | if (!device) { | |
77c556f6 AG |
2993 | /* No more volumes to dump on this resource. |
2994 | * Advance resource iterator. */ | |
2995 | pos = list_entry_rcu(resource->resources.next, | |
2996 | struct drbd_resource, resources); | |
2997 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 2998 | if (volume != 0) { |
71932efc LE |
2999 | /* If we reached the end of the list, |
3000 | * or only a single resource dump was requested, | |
3001 | * we are done. */ | |
77c556f6 | 3002 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3003 | goto out; |
543cc10b | 3004 | volume = 0; |
77c556f6 AG |
3005 | resource = pos; |
3006 | goto next_resource; | |
543cc10b LE |
3007 | } |
3008 | } | |
3009 | ||
98683650 | 3010 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3011 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3012 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3013 | if (!dh) | |
543cc10b LE |
3014 | goto out; |
3015 | ||
251b8f8e | 3016 | if (!device) { |
bde89a9e | 3017 | /* This is a connection without a single volume. |
367d675d LE |
3018 | * Suprisingly enough, it may have a network |
3019 | * configuration. */ | |
251b8f8e AG |
3020 | struct drbd_connection *connection; |
3021 | ||
543cc10b LE |
3022 | dh->minor = -1U; |
3023 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3024 | connection = the_only_connection(resource); |
3025 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3026 | goto cancel; |
251b8f8e AG |
3027 | if (connection) { |
3028 | struct net_conf *nc; | |
3029 | ||
3030 | nc = rcu_dereference(connection->net_conf); | |
3031 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3032 | goto cancel; | |
3033 | } | |
367d675d | 3034 | goto done; |
543cc10b | 3035 | } |
b411b363 | 3036 | |
0b0ba1ef | 3037 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3038 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3039 | |
b30ab791 | 3040 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3041 | dh->ret_code = NO_ERROR; |
3042 | ||
b30ab791 | 3043 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3044 | cancel: |
3b98c0c2 | 3045 | genlmsg_cancel(skb, dh); |
543cc10b | 3046 | goto out; |
3b98c0c2 | 3047 | } |
367d675d | 3048 | done: |
3b98c0c2 | 3049 | genlmsg_end(skb, dh); |
bde89a9e | 3050 | } |
b411b363 | 3051 | |
543cc10b | 3052 | out: |
c141ebda | 3053 | rcu_read_unlock(); |
543cc10b | 3054 | /* where to start the next iteration */ |
bde89a9e | 3055 | cb->args[0] = (long)pos; |
77c556f6 | 3056 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3057 | |
77c556f6 | 3058 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3059 | * Which will terminate the dump. */ |
3b98c0c2 | 3060 | return skb->len; |
b411b363 PR |
3061 | } |
3062 | ||
71932efc LE |
3063 | /* |
3064 | * Request status of all resources, or of all volumes within a single resource. | |
3065 | * | |
3066 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3067 | * Which means we cannot use the family->attrbuf or other such members, because | |
3068 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3069 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3070 | * | |
3071 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3072 | */ |
71932efc | 3073 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3074 | { |
71932efc LE |
3075 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3076 | struct nlattr *nla; | |
7c3063cc | 3077 | const char *resource_name; |
4bc76048 | 3078 | struct drbd_resource *resource; |
7c3063cc | 3079 | int maxtype; |
71932efc LE |
3080 | |
3081 | /* Is this a followup call? */ | |
3082 | if (cb->args[0]) { | |
3083 | /* ... of a single resource dump, | |
3084 | * and the resource iterator has been advanced already? */ | |
3085 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3086 | return 0; /* DONE. */ | |
3087 | goto dump; | |
3088 | } | |
3089 | ||
3090 | /* First call (from netlink_dump_start). We need to figure out | |
3091 | * which resource(s) the user wants us to dump. */ | |
3092 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3093 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3094 | DRBD_NLA_CFG_CONTEXT); | |
3095 | ||
3096 | /* No explicit context given. Dump all. */ | |
