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. */ | |
f221f4bc | 95 | static 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 | 548 | kref_get(&connection->kref); |
bbc1c5e8 LE |
549 | /* We may just have force_sig()'ed this thread |
550 | * to get it out of some blocking network function. | |
551 | * Clear signals; otherwise kthread_run(), which internally uses | |
552 | * wait_on_completion_killable(), will mistake our pending signal | |
553 | * for a new fatal signal and fail. */ | |
554 | flush_signals(current); | |
bde89a9e | 555 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); |
9dc9fbb3 | 556 | if (IS_ERR(opa)) { |
1ec861eb | 557 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 558 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 559 | } |
87f7be4c | 560 | } |
b411b363 | 561 | |
bf885f8a | 562 | enum drbd_state_rv |
44a4d551 | 563 | drbd_set_role(struct drbd_device *const device, enum drbd_role new_role, int force) |
b411b363 | 564 | { |
44a4d551 LE |
565 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
566 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; | |
b411b363 | 567 | const int max_tries = 4; |
bf885f8a | 568 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 569 | struct net_conf *nc; |
b411b363 PR |
570 | int try = 0; |
571 | int forced = 0; | |
572 | union drbd_state mask, val; | |
b411b363 | 573 | |
b6f85ef9 AG |
574 | if (new_role == R_PRIMARY) { |
575 | struct drbd_connection *connection; | |
576 | ||
577 | /* Detect dead peers as soon as possible. */ | |
578 | ||
579 | rcu_read_lock(); | |
580 | for_each_connection(connection, device->resource) | |
581 | request_ping(connection); | |
582 | rcu_read_unlock(); | |
583 | } | |
b411b363 | 584 | |
b30ab791 | 585 | mutex_lock(device->state_mutex); |
b411b363 PR |
586 | |
587 | mask.i = 0; mask.role = R_MASK; | |
588 | val.i = 0; val.role = new_role; | |
589 | ||
590 | while (try++ < max_tries) { | |
a8821531 | 591 | rv = _drbd_request_state_holding_state_mutex(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
592 | |
593 | /* in case we first succeeded to outdate, | |
594 | * but now suddenly could establish a connection */ | |
bf885f8a | 595 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
596 | val.pdsk = 0; |
597 | mask.pdsk = 0; | |
598 | continue; | |
599 | } | |
600 | ||
bf885f8a | 601 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
602 | (device->state.disk < D_UP_TO_DATE && |
603 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
604 | mask.disk = D_MASK; |
605 | val.disk = D_UP_TO_DATE; | |
606 | forced = 1; | |
607 | continue; | |
608 | } | |
609 | ||
bf885f8a | 610 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 611 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 612 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 613 | |
44a4d551 | 614 | if (conn_try_outdate_peer(connection)) { |
b411b363 PR |
615 | val.disk = D_UP_TO_DATE; |
616 | mask.disk = D_MASK; | |
617 | } | |
b411b363 PR |
618 | continue; |
619 | } | |
620 | ||
bf885f8a | 621 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 622 | goto out; |
bf885f8a | 623 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
44a4d551 | 624 | if (!conn_try_outdate_peer(connection) && force) { |
d0180171 | 625 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
626 | mask.pdsk = D_MASK; |
627 | val.pdsk = D_OUTDATED; | |
b411b363 | 628 | |
cb703454 | 629 | } |
b411b363 PR |
630 | continue; |
631 | } | |
bf885f8a | 632 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
633 | /* Maybe the peer is detected as dead very soon... |
634 | retry at most once more in this case. */ | |
44ed167d PR |
635 | int timeo; |
636 | rcu_read_lock(); | |
44a4d551 | 637 | nc = rcu_dereference(connection->net_conf); |
44ed167d PR |
638 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
639 | rcu_read_unlock(); | |
640 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
641 | if (try < max_tries) |
642 | try = max_tries - 1; | |
643 | continue; | |
644 | } | |
bf885f8a | 645 | if (rv < SS_SUCCESS) { |
b30ab791 | 646 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 647 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 648 | if (rv < SS_SUCCESS) |
3b98c0c2 | 649 | goto out; |
b411b363 PR |
650 | } |
651 | break; | |
652 | } | |
653 | ||
bf885f8a | 654 | if (rv < SS_SUCCESS) |
3b98c0c2 | 655 | goto out; |
b411b363 PR |
656 | |
657 | if (forced) | |
d0180171 | 658 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
659 | |
660 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 661 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 662 | |
b6dd1a89 LE |
663 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
664 | ||
b411b363 | 665 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
666 | if (get_ldev(device)) { |
667 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
668 | put_ldev(device); | |
b411b363 PR |
669 | } |
670 | } else { | |
66ce6dbc | 671 | mutex_lock(&device->resource->conf_update); |
44a4d551 | 672 | nc = connection->net_conf; |
44ed167d | 673 | if (nc) |
6139f60d | 674 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
66ce6dbc | 675 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 676 | |
b30ab791 AG |
677 | if (get_ldev(device)) { |
678 | if (((device->state.conn < C_CONNECTED || | |
679 | device->state.pdsk <= D_FAILED) | |
680 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
681 | drbd_uuid_new_current(device); | |
b411b363 | 682 | |
b30ab791 AG |
683 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
684 | put_ldev(device); | |
b411b363 PR |
685 | } |
686 | } | |
687 | ||
19f843aa LE |
688 | /* writeout of activity log covered areas of the bitmap |
689 | * to stable storage done in after state change already */ | |
b411b363 | 690 | |
b30ab791 | 691 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
692 | /* if this was forced, we should consider sync */ |
693 | if (forced) | |
44a4d551 LE |
694 | drbd_send_uuids(peer_device); |
695 | drbd_send_current_state(peer_device); | |
b411b363 PR |
696 | } |
697 | ||
b30ab791 | 698 | drbd_md_sync(device); |
720979fb | 699 | set_disk_ro(device->vdisk, new_role == R_SECONDARY); |
b30ab791 | 700 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 701 | out: |
b30ab791 | 702 | mutex_unlock(device->state_mutex); |
bf885f8a | 703 | return rv; |
b411b363 PR |
704 | } |
705 | ||
3b98c0c2 | 706 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 707 | { |
3b98c0c2 LE |
708 | return err == -ENOMSG ? "required attribute missing" : |
709 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 710 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 711 | "invalid attribute value"; |
ef50a3e3 | 712 | } |
b411b363 | 713 | |
3b98c0c2 | 714 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 715 | { |
a910b123 | 716 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
717 | struct set_role_parms parms; |
718 | int err; | |
719 | enum drbd_ret_code retcode; | |
b411b363 | 720 | |
a910b123 | 721 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
722 | if (!adm_ctx.reply_skb) |
723 | return retcode; | |
724 | if (retcode != NO_ERROR) | |
725 | goto out; | |
b411b363 | 726 | |
3b98c0c2 LE |
727 | memset(&parms, 0, sizeof(parms)); |
728 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 729 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
730 | if (err) { |
731 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 732 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
733 | goto out; |
734 | } | |
735 | } | |
9e276872 LE |
736 | genl_unlock(); |
737 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 738 | |
3b98c0c2 | 739 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 740 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 741 | else |
b30ab791 | 742 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
743 | |
744 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
745 | genl_lock(); | |
3b98c0c2 | 746 | out: |
a910b123 | 747 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
748 | return 0; |
749 | } | |
750 | ||
ae8bf312 LE |
751 | /* Initializes the md.*_offset members, so we are able to find |
752 | * the on disk meta data. | |
753 | * | |
754 | * We currently have two possible layouts: | |
755 | * external: | |
756 | * |----------- md_size_sect ------------------| | |
757 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
758 | * | al_offset == 8 | | |
759 | * | bm_offset = al_offset + X | | |
760 | * ==> bitmap sectors = md_size_sect - bm_offset | |
761 | * | |
762 | * internal: | |
763 | * |----------- md_size_sect ------------------| | |
764 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
765 | * | al_offset < 0 | | |
766 | * | bm_offset = al_offset - Y | | |
767 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
768 | * | |
769 | * Activity log size used to be fixed 32kB, | |
770 | * but is about to become configurable. | |
771 | */ | |
b30ab791 | 772 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
773 | struct drbd_backing_dev *bdev) |
774 | { | |
775 | sector_t md_size_sect = 0; | |
c04ccaa6 | 776 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 777 | |
3a4d4eb3 LE |
778 | bdev->md.md_offset = drbd_md_ss(bdev); |
779 | ||
68e41a43 | 780 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
781 | default: |
782 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 783 | bdev->md.md_size_sect = MD_128MB_SECT; |
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_FLEX_EXT: | |
788 | /* just occupy the full device; unit: sectors */ | |
789 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
790 | bdev->md.al_offset = MD_4kB_SECT; |
791 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
792 | break; |
793 | case DRBD_MD_INDEX_INTERNAL: | |
794 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 795 | /* al size is still fixed */ |
ae8bf312 | 796 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
797 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
798 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
799 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
800 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
801 | md_size_sect = ALIGN(md_size_sect, 8); | |
802 | ||
803 | /* plus the "drbd meta data super block", | |
804 | * and the activity log; */ | |
ae8bf312 | 805 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
806 | |
807 | bdev->md.md_size_sect = md_size_sect; | |
808 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 809 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
810 | break; |
811 | } | |
812 | } | |
813 | ||
4b0715f0 | 814 | /* input size is expected to be in KB */ |
b411b363 PR |
815 | char *ppsize(char *buf, unsigned long long size) |
816 | { | |
4b0715f0 LE |
817 | /* Needs 9 bytes at max including trailing NUL: |
818 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
819 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
820 | int base = 0; | |
4b0715f0 | 821 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
822 | /* shift + round */ |
823 | size = (size >> 10) + !!(size & (1<<9)); | |
824 | base++; | |
825 | } | |
4b0715f0 | 826 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
827 | |
828 | return buf; | |
829 | } | |
830 | ||
831 | /* there is still a theoretical deadlock when called from receiver | |
832 | * on an D_INCONSISTENT R_PRIMARY: | |
833 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
834 | * packet from remote to dec_ap_bio again. | |
835 | * receiver receive_sizes(), comes here, | |
836 | * waits for ap_bio_cnt == 0. -> deadlock. | |
837 | * but this cannot happen, actually, because: | |
838 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
839 | * (not connected, or bad/no disk on peer): | |
840 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
841 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
842 | * peer may not initiate a resize. | |
843 | */ | |
3b98c0c2 LE |
844 | /* Note these are not to be confused with |
845 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
846 | * which are (sub) state changes triggered by admin (drbdsetup), | |
847 | * and can be long lived. | |
b30ab791 | 848 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 849 | * and should be short-lived. */ |
b30ab791 | 850 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 851 | { |
b30ab791 AG |
852 | set_bit(SUSPEND_IO, &device->flags); |
853 | if (drbd_suspended(device)) | |
265be2d0 | 854 | return; |
b30ab791 | 855 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
856 | } |
857 | ||
b30ab791 | 858 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 859 | { |
b30ab791 AG |
860 | clear_bit(SUSPEND_IO, &device->flags); |
861 | wake_up(&device->misc_wait); | |
b411b363 PR |
862 | } |
863 | ||
864 | /** | |
865 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 866 | * @device: DRBD device. |
b411b363 PR |
867 | * |
868 | * Returns 0 on success, negative return values indicate errors. | |
869 | * You should call drbd_md_sync() after calling this function. | |
870 | */ | |
d752b269 | 871 | enum determine_dev_size |
b30ab791 | 872 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
873 | { |
874 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 875 | sector_t la_size_sect, u_size; |
b30ab791 | 876 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
877 | u32 prev_al_stripe_size_4k; |
878 | u32 prev_al_stripes; | |
b411b363 PR |
879 | sector_t size; |
880 | char ppb[10]; | |
d752b269 | 881 | void *buffer; |
b411b363 PR |
882 | |
883 | int md_moved, la_size_changed; | |
e96c9633 | 884 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
885 | |
886 | /* race: | |
887 | * application request passes inc_ap_bio, | |
888 | * but then cannot get an AL-reference. | |
889 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
890 | * | |
891 | * to avoid that: | |
892 | * Suspend IO right here. | |
893 | * still lock the act_log to not trigger ASSERTs there. | |
894 | */ | |
b30ab791 | 895 | drbd_suspend_io(device); |
e37d2438 | 896 | buffer = drbd_md_get_buffer(device, __func__); /* Lock meta-data IO */ |
d752b269 | 897 | if (!buffer) { |
b30ab791 | 898 | drbd_resume_io(device); |
d752b269 PR |
899 | return DS_ERROR; |
900 | } | |
b411b363 PR |
901 | |
902 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 903 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 904 | |
b30ab791 AG |
905 | prev_first_sect = drbd_md_first_sector(device->ldev); |
906 | prev_size = device->ldev->md.md_size_sect; | |
907 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 908 | |
d752b269 PR |
909 | if (rs) { |
910 | /* rs is non NULL if we should change the AL layout only */ | |
911 | ||
912 | prev_al_stripes = md->al_stripes; | |
913 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
914 | ||
915 | md->al_stripes = rs->al_stripes; | |
916 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
917 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
918 | } | |
919 | ||
b30ab791 | 920 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 921 | |
daeda1cc | 922 | rcu_read_lock(); |
b30ab791 | 923 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 924 | rcu_read_unlock(); |
b30ab791 | 925 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 926 | |
d752b269 PR |
927 | if (size < la_size_sect) { |
928 | if (rs && u_size == 0) { | |
929 | /* Remove "rs &&" later. This check should always be active, but | |
930 | right now the receiver expects the permissive behavior */ | |
d0180171 | 931 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
932 | "Use --size=%llus for explicit shrink.\n", |
933 | (unsigned long long)size); | |
934 | rv = DS_ERROR_SHRINK; | |
