]>
Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
b411b363 PR |
26 | #include <linux/module.h> |
27 | #include <linux/drbd.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/fs.h> | |
30 | #include <linux/file.h> | |
31 | #include <linux/slab.h> | |
b411b363 PR |
32 | #include <linux/blkpg.h> |
33 | #include <linux/cpumask.h> | |
34 | #include "drbd_int.h" | |
a3603a6e | 35 | #include "drbd_protocol.h" |
265be2d0 | 36 | #include "drbd_req.h" |
b411b363 | 37 | #include <asm/unaligned.h> |
b411b363 | 38 | #include <linux/drbd_limits.h> |
87f7be4c | 39 | #include <linux/kthread.h> |
b411b363 | 40 | |
3b98c0c2 LE |
41 | #include <net/genetlink.h> |
42 | ||
43 | /* .doit */ | |
44 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
45 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | ||
05a10ec7 AG |
47 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
48 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 49 | |
789c1b62 AG |
50 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
51 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 52 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
53 | |
54 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
55 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 56 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
57 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
58 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
62 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 71 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
72 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
73 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
74 | /* .dumpit */ | |
75 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
76 | ||
77 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 78 | #include "drbd_nla.h" |
3b98c0c2 LE |
79 | #include <linux/genl_magic_func.h> |
80 | ||
81 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
82 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
83 | ||
3b98c0c2 LE |
84 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) |
85 | { | |
86 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
87 | if (genlmsg_reply(skb, info)) | |
88 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 89 | } |
3b98c0c2 LE |
90 | |
91 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
92 | * reason it could fail was no space in skb, and there are 4k available. */ | |
a910b123 | 93 | int drbd_msg_put_info(struct sk_buff *skb, const char *info) |
3b98c0c2 | 94 | { |
3b98c0c2 LE |
95 | struct nlattr *nla; |
96 | int err = -EMSGSIZE; | |
97 | ||
98 | if (!info || !info[0]) | |
99 | return 0; | |
100 | ||
101 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
102 | if (!nla) | |
103 | return err; | |
104 | ||
105 | err = nla_put_string(skb, T_info_text, info); | |
106 | if (err) { | |
107 | nla_nest_cancel(skb, nla); | |
108 | return err; | |
109 | } else | |
110 | nla_nest_end(skb, nla); | |
111 | return 0; | |
b411b363 PR |
112 | } |
113 | ||
3b98c0c2 LE |
114 | /* This would be a good candidate for a "pre_doit" hook, |
115 | * and per-family private info->pointers. | |
116 | * But we need to stay compatible with older kernels. | |
117 | * If it returns successfully, adm_ctx members are valid. | |
9e276872 LE |
118 | * |
119 | * At this point, we still rely on the global genl_lock(). | |
120 | * If we want to avoid that, and allow "genl_family.parallel_ops", we may need | |
121 | * to add additional synchronization against object destruction/modification. | |
3b98c0c2 LE |
122 | */ |
123 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 124 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 125 | #define DRBD_ADM_NEED_CONNECTION 4 |
a910b123 LE |
126 | static int drbd_adm_prepare(struct drbd_config_context *adm_ctx, |
127 | struct sk_buff *skb, struct genl_info *info, unsigned flags) | |
3b98c0c2 LE |
128 | { |
129 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
130 | const u8 cmd = info->genlhdr->cmd; | |
131 | int err; | |
132 | ||
a910b123 | 133 | memset(adm_ctx, 0, sizeof(*adm_ctx)); |
3b98c0c2 LE |
134 | |
135 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 136 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
137 | return -EPERM; |
138 | ||
a910b123 LE |
139 | adm_ctx->reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); |
140 | if (!adm_ctx->reply_skb) { | |
1e2a2551 | 141 | err = -ENOMEM; |
3b98c0c2 | 142 | goto fail; |
1e2a2551 | 143 | } |
3b98c0c2 | 144 | |
a910b123 | 145 | adm_ctx->reply_dh = genlmsg_put_reply(adm_ctx->reply_skb, |
3b98c0c2 LE |
146 | info, &drbd_genl_family, 0, cmd); |
147 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
148 | * but anyways */ | |
a910b123 | 149 | if (!adm_ctx->reply_dh) { |
1e2a2551 | 150 | err = -ENOMEM; |
3b98c0c2 | 151 | goto fail; |
1e2a2551 | 152 | } |
3b98c0c2 | 153 | |
a910b123 LE |
154 | adm_ctx->reply_dh->minor = d_in->minor; |
155 | adm_ctx->reply_dh->ret_code = NO_ERROR; | |
3b98c0c2 | 156 | |
a910b123 | 157 | adm_ctx->volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
158 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
159 | struct nlattr *nla; | |
160 | /* parse and validate only */ | |
f399002e | 161 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
162 | if (err) |
163 | goto fail; | |
164 | ||
165 | /* It was present, and valid, | |
166 | * copy it over to the reply skb. */ | |
a910b123 | 167 | err = nla_put_nohdr(adm_ctx->reply_skb, |
3b98c0c2 LE |
168 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, |
169 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
170 | if (err) | |
171 | goto fail; | |
172 | ||
9e276872 | 173 | /* and assign stuff to the adm_ctx */ |
3b98c0c2 | 174 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; |
089c075d | 175 | if (nla) |
a910b123 | 176 | adm_ctx->volume = nla_get_u32(nla); |
7c3063cc | 177 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 178 | if (nla) |
a910b123 LE |
179 | adm_ctx->resource_name = nla_data(nla); |
180 | adm_ctx->my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; | |
181 | adm_ctx->peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
182 | if ((adm_ctx->my_addr && | |
183 | nla_len(adm_ctx->my_addr) > sizeof(adm_ctx->connection->my_addr)) || | |
184 | (adm_ctx->peer_addr && | |
185 | nla_len(adm_ctx->peer_addr) > sizeof(adm_ctx->connection->peer_addr))) { | |
089c075d AG |
186 | err = -EINVAL; |
187 | goto fail; | |
188 | } | |
189 | } | |
3b98c0c2 | 190 | |
a910b123 LE |
191 | adm_ctx->minor = d_in->minor; |
192 | adm_ctx->device = minor_to_device(d_in->minor); | |
9e276872 LE |
193 | |
194 | /* We are protected by the global genl_lock(). | |
195 | * But we may explicitly drop it/retake it in drbd_adm_set_role(), | |
196 | * so make sure this object stays around. */ | |
197 | if (adm_ctx->device) | |
198 | kref_get(&adm_ctx->device->kref); | |
199 | ||
a910b123 LE |
200 | if (adm_ctx->resource_name) { |
201 | adm_ctx->resource = drbd_find_resource(adm_ctx->resource_name); | |
4bc76048 | 202 | } |
3b98c0c2 | 203 | |
a910b123 LE |
204 | if (!adm_ctx->device && (flags & DRBD_ADM_NEED_MINOR)) { |
205 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown minor"); | |
3b98c0c2 LE |
206 | return ERR_MINOR_INVALID; |
207 | } | |
a910b123 LE |
208 | if (!adm_ctx->resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
209 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown resource"); | |
210 | if (adm_ctx->resource_name) | |
a10f6b8a | 211 | return ERR_RES_NOT_KNOWN; |
3b98c0c2 LE |
212 | return ERR_INVALID_REQUEST; |
213 | } | |
214 | ||
089c075d | 215 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
a910b123 LE |
216 | if (adm_ctx->resource) { |
217 | drbd_msg_put_info(adm_ctx->reply_skb, "no resource name expected"); | |
089c075d AG |
218 | return ERR_INVALID_REQUEST; |
219 | } | |
a910b123 LE |
220 | if (adm_ctx->device) { |
221 | drbd_msg_put_info(adm_ctx->reply_skb, "no minor number expected"); | |
089c075d AG |
222 | return ERR_INVALID_REQUEST; |
223 | } | |
a910b123 LE |
224 | if (adm_ctx->my_addr && adm_ctx->peer_addr) |
225 | adm_ctx->connection = conn_get_by_addrs(nla_data(adm_ctx->my_addr), | |
226 | nla_len(adm_ctx->my_addr), | |
227 | nla_data(adm_ctx->peer_addr), | |
228 | nla_len(adm_ctx->peer_addr)); | |
229 | if (!adm_ctx->connection) { | |
230 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown connection"); | |
089c075d AG |
231 | return ERR_INVALID_REQUEST; |
232 | } | |
233 | } | |
234 | ||
3b98c0c2 | 235 | /* some more paranoia, if the request was over-determined */ |
a910b123 LE |
236 | if (adm_ctx->device && adm_ctx->resource && |
237 | adm_ctx->device->resource != adm_ctx->resource) { | |
4bc76048 | 238 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", |
a910b123 LE |
239 | adm_ctx->minor, adm_ctx->resource->name, |
240 | adm_ctx->device->resource->name); | |
241 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists in different resource"); | |
527f4b24 LE |
242 | return ERR_INVALID_REQUEST; |
243 | } | |
a910b123 LE |
244 | if (adm_ctx->device && |
245 | adm_ctx->volume != VOLUME_UNSPECIFIED && | |
246 | adm_ctx->volume != adm_ctx->device->vnr) { | |
3b98c0c2 | 247 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
a910b123 LE |
248 | adm_ctx->minor, adm_ctx->volume, |
249 | adm_ctx->device->vnr, | |
250 | adm_ctx->device->resource->name); | |
251 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists as different volume"); | |
3b98c0c2 LE |
252 | return ERR_INVALID_REQUEST; |
253 | } | |
0ace9dfa | 254 | |
9e276872 LE |
255 | /* still, provide adm_ctx->resource always, if possible. */ |
256 | if (!adm_ctx->resource) { | |
257 | adm_ctx->resource = adm_ctx->device ? adm_ctx->device->resource | |
258 | : adm_ctx->connection ? adm_ctx->connection->resource : NULL; | |
259 | if (adm_ctx->resource) | |
260 | kref_get(&adm_ctx->resource->kref); | |
261 | } | |
262 | ||
3b98c0c2 LE |
263 | return NO_ERROR; |
264 | ||
265 | fail: | |
a910b123 LE |
266 | nlmsg_free(adm_ctx->reply_skb); |
267 | adm_ctx->reply_skb = NULL; | |
1e2a2551 | 268 | return err; |
3b98c0c2 LE |
269 | } |
270 | ||
a910b123 LE |
271 | static int drbd_adm_finish(struct drbd_config_context *adm_ctx, |
272 | struct genl_info *info, int retcode) | |
3b98c0c2 | 273 | { |
9e276872 LE |
274 | if (adm_ctx->device) { |
275 | kref_put(&adm_ctx->device->kref, drbd_destroy_device); | |
276 | adm_ctx->device = NULL; | |
277 | } | |
a910b123 LE |
278 | if (adm_ctx->connection) { |
279 | kref_put(&adm_ctx->connection->kref, &drbd_destroy_connection); | |
280 | adm_ctx->connection = NULL; | |
0ace9dfa | 281 | } |
a910b123 LE |
282 | if (adm_ctx->resource) { |
283 | kref_put(&adm_ctx->resource->kref, drbd_destroy_resource); | |
284 | adm_ctx->resource = NULL; | |
4bc76048 | 285 | } |
0ace9dfa | 286 | |
a910b123 | 287 | if (!adm_ctx->reply_skb) |
3b98c0c2 LE |
288 | return -ENOMEM; |
289 | ||
a910b123 LE |
290 | adm_ctx->reply_dh->ret_code = retcode; |
291 | drbd_adm_send_reply(adm_ctx->reply_skb, info); | |
3b98c0c2 LE |
292 | return 0; |
293 | } | |
b411b363 | 294 | |
bde89a9e | 295 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 296 | { |
6b75dced | 297 | char *afs; |
b411b363 | 298 | |
089c075d | 299 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 300 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
301 | return; |
302 | ||
bde89a9e | 303 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
304 | case AF_INET6: |
305 | afs = "ipv6"; | |
306 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 307 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 308 | break; |
089c075d AG |
309 | case AF_INET: |
310 | afs = "ipv4"; | |
311 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 312 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 313 | break; |
089c075d AG |
314 | default: |
315 | afs = "ssocks"; | |
316 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 317 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 318 | } |
089c075d | 319 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 320 | } |
b411b363 | 321 | |
b30ab791 | 322 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
323 | { |
324 | char *envp[] = { "HOME=/", | |
325 | "TERM=linux", | |
326 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
327 | (char[20]) { }, /* address family */ |
328 | (char[60]) { }, /* address */ | |
b411b363 | 329 | NULL }; |
6b75dced | 330 | char mb[12]; |
b411b363 | 331 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 332 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 333 | struct sib_info sib; |
b411b363 PR |
334 | int ret; |
335 | ||
bde89a9e AG |
336 | if (current == connection->worker.task) |
337 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 338 | |
b30ab791 | 339 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 340 | setup_khelper_env(connection, envp); |
b411b363 | 341 | |
1090c056 LE |
342 | /* The helper may take some time. |
343 | * write out any unsynced meta data changes now */ | |
b30ab791 | 344 | drbd_md_sync(device); |
1090c056 | 345 | |
d0180171 | 346 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
347 | sib.sib_reason = SIB_HELPER_PRE; |
348 | sib.helper_name = cmd; | |
b30ab791 | 349 | drbd_bcast_event(device, &sib); |
70834d30 | 350 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 351 | if (ret) |
d0180171 | 352 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
353 | usermode_helper, cmd, mb, |
354 | (ret >> 8) & 0xff, ret); | |
355 | else | |
d0180171 | 356 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
357 | usermode_helper, cmd, mb, |
358 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
359 | sib.sib_reason = SIB_HELPER_POST; |
360 | sib.helper_exit_code = ret; | |
b30ab791 | 361 | drbd_bcast_event(device, &sib); |
b411b363 | 362 | |
bde89a9e AG |
363 | if (current == connection->worker.task) |
364 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
365 | |
366 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
367 | ret = 0; | |
368 | ||
369 | return ret; | |
370 | } | |
371 | ||
bde89a9e | 372 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
373 | { |
374 | char *envp[] = { "HOME=/", | |
375 | "TERM=linux", | |
376 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
377 | (char[20]) { }, /* address family */ | |
378 | (char[60]) { }, /* address */ | |
379 | NULL }; | |
77c556f6 AG |
380 | char *resource_name = connection->resource->name; |
381 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
382 | int ret; |
383 | ||
bde89a9e AG |
384 | setup_khelper_env(connection, envp); |
385 | conn_md_sync(connection); | |
6b75dced | 386 | |
1ec861eb | 387 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced PR |
388 | /* TODO: conn_bcast_event() ?? */ |
389 | ||
98683650 | 390 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 391 | if (ret) |
1ec861eb | 392 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 393 | usermode_helper, cmd, resource_name, |
6b75dced PR |
394 | (ret >> 8) & 0xff, ret); |
395 | else | |
1ec861eb | 396 | drbd_info(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 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 400 | |
b411b363 PR |
401 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
402 | ret = 0; | |
403 | ||
404 | return ret; | |
405 | } | |
406 | ||
bde89a9e | 407 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 408 | { |
cb703454 | 409 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 410 | struct drbd_peer_device *peer_device; |
cb703454 PR |
411 | int vnr; |
412 | ||
695d08fa | 413 | rcu_read_lock(); |
c06ece6b AG |
414 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
415 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 416 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
417 | struct disk_conf *disk_conf = |
418 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
419 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 420 | put_ldev(device); |
cb703454 PR |
421 | } |
422 | } | |
695d08fa | 423 | rcu_read_unlock(); |
cb703454 | 424 | |
d7fe69c6 PR |
425 | if (fp == FP_NOT_AVAIL) { |
426 | /* IO Suspending works on the whole resource. | |
427 | Do it only for one device. */ | |
428 | vnr = 0; | |
429 | peer_device = idr_get_next(&connection->peer_devices, &vnr); | |
430 | drbd_change_state(peer_device->device, CS_VERBOSE | CS_HARD, NS(susp_fen, 0)); | |
431 | } | |
432 | ||
cb703454 PR |
433 | return fp; |
434 | } | |
435 | ||
bde89a9e | 436 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 437 | { |
28e448bb | 438 | unsigned int connect_cnt; |
cb703454 PR |
439 | union drbd_state mask = { }; |
440 | union drbd_state val = { }; | |
441 | enum drbd_fencing_p fp; | |
b411b363 PR |
442 | char *ex_to_string; |
443 | int r; | |
b411b363 | 444 | |
02df6fe1 | 445 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 446 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 447 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
02df6fe1 | 448 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 PR |
449 | return false; |
450 | } | |
b411b363 | 451 | |
bde89a9e | 452 | connect_cnt = connection->connect_cnt; |
0500813f | 453 | spin_unlock_irq(&connection->resource->req_lock); |
28e448bb | 454 | |
bde89a9e | 455 | fp = highest_fencing_policy(connection); |
cb703454 PR |
456 | switch (fp) { |
457 | case FP_NOT_AVAIL: | |
1ec861eb | 458 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 459 | goto out; |
cb703454 PR |
460 | case FP_DONT_CARE: |
461 | return true; | |
462 | default: ; | |
b411b363 PR |
463 | } |
464 | ||
bde89a9e | 465 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
466 | |
467 | switch ((r>>8) & 0xff) { | |
468 | case 3: /* peer is inconsistent */ | |
469 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
470 | mask.pdsk = D_MASK; |
471 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
472 | break; |
473 | case 4: /* peer got outdated, or was already outdated */ | |
474 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
475 | mask.pdsk = D_MASK; |
476 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
477 | break; |
478 | case 5: /* peer was down */ | |
bde89a9e | 479 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
