]>
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 | ||
f88c5d90 LE |
26 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
27 | ||
b411b363 PR |
28 | #include <linux/module.h> |
29 | #include <linux/drbd.h> | |
30 | #include <linux/in.h> | |
31 | #include <linux/fs.h> | |
32 | #include <linux/file.h> | |
33 | #include <linux/slab.h> | |
b411b363 PR |
34 | #include <linux/blkpg.h> |
35 | #include <linux/cpumask.h> | |
36 | #include "drbd_int.h" | |
a3603a6e | 37 | #include "drbd_protocol.h" |
265be2d0 | 38 | #include "drbd_req.h" |
b411b363 | 39 | #include <asm/unaligned.h> |
b411b363 | 40 | #include <linux/drbd_limits.h> |
87f7be4c | 41 | #include <linux/kthread.h> |
b411b363 | 42 | |
3b98c0c2 LE |
43 | #include <net/genetlink.h> |
44 | ||
45 | /* .doit */ | |
46 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
47 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
48 | ||
05a10ec7 AG |
49 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
50 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 51 | |
789c1b62 AG |
52 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
53 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 54 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
55 | |
56 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
57 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 58 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
59 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
60 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 61 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
62 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
63 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
72 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 73 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
74 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
75 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
76 | /* .dumpit */ | |
77 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
78 | ||
79 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 80 | #include "drbd_nla.h" |
3b98c0c2 LE |
81 | #include <linux/genl_magic_func.h> |
82 | ||
83 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
84 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
85 | ||
3b98c0c2 LE |
86 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) |
87 | { | |
88 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
89 | if (genlmsg_reply(skb, info)) | |
f88c5d90 | 90 | pr_err("error sending genl reply\n"); |
b411b363 | 91 | } |
3b98c0c2 LE |
92 | |
93 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
94 | * reason it could fail was no space in skb, and there are 4k available. */ | |
a910b123 | 95 | int drbd_msg_put_info(struct sk_buff *skb, const char *info) |
3b98c0c2 | 96 | { |
3b98c0c2 LE |
97 | struct nlattr *nla; |
98 | int err = -EMSGSIZE; | |
99 | ||
100 | if (!info || !info[0]) | |
101 | return 0; | |
102 | ||
103 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
104 | if (!nla) | |
105 | return err; | |
106 | ||
107 | err = nla_put_string(skb, T_info_text, info); | |
108 | if (err) { | |
109 | nla_nest_cancel(skb, nla); | |
110 | return err; | |
111 | } else | |
112 | nla_nest_end(skb, nla); | |
113 | return 0; | |
b411b363 PR |
114 | } |
115 | ||
3b98c0c2 LE |
116 | /* This would be a good candidate for a "pre_doit" hook, |
117 | * and per-family private info->pointers. | |
118 | * But we need to stay compatible with older kernels. | |
119 | * If it returns successfully, adm_ctx members are valid. | |
9e276872 LE |
120 | * |
121 | * At this point, we still rely on the global genl_lock(). | |
122 | * If we want to avoid that, and allow "genl_family.parallel_ops", we may need | |
123 | * to add additional synchronization against object destruction/modification. | |
3b98c0c2 LE |
124 | */ |
125 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 126 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 127 | #define DRBD_ADM_NEED_CONNECTION 4 |
a910b123 LE |
128 | static int drbd_adm_prepare(struct drbd_config_context *adm_ctx, |
129 | struct sk_buff *skb, struct genl_info *info, unsigned flags) | |
3b98c0c2 LE |
130 | { |
131 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
132 | const u8 cmd = info->genlhdr->cmd; | |
133 | int err; | |
134 | ||
a910b123 | 135 | memset(adm_ctx, 0, sizeof(*adm_ctx)); |
3b98c0c2 LE |
136 | |
137 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 138 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
139 | return -EPERM; |
140 | ||
a910b123 LE |
141 | adm_ctx->reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); |
142 | if (!adm_ctx->reply_skb) { | |
1e2a2551 | 143 | err = -ENOMEM; |
3b98c0c2 | 144 | goto fail; |
1e2a2551 | 145 | } |
3b98c0c2 | 146 | |
a910b123 | 147 | adm_ctx->reply_dh = genlmsg_put_reply(adm_ctx->reply_skb, |
3b98c0c2 LE |
148 | info, &drbd_genl_family, 0, cmd); |
149 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
150 | * but anyways */ | |
a910b123 | 151 | if (!adm_ctx->reply_dh) { |
1e2a2551 | 152 | err = -ENOMEM; |
3b98c0c2 | 153 | goto fail; |
1e2a2551 | 154 | } |
3b98c0c2 | 155 | |
a910b123 LE |
156 | adm_ctx->reply_dh->minor = d_in->minor; |
157 | adm_ctx->reply_dh->ret_code = NO_ERROR; | |
3b98c0c2 | 158 | |
a910b123 | 159 | adm_ctx->volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
160 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
161 | struct nlattr *nla; | |
162 | /* parse and validate only */ | |
f399002e | 163 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
164 | if (err) |
165 | goto fail; | |
166 | ||
167 | /* It was present, and valid, | |
168 | * copy it over to the reply skb. */ | |
a910b123 | 169 | err = nla_put_nohdr(adm_ctx->reply_skb, |
3b98c0c2 LE |
170 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, |
171 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
172 | if (err) | |
173 | goto fail; | |
174 | ||
9e276872 | 175 | /* and assign stuff to the adm_ctx */ |
3b98c0c2 | 176 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; |
089c075d | 177 | if (nla) |
a910b123 | 178 | adm_ctx->volume = nla_get_u32(nla); |
7c3063cc | 179 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 180 | if (nla) |
a910b123 LE |
181 | adm_ctx->resource_name = nla_data(nla); |
182 | adm_ctx->my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; | |
183 | adm_ctx->peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
184 | if ((adm_ctx->my_addr && | |
185 | nla_len(adm_ctx->my_addr) > sizeof(adm_ctx->connection->my_addr)) || | |
186 | (adm_ctx->peer_addr && | |
187 | nla_len(adm_ctx->peer_addr) > sizeof(adm_ctx->connection->peer_addr))) { | |
089c075d AG |
188 | err = -EINVAL; |
189 | goto fail; | |
190 | } | |
191 | } | |
3b98c0c2 | 192 | |
a910b123 LE |
193 | adm_ctx->minor = d_in->minor; |
194 | adm_ctx->device = minor_to_device(d_in->minor); | |
9e276872 LE |
195 | |
196 | /* We are protected by the global genl_lock(). | |
197 | * But we may explicitly drop it/retake it in drbd_adm_set_role(), | |
198 | * so make sure this object stays around. */ | |
199 | if (adm_ctx->device) | |
200 | kref_get(&adm_ctx->device->kref); | |
201 | ||
a910b123 LE |
202 | if (adm_ctx->resource_name) { |
203 | adm_ctx->resource = drbd_find_resource(adm_ctx->resource_name); | |
4bc76048 | 204 | } |
3b98c0c2 | 205 | |
a910b123 LE |
206 | if (!adm_ctx->device && (flags & DRBD_ADM_NEED_MINOR)) { |
207 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown minor"); | |
3b98c0c2 LE |
208 | return ERR_MINOR_INVALID; |
209 | } | |
a910b123 LE |
210 | if (!adm_ctx->resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
211 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown resource"); | |
212 | if (adm_ctx->resource_name) | |
a10f6b8a | 213 | return ERR_RES_NOT_KNOWN; |
3b98c0c2 LE |
214 | return ERR_INVALID_REQUEST; |
215 | } | |
216 | ||
089c075d | 217 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
a910b123 LE |
218 | if (adm_ctx->resource) { |
219 | drbd_msg_put_info(adm_ctx->reply_skb, "no resource name expected"); | |
089c075d AG |
220 | return ERR_INVALID_REQUEST; |
221 | } | |
a910b123 LE |
222 | if (adm_ctx->device) { |
223 | drbd_msg_put_info(adm_ctx->reply_skb, "no minor number expected"); | |
089c075d AG |
224 | return ERR_INVALID_REQUEST; |
225 | } | |
a910b123 LE |
226 | if (adm_ctx->my_addr && adm_ctx->peer_addr) |
227 | adm_ctx->connection = conn_get_by_addrs(nla_data(adm_ctx->my_addr), | |
228 | nla_len(adm_ctx->my_addr), | |
229 | nla_data(adm_ctx->peer_addr), | |
230 | nla_len(adm_ctx->peer_addr)); | |
231 | if (!adm_ctx->connection) { | |
232 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown connection"); | |
089c075d AG |
233 | return ERR_INVALID_REQUEST; |
234 | } | |
235 | } | |
236 | ||
3b98c0c2 | 237 | /* some more paranoia, if the request was over-determined */ |
a910b123 LE |
238 | if (adm_ctx->device && adm_ctx->resource && |
239 | adm_ctx->device->resource != adm_ctx->resource) { | |
4bc76048 | 240 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", |
a910b123 LE |
241 | adm_ctx->minor, adm_ctx->resource->name, |
242 | adm_ctx->device->resource->name); | |
243 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists in different resource"); | |
527f4b24 LE |
244 | return ERR_INVALID_REQUEST; |
245 | } | |
a910b123 LE |
246 | if (adm_ctx->device && |
247 | adm_ctx->volume != VOLUME_UNSPECIFIED && | |
248 | adm_ctx->volume != adm_ctx->device->vnr) { | |
3b98c0c2 | 249 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
a910b123 LE |
250 | adm_ctx->minor, adm_ctx->volume, |
251 | adm_ctx->device->vnr, | |
252 | adm_ctx->device->resource->name); | |
253 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists as different volume"); | |
3b98c0c2 LE |
254 | return ERR_INVALID_REQUEST; |
255 | } | |
0ace9dfa | 256 | |
9e276872 LE |
257 | /* still, provide adm_ctx->resource always, if possible. */ |
258 | if (!adm_ctx->resource) { | |
259 | adm_ctx->resource = adm_ctx->device ? adm_ctx->device->resource | |
260 | : adm_ctx->connection ? adm_ctx->connection->resource : NULL; | |
261 | if (adm_ctx->resource) | |
262 | kref_get(&adm_ctx->resource->kref); | |
263 | } | |
264 | ||
3b98c0c2 LE |
265 | return NO_ERROR; |
266 | ||
267 | fail: | |
a910b123 LE |
268 | nlmsg_free(adm_ctx->reply_skb); |
269 | adm_ctx->reply_skb = NULL; | |
1e2a2551 | 270 | return err; |
3b98c0c2 LE |
271 | } |
272 | ||
a910b123 LE |
273 | static int drbd_adm_finish(struct drbd_config_context *adm_ctx, |
274 | struct genl_info *info, int retcode) | |
3b98c0c2 | 275 | { |
9e276872 LE |
276 | if (adm_ctx->device) { |
277 | kref_put(&adm_ctx->device->kref, drbd_destroy_device); | |
278 | adm_ctx->device = NULL; | |
279 | } | |
a910b123 LE |
280 | if (adm_ctx->connection) { |
281 | kref_put(&adm_ctx->connection->kref, &drbd_destroy_connection); | |
282 | adm_ctx->connection = NULL; | |
0ace9dfa | 283 | } |
a910b123 LE |
284 | if (adm_ctx->resource) { |
285 | kref_put(&adm_ctx->resource->kref, drbd_destroy_resource); | |
286 | adm_ctx->resource = NULL; | |
4bc76048 | 287 | } |
0ace9dfa | 288 | |
a910b123 | 289 | if (!adm_ctx->reply_skb) |
3b98c0c2 LE |
290 | return -ENOMEM; |
291 | ||
a910b123 LE |
292 | adm_ctx->reply_dh->ret_code = retcode; |
293 | drbd_adm_send_reply(adm_ctx->reply_skb, info); | |
3b98c0c2 LE |
294 | return 0; |
295 | } | |
b411b363 | 296 | |
bde89a9e | 297 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 298 | { |
6b75dced | 299 | char *afs; |
b411b363 | 300 | |
089c075d | 301 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 302 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
303 | return; |
304 | ||
bde89a9e | 305 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
306 | case AF_INET6: |
307 | afs = "ipv6"; | |
308 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 309 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 310 | break; |
089c075d AG |
311 | case AF_INET: |
312 | afs = "ipv4"; | |
313 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 314 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 315 | break; |
089c075d AG |
316 | default: |
317 | afs = "ssocks"; | |
318 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 319 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 320 | } |
089c075d | 321 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 322 | } |
b411b363 | 323 | |
b30ab791 | 324 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
325 | { |
326 | char *envp[] = { "HOME=/", | |
327 | "TERM=linux", | |
328 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
329 | (char[20]) { }, /* address family */ |
330 | (char[60]) { }, /* address */ | |
b411b363 | 331 | NULL }; |
6b75dced | 332 | char mb[12]; |
b411b363 | 333 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 334 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 335 | struct sib_info sib; |
b411b363 PR |
336 | int ret; |
337 | ||
bde89a9e AG |
338 | if (current == connection->worker.task) |
339 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 340 | |
b30ab791 | 341 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 342 | setup_khelper_env(connection, envp); |
b411b363 | 343 | |
1090c056 LE |
344 | /* The helper may take some time. |
345 | * write out any unsynced meta data changes now */ | |
b30ab791 | 346 | drbd_md_sync(device); |
1090c056 | 347 | |
d0180171 | 348 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
349 | sib.sib_reason = SIB_HELPER_PRE; |
350 | sib.helper_name = cmd; | |
b30ab791 | 351 | drbd_bcast_event(device, &sib); |
70834d30 | 352 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 353 | if (ret) |
d0180171 | 354 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
355 | usermode_helper, cmd, mb, |
356 | (ret >> 8) & 0xff, ret); | |
357 | else | |
d0180171 | 358 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
359 | usermode_helper, cmd, mb, |
360 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
361 | sib.sib_reason = SIB_HELPER_POST; |
362 | sib.helper_exit_code = ret; | |
b30ab791 | 363 | drbd_bcast_event(device, &sib); |
b411b363 | 364 | |
bde89a9e AG |
365 | if (current == connection->worker.task) |
366 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
367 | |
368 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
369 | ret = 0; | |
370 | ||
371 | return ret; | |
372 | } | |
373 | ||
bde89a9e | 374 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
375 | { |
376 | char *envp[] = { "HOME=/", | |
377 | "TERM=linux", | |
378 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
379 | (char[20]) { }, /* address family */ | |
380 | (char[60]) { }, /* address */ | |
381 | NULL }; | |
77c556f6 AG |
382 | char *resource_name = connection->resource->name; |
383 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
384 | int ret; |
385 | ||
bde89a9e AG |
386 | setup_khelper_env(connection, envp); |
387 | conn_md_sync(connection); | |
6b75dced | 388 | |
1ec861eb | 389 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced PR |
390 | /* TODO: conn_bcast_event() ?? */ |
391 | ||
98683650 | 392 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 393 | if (ret) |
1ec861eb | 394 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 395 | usermode_helper, cmd, resource_name, |
6b75dced PR |
396 | (ret >> 8) & 0xff, ret); |
397 | else | |
1ec861eb | 398 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 399 | usermode_helper, cmd, resource_name, |
6b75dced PR |
400 | (ret >> 8) & 0xff, ret); |
401 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 402 | |
b411b363 PR |
403 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
404 | ret = 0; | |
405 | ||
406 | return ret; | |
407 | } | |
408 | ||
bde89a9e | 409 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 410 | { |
cb703454 | 411 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 412 | struct drbd_peer_device *peer_device; |
cb703454 PR |
413 | int vnr; |
414 | ||
695d08fa | 415 | rcu_read_lock(); |
c06ece6b AG |
416 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
417 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 418 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
419 | struct disk_conf *disk_conf = |
420 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
421 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 422 | put_ldev(device); |
cb703454 PR |
423 | } |
424 | } | |
695d08fa | 425 | rcu_read_unlock(); |
cb703454 | 426 | |
d7fe69c6 PR |
427 | if (fp == FP_NOT_AVAIL) { |
428 | /* IO Suspending works on the whole resource. | |
429 | Do it only for one device. */ | |
430 | vnr = 0; | |
431 | peer_device = idr_get_next(&connection->peer_devices, &vnr); | |
432 | drbd_change_state(peer_device->device, CS_VERBOSE | CS_HARD, NS(susp_fen, 0)); | |
433 | } | |
434 | ||
cb703454 PR |
435 | return fp; |
436 | } | |
437 | ||
bde89a9e | 438 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 439 | { |
28e448bb | 440 | unsigned int connect_cnt; |
cb703454 PR |
441 | union drbd_state mask = { }; |
442 | union drbd_state val = { }; | |
443 | enum drbd_fencing_p fp; | |
b411b363 PR |
444 | char *ex_to_string; |
445 | int r; | |
b411b363 | 446 | |
02df6fe1 | 447 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 448 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 449 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
02df6fe1 | 450 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 PR |
451 | return false; |
452 | } | |
b411b363 | 453 | |
bde89a9e | 454 | connect_cnt = connection->connect_cnt; |
0500813f | 455 | spin_unlock_irq(&connection->resource->req_lock); |
28e448bb | 456 | |
bde89a9e | 457 | fp = highest_fencing_policy(connection); |
cb703454 PR |
458 | switch (fp) { |
459 | case FP_NOT_AVAIL: | |
1ec861eb | 460 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 461 | goto out; |
cb703454 PR |
462 | case FP_DONT_CARE: |
463 | return true; | |
464 | default: ; | |
b411b363 PR |
465 | } |
466 | ||
bde89a9e | 467 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
468 | |
469 | switch ((r>>8) & 0xff) { | |
470 | case 3: /* peer is inconsistent */ | |
471 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
472 | mask.pdsk = D_MASK; |
473 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
474 | break; |
475 | case 4: /* peer got outdated, or was already outdated */ | |