935 | } | |
936 | if (u_size > size) | |
937 | rv = DS_ERROR_SPACE_MD; | |
938 | if (rv != DS_UNCHANGED) | |
939 | goto err_out; | |
940 | } | |
941 | ||
b30ab791 AG |
942 | if (drbd_get_capacity(device->this_bdev) != size || |
943 | drbd_bm_capacity(device) != size) { | |
b411b363 | 944 | int err; |
b30ab791 | 945 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
946 | if (unlikely(err)) { |
947 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 948 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 949 | if (size == 0) { |
d0180171 | 950 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
951 | "Could not allocate bitmap!\n"); |
952 | } else { | |
d0180171 | 953 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
954 | "Leaving size unchanged at size = %lu KB\n", |
955 | (unsigned long)size); | |
956 | } | |
e96c9633 | 957 | rv = DS_ERROR; |
b411b363 PR |
958 | } |
959 | /* racy, see comments above. */ | |
b30ab791 AG |
960 | drbd_set_my_capacity(device, size); |
961 | device->ldev->md.la_size_sect = size; | |
d0180171 | 962 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
963 | (unsigned long long)size>>1); |
964 | } | |
d752b269 PR |
965 | if (rv <= DS_ERROR) |
966 | goto err_out; | |
b411b363 | 967 | |
b30ab791 | 968 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 969 | |
b30ab791 AG |
970 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
971 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 972 | |
d752b269 PR |
973 | if (la_size_changed || md_moved || rs) { |
974 | u32 prev_flags; | |
24dccabb | 975 | |
fcb09674 LE |
976 | /* We do some synchronous IO below, which may take some time. |
977 | * Clear the timer, to avoid scary "timer expired!" messages, | |
978 | * "Superblock" is written out at least twice below, anyways. */ | |
979 | del_timer(&device->md_sync_timer); | |
b30ab791 | 980 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
981 | |
982 | prev_flags = md->flags; | |
983 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 984 | drbd_md_write(device, buffer); |
d752b269 | 985 | |
d0180171 | 986 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
987 | la_size_changed && md_moved ? "size changed and md moved" : |
988 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 989 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 990 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 991 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 992 | drbd_initialize_al(device, buffer); |
d752b269 PR |
993 | |
994 | md->flags = prev_flags; | |
b30ab791 | 995 | drbd_md_write(device, buffer); |
d752b269 PR |
996 | |
997 | if (rs) | |
d0180171 AG |
998 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
999 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
1000 | } |
1001 | ||
cccac985 | 1002 | if (size > la_size_sect) |
57737adc | 1003 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 1004 | if (size < la_size_sect) |
e96c9633 | 1005 | rv = DS_SHRUNK; |
d752b269 PR |
1006 | |
1007 | if (0) { | |
1008 | err_out: | |
1009 | if (rs) { | |
1010 | md->al_stripes = prev_al_stripes; | |
1011 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
1012 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
1013 | ||
b30ab791 | 1014 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
1015 | } |
1016 | } | |
b30ab791 AG |
1017 | lc_unlock(device->act_log); |
1018 | wake_up(&device->al_wait); | |
1019 | drbd_md_put_buffer(device); | |
1020 | drbd_resume_io(device); | |
b411b363 PR |
1021 | |
1022 | return rv; | |
1023 | } | |
1024 | ||
1025 | sector_t | |
b30ab791 | 1026 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1027 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1028 | { |
b30ab791 | 1029 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1030 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1031 | sector_t m_size; /* my size */ |
b411b363 PR |
1032 | sector_t size = 0; |
1033 | ||
1034 | m_size = drbd_get_max_capacity(bdev); | |
1035 | ||
b30ab791 | 1036 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1037 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1038 | p_size = m_size; |
1039 | } | |
1040 | ||
b411b363 PR |
1041 | if (p_size && m_size) { |
1042 | size = min_t(sector_t, p_size, m_size); | |
1043 | } else { | |
cccac985 LE |
1044 | if (la_size_sect) { |
1045 | size = la_size_sect; | |
b411b363 PR |
1046 | if (m_size && m_size < size) |
1047 | size = m_size; | |
1048 | if (p_size && p_size < size) | |
1049 | size = p_size; | |
1050 | } else { | |
1051 | if (m_size) | |
1052 | size = m_size; | |
1053 | if (p_size) | |
1054 | size = p_size; | |
1055 | } | |
1056 | } | |
1057 | ||
1058 | if (size == 0) | |
d0180171 | 1059 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1060 | |
1061 | if (u_size) { | |
1062 | if (u_size > size) | |
d0180171 | 1063 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1064 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1065 | else | |
1066 | size = u_size; | |
1067 | } | |
1068 | ||
1069 | return size; | |
1070 | } | |
1071 | ||
1072 | /** | |
1073 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1074 | * @device: DRBD device. |
b411b363 PR |
1075 | * |
1076 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1077 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1078 | * this function. | |
1079 | */ | |
b30ab791 | 1080 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1081 | { |
1082 | struct lru_cache *n, *t; | |
1083 | struct lc_element *e; | |
1084 | unsigned int in_use; | |
1085 | int i; | |
1086 | ||
b30ab791 AG |
1087 | if (device->act_log && |
1088 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1089 | return 0; |
1090 | ||
1091 | in_use = 0; | |
b30ab791 | 1092 | t = device->act_log; |
7ad651b5 | 1093 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1094 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1095 | |
1096 | if (n == NULL) { | |
d0180171 | 1097 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1098 | return -ENOMEM; |
1099 | } | |
b30ab791 | 1100 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1101 | if (t) { |
1102 | for (i = 0; i < t->nr_elements; i++) { | |
1103 | e = lc_element_by_index(t, i); | |
1104 | if (e->refcnt) | |
d0180171 | 1105 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1106 | e->lc_number, e->refcnt); |
1107 | in_use += e->refcnt; | |
1108 | } | |
1109 | } | |
1110 | if (!in_use) | |
b30ab791 AG |
1111 | device->act_log = n; |
1112 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1113 | if (in_use) { |
d0180171 | 1114 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1115 | lc_destroy(n); |
1116 | return -EBUSY; | |
1117 | } else { | |
1118 | if (t) | |
1119 | lc_destroy(t); | |
1120 | } | |
b30ab791 | 1121 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1122 | return 0; |
1123 | } | |
1124 | ||
8fe39aac PR |
1125 | static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev, |
1126 | unsigned int max_bio_size) | |
b411b363 | 1127 | { |
b30ab791 | 1128 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1129 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1130 | unsigned int max_segments = 0; | |
c1b3156f | 1131 | struct request_queue *b = NULL; |
99432fcc | 1132 | |
8fe39aac PR |
1133 | if (bdev) { |
1134 | b = bdev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1135 | |
1136 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1137 | rcu_read_lock(); |
b30ab791 | 1138 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1139 | rcu_read_unlock(); |
c1b3156f PR |
1140 | |
1141 | blk_set_stacking_limits(&q->limits); | |
1142 | blk_queue_max_write_same_sectors(q, 0); | |
99432fcc | 1143 | } |
b411b363 | 1144 | |
b411b363 | 1145 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1146 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1147 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1148 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1149 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1150 | |
c1b3156f | 1151 | if (b) { |
20c68fde LE |
1152 | struct drbd_connection *connection = first_peer_device(device)->connection; |
1153 | ||
1154 | if (blk_queue_discard(b) && | |
1155 | (connection->cstate < C_CONNECTED || connection->agreed_features & FF_TRIM)) { | |
20c68fde LE |
1156 | /* For now, don't allow more than one activity log extent worth of data |
1157 | * to be discarded in one go. We may need to rework drbd_al_begin_io() | |
1158 | * to allow for even larger discard ranges */ | |
2bb4cd5c | 1159 | blk_queue_max_discard_sectors(q, DRBD_MAX_DISCARD_SECTORS); |
20c68fde LE |
1160 | |
1161 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1162 | /* REALLY? Is stacking secdiscard "legal"? */ | |
1163 | if (blk_queue_secdiscard(b)) | |
1164 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1165 | } else { | |
2bb4cd5c | 1166 | blk_queue_max_discard_sectors(q, 0); |
20c68fde LE |
1167 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); |
1168 | queue_flag_clear_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1169 | } | |
99432fcc PR |
1170 | |
1171 | blk_queue_stack_limits(q, b); | |
1172 | ||
1173 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1174 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1175 | q->backing_dev_info.ra_pages, |
1176 | b->backing_dev_info.ra_pages); | |
1177 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1178 | } | |
99432fcc PR |
1179 | } |
1180 | } | |
1181 | ||
8fe39aac | 1182 | void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev) |
99432fcc | 1183 | { |
db141b2f | 1184 | unsigned int now, new, local, peer; |
99432fcc | 1185 | |
b30ab791 AG |
1186 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1187 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1188 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1189 | |
8fe39aac PR |
1190 | if (bdev) { |
1191 | local = queue_max_hw_sectors(bdev->backing_bdev->bd_disk->queue) << 9; | |
b30ab791 | 1192 | device->local_max_bio_size = local; |
b411b363 | 1193 | } |
db141b2f | 1194 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1195 | |
1196 | /* We may ignore peer limits if the peer is modern enough. | |
1197 | Because new from 8.3.8 onwards the peer can use multiple | |
1198 | BIOs for a single peer_request */ | |
b30ab791 | 1199 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1200 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1201 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1202 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1203 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1204 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1205 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1206 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1207 | else | |
99432fcc | 1208 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1209 | |
fa090e70 LE |
1210 | /* We may later detach and re-attach on a disconnected Primary. |
1211 | * Avoid this setting to jump back in that case. | |
1212 | * We want to store what we know the peer DRBD can handle, | |
1213 | * not what the peer IO backend can handle. */ | |
1214 | if (peer > device->peer_max_bio_size) | |
1215 | device->peer_max_bio_size = peer; | |
1216 | } | |
db141b2f | 1217 | new = min(local, peer); |
99432fcc | 1218 | |
b30ab791 | 1219 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1220 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1221 | |
1222 | if (new != now) | |
d0180171 | 1223 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1224 | |
8fe39aac | 1225 | drbd_setup_queue_param(device, bdev, new); |
b411b363 PR |
1226 | } |
1227 | ||
a18e9d1e | 1228 | /* Starts the worker thread */ |
bde89a9e | 1229 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1230 | { |
bde89a9e | 1231 | drbd_thread_start(&connection->worker); |
b5043c5e | 1232 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1233 | } |
1234 | ||
a18e9d1e | 1235 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1236 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1237 | { |
992d6e91 | 1238 | bool stop_threads; |
0500813f | 1239 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1240 | stop_threads = conn_all_vols_unconf(connection) && |
1241 | connection->cstate == C_STANDALONE; | |
0500813f | 1242 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1243 | if (stop_threads) { |
1244 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1245 | * in conn_disconnect() */ |
bde89a9e AG |
1246 | drbd_thread_stop(&connection->receiver); |
1247 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1248 | } |
b411b363 PR |
1249 | } |
1250 | ||
0778286a | 1251 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1252 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1253 | { |
1254 | int s = 0; | |
1255 | ||
b30ab791 | 1256 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1257 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1258 | return; |
1259 | } | |
1260 | ||
b30ab791 | 1261 | drbd_al_shrink(device); |
0500813f | 1262 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1263 | if (device->state.conn < C_CONNECTED) |
1264 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1265 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1266 | lc_unlock(device->act_log); |
0778286a PR |
1267 | |
1268 | if (s) | |
d0180171 | 1269 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1270 | } |
1271 | ||
5979e361 LE |
1272 | |
1273 | static bool should_set_defaults(struct genl_info *info) | |
1274 | { | |
1275 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1276 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1277 | } | |
1278 | ||
5bbcf5e6 | 1279 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1280 | { |
5bbcf5e6 LE |
1281 | /* This is limited by 16 bit "slot" numbers, |
1282 | * and by available on-disk context storage. | |
1283 | * | |
1284 | * Also (u16)~0 is special (denotes a "free" extent). | |
1285 | * | |
1286 | * One transaction occupies one 4kB on-disk block, | |
1287 | * we have n such blocks in the on disk ring buffer, | |
1288 | * the "current" transaction may fail (n-1), | |
1289 | * and there is 919 slot numbers context information per transaction. | |
1290 | * | |
1291 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1292 | * so cap there first. | |
1293 | */ | |
1294 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1295 | const unsigned int sufficient_on_disk = | |
1296 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1297 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1298 | |
5bbcf5e6 LE |
1299 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1300 | ||
1301 | if (al_size_4k > sufficient_on_disk) | |
1302 | return max_al_nr; | |
1303 | ||
1304 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1305 | } |
1306 | ||
70df7092 LE |
1307 | static bool write_ordering_changed(struct disk_conf *a, struct disk_conf *b) |
1308 | { | |
1309 | return a->disk_barrier != b->disk_barrier || | |
1310 | a->disk_flushes != b->disk_flushes || | |
1311 | a->disk_drain != b->disk_drain; | |
1312 | } | |
1313 | ||
f399002e LE |
1314 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1315 | { | |
a910b123 | 1316 | struct drbd_config_context adm_ctx; |
f399002e | 1317 | enum drbd_ret_code retcode; |
b30ab791 | 1318 | struct drbd_device *device; |
daeda1cc | 1319 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1320 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1321 | int err, fifo_size; |
f399002e | 1322 | |
a910b123 | 1323 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1324 | if (!adm_ctx.reply_skb) |
1325 | return retcode; | |
1326 | if (retcode != NO_ERROR) | |
9e276872 | 1327 | goto finish; |