480 | /* we will(have) create(d) a new UUID anyways... */ |
481 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
482 | mask.pdsk = D_MASK; |
483 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
484 | } else { |
485 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
486 | } |
487 | break; | |
488 | case 6: /* Peer is primary, voluntarily outdate myself. | |
489 | * This is useful when an unconnected R_SECONDARY is asked to | |
490 | * become R_PRIMARY, but finds the other peer being active. */ | |
491 | ex_to_string = "peer is active"; | |
1ec861eb | 492 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
493 | mask.disk = D_MASK; |
494 | val.disk = D_OUTDATED; | |
b411b363 PR |
495 | break; |
496 | case 7: | |
497 | if (fp != FP_STONITH) | |
1ec861eb | 498 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 499 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
500 | mask.pdsk = D_MASK; |
501 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
502 | break; |
503 | default: | |
504 | /* The script is broken ... */ | |
1ec861eb | 505 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 506 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
507 | } |
508 | ||
1ec861eb | 509 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 510 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 511 | |
cb703454 | 512 | out: |
fb22c402 | 513 | |
cb703454 | 514 | /* Not using |
bde89a9e | 515 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
516 | here, because we might were able to re-establish the connection in the |
517 | meantime. */ | |
0500813f | 518 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
519 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
520 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
521 | /* In case the connection was established and droped |
522 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 523 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 524 | else |
bde89a9e | 525 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 526 | } |
0500813f | 527 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 | 528 | |
bde89a9e | 529 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
530 | } |
531 | ||
87f7be4c PR |
532 | static int _try_outdate_peer_async(void *data) |
533 | { | |
bde89a9e | 534 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 535 | |
bde89a9e | 536 | conn_try_outdate_peer(connection); |
87f7be4c | 537 | |
05a10ec7 | 538 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
539 | return 0; |
540 | } | |
541 | ||
bde89a9e | 542 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
543 | { |
544 | struct task_struct *opa; | |
545 | ||
bde89a9e AG |
546 | kref_get(&connection->kref); |
547 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); | |
9dc9fbb3 | 548 | if (IS_ERR(opa)) { |
1ec861eb | 549 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 550 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 551 | } |
87f7be4c | 552 | } |
b411b363 | 553 | |
bf885f8a | 554 | enum drbd_state_rv |
b30ab791 | 555 | drbd_set_role(struct drbd_device *device, enum drbd_role new_role, int force) |
b411b363 PR |
556 | { |
557 | const int max_tries = 4; | |
bf885f8a | 558 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 559 | struct net_conf *nc; |
b411b363 PR |
560 | int try = 0; |
561 | int forced = 0; | |
562 | union drbd_state mask, val; | |
b411b363 | 563 | |
b6f85ef9 AG |
564 | if (new_role == R_PRIMARY) { |
565 | struct drbd_connection *connection; | |
566 | ||
567 | /* Detect dead peers as soon as possible. */ | |
568 | ||
569 | rcu_read_lock(); | |
570 | for_each_connection(connection, device->resource) | |
571 | request_ping(connection); | |
572 | rcu_read_unlock(); | |
573 | } | |
b411b363 | 574 | |
b30ab791 | 575 | mutex_lock(device->state_mutex); |
b411b363 PR |
576 | |
577 | mask.i = 0; mask.role = R_MASK; | |
578 | val.i = 0; val.role = new_role; | |
579 | ||
580 | while (try++ < max_tries) { | |
b30ab791 | 581 | rv = _drbd_request_state(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
582 | |
583 | /* in case we first succeeded to outdate, | |
584 | * but now suddenly could establish a connection */ | |
bf885f8a | 585 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
586 | val.pdsk = 0; |
587 | mask.pdsk = 0; | |
588 | continue; | |
589 | } | |
590 | ||
bf885f8a | 591 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
592 | (device->state.disk < D_UP_TO_DATE && |
593 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
594 | mask.disk = D_MASK; |
595 | val.disk = D_UP_TO_DATE; | |
596 | forced = 1; | |
597 | continue; | |
598 | } | |
599 | ||
bf885f8a | 600 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 601 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 602 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 603 | |
a6b32bc3 | 604 | if (conn_try_outdate_peer(first_peer_device(device)->connection)) { |
b411b363 PR |
605 | val.disk = D_UP_TO_DATE; |
606 | mask.disk = D_MASK; | |
607 | } | |
b411b363 PR |
608 | continue; |
609 | } | |
610 | ||
bf885f8a | 611 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 612 | goto out; |
bf885f8a | 613 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
a6b32bc3 | 614 | if (!conn_try_outdate_peer(first_peer_device(device)->connection) && force) { |
d0180171 | 615 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
616 | mask.pdsk = D_MASK; |
617 | val.pdsk = D_OUTDATED; | |
b411b363 | 618 | |
cb703454 | 619 | } |
b411b363 PR |
620 | continue; |
621 | } | |
bf885f8a | 622 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
623 | /* Maybe the peer is detected as dead very soon... |
624 | retry at most once more in this case. */ | |
44ed167d PR |
625 | int timeo; |
626 | rcu_read_lock(); | |
a6b32bc3 | 627 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
628 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
629 | rcu_read_unlock(); | |
630 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
631 | if (try < max_tries) |
632 | try = max_tries - 1; | |
633 | continue; | |
634 | } | |
bf885f8a | 635 | if (rv < SS_SUCCESS) { |
b30ab791 | 636 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 637 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 638 | if (rv < SS_SUCCESS) |
3b98c0c2 | 639 | goto out; |
b411b363 PR |
640 | } |
641 | break; | |
642 | } | |
643 | ||
bf885f8a | 644 | if (rv < SS_SUCCESS) |
3b98c0c2 | 645 | goto out; |
b411b363 PR |
646 | |
647 | if (forced) | |
d0180171 | 648 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
649 | |
650 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 651 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 652 | |
b6dd1a89 LE |
653 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
654 | ||
b411b363 | 655 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
656 | set_disk_ro(device->vdisk, true); |
657 | if (get_ldev(device)) { | |
658 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
659 | put_ldev(device); | |
b411b363 PR |
660 | } |
661 | } else { | |
9e276872 LE |
662 | /* Called from drbd_adm_set_role only. |
663 | * We are still holding the conf_update mutex. */ | |
a6b32bc3 | 664 | nc = first_peer_device(device)->connection->net_conf; |
44ed167d | 665 | if (nc) |
6139f60d | 666 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
91fd4dad | 667 | |
b30ab791 AG |
668 | set_disk_ro(device->vdisk, false); |
669 | if (get_ldev(device)) { | |
670 | if (((device->state.conn < C_CONNECTED || | |
671 | device->state.pdsk <= D_FAILED) | |
672 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
673 | drbd_uuid_new_current(device); | |
b411b363 | 674 | |
b30ab791 AG |
675 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
676 | put_ldev(device); | |
b411b363 PR |
677 | } |
678 | } | |
679 | ||
19f843aa LE |
680 | /* writeout of activity log covered areas of the bitmap |
681 | * to stable storage done in after state change already */ | |
b411b363 | 682 | |
b30ab791 | 683 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
684 | /* if this was forced, we should consider sync */ |
685 | if (forced) | |
69a22773 AG |
686 | drbd_send_uuids(first_peer_device(device)); |
687 | drbd_send_current_state(first_peer_device(device)); | |
b411b363 PR |
688 | } |
689 | ||
b30ab791 | 690 | drbd_md_sync(device); |
b411b363 | 691 | |
b30ab791 | 692 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 693 | out: |
b30ab791 | 694 | mutex_unlock(device->state_mutex); |
bf885f8a | 695 | return rv; |
b411b363 PR |
696 | } |
697 | ||
3b98c0c2 | 698 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 699 | { |
3b98c0c2 LE |
700 | return err == -ENOMSG ? "required attribute missing" : |
701 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 702 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 703 | "invalid attribute value"; |
ef50a3e3 | 704 | } |
b411b363 | 705 | |
3b98c0c2 | 706 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 707 | { |
a910b123 | 708 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
709 | struct set_role_parms parms; |
710 | int err; | |
711 | enum drbd_ret_code retcode; | |
b411b363 | 712 | |
a910b123 | 713 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
714 | if (!adm_ctx.reply_skb) |
715 | return retcode; | |
716 | if (retcode != NO_ERROR) | |
717 | goto out; | |
b411b363 | 718 | |
3b98c0c2 LE |
719 | memset(&parms, 0, sizeof(parms)); |
720 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 721 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
722 | if (err) { |
723 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 724 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
725 | goto out; |
726 | } | |
727 | } | |
9e276872 LE |
728 | genl_unlock(); |
729 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 730 | |
3b98c0c2 | 731 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 732 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 733 | else |
b30ab791 | 734 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
735 | |
736 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
737 | genl_lock(); | |
3b98c0c2 | 738 | out: |
a910b123 | 739 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
740 | return 0; |
741 | } | |
742 | ||
ae8bf312 LE |
743 | /* Initializes the md.*_offset members, so we are able to find |
744 | * the on disk meta data. | |
745 | * | |
746 | * We currently have two possible layouts: | |
747 | * external: | |
748 | * |----------- md_size_sect ------------------| | |
749 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
750 | * | al_offset == 8 | | |
751 | * | bm_offset = al_offset + X | | |
752 | * ==> bitmap sectors = md_size_sect - bm_offset | |
753 | * | |
754 | * internal: | |
755 | * |----------- md_size_sect ------------------| | |
756 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
757 | * | al_offset < 0 | | |
758 | * | bm_offset = al_offset - Y | | |
759 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
760 | * | |
761 | * Activity log size used to be fixed 32kB, | |
762 | * but is about to become configurable. | |
763 | */ | |
b30ab791 | 764 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
765 | struct drbd_backing_dev *bdev) |
766 | { | |
767 | sector_t md_size_sect = 0; | |
c04ccaa6 | 768 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 769 | |
3a4d4eb3 LE |
770 | bdev->md.md_offset = drbd_md_ss(bdev); |
771 | ||
68e41a43 | 772 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
773 | default: |
774 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 775 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
776 | bdev->md.al_offset = MD_4kB_SECT; |
777 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
778 | break; |
779 | case DRBD_MD_INDEX_FLEX_EXT: | |
780 | /* just occupy the full device; unit: sectors */ | |
781 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
782 | bdev->md.al_offset = MD_4kB_SECT; |
783 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
784 | break; |
785 | case DRBD_MD_INDEX_INTERNAL: | |
786 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 787 | /* al size is still fixed */ |
ae8bf312 | 788 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
789 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
790 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
791 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
792 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
793 | md_size_sect = ALIGN(md_size_sect, 8); | |
794 | ||
795 | /* plus the "drbd meta data super block", | |
796 | * and the activity log; */ | |
ae8bf312 | 797 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
798 | |
799 | bdev->md.md_size_sect = md_size_sect; | |
800 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 801 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
802 | break; |
803 | } | |
804 | } | |
805 | ||
4b0715f0 | 806 | /* input size is expected to be in KB */ |
b411b363 PR |
807 | char *ppsize(char *buf, unsigned long long size) |
808 | { | |
4b0715f0 LE |
809 | /* Needs 9 bytes at max including trailing NUL: |
810 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
811 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
812 | int base = 0; | |
4b0715f0 | 813 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
814 | /* shift + round */ |
815 | size = (size >> 10) + !!(size & (1<<9)); | |
816 | base++; | |
817 | } | |
4b0715f0 | 818 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
819 | |
820 | return buf; | |
821 | } | |
822 | ||
823 | /* there is still a theoretical deadlock when called from receiver | |
824 | * on an D_INCONSISTENT R_PRIMARY: | |
825 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
826 | * packet from remote to dec_ap_bio again. | |
827 | * receiver receive_sizes(), comes here, | |
828 | * waits for ap_bio_cnt == 0. -> deadlock. | |
829 | * but this cannot happen, actually, because: | |
830 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
831 | * (not connected, or bad/no disk on peer): | |
832 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
833 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
834 | * peer may not initiate a resize. | |
835 | */ | |
3b98c0c2 LE |
836 | /* Note these are not to be confused with |
837 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
838 | * which are (sub) state changes triggered by admin (drbdsetup), | |
839 | * and can be long lived. | |
b30ab791 | 840 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 841 | * and should be short-lived. */ |
b30ab791 | 842 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 843 | { |
b30ab791 AG |
844 | set_bit(SUSPEND_IO, &device->flags); |
845 | if (drbd_suspended(device)) | |
265be2d0 | 846 | return; |
b30ab791 | 847 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
848 | } |
849 | ||
b30ab791 | 850 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 851 | { |
b30ab791 AG |
852 | clear_bit(SUSPEND_IO, &device->flags); |
853 | wake_up(&device->misc_wait); | |
b411b363 PR |
854 | } |
855 | ||
856 | /** | |
857 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 858 | * @device: DRBD device. |
b411b363 PR |
859 | * |
860 | * Returns 0 on success, negative return values indicate errors. | |
861 | * You should call drbd_md_sync() after calling this function. | |
862 | */ | |
d752b269 | 863 | enum determine_dev_size |
b30ab791 | 864 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
865 | { |
866 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 867 | sector_t la_size_sect, u_size; |
b30ab791 | 868 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
869 | u32 prev_al_stripe_size_4k; |
870 | u32 prev_al_stripes; | |
b411b363 PR |
871 | sector_t size; |
872 | char ppb[10]; | |
d752b269 | 873 | void *buffer; |
b411b363 PR |
874 | |
875 | int md_moved, la_size_changed; | |
e96c9633 | 876 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
877 | |
878 | /* race: | |
879 | * application request passes inc_ap_bio, | |
880 | * but then cannot get an AL-reference. | |
881 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
882 | * | |
883 | * to avoid that: | |
884 | * Suspend IO right here. | |
885 | * still lock the act_log to not trigger ASSERTs there. | |
886 | */ | |
b30ab791 AG |
887 | drbd_suspend_io(device); |
888 | buffer = drbd_md_get_buffer(device); /* Lock meta-data IO */ | |
d752b269 | 889 | if (!buffer) { |
b30ab791 | 890 | drbd_resume_io(device); |
d752b269 PR |
891 | return DS_ERROR; |
892 | } | |
b411b363 PR |
893 | |
894 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 895 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 896 | |
b30ab791 AG |
897 | prev_first_sect = drbd_md_first_sector(device->ldev); |
898 | prev_size = device->ldev->md.md_size_sect; | |
899 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 900 | |
d752b269 PR |
901 | if (rs) { |
902 | /* rs is non NULL if we should change the AL layout only */ | |
903 | ||
904 | prev_al_stripes = md->al_stripes; | |
905 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
906 | ||
907 | md->al_stripes = rs->al_stripes; | |
908 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
909 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
910 | } | |
911 | ||
b30ab791 | 912 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 913 | |
daeda1cc | 914 | rcu_read_lock(); |
b30ab791 | 915 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 916 | rcu_read_unlock(); |
b30ab791 | 917 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 918 | |
d752b269 PR |
919 | if (size < la_size_sect) { |
920 | if (rs && u_size == 0) { | |
921 | /* Remove "rs &&" later. This check should always be active, but | |
922 | right now the receiver expects the permissive behavior */ | |
d0180171 | 923 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
924 | "Use --size=%llus for explicit shrink.\n", |
925 | (unsigned long long)size); | |
926 | rv = DS_ERROR_SHRINK; | |
927 | } | |
928 | if (u_size > size) | |
929 | rv = DS_ERROR_SPACE_MD; | |
930 | if (rv != DS_UNCHANGED) | |
931 | goto err_out; | |
932 | } | |
933 | ||
b30ab791 AG |
934 | if (drbd_get_capacity(device->this_bdev) != size || |