476 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
477 | mask.pdsk = D_MASK; |
478 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
479 | break; |
480 | case 5: /* peer was down */ | |
bde89a9e | 481 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
482 | /* we will(have) create(d) a new UUID anyways... */ |
483 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
484 | mask.pdsk = D_MASK; |
485 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
486 | } else { |
487 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
488 | } |
489 | break; | |
490 | case 6: /* Peer is primary, voluntarily outdate myself. | |
491 | * This is useful when an unconnected R_SECONDARY is asked to | |
492 | * become R_PRIMARY, but finds the other peer being active. */ | |
493 | ex_to_string = "peer is active"; | |
1ec861eb | 494 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
495 | mask.disk = D_MASK; |
496 | val.disk = D_OUTDATED; | |
b411b363 PR |
497 | break; |
498 | case 7: | |
499 | if (fp != FP_STONITH) | |
1ec861eb | 500 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 501 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
502 | mask.pdsk = D_MASK; |
503 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
504 | break; |
505 | default: | |
506 | /* The script is broken ... */ | |
1ec861eb | 507 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 508 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
509 | } |
510 | ||
1ec861eb | 511 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 512 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 513 | |
cb703454 | 514 | out: |
fb22c402 | 515 | |
cb703454 | 516 | /* Not using |
bde89a9e | 517 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
518 | here, because we might were able to re-establish the connection in the |
519 | meantime. */ | |
0500813f | 520 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
521 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
522 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
523 | /* In case the connection was established and droped |
524 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 525 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 526 | else |
bde89a9e | 527 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 528 | } |
0500813f | 529 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 | 530 | |
bde89a9e | 531 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
532 | } |
533 | ||
87f7be4c PR |
534 | static int _try_outdate_peer_async(void *data) |
535 | { | |
bde89a9e | 536 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 537 | |
bde89a9e | 538 | conn_try_outdate_peer(connection); |
87f7be4c | 539 | |
05a10ec7 | 540 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
541 | return 0; |
542 | } | |
543 | ||
bde89a9e | 544 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
545 | { |
546 | struct task_struct *opa; | |
547 | ||
bde89a9e AG |
548 | kref_get(&connection->kref); |
549 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); | |
9dc9fbb3 | 550 | if (IS_ERR(opa)) { |
1ec861eb | 551 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 552 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 553 | } |
87f7be4c | 554 | } |
b411b363 | 555 | |
bf885f8a | 556 | enum drbd_state_rv |
44a4d551 | 557 | drbd_set_role(struct drbd_device *const device, enum drbd_role new_role, int force) |
b411b363 | 558 | { |
44a4d551 LE |
559 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
560 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; | |
b411b363 | 561 | const int max_tries = 4; |
bf885f8a | 562 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 563 | struct net_conf *nc; |
b411b363 PR |
564 | int try = 0; |
565 | int forced = 0; | |
566 | union drbd_state mask, val; | |
b411b363 | 567 | |
b6f85ef9 AG |
568 | if (new_role == R_PRIMARY) { |
569 | struct drbd_connection *connection; | |
570 | ||
571 | /* Detect dead peers as soon as possible. */ | |
572 | ||
573 | rcu_read_lock(); | |
574 | for_each_connection(connection, device->resource) | |
575 | request_ping(connection); | |
576 | rcu_read_unlock(); | |
577 | } | |
b411b363 | 578 | |
b30ab791 | 579 | mutex_lock(device->state_mutex); |
b411b363 PR |
580 | |
581 | mask.i = 0; mask.role = R_MASK; | |
582 | val.i = 0; val.role = new_role; | |
583 | ||
584 | while (try++ < max_tries) { | |
b30ab791 | 585 | rv = _drbd_request_state(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
586 | |
587 | /* in case we first succeeded to outdate, | |
588 | * but now suddenly could establish a connection */ | |
bf885f8a | 589 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
590 | val.pdsk = 0; |
591 | mask.pdsk = 0; | |
592 | continue; | |
593 | } | |
594 | ||
bf885f8a | 595 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
596 | (device->state.disk < D_UP_TO_DATE && |
597 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
598 | mask.disk = D_MASK; |
599 | val.disk = D_UP_TO_DATE; | |
600 | forced = 1; | |
601 | continue; | |
602 | } | |
603 | ||
bf885f8a | 604 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 605 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 606 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 607 | |
44a4d551 | 608 | if (conn_try_outdate_peer(connection)) { |
b411b363 PR |
609 | val.disk = D_UP_TO_DATE; |
610 | mask.disk = D_MASK; | |
611 | } | |
b411b363 PR |
612 | continue; |
613 | } | |
614 | ||
bf885f8a | 615 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 616 | goto out; |
bf885f8a | 617 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
44a4d551 | 618 | if (!conn_try_outdate_peer(connection) && force) { |
d0180171 | 619 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
620 | mask.pdsk = D_MASK; |
621 | val.pdsk = D_OUTDATED; | |
b411b363 | 622 | |
cb703454 | 623 | } |
b411b363 PR |
624 | continue; |
625 | } | |
bf885f8a | 626 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
627 | /* Maybe the peer is detected as dead very soon... |
628 | retry at most once more in this case. */ | |
44ed167d PR |
629 | int timeo; |
630 | rcu_read_lock(); | |
44a4d551 | 631 | nc = rcu_dereference(connection->net_conf); |
44ed167d PR |
632 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
633 | rcu_read_unlock(); | |
634 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
635 | if (try < max_tries) |
636 | try = max_tries - 1; | |
637 | continue; | |
638 | } | |
bf885f8a | 639 | if (rv < SS_SUCCESS) { |
b30ab791 | 640 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 641 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 642 | if (rv < SS_SUCCESS) |
3b98c0c2 | 643 | goto out; |
b411b363 PR |
644 | } |
645 | break; | |
646 | } | |
647 | ||
bf885f8a | 648 | if (rv < SS_SUCCESS) |
3b98c0c2 | 649 | goto out; |
b411b363 PR |
650 | |
651 | if (forced) | |
d0180171 | 652 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
653 | |
654 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 655 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 656 | |
b6dd1a89 LE |
657 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
658 | ||
b411b363 | 659 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
660 | if (get_ldev(device)) { |
661 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
662 | put_ldev(device); | |
b411b363 PR |
663 | } |
664 | } else { | |
66ce6dbc | 665 | mutex_lock(&device->resource->conf_update); |
44a4d551 | 666 | nc = connection->net_conf; |
44ed167d | 667 | if (nc) |
6139f60d | 668 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
66ce6dbc | 669 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 670 | |
b30ab791 AG |
671 | if (get_ldev(device)) { |
672 | if (((device->state.conn < C_CONNECTED || | |
673 | device->state.pdsk <= D_FAILED) | |
674 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
675 | drbd_uuid_new_current(device); | |
b411b363 | 676 | |
b30ab791 AG |
677 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
678 | put_ldev(device); | |
b411b363 PR |
679 | } |
680 | } | |
681 | ||
19f843aa LE |
682 | /* writeout of activity log covered areas of the bitmap |
683 | * to stable storage done in after state change already */ | |
b411b363 | 684 | |
b30ab791 | 685 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
686 | /* if this was forced, we should consider sync */ |
687 | if (forced) | |
44a4d551 LE |
688 | drbd_send_uuids(peer_device); |
689 | drbd_send_current_state(peer_device); | |
b411b363 PR |
690 | } |
691 | ||
b30ab791 | 692 | drbd_md_sync(device); |
720979fb | 693 | set_disk_ro(device->vdisk, new_role == R_SECONDARY); |
b30ab791 | 694 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 695 | out: |
b30ab791 | 696 | mutex_unlock(device->state_mutex); |
bf885f8a | 697 | return rv; |
b411b363 PR |
698 | } |
699 | ||
3b98c0c2 | 700 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 701 | { |
3b98c0c2 LE |
702 | return err == -ENOMSG ? "required attribute missing" : |
703 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 704 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 705 | "invalid attribute value"; |
ef50a3e3 | 706 | } |
b411b363 | 707 | |
3b98c0c2 | 708 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 709 | { |
a910b123 | 710 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
711 | struct set_role_parms parms; |
712 | int err; | |
713 | enum drbd_ret_code retcode; | |
b411b363 | 714 | |
a910b123 | 715 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
716 | if (!adm_ctx.reply_skb) |
717 | return retcode; | |
718 | if (retcode != NO_ERROR) | |
719 | goto out; | |
b411b363 | 720 | |
3b98c0c2 LE |
721 | memset(&parms, 0, sizeof(parms)); |
722 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 723 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
724 | if (err) { |
725 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 726 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
727 | goto out; |
728 | } | |
729 | } | |
9e276872 LE |
730 | genl_unlock(); |
731 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 732 | |
3b98c0c2 | 733 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 734 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 735 | else |
b30ab791 | 736 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
737 | |
738 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
739 | genl_lock(); | |
3b98c0c2 | 740 | out: |
a910b123 | 741 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
742 | return 0; |
743 | } | |
744 | ||
ae8bf312 LE |
745 | /* Initializes the md.*_offset members, so we are able to find |
746 | * the on disk meta data. | |
747 | * | |
748 | * We currently have two possible layouts: | |
749 | * external: | |
750 | * |----------- md_size_sect ------------------| | |
751 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
752 | * | al_offset == 8 | | |
753 | * | bm_offset = al_offset + X | | |
754 | * ==> bitmap sectors = md_size_sect - bm_offset | |
755 | * | |
756 | * internal: | |
757 | * |----------- md_size_sect ------------------| | |
758 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
759 | * | al_offset < 0 | | |
760 | * | bm_offset = al_offset - Y | | |
761 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
762 | * | |
763 | * Activity log size used to be fixed 32kB, | |
764 | * but is about to become configurable. | |
765 | */ | |
b30ab791 | 766 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
767 | struct drbd_backing_dev *bdev) |
768 | { | |
769 | sector_t md_size_sect = 0; | |
c04ccaa6 | 770 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 771 | |
3a4d4eb3 LE |
772 | bdev->md.md_offset = drbd_md_ss(bdev); |
773 | ||
68e41a43 | 774 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
775 | default: |
776 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 777 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
778 | bdev->md.al_offset = MD_4kB_SECT; |
779 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
780 | break; |
781 | case DRBD_MD_INDEX_FLEX_EXT: | |
782 | /* just occupy the full device; unit: sectors */ | |
783 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
784 | bdev->md.al_offset = MD_4kB_SECT; |
785 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
786 | break; |
787 | case DRBD_MD_INDEX_INTERNAL: | |
788 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 789 | /* al size is still fixed */ |
ae8bf312 | 790 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
791 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
792 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
793 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
794 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
795 | md_size_sect = ALIGN(md_size_sect, 8); | |
796 | ||
797 | /* plus the "drbd meta data super block", | |
798 | * and the activity log; */ | |
ae8bf312 | 799 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
800 | |
801 | bdev->md.md_size_sect = md_size_sect; | |
802 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 803 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
804 | break; |
805 | } | |
806 | } | |
807 | ||
4b0715f0 | 808 | /* input size is expected to be in KB */ |
b411b363 PR |
809 | char *ppsize(char *buf, unsigned long long size) |
810 | { | |
4b0715f0 LE |
811 | /* Needs 9 bytes at max including trailing NUL: |
812 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
813 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
814 | int base = 0; | |
4b0715f0 | 815 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
816 | /* shift + round */ |
817 | size = (size >> 10) + !!(size & (1<<9)); | |
818 | base++; | |
819 | } | |
4b0715f0 | 820 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
821 | |
822 | return buf; | |
823 | } | |
824 | ||
825 | /* there is still a theoretical deadlock when called from receiver | |
826 | * on an D_INCONSISTENT R_PRIMARY: | |
827 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
828 | * packet from remote to dec_ap_bio again. | |
829 | * receiver receive_sizes(), comes here, | |
830 | * waits for ap_bio_cnt == 0. -> deadlock. | |
831 | * but this cannot happen, actually, because: | |
832 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
833 | * (not connected, or bad/no disk on peer): | |
834 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
835 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
836 | * peer may not initiate a resize. | |
837 | */ | |
3b98c0c2 LE |
838 | /* Note these are not to be confused with |
839 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
840 | * which are (sub) state changes triggered by admin (drbdsetup), | |
841 | * and can be long lived. | |
b30ab791 | 842 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 843 | * and should be short-lived. */ |
b30ab791 | 844 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 845 | { |
b30ab791 AG |
846 | set_bit(SUSPEND_IO, &device->flags); |
847 | if (drbd_suspended(device)) | |
265be2d0 | 848 | return; |
b30ab791 | 849 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
850 | } |
851 | ||
b30ab791 | 852 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 853 | { |
b30ab791 AG |
854 | clear_bit(SUSPEND_IO, &device->flags); |
855 | wake_up(&device->misc_wait); | |
b411b363 PR |
856 | } |
857 | ||
858 | /** | |
859 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 860 | * @device: DRBD device. |
b411b363 PR |
861 | * |
862 | * Returns 0 on success, negative return values indicate errors. | |
863 | * You should call drbd_md_sync() after calling this function. | |
864 | */ | |
d752b269 | 865 | enum determine_dev_size |
b30ab791 | 866 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
867 | { |
868 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 869 | sector_t la_size_sect, u_size; |
b30ab791 | 870 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
871 | u32 prev_al_stripe_size_4k; |
872 | u32 prev_al_stripes; | |
b411b363 PR |
873 | sector_t size; |
874 | char ppb[10]; | |
d752b269 | 875 | void *buffer; |
b411b363 PR |
876 | |
877 | int md_moved, la_size_changed; | |
e96c9633 | 878 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
879 | |
880 | /* race: | |
881 | * application request passes inc_ap_bio, | |
882 | * but then cannot get an AL-reference. | |
883 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
884 | * | |
885 | * to avoid that: | |
886 | * Suspend IO right here. | |
887 | * still lock the act_log to not trigger ASSERTs there. | |
888 | */ | |
b30ab791 AG |
889 | drbd_suspend_io(device); |
890 | buffer = drbd_md_get_buffer(device); /* Lock meta-data IO */ | |
d752b269 | 891 | if (!buffer) { |
b30ab791 | 892 | drbd_resume_io(device); |
d752b269 PR |
893 | return DS_ERROR; |
894 | } | |
b411b363 PR |
895 | |
896 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 897 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 898 | |
b30ab791 AG |
899 | prev_first_sect = drbd_md_first_sector(device->ldev); |
900 | prev_size = device->ldev->md.md_size_sect; | |
901 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 902 | |
d752b269 PR |
903 | if (rs) { |
904 | /* rs is non NULL if we should change the AL layout only */ | |
905 | ||
906 | prev_al_stripes = md->al_stripes; | |
907 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
908 | ||
909 | md->al_stripes = rs->al_stripes; | |
910 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
911 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
912 | } | |
913 | ||
b30ab791 | 914 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 915 | |
daeda1cc | 916 | rcu_read_lock(); |
b30ab791 | 917 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 918 | rcu_read_unlock(); |
b30ab791 | 919 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 920 | |
d752b269 PR |
921 | if (size < la_size_sect) { |
922 | if (rs && u_size == 0) { | |
923 | /* Remove "rs &&" later. This check should always be active, but | |
924 | right now the receiver expects the permissive behavior */ | |
d0180171 | 925 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
926 | "Use --size=%llus for explicit shrink.\n", |
927 | (unsigned long long)size); | |
928 | rv = DS_ERROR_SHRINK; | |
929 | } | |
930 | if (u_size > size) | |
931 | rv = DS_ERROR_SPACE_MD; | |
932 | if (rv != DS_UNCHANGED) | |
933 | goto err_out; | |
934 | } | |