f399002e | 1328 | |
b30ab791 | 1329 | device = adm_ctx.device; |
9e276872 | 1330 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1331 | |
1332 | /* we also need a disk | |
1333 | * to change the options on */ | |
b30ab791 | 1334 | if (!get_ldev(device)) { |
f399002e LE |
1335 | retcode = ERR_NO_DISK; |
1336 | goto out; | |
1337 | } | |
1338 | ||
daeda1cc | 1339 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1340 | if (!new_disk_conf) { |
f399002e LE |
1341 | retcode = ERR_NOMEM; |
1342 | goto fail; | |
1343 | } | |
1344 | ||
0500813f | 1345 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1346 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1347 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1348 | if (should_set_defaults(info)) |
b966b5dd | 1349 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1350 | |
5ecc72c3 | 1351 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1352 | if (err && err != -ENOMSG) { |
f399002e | 1353 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1354 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1355 | goto fail_unlock; |
f399002e LE |
1356 | } |
1357 | ||
5ecc72c3 LE |
1358 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1359 | new_disk_conf->resync_rate = 1; | |
f399002e | 1360 | |
5bbcf5e6 LE |
1361 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1362 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1363 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1364 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1365 | |
1366 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1367 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1368 | |
5ecc72c3 | 1369 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1370 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1371 | new_plan = fifo_alloc(fifo_size); |
1372 | if (!new_plan) { | |
d0180171 | 1373 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1374 | retcode = ERR_NOMEM; |
daeda1cc | 1375 | goto fail_unlock; |
f399002e LE |
1376 | } |
1377 | } | |
1378 | ||
b30ab791 AG |
1379 | drbd_suspend_io(device); |
1380 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1381 | drbd_al_shrink(device); | |
1382 | err = drbd_check_al_size(device, new_disk_conf); | |
1383 | lc_unlock(device->act_log); | |
1384 | wake_up(&device->al_wait); | |
1385 | drbd_resume_io(device); | |
f399002e LE |
1386 | |
1387 | if (err) { | |
1388 | retcode = ERR_NOMEM; | |
daeda1cc | 1389 | goto fail_unlock; |
f399002e LE |
1390 | } |
1391 | ||
dc97b708 | 1392 | write_lock_irq(&global_state_lock); |
b30ab791 | 1393 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1394 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1395 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1396 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1397 | } |
1398 | write_unlock_irq(&global_state_lock); | |
f399002e | 1399 | |
daeda1cc PR |
1400 | if (retcode != NO_ERROR) |
1401 | goto fail_unlock; | |
f399002e | 1402 | |
813472ce | 1403 | if (new_plan) { |
b30ab791 AG |
1404 | old_plan = device->rs_plan_s; |
1405 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1406 | } |
9958c857 | 1407 | |
0500813f | 1408 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1409 | |
9a51ab1c | 1410 | if (new_disk_conf->al_updates) |
b30ab791 | 1411 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1412 | else |
b30ab791 | 1413 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1414 | |
691631c0 | 1415 | if (new_disk_conf->md_flushes) |
b30ab791 | 1416 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1417 | else |
b30ab791 | 1418 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1419 | |
70df7092 LE |
1420 | if (write_ordering_changed(old_disk_conf, new_disk_conf)) |
1421 | drbd_bump_write_ordering(device->resource, NULL, WO_bdev_flush); | |
27eb13e9 | 1422 | |
b30ab791 | 1423 | drbd_md_sync(device); |
f399002e | 1424 | |
69a22773 AG |
1425 | if (device->state.conn >= C_CONNECTED) { |
1426 | struct drbd_peer_device *peer_device; | |
1427 | ||
1428 | for_each_peer_device(peer_device, device) | |
1429 | drbd_send_sync_param(peer_device); | |
1430 | } | |
f399002e | 1431 | |
daeda1cc PR |
1432 | synchronize_rcu(); |
1433 | kfree(old_disk_conf); | |
813472ce | 1434 | kfree(old_plan); |
b30ab791 | 1435 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1436 | goto success; |
1437 | ||
1438 | fail_unlock: | |
0500813f | 1439 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1440 | fail: |
5ecc72c3 | 1441 | kfree(new_disk_conf); |
813472ce | 1442 | kfree(new_plan); |
daeda1cc | 1443 | success: |
b30ab791 | 1444 | put_ldev(device); |
f399002e | 1445 | out: |
9e276872 LE |
1446 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1447 | finish: | |
a910b123 | 1448 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1449 | return 0; |
1450 | } | |
1451 | ||
3b98c0c2 | 1452 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1453 | { |
a910b123 | 1454 | struct drbd_config_context adm_ctx; |
b30ab791 | 1455 | struct drbd_device *device; |
44a4d551 LE |
1456 | struct drbd_peer_device *peer_device; |
1457 | struct drbd_connection *connection; | |
3b98c0c2 | 1458 | int err; |
116676ca | 1459 | enum drbd_ret_code retcode; |
b411b363 PR |
1460 | enum determine_dev_size dd; |
1461 | sector_t max_possible_sectors; | |
1462 | sector_t min_md_device_sectors; | |
1463 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1464 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1465 | struct block_device *bdev; |
b411b363 | 1466 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1467 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1468 | union drbd_state ns, os; |
f2024e7c | 1469 | enum drbd_state_rv rv; |
44ed167d | 1470 | struct net_conf *nc; |
b411b363 | 1471 | |
a910b123 | 1472 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1473 | if (!adm_ctx.reply_skb) |
1474 | return retcode; | |
1475 | if (retcode != NO_ERROR) | |
40cbf085 | 1476 | goto finish; |
b411b363 | 1477 | |
b30ab791 | 1478 | device = adm_ctx.device; |
9e276872 | 1479 | mutex_lock(&adm_ctx.resource->adm_mutex); |
44a4d551 LE |
1480 | peer_device = first_peer_device(device); |
1481 | connection = peer_device ? peer_device->connection : NULL; | |
1482 | conn_reconfig_start(connection); | |
b411b363 PR |
1483 | |
1484 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1485 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1486 | retcode = ERR_DISK_CONFIGURED; |
1487 | goto fail; | |
1488 | } | |
82f59cc6 LE |
1489 | /* It may just now have detached because of IO error. Make sure |
1490 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1491 | * e.g. if someone calls attach from the on-io-error handler, | |
1492 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
e334f550 | 1493 | wait_event(device->misc_wait, !test_bit(GOING_DISKLESS, &device->flags)); |
b411b363 | 1494 | |
383606e0 | 1495 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1496 | clear_bit(FORCE_DETACH, &device->flags); |
1497 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1498 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1499 | |
0029d624 | 1500 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1501 | device->rs_total = 0; |
1502 | device->rs_failed = 0; | |
1503 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1504 | |
3b98c0c2 | 1505 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1506 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1507 | if (!nbc) { | |
1508 | retcode = ERR_NOMEM; | |
1509 | goto fail; | |
1510 | } | |
9f2247bb PR |
1511 | spin_lock_init(&nbc->md.uuid_lock); |
1512 | ||
daeda1cc PR |
1513 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1514 | if (!new_disk_conf) { | |
1515 | retcode = ERR_NOMEM; | |
b411b363 PR |
1516 | goto fail; |
1517 | } | |
daeda1cc | 1518 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1519 | |
daeda1cc PR |
1520 | set_disk_conf_defaults(new_disk_conf); |
1521 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1522 | if (err) { |
b411b363 | 1523 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1524 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1525 | goto fail; |
1526 | } | |
1527 | ||
5bbcf5e6 LE |
1528 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1529 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1530 | |
9958c857 PR |
1531 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1532 | if (!new_plan) { | |
1533 | retcode = ERR_NOMEM; | |
1534 | goto fail; | |
1535 | } | |
1536 | ||
daeda1cc | 1537 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1538 | retcode = ERR_MD_IDX_INVALID; |
1539 | goto fail; | |
1540 | } | |
1541 | ||
a3f8f7dc | 1542 | write_lock_irq(&global_state_lock); |
b30ab791 | 1543 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1544 | write_unlock_irq(&global_state_lock); |
1545 | if (retcode != NO_ERROR) | |
1546 | goto fail; | |
1547 | ||
44ed167d | 1548 | rcu_read_lock(); |
44a4d551 | 1549 | nc = rcu_dereference(connection->net_conf); |
44ed167d | 1550 | if (nc) { |
daeda1cc | 1551 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1552 | rcu_read_unlock(); |
47ff2d0a PR |
1553 | retcode = ERR_STONITH_AND_PROT_A; |
1554 | goto fail; | |
1555 | } | |
1556 | } | |
44ed167d | 1557 | rcu_read_unlock(); |
47ff2d0a | 1558 | |
daeda1cc | 1559 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1560 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1561 | if (IS_ERR(bdev)) { |
d0180171 | 1562 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1563 | PTR_ERR(bdev)); |
b411b363 PR |
1564 | retcode = ERR_OPEN_DISK; |
1565 | goto fail; | |
1566 | } | |
e525fd89 TH |
1567 | nbc->backing_bdev = bdev; |
1568 | ||
1569 | /* | |
1570 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1571 | * drbd sharing one meta device. TODO in that case, paranoia | |
1572 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1573 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1574 | * should check it for you already; but if you don't, or | |
1575 | * someone fooled it, we need to double check here) | |
1576 | */ | |
daeda1cc | 1577 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1578 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1579 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1580 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1581 | if (IS_ERR(bdev)) { |
d0180171 | 1582 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1583 | PTR_ERR(bdev)); |
b411b363 PR |
1584 | retcode = ERR_OPEN_MD_DISK; |
1585 | goto fail; | |
1586 | } | |
e525fd89 | 1587 | nbc->md_bdev = bdev; |
b411b363 | 1588 | |
e525fd89 | 1589 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1590 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1591 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1592 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1593 | goto fail; |
1594 | } | |
1595 | ||
1596 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1597 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1598 | offsetof(struct bm_extent, lce)); |
1599 | if (!resync_lru) { | |
1600 | retcode = ERR_NOMEM; | |
e525fd89 | 1601 | goto fail; |
b411b363 PR |
1602 | } |
1603 | ||
c04ccaa6 LE |
1604 | /* Read our meta data super block early. |
1605 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1606 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1607 | if (retcode != NO_ERROR) |
1608 | goto fail; | |
b411b363 | 1609 | |
5bbcf5e6 LE |
1610 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1611 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1612 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1613 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1614 | ||
daeda1cc | 1615 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1616 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1617 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1618 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1619 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1620 | goto fail; |
b411b363 PR |
1621 | } |
1622 | ||
daeda1cc | 1623 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1624 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1625 | /* at least one MB, otherwise it does not make sense */ | |
1626 | min_md_device_sectors = (2<<10); | |
1627 | } else { | |
1628 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1629 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1630 | } |
1631 | ||
b411b363 | 1632 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1633 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1634 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1635 | "at least %llu sectors needed for this meta-disk type\n", |
1636 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1637 | goto fail; |
b411b363 PR |
1638 | } |
1639 | ||
1640 | /* Make sure the new disk is big enough | |
1641 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1642 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1643 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1644 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1645 | goto fail; |
b411b363 PR |
1646 | } |
1647 | ||
1648 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1649 | ||
1352994b | 1650 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1651 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1652 | "to currently maximum possible %llu sectors <==\n", |
1653 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1654 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1655 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1656 | "meta data may help <<==\n"); |
1657 | } | |
1658 | ||
b30ab791 | 1659 | drbd_suspend_io(device); |
b411b363 | 1660 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1661 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1662 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1663 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1664 | * As barriers are counted per resource, | |
1665 | * we'd need to suspend io on all devices of a resource. | |
1666 | */ | |
b30ab791 | 1667 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1668 | /* and for any other previously queued work */ |
44a4d551 | 1669 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 1670 | |
b30ab791 | 1671 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1672 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1673 | drbd_resume_io(device); |
f2024e7c | 1674 | if (rv < SS_SUCCESS) |
e525fd89 | 1675 | goto fail; |
b411b363 | 1676 | |
b30ab791 | 1677 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1678 | goto force_diskless; |
1679 | ||
b30ab791 AG |
1680 | if (!device->bitmap) { |
1681 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1682 | retcode = ERR_NOMEM; |
1683 | goto force_diskless_dec; | |
1684 | } | |
1685 | } | |
1686 | ||
b30ab791 | 1687 | if (device->state.conn < C_CONNECTED && |
babea49e | 1688 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1689 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1690 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1691 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1692 | retcode = ERR_DATA_NOT_CURRENT; |
1693 | goto force_diskless_dec; | |
1694 | } | |
1695 | ||
1696 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1697 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1698 | retcode = ERR_NOMEM; |
1699 | goto force_diskless_dec; | |
1700 | } | |
1701 | ||
1702 | /* Prevent shrinking of consistent devices ! */ | |
1703 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1704 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1705 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1706 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1707 | goto force_diskless_dec; |