3097 | if (!nla) | |
3098 | goto dump; | |
7c3063cc AG |
3099 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3100 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3101 | if (IS_ERR(nla)) | |
3102 | return PTR_ERR(nla); | |
71932efc LE |
3103 | /* context given, but no name present? */ |
3104 | if (!nla) | |
3105 | return -EINVAL; | |
7c3063cc | 3106 | resource_name = nla_data(nla); |
4bc76048 AG |
3107 | if (!*resource_name) |
3108 | return -ENODEV; | |
3109 | resource = drbd_find_resource(resource_name); | |
3110 | if (!resource) | |
71932efc LE |
3111 | return -ENODEV; |
3112 | ||
4bc76048 | 3113 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3114 | |
71932efc | 3115 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3116 | * only dump this connection. */ |
4bc76048 | 3117 | cb->args[0] = (long)resource; |
71932efc | 3118 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3119 | cb->args[2] = (long)resource; |
71932efc LE |
3120 | |
3121 | dump: | |
3122 | return get_one_status(skb, cb); | |
3123 | } | |
b411b363 | 3124 | |
3b98c0c2 | 3125 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3126 | { |
3b98c0c2 LE |
3127 | enum drbd_ret_code retcode; |
3128 | struct timeout_parms tp; | |
3129 | int err; | |
b411b363 | 3130 | |
3b98c0c2 LE |
3131 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3132 | if (!adm_ctx.reply_skb) | |
3133 | return retcode; | |
3134 | if (retcode != NO_ERROR) | |
3135 | goto out; | |
b411b363 | 3136 | |
3b98c0c2 | 3137 | tp.timeout_type = |
b30ab791 AG |
3138 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3139 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3140 | UT_DEFAULT; |
b411b363 | 3141 | |
3b98c0c2 LE |
3142 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3143 | if (err) { | |
3144 | nlmsg_free(adm_ctx.reply_skb); | |
3145 | return err; | |
3146 | } | |
3147 | out: | |
3148 | drbd_adm_finish(info, retcode); | |
3149 | return 0; | |
b411b363 PR |
3150 | } |
3151 | ||
3b98c0c2 | 3152 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3153 | { |
b30ab791 | 3154 | struct drbd_device *device; |
3b98c0c2 | 3155 | enum drbd_ret_code retcode; |
58ffa580 | 3156 | struct start_ov_parms parms; |
b411b363 | 3157 | |
3b98c0c2 LE |
3158 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3159 | if (!adm_ctx.reply_skb) | |
3160 | return retcode; | |
3161 | if (retcode != NO_ERROR) | |
3162 | goto out; | |
873b0d5f | 3163 | |
b30ab791 | 3164 | device = adm_ctx.device; |
58ffa580 LE |
3165 | |
3166 | /* resume from last known position, if possible */ | |
b30ab791 | 3167 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3168 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3169 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3170 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3171 | if (err) { |
3172 | retcode = ERR_MANDATORY_TAG; | |
3173 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3174 | goto out; | |
3175 | } | |
b411b363 | 3176 | } |
58ffa580 | 3177 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3178 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3179 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3180 | |
3181 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3182 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3183 | drbd_suspend_io(device); |
3184 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3185 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3186 | drbd_resume_io(device); | |
3b98c0c2 LE |
3187 | out: |
3188 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3189 | return 0; |
3190 | } | |
3191 | ||
3192 | ||
3b98c0c2 | 3193 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3194 | { |
b30ab791 | 3195 | struct drbd_device *device; |
3b98c0c2 | 3196 | enum drbd_ret_code retcode; |
b411b363 PR |
3197 | int skip_initial_sync = 0; |
3198 | int err; | |
3b98c0c2 | 3199 | struct new_c_uuid_parms args; |
b411b363 | 3200 | |
3b98c0c2 LE |
3201 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3202 | if (!adm_ctx.reply_skb) | |
3203 | return retcode; | |
3204 | if (retcode != NO_ERROR) | |
3205 | goto out_nolock; | |
b411b363 | 3206 | |
b30ab791 | 3207 | device = adm_ctx.device; |
3b98c0c2 LE |
3208 | memset(&args, 0, sizeof(args)); |