935 | ||
b30ab791 AG |
936 | if (drbd_get_capacity(device->this_bdev) != size || |
937 | drbd_bm_capacity(device) != size) { | |
b411b363 | 938 | int err; |
b30ab791 | 939 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
940 | if (unlikely(err)) { |
941 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 942 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 943 | if (size == 0) { |
d0180171 | 944 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
945 | "Could not allocate bitmap!\n"); |
946 | } else { | |
d0180171 | 947 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
948 | "Leaving size unchanged at size = %lu KB\n", |
949 | (unsigned long)size); | |
950 | } | |
e96c9633 | 951 | rv = DS_ERROR; |
b411b363 PR |
952 | } |
953 | /* racy, see comments above. */ | |
b30ab791 AG |
954 | drbd_set_my_capacity(device, size); |
955 | device->ldev->md.la_size_sect = size; | |
d0180171 | 956 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
957 | (unsigned long long)size>>1); |
958 | } | |
d752b269 PR |
959 | if (rv <= DS_ERROR) |
960 | goto err_out; | |
b411b363 | 961 | |
b30ab791 | 962 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 963 | |
b30ab791 AG |
964 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
965 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 966 | |
d752b269 PR |
967 | if (la_size_changed || md_moved || rs) { |
968 | u32 prev_flags; | |
24dccabb | 969 | |
fcb09674 LE |
970 | /* We do some synchronous IO below, which may take some time. |
971 | * Clear the timer, to avoid scary "timer expired!" messages, | |
972 | * "Superblock" is written out at least twice below, anyways. */ | |
973 | del_timer(&device->md_sync_timer); | |
b30ab791 | 974 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
975 | |
976 | prev_flags = md->flags; | |
977 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 978 | drbd_md_write(device, buffer); |
d752b269 | 979 | |
d0180171 | 980 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
981 | la_size_changed && md_moved ? "size changed and md moved" : |
982 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 983 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 984 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 985 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 986 | drbd_initialize_al(device, buffer); |
d752b269 PR |
987 | |
988 | md->flags = prev_flags; | |
b30ab791 | 989 | drbd_md_write(device, buffer); |
d752b269 PR |
990 | |
991 | if (rs) | |
d0180171 AG |
992 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
993 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
994 | } |
995 | ||
cccac985 | 996 | if (size > la_size_sect) |
57737adc | 997 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 998 | if (size < la_size_sect) |
e96c9633 | 999 | rv = DS_SHRUNK; |
d752b269 PR |
1000 | |
1001 | if (0) { | |
1002 | err_out: | |
1003 | if (rs) { | |
1004 | md->al_stripes = prev_al_stripes; | |
1005 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
1006 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
1007 | ||
b30ab791 | 1008 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
1009 | } |
1010 | } | |
b30ab791 AG |
1011 | lc_unlock(device->act_log); |
1012 | wake_up(&device->al_wait); | |
1013 | drbd_md_put_buffer(device); | |
1014 | drbd_resume_io(device); | |
b411b363 PR |
1015 | |
1016 | return rv; | |
1017 | } | |
1018 | ||
1019 | sector_t | |
b30ab791 | 1020 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1021 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1022 | { |
b30ab791 | 1023 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1024 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1025 | sector_t m_size; /* my size */ |
b411b363 PR |
1026 | sector_t size = 0; |
1027 | ||
1028 | m_size = drbd_get_max_capacity(bdev); | |
1029 | ||
b30ab791 | 1030 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1031 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1032 | p_size = m_size; |
1033 | } | |
1034 | ||
b411b363 PR |
1035 | if (p_size && m_size) { |
1036 | size = min_t(sector_t, p_size, m_size); | |
1037 | } else { | |
cccac985 LE |
1038 | if (la_size_sect) { |
1039 | size = la_size_sect; | |
b411b363 PR |
1040 | if (m_size && m_size < size) |
1041 | size = m_size; | |
1042 | if (p_size && p_size < size) | |
1043 | size = p_size; | |
1044 | } else { | |
1045 | if (m_size) | |
1046 | size = m_size; | |
1047 | if (p_size) | |
1048 | size = p_size; | |
1049 | } | |
1050 | } | |
1051 | ||
1052 | if (size == 0) | |
d0180171 | 1053 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1054 | |
1055 | if (u_size) { | |
1056 | if (u_size > size) | |
d0180171 | 1057 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1058 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1059 | else | |
1060 | size = u_size; | |
1061 | } | |
1062 | ||
1063 | return size; | |
1064 | } | |
1065 | ||
1066 | /** | |
1067 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1068 | * @device: DRBD device. |
b411b363 PR |
1069 | * |
1070 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1071 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1072 | * this function. | |
1073 | */ | |
b30ab791 | 1074 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1075 | { |
1076 | struct lru_cache *n, *t; | |
1077 | struct lc_element *e; | |
1078 | unsigned int in_use; | |
1079 | int i; | |
1080 | ||
b30ab791 AG |
1081 | if (device->act_log && |
1082 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1083 | return 0; |
1084 | ||
1085 | in_use = 0; | |
b30ab791 | 1086 | t = device->act_log; |
7ad651b5 | 1087 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1088 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1089 | |
1090 | if (n == NULL) { | |
d0180171 | 1091 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1092 | return -ENOMEM; |
1093 | } | |
b30ab791 | 1094 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1095 | if (t) { |
1096 | for (i = 0; i < t->nr_elements; i++) { | |
1097 | e = lc_element_by_index(t, i); | |
1098 | if (e->refcnt) | |
d0180171 | 1099 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1100 | e->lc_number, e->refcnt); |
1101 | in_use += e->refcnt; | |
1102 | } | |
1103 | } | |
1104 | if (!in_use) | |
b30ab791 AG |
1105 | device->act_log = n; |
1106 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1107 | if (in_use) { |
d0180171 | 1108 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1109 | lc_destroy(n); |
1110 | return -EBUSY; | |
1111 | } else { | |
1112 | if (t) | |
1113 | lc_destroy(t); | |
1114 | } | |
b30ab791 | 1115 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1116 | return 0; |
1117 | } | |
1118 | ||
8fe39aac PR |
1119 | static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev, |
1120 | unsigned int max_bio_size) | |
b411b363 | 1121 | { |
b30ab791 | 1122 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1123 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1124 | unsigned int max_segments = 0; | |
c1b3156f | 1125 | struct request_queue *b = NULL; |
99432fcc | 1126 | |
8fe39aac PR |
1127 | if (bdev) { |
1128 | b = bdev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1129 | |
1130 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1131 | rcu_read_lock(); |
b30ab791 | 1132 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1133 | rcu_read_unlock(); |
c1b3156f PR |
1134 | |
1135 | blk_set_stacking_limits(&q->limits); | |
1136 | blk_queue_max_write_same_sectors(q, 0); | |
99432fcc | 1137 | } |
b411b363 | 1138 | |
b411b363 | 1139 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1140 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1141 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1142 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1143 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1144 | |
c1b3156f | 1145 | if (b) { |
20c68fde LE |
1146 | struct drbd_connection *connection = first_peer_device(device)->connection; |
1147 | ||
1148 | if (blk_queue_discard(b) && | |
1149 | (connection->cstate < C_CONNECTED || connection->agreed_features & FF_TRIM)) { | |
20c68fde LE |
1150 | /* For now, don't allow more than one activity log extent worth of data |
1151 | * to be discarded in one go. We may need to rework drbd_al_begin_io() | |
1152 | * to allow for even larger discard ranges */ | |
1153 | q->limits.max_discard_sectors = DRBD_MAX_DISCARD_SECTORS; | |
1154 | ||
1155 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1156 | /* REALLY? Is stacking secdiscard "legal"? */ | |
1157 | if (blk_queue_secdiscard(b)) | |
1158 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1159 | } else { | |
1160 | q->limits.max_discard_sectors = 0; | |
1161 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1162 | queue_flag_clear_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1163 | } | |
99432fcc PR |
1164 | |
1165 | blk_queue_stack_limits(q, b); | |
1166 | ||
1167 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1168 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1169 | q->backing_dev_info.ra_pages, |
1170 | b->backing_dev_info.ra_pages); | |
1171 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1172 | } | |
99432fcc PR |
1173 | } |
1174 | } | |
1175 | ||
8fe39aac | 1176 | void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev) |
99432fcc | 1177 | { |
db141b2f | 1178 | unsigned int now, new, local, peer; |
99432fcc | 1179 | |
b30ab791 AG |
1180 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1181 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1182 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1183 | |
8fe39aac PR |
1184 | if (bdev) { |
1185 | local = queue_max_hw_sectors(bdev->backing_bdev->bd_disk->queue) << 9; | |
b30ab791 | 1186 | device->local_max_bio_size = local; |
b411b363 | 1187 | } |
db141b2f | 1188 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1189 | |
1190 | /* We may ignore peer limits if the peer is modern enough. | |
1191 | Because new from 8.3.8 onwards the peer can use multiple | |
1192 | BIOs for a single peer_request */ | |
b30ab791 | 1193 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1194 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1195 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1196 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1197 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1198 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1199 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1200 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1201 | else | |
99432fcc | 1202 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1203 | |
fa090e70 LE |
1204 | /* We may later detach and re-attach on a disconnected Primary. |
1205 | * Avoid this setting to jump back in that case. | |
1206 | * We want to store what we know the peer DRBD can handle, | |
1207 | * not what the peer IO backend can handle. */ | |
1208 | if (peer > device->peer_max_bio_size) | |
1209 | device->peer_max_bio_size = peer; | |
1210 | } | |
db141b2f | 1211 | new = min(local, peer); |
99432fcc | 1212 | |
b30ab791 | 1213 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1214 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1215 | |
1216 | if (new != now) | |
d0180171 | 1217 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1218 | |
8fe39aac | 1219 | drbd_setup_queue_param(device, bdev, new); |
b411b363 PR |
1220 | } |
1221 | ||
a18e9d1e | 1222 | /* Starts the worker thread */ |
bde89a9e | 1223 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1224 | { |
bde89a9e | 1225 | drbd_thread_start(&connection->worker); |
b5043c5e | 1226 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1227 | } |
1228 | ||
a18e9d1e | 1229 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1230 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1231 | { |
992d6e91 | 1232 | bool stop_threads; |
0500813f | 1233 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1234 | stop_threads = conn_all_vols_unconf(connection) && |
1235 | connection->cstate == C_STANDALONE; | |
0500813f | 1236 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1237 | if (stop_threads) { |
1238 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1239 | * in conn_disconnect() */ |
bde89a9e AG |
1240 | drbd_thread_stop(&connection->receiver); |
1241 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1242 | } |
b411b363 PR |
1243 | } |
1244 | ||
0778286a | 1245 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1246 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1247 | { |
1248 | int s = 0; | |
1249 | ||
b30ab791 | 1250 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1251 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1252 | return; |
1253 | } | |
1254 | ||
b30ab791 | 1255 | drbd_al_shrink(device); |
0500813f | 1256 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1257 | if (device->state.conn < C_CONNECTED) |
1258 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1259 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1260 | lc_unlock(device->act_log); |
0778286a PR |
1261 | |
1262 | if (s) | |
d0180171 | 1263 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1264 | } |
1265 | ||
5979e361 LE |
1266 | |
1267 | static bool should_set_defaults(struct genl_info *info) | |
1268 | { | |
1269 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1270 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1271 | } | |
1272 | ||
5bbcf5e6 | 1273 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1274 | { |
5bbcf5e6 LE |
1275 | /* This is limited by 16 bit "slot" numbers, |
1276 | * and by available on-disk context storage. | |
1277 | * | |
1278 | * Also (u16)~0 is special (denotes a "free" extent). | |
1279 | * | |
1280 | * One transaction occupies one 4kB on-disk block, | |
1281 | * we have n such blocks in the on disk ring buffer, | |
1282 | * the "current" transaction may fail (n-1), | |
1283 | * and there is 919 slot numbers context information per transaction. | |
1284 | * | |
1285 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1286 | * so cap there first. | |
1287 | */ | |
1288 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1289 | const unsigned int sufficient_on_disk = | |
1290 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1291 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1292 | |
5bbcf5e6 LE |
1293 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1294 | ||
1295 | if (al_size_4k > sufficient_on_disk) | |
1296 | return max_al_nr; | |
1297 | ||
1298 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1299 | } |
1300 | ||
70df7092 LE |
1301 | static bool write_ordering_changed(struct disk_conf *a, struct disk_conf *b) |
1302 | { | |
1303 | return a->disk_barrier != b->disk_barrier || | |
1304 | a->disk_flushes != b->disk_flushes || | |
1305 | a->disk_drain != b->disk_drain; | |
1306 | } | |
1307 | ||
f399002e LE |
1308 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1309 | { | |
a910b123 | 1310 | struct drbd_config_context adm_ctx; |
f399002e | 1311 | enum drbd_ret_code retcode; |
b30ab791 | 1312 | struct drbd_device *device; |
daeda1cc | 1313 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1314 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1315 | int err, fifo_size; |
f399002e | 1316 | |
a910b123 | 1317 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1318 | if (!adm_ctx.reply_skb) |
1319 | return retcode; | |
1320 | if (retcode != NO_ERROR) | |
9e276872 | 1321 | goto finish; |
f399002e | 1322 | |
b30ab791 | 1323 | device = adm_ctx.device; |
9e276872 | 1324 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1325 | |
1326 | /* we also need a disk | |
1327 | * to change the options on */ | |
b30ab791 | 1328 | if (!get_ldev(device)) { |
f399002e LE |
1329 | retcode = ERR_NO_DISK; |
1330 | goto out; | |
1331 | } | |
1332 | ||
daeda1cc | 1333 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1334 | if (!new_disk_conf) { |
f399002e LE |
1335 | retcode = ERR_NOMEM; |
1336 | goto fail; | |
1337 | } | |
1338 | ||
0500813f | 1339 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1340 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1341 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1342 | if (should_set_defaults(info)) |
b966b5dd | 1343 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1344 | |
5ecc72c3 | 1345 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1346 | if (err && err != -ENOMSG) { |
f399002e | 1347 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1348 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1349 | goto fail_unlock; |
f399002e LE |
1350 | } |
1351 | ||
5ecc72c3 LE |
1352 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1353 | new_disk_conf->resync_rate = 1; | |
f399002e | 1354 | |
5bbcf5e6 LE |
1355 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1356 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1357 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1358 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1359 | |
1360 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1361 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1362 | |
5ecc72c3 | 1363 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1364 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1365 | new_plan = fifo_alloc(fifo_size); |
1366 | if (!new_plan) { | |
d0180171 | 1367 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1368 | retcode = ERR_NOMEM; |
daeda1cc | 1369 | goto fail_unlock; |
f399002e LE |
1370 | } |
1371 | } | |
1372 | ||
b30ab791 AG |
1373 | drbd_suspend_io(device); |
1374 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1375 | drbd_al_shrink(device); | |
1376 | err = drbd_check_al_size(device, new_disk_conf); | |
1377 | lc_unlock(device->act_log); | |
1378 | wake_up(&device->al_wait); | |
1379 | drbd_resume_io(device); | |
f399002e LE |
1380 | |
1381 | if (err) { | |
1382 | retcode = ERR_NOMEM; | |
daeda1cc | 1383 | goto fail_unlock; |
f399002e LE |
1384 | } |
1385 | ||
dc97b708 | 1386 | write_lock_irq(&global_state_lock); |
b30ab791 | 1387 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1388 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1389 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1390 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1391 | } |