935 | drbd_bm_capacity(device) != size) { | |
b411b363 | 936 | int err; |
b30ab791 | 937 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
938 | if (unlikely(err)) { |
939 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 940 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 941 | if (size == 0) { |
d0180171 | 942 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
943 | "Could not allocate bitmap!\n"); |
944 | } else { | |
d0180171 | 945 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
946 | "Leaving size unchanged at size = %lu KB\n", |
947 | (unsigned long)size); | |
948 | } | |
e96c9633 | 949 | rv = DS_ERROR; |
b411b363 PR |
950 | } |
951 | /* racy, see comments above. */ | |
b30ab791 AG |
952 | drbd_set_my_capacity(device, size); |
953 | device->ldev->md.la_size_sect = size; | |
d0180171 | 954 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
955 | (unsigned long long)size>>1); |
956 | } | |
d752b269 PR |
957 | if (rv <= DS_ERROR) |
958 | goto err_out; | |
b411b363 | 959 | |
b30ab791 | 960 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 961 | |
b30ab791 AG |
962 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
963 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 964 | |
d752b269 PR |
965 | if (la_size_changed || md_moved || rs) { |
966 | u32 prev_flags; | |
24dccabb | 967 | |
b30ab791 | 968 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
969 | |
970 | prev_flags = md->flags; | |
971 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 972 | drbd_md_write(device, buffer); |
d752b269 | 973 | |
d0180171 | 974 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
975 | la_size_changed && md_moved ? "size changed and md moved" : |
976 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 977 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 978 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 979 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 980 | drbd_initialize_al(device, buffer); |
d752b269 PR |
981 | |
982 | md->flags = prev_flags; | |
b30ab791 | 983 | drbd_md_write(device, buffer); |
d752b269 PR |
984 | |
985 | if (rs) | |
d0180171 AG |
986 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
987 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
988 | } |
989 | ||
cccac985 | 990 | if (size > la_size_sect) |
57737adc | 991 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 992 | if (size < la_size_sect) |
e96c9633 | 993 | rv = DS_SHRUNK; |
d752b269 PR |
994 | |
995 | if (0) { | |
996 | err_out: | |
997 | if (rs) { | |
998 | md->al_stripes = prev_al_stripes; | |
999 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
1000 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
1001 | ||
b30ab791 | 1002 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
1003 | } |
1004 | } | |
b30ab791 AG |
1005 | lc_unlock(device->act_log); |
1006 | wake_up(&device->al_wait); | |
1007 | drbd_md_put_buffer(device); | |
1008 | drbd_resume_io(device); | |
b411b363 PR |
1009 | |
1010 | return rv; | |
1011 | } | |
1012 | ||
1013 | sector_t | |
b30ab791 | 1014 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1015 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1016 | { |
b30ab791 | 1017 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1018 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1019 | sector_t m_size; /* my size */ |
b411b363 PR |
1020 | sector_t size = 0; |
1021 | ||
1022 | m_size = drbd_get_max_capacity(bdev); | |
1023 | ||
b30ab791 | 1024 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1025 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1026 | p_size = m_size; |
1027 | } | |
1028 | ||
b411b363 PR |
1029 | if (p_size && m_size) { |
1030 | size = min_t(sector_t, p_size, m_size); | |
1031 | } else { | |
cccac985 LE |
1032 | if (la_size_sect) { |
1033 | size = la_size_sect; | |
b411b363 PR |
1034 | if (m_size && m_size < size) |
1035 | size = m_size; | |
1036 | if (p_size && p_size < size) | |
1037 | size = p_size; | |
1038 | } else { | |
1039 | if (m_size) | |
1040 | size = m_size; | |
1041 | if (p_size) | |
1042 | size = p_size; | |
1043 | } | |
1044 | } | |
1045 | ||
1046 | if (size == 0) | |
d0180171 | 1047 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1048 | |
1049 | if (u_size) { | |
1050 | if (u_size > size) | |
d0180171 | 1051 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1052 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1053 | else | |
1054 | size = u_size; | |
1055 | } | |
1056 | ||
1057 | return size; | |
1058 | } | |
1059 | ||
1060 | /** | |
1061 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1062 | * @device: DRBD device. |
b411b363 PR |
1063 | * |
1064 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1065 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1066 | * this function. | |
1067 | */ | |
b30ab791 | 1068 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1069 | { |
1070 | struct lru_cache *n, *t; | |
1071 | struct lc_element *e; | |
1072 | unsigned int in_use; | |
1073 | int i; | |
1074 | ||
b30ab791 AG |
1075 | if (device->act_log && |
1076 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1077 | return 0; |
1078 | ||
1079 | in_use = 0; | |
b30ab791 | 1080 | t = device->act_log; |
7ad651b5 | 1081 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1082 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1083 | |
1084 | if (n == NULL) { | |
d0180171 | 1085 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1086 | return -ENOMEM; |
1087 | } | |
b30ab791 | 1088 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1089 | if (t) { |
1090 | for (i = 0; i < t->nr_elements; i++) { | |
1091 | e = lc_element_by_index(t, i); | |
1092 | if (e->refcnt) | |
d0180171 | 1093 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1094 | e->lc_number, e->refcnt); |
1095 | in_use += e->refcnt; | |
1096 | } | |
1097 | } | |
1098 | if (!in_use) | |
b30ab791 AG |
1099 | device->act_log = n; |
1100 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1101 | if (in_use) { |
d0180171 | 1102 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1103 | lc_destroy(n); |
1104 | return -EBUSY; | |
1105 | } else { | |
1106 | if (t) | |
1107 | lc_destroy(t); | |
1108 | } | |
b30ab791 | 1109 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1110 | return 0; |
1111 | } | |
1112 | ||
b30ab791 | 1113 | static void drbd_setup_queue_param(struct drbd_device *device, unsigned int max_bio_size) |
b411b363 | 1114 | { |
b30ab791 | 1115 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1116 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1117 | unsigned int max_segments = 0; | |
c1b3156f | 1118 | struct request_queue *b = NULL; |
99432fcc | 1119 | |
b30ab791 | 1120 | if (get_ldev_if_state(device, D_ATTACHING)) { |
c1b3156f | 1121 | b = device->ldev->backing_bdev->bd_disk->queue; |
99432fcc PR |
1122 | |
1123 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1124 | rcu_read_lock(); |
b30ab791 | 1125 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1126 | rcu_read_unlock(); |
c1b3156f PR |
1127 | |
1128 | blk_set_stacking_limits(&q->limits); | |
1129 | blk_queue_max_write_same_sectors(q, 0); | |
99432fcc | 1130 | } |
b411b363 | 1131 | |
b411b363 | 1132 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1133 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1134 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1135 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1136 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1137 | |
c1b3156f | 1138 | if (b) { |
20c68fde LE |
1139 | struct drbd_connection *connection = first_peer_device(device)->connection; |
1140 | ||
1141 | if (blk_queue_discard(b) && | |
1142 | (connection->cstate < C_CONNECTED || connection->agreed_features & FF_TRIM)) { | |
20c68fde LE |
1143 | /* For now, don't allow more than one activity log extent worth of data |
1144 | * to be discarded in one go. We may need to rework drbd_al_begin_io() | |
1145 | * to allow for even larger discard ranges */ | |
1146 | q->limits.max_discard_sectors = DRBD_MAX_DISCARD_SECTORS; | |
1147 | ||
1148 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1149 | /* REALLY? Is stacking secdiscard "legal"? */ | |
1150 | if (blk_queue_secdiscard(b)) | |
1151 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1152 | } else { | |
1153 | q->limits.max_discard_sectors = 0; | |
1154 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1155 | queue_flag_clear_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1156 | } | |
99432fcc PR |
1157 | |
1158 | blk_queue_stack_limits(q, b); | |
1159 | ||
1160 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1161 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1162 | q->backing_dev_info.ra_pages, |
1163 | b->backing_dev_info.ra_pages); | |
1164 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1165 | } | |
b30ab791 | 1166 | put_ldev(device); |
99432fcc PR |
1167 | } |
1168 | } | |
1169 | ||
b30ab791 | 1170 | void drbd_reconsider_max_bio_size(struct drbd_device *device) |
99432fcc | 1171 | { |
db141b2f | 1172 | unsigned int now, new, local, peer; |
99432fcc | 1173 | |
b30ab791 AG |
1174 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1175 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1176 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1177 | |
b30ab791 AG |
1178 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1179 | local = queue_max_hw_sectors(device->ldev->backing_bdev->bd_disk->queue) << 9; | |
1180 | device->local_max_bio_size = local; | |
1181 | put_ldev(device); | |
b411b363 | 1182 | } |
db141b2f | 1183 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1184 | |
1185 | /* We may ignore peer limits if the peer is modern enough. | |
1186 | Because new from 8.3.8 onwards the peer can use multiple | |
1187 | BIOs for a single peer_request */ | |
b30ab791 | 1188 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1189 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1190 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1191 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1192 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1193 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1194 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1195 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1196 | else | |
99432fcc | 1197 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1198 | |
fa090e70 LE |
1199 | /* We may later detach and re-attach on a disconnected Primary. |
1200 | * Avoid this setting to jump back in that case. | |
1201 | * We want to store what we know the peer DRBD can handle, | |
1202 | * not what the peer IO backend can handle. */ | |
1203 | if (peer > device->peer_max_bio_size) | |
1204 | device->peer_max_bio_size = peer; | |
1205 | } | |
db141b2f | 1206 | new = min(local, peer); |
99432fcc | 1207 | |
b30ab791 | 1208 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1209 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1210 | |
1211 | if (new != now) | |
d0180171 | 1212 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1213 | |
b30ab791 | 1214 | drbd_setup_queue_param(device, new); |
b411b363 PR |
1215 | } |
1216 | ||
a18e9d1e | 1217 | /* Starts the worker thread */ |
bde89a9e | 1218 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1219 | { |
bde89a9e | 1220 | drbd_thread_start(&connection->worker); |
b5043c5e | 1221 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1222 | } |
1223 | ||
a18e9d1e | 1224 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1225 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1226 | { |
992d6e91 | 1227 | bool stop_threads; |
0500813f | 1228 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1229 | stop_threads = conn_all_vols_unconf(connection) && |
1230 | connection->cstate == C_STANDALONE; | |
0500813f | 1231 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1232 | if (stop_threads) { |
1233 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1234 | * in conn_disconnect() */ |
bde89a9e AG |
1235 | drbd_thread_stop(&connection->receiver); |
1236 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1237 | } |
b411b363 PR |
1238 | } |
1239 | ||
0778286a | 1240 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1241 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1242 | { |
1243 | int s = 0; | |
1244 | ||
b30ab791 | 1245 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1246 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1247 | return; |
1248 | } | |
1249 | ||
b30ab791 | 1250 | drbd_al_shrink(device); |
0500813f | 1251 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1252 | if (device->state.conn < C_CONNECTED) |
1253 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1254 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1255 | lc_unlock(device->act_log); |
0778286a PR |
1256 | |
1257 | if (s) | |
d0180171 | 1258 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1259 | } |
1260 | ||
5979e361 LE |
1261 | |
1262 | static bool should_set_defaults(struct genl_info *info) | |
1263 | { | |
1264 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1265 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1266 | } | |
1267 | ||
5bbcf5e6 | 1268 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1269 | { |
5bbcf5e6 LE |
1270 | /* This is limited by 16 bit "slot" numbers, |
1271 | * and by available on-disk context storage. | |
1272 | * | |
1273 | * Also (u16)~0 is special (denotes a "free" extent). | |
1274 | * | |
1275 | * One transaction occupies one 4kB on-disk block, | |
1276 | * we have n such blocks in the on disk ring buffer, | |
1277 | * the "current" transaction may fail (n-1), | |
1278 | * and there is 919 slot numbers context information per transaction. | |
1279 | * | |
1280 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1281 | * so cap there first. | |
1282 | */ | |
1283 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1284 | const unsigned int sufficient_on_disk = | |
1285 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1286 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1287 | |
5bbcf5e6 LE |
1288 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1289 | ||
1290 | if (al_size_4k > sufficient_on_disk) | |
1291 | return max_al_nr; | |
1292 | ||
1293 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1294 | } |
1295 | ||
f399002e LE |
1296 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1297 | { | |
a910b123 | 1298 | struct drbd_config_context adm_ctx; |
f399002e | 1299 | enum drbd_ret_code retcode; |
b30ab791 | 1300 | struct drbd_device *device; |
daeda1cc | 1301 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1302 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1303 | int err, fifo_size; |
f399002e | 1304 | |
a910b123 | 1305 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1306 | if (!adm_ctx.reply_skb) |
1307 | return retcode; | |
1308 | if (retcode != NO_ERROR) | |
9e276872 | 1309 | goto finish; |
f399002e | 1310 | |
b30ab791 | 1311 | device = adm_ctx.device; |
9e276872 | 1312 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1313 | |
1314 | /* we also need a disk | |
1315 | * to change the options on */ | |
b30ab791 | 1316 | if (!get_ldev(device)) { |
f399002e LE |
1317 | retcode = ERR_NO_DISK; |
1318 | goto out; | |
1319 | } | |
1320 | ||
daeda1cc | 1321 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1322 | if (!new_disk_conf) { |
f399002e LE |
1323 | retcode = ERR_NOMEM; |
1324 | goto fail; | |
1325 | } | |
1326 | ||
0500813f | 1327 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1328 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1329 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1330 | if (should_set_defaults(info)) |
b966b5dd | 1331 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1332 | |
5ecc72c3 | 1333 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1334 | if (err && err != -ENOMSG) { |
f399002e | 1335 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1336 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1337 | goto fail_unlock; |
f399002e LE |
1338 | } |
1339 | ||
5ecc72c3 LE |
1340 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1341 | new_disk_conf->resync_rate = 1; | |
f399002e | 1342 | |
5bbcf5e6 LE |
1343 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1344 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1345 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1346 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1347 | |
1348 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1349 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1350 | |
5ecc72c3 | 1351 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1352 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1353 | new_plan = fifo_alloc(fifo_size); |
1354 | if (!new_plan) { | |
d0180171 | 1355 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1356 | retcode = ERR_NOMEM; |
daeda1cc | 1357 | goto fail_unlock; |
f399002e LE |
1358 | } |
1359 | } | |
1360 | ||
b30ab791 AG |
1361 | drbd_suspend_io(device); |
1362 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1363 | drbd_al_shrink(device); | |
1364 | err = drbd_check_al_size(device, new_disk_conf); | |
1365 | lc_unlock(device->act_log); | |
1366 | wake_up(&device->al_wait); | |
1367 | drbd_resume_io(device); | |
f399002e LE |
1368 | |
1369 | if (err) { | |
1370 | retcode = ERR_NOMEM; | |
daeda1cc | 1371 | goto fail_unlock; |
f399002e LE |
1372 | } |
1373 | ||
dc97b708 | 1374 | write_lock_irq(&global_state_lock); |
b30ab791 | 1375 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1376 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1377 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1378 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1379 | } |
1380 | write_unlock_irq(&global_state_lock); | |
f399002e | 1381 | |
daeda1cc PR |
1382 | if (retcode != NO_ERROR) |
1383 | goto fail_unlock; | |
f399002e | 1384 | |
813472ce | 1385 | if (new_plan) { |
b30ab791 AG |
1386 | old_plan = device->rs_plan_s; |
1387 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1388 | } |
9958c857 | 1389 | |
0500813f | 1390 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1391 | |
