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" |
a2972846 | 39 | #include "drbd_state_change.h" |
b411b363 | 40 | #include <asm/unaligned.h> |
b411b363 | 41 | #include <linux/drbd_limits.h> |
87f7be4c | 42 | #include <linux/kthread.h> |
b411b363 | 43 | |
3b98c0c2 LE |
44 | #include <net/genetlink.h> |
45 | ||
46 | /* .doit */ | |
47 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
48 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
49 | ||
05a10ec7 AG |
50 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
51 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 52 | |
789c1b62 AG |
53 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
54 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 55 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
56 | |
57 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
58 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 62 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
63 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
64 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
72 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
73 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 74 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
75 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
76 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
77 | /* .dumpit */ | |
78 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
a2972846 | 79 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb); |
3b98c0c2 LE |
80 | |
81 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 82 | #include "drbd_nla.h" |
3b98c0c2 LE |
83 | #include <linux/genl_magic_func.h> |
84 | ||
a2972846 AG |
85 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
86 | static atomic_t notify_genl_seq = ATOMIC_INIT(2); /* two. */ | |
87 | ||
88 | DEFINE_MUTEX(notification_mutex); | |
89 | ||
3b98c0c2 | 90 | /* used blkdev_get_by_path, to claim our meta data device(s) */ |
b411b363 PR |
91 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
92 | ||
3b98c0c2 LE |
93 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) |
94 | { | |
95 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
96 | if (genlmsg_reply(skb, info)) | |
f88c5d90 | 97 | pr_err("error sending genl reply\n"); |
b411b363 | 98 | } |
3b98c0c2 LE |
99 | |
100 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
101 | * reason it could fail was no space in skb, and there are 4k available. */ | |
f221f4bc | 102 | static int drbd_msg_put_info(struct sk_buff *skb, const char *info) |
3b98c0c2 | 103 | { |
3b98c0c2 LE |
104 | struct nlattr *nla; |
105 | int err = -EMSGSIZE; | |
106 | ||
107 | if (!info || !info[0]) | |
108 | return 0; | |
109 | ||
110 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
111 | if (!nla) | |
112 | return err; | |
113 | ||
114 | err = nla_put_string(skb, T_info_text, info); | |
115 | if (err) { | |
116 | nla_nest_cancel(skb, nla); | |
117 | return err; | |
118 | } else | |
119 | nla_nest_end(skb, nla); | |
120 | return 0; | |
b411b363 PR |
121 | } |
122 | ||
3b98c0c2 LE |
123 | /* This would be a good candidate for a "pre_doit" hook, |
124 | * and per-family private info->pointers. | |
125 | * But we need to stay compatible with older kernels. | |
126 | * If it returns successfully, adm_ctx members are valid. | |
9e276872 LE |
127 | * |
128 | * At this point, we still rely on the global genl_lock(). | |
129 | * If we want to avoid that, and allow "genl_family.parallel_ops", we may need | |
130 | * to add additional synchronization against object destruction/modification. | |
3b98c0c2 LE |
131 | */ |
132 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 133 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 134 | #define DRBD_ADM_NEED_CONNECTION 4 |
a910b123 LE |
135 | static int drbd_adm_prepare(struct drbd_config_context *adm_ctx, |
136 | struct sk_buff *skb, struct genl_info *info, unsigned flags) | |
3b98c0c2 LE |
137 | { |
138 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
139 | const u8 cmd = info->genlhdr->cmd; | |
140 | int err; | |
141 | ||
a910b123 | 142 | memset(adm_ctx, 0, sizeof(*adm_ctx)); |
3b98c0c2 LE |
143 | |
144 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 145 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
146 | return -EPERM; |
147 | ||
a910b123 LE |