1392 | write_unlock_irq(&global_state_lock); | |
f399002e | 1393 | |
daeda1cc PR |
1394 | if (retcode != NO_ERROR) |
1395 | goto fail_unlock; | |
f399002e | 1396 | |
813472ce | 1397 | if (new_plan) { |
b30ab791 AG |
1398 | old_plan = device->rs_plan_s; |
1399 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1400 | } |
9958c857 | 1401 | |
0500813f | 1402 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1403 | |
9a51ab1c | 1404 | if (new_disk_conf->al_updates) |
b30ab791 | 1405 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1406 | else |
b30ab791 | 1407 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1408 | |
691631c0 | 1409 | if (new_disk_conf->md_flushes) |
b30ab791 | 1410 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1411 | else |
b30ab791 | 1412 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1413 | |
70df7092 LE |
1414 | if (write_ordering_changed(old_disk_conf, new_disk_conf)) |
1415 | drbd_bump_write_ordering(device->resource, NULL, WO_bdev_flush); | |
27eb13e9 | 1416 | |
b30ab791 | 1417 | drbd_md_sync(device); |
f399002e | 1418 | |
69a22773 AG |
1419 | if (device->state.conn >= C_CONNECTED) { |
1420 | struct drbd_peer_device *peer_device; | |
1421 | ||
1422 | for_each_peer_device(peer_device, device) | |
1423 | drbd_send_sync_param(peer_device); | |
1424 | } | |
f399002e | 1425 | |
daeda1cc PR |
1426 | synchronize_rcu(); |
1427 | kfree(old_disk_conf); | |
813472ce | 1428 | kfree(old_plan); |
b30ab791 | 1429 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1430 | goto success; |
1431 | ||
1432 | fail_unlock: | |
0500813f | 1433 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1434 | fail: |
5ecc72c3 | 1435 | kfree(new_disk_conf); |
813472ce | 1436 | kfree(new_plan); |
daeda1cc | 1437 | success: |
b30ab791 | 1438 | put_ldev(device); |
f399002e | 1439 | out: |
9e276872 LE |
1440 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1441 | finish: | |
a910b123 | 1442 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1443 | return 0; |
1444 | } | |
1445 | ||
3b98c0c2 | 1446 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1447 | { |
a910b123 | 1448 | struct drbd_config_context adm_ctx; |
b30ab791 | 1449 | struct drbd_device *device; |
44a4d551 LE |
1450 | struct drbd_peer_device *peer_device; |
1451 | struct drbd_connection *connection; | |
3b98c0c2 | 1452 | int err; |
116676ca | 1453 | enum drbd_ret_code retcode; |
b411b363 PR |
1454 | enum determine_dev_size dd; |
1455 | sector_t max_possible_sectors; | |
1456 | sector_t min_md_device_sectors; | |
1457 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1458 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1459 | struct block_device *bdev; |
b411b363 | 1460 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1461 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1462 | union drbd_state ns, os; |
f2024e7c | 1463 | enum drbd_state_rv rv; |
44ed167d | 1464 | struct net_conf *nc; |
b411b363 | 1465 | |
a910b123 | 1466 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1467 | if (!adm_ctx.reply_skb) |
1468 | return retcode; | |
1469 | if (retcode != NO_ERROR) | |
40cbf085 | 1470 | goto finish; |
b411b363 | 1471 | |
b30ab791 | 1472 | device = adm_ctx.device; |
9e276872 | 1473 | mutex_lock(&adm_ctx.resource->adm_mutex); |
44a4d551 LE |
1474 | peer_device = first_peer_device(device); |
1475 | connection = peer_device ? peer_device->connection : NULL; | |
1476 | conn_reconfig_start(connection); | |
b411b363 PR |
1477 | |
1478 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1479 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1480 | retcode = ERR_DISK_CONFIGURED; |
1481 | goto fail; | |
1482 | } | |
82f59cc6 LE |
1483 | /* It may just now have detached because of IO error. Make sure |
1484 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1485 | * e.g. if someone calls attach from the on-io-error handler, | |
1486 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
e334f550 | 1487 | wait_event(device->misc_wait, !test_bit(GOING_DISKLESS, &device->flags)); |
b411b363 | 1488 | |
383606e0 | 1489 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1490 | clear_bit(FORCE_DETACH, &device->flags); |
1491 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1492 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1493 | |
0029d624 | 1494 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1495 | device->rs_total = 0; |
1496 | device->rs_failed = 0; | |
1497 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1498 | |
3b98c0c2 | 1499 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1500 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1501 | if (!nbc) { | |
1502 | retcode = ERR_NOMEM; | |
1503 | goto fail; | |
1504 | } | |
9f2247bb PR |
1505 | spin_lock_init(&nbc->md.uuid_lock); |
1506 | ||
daeda1cc PR |
1507 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1508 | if (!new_disk_conf) { | |
1509 | retcode = ERR_NOMEM; | |
b411b363 PR |
1510 | goto fail; |
1511 | } | |
daeda1cc | 1512 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1513 | |
daeda1cc PR |
1514 | set_disk_conf_defaults(new_disk_conf); |
1515 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1516 | if (err) { |
b411b363 | 1517 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1518 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1519 | goto fail; |
1520 | } | |
1521 | ||
5bbcf5e6 LE |
1522 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1523 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1524 | |
9958c857 PR |
1525 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1526 | if (!new_plan) { | |
1527 | retcode = ERR_NOMEM; | |
1528 | goto fail; | |
1529 | } | |
1530 | ||
daeda1cc | 1531 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1532 | retcode = ERR_MD_IDX_INVALID; |
1533 | goto fail; | |
1534 | } | |
1535 | ||
a3f8f7dc | 1536 | write_lock_irq(&global_state_lock); |
b30ab791 | 1537 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1538 | write_unlock_irq(&global_state_lock); |
1539 | if (retcode != NO_ERROR) | |
1540 | goto fail; | |
1541 | ||
44ed167d | 1542 | rcu_read_lock(); |
44a4d551 | 1543 | nc = rcu_dereference(connection->net_conf); |
44ed167d | 1544 | if (nc) { |
daeda1cc | 1545 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1546 | rcu_read_unlock(); |
47ff2d0a PR |
1547 | retcode = ERR_STONITH_AND_PROT_A; |
1548 | goto fail; | |
1549 | } | |
1550 | } | |
44ed167d | 1551 | rcu_read_unlock(); |
47ff2d0a | 1552 | |
daeda1cc | 1553 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1554 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1555 | if (IS_ERR(bdev)) { |
d0180171 | 1556 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1557 | PTR_ERR(bdev)); |
b411b363 PR |
1558 | retcode = ERR_OPEN_DISK; |
1559 | goto fail; | |
1560 | } | |
e525fd89 TH |
1561 | nbc->backing_bdev = bdev; |
1562 | ||
1563 | /* | |
1564 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1565 | * drbd sharing one meta device. TODO in that case, paranoia | |
1566 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1567 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1568 | * should check it for you already; but if you don't, or | |
1569 | * someone fooled it, we need to double check here) | |
1570 | */ | |
daeda1cc | 1571 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1572 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1573 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1574 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1575 | if (IS_ERR(bdev)) { |
d0180171 | 1576 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1577 | PTR_ERR(bdev)); |
b411b363 PR |
1578 | retcode = ERR_OPEN_MD_DISK; |
1579 | goto fail; | |
1580 | } | |
e525fd89 | 1581 | nbc->md_bdev = bdev; |
b411b363 | 1582 | |
e525fd89 | 1583 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1584 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1585 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1586 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1587 | goto fail; |
1588 | } | |
1589 | ||
1590 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1591 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1592 | offsetof(struct bm_extent, lce)); |
1593 | if (!resync_lru) { | |
1594 | retcode = ERR_NOMEM; | |
e525fd89 | 1595 | goto fail; |
b411b363 PR |
1596 | } |
1597 | ||
c04ccaa6 LE |
1598 | /* Read our meta data super block early. |
1599 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1600 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1601 | if (retcode != NO_ERROR) |
1602 | goto fail; | |
b411b363 | 1603 | |
5bbcf5e6 LE |
1604 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1605 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1606 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1607 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1608 | ||
daeda1cc | 1609 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1610 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1611 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1612 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1613 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1614 | goto fail; |
b411b363 PR |
1615 | } |
1616 | ||
daeda1cc | 1617 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1618 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1619 | /* at least one MB, otherwise it does not make sense */ | |
1620 | min_md_device_sectors = (2<<10); | |
1621 | } else { | |
1622 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1623 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1624 | } |
1625 | ||
b411b363 | 1626 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1627 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1628 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1629 | "at least %llu sectors needed for this meta-disk type\n", |
1630 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1631 | goto fail; |
b411b363 PR |
1632 | } |
1633 | ||
1634 | /* Make sure the new disk is big enough | |
1635 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1636 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1637 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1638 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1639 | goto fail; |
b411b363 PR |
1640 | } |
1641 | ||
1642 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1643 | ||
1352994b | 1644 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1645 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1646 | "to currently maximum possible %llu sectors <==\n", |
1647 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1648 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1649 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1650 | "meta data may help <<==\n"); |
1651 | } | |
1652 | ||
b30ab791 | 1653 | drbd_suspend_io(device); |
b411b363 | 1654 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1655 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1656 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1657 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1658 | * As barriers are counted per resource, | |
1659 | * we'd need to suspend io on all devices of a resource. | |
1660 | */ | |
b30ab791 | 1661 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1662 | /* and for any other previously queued work */ |
44a4d551 | 1663 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 1664 | |
b30ab791 | 1665 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1666 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1667 | drbd_resume_io(device); |
f2024e7c | 1668 | if (rv < SS_SUCCESS) |
e525fd89 | 1669 | goto fail; |
b411b363 | 1670 | |
b30ab791 | 1671 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1672 | goto force_diskless; |
1673 | ||
b30ab791 AG |
1674 | if (!device->bitmap) { |
1675 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1676 | retcode = ERR_NOMEM; |
1677 | goto force_diskless_dec; | |
1678 | } | |
1679 | } | |
1680 | ||
b30ab791 | 1681 | if (device->state.conn < C_CONNECTED && |
babea49e | 1682 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1683 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1684 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1685 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1686 | retcode = ERR_DATA_NOT_CURRENT; |
1687 | goto force_diskless_dec; | |
1688 | } | |
1689 | ||
1690 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1691 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1692 | retcode = ERR_NOMEM; |
1693 | goto force_diskless_dec; | |
1694 | } | |
1695 | ||
1696 | /* Prevent shrinking of consistent devices ! */ | |
1697 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1698 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1699 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1700 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1701 | goto force_diskless_dec; |
1702 | } | |
1703 | ||
b411b363 PR |
1704 | /* Reset the "barriers don't work" bits here, then force meta data to |
1705 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1706 | if (new_disk_conf->md_flushes) |
b30ab791 | 1707 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1708 | else |
b30ab791 | 1709 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1710 | |
1711 | /* Point of no return reached. | |
1712 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1713 | * now device takes over responsibility, and the state engine should |
b411b363 | 1714 | * clean it up somewhere. */ |
0b0ba1ef | 1715 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1716 | device->ldev = nbc; |
1717 | device->resync = resync_lru; | |
1718 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1719 | nbc = NULL; |
1720 | resync_lru = NULL; | |
daeda1cc | 1721 | new_disk_conf = NULL; |
9958c857 | 1722 | new_plan = NULL; |
b411b363 | 1723 | |
8fe39aac | 1724 | drbd_bump_write_ordering(device->resource, device->ldev, WO_bdev_flush); |
b411b363 | 1725 | |
b30ab791 AG |
1726 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1727 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1728 | else |
b30ab791 | 1729 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1730 | |
b30ab791 | 1731 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1732 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1733 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1734 | |
b30ab791 AG |
1735 | device->send_cnt = 0; |
1736 | device->recv_cnt = 0; | |
1737 | device->read_cnt = 0; | |
1738 | device->writ_cnt = 0; | |
b411b363 | 1739 | |
8fe39aac | 1740 | drbd_reconsider_max_bio_size(device, device->ldev); |
b411b363 PR |
1741 | |
1742 | /* If I am currently not R_PRIMARY, | |
1743 | * but meta data primary indicator is set, | |
1744 | * I just now recover from a hard crash, | |
1745 | * and have been R_PRIMARY before that crash. | |
1746 | * | |
1747 | * Now, if I had no connection before that crash | |
1748 | * (have been degraded R_PRIMARY), chances are that | |
1749 | * I won't find my peer now either. | |
1750 | * | |
1751 | * In that case, and _only_ in that case, | |
1752 | * we use the degr-wfc-timeout instead of the default, | |
1753 | * so we can automatically recover from a crash of a | |
1754 | * degraded but active "cluster" after a certain timeout. | |
1755 | */ | |
b30ab791 AG |
1756 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1757 | if (device->state.role != R_PRIMARY && | |
1758 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1759 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1760 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1761 | |
b30ab791 | 1762 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1763 | if (dd <= DS_ERROR) { |
b411b363 PR |
1764 | retcode = ERR_NOMEM_BITMAP; |
1765 | goto force_diskless_dec; | |
e96c9633 | 1766 | } else if (dd == DS_GREW) |
b30ab791 | 1767 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1768 | |
b30ab791 AG |
1769 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1770 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1771 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1772 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1773 | "(aka FullSync)\n"); |
b30ab791 | 1774 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1775 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1776 | retcode = ERR_IO_MD_DISK; |
1777 | goto force_diskless_dec; | |
1778 | } | |
1779 | } else { | |
b30ab791 | 1780 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1781 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1782 | retcode = ERR_IO_MD_DISK; |
1783 | goto force_diskless_dec; | |
1784 | } | |
b411b363 PR |
1785 | } |
1786 | ||
b30ab791 AG |
1787 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1788 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1789 | |
0500813f | 1790 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1791 | os = drbd_read_state(device); |
78bae59b | 1792 | ns = os; |
b411b363 PR |
1793 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1794 | otherwise investigate MDF_WasUpToDate... | |
1795 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1796 | otherwise into D_CONSISTENT state. | |
1797 | */ | |
b30ab791 AG |
1798 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1799 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1800 | ns.disk = D_CONSISTENT; |
1801 | else | |
1802 | ns.disk = D_OUTDATED; | |
1803 | } else { | |
1804 | ns.disk = D_INCONSISTENT; | |
1805 | } | |
1806 | ||
b30ab791 | 1807 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1808 | ns.pdsk = D_OUTDATED; |
1809 | ||
daeda1cc PR |
1810 | rcu_read_lock(); |
1811 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1812 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1813 | ns.disk = D_UP_TO_DATE; |
1814 | ||
1815 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1816 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1817 | this point, because drbd_request_state() modifies these | |
1818 | flags. */ | |
1819 | ||
b30ab791 AG |
1820 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1821 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1822 | else |
b30ab791 | 1823 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1824 | |
1825 | rcu_read_unlock(); | |
1826 | ||
b411b363 PR |