9a51ab1c | 1392 | if (new_disk_conf->al_updates) |
b30ab791 | 1393 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1394 | else |
b30ab791 | 1395 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1396 | |
691631c0 | 1397 | if (new_disk_conf->md_flushes) |
b30ab791 | 1398 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1399 | else |
b30ab791 | 1400 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1401 | |
a6b32bc3 | 1402 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
27eb13e9 | 1403 | |
b30ab791 | 1404 | drbd_md_sync(device); |
f399002e | 1405 | |
69a22773 AG |
1406 | if (device->state.conn >= C_CONNECTED) { |
1407 | struct drbd_peer_device *peer_device; | |
1408 | ||
1409 | for_each_peer_device(peer_device, device) | |
1410 | drbd_send_sync_param(peer_device); | |
1411 | } | |
f399002e | 1412 | |
daeda1cc PR |
1413 | synchronize_rcu(); |
1414 | kfree(old_disk_conf); | |
813472ce | 1415 | kfree(old_plan); |
b30ab791 | 1416 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1417 | goto success; |
1418 | ||
1419 | fail_unlock: | |
0500813f | 1420 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1421 | fail: |
5ecc72c3 | 1422 | kfree(new_disk_conf); |
813472ce | 1423 | kfree(new_plan); |
daeda1cc | 1424 | success: |
b30ab791 | 1425 | put_ldev(device); |
f399002e | 1426 | out: |
9e276872 LE |
1427 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1428 | finish: | |
a910b123 | 1429 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1430 | return 0; |
1431 | } | |
1432 | ||
3b98c0c2 | 1433 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1434 | { |
a910b123 | 1435 | struct drbd_config_context adm_ctx; |
b30ab791 | 1436 | struct drbd_device *device; |
3b98c0c2 | 1437 | int err; |
116676ca | 1438 | enum drbd_ret_code retcode; |
b411b363 PR |
1439 | enum determine_dev_size dd; |
1440 | sector_t max_possible_sectors; | |
1441 | sector_t min_md_device_sectors; | |
1442 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1443 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1444 | struct block_device *bdev; |
b411b363 | 1445 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1446 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1447 | union drbd_state ns, os; |
f2024e7c | 1448 | enum drbd_state_rv rv; |
44ed167d | 1449 | struct net_conf *nc; |
b411b363 | 1450 | |
a910b123 | 1451 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1452 | if (!adm_ctx.reply_skb) |
1453 | return retcode; | |
1454 | if (retcode != NO_ERROR) | |
40cbf085 | 1455 | goto finish; |
b411b363 | 1456 | |
b30ab791 | 1457 | device = adm_ctx.device; |
9e276872 | 1458 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a6b32bc3 | 1459 | conn_reconfig_start(first_peer_device(device)->connection); |
b411b363 PR |
1460 | |
1461 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1462 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1463 | retcode = ERR_DISK_CONFIGURED; |
1464 | goto fail; | |
1465 | } | |
82f59cc6 LE |
1466 | /* It may just now have detached because of IO error. Make sure |
1467 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1468 | * e.g. if someone calls attach from the on-io-error handler, | |
1469 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
b30ab791 | 1470 | wait_event(device->misc_wait, !atomic_read(&device->local_cnt)); |
b411b363 | 1471 | |
383606e0 | 1472 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1473 | clear_bit(FORCE_DETACH, &device->flags); |
1474 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1475 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1476 | |
0029d624 | 1477 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1478 | device->rs_total = 0; |
1479 | device->rs_failed = 0; | |
1480 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1481 | |
3b98c0c2 | 1482 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1483 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1484 | if (!nbc) { | |
1485 | retcode = ERR_NOMEM; | |
1486 | goto fail; | |
1487 | } | |
9f2247bb PR |
1488 | spin_lock_init(&nbc->md.uuid_lock); |
1489 | ||
daeda1cc PR |
1490 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1491 | if (!new_disk_conf) { | |
1492 | retcode = ERR_NOMEM; | |
b411b363 PR |
1493 | goto fail; |
1494 | } | |
daeda1cc | 1495 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1496 | |
daeda1cc PR |
1497 | set_disk_conf_defaults(new_disk_conf); |
1498 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1499 | if (err) { |
b411b363 | 1500 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1501 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1502 | goto fail; |
1503 | } | |
1504 | ||
5bbcf5e6 LE |
1505 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1506 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1507 | |
9958c857 PR |
1508 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1509 | if (!new_plan) { | |
1510 | retcode = ERR_NOMEM; | |
1511 | goto fail; | |
1512 | } | |
1513 | ||
daeda1cc | 1514 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1515 | retcode = ERR_MD_IDX_INVALID; |
1516 | goto fail; | |
1517 | } | |
1518 | ||
a3f8f7dc | 1519 | write_lock_irq(&global_state_lock); |
b30ab791 | 1520 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1521 | write_unlock_irq(&global_state_lock); |
1522 | if (retcode != NO_ERROR) | |
1523 | goto fail; | |
1524 | ||
44ed167d | 1525 | rcu_read_lock(); |
a6b32bc3 | 1526 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d | 1527 | if (nc) { |
daeda1cc | 1528 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1529 | rcu_read_unlock(); |
47ff2d0a PR |
1530 | retcode = ERR_STONITH_AND_PROT_A; |
1531 | goto fail; | |
1532 | } | |
1533 | } | |
44ed167d | 1534 | rcu_read_unlock(); |
47ff2d0a | 1535 | |
daeda1cc | 1536 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1537 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1538 | if (IS_ERR(bdev)) { |
d0180171 | 1539 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1540 | PTR_ERR(bdev)); |
b411b363 PR |
1541 | retcode = ERR_OPEN_DISK; |
1542 | goto fail; | |
1543 | } | |
e525fd89 TH |
1544 | nbc->backing_bdev = bdev; |
1545 | ||
1546 | /* | |
1547 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1548 | * drbd sharing one meta device. TODO in that case, paranoia | |
1549 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1550 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1551 | * should check it for you already; but if you don't, or | |
1552 | * someone fooled it, we need to double check here) | |
1553 | */ | |
daeda1cc | 1554 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1555 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1556 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1557 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1558 | if (IS_ERR(bdev)) { |
d0180171 | 1559 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1560 | PTR_ERR(bdev)); |
b411b363 PR |
1561 | retcode = ERR_OPEN_MD_DISK; |
1562 | goto fail; | |
1563 | } | |
e525fd89 | 1564 | nbc->md_bdev = bdev; |
b411b363 | 1565 | |
e525fd89 | 1566 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1567 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1568 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1569 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1570 | goto fail; |
1571 | } | |
1572 | ||
1573 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1574 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1575 | offsetof(struct bm_extent, lce)); |
1576 | if (!resync_lru) { | |
1577 | retcode = ERR_NOMEM; | |
e525fd89 | 1578 | goto fail; |
b411b363 PR |
1579 | } |
1580 | ||
c04ccaa6 LE |
1581 | /* Read our meta data super block early. |
1582 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1583 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1584 | if (retcode != NO_ERROR) |
1585 | goto fail; | |
b411b363 | 1586 | |
5bbcf5e6 LE |
1587 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1588 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1589 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1590 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1591 | ||
daeda1cc | 1592 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1593 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1594 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1595 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1596 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1597 | goto fail; |
b411b363 PR |
1598 | } |
1599 | ||
daeda1cc | 1600 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1601 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1602 | /* at least one MB, otherwise it does not make sense */ | |
1603 | min_md_device_sectors = (2<<10); | |
1604 | } else { | |
1605 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1606 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1607 | } |
1608 | ||
b411b363 | 1609 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1610 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1611 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1612 | "at least %llu sectors needed for this meta-disk type\n", |
1613 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1614 | goto fail; |
b411b363 PR |
1615 | } |
1616 | ||
1617 | /* Make sure the new disk is big enough | |
1618 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1619 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1620 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1621 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1622 | goto fail; |
b411b363 PR |
1623 | } |
1624 | ||
1625 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1626 | ||
1352994b | 1627 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1628 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1629 | "to currently maximum possible %llu sectors <==\n", |
1630 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1631 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1632 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1633 | "meta data may help <<==\n"); |
1634 | } | |
1635 | ||
b30ab791 | 1636 | drbd_suspend_io(device); |
b411b363 | 1637 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1638 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1639 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1640 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1641 | * As barriers are counted per resource, | |
1642 | * we'd need to suspend io on all devices of a resource. | |
1643 | */ | |
b30ab791 | 1644 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1645 | /* and for any other previously queued work */ |
b5043c5e | 1646 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
b411b363 | 1647 | |
b30ab791 | 1648 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1649 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1650 | drbd_resume_io(device); |
f2024e7c | 1651 | if (rv < SS_SUCCESS) |
e525fd89 | 1652 | goto fail; |
b411b363 | 1653 | |
b30ab791 | 1654 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1655 | goto force_diskless; |
1656 | ||
b30ab791 AG |
1657 | if (!device->bitmap) { |
1658 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1659 | retcode = ERR_NOMEM; |
1660 | goto force_diskless_dec; | |
1661 | } | |
1662 | } | |
1663 | ||
b30ab791 | 1664 | if (device->state.conn < C_CONNECTED && |
babea49e | 1665 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1666 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1667 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1668 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1669 | retcode = ERR_DATA_NOT_CURRENT; |
1670 | goto force_diskless_dec; | |
1671 | } | |
1672 | ||
1673 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1674 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1675 | retcode = ERR_NOMEM; |
1676 | goto force_diskless_dec; | |
1677 | } | |
1678 | ||
1679 | /* Prevent shrinking of consistent devices ! */ | |
1680 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1681 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1682 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1683 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1684 | goto force_diskless_dec; |
1685 | } | |
1686 | ||
b411b363 PR |
1687 | /* Reset the "barriers don't work" bits here, then force meta data to |
1688 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1689 | if (new_disk_conf->md_flushes) |
b30ab791 | 1690 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1691 | else |
b30ab791 | 1692 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1693 | |
1694 | /* Point of no return reached. | |
1695 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1696 | * now device takes over responsibility, and the state engine should |
b411b363 | 1697 | * clean it up somewhere. */ |
0b0ba1ef | 1698 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1699 | device->ldev = nbc; |
1700 | device->resync = resync_lru; | |
1701 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1702 | nbc = NULL; |
1703 | resync_lru = NULL; | |
daeda1cc | 1704 | new_disk_conf = NULL; |
9958c857 | 1705 | new_plan = NULL; |
b411b363 | 1706 | |
a6b32bc3 | 1707 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
b411b363 | 1708 | |
b30ab791 AG |
1709 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1710 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1711 | else |
b30ab791 | 1712 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1713 | |
b30ab791 | 1714 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1715 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1716 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1717 | |
b30ab791 AG |
1718 | device->send_cnt = 0; |
1719 | device->recv_cnt = 0; | |
1720 | device->read_cnt = 0; | |
1721 | device->writ_cnt = 0; | |
b411b363 | 1722 | |
b30ab791 | 1723 | drbd_reconsider_max_bio_size(device); |
b411b363 PR |
1724 | |
1725 | /* If I am currently not R_PRIMARY, | |
1726 | * but meta data primary indicator is set, | |
1727 | * I just now recover from a hard crash, | |
1728 | * and have been R_PRIMARY before that crash. | |
1729 | * | |
1730 | * Now, if I had no connection before that crash | |
1731 | * (have been degraded R_PRIMARY), chances are that | |
1732 | * I won't find my peer now either. | |
1733 | * | |
1734 | * In that case, and _only_ in that case, | |
1735 | * we use the degr-wfc-timeout instead of the default, | |
1736 | * so we can automatically recover from a crash of a | |
1737 | * degraded but active "cluster" after a certain timeout. | |
1738 | */ | |
b30ab791 AG |
1739 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1740 | if (device->state.role != R_PRIMARY && | |
1741 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1742 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1743 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1744 | |
b30ab791 | 1745 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1746 | if (dd <= DS_ERROR) { |
b411b363 PR |
1747 | retcode = ERR_NOMEM_BITMAP; |
1748 | goto force_diskless_dec; | |
e96c9633 | 1749 | } else if (dd == DS_GREW) |
b30ab791 | 1750 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1751 | |
b30ab791 AG |
1752 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1753 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1754 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1755 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1756 | "(aka FullSync)\n"); |
b30ab791 | 1757 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1758 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1759 | retcode = ERR_IO_MD_DISK; |
1760 | goto force_diskless_dec; | |
1761 | } | |
1762 | } else { | |
b30ab791 | 1763 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1764 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1765 | retcode = ERR_IO_MD_DISK; |
1766 | goto force_diskless_dec; | |
1767 | } | |
b411b363 PR |
1768 | } |
1769 | ||
b30ab791 AG |
1770 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1771 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1772 | |
0500813f | 1773 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1774 | os = drbd_read_state(device); |
78bae59b | 1775 | ns = os; |
b411b363 PR |
1776 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1777 | otherwise investigate MDF_WasUpToDate... | |
1778 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1779 | otherwise into D_CONSISTENT state. | |
1780 | */ | |
b30ab791 AG |
1781 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1782 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1783 | ns.disk = D_CONSISTENT; |
1784 | else | |
1785 | ns.disk = D_OUTDATED; | |
1786 | } else { | |
1787 | ns.disk = D_INCONSISTENT; | |
1788 | } | |
1789 | ||
b30ab791 | 1790 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1791 | ns.pdsk = D_OUTDATED; |
1792 | ||
daeda1cc PR |
1793 | rcu_read_lock(); |
1794 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1795 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1796 | ns.disk = D_UP_TO_DATE; |
1797 | ||
1798 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1799 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1800 | this point, because drbd_request_state() modifies these | |
1801 | flags. */ | |
1802 | ||
b30ab791 AG |
1803 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1804 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1805 | else |
b30ab791 | 1806 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1807 | |
1808 | rcu_read_unlock(); | |
1809 | ||
b411b363 PR |
1810 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1811 | state after the negotiation phase. */ | |
b30ab791 AG |
1812 | if (device->state.conn == C_CONNECTED) { |
1813 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1814 | ns.i = os.i; |
1815 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1816 | |
1817 | /* We expect to receive up-to-date UUIDs soon. | |
1818 | To avoid a race in receive_state, free p_uuid while | |
1819 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1820 | kfree(device->p_uuid); |
1821 | device->p_uuid = NULL; | |
b411b363 PR |
1822 | } |