148 | adm_ctx->reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); |
149 | if (!adm_ctx->reply_skb) { | |
1e2a2551 | 150 | err = -ENOMEM; |
3b98c0c2 | 151 | goto fail; |
1e2a2551 | 152 | } |
3b98c0c2 | 153 | |
a910b123 | 154 | adm_ctx->reply_dh = genlmsg_put_reply(adm_ctx->reply_skb, |
3b98c0c2 LE |
155 | info, &drbd_genl_family, 0, cmd); |
156 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
157 | * but anyways */ | |
a910b123 | 158 | if (!adm_ctx->reply_dh) { |
1e2a2551 | 159 | err = -ENOMEM; |
3b98c0c2 | 160 | goto fail; |
1e2a2551 | 161 | } |
3b98c0c2 | 162 | |
a910b123 LE |
163 | adm_ctx->reply_dh->minor = d_in->minor; |
164 | adm_ctx->reply_dh->ret_code = NO_ERROR; | |
3b98c0c2 | 165 | |
a910b123 | 166 | adm_ctx->volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
167 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
168 | struct nlattr *nla; | |
169 | /* parse and validate only */ | |
f399002e | 170 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
171 | if (err) |
172 | goto fail; | |
173 | ||
174 | /* It was present, and valid, | |
175 | * copy it over to the reply skb. */ | |
a910b123 | 176 | err = nla_put_nohdr(adm_ctx->reply_skb, |
3b98c0c2 LE |
177 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, |
178 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
179 | if (err) | |
180 | goto fail; | |
181 | ||
9e276872 | 182 | /* and assign stuff to the adm_ctx */ |
3b98c0c2 | 183 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; |
089c075d | 184 | if (nla) |
a910b123 | 185 | adm_ctx->volume = nla_get_u32(nla); |
7c3063cc | 186 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 187 | if (nla) |
a910b123 LE |
188 | adm_ctx->resource_name = nla_data(nla); |
189 | adm_ctx->my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; | |
190 | adm_ctx->peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
191 | if ((adm_ctx->my_addr && | |
192 | nla_len(adm_ctx->my_addr) > sizeof(adm_ctx->connection->my_addr)) || | |
193 | (adm_ctx->peer_addr && | |
194 | nla_len(adm_ctx->peer_addr) > sizeof(adm_ctx->connection->peer_addr))) { | |
089c075d AG |
195 | err = -EINVAL; |
196 | goto fail; | |
197 | } | |
198 | } | |
3b98c0c2 | 199 | |
a910b123 LE |
200 | adm_ctx->minor = d_in->minor; |
201 | adm_ctx->device = minor_to_device(d_in->minor); | |
9e276872 LE |
202 | |
203 | /* We are protected by the global genl_lock(). | |
204 | * But we may explicitly drop it/retake it in drbd_adm_set_role(), | |
205 | * so make sure this object stays around. */ | |
206 | if (adm_ctx->device) | |
207 | kref_get(&adm_ctx->device->kref); | |
208 | ||
a910b123 LE |
209 | if (adm_ctx->resource_name) { |
210 | adm_ctx->resource = drbd_find_resource(adm_ctx->resource_name); | |
4bc76048 | 211 | } |
3b98c0c2 | 212 | |
a910b123 LE |
213 | if (!adm_ctx->device && (flags & DRBD_ADM_NEED_MINOR)) { |
214 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown minor"); | |
3b98c0c2 LE |
215 | return ERR_MINOR_INVALID; |
216 | } | |
a910b123 LE |
217 | if (!adm_ctx->resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
218 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown resource"); | |
219 | if (adm_ctx->resource_name) | |
a10f6b8a | 220 | return ERR_RES_NOT_KNOWN; |
3b98c0c2 LE |
221 | return ERR_INVALID_REQUEST; |
222 | } | |
223 | ||
089c075d | 224 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
a910b123 LE |
225 | if (adm_ctx->resource) { |
226 | drbd_msg_put_info(adm_ctx->reply_skb, "no resource name expected"); | |
089c075d AG |
227 | return ERR_INVALID_REQUEST; |
228 | } | |
a910b123 LE |
229 | if (adm_ctx->device) { |
230 | drbd_msg_put_info(adm_ctx->reply_skb, "no minor number expected"); | |
089c075d AG |
231 | return ERR_INVALID_REQUEST; |
232 | } | |
a910b123 LE |
233 | if (adm_ctx->my_addr && adm_ctx->peer_addr) |
234 | adm_ctx->connection = conn_get_by_addrs(nla_data(adm_ctx->my_addr), | |
235 | nla_len(adm_ctx->my_addr), | |
236 | nla_data(adm_ctx->peer_addr), | |
237 | nla_len(adm_ctx->peer_addr)); | |
238 | if (!adm_ctx->connection) { | |
239 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown connection"); | |
089c075d AG |
240 | return ERR_INVALID_REQUEST; |
241 | } | |
242 | } | |
243 | ||
3b98c0c2 | 244 | /* some more paranoia, if the request was over-determined */ |
a910b123 LE |
245 | if (adm_ctx->device && adm_ctx->resource && |
246 | adm_ctx->device->resource != adm_ctx->resource) { | |