1708 | } | |
1709 | ||
b411b363 PR |
1710 | /* Reset the "barriers don't work" bits here, then force meta data to |
1711 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1712 | if (new_disk_conf->md_flushes) |
b30ab791 | 1713 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1714 | else |
b30ab791 | 1715 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1716 | |
1717 | /* Point of no return reached. | |
1718 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1719 | * now device takes over responsibility, and the state engine should |
b411b363 | 1720 | * clean it up somewhere. */ |
0b0ba1ef | 1721 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1722 | device->ldev = nbc; |
1723 | device->resync = resync_lru; | |
1724 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1725 | nbc = NULL; |
1726 | resync_lru = NULL; | |
daeda1cc | 1727 | new_disk_conf = NULL; |
9958c857 | 1728 | new_plan = NULL; |
b411b363 | 1729 | |
8fe39aac | 1730 | drbd_bump_write_ordering(device->resource, device->ldev, WO_bdev_flush); |
b411b363 | 1731 | |
b30ab791 AG |
1732 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1733 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1734 | else |
b30ab791 | 1735 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1736 | |
b30ab791 | 1737 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1738 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1739 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1740 | |
b30ab791 AG |
1741 | device->send_cnt = 0; |
1742 | device->recv_cnt = 0; | |
1743 | device->read_cnt = 0; | |
1744 | device->writ_cnt = 0; | |
b411b363 | 1745 | |
8fe39aac | 1746 | drbd_reconsider_max_bio_size(device, device->ldev); |
b411b363 PR |
1747 | |
1748 | /* If I am currently not R_PRIMARY, | |
1749 | * but meta data primary indicator is set, | |
1750 | * I just now recover from a hard crash, | |
1751 | * and have been R_PRIMARY before that crash. | |
1752 | * | |
1753 | * Now, if I had no connection before that crash | |
1754 | * (have been degraded R_PRIMARY), chances are that | |
1755 | * I won't find my peer now either. | |
1756 | * | |
1757 | * In that case, and _only_ in that case, | |
1758 | * we use the degr-wfc-timeout instead of the default, | |
1759 | * so we can automatically recover from a crash of a | |
1760 | * degraded but active "cluster" after a certain timeout. | |
1761 | */ | |
b30ab791 AG |
1762 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1763 | if (device->state.role != R_PRIMARY && | |
1764 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1765 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1766 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1767 | |
b30ab791 | 1768 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1769 | if (dd <= DS_ERROR) { |
b411b363 PR |
1770 | retcode = ERR_NOMEM_BITMAP; |
1771 | goto force_diskless_dec; | |
e96c9633 | 1772 | } else if (dd == DS_GREW) |
b30ab791 | 1773 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1774 | |
b30ab791 AG |
1775 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1776 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1777 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1778 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1779 | "(aka FullSync)\n"); |
b30ab791 | 1780 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1781 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1782 | retcode = ERR_IO_MD_DISK; |
1783 | goto force_diskless_dec; | |
1784 | } | |
1785 | } else { | |
b30ab791 | 1786 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1787 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1788 | retcode = ERR_IO_MD_DISK; |
1789 | goto force_diskless_dec; | |
1790 | } | |
b411b363 PR |
1791 | } |
1792 | ||
b30ab791 AG |
1793 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1794 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1795 | |
0500813f | 1796 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1797 | os = drbd_read_state(device); |
78bae59b | 1798 | ns = os; |
b411b363 PR |
1799 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1800 | otherwise investigate MDF_WasUpToDate... | |
1801 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1802 | otherwise into D_CONSISTENT state. | |
1803 | */ | |
b30ab791 AG |
1804 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1805 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1806 | ns.disk = D_CONSISTENT; |
1807 | else | |
1808 | ns.disk = D_OUTDATED; | |
1809 | } else { | |
1810 | ns.disk = D_INCONSISTENT; | |
1811 | } | |
1812 | ||
b30ab791 | 1813 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1814 | ns.pdsk = D_OUTDATED; |
1815 | ||
daeda1cc PR |
1816 | rcu_read_lock(); |
1817 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1818 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1819 | ns.disk = D_UP_TO_DATE; |
1820 | ||
1821 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1822 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1823 | this point, because drbd_request_state() modifies these | |
1824 | flags. */ | |
1825 | ||
b30ab791 AG |
1826 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1827 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1828 | else |
b30ab791 | 1829 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1830 | |
1831 | rcu_read_unlock(); | |
1832 | ||
b411b363 PR |
1833 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1834 | state after the negotiation phase. */ | |
b30ab791 AG |
1835 | if (device->state.conn == C_CONNECTED) { |
1836 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1837 | ns.i = os.i; |
1838 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1839 | |
1840 | /* We expect to receive up-to-date UUIDs soon. | |
1841 | To avoid a race in receive_state, free p_uuid while | |
1842 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1843 | kfree(device->p_uuid); |
1844 | device->p_uuid = NULL; | |
b411b363 PR |
1845 | } |
1846 | ||
b30ab791 | 1847 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1848 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1849 | |
1850 | if (rv < SS_SUCCESS) | |
1851 | goto force_diskless_dec; | |
1852 | ||
b30ab791 | 1853 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1854 | |
b30ab791 AG |
1855 | if (device->state.role == R_PRIMARY) |
1856 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1857 | else |
b30ab791 | 1858 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1859 | |
b30ab791 AG |
1860 | drbd_md_mark_dirty(device); |
1861 | drbd_md_sync(device); | |
b411b363 | 1862 | |
b30ab791 AG |
1863 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1864 | put_ldev(device); | |
44a4d551 | 1865 | conn_reconfig_done(connection); |
9e276872 | 1866 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 1867 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1868 | return 0; |
1869 | ||
1870 | force_diskless_dec: | |
b30ab791 | 1871 | put_ldev(device); |
b411b363 | 1872 | force_diskless: |
b30ab791 AG |
1873 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1874 | drbd_md_sync(device); | |
b411b363 | 1875 | fail: |
44a4d551 | 1876 | conn_reconfig_done(connection); |
b411b363 | 1877 | if (nbc) { |
e525fd89 TH |
1878 | if (nbc->backing_bdev) |
1879 | blkdev_put(nbc->backing_bdev, | |
1880 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1881 | if (nbc->md_bdev) | |
1882 | blkdev_put(nbc->md_bdev, | |
1883 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1884 | kfree(nbc); |
1885 | } | |
daeda1cc | 1886 | kfree(new_disk_conf); |
b411b363 | 1887 | lc_destroy(resync_lru); |
9958c857 | 1888 | kfree(new_plan); |
9e276872 | 1889 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 1890 | finish: |
a910b123 | 1891 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1892 | return 0; |
1893 | } | |
1894 | ||
b30ab791 | 1895 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1896 | { |
19f83c76 | 1897 | enum drbd_state_rv retcode; |
9a0d9d03 | 1898 | int ret; |
02ee8f95 | 1899 | |
cdfda633 | 1900 | if (force) { |
b30ab791 AG |
1901 | set_bit(FORCE_DETACH, &device->flags); |
1902 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1903 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1904 | goto out; |
1905 | } | |
1906 | ||
b30ab791 | 1907 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
e37d2438 | 1908 | drbd_md_get_buffer(device, __func__); /* make sure there is no in-flight meta-data IO */ |
b30ab791 AG |
1909 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); |
1910 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1911 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1912 | ret = wait_event_interruptible(device->misc_wait, |
1913 | device->state.disk != D_FAILED); | |
1914 | drbd_resume_io(device); | |
9b2f61ae | 1915 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1916 | retcode = SS_NOTHING_TO_DO; |
1917 | if (ret) | |
1918 | retcode = ERR_INTR; | |
02ee8f95 | 1919 | out: |
85f75dd7 | 1920 | return retcode; |
b411b363 PR |
1921 | } |
1922 | ||
82f59cc6 LE |
1923 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1924 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1925 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1926 | * internal references as well. | |
1927 | * Only then we have finally detached. */ | |
3b98c0c2 | 1928 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1929 | { |
a910b123 | 1930 | struct drbd_config_context adm_ctx; |
116676ca | 1931 | enum drbd_ret_code retcode; |
cdfda633 PR |
1932 | struct detach_parms parms = { }; |
1933 | int err; | |
b411b363 | 1934 | |
a910b123 | 1935 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1936 | if (!adm_ctx.reply_skb) |
1937 | return retcode; | |
1938 | if (retcode != NO_ERROR) | |
1939 | goto out; | |
b411b363 | 1940 | |
cdfda633 PR |
1941 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1942 | err = detach_parms_from_attrs(&parms, info); | |
1943 | if (err) { | |
1944 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 1945 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
1946 | goto out; |
1947 | } | |
b411b363 PR |
1948 | } |
1949 | ||
9e276872 | 1950 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 1951 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 1952 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 1953 | out: |
a910b123 | 1954 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1955 | return 0; |
1956 | } | |
b411b363 | 1957 | |
bde89a9e | 1958 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1959 | { |
c06ece6b | 1960 | struct drbd_peer_device *peer_device; |
695d08fa | 1961 | bool rv = false; |
f399002e LE |
1962 | int vnr; |
1963 | ||
695d08fa | 1964 | rcu_read_lock(); |
c06ece6b AG |
1965 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1966 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1967 | if (device->state.conn == C_SYNC_SOURCE || |
1968 | device->state.conn == C_SYNC_TARGET || | |
1969 | device->state.conn == C_PAUSED_SYNC_S || | |
1970 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1971 | rv = true; |
1972 | break; | |
1973 | } | |
b411b363 | 1974 | } |
695d08fa | 1975 | rcu_read_unlock(); |
b411b363 | 1976 | |
695d08fa | 1977 | return rv; |
f399002e | 1978 | } |
47ff2d0a | 1979 | |
bde89a9e | 1980 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1981 | { |
c06ece6b | 1982 | struct drbd_peer_device *peer_device; |
695d08fa | 1983 | bool rv = false; |
f399002e LE |
1984 | int vnr; |
1985 | ||
695d08fa | 1986 | rcu_read_lock(); |
c06ece6b AG |
1987 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1988 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1989 | if (device->state.conn == C_VERIFY_S || |
1990 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1991 | rv = true; |
1992 | break; | |
47ff2d0a PR |
1993 | } |
1994 | } | |
695d08fa | 1995 | rcu_read_unlock(); |
b411b363 | 1996 | |
695d08fa | 1997 | return rv; |
f399002e | 1998 | } |
422028b1 | 1999 | |
cd64397c | 2000 | static enum drbd_ret_code |
270eb5c9 | 2001 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 2002 | { |
c06ece6b | 2003 | struct drbd_peer_device *peer_device; |
cd64397c | 2004 | int i; |
b411b363 | 2005 | |
270eb5c9 AG |
2006 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
2007 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 2008 | return ERR_NEED_APV_100; |
b411b363 | 2009 | |
270eb5c9 | 2010 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
2011 | return ERR_NEED_APV_100; |
2012 | ||
270eb5c9 | 2013 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 2014 | return ERR_NEED_APV_100; |
b411b363 PR |
2015 | } |
2016 | ||
270eb5c9 | 2017 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
2018 | conn_highest_role(connection) == R_PRIMARY && |
2019 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 2020 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 2021 | |
270eb5c9 AG |
2022 | if (new_net_conf->two_primaries && |
2023 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2024 | return ERR_NOT_PROTO_C; |
2025 | ||
c06ece6b AG |
2026 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2027 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2028 | if (get_ldev(device)) { |
2029 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2030 | put_ldev(device); | |
270eb5c9 | 2031 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2032 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2033 | } |
270eb5c9 | 2034 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2035 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2036 | } |
2037 | ||
270eb5c9 | 2038 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2039 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2040 | |
cd64397c PR |
2041 | return NO_ERROR; |
2042 | } | |
b411b363 | 2043 | |
44ed167d | 2044 | static enum drbd_ret_code |
270eb5c9 | 2045 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
2046 | { |
2047 | static enum drbd_ret_code rv; | |
c06ece6b | 2048 | struct drbd_peer_device *peer_device; |
44ed167d | 2049 | int i; |
b411b363 | 2050 | |
44ed167d | 2051 | rcu_read_lock(); |
270eb5c9 | 2052 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2053 | rcu_read_unlock(); |
b411b363 | 2054 | |
179e20b8 | 2055 | /* connection->peer_devices protected by genl_lock() here */ |
c06ece6b AG |
2056 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2057 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2058 | if (!device->bitmap) { |
2059 | if (drbd_bm_init(device)) | |
44ed167d | 2060 | return ERR_NOMEM; |
b411b363 PR |
2061 | } |
2062 | } | |
2063 | ||
44ed167d PR |
2064 | return rv; |
2065 | } | |
b411b363 | 2066 | |
0fd0ea06 PR |
2067 | struct crypto { |
2068 | struct crypto_hash *verify_tfm; | |
2069 | struct crypto_hash *csums_tfm; | |
2070 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2071 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2072 | }; |
b411b363 | 2073 | |
0fd0ea06 | 2074 | static int |
4b6ad6d4 | 2075 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2076 | { |
2077 | if (!tfm_name[0]) | |
2078 | return NO_ERROR; | |
b411b363 | 2079 | |
0fd0ea06 PR |
2080 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2081 | if (IS_ERR(*tfm)) { | |
2082 | *tfm = NULL; | |
2083 | return err_alg; | |
b411b363 | 2084 | } |
b411b363 | 2085 | |
0fd0ea06 PR |
2086 | return NO_ERROR; |
2087 | } | |
b411b363 | 2088 | |
0fd0ea06 | 2089 | static enum drbd_ret_code |
270eb5c9 | 2090 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2091 | { |
2092 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2093 | enum drbd_ret_code rv; | |