1823 | ||
b30ab791 | 1824 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1825 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1826 | |
1827 | if (rv < SS_SUCCESS) | |
1828 | goto force_diskless_dec; | |
1829 | ||
b30ab791 | 1830 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1831 | |
b30ab791 AG |
1832 | if (device->state.role == R_PRIMARY) |
1833 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1834 | else |
b30ab791 | 1835 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1836 | |
b30ab791 AG |
1837 | drbd_md_mark_dirty(device); |
1838 | drbd_md_sync(device); | |
b411b363 | 1839 | |
b30ab791 AG |
1840 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1841 | put_ldev(device); | |
a6b32bc3 | 1842 | conn_reconfig_done(first_peer_device(device)->connection); |
9e276872 | 1843 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 1844 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1845 | return 0; |
1846 | ||
1847 | force_diskless_dec: | |
b30ab791 | 1848 | put_ldev(device); |
b411b363 | 1849 | force_diskless: |
b30ab791 AG |
1850 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1851 | drbd_md_sync(device); | |
b411b363 | 1852 | fail: |
a6b32bc3 | 1853 | conn_reconfig_done(first_peer_device(device)->connection); |
b411b363 | 1854 | if (nbc) { |
e525fd89 TH |
1855 | if (nbc->backing_bdev) |
1856 | blkdev_put(nbc->backing_bdev, | |
1857 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1858 | if (nbc->md_bdev) | |
1859 | blkdev_put(nbc->md_bdev, | |
1860 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1861 | kfree(nbc); |
1862 | } | |
daeda1cc | 1863 | kfree(new_disk_conf); |
b411b363 | 1864 | lc_destroy(resync_lru); |
9958c857 | 1865 | kfree(new_plan); |
9e276872 | 1866 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 1867 | finish: |
a910b123 | 1868 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1869 | return 0; |
1870 | } | |
1871 | ||
b30ab791 | 1872 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1873 | { |
19f83c76 | 1874 | enum drbd_state_rv retcode; |
9a0d9d03 | 1875 | int ret; |
02ee8f95 | 1876 | |
cdfda633 | 1877 | if (force) { |
b30ab791 AG |
1878 | set_bit(FORCE_DETACH, &device->flags); |
1879 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1880 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1881 | goto out; |
1882 | } | |
1883 | ||
b30ab791 AG |
1884 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
1885 | drbd_md_get_buffer(device); /* make sure there is no in-flight meta-data IO */ | |
1886 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
1887 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1888 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1889 | ret = wait_event_interruptible(device->misc_wait, |
1890 | device->state.disk != D_FAILED); | |
1891 | drbd_resume_io(device); | |
9b2f61ae | 1892 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1893 | retcode = SS_NOTHING_TO_DO; |
1894 | if (ret) | |
1895 | retcode = ERR_INTR; | |
02ee8f95 | 1896 | out: |
85f75dd7 | 1897 | return retcode; |
b411b363 PR |
1898 | } |
1899 | ||
82f59cc6 LE |
1900 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1901 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1902 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1903 | * internal references as well. | |
1904 | * Only then we have finally detached. */ | |
3b98c0c2 | 1905 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1906 | { |
a910b123 | 1907 | struct drbd_config_context adm_ctx; |
116676ca | 1908 | enum drbd_ret_code retcode; |
cdfda633 PR |
1909 | struct detach_parms parms = { }; |
1910 | int err; | |
b411b363 | 1911 | |
a910b123 | 1912 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1913 | if (!adm_ctx.reply_skb) |
1914 | return retcode; | |
1915 | if (retcode != NO_ERROR) | |
1916 | goto out; | |
b411b363 | 1917 | |
cdfda633 PR |
1918 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1919 | err = detach_parms_from_attrs(&parms, info); | |
1920 | if (err) { | |
1921 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 1922 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
1923 | goto out; |
1924 | } | |
b411b363 PR |
1925 | } |
1926 | ||
9e276872 | 1927 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 1928 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 1929 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 1930 | out: |
a910b123 | 1931 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1932 | return 0; |
1933 | } | |
b411b363 | 1934 | |
bde89a9e | 1935 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1936 | { |
c06ece6b | 1937 | struct drbd_peer_device *peer_device; |
695d08fa | 1938 | bool rv = false; |
f399002e LE |
1939 | int vnr; |
1940 | ||
695d08fa | 1941 | rcu_read_lock(); |
c06ece6b AG |
1942 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1943 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1944 | if (device->state.conn == C_SYNC_SOURCE || |
1945 | device->state.conn == C_SYNC_TARGET || | |
1946 | device->state.conn == C_PAUSED_SYNC_S || | |
1947 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1948 | rv = true; |
1949 | break; | |
1950 | } | |
b411b363 | 1951 | } |
695d08fa | 1952 | rcu_read_unlock(); |
b411b363 | 1953 | |
695d08fa | 1954 | return rv; |
f399002e | 1955 | } |
47ff2d0a | 1956 | |
bde89a9e | 1957 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1958 | { |
c06ece6b | 1959 | struct drbd_peer_device *peer_device; |
695d08fa | 1960 | bool rv = false; |
f399002e LE |
1961 | int vnr; |
1962 | ||
695d08fa | 1963 | rcu_read_lock(); |
c06ece6b AG |
1964 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1965 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1966 | if (device->state.conn == C_VERIFY_S || |
1967 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1968 | rv = true; |
1969 | break; | |
47ff2d0a PR |
1970 | } |
1971 | } | |
695d08fa | 1972 | rcu_read_unlock(); |
b411b363 | 1973 | |
695d08fa | 1974 | return rv; |
f399002e | 1975 | } |
422028b1 | 1976 | |
cd64397c | 1977 | static enum drbd_ret_code |
270eb5c9 | 1978 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 1979 | { |
c06ece6b | 1980 | struct drbd_peer_device *peer_device; |
cd64397c | 1981 | int i; |
b411b363 | 1982 | |
270eb5c9 AG |
1983 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
1984 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 1985 | return ERR_NEED_APV_100; |
b411b363 | 1986 | |
270eb5c9 | 1987 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
1988 | return ERR_NEED_APV_100; |
1989 | ||
270eb5c9 | 1990 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 1991 | return ERR_NEED_APV_100; |
b411b363 PR |
1992 | } |
1993 | ||
270eb5c9 | 1994 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
1995 | conn_highest_role(connection) == R_PRIMARY && |
1996 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 1997 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 1998 | |
270eb5c9 AG |
1999 | if (new_net_conf->two_primaries && |
2000 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2001 | return ERR_NOT_PROTO_C; |
2002 | ||
c06ece6b AG |
2003 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2004 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2005 | if (get_ldev(device)) { |
2006 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2007 | put_ldev(device); | |
270eb5c9 | 2008 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2009 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2010 | } |
270eb5c9 | 2011 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2012 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2013 | } |
2014 | ||
270eb5c9 | 2015 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2016 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2017 | |
cd64397c PR |
2018 | return NO_ERROR; |
2019 | } | |
b411b363 | 2020 | |
44ed167d | 2021 | static enum drbd_ret_code |
270eb5c9 | 2022 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
2023 | { |
2024 | static enum drbd_ret_code rv; | |
c06ece6b | 2025 | struct drbd_peer_device *peer_device; |
44ed167d | 2026 | int i; |
b411b363 | 2027 | |
44ed167d | 2028 | rcu_read_lock(); |
270eb5c9 | 2029 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2030 | rcu_read_unlock(); |
b411b363 | 2031 | |
bde89a9e | 2032 | /* connection->volumes protected by genl_lock() here */ |
c06ece6b AG |
2033 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2034 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2035 | if (!device->bitmap) { |
2036 | if (drbd_bm_init(device)) | |
44ed167d | 2037 | return ERR_NOMEM; |
b411b363 PR |
2038 | } |
2039 | } | |
2040 | ||
44ed167d PR |
2041 | return rv; |
2042 | } | |
b411b363 | 2043 | |
0fd0ea06 PR |
2044 | struct crypto { |
2045 | struct crypto_hash *verify_tfm; | |
2046 | struct crypto_hash *csums_tfm; | |
2047 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2048 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2049 | }; |
b411b363 | 2050 | |
0fd0ea06 | 2051 | static int |
4b6ad6d4 | 2052 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2053 | { |
2054 | if (!tfm_name[0]) | |
2055 | return NO_ERROR; | |
b411b363 | 2056 | |
0fd0ea06 PR |
2057 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2058 | if (IS_ERR(*tfm)) { | |
2059 | *tfm = NULL; | |
2060 | return err_alg; | |
b411b363 | 2061 | } |
b411b363 | 2062 | |
0fd0ea06 PR |
2063 | return NO_ERROR; |
2064 | } | |
b411b363 | 2065 | |
0fd0ea06 | 2066 | static enum drbd_ret_code |
270eb5c9 | 2067 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2068 | { |
2069 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2070 | enum drbd_ret_code rv; | |
0fd0ea06 | 2071 | |
270eb5c9 | 2072 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2073 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2074 | if (rv != NO_ERROR) |
2075 | return rv; | |
270eb5c9 | 2076 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2077 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2078 | if (rv != NO_ERROR) |
2079 | return rv; | |
270eb5c9 | 2080 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2081 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2082 | if (rv != NO_ERROR) |
2083 | return rv; | |
270eb5c9 | 2084 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2085 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2086 | new_net_conf->cram_hmac_alg); |
b411b363 | 2087 | |
4b6ad6d4 AG |
2088 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2089 | ERR_AUTH_ALG); | |
b411b363 PR |
2090 | } |
2091 | ||
0fd0ea06 PR |
2092 | return rv; |
2093 | } | |
b411b363 | 2094 | |
0fd0ea06 PR |
2095 | static void free_crypto(struct crypto *crypto) |
2096 | { | |
0fd0ea06 | 2097 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2098 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2099 | crypto_free_hash(crypto->csums_tfm); |
2100 | crypto_free_hash(crypto->verify_tfm); | |
2101 | } | |
b411b363 | 2102 | |
f399002e LE |
2103 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2104 | { | |
a910b123 | 2105 | struct drbd_config_context adm_ctx; |
f399002e | 2106 | enum drbd_ret_code retcode; |
bde89a9e | 2107 | struct drbd_connection *connection; |
270eb5c9 | 2108 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2109 | int err; |
2110 | int ovr; /* online verify running */ | |
2111 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2112 | struct crypto crypto = { }; |
b411b363 | 2113 | |
a910b123 | 2114 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2115 | if (!adm_ctx.reply_skb) |
2116 | return retcode; | |
2117 | if (retcode != NO_ERROR) | |
9e276872 | 2118 | goto finish; |
b411b363 | 2119 | |
bde89a9e | 2120 | connection = adm_ctx.connection; |
9e276872 | 2121 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2122 | |
270eb5c9 AG |
2123 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2124 | if (!new_net_conf) { | |
f399002e LE |
2125 | retcode = ERR_NOMEM; |
2126 | goto out; | |
2127 | } | |
b411b363 | 2128 | |
bde89a9e | 2129 | conn_reconfig_start(connection); |
b411b363 | 2130 | |
bde89a9e | 2131 | mutex_lock(&connection->data.mutex); |
0500813f | 2132 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2133 | old_net_conf = connection->net_conf; |
2561b9c1 | 2134 | |
270eb5c9 | 2135 | if (!old_net_conf) { |
a910b123 | 2136 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2137 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2138 | goto fail; |
2139 | } | |
2140 | ||
270eb5c9 | 2141 | *new_net_conf = *old_net_conf; |
5979e361 | 2142 | if (should_set_defaults(info)) |
270eb5c9 | 2143 | set_net_conf_defaults(new_net_conf); |
f399002e | 2144 | |
270eb5c9 | 2145 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2146 | if (err && err != -ENOMSG) { |
f399002e | 2147 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2148 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2149 | goto fail; |
2561b9c1 | 2150 | } |
b411b363 | 2151 | |
270eb5c9 | 2152 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2153 | if (retcode != NO_ERROR) |
2154 | goto fail; | |
b411b363 | 2155 | |
f399002e | 2156 | /* re-sync running */ |
bde89a9e | 2157 | rsr = conn_resync_running(connection); |
270eb5c9 | 2158 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2159 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2160 | goto fail; |
b411b363 PR |
2161 | } |
2162 | ||
f399002e | 2163 | /* online verify running */ |
bde89a9e | 2164 | ovr = conn_ov_running(connection); |
270eb5c9 | 2165 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2166 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2167 | goto fail; |
f399002e | 2168 | } |
b411b363 | 2169 | |
270eb5c9 | 2170 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2171 | if (retcode != NO_ERROR) |
b411b363 | 2172 | goto fail; |
f399002e | 2173 | |
270eb5c9 | 2174 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2175 | |
2176 | if (!rsr) { | |
bde89a9e AG |
2177 | crypto_free_hash(connection->csums_tfm); |
2178 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2179 | crypto.csums_tfm = NULL; |
f399002e LE |
2180 | } |
2181 | if (!ovr) { | |
bde89a9e AG |
2182 | crypto_free_hash(connection->verify_tfm); |
2183 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2184 | crypto.verify_tfm = NULL; |
b411b363 PR |
2185 | } |
2186 | ||
bde89a9e AG |
2187 | crypto_free_hash(connection->integrity_tfm); |
2188 | connection->integrity_tfm = crypto.integrity_tfm; | |
2189 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2190 | /* Do this without trying to take connection->data.mutex again. */ | |
2191 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2192 | |
bde89a9e AG |
2193 | crypto_free_hash(connection->cram_hmac_tfm); |
2194 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2195 | |
0500813f | 2196 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2197 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2198 | synchronize_rcu(); |
270eb5c9 | 2199 | kfree(old_net_conf); |
91fd4dad | 2200 | |
69a22773 AG |
2201 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2202 | struct drbd_peer_device *peer_device; | |
2203 | int vnr; | |
2204 | ||
2205 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2206 | drbd_send_sync_param(peer_device); | |
2207 | } | |
f399002e | 2208 | |
91fd4dad PR |
2209 | goto done; |
2210 | ||
b411b363 | 2211 | fail: |
0500813f | 2212 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2213 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2214 | free_crypto(&crypto); |
270eb5c9 | 2215 | kfree(new_net_conf); |
91fd4dad | 2216 | done: |
bde89a9e | 2217 | conn_reconfig_done(connection); |
f399002e | 2218 | out: |
9e276872 LE |
2219 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2220 | finish: | |
a910b123 | 2221 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2222 | return 0; |
2223 | } | |
2224 | ||
3b98c0c2 | 2225 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2226 | { |
a910b123 | 2227 | struct drbd_config_context adm_ctx; |
c06ece6b | 2228 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2229 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2230 | struct crypto crypto = { }; |
77c556f6 | 2231 | struct drbd_resource *resource; |
bde89a9e | 2232 | struct drbd_connection *connection; |
3b98c0c2 LE |
2233 | enum drbd_ret_code retcode; |
2234 | int i; | |
2235 | int err; | |
b411b363 | 2236 | |
a910b123 | 2237 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2238 | |
3b98c0c2 LE |
2239 | if (!adm_ctx.reply_skb) |
2240 | return retcode; | |
2241 | if (retcode != NO_ERROR) | |
2242 | goto out; | |
089c075d | 2243 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2244 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2245 | retcode = ERR_INVALID_REQUEST; |
2246 | goto out; | |
2247 | } | |
b411b363 | 2248 | |
089c075d AG |
2249 | /* No need for _rcu here. All reconfiguration is |
2250 | * strictly serialized on genl_lock(). We are protected against | |
2251 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2252 | for_each_resource(resource, &drbd_resources) { |
2253 | for_each_connection(connection, resource) { | |
2254 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2255 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2256 | connection->my_addr_len)) { | |
2257 | retcode = ERR_LOCAL_ADDR; | |
2258 | goto out; | |
2259 | } | |
b411b363 | 2260 | |
77c556f6 AG |
2261 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2262 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2263 | connection->peer_addr_len)) { | |
2264 | retcode = ERR_PEER_ADDR; | |
2265 | goto out; | |
2266 | } | |
089c075d | 2267 | } |
b411b363 PR |
2268 | } |
2269 | ||
9e276872 | 2270 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2271 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2272 | conn_reconfig_start(connection); |
b411b363 | 2273 | |
bde89a9e | 2274 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2275 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2276 | goto fail; |
2277 | } | |
2278 | ||
a209b4ae | 2279 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2280 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2281 | if (!new_net_conf) { | |