3209 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3210 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3211 | if (err) { |
3212 | retcode = ERR_MANDATORY_TAG; | |
3213 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3214 | goto out_nolock; | |
3215 | } | |
b411b363 PR |
3216 | } |
3217 | ||
b30ab791 | 3218 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3219 | |
b30ab791 | 3220 | if (!get_ldev(device)) { |
b411b363 PR |
3221 | retcode = ERR_NO_DISK; |
3222 | goto out; | |
3223 | } | |
3224 | ||
3225 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3226 | if (device->state.conn == C_CONNECTED && |
3227 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3228 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3229 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3230 | skip_initial_sync = 1; |
b30ab791 | 3231 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3232 | retcode = ERR_CONNECTED; |
3233 | goto out_dec; | |
3234 | } | |
3235 | ||
b30ab791 AG |
3236 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3237 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3238 | |
3239 | if (args.clear_bm) { | |
b30ab791 | 3240 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3241 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3242 | if (err) { |
d0180171 | 3243 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3244 | retcode = ERR_IO_MD_DISK; |
3245 | } | |
3246 | if (skip_initial_sync) { | |
b30ab791 AG |
3247 | drbd_send_uuids_skip_initial_sync(device); |
3248 | _drbd_uuid_set(device, UI_BITMAP, 0); | |
3249 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3250 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3251 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3252 | CS_VERBOSE, NULL); |
0500813f | 3253 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3254 | } |
3255 | } | |
3256 | ||
b30ab791 | 3257 | drbd_md_sync(device); |
b411b363 | 3258 | out_dec: |
b30ab791 | 3259 | put_ldev(device); |
b411b363 | 3260 | out: |
b30ab791 | 3261 | mutex_unlock(device->state_mutex); |
3b98c0c2 LE |
3262 | out_nolock: |
3263 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3264 | return 0; |
3265 | } | |
3266 | ||
3b98c0c2 | 3267 | static enum drbd_ret_code |
7c3063cc | 3268 | drbd_check_resource_name(const char *name) |
b411b363 | 3269 | { |
3b98c0c2 | 3270 | if (!name || !name[0]) { |
7c3063cc | 3271 | drbd_msg_put_info("resource name missing"); |
3b98c0c2 | 3272 | return ERR_MANDATORY_TAG; |
b411b363 | 3273 | } |
3b98c0c2 LE |
3274 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3275 | * we must not allow slashes */ | |
3276 | if (strchr(name, '/')) { | |
7c3063cc | 3277 | drbd_msg_put_info("invalid resource name"); |
3b98c0c2 | 3278 | return ERR_INVALID_REQUEST; |
b411b363 | 3279 | } |
3b98c0c2 | 3280 | return NO_ERROR; |
774b3055 | 3281 | } |
b411b363 | 3282 | |
789c1b62 | 3283 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3284 | { |
3b98c0c2 | 3285 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3286 | struct res_opts res_opts; |
3287 | int err; | |
b411b363 | 3288 | |
3b98c0c2 LE |
3289 | retcode = drbd_adm_prepare(skb, info, 0); |
3290 | if (!adm_ctx.reply_skb) | |
3291 | return retcode; | |
3292 | if (retcode != NO_ERROR) | |
3293 | goto out; | |
b411b363 | 3294 | |
afbbfa88 AG |
3295 | set_res_opts_defaults(&res_opts); |
3296 | err = res_opts_from_attrs(&res_opts, info); | |
3297 | if (err && err != -ENOMSG) { | |
3298 | retcode = ERR_MANDATORY_TAG; | |
3299 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3300 | goto out; | |
b411b363 PR |
3301 | } |
3302 | ||
7c3063cc | 3303 | retcode = drbd_check_resource_name(adm_ctx.resource_name); |
3b98c0c2 LE |
3304 | if (retcode != NO_ERROR) |
3305 | goto out; | |
b411b363 | 3306 | |
5c661042 | 3307 | if (adm_ctx.resource) { |
38f19616 LE |
3308 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3309 | retcode = ERR_INVALID_REQUEST; | |
789c1b62 | 3310 | drbd_msg_put_info("resource exists"); |
38f19616 LE |
3311 | } |
3312 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3313 | goto out; |
b411b363 | 3314 | } |
b411b363 | 3315 | |
afbbfa88 | 3316 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3317 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