1827 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1828 | state after the negotiation phase. */ | |
b30ab791 AG |
1829 | if (device->state.conn == C_CONNECTED) { |
1830 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1831 | ns.i = os.i; |
1832 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1833 | |
1834 | /* We expect to receive up-to-date UUIDs soon. | |
1835 | To avoid a race in receive_state, free p_uuid while | |
1836 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1837 | kfree(device->p_uuid); |
1838 | device->p_uuid = NULL; | |
b411b363 PR |
1839 | } |
1840 | ||
b30ab791 | 1841 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1842 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1843 | |
1844 | if (rv < SS_SUCCESS) | |
1845 | goto force_diskless_dec; | |
1846 | ||
b30ab791 | 1847 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1848 | |
b30ab791 AG |
1849 | if (device->state.role == R_PRIMARY) |
1850 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1851 | else |
b30ab791 | 1852 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1853 | |
b30ab791 AG |
1854 | drbd_md_mark_dirty(device); |
1855 | drbd_md_sync(device); | |
b411b363 | 1856 | |
b30ab791 AG |
1857 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1858 | put_ldev(device); | |
44a4d551 | 1859 | conn_reconfig_done(connection); |
9e276872 | 1860 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 1861 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1862 | return 0; |
1863 | ||
1864 | force_diskless_dec: | |
b30ab791 | 1865 | put_ldev(device); |
b411b363 | 1866 | force_diskless: |
b30ab791 AG |
1867 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1868 | drbd_md_sync(device); | |
b411b363 | 1869 | fail: |
44a4d551 | 1870 | conn_reconfig_done(connection); |
b411b363 | 1871 | if (nbc) { |
e525fd89 TH |
1872 | if (nbc->backing_bdev) |
1873 | blkdev_put(nbc->backing_bdev, | |
1874 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1875 | if (nbc->md_bdev) | |
1876 | blkdev_put(nbc->md_bdev, | |
1877 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1878 | kfree(nbc); |
1879 | } | |
daeda1cc | 1880 | kfree(new_disk_conf); |
b411b363 | 1881 | lc_destroy(resync_lru); |
9958c857 | 1882 | kfree(new_plan); |
9e276872 | 1883 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 1884 | finish: |
a910b123 | 1885 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1886 | return 0; |
1887 | } | |
1888 | ||
b30ab791 | 1889 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1890 | { |
19f83c76 | 1891 | enum drbd_state_rv retcode; |
9a0d9d03 | 1892 | int ret; |
02ee8f95 | 1893 | |
cdfda633 | 1894 | if (force) { |
b30ab791 AG |
1895 | set_bit(FORCE_DETACH, &device->flags); |
1896 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1897 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1898 | goto out; |
1899 | } | |
1900 | ||
b30ab791 AG |
1901 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
1902 | drbd_md_get_buffer(device); /* make sure there is no in-flight meta-data IO */ | |
1903 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
1904 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1905 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1906 | ret = wait_event_interruptible(device->misc_wait, |
1907 | device->state.disk != D_FAILED); | |
1908 | drbd_resume_io(device); | |
9b2f61ae | 1909 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1910 | retcode = SS_NOTHING_TO_DO; |
1911 | if (ret) | |
1912 | retcode = ERR_INTR; | |
02ee8f95 | 1913 | out: |
85f75dd7 | 1914 | return retcode; |
b411b363 PR |
1915 | } |
1916 | ||
82f59cc6 LE |
1917 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1918 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1919 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1920 | * internal references as well. | |
1921 | * Only then we have finally detached. */ | |
3b98c0c2 | 1922 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1923 | { |
a910b123 | 1924 | struct drbd_config_context adm_ctx; |
116676ca | 1925 | enum drbd_ret_code retcode; |
cdfda633 PR |
1926 | struct detach_parms parms = { }; |
1927 | int err; | |
b411b363 | 1928 | |
a910b123 | 1929 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1930 | if (!adm_ctx.reply_skb) |
1931 | return retcode; | |
1932 | if (retcode != NO_ERROR) | |
1933 | goto out; | |
b411b363 | 1934 | |
cdfda633 PR |
1935 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1936 | err = detach_parms_from_attrs(&parms, info); | |
1937 | if (err) { | |
1938 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 1939 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
1940 | goto out; |
1941 | } | |
b411b363 PR |
1942 | } |
1943 | ||
9e276872 | 1944 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 1945 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 1946 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 1947 | out: |
a910b123 | 1948 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1949 | return 0; |
1950 | } | |
b411b363 | 1951 | |
bde89a9e | 1952 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1953 | { |
c06ece6b | 1954 | struct drbd_peer_device *peer_device; |
695d08fa | 1955 | bool rv = false; |
f399002e LE |
1956 | int vnr; |
1957 | ||
695d08fa | 1958 | rcu_read_lock(); |
c06ece6b AG |
1959 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1960 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1961 | if (device->state.conn == C_SYNC_SOURCE || |
1962 | device->state.conn == C_SYNC_TARGET || | |
1963 | device->state.conn == C_PAUSED_SYNC_S || | |
1964 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1965 | rv = true; |
1966 | break; | |
1967 | } | |
b411b363 | 1968 | } |
695d08fa | 1969 | rcu_read_unlock(); |
b411b363 | 1970 | |
695d08fa | 1971 | return rv; |
f399002e | 1972 | } |
47ff2d0a | 1973 | |
bde89a9e | 1974 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1975 | { |
c06ece6b | 1976 | struct drbd_peer_device *peer_device; |
695d08fa | 1977 | bool rv = false; |
f399002e LE |
1978 | int vnr; |
1979 | ||
695d08fa | 1980 | rcu_read_lock(); |
c06ece6b AG |
1981 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1982 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1983 | if (device->state.conn == C_VERIFY_S || |
1984 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1985 | rv = true; |
1986 | break; | |
47ff2d0a PR |
1987 | } |
1988 | } | |
695d08fa | 1989 | rcu_read_unlock(); |
b411b363 | 1990 | |
695d08fa | 1991 | return rv; |
f399002e | 1992 | } |
422028b1 | 1993 | |
cd64397c | 1994 | static enum drbd_ret_code |
270eb5c9 | 1995 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 1996 | { |
c06ece6b | 1997 | struct drbd_peer_device *peer_device; |
cd64397c | 1998 | int i; |
b411b363 | 1999 | |
270eb5c9 AG |
2000 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
2001 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 2002 | return ERR_NEED_APV_100; |
b411b363 | 2003 | |
270eb5c9 | 2004 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
2005 | return ERR_NEED_APV_100; |
2006 | ||
270eb5c9 | 2007 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 2008 | return ERR_NEED_APV_100; |
b411b363 PR |
2009 | } |
2010 | ||
270eb5c9 | 2011 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
2012 | conn_highest_role(connection) == R_PRIMARY && |
2013 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 2014 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 2015 | |
270eb5c9 AG |
2016 | if (new_net_conf->two_primaries && |
2017 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2018 | return ERR_NOT_PROTO_C; |
2019 | ||
c06ece6b AG |
2020 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2021 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2022 | if (get_ldev(device)) { |
2023 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2024 | put_ldev(device); | |
270eb5c9 | 2025 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2026 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2027 | } |
270eb5c9 | 2028 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2029 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2030 | } |
2031 | ||
270eb5c9 | 2032 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2033 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2034 | |
cd64397c PR |
2035 | return NO_ERROR; |
2036 | } | |
b411b363 | 2037 | |
44ed167d | 2038 | static enum drbd_ret_code |
270eb5c9 | 2039 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
2040 | { |
2041 | static enum drbd_ret_code rv; | |
c06ece6b | 2042 | struct drbd_peer_device *peer_device; |
44ed167d | 2043 | int i; |
b411b363 | 2044 | |
44ed167d | 2045 | rcu_read_lock(); |
270eb5c9 | 2046 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2047 | rcu_read_unlock(); |
b411b363 | 2048 | |
bde89a9e | 2049 | /* connection->volumes protected by genl_lock() here */ |
c06ece6b AG |
2050 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2051 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2052 | if (!device->bitmap) { |
2053 | if (drbd_bm_init(device)) | |
44ed167d | 2054 | return ERR_NOMEM; |
b411b363 PR |
2055 | } |
2056 | } | |
2057 | ||
44ed167d PR |
2058 | return rv; |
2059 | } | |
b411b363 | 2060 | |
0fd0ea06 PR |
2061 | struct crypto { |
2062 | struct crypto_hash *verify_tfm; | |
2063 | struct crypto_hash *csums_tfm; | |
2064 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2065 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2066 | }; |
b411b363 | 2067 | |
0fd0ea06 | 2068 | static int |
4b6ad6d4 | 2069 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2070 | { |
2071 | if (!tfm_name[0]) | |
2072 | return NO_ERROR; | |
b411b363 | 2073 | |
0fd0ea06 PR |
2074 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2075 | if (IS_ERR(*tfm)) { | |
2076 | *tfm = NULL; | |
2077 | return err_alg; | |
b411b363 | 2078 | } |
b411b363 | 2079 | |
0fd0ea06 PR |
2080 | return NO_ERROR; |
2081 | } | |
b411b363 | 2082 | |
0fd0ea06 | 2083 | static enum drbd_ret_code |
270eb5c9 | 2084 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2085 | { |
2086 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2087 | enum drbd_ret_code rv; | |
0fd0ea06 | 2088 | |
270eb5c9 | 2089 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2090 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2091 | if (rv != NO_ERROR) |
2092 | return rv; | |
270eb5c9 | 2093 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2094 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2095 | if (rv != NO_ERROR) |
2096 | return rv; | |
270eb5c9 | 2097 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2098 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2099 | if (rv != NO_ERROR) |
2100 | return rv; | |
270eb5c9 | 2101 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2102 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2103 | new_net_conf->cram_hmac_alg); |
b411b363 | 2104 | |
4b6ad6d4 AG |
2105 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2106 | ERR_AUTH_ALG); | |
b411b363 PR |
2107 | } |
2108 | ||
0fd0ea06 PR |
2109 | return rv; |
2110 | } | |
b411b363 | 2111 | |
0fd0ea06 PR |
2112 | static void free_crypto(struct crypto *crypto) |
2113 | { | |
0fd0ea06 | 2114 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2115 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2116 | crypto_free_hash(crypto->csums_tfm); |
2117 | crypto_free_hash(crypto->verify_tfm); | |
2118 | } | |
b411b363 | 2119 | |
f399002e LE |
2120 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2121 | { | |
a910b123 | 2122 | struct drbd_config_context adm_ctx; |
f399002e | 2123 | enum drbd_ret_code retcode; |
bde89a9e | 2124 | struct drbd_connection *connection; |
270eb5c9 | 2125 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2126 | int err; |
2127 | int ovr; /* online verify running */ | |
2128 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2129 | struct crypto crypto = { }; |
b411b363 | 2130 | |
a910b123 | 2131 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2132 | if (!adm_ctx.reply_skb) |
2133 | return retcode; | |
2134 | if (retcode != NO_ERROR) | |
9e276872 | 2135 | goto finish; |
b411b363 | 2136 | |
bde89a9e | 2137 | connection = adm_ctx.connection; |
9e276872 | 2138 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2139 | |
270eb5c9 AG |
2140 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2141 | if (!new_net_conf) { | |
f399002e LE |
2142 | retcode = ERR_NOMEM; |
2143 | goto out; | |
2144 | } | |
b411b363 | 2145 | |
bde89a9e | 2146 | conn_reconfig_start(connection); |
b411b363 | 2147 | |
bde89a9e | 2148 | mutex_lock(&connection->data.mutex); |
0500813f | 2149 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2150 | old_net_conf = connection->net_conf; |
2561b9c1 | 2151 | |
270eb5c9 | 2152 | if (!old_net_conf) { |
a910b123 | 2153 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2154 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2155 | goto fail; |
2156 | } | |
2157 | ||
270eb5c9 | 2158 | *new_net_conf = *old_net_conf; |
5979e361 | 2159 | if (should_set_defaults(info)) |
270eb5c9 | 2160 | set_net_conf_defaults(new_net_conf); |
f399002e | 2161 | |
270eb5c9 | 2162 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2163 | if (err && err != -ENOMSG) { |
f399002e | 2164 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2165 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2166 | goto fail; |
2561b9c1 | 2167 | } |
b411b363 | 2168 | |
270eb5c9 | 2169 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2170 | if (retcode != NO_ERROR) |
2171 | goto fail; | |
b411b363 | 2172 | |
f399002e | 2173 | /* re-sync running */ |
bde89a9e | 2174 | rsr = conn_resync_running(connection); |
270eb5c9 | 2175 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2176 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2177 | goto fail; |
b411b363 PR |
2178 | } |
2179 | ||
f399002e | 2180 | /* online verify running */ |
bde89a9e | 2181 | ovr = conn_ov_running(connection); |
270eb5c9 | 2182 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2183 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2184 | goto fail; |
f399002e | 2185 | } |
b411b363 | 2186 | |
270eb5c9 | 2187 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2188 | if (retcode != NO_ERROR) |
b411b363 | 2189 | goto fail; |
f399002e | 2190 | |
270eb5c9 | 2191 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2192 | |
2193 | if (!rsr) { | |
bde89a9e AG |
2194 | crypto_free_hash(connection->csums_tfm); |
2195 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2196 | crypto.csums_tfm = NULL; |
f399002e LE |
2197 | } |
2198 | if (!ovr) { | |
bde89a9e AG |
2199 | crypto_free_hash(connection->verify_tfm); |
2200 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2201 | crypto.verify_tfm = NULL; |
b411b363 PR |
2202 | } |
2203 | ||
bde89a9e AG |
2204 | crypto_free_hash(connection->integrity_tfm); |
2205 | connection->integrity_tfm = crypto.integrity_tfm; | |
2206 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2207 | /* Do this without trying to take connection->data.mutex again. */ | |
2208 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2209 | |
bde89a9e AG |
2210 | crypto_free_hash(connection->cram_hmac_tfm); |
2211 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2212 | |
0500813f | 2213 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2214 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2215 | synchronize_rcu(); |
270eb5c9 | 2216 | kfree(old_net_conf); |
91fd4dad | 2217 | |
69a22773 AG |
2218 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2219 | struct drbd_peer_device *peer_device; | |
2220 | int vnr; | |
2221 | ||
2222 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2223 | drbd_send_sync_param(peer_device); | |
2224 | } | |
f399002e | 2225 | |
91fd4dad PR |
2226 | goto done; |
2227 | ||
b411b363 | 2228 | fail: |
0500813f | 2229 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2230 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2231 | free_crypto(&crypto); |
270eb5c9 | 2232 | kfree(new_net_conf); |
91fd4dad | 2233 | done: |
bde89a9e | 2234 | conn_reconfig_done(connection); |
f399002e | 2235 | out: |
9e276872 LE |
2236 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2237 | finish: | |
a910b123 | 2238 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2239 | return 0; |
2240 | } | |
2241 | ||
3b98c0c2 | 2242 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2243 | { |
a910b123 | 2244 | struct drbd_config_context adm_ctx; |
c06ece6b | 2245 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2246 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2247 | struct crypto crypto = { }; |
77c556f6 | 2248 | struct drbd_resource *resource; |
bde89a9e | 2249 | struct drbd_connection *connection; |
3b98c0c2 LE |
2250 | enum drbd_ret_code retcode; |
2251 | int i; | |
2252 | int err; | |
b411b363 | 2253 | |
a910b123 | 2254 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2255 | |
3b98c0c2 LE |
2256 | if (!adm_ctx.reply_skb) |
2257 | return retcode; | |
2258 | if (retcode != NO_ERROR) | |
2259 | goto out; | |
089c075d | 2260 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2261 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2262 | retcode = ERR_INVALID_REQUEST; |
2263 | goto out; | |
2264 | } | |
b411b363 | 2265 | |
089c075d AG |
2266 | /* No need for _rcu here. All reconfiguration is |
2267 | * strictly serialized on genl_lock(). We are protected against | |
2268 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2269 | for_each_resource(resource, &drbd_resources) { |
2270 | for_each_connection(connection, resource) { | |
2271 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2272 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2273 | connection->my_addr_len)) { | |
2274 | retcode = ERR_LOCAL_ADDR; | |
2275 | goto out; | |
2276 | } | |
b411b363 | 2277 | |
77c556f6 AG |
2278 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2279 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2280 | connection->peer_addr_len)) { | |
2281 | retcode = ERR_PEER_ADDR; | |
2282 | goto out; | |
2283 | } | |
089c075d | 2284 | } |
b411b363 PR |
2285 | } |
2286 | ||
9e276872 | 2287 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2288 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2289 | conn_reconfig_start(connection); |