b411b363 | 2282 | retcode = ERR_NOMEM; |
b411b363 PR |
2283 | goto fail; |
2284 | } | |
2285 | ||
270eb5c9 | 2286 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2287 | |
270eb5c9 | 2288 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2289 | if (err && err != -ENOMSG) { |
b411b363 | 2290 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2291 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2292 | goto fail; |
2293 | } | |
2294 | ||
270eb5c9 | 2295 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2296 | if (retcode != NO_ERROR) |
422028b1 | 2297 | goto fail; |
b411b363 | 2298 | |
270eb5c9 | 2299 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2300 | if (retcode != NO_ERROR) |
2301 | goto fail; | |
b411b363 | 2302 | |
270eb5c9 | 2303 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2304 | |
b5043c5e | 2305 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2306 | |
0500813f | 2307 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2308 | old_net_conf = connection->net_conf; |
2309 | if (old_net_conf) { | |
b411b363 | 2310 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2311 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2312 | goto fail; |
2313 | } | |
270eb5c9 | 2314 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2315 | |
bde89a9e AG |
2316 | conn_free_crypto(connection); |
2317 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2318 | connection->integrity_tfm = crypto.integrity_tfm; | |
2319 | connection->csums_tfm = crypto.csums_tfm; | |
2320 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2321 | |
bde89a9e AG |
2322 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2323 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2324 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2325 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2326 | |
0500813f | 2327 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2328 | |
695d08fa | 2329 | rcu_read_lock(); |
c06ece6b AG |
2330 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2331 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2332 | device->send_cnt = 0; |
2333 | device->recv_cnt = 0; | |
b411b363 | 2334 | } |
695d08fa | 2335 | rcu_read_unlock(); |
b411b363 | 2336 | |
bde89a9e | 2337 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2338 | |
bde89a9e | 2339 | conn_reconfig_done(connection); |
9e276872 | 2340 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2341 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2342 | return 0; |
b411b363 | 2343 | |
b411b363 | 2344 | fail: |
0fd0ea06 | 2345 | free_crypto(&crypto); |
270eb5c9 | 2346 | kfree(new_net_conf); |
b411b363 | 2347 | |
bde89a9e | 2348 | conn_reconfig_done(connection); |
9e276872 | 2349 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2350 | out: |
a910b123 | 2351 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2352 | return 0; |
2353 | } | |
2354 | ||
bde89a9e | 2355 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2356 | { |
2357 | enum drbd_state_rv rv; | |
85f75dd7 | 2358 | |
bde89a9e | 2359 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2360 | force ? CS_HARD : 0); |
85f75dd7 LE |
2361 | |
2362 | switch (rv) { | |
2363 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2364 | break; |
85f75dd7 LE |
2365 | case SS_ALREADY_STANDALONE: |
2366 | return SS_SUCCESS; | |
2367 | case SS_PRIMARY_NOP: | |
2368 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2369 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2370 | |
2371 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2372 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2373 | |
85f75dd7 LE |
2374 | break; |
2375 | case SS_CW_FAILED_BY_PEER: | |
2376 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2377 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2378 | disk, D_OUTDATED), 0); |
2379 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2380 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2381 | CS_HARD); |
b411b363 | 2382 | } |
85f75dd7 LE |
2383 | break; |
2384 | default:; | |
2385 | /* no special handling necessary */ | |
2386 | } | |
2387 | ||
f3dfa40a LE |
2388 | if (rv >= SS_SUCCESS) { |
2389 | enum drbd_state_rv rv2; | |
2390 | /* No one else can reconfigure the network while I am here. | |
2391 | * The state handling only uses drbd_thread_stop_nowait(), | |
2392 | * we want to really wait here until the receiver is no more. | |
2393 | */ | |
9693da23 | 2394 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2395 | |
2396 | /* Race breaker. This additional state change request may be | |
2397 | * necessary, if this was a forced disconnect during a receiver | |
2398 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2399 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2400 | * C_STANDALONE already, now, and this becomes a no-op. |
2401 | */ | |
bde89a9e | 2402 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2403 | CS_VERBOSE | CS_HARD); |
2404 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2405 | drbd_err(connection, |
f3dfa40a LE |
2406 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2407 | rv2); | |
b411b363 | 2408 | } |
85f75dd7 LE |
2409 | return rv; |
2410 | } | |
b411b363 | 2411 | |
3b98c0c2 | 2412 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2413 | { |
a910b123 | 2414 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2415 | struct disconnect_parms parms; |
bde89a9e | 2416 | struct drbd_connection *connection; |
85f75dd7 | 2417 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2418 | enum drbd_ret_code retcode; |
2419 | int err; | |
2561b9c1 | 2420 | |
a910b123 | 2421 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2422 | if (!adm_ctx.reply_skb) |
2423 | return retcode; | |
2424 | if (retcode != NO_ERROR) | |
2561b9c1 | 2425 | goto fail; |
b411b363 | 2426 | |
bde89a9e | 2427 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2428 | memset(&parms, 0, sizeof(parms)); |
2429 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2430 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2431 | if (err) { |
2432 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2433 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2434 | goto fail; |
2435 | } | |
2436 | } | |
2437 | ||
9e276872 | 2438 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2439 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2440 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2441 | retcode = rv; /* FIXME: Type mismatch. */ |
2442 | else | |
2443 | retcode = NO_ERROR; | |
9e276872 | 2444 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2445 | fail: |
a910b123 | 2446 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2447 | return 0; |
2448 | } | |
2449 | ||
b30ab791 | 2450 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2451 | { |
2452 | int iass; /* I am sync source */ | |
2453 | ||
d0180171 | 2454 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2455 | if (device->state.role != device->state.peer) |
2456 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2457 | else |
a6b32bc3 | 2458 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2459 | |
2460 | if (iass) | |
b30ab791 | 2461 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2462 | else |
b30ab791 | 2463 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2464 | } |
2465 | ||
3b98c0c2 | 2466 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2467 | { |
a910b123 | 2468 | struct drbd_config_context adm_ctx; |
daeda1cc | 2469 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2470 | struct resize_parms rs; |
b30ab791 | 2471 | struct drbd_device *device; |
3b98c0c2 | 2472 | enum drbd_ret_code retcode; |
b411b363 | 2473 | enum determine_dev_size dd; |
d752b269 | 2474 | bool change_al_layout = false; |
6495d2c6 | 2475 | enum dds_flags ddsf; |
daeda1cc | 2476 | sector_t u_size; |
3b98c0c2 | 2477 | int err; |
b411b363 | 2478 | |
a910b123 | 2479 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2480 | if (!adm_ctx.reply_skb) |
2481 | return retcode; | |
2482 | if (retcode != NO_ERROR) | |
9e276872 | 2483 | goto finish; |
3b98c0c2 | 2484 | |
9e276872 | 2485 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2486 | device = adm_ctx.device; |
2487 | if (!get_ldev(device)) { | |
d752b269 PR |
2488 | retcode = ERR_NO_DISK; |
2489 | goto fail; | |
2490 | } | |
2491 | ||
3b98c0c2 | 2492 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2493 | rs.al_stripes = device->ldev->md.al_stripes; |
2494 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2495 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2496 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2497 | if (err) { |
3b98c0c2 | 2498 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2499 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2500 | goto fail_ldev; |
b411b363 PR |
2501 | } |
2502 | } | |
2503 | ||
b30ab791 | 2504 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2505 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2506 | goto fail_ldev; |
b411b363 | 2507 | } |
b411b363 | 2508 | |
b30ab791 AG |
2509 | if (device->state.role == R_SECONDARY && |
2510 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2511 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2512 | goto fail_ldev; |
b411b363 | 2513 | } |
b411b363 | 2514 | |
a6b32bc3 | 2515 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2516 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2517 | goto fail_ldev; |
6495d2c6 PR |
2518 | } |
2519 | ||
daeda1cc | 2520 | rcu_read_lock(); |
b30ab791 | 2521 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2522 | rcu_read_unlock(); |
2523 | if (u_size != (sector_t)rs.resize_size) { | |
2524 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2525 | if (!new_disk_conf) { | |
778f271d | 2526 | retcode = ERR_NOMEM; |
9bcd2521 | 2527 | goto fail_ldev; |
778f271d PR |
2528 | } |
2529 | } | |
2530 | ||
b30ab791 AG |
2531 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2532 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2533 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2534 | ||
2535 | if (al_size_k > (16 * 1024 * 1024)) { | |
2536 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2537 | goto fail_ldev; | |
2538 | } | |
2539 | ||
2540 | if (al_size_k < MD_32kB_SECT/2) { | |
2541 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2542 | goto fail_ldev; | |
2543 | } | |
2544 | ||
cdc6af8d | 2545 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2546 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2547 | goto fail_ldev; | |
2548 | } | |
2549 | ||
2550 | change_al_layout = true; | |
2551 | } | |
2552 | ||
b30ab791 AG |
2553 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2554 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2555 | |
daeda1cc | 2556 | if (new_disk_conf) { |
0500813f | 2557 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2558 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2559 | *new_disk_conf = *old_disk_conf; |
2560 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2561 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2562 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2563 | synchronize_rcu(); |
2564 | kfree(old_disk_conf); | |
b411b363 PR |
2565 | } |
2566 | ||
6495d2c6 | 2567 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2568 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2569 | drbd_md_sync(device); | |
2570 | put_ldev(device); | |
e96c9633 | 2571 | if (dd == DS_ERROR) { |
b411b363 PR |
2572 | retcode = ERR_NOMEM_BITMAP; |
2573 | goto fail; | |
d752b269 PR |
2574 | } else if (dd == DS_ERROR_SPACE_MD) { |
2575 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2576 | goto fail; | |
2577 | } else if (dd == DS_ERROR_SHRINK) { | |
2578 | retcode = ERR_IMPLICIT_SHRINK; | |
2579 | goto fail; | |
b411b363 | 2580 | } |
778f271d | 2581 | |
b30ab791 | 2582 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2583 | if (dd == DS_GREW) |
b30ab791 | 2584 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2585 | |
69a22773 AG |
2586 | drbd_send_uuids(first_peer_device(device)); |
2587 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2588 | } |
2589 | ||
b411b363 | 2590 | fail: |
9e276872 LE |
2591 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2592 | finish: | |
a910b123 | 2593 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2594 | return 0; |
b411b363 | 2595 | |
9bcd2521 | 2596 | fail_ldev: |
b30ab791 | 2597 | put_ldev(device); |
9bcd2521 | 2598 | goto fail; |
b411b363 | 2599 | } |
b411b363 | 2600 | |
f399002e | 2601 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2602 | { |
a910b123 | 2603 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2604 | enum drbd_ret_code retcode; |
b57a1e27 | 2605 | struct res_opts res_opts; |
f399002e | 2606 | int err; |
b411b363 | 2607 | |
a910b123 | 2608 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2609 | if (!adm_ctx.reply_skb) |
2610 | return retcode; | |
2611 | if (retcode != NO_ERROR) | |
2612 | goto fail; | |
b411b363 | 2613 | |
eb6bea67 | 2614 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2615 | if (should_set_defaults(info)) |
b966b5dd | 2616 | set_res_opts_defaults(&res_opts); |
b411b363 | 2617 | |
b57a1e27 | 2618 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2619 | if (err && err != -ENOMSG) { |
b411b363 | 2620 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2621 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2622 | goto fail; |
2623 | } | |
2624 | ||
9e276872 | 2625 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2626 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2627 | if (err) { |
2628 | retcode = ERR_INVALID_REQUEST; | |
2629 | if (err == -ENOMEM) | |
2630 | retcode = ERR_NOMEM; | |
b411b363 | 2631 | } |
9e276872 | 2632 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2633 | |
b411b363 | 2634 | fail: |
a910b123 | 2635 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2636 | return 0; |
2637 | } | |
2638 | ||
3b98c0c2 | 2639 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2640 | { |
a910b123 | 2641 | struct drbd_config_context adm_ctx; |
b30ab791 | 2642 | struct drbd_device *device; |
3b98c0c2 LE |
2643 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2644 | ||
a910b123 | 2645 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2646 | if (!adm_ctx.reply_skb) |
2647 | return retcode; | |
2648 | if (retcode != NO_ERROR) | |
2649 | goto out; | |
2650 | ||
9e276872 | 2651 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2652 | device = adm_ctx.device; |
b411b363 | 2653 | |
194bfb32 | 2654 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2655 | * resync just being finished, wait for it before requesting a new resync. |
2656 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2657 | drbd_suspend_io(device); |
2658 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2659 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2660 | |
0b2dafcd PR |
2661 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2662 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2663 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2664 | */ |
b30ab791 AG |
2665 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2666 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2667 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2668 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2669 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2670 | retcode = ERR_IO_MD_DISK; | |
2671 | } | |
2672 | } else | |
b30ab791 AG |
2673 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2674 | drbd_resume_io(device); | |
9e276872 | 2675 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2676 | out: |
a910b123 | 2677 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2678 | return 0; |
2679 | } | |
2680 | ||
3b98c0c2 LE |
2681 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2682 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2683 | { |
a910b123 | 2684 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2685 | enum drbd_ret_code retcode; |
194bfb32 | 2686 | |
a910b123 | 2687 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2688 | if (!adm_ctx.reply_skb) |
2689 | return retcode; | |
2690 | if (retcode != NO_ERROR) | |
2691 | goto out; | |
b411b363 | 2692 | |
9e276872 | 2693 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2694 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 2695 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2696 | out: |
a910b123 | 2697 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2698 | return 0; |
2699 | } | |
2700 | ||
b30ab791 | 2701 | static int drbd_bmio_set_susp_al(struct drbd_device *device) |
0778286a PR |
2702 | { |
2703 | int rv; | |
2704 | ||
b30ab791 AG |
2705 | rv = drbd_bmio_set_n_write(device); |
2706 | drbd_suspend_al(device); | |
0778286a PR |
2707 | return rv; |
2708 | } | |
2709 | ||
3b98c0c2 | 2710 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2711 | { |
a910b123 | 2712 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 2713 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2714 | struct drbd_device *device; |
25b0d6c8 | 2715 | |
a910b123 | 2716 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
2717 | if (!adm_ctx.reply_skb) |
2718 | return retcode; | |
2719 | if (retcode != NO_ERROR) | |
2720 | goto out; | |
2721 | ||
9e276872 | 2722 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2723 | device = adm_ctx.device; |
b411b363 | 2724 | |
194bfb32 | 2725 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2726 | * resync just being finished, wait for it before requesting a new resync. |
2727 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2728 | drbd_suspend_io(device); |
2729 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2730 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2731 | |
0b2dafcd PR |
2732 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2733 | * in the bitmap. Otherwise, try to start a resync handshake | |
2734 | * as sync source for full sync. | |
2735 | */ | |
b30ab791 | 2736 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2737 | /* The peer will get a resync upon connect anyways. Just make that |
2738 | into a full resync. */ | |