0fd0ea06 | 2094 | |
270eb5c9 | 2095 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2096 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2097 | if (rv != NO_ERROR) |
2098 | return rv; | |
270eb5c9 | 2099 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2100 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2101 | if (rv != NO_ERROR) |
2102 | return rv; | |
270eb5c9 | 2103 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2104 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2105 | if (rv != NO_ERROR) |
2106 | return rv; | |
270eb5c9 | 2107 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2108 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2109 | new_net_conf->cram_hmac_alg); |
b411b363 | 2110 | |
4b6ad6d4 AG |
2111 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2112 | ERR_AUTH_ALG); | |
b411b363 PR |
2113 | } |
2114 | ||
0fd0ea06 PR |
2115 | return rv; |
2116 | } | |
b411b363 | 2117 | |
0fd0ea06 PR |
2118 | static void free_crypto(struct crypto *crypto) |
2119 | { | |
0fd0ea06 | 2120 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2121 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2122 | crypto_free_hash(crypto->csums_tfm); |
2123 | crypto_free_hash(crypto->verify_tfm); | |
2124 | } | |
b411b363 | 2125 | |
f399002e LE |
2126 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2127 | { | |
a910b123 | 2128 | struct drbd_config_context adm_ctx; |
f399002e | 2129 | enum drbd_ret_code retcode; |
bde89a9e | 2130 | struct drbd_connection *connection; |
270eb5c9 | 2131 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2132 | int err; |
2133 | int ovr; /* online verify running */ | |
2134 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2135 | struct crypto crypto = { }; |
b411b363 | 2136 | |
a910b123 | 2137 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2138 | if (!adm_ctx.reply_skb) |
2139 | return retcode; | |
2140 | if (retcode != NO_ERROR) | |
9e276872 | 2141 | goto finish; |
b411b363 | 2142 | |
bde89a9e | 2143 | connection = adm_ctx.connection; |
9e276872 | 2144 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2145 | |
270eb5c9 AG |
2146 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2147 | if (!new_net_conf) { | |
f399002e LE |
2148 | retcode = ERR_NOMEM; |
2149 | goto out; | |
2150 | } | |
b411b363 | 2151 | |
bde89a9e | 2152 | conn_reconfig_start(connection); |
b411b363 | 2153 | |
bde89a9e | 2154 | mutex_lock(&connection->data.mutex); |
0500813f | 2155 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2156 | old_net_conf = connection->net_conf; |
2561b9c1 | 2157 | |
270eb5c9 | 2158 | if (!old_net_conf) { |
a910b123 | 2159 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2160 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2161 | goto fail; |
2162 | } | |
2163 | ||
270eb5c9 | 2164 | *new_net_conf = *old_net_conf; |
5979e361 | 2165 | if (should_set_defaults(info)) |
270eb5c9 | 2166 | set_net_conf_defaults(new_net_conf); |
f399002e | 2167 | |
270eb5c9 | 2168 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2169 | if (err && err != -ENOMSG) { |
f399002e | 2170 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2171 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2172 | goto fail; |
2561b9c1 | 2173 | } |
b411b363 | 2174 | |
270eb5c9 | 2175 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2176 | if (retcode != NO_ERROR) |
2177 | goto fail; | |
b411b363 | 2178 | |
f399002e | 2179 | /* re-sync running */ |
bde89a9e | 2180 | rsr = conn_resync_running(connection); |
270eb5c9 | 2181 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2182 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2183 | goto fail; |
b411b363 PR |
2184 | } |
2185 | ||
f399002e | 2186 | /* online verify running */ |
bde89a9e | 2187 | ovr = conn_ov_running(connection); |
270eb5c9 | 2188 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2189 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2190 | goto fail; |
f399002e | 2191 | } |
b411b363 | 2192 | |
270eb5c9 | 2193 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2194 | if (retcode != NO_ERROR) |
b411b363 | 2195 | goto fail; |
f399002e | 2196 | |
270eb5c9 | 2197 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2198 | |
2199 | if (!rsr) { | |
bde89a9e AG |
2200 | crypto_free_hash(connection->csums_tfm); |
2201 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2202 | crypto.csums_tfm = NULL; |
f399002e LE |
2203 | } |
2204 | if (!ovr) { | |
bde89a9e AG |
2205 | crypto_free_hash(connection->verify_tfm); |
2206 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2207 | crypto.verify_tfm = NULL; |
b411b363 PR |
2208 | } |
2209 | ||
bde89a9e AG |
2210 | crypto_free_hash(connection->integrity_tfm); |
2211 | connection->integrity_tfm = crypto.integrity_tfm; | |
2212 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2213 | /* Do this without trying to take connection->data.mutex again. */ | |
2214 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2215 | |
bde89a9e AG |
2216 | crypto_free_hash(connection->cram_hmac_tfm); |
2217 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2218 | |
0500813f | 2219 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2220 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2221 | synchronize_rcu(); |
270eb5c9 | 2222 | kfree(old_net_conf); |
91fd4dad | 2223 | |
69a22773 AG |
2224 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2225 | struct drbd_peer_device *peer_device; | |
2226 | int vnr; | |
2227 | ||
2228 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2229 | drbd_send_sync_param(peer_device); | |
2230 | } | |
f399002e | 2231 | |
91fd4dad PR |
2232 | goto done; |
2233 | ||
b411b363 | 2234 | fail: |
0500813f | 2235 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2236 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2237 | free_crypto(&crypto); |
270eb5c9 | 2238 | kfree(new_net_conf); |
91fd4dad | 2239 | done: |
bde89a9e | 2240 | conn_reconfig_done(connection); |
f399002e | 2241 | out: |
9e276872 LE |
2242 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2243 | finish: | |
a910b123 | 2244 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2245 | return 0; |
2246 | } | |
2247 | ||
3b98c0c2 | 2248 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2249 | { |
a910b123 | 2250 | struct drbd_config_context adm_ctx; |
c06ece6b | 2251 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2252 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2253 | struct crypto crypto = { }; |
77c556f6 | 2254 | struct drbd_resource *resource; |
bde89a9e | 2255 | struct drbd_connection *connection; |
3b98c0c2 LE |
2256 | enum drbd_ret_code retcode; |
2257 | int i; | |
2258 | int err; | |
b411b363 | 2259 | |
a910b123 | 2260 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2261 | |
3b98c0c2 LE |
2262 | if (!adm_ctx.reply_skb) |
2263 | return retcode; | |
2264 | if (retcode != NO_ERROR) | |
2265 | goto out; | |
089c075d | 2266 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2267 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2268 | retcode = ERR_INVALID_REQUEST; |
2269 | goto out; | |
2270 | } | |
b411b363 | 2271 | |
089c075d AG |
2272 | /* No need for _rcu here. All reconfiguration is |
2273 | * strictly serialized on genl_lock(). We are protected against | |
2274 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2275 | for_each_resource(resource, &drbd_resources) { |
2276 | for_each_connection(connection, resource) { | |
2277 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2278 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2279 | connection->my_addr_len)) { | |
2280 | retcode = ERR_LOCAL_ADDR; | |
2281 | goto out; | |
2282 | } | |
b411b363 | 2283 | |
77c556f6 AG |
2284 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2285 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2286 | connection->peer_addr_len)) { | |
2287 | retcode = ERR_PEER_ADDR; | |
2288 | goto out; | |
2289 | } | |
089c075d | 2290 | } |
b411b363 PR |
2291 | } |
2292 | ||
9e276872 | 2293 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2294 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2295 | conn_reconfig_start(connection); |
b411b363 | 2296 | |
bde89a9e | 2297 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2298 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2299 | goto fail; |
2300 | } | |
2301 | ||
a209b4ae | 2302 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2303 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2304 | if (!new_net_conf) { | |
b411b363 | 2305 | retcode = ERR_NOMEM; |
b411b363 PR |
2306 | goto fail; |
2307 | } | |
2308 | ||
270eb5c9 | 2309 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2310 | |
270eb5c9 | 2311 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2312 | if (err && err != -ENOMSG) { |
b411b363 | 2313 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2314 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2315 | goto fail; |
2316 | } | |
2317 | ||
270eb5c9 | 2318 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2319 | if (retcode != NO_ERROR) |
422028b1 | 2320 | goto fail; |
b411b363 | 2321 | |
270eb5c9 | 2322 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2323 | if (retcode != NO_ERROR) |
2324 | goto fail; | |
b411b363 | 2325 | |
270eb5c9 | 2326 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2327 | |
b5043c5e | 2328 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2329 | |
0500813f | 2330 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2331 | old_net_conf = connection->net_conf; |
2332 | if (old_net_conf) { | |
b411b363 | 2333 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2334 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2335 | goto fail; |
2336 | } | |
270eb5c9 | 2337 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2338 | |
bde89a9e AG |
2339 | conn_free_crypto(connection); |
2340 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2341 | connection->integrity_tfm = crypto.integrity_tfm; | |
2342 | connection->csums_tfm = crypto.csums_tfm; | |
2343 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2344 | |
bde89a9e AG |
2345 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2346 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2347 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2348 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2349 | |
0500813f | 2350 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2351 | |
695d08fa | 2352 | rcu_read_lock(); |
c06ece6b AG |
2353 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2354 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2355 | device->send_cnt = 0; |
2356 | device->recv_cnt = 0; | |
b411b363 | 2357 | } |
695d08fa | 2358 | rcu_read_unlock(); |
b411b363 | 2359 | |
bde89a9e | 2360 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2361 | |
bde89a9e | 2362 | conn_reconfig_done(connection); |
9e276872 | 2363 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2364 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2365 | return 0; |
b411b363 | 2366 | |
b411b363 | 2367 | fail: |
0fd0ea06 | 2368 | free_crypto(&crypto); |
270eb5c9 | 2369 | kfree(new_net_conf); |
b411b363 | 2370 | |
bde89a9e | 2371 | conn_reconfig_done(connection); |
9e276872 | 2372 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2373 | out: |
a910b123 | 2374 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2375 | return 0; |
2376 | } | |
2377 | ||
bde89a9e | 2378 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2379 | { |
2380 | enum drbd_state_rv rv; | |
85f75dd7 | 2381 | |
bde89a9e | 2382 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2383 | force ? CS_HARD : 0); |
85f75dd7 LE |
2384 | |
2385 | switch (rv) { | |
2386 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2387 | break; |
85f75dd7 LE |
2388 | case SS_ALREADY_STANDALONE: |
2389 | return SS_SUCCESS; | |
2390 | case SS_PRIMARY_NOP: | |
2391 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2392 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2393 | |
2394 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2395 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2396 | |
85f75dd7 LE |
2397 | break; |
2398 | case SS_CW_FAILED_BY_PEER: | |
2399 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2400 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2401 | disk, D_OUTDATED), 0); |
2402 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2403 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2404 | CS_HARD); |
b411b363 | 2405 | } |
85f75dd7 LE |
2406 | break; |
2407 | default:; | |
2408 | /* no special handling necessary */ | |
2409 | } | |
2410 | ||
f3dfa40a LE |
2411 | if (rv >= SS_SUCCESS) { |
2412 | enum drbd_state_rv rv2; | |
2413 | /* No one else can reconfigure the network while I am here. | |
2414 | * The state handling only uses drbd_thread_stop_nowait(), | |
2415 | * we want to really wait here until the receiver is no more. | |
2416 | */ | |
9693da23 | 2417 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2418 | |
2419 | /* Race breaker. This additional state change request may be | |
2420 | * necessary, if this was a forced disconnect during a receiver | |
2421 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2422 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2423 | * C_STANDALONE already, now, and this becomes a no-op. |
2424 | */ | |
bde89a9e | 2425 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2426 | CS_VERBOSE | CS_HARD); |
2427 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2428 | drbd_err(connection, |
f3dfa40a LE |
2429 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2430 | rv2); | |
b411b363 | 2431 | } |
85f75dd7 LE |
2432 | return rv; |
2433 | } | |
b411b363 | 2434 | |
3b98c0c2 | 2435 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2436 | { |
a910b123 | 2437 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2438 | struct disconnect_parms parms; |
bde89a9e | 2439 | struct drbd_connection *connection; |
85f75dd7 | 2440 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2441 | enum drbd_ret_code retcode; |
2442 | int err; | |
2561b9c1 | 2443 | |
a910b123 | 2444 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2445 | if (!adm_ctx.reply_skb) |
2446 | return retcode; | |
2447 | if (retcode != NO_ERROR) | |
2561b9c1 | 2448 | goto fail; |
b411b363 | 2449 | |
bde89a9e | 2450 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2451 | memset(&parms, 0, sizeof(parms)); |
2452 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2453 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2454 | if (err) { |
2455 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2456 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2457 | goto fail; |
2458 | } | |
2459 | } | |
2460 | ||
9e276872 | 2461 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2462 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2463 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2464 | retcode = rv; /* FIXME: Type mismatch. */ |
2465 | else | |
2466 | retcode = NO_ERROR; | |
9e276872 | 2467 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2468 | fail: |
a910b123 | 2469 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2470 | return 0; |
2471 | } | |
2472 | ||
b30ab791 | 2473 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2474 | { |
2475 | int iass; /* I am sync source */ | |
2476 | ||
d0180171 | 2477 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2478 | if (device->state.role != device->state.peer) |
2479 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2480 | else |
a6b32bc3 | 2481 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2482 | |
2483 | if (iass) | |