b411b363 | 2290 | |
bde89a9e | 2291 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2292 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2293 | goto fail; |
2294 | } | |
2295 | ||
a209b4ae | 2296 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2297 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2298 | if (!new_net_conf) { | |
b411b363 | 2299 | retcode = ERR_NOMEM; |
b411b363 PR |
2300 | goto fail; |
2301 | } | |
2302 | ||
270eb5c9 | 2303 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2304 | |
270eb5c9 | 2305 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2306 | if (err && err != -ENOMSG) { |
b411b363 | 2307 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2308 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2309 | goto fail; |
2310 | } | |
2311 | ||
270eb5c9 | 2312 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2313 | if (retcode != NO_ERROR) |
422028b1 | 2314 | goto fail; |
b411b363 | 2315 | |
270eb5c9 | 2316 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2317 | if (retcode != NO_ERROR) |
2318 | goto fail; | |
b411b363 | 2319 | |
270eb5c9 | 2320 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2321 | |
b5043c5e | 2322 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2323 | |
0500813f | 2324 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2325 | old_net_conf = connection->net_conf; |
2326 | if (old_net_conf) { | |
b411b363 | 2327 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2328 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2329 | goto fail; |
2330 | } | |
270eb5c9 | 2331 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2332 | |
bde89a9e AG |
2333 | conn_free_crypto(connection); |
2334 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2335 | connection->integrity_tfm = crypto.integrity_tfm; | |
2336 | connection->csums_tfm = crypto.csums_tfm; | |
2337 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2338 | |
bde89a9e AG |
2339 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2340 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2341 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2342 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2343 | |
0500813f | 2344 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2345 | |
695d08fa | 2346 | rcu_read_lock(); |
c06ece6b AG |
2347 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2348 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2349 | device->send_cnt = 0; |
2350 | device->recv_cnt = 0; | |
b411b363 | 2351 | } |
695d08fa | 2352 | rcu_read_unlock(); |
b411b363 | 2353 | |
bde89a9e | 2354 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2355 | |
bde89a9e | 2356 | conn_reconfig_done(connection); |
9e276872 | 2357 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2358 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2359 | return 0; |
b411b363 | 2360 | |
b411b363 | 2361 | fail: |
0fd0ea06 | 2362 | free_crypto(&crypto); |
270eb5c9 | 2363 | kfree(new_net_conf); |
b411b363 | 2364 | |
bde89a9e | 2365 | conn_reconfig_done(connection); |
9e276872 | 2366 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2367 | out: |
a910b123 | 2368 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2369 | return 0; |
2370 | } | |
2371 | ||
bde89a9e | 2372 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2373 | { |
2374 | enum drbd_state_rv rv; | |
85f75dd7 | 2375 | |
bde89a9e | 2376 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2377 | force ? CS_HARD : 0); |
85f75dd7 LE |
2378 | |
2379 | switch (rv) { | |
2380 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2381 | break; |
85f75dd7 LE |
2382 | case SS_ALREADY_STANDALONE: |
2383 | return SS_SUCCESS; | |
2384 | case SS_PRIMARY_NOP: | |
2385 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2386 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2387 | |
2388 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2389 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2390 | |
85f75dd7 LE |
2391 | break; |
2392 | case SS_CW_FAILED_BY_PEER: | |
2393 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2394 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2395 | disk, D_OUTDATED), 0); |
2396 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2397 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2398 | CS_HARD); |
b411b363 | 2399 | } |
85f75dd7 LE |
2400 | break; |
2401 | default:; | |
2402 | /* no special handling necessary */ | |
2403 | } | |
2404 | ||
f3dfa40a LE |
2405 | if (rv >= SS_SUCCESS) { |
2406 | enum drbd_state_rv rv2; | |
2407 | /* No one else can reconfigure the network while I am here. | |
2408 | * The state handling only uses drbd_thread_stop_nowait(), | |
2409 | * we want to really wait here until the receiver is no more. | |
2410 | */ | |
9693da23 | 2411 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2412 | |
2413 | /* Race breaker. This additional state change request may be | |
2414 | * necessary, if this was a forced disconnect during a receiver | |
2415 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2416 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2417 | * C_STANDALONE already, now, and this becomes a no-op. |
2418 | */ | |
bde89a9e | 2419 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2420 | CS_VERBOSE | CS_HARD); |
2421 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2422 | drbd_err(connection, |
f3dfa40a LE |
2423 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2424 | rv2); | |
b411b363 | 2425 | } |
85f75dd7 LE |
2426 | return rv; |
2427 | } | |
b411b363 | 2428 | |
3b98c0c2 | 2429 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2430 | { |
a910b123 | 2431 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2432 | struct disconnect_parms parms; |
bde89a9e | 2433 | struct drbd_connection *connection; |
85f75dd7 | 2434 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2435 | enum drbd_ret_code retcode; |
2436 | int err; | |
2561b9c1 | 2437 | |
a910b123 | 2438 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2439 | if (!adm_ctx.reply_skb) |
2440 | return retcode; | |
2441 | if (retcode != NO_ERROR) | |
2561b9c1 | 2442 | goto fail; |
b411b363 | 2443 | |
bde89a9e | 2444 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2445 | memset(&parms, 0, sizeof(parms)); |
2446 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2447 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2448 | if (err) { |
2449 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2450 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2451 | goto fail; |
2452 | } | |
2453 | } | |
2454 | ||
9e276872 | 2455 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2456 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2457 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2458 | retcode = rv; /* FIXME: Type mismatch. */ |
2459 | else | |
2460 | retcode = NO_ERROR; | |
9e276872 | 2461 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2462 | fail: |
a910b123 | 2463 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2464 | return 0; |
2465 | } | |
2466 | ||
b30ab791 | 2467 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2468 | { |
2469 | int iass; /* I am sync source */ | |
2470 | ||
d0180171 | 2471 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2472 | if (device->state.role != device->state.peer) |
2473 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2474 | else |
a6b32bc3 | 2475 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2476 | |
2477 | if (iass) | |
b30ab791 | 2478 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2479 | else |
b30ab791 | 2480 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2481 | } |
2482 | ||
3b98c0c2 | 2483 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2484 | { |
a910b123 | 2485 | struct drbd_config_context adm_ctx; |
daeda1cc | 2486 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2487 | struct resize_parms rs; |
b30ab791 | 2488 | struct drbd_device *device; |
3b98c0c2 | 2489 | enum drbd_ret_code retcode; |
b411b363 | 2490 | enum determine_dev_size dd; |
d752b269 | 2491 | bool change_al_layout = false; |
6495d2c6 | 2492 | enum dds_flags ddsf; |
daeda1cc | 2493 | sector_t u_size; |
3b98c0c2 | 2494 | int err; |
b411b363 | 2495 | |
a910b123 | 2496 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2497 | if (!adm_ctx.reply_skb) |
2498 | return retcode; | |
2499 | if (retcode != NO_ERROR) | |
9e276872 | 2500 | goto finish; |
3b98c0c2 | 2501 | |
9e276872 | 2502 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2503 | device = adm_ctx.device; |
2504 | if (!get_ldev(device)) { | |
d752b269 PR |
2505 | retcode = ERR_NO_DISK; |
2506 | goto fail; | |
2507 | } | |
2508 | ||
3b98c0c2 | 2509 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2510 | rs.al_stripes = device->ldev->md.al_stripes; |
2511 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2512 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2513 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2514 | if (err) { |
3b98c0c2 | 2515 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2516 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2517 | goto fail_ldev; |
b411b363 PR |
2518 | } |
2519 | } | |
2520 | ||
b30ab791 | 2521 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2522 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2523 | goto fail_ldev; |
b411b363 | 2524 | } |
b411b363 | 2525 | |
b30ab791 AG |
2526 | if (device->state.role == R_SECONDARY && |
2527 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2528 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2529 | goto fail_ldev; |
b411b363 | 2530 | } |
b411b363 | 2531 | |
a6b32bc3 | 2532 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2533 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2534 | goto fail_ldev; |
6495d2c6 PR |
2535 | } |
2536 | ||
daeda1cc | 2537 | rcu_read_lock(); |
b30ab791 | 2538 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2539 | rcu_read_unlock(); |
2540 | if (u_size != (sector_t)rs.resize_size) { | |
2541 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2542 | if (!new_disk_conf) { | |
778f271d | 2543 | retcode = ERR_NOMEM; |
9bcd2521 | 2544 | goto fail_ldev; |
778f271d PR |
2545 | } |
2546 | } | |
2547 | ||
b30ab791 AG |
2548 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2549 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2550 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2551 | ||
2552 | if (al_size_k > (16 * 1024 * 1024)) { | |
2553 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2554 | goto fail_ldev; | |
2555 | } | |
2556 | ||
2557 | if (al_size_k < MD_32kB_SECT/2) { | |
2558 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2559 | goto fail_ldev; | |
2560 | } | |
2561 | ||
cdc6af8d | 2562 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2563 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2564 | goto fail_ldev; | |
2565 | } | |
2566 | ||
2567 | change_al_layout = true; | |
2568 | } | |
2569 | ||
b30ab791 AG |
2570 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2571 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2572 | |
daeda1cc | 2573 | if (new_disk_conf) { |
0500813f | 2574 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2575 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2576 | *new_disk_conf = *old_disk_conf; |
2577 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2578 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2579 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2580 | synchronize_rcu(); |
2581 | kfree(old_disk_conf); | |
b411b363 PR |
2582 | } |
2583 | ||
6495d2c6 | 2584 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2585 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2586 | drbd_md_sync(device); | |
2587 | put_ldev(device); | |
e96c9633 | 2588 | if (dd == DS_ERROR) { |
b411b363 PR |
2589 | retcode = ERR_NOMEM_BITMAP; |
2590 | goto fail; | |
d752b269 PR |
2591 | } else if (dd == DS_ERROR_SPACE_MD) { |
2592 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2593 | goto fail; | |
2594 | } else if (dd == DS_ERROR_SHRINK) { | |
2595 | retcode = ERR_IMPLICIT_SHRINK; | |
2596 | goto fail; | |
b411b363 | 2597 | } |
778f271d | 2598 | |
b30ab791 | 2599 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2600 | if (dd == DS_GREW) |
b30ab791 | 2601 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2602 | |
69a22773 AG |
2603 | drbd_send_uuids(first_peer_device(device)); |
2604 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2605 | } |
2606 | ||
b411b363 | 2607 | fail: |
9e276872 LE |
2608 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2609 | finish: | |
a910b123 | 2610 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2611 | return 0; |
b411b363 | 2612 | |
9bcd2521 | 2613 | fail_ldev: |
b30ab791 | 2614 | put_ldev(device); |
9bcd2521 | 2615 | goto fail; |
b411b363 | 2616 | } |
b411b363 | 2617 | |
f399002e | 2618 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2619 | { |
a910b123 | 2620 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2621 | enum drbd_ret_code retcode; |
b57a1e27 | 2622 | struct res_opts res_opts; |
f399002e | 2623 | int err; |
b411b363 | 2624 | |
a910b123 | 2625 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2626 | if (!adm_ctx.reply_skb) |
2627 | return retcode; | |
2628 | if (retcode != NO_ERROR) | |
2629 | goto fail; | |
b411b363 | 2630 | |
eb6bea67 | 2631 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2632 | if (should_set_defaults(info)) |
b966b5dd | 2633 | set_res_opts_defaults(&res_opts); |
b411b363 | 2634 | |
b57a1e27 | 2635 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2636 | if (err && err != -ENOMSG) { |
b411b363 | 2637 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2638 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2639 | goto fail; |
2640 | } | |
2641 | ||
9e276872 | 2642 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2643 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2644 | if (err) { |
2645 | retcode = ERR_INVALID_REQUEST; | |
2646 | if (err == -ENOMEM) | |
2647 | retcode = ERR_NOMEM; | |
b411b363 | 2648 | } |
9e276872 | 2649 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2650 | |
b411b363 | 2651 | fail: |
a910b123 | 2652 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2653 | return 0; |
2654 | } | |
2655 | ||
3b98c0c2 | 2656 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2657 | { |
a910b123 | 2658 | struct drbd_config_context adm_ctx; |
b30ab791 | 2659 | struct drbd_device *device; |
3b98c0c2 LE |
2660 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2661 | ||
a910b123 | 2662 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2663 | if (!adm_ctx.reply_skb) |
2664 | return retcode; | |
2665 | if (retcode != NO_ERROR) | |
2666 | goto out; | |
2667 | ||
b30ab791 | 2668 | device = adm_ctx.device; |
8fe39aac PR |
2669 | if (!get_ldev(device)) { |
2670 | retcode = ERR_NO_DISK; | |
2671 | goto out; | |
2672 | } | |
2673 | ||
2674 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2675 | |
194bfb32 | 2676 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2677 | * resync just being finished, wait for it before requesting a new resync. |
2678 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2679 | drbd_suspend_io(device); |
2680 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2681 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2682 | |
0b2dafcd PR |
2683 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2684 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2685 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2686 | */ |
b30ab791 AG |
2687 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2688 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2689 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2690 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2691 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2692 | retcode = ERR_IO_MD_DISK; | |
2693 | } | |
2694 | } else | |
b30ab791 AG |
2695 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2696 | drbd_resume_io(device); | |
9e276872 | 2697 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2698 | put_ldev(device); |
3b98c0c2 | 2699 | out: |
a910b123 | 2700 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2701 | return 0; |
2702 | } | |
2703 | ||
3b98c0c2 LE |
2704 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2705 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2706 | { |
a910b123 | 2707 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2708 | enum drbd_ret_code retcode; |
194bfb32 | 2709 | |
a910b123 | 2710 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2711 | if (!adm_ctx.reply_skb) |
2712 | return retcode; | |
2713 | if (retcode != NO_ERROR) | |
2714 | goto out; | |
b411b363 | 2715 | |
9e276872 | 2716 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2717 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 2718 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2719 | out: |
a910b123 | 2720 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2721 | return 0; |
2722 | } | |
2723 | ||
8fe39aac | 2724 | static int drbd_bmio_set_susp_al(struct drbd_device *device) __must_hold(local) |
0778286a PR |
2725 | { |
2726 | int rv; | |
2727 | ||
b30ab791 AG |
2728 | rv = drbd_bmio_set_n_write(device); |
2729 | drbd_suspend_al(device); | |
0778286a PR |
2730 | return rv; |
2731 | } | |
2732 | ||
3b98c0c2 | 2733 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2734 | { |
a910b123 | 2735 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 2736 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2737 | struct drbd_device *device; |
25b0d6c8 | 2738 | |
a910b123 | 2739 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
2740 | if (!adm_ctx.reply_skb) |
2741 | return retcode; | |
2742 | if (retcode != NO_ERROR) | |
2743 | goto out; | |
2744 | ||
b30ab791 | 2745 | device = adm_ctx.device; |
8fe39aac PR |
2746 | if (!get_ldev(device)) { |
2747 | retcode = ERR_NO_DISK; | |
2748 | goto out; | |
2749 | } | |
2750 | ||
2751 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2752 | |
194bfb32 | 2753 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2754 | * resync just being finished, wait for it before requesting a new resync. |
2755 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2756 | drbd_suspend_io(device); |