b30ab791 | 2739 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2740 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2741 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2742 | "set_n_write from invalidate_peer", |
2743 | BM_LOCKED_SET_ALLOWED)) | |
2744 | retcode = ERR_IO_MD_DISK; | |
2745 | } | |
2746 | } else | |
b30ab791 AG |
2747 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2748 | drbd_resume_io(device); | |
9e276872 | 2749 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
25b0d6c8 | 2750 | out: |
a910b123 | 2751 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2752 | return 0; |
2753 | } | |
2754 | ||
3b98c0c2 | 2755 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2756 | { |
a910b123 | 2757 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2758 | enum drbd_ret_code retcode; |
b411b363 | 2759 | |
a910b123 | 2760 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2761 | if (!adm_ctx.reply_skb) |
2762 | return retcode; | |
2763 | if (retcode != NO_ERROR) | |
2764 | goto out; | |
b411b363 | 2765 | |
9e276872 | 2766 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2767 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 2768 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 2769 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2770 | out: |
a910b123 | 2771 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2772 | return 0; |
2773 | } | |
2774 | ||
3b98c0c2 | 2775 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2776 | { |
a910b123 | 2777 | struct drbd_config_context adm_ctx; |
da9fbc27 | 2778 | union drbd_dev_state s; |
3b98c0c2 LE |
2779 | enum drbd_ret_code retcode; |
2780 | ||
a910b123 | 2781 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2782 | if (!adm_ctx.reply_skb) |
2783 | return retcode; | |
2784 | if (retcode != NO_ERROR) | |
2785 | goto out; | |
b411b363 | 2786 | |
9e276872 | 2787 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2788 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2789 | s = adm_ctx.device->state; | |
cd88d030 PR |
2790 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2791 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2792 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2793 | } else { | |
2794 | retcode = ERR_PAUSE_IS_CLEAR; | |
2795 | } | |
2796 | } | |
9e276872 | 2797 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2798 | out: |
a910b123 | 2799 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2800 | return 0; |
2801 | } | |
2802 | ||
3b98c0c2 | 2803 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2804 | { |
3b98c0c2 | 2805 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2806 | } |
2807 | ||
3b98c0c2 | 2808 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2809 | { |
a910b123 | 2810 | struct drbd_config_context adm_ctx; |
b30ab791 | 2811 | struct drbd_device *device; |
3b98c0c2 LE |
2812 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2813 | ||
a910b123 | 2814 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2815 | if (!adm_ctx.reply_skb) |
2816 | return retcode; | |
2817 | if (retcode != NO_ERROR) | |
2818 | goto out; | |
2819 | ||
9e276872 | 2820 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2821 | device = adm_ctx.device; |
2822 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2823 | drbd_uuid_new_current(device); | |
2824 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2825 | } |
b30ab791 AG |
2826 | drbd_suspend_io(device); |
2827 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2828 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2829 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2830 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2831 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2832 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2833 | } |
b30ab791 | 2834 | drbd_resume_io(device); |
9e276872 | 2835 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2836 | out: |
a910b123 | 2837 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2838 | return 0; |
2839 | } | |
2840 | ||
3b98c0c2 | 2841 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2842 | { |
3b98c0c2 | 2843 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2844 | } |
2845 | ||
251b8f8e AG |
2846 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2847 | struct drbd_resource *resource, | |
2848 | struct drbd_connection *connection, | |
2849 | struct drbd_device *device) | |
b411b363 | 2850 | { |
543cc10b LE |
2851 | struct nlattr *nla; |
2852 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2853 | if (!nla) | |
2854 | goto nla_put_failure; | |
251b8f8e AG |
2855 | if (device && |
2856 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2857 | goto nla_put_failure; |
f597f6b8 | 2858 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 2859 | goto nla_put_failure; |
251b8f8e AG |
2860 | if (connection) { |
2861 | if (connection->my_addr_len && | |
2862 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2863 | goto nla_put_failure; | |
2864 | if (connection->peer_addr_len && | |
2865 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2866 | goto nla_put_failure; | |
2867 | } | |
543cc10b LE |
2868 | nla_nest_end(skb, nla); |
2869 | return 0; | |
b411b363 | 2870 | |
543cc10b LE |
2871 | nla_put_failure: |
2872 | if (nla) | |
2873 | nla_nest_cancel(skb, nla); | |
2874 | return -EMSGSIZE; | |
2875 | } | |
b411b363 | 2876 | |
251b8f8e AG |
2877 | /* |
2878 | * Return the connection of @resource if @resource has exactly one connection. | |
2879 | */ | |
2880 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2881 | { | |
2882 | struct list_head *connections = &resource->connections; | |
2883 | ||
2884 | if (list_empty(connections) || connections->next->next != connections) | |
2885 | return NULL; | |
2886 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2887 | } | |
2888 | ||
2889 | int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, | |
3b98c0c2 | 2890 | const struct sib_info *sib) |
b411b363 | 2891 | { |
251b8f8e | 2892 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2893 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2894 | struct nlattr *nla; | |
2895 | int got_ldev; | |
3b98c0c2 LE |
2896 | int err = 0; |
2897 | int exclude_sensitive; | |
2898 | ||
2899 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2900 | * to. So we better exclude_sensitive information. | |
2901 | * | |
2902 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2903 | * in the context of the requesting user process. Exclude sensitive | |
2904 | * information, unless current has superuser. | |
2905 | * | |
2906 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2907 | * relies on the current implementation of netlink_dump(), which | |
2908 | * executes the dump callback successively from netlink_recvmsg(), | |
2909 | * always in the context of the receiving process */ | |
2910 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2911 | ||
b30ab791 | 2912 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2913 | |
2914 | /* We need to add connection name and volume number information still. | |
2915 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 2916 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 2917 | goto nla_put_failure; |
3b98c0c2 | 2918 | |
eb6bea67 | 2919 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
2920 | goto nla_put_failure; |
2921 | ||
daeda1cc | 2922 | rcu_read_lock(); |
f9eb7bf4 AG |
2923 | if (got_ldev) { |
2924 | struct disk_conf *disk_conf; | |
44ed167d | 2925 | |
b30ab791 | 2926 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2927 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2928 | } | |
2929 | if (!err) { | |
2930 | struct net_conf *nc; | |
2931 | ||
a6b32bc3 | 2932 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2933 | if (nc) |
2934 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2935 | } | |
44ed167d PR |
2936 | rcu_read_unlock(); |
2937 | if (err) | |
2938 | goto nla_put_failure; | |
3b98c0c2 | 2939 | |
3b98c0c2 LE |
2940 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2941 | if (!nla) | |
2942 | goto nla_put_failure; | |
26ec9287 | 2943 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2944 | nla_put_u32(skb, T_current_state, device->state.i) || |
2945 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2946 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2947 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2948 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2949 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2950 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2951 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2952 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2953 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2954 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2955 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2956 | goto nla_put_failure; |
3b98c0c2 LE |
2957 | |
2958 | if (got_ldev) { | |
39a1aa7f | 2959 | int err; |
b411b363 | 2960 | |
b30ab791 AG |
2961 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2962 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2963 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2964 | |
2965 | if (err) | |
2966 | goto nla_put_failure; | |
2967 | ||
b30ab791 AG |
2968 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
2969 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
2970 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 2971 | goto nla_put_failure; |
b30ab791 AG |
2972 | if (C_SYNC_SOURCE <= device->state.conn && |
2973 | C_PAUSED_SYNC_T >= device->state.conn) { | |
2974 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
2975 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 2976 | goto nla_put_failure; |
3b98c0c2 | 2977 | } |
b411b363 | 2978 | } |
b411b363 | 2979 | |
3b98c0c2 LE |
2980 | if (sib) { |
2981 | switch(sib->sib_reason) { | |
2982 | case SIB_SYNC_PROGRESS: | |
2983 | case SIB_GET_STATUS_REPLY: | |
2984 | break; | |
2985 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2986 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2987 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2988 | goto nla_put_failure; | |
3b98c0c2 LE |
2989 | break; |
2990 | case SIB_HELPER_POST: | |
26ec9287 AG |
2991 | if (nla_put_u32(skb, T_helper_exit_code, |
2992 | sib->helper_exit_code)) | |
2993 | goto nla_put_failure; | |
3b98c0c2 LE |
2994 | /* fall through */ |
2995 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2996 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2997 | goto nla_put_failure; | |
3b98c0c2 LE |
2998 | break; |
2999 | } | |
b411b363 | 3000 | } |
3b98c0c2 | 3001 | nla_nest_end(skb, nla); |
b411b363 | 3002 | |
3b98c0c2 LE |
3003 | if (0) |
3004 | nla_put_failure: | |
3005 | err = -EMSGSIZE; | |
3006 | if (got_ldev) | |
b30ab791 | 3007 | put_ldev(device); |
3b98c0c2 | 3008 | return err; |
b411b363 PR |
3009 | } |
3010 | ||
3b98c0c2 | 3011 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3012 | { |
a910b123 | 3013 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3014 | enum drbd_ret_code retcode; |
3015 | int err; | |
b411b363 | 3016 | |
a910b123 | 3017 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3018 | if (!adm_ctx.reply_skb) |
3019 | return retcode; | |
3020 | if (retcode != NO_ERROR) | |
3021 | goto out; | |
b411b363 | 3022 | |
b30ab791 | 3023 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3024 | if (err) { |
3025 | nlmsg_free(adm_ctx.reply_skb); | |
3026 | return err; | |
b411b363 | 3027 | } |
3b98c0c2 | 3028 | out: |
a910b123 | 3029 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3030 | return 0; |
b411b363 PR |
3031 | } |
3032 | ||
4b7a530f | 3033 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3034 | { |
b30ab791 | 3035 | struct drbd_device *device; |
3b98c0c2 | 3036 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3037 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3038 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3039 | struct drbd_resource *tmp; |
543cc10b LE |
3040 | unsigned volume = cb->args[1]; |
3041 | ||
3042 | /* Open coded, deferred, iteration: | |
77c556f6 | 3043 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3044 | * connection = "first connection of resource or undefined"; |
3045 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3046 | * ... |
3047 | * } | |
3048 | * } | |
77c556f6 | 3049 | * where resource is cb->args[0]; |
543cc10b LE |
3050 | * and i is cb->args[1]; |
3051 | * | |
71932efc LE |
3052 | * cb->args[2] indicates if we shall loop over all resources, |
3053 | * or just dump all volumes of a single resource. | |
3054 | * | |
3b98c0c2 LE |
3055 | * This may miss entries inserted after this dump started, |
3056 | * or entries deleted before they are reached. | |
543cc10b | 3057 | * |
b30ab791 | 3058 | * We need to make sure the device won't disappear while |
543cc10b LE |
3059 | * we are looking at it, and revalidate our iterators |
3060 | * on each iteration. | |
3061 | */ | |
b411b363 | 3062 | |
05a10ec7 | 3063 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3064 | rcu_read_lock(); |
543cc10b | 3065 | /* revalidate iterator position */ |
77c556f6 | 3066 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3067 | if (pos == NULL) { |
3068 | /* first iteration */ | |
3069 | pos = tmp; | |
77c556f6 | 3070 | resource = pos; |
543cc10b LE |
3071 | break; |
3072 | } | |
3073 | if (tmp == pos) { | |
77c556f6 | 3074 | resource = pos; |
543cc10b LE |
3075 | break; |
3076 | } | |
b411b363 | 3077 | } |
77c556f6 AG |
3078 | if (resource) { |
3079 | next_resource: | |
251b8f8e AG |
3080 | device = idr_get_next(&resource->devices, &volume); |
3081 | if (!device) { | |
77c556f6 AG |
3082 | /* No more volumes to dump on this resource. |
3083 | * Advance resource iterator. */ | |
3084 | pos = list_entry_rcu(resource->resources.next, | |
3085 | struct drbd_resource, resources); | |
3086 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3087 | if (volume != 0) { |
71932efc LE |
3088 | /* If we reached the end of the list, |
3089 | * or only a single resource dump was requested, | |
3090 | * we are done. */ | |
77c556f6 | 3091 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3092 | goto out; |
543cc10b | 3093 | volume = 0; |
77c556f6 AG |
3094 | resource = pos; |
3095 | goto next_resource; | |
543cc10b LE |
3096 | } |
3097 | } | |
3098 | ||
98683650 | 3099 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3100 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3101 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3102 | if (!dh) | |
543cc10b LE |
3103 | goto out; |
3104 | ||
251b8f8e | 3105 | if (!device) { |
bde89a9e | 3106 | /* This is a connection without a single volume. |
367d675d LE |
3107 | * Suprisingly enough, it may have a network |
3108 | * configuration. */ | |
251b8f8e AG |
3109 | struct drbd_connection *connection; |
3110 | ||
543cc10b LE |
3111 | dh->minor = -1U; |
3112 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3113 | connection = the_only_connection(resource); |
3114 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3115 | goto cancel; |
251b8f8e AG |
3116 | if (connection) { |
3117 | struct net_conf *nc; | |
3118 | ||
3119 | nc = rcu_dereference(connection->net_conf); | |
3120 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3121 | goto cancel; | |
3122 | } | |
367d675d | 3123 | goto done; |
543cc10b | 3124 | } |
b411b363 | 3125 | |
0b0ba1ef | 3126 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3127 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3128 | |
b30ab791 | 3129 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3130 | dh->ret_code = NO_ERROR; |
3131 | ||
b30ab791 | 3132 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3133 | cancel: |
3b98c0c2 | 3134 | genlmsg_cancel(skb, dh); |
543cc10b | 3135 | goto out; |
3b98c0c2 | 3136 | } |
367d675d | 3137 | done: |
3b98c0c2 | 3138 | genlmsg_end(skb, dh); |
bde89a9e | 3139 | } |
b411b363 | 3140 | |
543cc10b | 3141 | out: |
c141ebda | 3142 | rcu_read_unlock(); |
543cc10b | 3143 | /* where to start the next iteration */ |
bde89a9e | 3144 | cb->args[0] = (long)pos; |
77c556f6 | 3145 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3146 | |
77c556f6 | 3147 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3148 | * Which will terminate the dump. */ |
3b98c0c2 | 3149 | return skb->len; |
b411b363 PR |
3150 | } |
3151 | ||
71932efc LE |
3152 | /* |
3153 | * Request status of all resources, or of all volumes within a single resource. | |
3154 | * | |
3155 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3156 | * Which means we cannot use the family->attrbuf or other such members, because | |
3157 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3158 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3159 | * | |
3160 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3161 | */ |
71932efc | 3162 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3163 | { |
71932efc LE |
3164 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3165 | struct nlattr *nla; | |
7c3063cc | 3166 | const char *resource_name; |
4bc76048 | 3167 | struct drbd_resource *resource; |
7c3063cc | 3168 | int maxtype; |
71932efc LE |
3169 | |
3170 | /* Is this a followup call? */ | |
3171 | if (cb->args[0]) { | |
3172 | /* ... of a single resource dump, | |
3173 | * and the resource iterator has been advanced already? */ | |
3174 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3175 | return 0; /* DONE. */ | |
3176 | goto dump; | |
3177 | } | |
3178 | ||
3179 | /* First call (from netlink_dump_start). We need to figure out | |
3180 | * which resource(s) the user wants us to dump. */ | |
3181 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3182 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3183 | DRBD_NLA_CFG_CONTEXT); | |
3184 | ||
3185 | /* No explicit context given. Dump all. */ | |
3186 | if (!nla) | |
3187 | goto dump; | |
7c3063cc AG |
3188 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3189 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3190 | if (IS_ERR(nla)) | |
3191 | return PTR_ERR(nla); | |