3318 | out: |
3319 | drbd_adm_finish(info, retcode); | |
3320 | return 0; | |
b411b363 PR |
3321 | } |
3322 | ||
05a10ec7 | 3323 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3324 | { |
3b98c0c2 LE |
3325 | struct drbd_genlmsghdr *dh = info->userhdr; |
3326 | enum drbd_ret_code retcode; | |
b411b363 | 3327 | |
44e52cfa | 3328 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3329 | if (!adm_ctx.reply_skb) |
3330 | return retcode; | |
3331 | if (retcode != NO_ERROR) | |
3332 | goto out; | |
b411b363 | 3333 | |
f2257a56 | 3334 | if (dh->minor > MINORMASK) { |
3b98c0c2 LE |
3335 | drbd_msg_put_info("requested minor out of range"); |
3336 | retcode = ERR_INVALID_REQUEST; | |
3337 | goto out; | |
b411b363 | 3338 | } |
0c8e36d9 | 3339 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
3340 | drbd_msg_put_info("requested volume id out of range"); |
3341 | retcode = ERR_INVALID_REQUEST; | |
3342 | goto out; | |
b411b363 | 3343 | } |
b411b363 | 3344 | |
38f19616 | 3345 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3346 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3347 | if (adm_ctx.device) { |
38f19616 LE |
3348 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
3349 | retcode = ERR_MINOR_EXISTS; | |
3350 | /* else: still NO_ERROR */ | |
3351 | goto out; | |
b411b363 | 3352 | } |
38f19616 | 3353 | |
59515a2e | 3354 | retcode = drbd_create_device(adm_ctx.resource, dh->minor, adm_ctx.volume); |
3b98c0c2 LE |
3355 | out: |
3356 | drbd_adm_finish(info, retcode); | |
3357 | return 0; | |
b411b363 PR |
3358 | } |
3359 | ||
05a10ec7 | 3360 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3361 | { |
b30ab791 AG |
3362 | if (device->state.disk == D_DISKLESS && |
3363 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3364 | * we may want to delete a minor from a live replication group. |
3365 | */ | |
b30ab791 AG |
3366 | device->state.role == R_SECONDARY) { |
3367 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3368 | CS_VERBOSE + CS_WAIT_COMPLETE); |
f82795d6 | 3369 | drbd_delete_device(device); |
85f75dd7 LE |
3370 | return NO_ERROR; |
3371 | } else | |
3372 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3373 | } |
3374 | ||
05a10ec7 | 3375 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3376 | { |
3b98c0c2 | 3377 | enum drbd_ret_code retcode; |
b411b363 | 3378 | |
3b98c0c2 LE |
3379 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3380 | if (!adm_ctx.reply_skb) | |
3381 | return retcode; | |
3382 | if (retcode != NO_ERROR) | |
3383 | goto out; | |
b411b363 | 3384 | |
05a10ec7 | 3385 | retcode = adm_del_minor(adm_ctx.device); |
85f75dd7 LE |
3386 | out: |
3387 | drbd_adm_finish(info, retcode); | |
3388 | return 0; | |
b411b363 PR |
3389 | } |
3390 | ||
85f75dd7 | 3391 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3392 | { |
b6f85ef9 AG |
3393 | struct drbd_resource *resource; |
3394 | struct drbd_connection *connection; | |
3395 | struct drbd_device *device; | |
f3dfa40a | 3396 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3397 | unsigned i; |
b411b363 | 3398 | |
a10f6b8a | 3399 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3400 | if (!adm_ctx.reply_skb) |
3401 | return retcode; | |
3402 | if (retcode != NO_ERROR) | |
3403 | goto out; | |
b411b363 | 3404 | |
b6f85ef9 | 3405 | resource = adm_ctx.resource; |
85f75dd7 | 3406 | /* demote */ |
b6f85ef9 AG |
3407 | for_each_connection(connection, resource) { |
3408 | struct drbd_peer_device *peer_device; | |
3409 | ||
3410 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3411 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3412 | if (retcode < SS_SUCCESS) { | |
3413 | drbd_msg_put_info("failed to demote"); | |
3414 | goto out; | |
3415 | } | |
3416 | } | |
3417 | ||
3418 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3419 | if (retcode < SS_SUCCESS) { |
b6f85ef9 | 3420 | drbd_msg_put_info("failed to disconnect"); |
c141ebda | 3421 | goto out; |
85f75dd7 | 3422 | } |
b411b363 | 3423 | } |
b411b363 | 3424 | |
85f75dd7 | 3425 | /* detach */ |
b6f85ef9 AG |
3426 | idr_for_each_entry(&resource->devices, device, i) { |
3427 | retcode = adm_detach(device, 0); | |
27012382 | 3428 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