b30ab791 | 2484 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2485 | else |
b30ab791 | 2486 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2487 | } |
2488 | ||
3b98c0c2 | 2489 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2490 | { |
a910b123 | 2491 | struct drbd_config_context adm_ctx; |
daeda1cc | 2492 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2493 | struct resize_parms rs; |
b30ab791 | 2494 | struct drbd_device *device; |
3b98c0c2 | 2495 | enum drbd_ret_code retcode; |
b411b363 | 2496 | enum determine_dev_size dd; |
d752b269 | 2497 | bool change_al_layout = false; |
6495d2c6 | 2498 | enum dds_flags ddsf; |
daeda1cc | 2499 | sector_t u_size; |
3b98c0c2 | 2500 | int err; |
b411b363 | 2501 | |
a910b123 | 2502 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2503 | if (!adm_ctx.reply_skb) |
2504 | return retcode; | |
2505 | if (retcode != NO_ERROR) | |
9e276872 | 2506 | goto finish; |
3b98c0c2 | 2507 | |
9e276872 | 2508 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2509 | device = adm_ctx.device; |
2510 | if (!get_ldev(device)) { | |
d752b269 PR |
2511 | retcode = ERR_NO_DISK; |
2512 | goto fail; | |
2513 | } | |
2514 | ||
3b98c0c2 | 2515 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2516 | rs.al_stripes = device->ldev->md.al_stripes; |
2517 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2518 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2519 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2520 | if (err) { |
3b98c0c2 | 2521 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2522 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2523 | goto fail_ldev; |
b411b363 PR |
2524 | } |
2525 | } | |
2526 | ||
b30ab791 | 2527 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2528 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2529 | goto fail_ldev; |
b411b363 | 2530 | } |
b411b363 | 2531 | |
b30ab791 AG |
2532 | if (device->state.role == R_SECONDARY && |
2533 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2534 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2535 | goto fail_ldev; |
b411b363 | 2536 | } |
b411b363 | 2537 | |
a6b32bc3 | 2538 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2539 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2540 | goto fail_ldev; |
6495d2c6 PR |
2541 | } |
2542 | ||
daeda1cc | 2543 | rcu_read_lock(); |
b30ab791 | 2544 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2545 | rcu_read_unlock(); |
2546 | if (u_size != (sector_t)rs.resize_size) { | |
2547 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2548 | if (!new_disk_conf) { | |
778f271d | 2549 | retcode = ERR_NOMEM; |
9bcd2521 | 2550 | goto fail_ldev; |
778f271d PR |
2551 | } |
2552 | } | |
2553 | ||
b30ab791 AG |
2554 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2555 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2556 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2557 | ||
2558 | if (al_size_k > (16 * 1024 * 1024)) { | |
2559 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2560 | goto fail_ldev; | |
2561 | } | |
2562 | ||
2563 | if (al_size_k < MD_32kB_SECT/2) { | |
2564 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2565 | goto fail_ldev; | |
2566 | } | |
2567 | ||
cdc6af8d | 2568 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2569 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2570 | goto fail_ldev; | |
2571 | } | |
2572 | ||
2573 | change_al_layout = true; | |
2574 | } | |
2575 | ||
b30ab791 AG |
2576 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2577 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2578 | |
daeda1cc | 2579 | if (new_disk_conf) { |
0500813f | 2580 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2581 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2582 | *new_disk_conf = *old_disk_conf; |
2583 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2584 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2585 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2586 | synchronize_rcu(); |
2587 | kfree(old_disk_conf); | |
b411b363 PR |
2588 | } |
2589 | ||
6495d2c6 | 2590 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2591 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2592 | drbd_md_sync(device); | |
2593 | put_ldev(device); | |
e96c9633 | 2594 | if (dd == DS_ERROR) { |
b411b363 PR |
2595 | retcode = ERR_NOMEM_BITMAP; |
2596 | goto fail; | |
d752b269 PR |
2597 | } else if (dd == DS_ERROR_SPACE_MD) { |
2598 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2599 | goto fail; | |
2600 | } else if (dd == DS_ERROR_SHRINK) { | |
2601 | retcode = ERR_IMPLICIT_SHRINK; | |
2602 | goto fail; | |
b411b363 | 2603 | } |
778f271d | 2604 | |
b30ab791 | 2605 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2606 | if (dd == DS_GREW) |
b30ab791 | 2607 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2608 | |
69a22773 AG |
2609 | drbd_send_uuids(first_peer_device(device)); |
2610 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2611 | } |
2612 | ||
b411b363 | 2613 | fail: |
9e276872 LE |
2614 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2615 | finish: | |
a910b123 | 2616 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2617 | return 0; |
b411b363 | 2618 | |
9bcd2521 | 2619 | fail_ldev: |
b30ab791 | 2620 | put_ldev(device); |
9bcd2521 | 2621 | goto fail; |
b411b363 | 2622 | } |
b411b363 | 2623 | |
f399002e | 2624 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2625 | { |
a910b123 | 2626 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2627 | enum drbd_ret_code retcode; |
b57a1e27 | 2628 | struct res_opts res_opts; |
f399002e | 2629 | int err; |
b411b363 | 2630 | |
a910b123 | 2631 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2632 | if (!adm_ctx.reply_skb) |
2633 | return retcode; | |
2634 | if (retcode != NO_ERROR) | |
2635 | goto fail; | |
b411b363 | 2636 | |
eb6bea67 | 2637 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2638 | if (should_set_defaults(info)) |
b966b5dd | 2639 | set_res_opts_defaults(&res_opts); |
b411b363 | 2640 | |
b57a1e27 | 2641 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2642 | if (err && err != -ENOMSG) { |
b411b363 | 2643 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2644 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2645 | goto fail; |
2646 | } | |
2647 | ||
9e276872 | 2648 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2649 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2650 | if (err) { |
2651 | retcode = ERR_INVALID_REQUEST; | |
2652 | if (err == -ENOMEM) | |
2653 | retcode = ERR_NOMEM; | |
b411b363 | 2654 | } |
9e276872 | 2655 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2656 | |
b411b363 | 2657 | fail: |
a910b123 | 2658 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2659 | return 0; |
2660 | } | |
2661 | ||
3b98c0c2 | 2662 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2663 | { |
a910b123 | 2664 | struct drbd_config_context adm_ctx; |
b30ab791 | 2665 | struct drbd_device *device; |
3b98c0c2 LE |
2666 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2667 | ||
a910b123 | 2668 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2669 | if (!adm_ctx.reply_skb) |
2670 | return retcode; | |
2671 | if (retcode != NO_ERROR) | |
2672 | goto out; | |
2673 | ||
b30ab791 | 2674 | device = adm_ctx.device; |
8fe39aac PR |
2675 | if (!get_ldev(device)) { |
2676 | retcode = ERR_NO_DISK; | |
2677 | goto out; | |
2678 | } | |
2679 | ||
2680 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2681 | |
194bfb32 | 2682 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2683 | * resync just being finished, wait for it before requesting a new resync. |
2684 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2685 | drbd_suspend_io(device); |
2686 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2687 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2688 | |
0b2dafcd PR |
2689 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2690 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2691 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2692 | */ |
b30ab791 AG |
2693 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2694 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2695 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2696 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2697 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2698 | retcode = ERR_IO_MD_DISK; | |
2699 | } | |
2700 | } else | |
b30ab791 AG |
2701 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2702 | drbd_resume_io(device); | |
9e276872 | 2703 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2704 | put_ldev(device); |
3b98c0c2 | 2705 | out: |
a910b123 | 2706 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2707 | return 0; |
2708 | } | |
2709 | ||
3b98c0c2 LE |
2710 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2711 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2712 | { |
a910b123 | 2713 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2714 | enum drbd_ret_code retcode; |
194bfb32 | 2715 | |
a910b123 | 2716 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2717 | if (!adm_ctx.reply_skb) |
2718 | return retcode; | |
2719 | if (retcode != NO_ERROR) | |
2720 | goto out; | |
b411b363 | 2721 | |
9e276872 | 2722 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2723 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 2724 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2725 | out: |
a910b123 | 2726 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2727 | return 0; |
2728 | } | |
2729 | ||
8fe39aac | 2730 | static int drbd_bmio_set_susp_al(struct drbd_device *device) __must_hold(local) |
0778286a PR |
2731 | { |
2732 | int rv; | |
2733 | ||
b30ab791 AG |
2734 | rv = drbd_bmio_set_n_write(device); |
2735 | drbd_suspend_al(device); | |
0778286a PR |
2736 | return rv; |
2737 | } | |
2738 | ||
3b98c0c2 | 2739 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2740 | { |
a910b123 | 2741 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 2742 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2743 | struct drbd_device *device; |
25b0d6c8 | 2744 | |
a910b123 | 2745 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
2746 | if (!adm_ctx.reply_skb) |
2747 | return retcode; | |
2748 | if (retcode != NO_ERROR) | |
2749 | goto out; | |
2750 | ||
b30ab791 | 2751 | device = adm_ctx.device; |
8fe39aac PR |
2752 | if (!get_ldev(device)) { |
2753 | retcode = ERR_NO_DISK; | |
2754 | goto out; | |
2755 | } | |
2756 | ||
2757 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2758 | |
194bfb32 | 2759 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2760 | * resync just being finished, wait for it before requesting a new resync. |
2761 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2762 | drbd_suspend_io(device); |
2763 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2764 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2765 | |
0b2dafcd PR |
2766 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2767 | * in the bitmap. Otherwise, try to start a resync handshake | |
2768 | * as sync source for full sync. | |
2769 | */ | |
b30ab791 | 2770 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2771 | /* The peer will get a resync upon connect anyways. Just make that |
2772 | into a full resync. */ | |
b30ab791 | 2773 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2774 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2775 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2776 | "set_n_write from invalidate_peer", |
2777 | BM_LOCKED_SET_ALLOWED)) | |
2778 | retcode = ERR_IO_MD_DISK; | |
2779 | } | |
2780 | } else | |
b30ab791 AG |
2781 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2782 | drbd_resume_io(device); | |
9e276872 | 2783 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2784 | put_ldev(device); |
25b0d6c8 | 2785 | out: |
a910b123 | 2786 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2787 | return 0; |
2788 | } | |
2789 | ||
3b98c0c2 | 2790 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2791 | { |
a910b123 | 2792 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2793 | enum drbd_ret_code retcode; |
b411b363 | 2794 | |
a910b123 | 2795 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2796 | if (!adm_ctx.reply_skb) |
2797 | return retcode; | |
2798 | if (retcode != NO_ERROR) | |
2799 | goto out; | |
b411b363 | 2800 | |
9e276872 | 2801 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2802 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 2803 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 2804 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2805 | out: |
a910b123 | 2806 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2807 | return 0; |
2808 | } | |
2809 | ||
3b98c0c2 | 2810 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2811 | { |
a910b123 | 2812 | struct drbd_config_context adm_ctx; |
da9fbc27 | 2813 | union drbd_dev_state s; |
3b98c0c2 LE |
2814 | enum drbd_ret_code retcode; |
2815 | ||
a910b123 | 2816 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2817 | if (!adm_ctx.reply_skb) |
2818 | return retcode; | |
2819 | if (retcode != NO_ERROR) | |
2820 | goto out; | |
b411b363 | 2821 | |
9e276872 | 2822 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2823 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2824 | s = adm_ctx.device->state; | |
cd88d030 PR |
2825 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2826 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2827 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2828 | } else { | |
2829 | retcode = ERR_PAUSE_IS_CLEAR; | |
2830 | } | |
2831 | } | |
9e276872 | 2832 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2833 | out: |
a910b123 | 2834 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2835 | return 0; |
2836 | } | |
2837 | ||
3b98c0c2 | 2838 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2839 | { |
3b98c0c2 | 2840 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2841 | } |
2842 | ||
3b98c0c2 | 2843 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2844 | { |
a910b123 | 2845 | struct drbd_config_context adm_ctx; |
b30ab791 | 2846 | struct drbd_device *device; |
3b98c0c2 LE |
2847 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2848 | ||
a910b123 | 2849 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2850 | if (!adm_ctx.reply_skb) |
2851 | return retcode; | |
2852 | if (retcode != NO_ERROR) | |
2853 | goto out; | |
2854 | ||
9e276872 | 2855 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2856 | device = adm_ctx.device; |
2857 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2858 | drbd_uuid_new_current(device); | |
2859 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2860 | } |
b30ab791 AG |
2861 | drbd_suspend_io(device); |
2862 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2863 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2864 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2865 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2866 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2867 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2868 | } |
b30ab791 | 2869 | drbd_resume_io(device); |
9e276872 | 2870 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2871 | out: |
a910b123 | 2872 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2873 | return 0; |
2874 | } | |
2875 | ||
3b98c0c2 | 2876 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2877 | { |
3b98c0c2 | 2878 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2879 | } |
2880 | ||
251b8f8e AG |
2881 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2882 | struct drbd_resource *resource, | |