2757 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2758 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2759 | |
0b2dafcd PR |
2760 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2761 | * in the bitmap. Otherwise, try to start a resync handshake | |
2762 | * as sync source for full sync. | |
2763 | */ | |
b30ab791 | 2764 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2765 | /* The peer will get a resync upon connect anyways. Just make that |
2766 | into a full resync. */ | |
b30ab791 | 2767 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2768 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2769 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2770 | "set_n_write from invalidate_peer", |
2771 | BM_LOCKED_SET_ALLOWED)) | |
2772 | retcode = ERR_IO_MD_DISK; | |
2773 | } | |
2774 | } else | |
b30ab791 AG |
2775 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2776 | drbd_resume_io(device); | |
9e276872 | 2777 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2778 | put_ldev(device); |
25b0d6c8 | 2779 | out: |
a910b123 | 2780 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2781 | return 0; |
2782 | } | |
2783 | ||
3b98c0c2 | 2784 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2785 | { |
a910b123 | 2786 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2787 | enum drbd_ret_code retcode; |
b411b363 | 2788 | |
a910b123 | 2789 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2790 | if (!adm_ctx.reply_skb) |
2791 | return retcode; | |
2792 | if (retcode != NO_ERROR) | |
2793 | goto out; | |
b411b363 | 2794 | |
9e276872 | 2795 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2796 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 2797 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 2798 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2799 | out: |
a910b123 | 2800 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2801 | return 0; |
2802 | } | |
2803 | ||
3b98c0c2 | 2804 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2805 | { |
a910b123 | 2806 | struct drbd_config_context adm_ctx; |
da9fbc27 | 2807 | union drbd_dev_state s; |
3b98c0c2 LE |
2808 | enum drbd_ret_code retcode; |
2809 | ||
a910b123 | 2810 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2811 | if (!adm_ctx.reply_skb) |
2812 | return retcode; | |
2813 | if (retcode != NO_ERROR) | |
2814 | goto out; | |
b411b363 | 2815 | |
9e276872 | 2816 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2817 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2818 | s = adm_ctx.device->state; | |
cd88d030 PR |
2819 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2820 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2821 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2822 | } else { | |
2823 | retcode = ERR_PAUSE_IS_CLEAR; | |
2824 | } | |
2825 | } | |
9e276872 | 2826 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2827 | out: |
a910b123 | 2828 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2829 | return 0; |
2830 | } | |
2831 | ||
3b98c0c2 | 2832 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2833 | { |
3b98c0c2 | 2834 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2835 | } |
2836 | ||
3b98c0c2 | 2837 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2838 | { |
a910b123 | 2839 | struct drbd_config_context adm_ctx; |
b30ab791 | 2840 | struct drbd_device *device; |
3b98c0c2 LE |
2841 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2842 | ||
a910b123 | 2843 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2844 | if (!adm_ctx.reply_skb) |
2845 | return retcode; | |
2846 | if (retcode != NO_ERROR) | |
2847 | goto out; | |
2848 | ||
9e276872 | 2849 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2850 | device = adm_ctx.device; |
2851 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2852 | drbd_uuid_new_current(device); | |
2853 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2854 | } |
b30ab791 AG |
2855 | drbd_suspend_io(device); |
2856 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2857 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2858 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2859 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2860 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2861 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2862 | } |
b30ab791 | 2863 | drbd_resume_io(device); |
9e276872 | 2864 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2865 | out: |
a910b123 | 2866 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2867 | return 0; |
2868 | } | |
2869 | ||
3b98c0c2 | 2870 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2871 | { |
3b98c0c2 | 2872 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2873 | } |
2874 | ||
251b8f8e AG |
2875 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2876 | struct drbd_resource *resource, | |
2877 | struct drbd_connection *connection, | |
2878 | struct drbd_device *device) | |
b411b363 | 2879 | { |
543cc10b LE |
2880 | struct nlattr *nla; |
2881 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2882 | if (!nla) | |
2883 | goto nla_put_failure; | |
251b8f8e AG |
2884 | if (device && |
2885 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2886 | goto nla_put_failure; |
f597f6b8 | 2887 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 2888 | goto nla_put_failure; |
251b8f8e AG |
2889 | if (connection) { |
2890 | if (connection->my_addr_len && | |
2891 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2892 | goto nla_put_failure; | |
2893 | if (connection->peer_addr_len && | |
2894 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2895 | goto nla_put_failure; | |
2896 | } | |
543cc10b LE |
2897 | nla_nest_end(skb, nla); |
2898 | return 0; | |
b411b363 | 2899 | |
543cc10b LE |
2900 | nla_put_failure: |
2901 | if (nla) | |
2902 | nla_nest_cancel(skb, nla); | |
2903 | return -EMSGSIZE; | |
2904 | } | |
b411b363 | 2905 | |
251b8f8e AG |
2906 | /* |
2907 | * Return the connection of @resource if @resource has exactly one connection. | |
2908 | */ | |
2909 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2910 | { | |
2911 | struct list_head *connections = &resource->connections; | |
2912 | ||
2913 | if (list_empty(connections) || connections->next->next != connections) | |
2914 | return NULL; | |
2915 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2916 | } | |
2917 | ||
8ce953aa | 2918 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 2919 | const struct sib_info *sib) |
b411b363 | 2920 | { |
251b8f8e | 2921 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2922 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2923 | struct nlattr *nla; | |
2924 | int got_ldev; | |
3b98c0c2 LE |
2925 | int err = 0; |
2926 | int exclude_sensitive; | |
2927 | ||
2928 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2929 | * to. So we better exclude_sensitive information. | |
2930 | * | |
2931 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2932 | * in the context of the requesting user process. Exclude sensitive | |
2933 | * information, unless current has superuser. | |
2934 | * | |
2935 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2936 | * relies on the current implementation of netlink_dump(), which | |
2937 | * executes the dump callback successively from netlink_recvmsg(), | |
2938 | * always in the context of the receiving process */ | |
2939 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2940 | ||
b30ab791 | 2941 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2942 | |
2943 | /* We need to add connection name and volume number information still. | |
2944 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 2945 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 2946 | goto nla_put_failure; |
3b98c0c2 | 2947 | |
eb6bea67 | 2948 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
2949 | goto nla_put_failure; |
2950 | ||
daeda1cc | 2951 | rcu_read_lock(); |
f9eb7bf4 AG |
2952 | if (got_ldev) { |
2953 | struct disk_conf *disk_conf; | |
44ed167d | 2954 | |
b30ab791 | 2955 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2956 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2957 | } | |
2958 | if (!err) { | |
2959 | struct net_conf *nc; | |
2960 | ||
a6b32bc3 | 2961 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2962 | if (nc) |
2963 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2964 | } | |
44ed167d PR |
2965 | rcu_read_unlock(); |
2966 | if (err) | |
2967 | goto nla_put_failure; | |
3b98c0c2 | 2968 | |
3b98c0c2 LE |
2969 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2970 | if (!nla) | |
2971 | goto nla_put_failure; | |
26ec9287 | 2972 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2973 | nla_put_u32(skb, T_current_state, device->state.i) || |
2974 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2975 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2976 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2977 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2978 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2979 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2980 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2981 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2982 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2983 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2984 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2985 | goto nla_put_failure; |
3b98c0c2 LE |
2986 | |
2987 | if (got_ldev) { | |
39a1aa7f | 2988 | int err; |
b411b363 | 2989 | |
b30ab791 AG |
2990 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2991 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2992 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2993 | |
2994 | if (err) | |
2995 | goto nla_put_failure; | |
2996 | ||
b30ab791 AG |
2997 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
2998 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
2999 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 3000 | goto nla_put_failure; |
b30ab791 AG |
3001 | if (C_SYNC_SOURCE <= device->state.conn && |
3002 | C_PAUSED_SYNC_T >= device->state.conn) { | |
3003 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
3004 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 3005 | goto nla_put_failure; |
3b98c0c2 | 3006 | } |
b411b363 | 3007 | } |
b411b363 | 3008 | |
3b98c0c2 LE |
3009 | if (sib) { |
3010 | switch(sib->sib_reason) { | |
3011 | case SIB_SYNC_PROGRESS: | |
3012 | case SIB_GET_STATUS_REPLY: | |
3013 | break; | |
3014 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
3015 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
3016 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
3017 | goto nla_put_failure; | |
3b98c0c2 LE |
3018 | break; |
3019 | case SIB_HELPER_POST: | |
26ec9287 AG |
3020 | if (nla_put_u32(skb, T_helper_exit_code, |
3021 | sib->helper_exit_code)) | |
3022 | goto nla_put_failure; | |
3b98c0c2 LE |
3023 | /* fall through */ |
3024 | case SIB_HELPER_PRE: | |
26ec9287 AG |
3025 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
3026 | goto nla_put_failure; | |
3b98c0c2 LE |
3027 | break; |
3028 | } | |
b411b363 | 3029 | } |
3b98c0c2 | 3030 | nla_nest_end(skb, nla); |
b411b363 | 3031 | |
3b98c0c2 LE |
3032 | if (0) |
3033 | nla_put_failure: | |
3034 | err = -EMSGSIZE; | |
3035 | if (got_ldev) | |
b30ab791 | 3036 | put_ldev(device); |
3b98c0c2 | 3037 | return err; |
b411b363 PR |
3038 | } |
3039 | ||
3b98c0c2 | 3040 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3041 | { |
a910b123 | 3042 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3043 | enum drbd_ret_code retcode; |
3044 | int err; | |
b411b363 | 3045 | |
a910b123 | 3046 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3047 | if (!adm_ctx.reply_skb) |
3048 | return retcode; | |
3049 | if (retcode != NO_ERROR) | |
3050 | goto out; | |
b411b363 | 3051 | |
b30ab791 | 3052 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3053 | if (err) { |
3054 | nlmsg_free(adm_ctx.reply_skb); | |
3055 | return err; | |
b411b363 | 3056 | } |
3b98c0c2 | 3057 | out: |
a910b123 | 3058 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3059 | return 0; |
b411b363 PR |
3060 | } |
3061 | ||
4b7a530f | 3062 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3063 | { |
b30ab791 | 3064 | struct drbd_device *device; |
3b98c0c2 | 3065 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3066 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3067 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3068 | struct drbd_resource *tmp; |
543cc10b LE |
3069 | unsigned volume = cb->args[1]; |
3070 | ||
3071 | /* Open coded, deferred, iteration: | |
77c556f6 | 3072 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3073 | * connection = "first connection of resource or undefined"; |
3074 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3075 | * ... |
3076 | * } | |
3077 | * } | |
77c556f6 | 3078 | * where resource is cb->args[0]; |
543cc10b LE |
3079 | * and i is cb->args[1]; |
3080 | * | |
71932efc LE |
3081 | * cb->args[2] indicates if we shall loop over all resources, |
3082 | * or just dump all volumes of a single resource. | |
3083 | * | |
3b98c0c2 LE |
3084 | * This may miss entries inserted after this dump started, |
3085 | * or entries deleted before they are reached. | |
543cc10b | 3086 | * |
b30ab791 | 3087 | * We need to make sure the device won't disappear while |
543cc10b LE |
3088 | * we are looking at it, and revalidate our iterators |
3089 | * on each iteration. | |
3090 | */ | |
b411b363 | 3091 | |
05a10ec7 | 3092 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3093 | rcu_read_lock(); |
543cc10b | 3094 | /* revalidate iterator position */ |
77c556f6 | 3095 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3096 | if (pos == NULL) { |
3097 | /* first iteration */ | |
3098 | pos = tmp; | |
77c556f6 | 3099 | resource = pos; |
543cc10b LE |
3100 | break; |
3101 | } | |
3102 | if (tmp == pos) { | |
77c556f6 | 3103 | resource = pos; |
543cc10b LE |
3104 | break; |
3105 | } | |
b411b363 | 3106 | } |
77c556f6 AG |
3107 | if (resource) { |
3108 | next_resource: | |
251b8f8e AG |
3109 | device = idr_get_next(&resource->devices, &volume); |
3110 | if (!device) { | |
77c556f6 AG |
3111 | /* No more volumes to dump on this resource. |
3112 | * Advance resource iterator. */ | |
3113 | pos = list_entry_rcu(resource->resources.next, | |
3114 | struct drbd_resource, resources); | |
3115 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3116 | if (volume != 0) { |
71932efc LE |
3117 | /* If we reached the end of the list, |
3118 | * or only a single resource dump was requested, | |
3119 | * we are done. */ | |
77c556f6 | 3120 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3121 | goto out; |
543cc10b | 3122 | volume = 0; |
77c556f6 AG |
3123 | resource = pos; |
3124 | goto next_resource; | |
543cc10b LE |
3125 | } |
3126 | } | |
3127 | ||
98683650 | 3128 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3129 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3130 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3131 | if (!dh) | |
543cc10b LE |
3132 | goto out; |
3133 | ||
251b8f8e | 3134 | if (!device) { |
bde89a9e | 3135 | /* This is a connection without a single volume. |
367d675d LE |
3136 | * Suprisingly enough, it may have a network |
3137 | * configuration. */ | |
251b8f8e AG |
3138 | struct drbd_connection *connection; |
3139 | ||
543cc10b LE |
3140 | dh->minor = -1U; |
3141 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3142 | connection = the_only_connection(resource); |
3143 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3144 | goto cancel; |
251b8f8e AG |
3145 | if (connection) { |
3146 | struct net_conf *nc; | |
3147 | ||
3148 | nc = rcu_dereference(connection->net_conf); | |
3149 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3150 | goto cancel; | |
3151 | } | |
367d675d | 3152 | goto done; |
543cc10b | 3153 | } |
b411b363 | 3154 | |
0b0ba1ef | 3155 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3156 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3157 | |
b30ab791 | 3158 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3159 | dh->ret_code = NO_ERROR; |
3160 | ||
b30ab791 | 3161 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3162 | cancel: |
3b98c0c2 | 3163 | genlmsg_cancel(skb, dh); |
543cc10b | 3164 | goto out; |
3b98c0c2 | 3165 | } |
367d675d | 3166 | done: |
3b98c0c2 | 3167 | genlmsg_end(skb, dh); |
bde89a9e | 3168 | } |
b411b363 | 3169 | |
543cc10b | 3170 | out: |
c141ebda | 3171 | rcu_read_unlock(); |
543cc10b | 3172 | /* where to start the next iteration */ |
bde89a9e | 3173 | cb->args[0] = (long)pos; |
77c556f6 | 3174 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3175 | |
77c556f6 | 3176 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3177 | * Which will terminate the dump. */ |
3b98c0c2 | 3178 | return skb->len; |
b411b363 PR |
3179 | } |
3180 | ||
71932efc LE |
3181 | /* |
3182 | * Request status of all resources, or of all volumes within a single resource. | |
3183 | * | |
3184 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3185 | * Which means we cannot use the family->attrbuf or other such members, because | |
3186 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3187 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3188 | * | |
3189 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3190 | */ |
71932efc | 3191 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3192 | { |
71932efc LE |
3193 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3194 | struct nlattr *nla; | |
7c3063cc | 3195 | const char *resource_name; |
4bc76048 | 3196 | struct drbd_resource *resource; |
7c3063cc | 3197 | int maxtype; |
71932efc LE |
3198 | |
3199 | /* Is this a followup call? */ | |
3200 | if (cb->args[0]) { | |
3201 | /* ... of a single resource dump, | |
3202 | * and the resource iterator has been advanced already? */ | |
3203 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3204 | return 0; /* DONE. */ | |
3205 | goto dump; | |
3206 | } | |
3207 | ||
3208 | /* First call (from netlink_dump_start). We need to figure out | |
3209 | * which resource(s) the user wants us to dump. */ | |