4bc76048 | 247 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", |
a910b123 LE |
248 | adm_ctx->minor, adm_ctx->resource->name, |
249 | adm_ctx->device->resource->name); | |
250 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists in different resource"); | |
527f4b24 LE |
251 | return ERR_INVALID_REQUEST; |
252 | } | |
a910b123 LE |
253 | if (adm_ctx->device && |
254 | adm_ctx->volume != VOLUME_UNSPECIFIED && | |
255 | adm_ctx->volume != adm_ctx->device->vnr) { | |
3b98c0c2 | 256 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
a910b123 LE |
257 | adm_ctx->minor, adm_ctx->volume, |
258 | adm_ctx->device->vnr, | |
259 | adm_ctx->device->resource->name); | |
260 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists as different volume"); | |
3b98c0c2 LE |
261 | return ERR_INVALID_REQUEST; |
262 | } | |
0ace9dfa | 263 | |
9e276872 LE |
264 | /* still, provide adm_ctx->resource always, if possible. */ |
265 | if (!adm_ctx->resource) { | |
266 | adm_ctx->resource = adm_ctx->device ? adm_ctx->device->resource | |
267 | : adm_ctx->connection ? adm_ctx->connection->resource : NULL; | |
268 | if (adm_ctx->resource) | |
269 | kref_get(&adm_ctx->resource->kref); | |
270 | } | |
271 | ||
3b98c0c2 LE |
272 | return NO_ERROR; |
273 | ||
274 | fail: | |
a910b123 LE |
275 | nlmsg_free(adm_ctx->reply_skb); |
276 | adm_ctx->reply_skb = NULL; | |
1e2a2551 | 277 | return err; |
3b98c0c2 LE |
278 | } |
279 | ||
a910b123 LE |
280 | static int drbd_adm_finish(struct drbd_config_context *adm_ctx, |
281 | struct genl_info *info, int retcode) | |
3b98c0c2 | 282 | { |
9e276872 LE |
283 | if (adm_ctx->device) { |
284 | kref_put(&adm_ctx->device->kref, drbd_destroy_device); | |
285 | adm_ctx->device = NULL; | |
286 | } | |
a910b123 LE |
287 | if (adm_ctx->connection) { |
288 | kref_put(&adm_ctx->connection->kref, &drbd_destroy_connection); | |
289 | adm_ctx->connection = NULL; | |
0ace9dfa | 290 | } |
a910b123 LE |
291 | if (adm_ctx->resource) { |
292 | kref_put(&adm_ctx->resource->kref, drbd_destroy_resource); | |
293 | adm_ctx->resource = NULL; | |
4bc76048 | 294 | } |
0ace9dfa | 295 | |
a910b123 | 296 | if (!adm_ctx->reply_skb) |
3b98c0c2 LE |
297 | return -ENOMEM; |
298 | ||
a910b123 LE |
299 | adm_ctx->reply_dh->ret_code = retcode; |
300 | drbd_adm_send_reply(adm_ctx->reply_skb, info); | |
3b98c0c2 LE |
301 | return 0; |
302 | } | |
b411b363 | 303 | |
bde89a9e | 304 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 305 | { |
6b75dced | 306 | char *afs; |
b411b363 | 307 | |
089c075d | 308 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 309 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
310 | return; |
311 | ||
bde89a9e | 312 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
313 | case AF_INET6: |
314 | afs = "ipv6"; | |
315 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 316 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 317 | break; |
089c075d AG |
318 | case AF_INET: |
319 | afs = "ipv4"; | |
320 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 321 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 322 | break; |
089c075d AG |
323 | default: |
324 | afs = "ssocks"; | |
325 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 326 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 327 | } |
089c075d | 328 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 329 | } |
b411b363 | 330 | |
b30ab791 | 331 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
332 | { |
333 | char *envp[] = { "HOME=/", | |
334 | "TERM=linux", | |
335 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
336 | (char[20]) { }, /* address family */ |
337 | (char[60]) { }, /* address */ | |
b411b363 | 338 | NULL }; |
6b75dced | 339 | char mb[12]; |
b411b363 | 340 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 341 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 342 | struct sib_info sib; |
b411b363 PR |
343 | int ret; |
344 | ||
bde89a9e AG |
345 | if (current == connection->worker.task) |
346 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 347 | |
b30ab791 | 348 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 349 | setup_khelper_env(connection, envp); |
b411b363 | 350 | |
1090c056 LE |
351 | /* The helper may take some time. |
352 | * write out any unsynced meta data changes now */ | |
b30ab791 | 353 | drbd_md_sync(device); |