2883 | struct drbd_connection *connection, | |
2884 | struct drbd_device *device) | |
b411b363 | 2885 | { |
543cc10b LE |
2886 | struct nlattr *nla; |
2887 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2888 | if (!nla) | |
2889 | goto nla_put_failure; | |
251b8f8e AG |
2890 | if (device && |
2891 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2892 | goto nla_put_failure; |
f597f6b8 | 2893 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 2894 | goto nla_put_failure; |
251b8f8e AG |
2895 | if (connection) { |
2896 | if (connection->my_addr_len && | |
2897 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2898 | goto nla_put_failure; | |
2899 | if (connection->peer_addr_len && | |
2900 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2901 | goto nla_put_failure; | |
2902 | } | |
543cc10b LE |
2903 | nla_nest_end(skb, nla); |
2904 | return 0; | |
b411b363 | 2905 | |
543cc10b LE |
2906 | nla_put_failure: |
2907 | if (nla) | |
2908 | nla_nest_cancel(skb, nla); | |
2909 | return -EMSGSIZE; | |
2910 | } | |
b411b363 | 2911 | |
251b8f8e AG |
2912 | /* |
2913 | * Return the connection of @resource if @resource has exactly one connection. | |
2914 | */ | |
2915 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2916 | { | |
2917 | struct list_head *connections = &resource->connections; | |
2918 | ||
2919 | if (list_empty(connections) || connections->next->next != connections) | |
2920 | return NULL; | |
2921 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2922 | } | |
2923 | ||
8ce953aa | 2924 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 2925 | const struct sib_info *sib) |
b411b363 | 2926 | { |
251b8f8e | 2927 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2928 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2929 | struct nlattr *nla; | |
2930 | int got_ldev; | |
3b98c0c2 LE |
2931 | int err = 0; |
2932 | int exclude_sensitive; | |
2933 | ||
2934 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2935 | * to. So we better exclude_sensitive information. | |
2936 | * | |
2937 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2938 | * in the context of the requesting user process. Exclude sensitive | |
2939 | * information, unless current has superuser. | |
2940 | * | |
2941 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2942 | * relies on the current implementation of netlink_dump(), which | |
2943 | * executes the dump callback successively from netlink_recvmsg(), | |
2944 | * always in the context of the receiving process */ | |
2945 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2946 | ||
b30ab791 | 2947 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2948 | |
2949 | /* We need to add connection name and volume number information still. | |
2950 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 2951 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 2952 | goto nla_put_failure; |
3b98c0c2 | 2953 | |
eb6bea67 | 2954 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
2955 | goto nla_put_failure; |
2956 | ||
daeda1cc | 2957 | rcu_read_lock(); |
f9eb7bf4 AG |
2958 | if (got_ldev) { |
2959 | struct disk_conf *disk_conf; | |
44ed167d | 2960 | |
b30ab791 | 2961 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2962 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2963 | } | |
2964 | if (!err) { | |
2965 | struct net_conf *nc; | |
2966 | ||
a6b32bc3 | 2967 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2968 | if (nc) |
2969 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2970 | } | |
44ed167d PR |
2971 | rcu_read_unlock(); |
2972 | if (err) | |
2973 | goto nla_put_failure; | |
3b98c0c2 | 2974 | |
3b98c0c2 LE |
2975 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2976 | if (!nla) | |
2977 | goto nla_put_failure; | |
26ec9287 | 2978 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2979 | nla_put_u32(skb, T_current_state, device->state.i) || |
2980 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2981 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2982 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2983 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2984 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2985 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2986 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2987 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2988 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2989 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2990 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2991 | goto nla_put_failure; |
3b98c0c2 LE |
2992 | |
2993 | if (got_ldev) { | |
39a1aa7f | 2994 | int err; |
b411b363 | 2995 | |
b30ab791 AG |
2996 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2997 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2998 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2999 | |
3000 | if (err) | |
3001 | goto nla_put_failure; | |
3002 | ||
b30ab791 AG |
3003 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
3004 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
3005 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 3006 | goto nla_put_failure; |
b30ab791 AG |
3007 | if (C_SYNC_SOURCE <= device->state.conn && |
3008 | C_PAUSED_SYNC_T >= device->state.conn) { | |
3009 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
3010 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 3011 | goto nla_put_failure; |
3b98c0c2 | 3012 | } |
b411b363 | 3013 | } |
b411b363 | 3014 | |
3b98c0c2 LE |
3015 | if (sib) { |
3016 | switch(sib->sib_reason) { | |
3017 | case SIB_SYNC_PROGRESS: | |
3018 | case SIB_GET_STATUS_REPLY: | |
3019 | break; | |
3020 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
3021 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
3022 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
3023 | goto nla_put_failure; | |
3b98c0c2 LE |
3024 | break; |
3025 | case SIB_HELPER_POST: | |
26ec9287 AG |
3026 | if (nla_put_u32(skb, T_helper_exit_code, |
3027 | sib->helper_exit_code)) | |
3028 | goto nla_put_failure; | |
3b98c0c2 LE |
3029 | /* fall through */ |
3030 | case SIB_HELPER_PRE: | |
26ec9287 AG |
3031 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
3032 | goto nla_put_failure; | |
3b98c0c2 LE |
3033 | break; |
3034 | } | |
b411b363 | 3035 | } |
3b98c0c2 | 3036 | nla_nest_end(skb, nla); |
b411b363 | 3037 | |
3b98c0c2 LE |
3038 | if (0) |
3039 | nla_put_failure: | |
3040 | err = -EMSGSIZE; | |
3041 | if (got_ldev) | |
b30ab791 | 3042 | put_ldev(device); |
3b98c0c2 | 3043 | return err; |
b411b363 PR |
3044 | } |
3045 | ||
3b98c0c2 | 3046 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3047 | { |
a910b123 | 3048 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3049 | enum drbd_ret_code retcode; |
3050 | int err; | |
b411b363 | 3051 | |
a910b123 | 3052 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3053 | if (!adm_ctx.reply_skb) |
3054 | return retcode; | |
3055 | if (retcode != NO_ERROR) | |
3056 | goto out; | |
b411b363 | 3057 | |
b30ab791 | 3058 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3059 | if (err) { |
3060 | nlmsg_free(adm_ctx.reply_skb); | |
3061 | return err; | |
b411b363 | 3062 | } |
3b98c0c2 | 3063 | out: |
a910b123 | 3064 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3065 | return 0; |
b411b363 PR |
3066 | } |
3067 | ||
4b7a530f | 3068 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3069 | { |
b30ab791 | 3070 | struct drbd_device *device; |
3b98c0c2 | 3071 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3072 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3073 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3074 | struct drbd_resource *tmp; |
543cc10b LE |
3075 | unsigned volume = cb->args[1]; |
3076 | ||
3077 | /* Open coded, deferred, iteration: | |
77c556f6 | 3078 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3079 | * connection = "first connection of resource or undefined"; |
3080 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3081 | * ... |
3082 | * } | |
3083 | * } | |
77c556f6 | 3084 | * where resource is cb->args[0]; |
543cc10b LE |
3085 | * and i is cb->args[1]; |
3086 | * | |
71932efc LE |
3087 | * cb->args[2] indicates if we shall loop over all resources, |
3088 | * or just dump all volumes of a single resource. | |
3089 | * | |
3b98c0c2 LE |
3090 | * This may miss entries inserted after this dump started, |
3091 | * or entries deleted before they are reached. | |
543cc10b | 3092 | * |
b30ab791 | 3093 | * We need to make sure the device won't disappear while |
543cc10b LE |
3094 | * we are looking at it, and revalidate our iterators |
3095 | * on each iteration. | |
3096 | */ | |
b411b363 | 3097 | |
05a10ec7 | 3098 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3099 | rcu_read_lock(); |
543cc10b | 3100 | /* revalidate iterator position */ |
77c556f6 | 3101 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3102 | if (pos == NULL) { |
3103 | /* first iteration */ | |
3104 | pos = tmp; | |
77c556f6 | 3105 | resource = pos; |
543cc10b LE |
3106 | break; |
3107 | } | |
3108 | if (tmp == pos) { | |
77c556f6 | 3109 | resource = pos; |
543cc10b LE |
3110 | break; |
3111 | } | |
b411b363 | 3112 | } |
77c556f6 AG |
3113 | if (resource) { |
3114 | next_resource: | |
251b8f8e AG |
3115 | device = idr_get_next(&resource->devices, &volume); |
3116 | if (!device) { | |
77c556f6 AG |
3117 | /* No more volumes to dump on this resource. |
3118 | * Advance resource iterator. */ | |
3119 | pos = list_entry_rcu(resource->resources.next, | |
3120 | struct drbd_resource, resources); | |
3121 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3122 | if (volume != 0) { |
71932efc LE |
3123 | /* If we reached the end of the list, |
3124 | * or only a single resource dump was requested, | |
3125 | * we are done. */ | |
77c556f6 | 3126 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3127 | goto out; |
543cc10b | 3128 | volume = 0; |
77c556f6 AG |
3129 | resource = pos; |
3130 | goto next_resource; | |
543cc10b LE |
3131 | } |
3132 | } | |
3133 | ||
98683650 | 3134 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3135 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3136 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3137 | if (!dh) | |
543cc10b LE |
3138 | goto out; |
3139 | ||
251b8f8e | 3140 | if (!device) { |
bde89a9e | 3141 | /* This is a connection without a single volume. |
367d675d LE |
3142 | * Suprisingly enough, it may have a network |
3143 | * configuration. */ | |
251b8f8e AG |
3144 | struct drbd_connection *connection; |
3145 | ||
543cc10b LE |
3146 | dh->minor = -1U; |
3147 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3148 | connection = the_only_connection(resource); |
3149 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3150 | goto cancel; |
251b8f8e AG |
3151 | if (connection) { |
3152 | struct net_conf *nc; | |
3153 | ||
3154 | nc = rcu_dereference(connection->net_conf); | |
3155 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3156 | goto cancel; | |
3157 | } | |
367d675d | 3158 | goto done; |
543cc10b | 3159 | } |
b411b363 | 3160 | |
0b0ba1ef | 3161 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3162 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3163 | |
b30ab791 | 3164 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3165 | dh->ret_code = NO_ERROR; |
3166 | ||
b30ab791 | 3167 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3168 | cancel: |
3b98c0c2 | 3169 | genlmsg_cancel(skb, dh); |
543cc10b | 3170 | goto out; |
3b98c0c2 | 3171 | } |
367d675d | 3172 | done: |
3b98c0c2 | 3173 | genlmsg_end(skb, dh); |
bde89a9e | 3174 | } |
b411b363 | 3175 | |
543cc10b | 3176 | out: |
c141ebda | 3177 | rcu_read_unlock(); |
543cc10b | 3178 | /* where to start the next iteration */ |
bde89a9e | 3179 | cb->args[0] = (long)pos; |
77c556f6 | 3180 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3181 | |
77c556f6 | 3182 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3183 | * Which will terminate the dump. */ |
3b98c0c2 | 3184 | return skb->len; |
b411b363 PR |
3185 | } |
3186 | ||
71932efc LE |
3187 | /* |
3188 | * Request status of all resources, or of all volumes within a single resource. | |
3189 | * | |
3190 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3191 | * Which means we cannot use the family->attrbuf or other such members, because | |
3192 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3193 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3194 | * | |
3195 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3196 | */ |
71932efc | 3197 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3198 | { |
71932efc LE |
3199 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3200 | struct nlattr *nla; | |
7c3063cc | 3201 | const char *resource_name; |
4bc76048 | 3202 | struct drbd_resource *resource; |
7c3063cc | 3203 | int maxtype; |
71932efc LE |
3204 | |
3205 | /* Is this a followup call? */ | |
3206 | if (cb->args[0]) { | |
3207 | /* ... of a single resource dump, | |
3208 | * and the resource iterator has been advanced already? */ | |
3209 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3210 | return 0; /* DONE. */ | |
3211 | goto dump; | |
3212 | } | |
3213 | ||
3214 | /* First call (from netlink_dump_start). We need to figure out | |
3215 | * which resource(s) the user wants us to dump. */ | |
3216 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3217 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3218 | DRBD_NLA_CFG_CONTEXT); | |
3219 | ||
3220 | /* No explicit context given. Dump all. */ | |
3221 | if (!nla) | |
3222 | goto dump; | |
7c3063cc AG |
3223 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3224 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3225 | if (IS_ERR(nla)) | |
3226 | return PTR_ERR(nla); | |
71932efc LE |
3227 | /* context given, but no name present? */ |
3228 | if (!nla) | |
3229 | return -EINVAL; | |
7c3063cc | 3230 | resource_name = nla_data(nla); |
4bc76048 AG |
3231 | if (!*resource_name) |
3232 | return -ENODEV; | |
3233 | resource = drbd_find_resource(resource_name); | |
3234 | if (!resource) | |
71932efc LE |
3235 | return -ENODEV; |
3236 | ||
4bc76048 | 3237 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3238 | |
71932efc | 3239 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3240 | * only dump this connection. */ |
4bc76048 | 3241 | cb->args[0] = (long)resource; |
71932efc | 3242 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3243 | cb->args[2] = (long)resource; |
71932efc LE |
3244 | |
3245 | dump: | |
3246 | return get_one_status(skb, cb); | |
3247 | } | |
b411b363 | 3248 | |
3b98c0c2 | 3249 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3250 | { |
a910b123 | 3251 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3252 | enum drbd_ret_code retcode; |
3253 | struct timeout_parms tp; | |
3254 | int err; | |
b411b363 | 3255 | |
a910b123 | 3256 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3257 | if (!adm_ctx.reply_skb) |
3258 | return retcode; | |
3259 | if (retcode != NO_ERROR) | |
3260 | goto out; | |
b411b363 | 3261 | |
3b98c0c2 | 3262 | tp.timeout_type = |
b30ab791 AG |
3263 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3264 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3265 | UT_DEFAULT; |
b411b363 | 3266 | |
3b98c0c2 LE |
3267 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3268 | if (err) { | |
3269 | nlmsg_free(adm_ctx.reply_skb); | |
3270 | return err; | |
3271 | } | |
3272 | out: | |
a910b123 | 3273 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3274 | return 0; |
b411b363 PR |
3275 | } |
3276 | ||
3b98c0c2 | 3277 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3278 | { |