3210 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3211 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3212 | DRBD_NLA_CFG_CONTEXT); | |
3213 | ||
3214 | /* No explicit context given. Dump all. */ | |
3215 | if (!nla) | |
3216 | goto dump; | |
7c3063cc AG |
3217 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3218 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3219 | if (IS_ERR(nla)) | |
3220 | return PTR_ERR(nla); | |
71932efc LE |
3221 | /* context given, but no name present? */ |
3222 | if (!nla) | |
3223 | return -EINVAL; | |
7c3063cc | 3224 | resource_name = nla_data(nla); |
4bc76048 AG |
3225 | if (!*resource_name) |
3226 | return -ENODEV; | |
3227 | resource = drbd_find_resource(resource_name); | |
3228 | if (!resource) | |
71932efc LE |
3229 | return -ENODEV; |
3230 | ||
4bc76048 | 3231 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3232 | |
71932efc | 3233 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3234 | * only dump this connection. */ |
4bc76048 | 3235 | cb->args[0] = (long)resource; |
71932efc | 3236 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3237 | cb->args[2] = (long)resource; |
71932efc LE |
3238 | |
3239 | dump: | |
3240 | return get_one_status(skb, cb); | |
3241 | } | |
b411b363 | 3242 | |
3b98c0c2 | 3243 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3244 | { |
a910b123 | 3245 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3246 | enum drbd_ret_code retcode; |
3247 | struct timeout_parms tp; | |
3248 | int err; | |
b411b363 | 3249 | |
a910b123 | 3250 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3251 | if (!adm_ctx.reply_skb) |
3252 | return retcode; | |
3253 | if (retcode != NO_ERROR) | |
3254 | goto out; | |
b411b363 | 3255 | |
3b98c0c2 | 3256 | tp.timeout_type = |
b30ab791 AG |
3257 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3258 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3259 | UT_DEFAULT; |
b411b363 | 3260 | |
3b98c0c2 LE |
3261 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3262 | if (err) { | |
3263 | nlmsg_free(adm_ctx.reply_skb); | |
3264 | return err; | |
3265 | } | |
3266 | out: | |
a910b123 | 3267 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3268 | return 0; |
b411b363 PR |
3269 | } |
3270 | ||
3b98c0c2 | 3271 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3272 | { |
a910b123 | 3273 | struct drbd_config_context adm_ctx; |
b30ab791 | 3274 | struct drbd_device *device; |
3b98c0c2 | 3275 | enum drbd_ret_code retcode; |
58ffa580 | 3276 | struct start_ov_parms parms; |
b411b363 | 3277 | |
a910b123 | 3278 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3279 | if (!adm_ctx.reply_skb) |
3280 | return retcode; | |
3281 | if (retcode != NO_ERROR) | |
3282 | goto out; | |
873b0d5f | 3283 | |
b30ab791 | 3284 | device = adm_ctx.device; |
58ffa580 LE |
3285 | |
3286 | /* resume from last known position, if possible */ | |
b30ab791 | 3287 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3288 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3289 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3290 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3291 | if (err) { |
3292 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3293 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3294 | goto out; |
3295 | } | |
b411b363 | 3296 | } |
9e276872 LE |
3297 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3298 | ||
58ffa580 | 3299 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3300 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3301 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3302 | |
3303 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3304 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3305 | drbd_suspend_io(device); |
3306 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3307 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3308 | drbd_resume_io(device); | |
9e276872 LE |
3309 | |
3310 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 3311 | out: |
a910b123 | 3312 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3313 | return 0; |
3314 | } | |
3315 | ||
3316 | ||
3b98c0c2 | 3317 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3318 | { |
a910b123 | 3319 | struct drbd_config_context adm_ctx; |
b30ab791 | 3320 | struct drbd_device *device; |
3b98c0c2 | 3321 | enum drbd_ret_code retcode; |
b411b363 PR |
3322 | int skip_initial_sync = 0; |
3323 | int err; | |
3b98c0c2 | 3324 | struct new_c_uuid_parms args; |
b411b363 | 3325 | |
a910b123 | 3326 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3327 | if (!adm_ctx.reply_skb) |
3328 | return retcode; | |
3329 | if (retcode != NO_ERROR) | |
3330 | goto out_nolock; | |
b411b363 | 3331 | |
b30ab791 | 3332 | device = adm_ctx.device; |
3b98c0c2 LE |
3333 | memset(&args, 0, sizeof(args)); |
3334 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3335 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3336 | if (err) { |
3337 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3338 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3339 | goto out_nolock; |
3340 | } | |
b411b363 PR |
3341 | } |
3342 | ||
9e276872 | 3343 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3344 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3345 | |
b30ab791 | 3346 | if (!get_ldev(device)) { |
b411b363 PR |
3347 | retcode = ERR_NO_DISK; |
3348 | goto out; | |
3349 | } | |
3350 | ||
3351 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3352 | if (device->state.conn == C_CONNECTED && |
3353 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3354 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3355 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3356 | skip_initial_sync = 1; |
b30ab791 | 3357 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3358 | retcode = ERR_CONNECTED; |
3359 | goto out_dec; | |
3360 | } | |
3361 | ||
b30ab791 AG |
3362 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3363 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3364 | |
3365 | if (args.clear_bm) { | |
b30ab791 | 3366 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3367 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3368 | if (err) { |
d0180171 | 3369 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3370 | retcode = ERR_IO_MD_DISK; |
3371 | } | |
3372 | if (skip_initial_sync) { | |
69a22773 | 3373 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
3374 | _drbd_uuid_set(device, UI_BITMAP, 0); |
3375 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3376 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3377 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3378 | CS_VERBOSE, NULL); |
0500813f | 3379 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3380 | } |
3381 | } | |
3382 | ||
b30ab791 | 3383 | drbd_md_sync(device); |
b411b363 | 3384 | out_dec: |
b30ab791 | 3385 | put_ldev(device); |
b411b363 | 3386 | out: |
b30ab791 | 3387 | mutex_unlock(device->state_mutex); |
9e276872 | 3388 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3389 | out_nolock: |
a910b123 | 3390 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3391 | return 0; |
3392 | } | |
3393 | ||
3b98c0c2 | 3394 | static enum drbd_ret_code |
a910b123 | 3395 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 3396 | { |
a910b123 | 3397 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 3398 | if (!name || !name[0]) { |
a910b123 | 3399 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 3400 | return ERR_MANDATORY_TAG; |
b411b363 | 3401 | } |
3b98c0c2 LE |
3402 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3403 | * we must not allow slashes */ | |
3404 | if (strchr(name, '/')) { | |
a910b123 | 3405 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 3406 | return ERR_INVALID_REQUEST; |
b411b363 | 3407 | } |
3b98c0c2 | 3408 | return NO_ERROR; |
774b3055 | 3409 | } |
b411b363 | 3410 | |
789c1b62 | 3411 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3412 | { |
a910b123 | 3413 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3414 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3415 | struct res_opts res_opts; |
3416 | int err; | |
b411b363 | 3417 | |
a910b123 | 3418 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
3419 | if (!adm_ctx.reply_skb) |
3420 | return retcode; | |
3421 | if (retcode != NO_ERROR) | |
3422 | goto out; | |
b411b363 | 3423 | |
afbbfa88 AG |
3424 | set_res_opts_defaults(&res_opts); |
3425 | err = res_opts_from_attrs(&res_opts, info); | |
3426 | if (err && err != -ENOMSG) { | |
3427 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3428 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 3429 | goto out; |
b411b363 PR |
3430 | } |
3431 | ||
a910b123 | 3432 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
3433 | if (retcode != NO_ERROR) |
3434 | goto out; | |
b411b363 | 3435 | |
5c661042 | 3436 | if (adm_ctx.resource) { |
38f19616 LE |
3437 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3438 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 3439 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
3440 | } |
3441 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3442 | goto out; |
b411b363 | 3443 | } |
b411b363 | 3444 | |
9e276872 | 3445 | /* not yet safe for genl_family.parallel_ops */ |
afbbfa88 | 3446 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3447 | retcode = ERR_NOMEM; |
3b98c0c2 | 3448 | out: |
a910b123 | 3449 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3450 | return 0; |
b411b363 PR |
3451 | } |
3452 | ||
05a10ec7 | 3453 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3454 | { |
a910b123 | 3455 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3456 | struct drbd_genlmsghdr *dh = info->userhdr; |
3457 | enum drbd_ret_code retcode; | |
b411b363 | 3458 | |
a910b123 | 3459 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3460 | if (!adm_ctx.reply_skb) |
3461 | return retcode; | |
3462 | if (retcode != NO_ERROR) | |
3463 | goto out; | |
b411b363 | 3464 | |
f2257a56 | 3465 | if (dh->minor > MINORMASK) { |
a910b123 | 3466 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
3467 | retcode = ERR_INVALID_REQUEST; |
3468 | goto out; | |
b411b363 | 3469 | } |
0c8e36d9 | 3470 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 3471 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
3472 | retcode = ERR_INVALID_REQUEST; |
3473 | goto out; | |
b411b363 | 3474 | } |
b411b363 | 3475 | |
38f19616 | 3476 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3477 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3478 | if (adm_ctx.device) { |
38f19616 LE |
3479 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
3480 | retcode = ERR_MINOR_EXISTS; | |
3481 | /* else: still NO_ERROR */ | |
3482 | goto out; | |
b411b363 | 3483 | } |
38f19616 | 3484 | |
9e276872 | 3485 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 3486 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
9e276872 | 3487 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3488 | out: |
a910b123 | 3489 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3490 | return 0; |
b411b363 PR |
3491 | } |
3492 | ||
05a10ec7 | 3493 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3494 | { |
b30ab791 AG |
3495 | if (device->state.disk == D_DISKLESS && |
3496 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3497 | * we may want to delete a minor from a live replication group. |
3498 | */ | |
b30ab791 AG |
3499 | device->state.role == R_SECONDARY) { |
3500 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3501 | CS_VERBOSE + CS_WAIT_COMPLETE); |
f82795d6 | 3502 | drbd_delete_device(device); |
85f75dd7 LE |
3503 | return NO_ERROR; |
3504 | } else | |
3505 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3506 | } |
3507 | ||
05a10ec7 | 3508 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3509 | { |
a910b123 | 3510 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3511 | enum drbd_ret_code retcode; |
b411b363 | 3512 | |
a910b123 | 3513 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3514 | if (!adm_ctx.reply_skb) |
3515 | return retcode; | |
3516 | if (retcode != NO_ERROR) | |
3517 | goto out; | |
b411b363 | 3518 | |
9e276872 | 3519 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 3520 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 3521 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 3522 | out: |
a910b123 | 3523 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 3524 | return 0; |
b411b363 PR |
3525 | } |
3526 | ||
85f75dd7 | 3527 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3528 | { |
a910b123 | 3529 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
3530 | struct drbd_resource *resource; |
3531 | struct drbd_connection *connection; | |
3532 | struct drbd_device *device; | |
f3dfa40a | 3533 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3534 | unsigned i; |
b411b363 | 3535 | |
a910b123 | 3536 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3537 | if (!adm_ctx.reply_skb) |
3538 | return retcode; | |
3539 | if (retcode != NO_ERROR) | |
9e276872 | 3540 | goto finish; |
b411b363 | 3541 | |
b6f85ef9 | 3542 | resource = adm_ctx.resource; |
9e276872 | 3543 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 3544 | /* demote */ |
b6f85ef9 AG |
3545 | for_each_connection(connection, resource) { |
3546 | struct drbd_peer_device *peer_device; | |
3547 | ||
3548 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3549 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3550 | if (retcode < SS_SUCCESS) { | |
a910b123 | 3551 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
3552 | goto out; |
3553 | } | |
3554 | } | |
3555 | ||
3556 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3557 | if (retcode < SS_SUCCESS) { |
a910b123 | 3558 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 3559 | goto out; |
85f75dd7 | 3560 | } |
b411b363 | 3561 | } |
b411b363 | 3562 | |
85f75dd7 | 3563 | /* detach */ |
b6f85ef9 AG |
3564 | idr_for_each_entry(&resource->devices, device, i) { |
3565 | retcode = adm_detach(device, 0); | |
27012382 | 3566 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 3567 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 3568 | goto out; |
85f75dd7 LE |
3569 | } |
3570 | } | |
b411b363 | 3571 | |
bde89a9e | 3572 | /* If we reach this, all volumes (of this connection) are Secondary, |
f3dfa40a | 3573 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have |
c141ebda | 3574 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
b6f85ef9 AG |
3575 | for_each_connection(connection, resource) |
3576 | drbd_thread_stop(&connection->worker); | |
b411b363 | 3577 | |
f3dfa40a | 3578 | /* Now, nothing can fail anymore */ |
b411b363 | 3579 | |
85f75dd7 | 3580 | /* delete volumes */ |
b6f85ef9 AG |
3581 | idr_for_each_entry(&resource->devices, device, i) { |
3582 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3583 | if (retcode != NO_ERROR) { |
3584 | /* "can not happen" */ | |
a910b123 | 3585 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 3586 | goto out; |
85f75dd7 LE |
3587 | } |
3588 | } | |
b411b363 | 3589 | |
b6f85ef9 AG |
3590 | list_del_rcu(&resource->resources); |
3591 | synchronize_rcu(); | |
3592 | drbd_free_resource(resource); | |
3593 | retcode = NO_ERROR; | |
3b98c0c2 | 3594 | out: |
9e276872 LE |
3595 | mutex_unlock(&resource->adm_mutex); |
3596 | finish: | |
a910b123 | 3597 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3598 | return 0; |
b411b363 PR |
3599 | } |
3600 | ||
789c1b62 | 3601 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3602 | { |
a910b123 | 3603 | struct drbd_config_context adm_ctx; |
77c556f6 AG |
3604 | struct drbd_resource *resource; |
3605 | struct drbd_connection *connection; | |
3b98c0c2 | 3606 | enum drbd_ret_code retcode; |
b411b363 | 3607 | |
a910b123 | 3608 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3609 | if (!adm_ctx.reply_skb) |
3610 | return retcode; | |
3611 | if (retcode != NO_ERROR) | |
9e276872 | 3612 | goto finish; |
b411b363 | 3613 | |
77c556f6 | 3614 | resource = adm_ctx.resource; |
9e276872 | 3615 | mutex_lock(&resource->adm_mutex); |
77c556f6 AG |
3616 | for_each_connection(connection, resource) { |
3617 | if (connection->cstate > C_STANDALONE) { | |
3618 | retcode = ERR_NET_CONFIGURED; | |
3619 | goto out; | |
3620 | } | |
3621 | } | |
3622 | if (!idr_is_empty(&resource->devices)) { | |
789c1b62 | 3623 | retcode = ERR_RES_IN_USE; |
77c556f6 | 3624 | goto out; |
b411b363 PR |
3625 | } |
3626 | ||
77c556f6 AG |
3627 | list_del_rcu(&resource->resources); |
3628 | for_each_connection(connection, resource) | |
3629 | drbd_thread_stop(&connection->worker); | |
3630 | synchronize_rcu(); | |
3631 | drbd_free_resource(resource); | |
3632 | retcode = NO_ERROR; | |
3b98c0c2 | 3633 | out: |
9e276872 LE |
3634 | mutex_unlock(&resource->adm_mutex); |
3635 | finish: | |
a910b123 | 3636 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3637 | return 0; |
3638 | } | |
3639 | ||
b30ab791 | 3640 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3641 | { |
3b98c0c2 LE |
3642 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3643 | struct sk_buff *msg; | |
3644 | struct drbd_genlmsghdr *d_out; | |
3645 | unsigned seq; | |
3646 | int err = -ENOMEM; | |
3647 | ||
3648 | seq = atomic_inc_return(&drbd_genl_seq); | |
3649 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3650 | if (!msg) | |
3651 | goto failed; | |
3652 | ||
3653 | err = -EMSGSIZE; | |
3654 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3655 | if (!d_out) /* cannot happen, but anyways. */ | |
3656 | goto nla_put_failure; | |
b30ab791 | 3657 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3658 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3659 | |
b30ab791 | 3660 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3661 | goto nla_put_failure; |
3662 | genlmsg_end(msg, d_out); | |
3663 | err = drbd_genl_multicast_events(msg, 0); | |
3664 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3665 | if (err && err != -ESRCH) | |
3666 | goto failed; | |
b411b363 | 3667 | |
3b98c0c2 | 3668 | return; |
b411b363 | 3669 | |
3b98c0c2 LE |
3670 | nla_put_failure: |
3671 | nlmsg_free(msg); | |
3672 | failed: | |
d0180171 | 3673 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3674 | "Event seq:%u sib_reason:%u\n", |
3675 | err, seq, sib->sib_reason); | |
b411b363 | 3676 | } |