1090c056 | 354 | |
d0180171 | 355 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
356 | sib.sib_reason = SIB_HELPER_PRE; |
357 | sib.helper_name = cmd; | |
b30ab791 | 358 | drbd_bcast_event(device, &sib); |
a2972846 | 359 | notify_helper(NOTIFY_CALL, device, connection, cmd, 0); |
70834d30 | 360 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 361 | if (ret) |
d0180171 | 362 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
363 | usermode_helper, cmd, mb, |
364 | (ret >> 8) & 0xff, ret); | |
365 | else | |
d0180171 | 366 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
367 | usermode_helper, cmd, mb, |
368 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
369 | sib.sib_reason = SIB_HELPER_POST; |
370 | sib.helper_exit_code = ret; | |
b30ab791 | 371 | drbd_bcast_event(device, &sib); |
a2972846 | 372 | notify_helper(NOTIFY_RESPONSE, device, connection, cmd, ret); |
b411b363 | 373 | |
bde89a9e AG |
374 | if (current == connection->worker.task) |
375 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
376 | |
377 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
378 | ret = 0; | |
379 | ||
380 | return ret; | |
381 | } | |
382 | ||
bde89a9e | 383 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
384 | { |
385 | char *envp[] = { "HOME=/", | |
386 | "TERM=linux", | |
387 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
388 | (char[20]) { }, /* address family */ | |
389 | (char[60]) { }, /* address */ | |
390 | NULL }; | |
77c556f6 AG |
391 | char *resource_name = connection->resource->name; |
392 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
393 | int ret; |
394 | ||
bde89a9e AG |
395 | setup_khelper_env(connection, envp); |
396 | conn_md_sync(connection); | |
6b75dced | 397 | |
1ec861eb | 398 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced | 399 | /* TODO: conn_bcast_event() ?? */ |
a2972846 | 400 | notify_helper(NOTIFY_CALL, NULL, connection, cmd, 0); |
6b75dced | 401 | |
98683650 | 402 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 403 | if (ret) |
1ec861eb | 404 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 405 | usermode_helper, cmd, resource_name, |
6b75dced PR |
406 | (ret >> 8) & 0xff, ret); |
407 | else | |
1ec861eb | 408 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 409 | usermode_helper, cmd, resource_name, |
6b75dced PR |
410 | (ret >> 8) & 0xff, ret); |
411 | /* TODO: conn_bcast_event() ?? */ | |
a2972846 | 412 | notify_helper(NOTIFY_RESPONSE, NULL, connection, cmd, ret); |
c2ba686f | 413 | |
b411b363 PR |
414 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
415 | ret = 0; | |
416 | ||
417 | return ret; | |
418 | } | |
419 | ||
bde89a9e | 420 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 421 | { |
cb703454 | 422 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 423 | struct drbd_peer_device *peer_device; |
cb703454 PR |
424 | int vnr; |
425 | ||
695d08fa | 426 | rcu_read_lock(); |
c06ece6b AG |
427 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
428 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 429 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
430 | struct disk_conf *disk_conf = |
431 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
432 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 433 | put_ldev(device); |
cb703454 PR |
434 | } |
435 | } | |
695d08fa | 436 | rcu_read_unlock(); |
cb703454 | 437 | |
d7fe69c6 PR |
438 | if (fp == FP_NOT_AVAIL) { |
439 | /* IO Suspending works on the whole resource. | |
440 | Do it only for one device. */ | |
441 | vnr = 0; | |
442 | peer_device = idr_get_next(&connection->peer_devices, &vnr); | |
443 | drbd_change_state(peer_device->device, CS_VERBOSE | CS_HARD, NS(susp_fen, 0)); | |
444 | } | |
445 | ||
cb703454 PR |
446 | return fp; |
447 | } | |
448 | ||
bde89a9e | 449 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 450 | { |
28e448bb | 451 | unsigned int connect_cnt; |
cb703454 PR |
452 | union drbd_state mask = { }; |
453 | union drbd_state val = { }; | |
454 | enum drbd_fencing_p fp; | |
b411b363 PR |
455 | char *ex_to_string; |
456 | int r; | |
b411b363 | 457 | |
02df6fe1 | 458 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 459 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 460 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
02df6fe1 | 461 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 PR |