a910b123 | 3279 | struct drbd_config_context adm_ctx; |
b30ab791 | 3280 | struct drbd_device *device; |
3b98c0c2 | 3281 | enum drbd_ret_code retcode; |
58ffa580 | 3282 | struct start_ov_parms parms; |
b411b363 | 3283 | |
a910b123 | 3284 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3285 | if (!adm_ctx.reply_skb) |
3286 | return retcode; | |
3287 | if (retcode != NO_ERROR) | |
3288 | goto out; | |
873b0d5f | 3289 | |
b30ab791 | 3290 | device = adm_ctx.device; |
58ffa580 LE |
3291 | |
3292 | /* resume from last known position, if possible */ | |
b30ab791 | 3293 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3294 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3295 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3296 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3297 | if (err) { |
3298 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3299 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3300 | goto out; |
3301 | } | |
b411b363 | 3302 | } |
9e276872 LE |
3303 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3304 | ||
58ffa580 | 3305 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3306 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3307 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3308 | |
3309 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3310 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3311 | drbd_suspend_io(device); |
3312 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3313 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3314 | drbd_resume_io(device); | |
9e276872 LE |
3315 | |
3316 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 3317 | out: |
a910b123 | 3318 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3319 | return 0; |
3320 | } | |
3321 | ||
3322 | ||
3b98c0c2 | 3323 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3324 | { |
a910b123 | 3325 | struct drbd_config_context adm_ctx; |
b30ab791 | 3326 | struct drbd_device *device; |
3b98c0c2 | 3327 | enum drbd_ret_code retcode; |
b411b363 PR |
3328 | int skip_initial_sync = 0; |
3329 | int err; | |
3b98c0c2 | 3330 | struct new_c_uuid_parms args; |
b411b363 | 3331 | |
a910b123 | 3332 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3333 | if (!adm_ctx.reply_skb) |
3334 | return retcode; | |
3335 | if (retcode != NO_ERROR) | |
3336 | goto out_nolock; | |
b411b363 | 3337 | |
b30ab791 | 3338 | device = adm_ctx.device; |
3b98c0c2 LE |
3339 | memset(&args, 0, sizeof(args)); |
3340 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3341 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3342 | if (err) { |
3343 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3344 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3345 | goto out_nolock; |
3346 | } | |
b411b363 PR |
3347 | } |
3348 | ||
9e276872 | 3349 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3350 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3351 | |
b30ab791 | 3352 | if (!get_ldev(device)) { |
b411b363 PR |
3353 | retcode = ERR_NO_DISK; |
3354 | goto out; | |
3355 | } | |
3356 | ||
3357 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3358 | if (device->state.conn == C_CONNECTED && |
3359 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3360 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3361 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3362 | skip_initial_sync = 1; |
b30ab791 | 3363 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3364 | retcode = ERR_CONNECTED; |
3365 | goto out_dec; | |
3366 | } | |
3367 | ||
b30ab791 AG |
3368 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3369 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3370 | |
3371 | if (args.clear_bm) { | |
b30ab791 | 3372 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3373 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3374 | if (err) { |
d0180171 | 3375 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3376 | retcode = ERR_IO_MD_DISK; |
3377 | } | |
3378 | if (skip_initial_sync) { | |
69a22773 | 3379 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
3380 | _drbd_uuid_set(device, UI_BITMAP, 0); |
3381 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3382 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3383 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3384 | CS_VERBOSE, NULL); |
0500813f | 3385 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3386 | } |
3387 | } | |
3388 | ||
b30ab791 | 3389 | drbd_md_sync(device); |
b411b363 | 3390 | out_dec: |
b30ab791 | 3391 | put_ldev(device); |
b411b363 | 3392 | out: |
b30ab791 | 3393 | mutex_unlock(device->state_mutex); |
9e276872 | 3394 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3395 | out_nolock: |
a910b123 | 3396 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3397 | return 0; |
3398 | } | |
3399 | ||
3b98c0c2 | 3400 | static enum drbd_ret_code |
a910b123 | 3401 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 3402 | { |
a910b123 | 3403 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 3404 | if (!name || !name[0]) { |
a910b123 | 3405 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 3406 | return ERR_MANDATORY_TAG; |
b411b363 | 3407 | } |
3b98c0c2 LE |
3408 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3409 | * we must not allow slashes */ | |
3410 | if (strchr(name, '/')) { | |
a910b123 | 3411 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 3412 | return ERR_INVALID_REQUEST; |
b411b363 | 3413 | } |
3b98c0c2 | 3414 | return NO_ERROR; |
774b3055 | 3415 | } |
b411b363 | 3416 | |
789c1b62 | 3417 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3418 | { |
a910b123 | 3419 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3420 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3421 | struct res_opts res_opts; |
3422 | int err; | |
b411b363 | 3423 | |
a910b123 | 3424 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
3425 | if (!adm_ctx.reply_skb) |
3426 | return retcode; | |
3427 | if (retcode != NO_ERROR) | |
3428 | goto out; | |
b411b363 | 3429 | |
afbbfa88 AG |
3430 | set_res_opts_defaults(&res_opts); |
3431 | err = res_opts_from_attrs(&res_opts, info); | |
3432 | if (err && err != -ENOMSG) { | |
3433 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3434 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 3435 | goto out; |
b411b363 PR |
3436 | } |
3437 | ||
a910b123 | 3438 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
3439 | if (retcode != NO_ERROR) |
3440 | goto out; | |
b411b363 | 3441 | |
5c661042 | 3442 | if (adm_ctx.resource) { |
38f19616 LE |
3443 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3444 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 3445 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
3446 | } |
3447 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3448 | goto out; |
b411b363 | 3449 | } |
b411b363 | 3450 | |
9e276872 | 3451 | /* not yet safe for genl_family.parallel_ops */ |
afbbfa88 | 3452 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3453 | retcode = ERR_NOMEM; |
3b98c0c2 | 3454 | out: |
a910b123 | 3455 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3456 | return 0; |
b411b363 PR |
3457 | } |
3458 | ||
05a10ec7 | 3459 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3460 | { |
a910b123 | 3461 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3462 | struct drbd_genlmsghdr *dh = info->userhdr; |
3463 | enum drbd_ret_code retcode; | |
b411b363 | 3464 | |
a910b123 | 3465 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3466 | if (!adm_ctx.reply_skb) |
3467 | return retcode; | |
3468 | if (retcode != NO_ERROR) | |
3469 | goto out; | |
b411b363 | 3470 | |
f2257a56 | 3471 | if (dh->minor > MINORMASK) { |
a910b123 | 3472 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
3473 | retcode = ERR_INVALID_REQUEST; |
3474 | goto out; | |
b411b363 | 3475 | } |
0c8e36d9 | 3476 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 3477 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
3478 | retcode = ERR_INVALID_REQUEST; |
3479 | goto out; | |
b411b363 | 3480 | } |
b411b363 | 3481 | |
38f19616 | 3482 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3483 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3484 | if (adm_ctx.device) { |
38f19616 | 3485 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
179e20b8 | 3486 | retcode = ERR_MINOR_OR_VOLUME_EXISTS; |
38f19616 LE |
3487 | /* else: still NO_ERROR */ |
3488 | goto out; | |
b411b363 | 3489 | } |
38f19616 | 3490 | |
9e276872 | 3491 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 3492 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
9e276872 | 3493 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3494 | out: |
a910b123 | 3495 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3496 | return 0; |
b411b363 PR |
3497 | } |
3498 | ||
05a10ec7 | 3499 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3500 | { |
b30ab791 AG |
3501 | if (device->state.disk == D_DISKLESS && |
3502 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3503 | * we may want to delete a minor from a live replication group. |
3504 | */ | |
b30ab791 AG |
3505 | device->state.role == R_SECONDARY) { |
3506 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3507 | CS_VERBOSE + CS_WAIT_COMPLETE); |
f82795d6 | 3508 | drbd_delete_device(device); |
85f75dd7 LE |
3509 | return NO_ERROR; |
3510 | } else | |
3511 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3512 | } |
3513 | ||
05a10ec7 | 3514 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3515 | { |
a910b123 | 3516 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3517 | enum drbd_ret_code retcode; |
b411b363 | 3518 | |
a910b123 | 3519 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3520 | if (!adm_ctx.reply_skb) |
3521 | return retcode; | |
3522 | if (retcode != NO_ERROR) | |
3523 | goto out; | |
b411b363 | 3524 | |
9e276872 | 3525 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 3526 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 3527 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 3528 | out: |
a910b123 | 3529 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 3530 | return 0; |
b411b363 PR |
3531 | } |
3532 | ||
179e20b8 AG |
3533 | static int adm_del_resource(struct drbd_resource *resource) |
3534 | { | |
3535 | struct drbd_connection *connection; | |
3536 | ||
3537 | for_each_connection(connection, resource) { | |
3538 | if (connection->cstate > C_STANDALONE) | |
3539 | return ERR_NET_CONFIGURED; | |
3540 | } | |
3541 | if (!idr_is_empty(&resource->devices)) | |
3542 | return ERR_RES_IN_USE; | |
3543 | ||
3544 | list_del_rcu(&resource->resources); | |
3545 | /* Make sure all threads have actually stopped: state handling only | |
3546 | * does drbd_thread_stop_nowait(). */ | |
3547 | list_for_each_entry(connection, &resource->connections, connections) | |
3548 | drbd_thread_stop(&connection->worker); | |
3549 | synchronize_rcu(); | |
3550 | drbd_free_resource(resource); | |
3551 | return NO_ERROR; | |
3552 | } | |
3553 | ||
85f75dd7 | 3554 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3555 | { |
a910b123 | 3556 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
3557 | struct drbd_resource *resource; |
3558 | struct drbd_connection *connection; | |
3559 | struct drbd_device *device; | |
f3dfa40a | 3560 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3561 | unsigned i; |
b411b363 | 3562 | |
a910b123 | 3563 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3564 | if (!adm_ctx.reply_skb) |
3565 | return retcode; | |
3566 | if (retcode != NO_ERROR) | |
9e276872 | 3567 | goto finish; |
b411b363 | 3568 | |
b6f85ef9 | 3569 | resource = adm_ctx.resource; |
9e276872 | 3570 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 3571 | /* demote */ |
b6f85ef9 AG |
3572 | for_each_connection(connection, resource) { |
3573 | struct drbd_peer_device *peer_device; | |
3574 | ||
3575 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3576 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3577 | if (retcode < SS_SUCCESS) { | |
a910b123 | 3578 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
3579 | goto out; |
3580 | } | |
3581 | } | |
3582 | ||
3583 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3584 | if (retcode < SS_SUCCESS) { |
a910b123 | 3585 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 3586 | goto out; |
85f75dd7 | 3587 | } |
b411b363 | 3588 | } |
b411b363 | 3589 | |
85f75dd7 | 3590 | /* detach */ |
b6f85ef9 AG |
3591 | idr_for_each_entry(&resource->devices, device, i) { |
3592 | retcode = adm_detach(device, 0); | |
27012382 | 3593 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 3594 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 3595 | goto out; |
85f75dd7 LE |
3596 | } |
3597 | } | |
b411b363 | 3598 | |
85f75dd7 | 3599 | /* delete volumes */ |
b6f85ef9 AG |
3600 | idr_for_each_entry(&resource->devices, device, i) { |
3601 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3602 | if (retcode != NO_ERROR) { |
3603 | /* "can not happen" */ | |
a910b123 | 3604 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 3605 | goto out; |
85f75dd7 LE |
3606 | } |
3607 | } | |
b411b363 | 3608 | |
179e20b8 | 3609 | retcode = adm_del_resource(resource); |
3b98c0c2 | 3610 | out: |
9e276872 LE |
3611 | mutex_unlock(&resource->adm_mutex); |
3612 | finish: | |
a910b123 | 3613 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3614 | return 0; |
b411b363 PR |
3615 | } |
3616 | ||
789c1b62 | 3617 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3618 | { |
a910b123 | 3619 | struct drbd_config_context adm_ctx; |
77c556f6 | 3620 | struct drbd_resource *resource; |
3b98c0c2 | 3621 | enum drbd_ret_code retcode; |
b411b363 | 3622 | |
a910b123 | 3623 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3624 | if (!adm_ctx.reply_skb) |
3625 | return retcode; | |
3626 | if (retcode != NO_ERROR) | |
9e276872 | 3627 | goto finish; |
77c556f6 | 3628 | resource = adm_ctx.resource; |
b411b363 | 3629 | |
179e20b8 AG |
3630 | mutex_lock(&resource->adm_mutex); |
3631 | retcode = adm_del_resource(resource); | |
9e276872 LE |
3632 | mutex_unlock(&resource->adm_mutex); |
3633 | finish: | |
a910b123 | 3634 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3635 | return 0; |
3636 | } | |
3637 | ||
b30ab791 | 3638 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3639 | { |
3b98c0c2 LE |
3640 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3641 | struct sk_buff *msg; | |
3642 | struct drbd_genlmsghdr *d_out; | |
3643 | unsigned seq; | |
3644 | int err = -ENOMEM; | |
3645 | ||
3646 | seq = atomic_inc_return(&drbd_genl_seq); | |
3647 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3648 | if (!msg) | |
3649 | goto failed; | |
3650 | ||
3651 | err = -EMSGSIZE; | |
3652 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3653 | if (!d_out) /* cannot happen, but anyways. */ | |
3654 | goto nla_put_failure; | |
b30ab791 | 3655 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3656 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3657 | |
b30ab791 | 3658 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3659 | goto nla_put_failure; |
3660 | genlmsg_end(msg, d_out); | |
3661 | err = drbd_genl_multicast_events(msg, 0); | |
3662 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3663 | if (err && err != -ESRCH) | |
3664 | goto failed; | |
b411b363 | 3665 | |
3b98c0c2 | 3666 | return; |
b411b363 | 3667 | |
3b98c0c2 LE |
3668 | nla_put_failure: |
3669 | nlmsg_free(msg); | |
3670 | failed: | |
d0180171 | 3671 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3672 | "Event seq:%u sib_reason:%u\n", |
3673 | err, seq, sib->sib_reason); | |
b411b363 | 3674 | } |