71932efc LE |
3192 | /* context given, but no name present? */ |
3193 | if (!nla) | |
3194 | return -EINVAL; | |
7c3063cc | 3195 | resource_name = nla_data(nla); |
4bc76048 AG |
3196 | if (!*resource_name) |
3197 | return -ENODEV; | |
3198 | resource = drbd_find_resource(resource_name); | |
3199 | if (!resource) | |
71932efc LE |
3200 | return -ENODEV; |
3201 | ||
4bc76048 | 3202 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3203 | |
71932efc | 3204 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3205 | * only dump this connection. */ |
4bc76048 | 3206 | cb->args[0] = (long)resource; |
71932efc | 3207 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3208 | cb->args[2] = (long)resource; |
71932efc LE |
3209 | |
3210 | dump: | |
3211 | return get_one_status(skb, cb); | |
3212 | } | |
b411b363 | 3213 | |
3b98c0c2 | 3214 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3215 | { |
a910b123 | 3216 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3217 | enum drbd_ret_code retcode; |
3218 | struct timeout_parms tp; | |
3219 | int err; | |
b411b363 | 3220 | |
a910b123 | 3221 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3222 | if (!adm_ctx.reply_skb) |
3223 | return retcode; | |
3224 | if (retcode != NO_ERROR) | |
3225 | goto out; | |
b411b363 | 3226 | |
3b98c0c2 | 3227 | tp.timeout_type = |
b30ab791 AG |
3228 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3229 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3230 | UT_DEFAULT; |
b411b363 | 3231 | |
3b98c0c2 LE |
3232 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3233 | if (err) { | |
3234 | nlmsg_free(adm_ctx.reply_skb); | |
3235 | return err; | |
3236 | } | |
3237 | out: | |
a910b123 | 3238 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3239 | return 0; |
b411b363 PR |
3240 | } |
3241 | ||
3b98c0c2 | 3242 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3243 | { |
a910b123 | 3244 | struct drbd_config_context adm_ctx; |
b30ab791 | 3245 | struct drbd_device *device; |
3b98c0c2 | 3246 | enum drbd_ret_code retcode; |
58ffa580 | 3247 | struct start_ov_parms parms; |
b411b363 | 3248 | |
a910b123 | 3249 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3250 | if (!adm_ctx.reply_skb) |
3251 | return retcode; | |
3252 | if (retcode != NO_ERROR) | |
3253 | goto out; | |
873b0d5f | 3254 | |
b30ab791 | 3255 | device = adm_ctx.device; |
58ffa580 LE |
3256 | |
3257 | /* resume from last known position, if possible */ | |
b30ab791 | 3258 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3259 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3260 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3261 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3262 | if (err) { |
3263 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3264 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3265 | goto out; |
3266 | } | |
b411b363 | 3267 | } |
9e276872 LE |
3268 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3269 | ||
58ffa580 | 3270 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3271 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3272 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3273 | |
3274 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3275 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3276 | drbd_suspend_io(device); |
3277 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3278 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3279 | drbd_resume_io(device); | |
9e276872 LE |
3280 | |
3281 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 3282 | out: |
a910b123 | 3283 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3284 | return 0; |
3285 | } | |
3286 | ||
3287 | ||
3b98c0c2 | 3288 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3289 | { |
a910b123 | 3290 | struct drbd_config_context adm_ctx; |
b30ab791 | 3291 | struct drbd_device *device; |
3b98c0c2 | 3292 | enum drbd_ret_code retcode; |
b411b363 PR |
3293 | int skip_initial_sync = 0; |
3294 | int err; | |
3b98c0c2 | 3295 | struct new_c_uuid_parms args; |
b411b363 | 3296 | |
a910b123 | 3297 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3298 | if (!adm_ctx.reply_skb) |
3299 | return retcode; | |
3300 | if (retcode != NO_ERROR) | |
3301 | goto out_nolock; | |
b411b363 | 3302 | |
b30ab791 | 3303 | device = adm_ctx.device; |
3b98c0c2 LE |
3304 | memset(&args, 0, sizeof(args)); |
3305 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3306 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3307 | if (err) { |
3308 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3309 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3310 | goto out_nolock; |
3311 | } | |
b411b363 PR |
3312 | } |
3313 | ||
9e276872 | 3314 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3315 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3316 | |
b30ab791 | 3317 | if (!get_ldev(device)) { |
b411b363 PR |
3318 | retcode = ERR_NO_DISK; |
3319 | goto out; | |
3320 | } | |
3321 | ||
3322 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3323 | if (device->state.conn == C_CONNECTED && |
3324 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3325 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3326 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3327 | skip_initial_sync = 1; |
b30ab791 | 3328 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3329 | retcode = ERR_CONNECTED; |
3330 | goto out_dec; | |
3331 | } | |
3332 | ||
b30ab791 AG |
3333 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3334 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3335 | |
3336 | if (args.clear_bm) { | |
b30ab791 | 3337 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3338 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3339 | if (err) { |
d0180171 | 3340 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3341 | retcode = ERR_IO_MD_DISK; |
3342 | } | |
3343 | if (skip_initial_sync) { | |
69a22773 | 3344 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
3345 | _drbd_uuid_set(device, UI_BITMAP, 0); |
3346 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3347 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3348 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3349 | CS_VERBOSE, NULL); |
0500813f | 3350 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3351 | } |
3352 | } | |
3353 | ||
b30ab791 | 3354 | drbd_md_sync(device); |
b411b363 | 3355 | out_dec: |
b30ab791 | 3356 | put_ldev(device); |
b411b363 | 3357 | out: |
b30ab791 | 3358 | mutex_unlock(device->state_mutex); |
9e276872 | 3359 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3360 | out_nolock: |
a910b123 | 3361 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3362 | return 0; |
3363 | } | |
3364 | ||
3b98c0c2 | 3365 | static enum drbd_ret_code |
a910b123 | 3366 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 3367 | { |
a910b123 | 3368 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 3369 | if (!name || !name[0]) { |
a910b123 | 3370 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 3371 | return ERR_MANDATORY_TAG; |
b411b363 | 3372 | } |
3b98c0c2 LE |
3373 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3374 | * we must not allow slashes */ | |
3375 | if (strchr(name, '/')) { | |
a910b123 | 3376 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 3377 | return ERR_INVALID_REQUEST; |
b411b363 | 3378 | } |
3b98c0c2 | 3379 | return NO_ERROR; |
774b3055 | 3380 | } |
b411b363 | 3381 | |
789c1b62 | 3382 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3383 | { |
a910b123 | 3384 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3385 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3386 | struct res_opts res_opts; |
3387 | int err; | |
b411b363 | 3388 | |
a910b123 | 3389 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
3390 | if (!adm_ctx.reply_skb) |
3391 | return retcode; | |
3392 | if (retcode != NO_ERROR) | |
3393 | goto out; | |
b411b363 | 3394 | |
afbbfa88 AG |
3395 | set_res_opts_defaults(&res_opts); |
3396 | err = res_opts_from_attrs(&res_opts, info); | |
3397 | if (err && err != -ENOMSG) { | |
3398 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3399 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 3400 | goto out; |
b411b363 PR |
3401 | } |
3402 | ||
a910b123 | 3403 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
3404 | if (retcode != NO_ERROR) |
3405 | goto out; | |
b411b363 | 3406 | |
5c661042 | 3407 | if (adm_ctx.resource) { |
38f19616 LE |
3408 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3409 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 3410 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
3411 | } |
3412 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3413 | goto out; |
b411b363 | 3414 | } |
b411b363 | 3415 | |
9e276872 | 3416 | /* not yet safe for genl_family.parallel_ops */ |
afbbfa88 | 3417 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3418 | retcode = ERR_NOMEM; |
3b98c0c2 | 3419 | out: |
a910b123 | 3420 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3421 | return 0; |
b411b363 PR |
3422 | } |
3423 | ||
05a10ec7 | 3424 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3425 | { |
a910b123 | 3426 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3427 | struct drbd_genlmsghdr *dh = info->userhdr; |
3428 | enum drbd_ret_code retcode; | |
b411b363 | 3429 | |
a910b123 | 3430 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3431 | if (!adm_ctx.reply_skb) |
3432 | return retcode; | |
3433 | if (retcode != NO_ERROR) | |
3434 | goto out; | |
b411b363 | 3435 | |
f2257a56 | 3436 | if (dh->minor > MINORMASK) { |
a910b123 | 3437 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
3438 | retcode = ERR_INVALID_REQUEST; |
3439 | goto out; | |
b411b363 | 3440 | } |
0c8e36d9 | 3441 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 3442 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
3443 | retcode = ERR_INVALID_REQUEST; |
3444 | goto out; | |
b411b363 | 3445 | } |
b411b363 | 3446 | |
38f19616 | 3447 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3448 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3449 | if (adm_ctx.device) { |
38f19616 LE |
3450 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
3451 | retcode = ERR_MINOR_EXISTS; | |
3452 | /* else: still NO_ERROR */ | |
3453 | goto out; | |
b411b363 | 3454 | } |
38f19616 | 3455 | |
9e276872 | 3456 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 3457 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
9e276872 | 3458 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3459 | out: |
a910b123 | 3460 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3461 | return 0; |
b411b363 PR |
3462 | } |
3463 | ||
05a10ec7 | 3464 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3465 | { |
b30ab791 AG |
3466 | if (device->state.disk == D_DISKLESS && |
3467 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3468 | * we may want to delete a minor from a live replication group. |
3469 | */ | |
b30ab791 AG |
3470 | device->state.role == R_SECONDARY) { |
3471 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3472 | CS_VERBOSE + CS_WAIT_COMPLETE); |
f82795d6 | 3473 | drbd_delete_device(device); |
85f75dd7 LE |
3474 | return NO_ERROR; |
3475 | } else | |
3476 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3477 | } |
3478 | ||
05a10ec7 | 3479 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3480 | { |
a910b123 | 3481 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3482 | enum drbd_ret_code retcode; |
b411b363 | 3483 | |
a910b123 | 3484 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3485 | if (!adm_ctx.reply_skb) |
3486 | return retcode; | |
3487 | if (retcode != NO_ERROR) | |
3488 | goto out; | |
b411b363 | 3489 | |
9e276872 | 3490 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 3491 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 3492 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 3493 | out: |
a910b123 | 3494 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 3495 | return 0; |
b411b363 PR |
3496 | } |
3497 | ||
85f75dd7 | 3498 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3499 | { |
a910b123 | 3500 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
3501 | struct drbd_resource *resource; |
3502 | struct drbd_connection *connection; | |
3503 | struct drbd_device *device; | |
f3dfa40a | 3504 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3505 | unsigned i; |
b411b363 | 3506 | |
a910b123 | 3507 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3508 | if (!adm_ctx.reply_skb) |
3509 | return retcode; | |
3510 | if (retcode != NO_ERROR) | |
9e276872 | 3511 | goto finish; |
b411b363 | 3512 | |
b6f85ef9 | 3513 | resource = adm_ctx.resource; |
9e276872 | 3514 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 3515 | /* demote */ |
b6f85ef9 AG |
3516 | for_each_connection(connection, resource) { |
3517 | struct drbd_peer_device *peer_device; | |
3518 | ||
3519 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3520 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3521 | if (retcode < SS_SUCCESS) { | |
a910b123 | 3522 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
3523 | goto out; |
3524 | } | |
3525 | } | |
3526 | ||
3527 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3528 | if (retcode < SS_SUCCESS) { |
a910b123 | 3529 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 3530 | goto out; |
85f75dd7 | 3531 | } |
b411b363 | 3532 | } |
b411b363 | 3533 | |
85f75dd7 | 3534 | /* detach */ |
b6f85ef9 AG |
3535 | idr_for_each_entry(&resource->devices, device, i) { |
3536 | retcode = adm_detach(device, 0); | |
27012382 | 3537 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 3538 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 3539 | goto out; |
85f75dd7 LE |
3540 | } |
3541 | } | |
b411b363 | 3542 | |
bde89a9e | 3543 | /* If we reach this, all volumes (of this connection) are Secondary, |
f3dfa40a | 3544 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have |
c141ebda | 3545 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
b6f85ef9 AG |
3546 | for_each_connection(connection, resource) |
3547 | drbd_thread_stop(&connection->worker); | |
b411b363 | 3548 | |
f3dfa40a | 3549 | /* Now, nothing can fail anymore */ |
b411b363 | 3550 | |
85f75dd7 | 3551 | /* delete volumes */ |
b6f85ef9 AG |
3552 | idr_for_each_entry(&resource->devices, device, i) { |
3553 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3554 | if (retcode != NO_ERROR) { |
3555 | /* "can not happen" */ | |
a910b123 | 3556 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 3557 | goto out; |
85f75dd7 LE |
3558 | } |
3559 | } | |
b411b363 | 3560 | |
b6f85ef9 AG |
3561 | list_del_rcu(&resource->resources); |
3562 | synchronize_rcu(); | |
3563 | drbd_free_resource(resource); | |
3564 | retcode = NO_ERROR; | |
3b98c0c2 | 3565 | out: |
9e276872 LE |
3566 | mutex_unlock(&resource->adm_mutex); |
3567 | finish: | |
a910b123 | 3568 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3569 | return 0; |
b411b363 PR |
3570 | } |
3571 | ||
789c1b62 | 3572 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3573 | { |
a910b123 | 3574 | struct drbd_config_context adm_ctx; |
77c556f6 AG |
3575 | struct drbd_resource *resource; |
3576 | struct drbd_connection *connection; | |
3b98c0c2 | 3577 | enum drbd_ret_code retcode; |
b411b363 | 3578 | |
a910b123 | 3579 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3580 | if (!adm_ctx.reply_skb) |
3581 | return retcode; | |
3582 | if (retcode != NO_ERROR) | |
9e276872 | 3583 | goto finish; |
b411b363 | 3584 | |
77c556f6 | 3585 | resource = adm_ctx.resource; |
9e276872 | 3586 | mutex_lock(&resource->adm_mutex); |
77c556f6 AG |
3587 | for_each_connection(connection, resource) { |
3588 | if (connection->cstate > C_STANDALONE) { | |
3589 | retcode = ERR_NET_CONFIGURED; | |
3590 | goto out; | |
3591 | } | |
3592 | } | |
3593 | if (!idr_is_empty(&resource->devices)) { | |
789c1b62 | 3594 | retcode = ERR_RES_IN_USE; |
77c556f6 | 3595 | goto out; |
b411b363 PR |
3596 | } |
3597 | ||
77c556f6 AG |
3598 | list_del_rcu(&resource->resources); |
3599 | for_each_connection(connection, resource) | |
3600 | drbd_thread_stop(&connection->worker); | |
3601 | synchronize_rcu(); | |
3602 | drbd_free_resource(resource); | |
3603 | retcode = NO_ERROR; | |
3b98c0c2 | 3604 | out: |
9e276872 LE |
3605 | mutex_unlock(&resource->adm_mutex); |
3606 | finish: | |
a910b123 | 3607 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3608 | return 0; |
3609 | } | |
3610 | ||
b30ab791 | 3611 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3612 | { |
3b98c0c2 LE |
3613 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3614 | struct sk_buff *msg; | |
3615 | struct drbd_genlmsghdr *d_out; | |
3616 | unsigned seq; | |
3617 | int err = -ENOMEM; | |
3618 | ||
ef86b779 | 3619 | if (sib->sib_reason == SIB_SYNC_PROGRESS) { |
b30ab791 AG |
3620 | if (time_after(jiffies, device->rs_last_bcast + HZ)) |
3621 | device->rs_last_bcast = jiffies; | |
ef86b779 PR |
3622 | else |
3623 | return; | |
3624 | } | |
b411b363 | 3625 | |
3b98c0c2 LE |
3626 | seq = atomic_inc_return(&drbd_genl_seq); |
3627 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3628 | if (!msg) | |
3629 | goto failed; | |
3630 | ||
3631 | err = -EMSGSIZE; | |
3632 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3633 | if (!d_out) /* cannot happen, but anyways. */ | |
3634 | goto nla_put_failure; | |
b30ab791 | 3635 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3636 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3637 | |
b30ab791 | 3638 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3639 | goto nla_put_failure; |
3640 | genlmsg_end(msg, d_out); | |
3641 | err = drbd_genl_multicast_events(msg, 0); | |
3642 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3643 | if (err && err != -ESRCH) | |
3644 | goto failed; | |
b411b363 | 3645 | |
3b98c0c2 | 3646 | return; |
b411b363 | 3647 | |
3b98c0c2 LE |
3648 | nla_put_failure: |
3649 | nlmsg_free(msg); | |
3650 | failed: | |
d0180171 | 3651 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3652 | "Event seq:%u sib_reason:%u\n", |
3653 | err, seq, sib->sib_reason); | |
b411b363 | 3654 | } |