462 | return false; |
463 | } | |
b411b363 | 464 | |
bde89a9e | 465 | connect_cnt = connection->connect_cnt; |
0500813f | 466 | spin_unlock_irq(&connection->resource->req_lock); |
28e448bb | 467 | |
bde89a9e | 468 | fp = highest_fencing_policy(connection); |
cb703454 PR |
469 | switch (fp) { |
470 | case FP_NOT_AVAIL: | |
1ec861eb | 471 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 472 | goto out; |
cb703454 PR |
473 | case FP_DONT_CARE: |
474 | return true; | |
475 | default: ; | |
b411b363 PR |
476 | } |
477 | ||
bde89a9e | 478 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
479 | |
480 | switch ((r>>8) & 0xff) { | |
481 | case 3: /* peer is inconsistent */ | |
482 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
483 | mask.pdsk = D_MASK; |
484 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
485 | break; |
486 | case 4: /* peer got outdated, or was already outdated */ | |
487 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
488 | mask.pdsk = D_MASK; |
489 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
490 | break; |
491 | case 5: /* peer was down */ | |
bde89a9e | 492 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
493 | /* we will(have) create(d) a new UUID anyways... */ |
494 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
495 | mask.pdsk = D_MASK; |
496 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
497 | } else { |
498 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
499 | } |
500 | break; | |
501 | case 6: /* Peer is primary, voluntarily outdate myself. | |
502 | * This is useful when an unconnected R_SECONDARY is asked to | |
503 | * become R_PRIMARY, but finds the other peer being active. */ | |
504 | ex_to_string = "peer is active"; | |
1ec861eb | 505 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
506 | mask.disk = D_MASK; |
507 | val.disk = D_OUTDATED; | |
b411b363 PR |
508 | break; |
509 | case 7: | |
510 | if (fp != FP_STONITH) | |
1ec861eb | 511 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 512 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
513 | mask.pdsk = D_MASK; |
514 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
515 | break; |
516 | default: | |
517 | /* The script is broken ... */ | |
1ec861eb | 518 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 519 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
520 | } |
521 | ||
1ec861eb | 522 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 523 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 524 | |
cb703454 | 525 | out: |
fb22c402 | 526 | |
cb703454 | 527 | /* Not using |
bde89a9e | 528 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
529 | here, because we might were able to re-establish the connection in the |
530 | meantime. */ | |
0500813f | 531 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
532 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
533 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
534 | /* In case the connection was established and droped |
535 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 536 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 537 | else |
bde89a9e | 538 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 539 | } |
0500813f | 540 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 | 541 | |
bde89a9e | 542 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
543 | } |
544 | ||
87f7be4c PR |
545 | static int _try_outdate_peer_async(void *data) |
546 | { | |
bde89a9e | 547 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 548 | |
bde89a9e | 549 | conn_try_outdate_peer(connection); |
87f7be4c | 550 | |
05a10ec7 | 551 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
552 | return 0; |
553 | } | |
554 | ||
bde89a9e | 555 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
556 | { |
557 | struct task_struct *opa; | |
558 | ||
bde89a9e | 559 | kref_get(&connection->kref); |
bbc1c5e8 LE |
560 | /* We may just have force_sig()'ed this thread |
561 | * to get it out of some blocking network function. | |
562 | * Clear signals; otherwise kthread_run(), which internally uses | |
563 | * wait_on_completion_killable(), will mistake our pending signal | |
564 | * for a new fatal signal and fail. */ | |
565 | flush_signals(current); | |
bde89a9e | 566 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); |