85f75dd7 | 3429 | drbd_msg_put_info("failed to detach"); |
c141ebda | 3430 | goto out; |
85f75dd7 LE |
3431 | } |
3432 | } | |
b411b363 | 3433 | |
bde89a9e | 3434 | /* If we reach this, all volumes (of this connection) are Secondary, |
f3dfa40a | 3435 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have |
c141ebda | 3436 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
b6f85ef9 AG |
3437 | for_each_connection(connection, resource) |
3438 | drbd_thread_stop(&connection->worker); | |
b411b363 | 3439 | |
f3dfa40a | 3440 | /* Now, nothing can fail anymore */ |
b411b363 | 3441 | |
85f75dd7 | 3442 | /* delete volumes */ |
b6f85ef9 AG |
3443 | idr_for_each_entry(&resource->devices, device, i) { |
3444 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3445 | if (retcode != NO_ERROR) { |
3446 | /* "can not happen" */ | |
3447 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 | 3448 | goto out; |
85f75dd7 LE |
3449 | } |
3450 | } | |
b411b363 | 3451 | |
b6f85ef9 AG |
3452 | list_del_rcu(&resource->resources); |
3453 | synchronize_rcu(); | |
3454 | drbd_free_resource(resource); | |
3455 | retcode = NO_ERROR; | |
b411b363 | 3456 | |
3b98c0c2 LE |
3457 | out: |
3458 | drbd_adm_finish(info, retcode); | |
3459 | return 0; | |
b411b363 PR |
3460 | } |
3461 | ||
789c1b62 | 3462 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3463 | { |
77c556f6 AG |
3464 | struct drbd_resource *resource; |
3465 | struct drbd_connection *connection; | |
3b98c0c2 | 3466 | enum drbd_ret_code retcode; |
b411b363 | 3467 | |
44e52cfa | 3468 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3469 | if (!adm_ctx.reply_skb) |
3470 | return retcode; | |
3471 | if (retcode != NO_ERROR) | |
3472 | goto out; | |
b411b363 | 3473 | |
77c556f6 AG |
3474 | resource = adm_ctx.resource; |
3475 | for_each_connection(connection, resource) { | |
3476 | if (connection->cstate > C_STANDALONE) { | |
3477 | retcode = ERR_NET_CONFIGURED; | |
3478 | goto out; | |
3479 | } | |
3480 | } | |
3481 | if (!idr_is_empty(&resource->devices)) { | |
789c1b62 | 3482 | retcode = ERR_RES_IN_USE; |
77c556f6 | 3483 | goto out; |
b411b363 PR |
3484 | } |
3485 | ||
77c556f6 AG |
3486 | list_del_rcu(&resource->resources); |
3487 | for_each_connection(connection, resource) | |
3488 | drbd_thread_stop(&connection->worker); | |
3489 | synchronize_rcu(); | |
3490 | drbd_free_resource(resource); | |
3491 | retcode = NO_ERROR; | |
3b98c0c2 LE |
3492 | out: |
3493 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3494 | return 0; |
3495 | } | |
3496 | ||
b30ab791 | 3497 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3498 | { |
3b98c0c2 LE |
3499 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3500 | struct sk_buff *msg; | |
3501 | struct drbd_genlmsghdr *d_out; | |
3502 | unsigned seq; | |
3503 | int err = -ENOMEM; | |
3504 | ||
ef86b779 | 3505 | if (sib->sib_reason == SIB_SYNC_PROGRESS) { |
b30ab791 AG |
3506 | if (time_after(jiffies, device->rs_last_bcast + HZ)) |
3507 | device->rs_last_bcast = jiffies; | |
ef86b779 PR |
3508 | else |
3509 | return; | |
3510 | } | |
b411b363 | 3511 | |
3b98c0c2 LE |
3512 | seq = atomic_inc_return(&drbd_genl_seq); |
3513 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3514 | if (!msg) | |
3515 | goto failed; | |
3516 | ||
3517 | err = -EMSGSIZE; | |
3518 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3519 | if (!d_out) /* cannot happen, but anyways. */ | |
3520 | goto nla_put_failure; | |
b30ab791 | 3521 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3522 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3523 | |
b30ab791 | 3524 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3525 | goto nla_put_failure; |
3526 | genlmsg_end(msg, d_out); | |
3527 | err = drbd_genl_multicast_events(msg, 0); | |
3528 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3529 | if (err && err != -ESRCH) | |
3530 | goto failed; | |
b411b363 | 3531 | |
3b98c0c2 | 3532 | return; |
b411b363 | 3533 | |
3b98c0c2 LE |
3534 | nla_put_failure: |
3535 | nlmsg_free(msg); | |
3536 | failed: | |
d0180171 | 3537 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3538 | "Event seq:%u sib_reason:%u\n", |
3539 | err, seq, sib->sib_reason); | |
b411b363 | 3540 | } |