9dc9fbb3 | 567 | if (IS_ERR(opa)) { |
1ec861eb | 568 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 569 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 570 | } |
87f7be4c | 571 | } |
b411b363 | 572 | |
bf885f8a | 573 | enum drbd_state_rv |
44a4d551 | 574 | drbd_set_role(struct drbd_device *const device, enum drbd_role new_role, int force) |
b411b363 | 575 | { |
44a4d551 LE |
576 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
577 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; | |
b411b363 | 578 | const int max_tries = 4; |
bf885f8a | 579 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 580 | struct net_conf *nc; |
b411b363 PR |
581 | int try = 0; |
582 | int forced = 0; | |
583 | union drbd_state mask, val; | |
b411b363 | 584 | |
b6f85ef9 AG |
585 | if (new_role == R_PRIMARY) { |
586 | struct drbd_connection *connection; | |
587 | ||
588 | /* Detect dead peers as soon as possible. */ | |
589 | ||
590 | rcu_read_lock(); | |
591 | for_each_connection(connection, device->resource) | |
592 | request_ping(connection); | |
593 | rcu_read_unlock(); | |
594 | } | |
b411b363 | 595 | |
b30ab791 | 596 | mutex_lock(device->state_mutex); |
b411b363 PR |
597 | |
598 | mask.i = 0; mask.role = R_MASK; | |
599 | val.i = 0; val.role = new_role; | |
600 | ||
601 | while (try++ < max_tries) { | |
a8821531 | 602 | rv = _drbd_request_state_holding_state_mutex(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
603 | |
604 | /* in case we first succeeded to outdate, | |
605 | * but now suddenly could establish a connection */ | |
bf885f8a | 606 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
607 | val.pdsk = 0; |
608 | mask.pdsk = 0; | |
609 | continue; | |
610 | } | |
611 | ||
bf885f8a | 612 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
613 | (device->state.disk < D_UP_TO_DATE && |
614 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
615 | mask.disk = D_MASK; |
616 | val.disk = D_UP_TO_DATE; | |
617 | forced = 1; | |
618 | continue; | |
619 | } | |
620 | ||
bf885f8a | 621 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 622 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 623 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 624 | |
44a4d551 | 625 | if (conn_try_outdate_peer(connection)) { |
b411b363 PR |
626 | val.disk = D_UP_TO_DATE; |
627 | mask.disk = D_MASK; | |
628 | } | |
b411b363 PR |
629 | continue; |
630 | } | |
631 | ||
bf885f8a | 632 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 633 | goto out; |
bf885f8a | 634 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
44a4d551 | 635 | if (!conn_try_outdate_peer(connection) && force) { |
d0180171 | 636 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
637 | mask.pdsk = D_MASK; |
638 | val.pdsk = D_OUTDATED; | |
b411b363 | 639 | |
cb703454 | 640 | } |
b411b363 PR |
641 | continue; |
642 | } | |
bf885f8a | 643 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
644 | /* Maybe the peer is detected as dead very soon... |
645 | retry at most once more in this case. */ | |
44ed167d PR |
646 | int timeo; |
647 | rcu_read_lock(); | |
44a4d551 | 648 | nc = rcu_dereference(connection->net_conf); |
44ed167d PR |
649 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
650 | rcu_read_unlock(); | |
651 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
652 | if (try < max_tries) |
653 | try = max_tries - 1; | |
654 | continue; | |
655 | } | |
bf885f8a | 656 | if (rv < SS_SUCCESS) { |
b30ab791 | 657 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 658 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 659 | if (rv < SS_SUCCESS) |
3b98c0c2 | 660 | goto out; |
b411b363 PR |
661 | } |
662 | break; | |
663 | } | |
664 | ||
bf885f8a | 665 | if (rv < SS_SUCCESS) |
3b98c0c2 | 666 | goto out; |
b411b363 PR |
667 | |
668 | if (forced) | |
d0180171 | 669 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
670 | |
671 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 672 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 673 | |
b6dd1a89 LE |
674 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
675 | ||
b411b363 | 676 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
677 | if (get_ldev(device)) { |
678 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
679 | put_ldev(device); | |
b411b363 PR |
680 | } |
681 | } else { | |
66ce6dbc | 682 | mutex_lock(&device->resource->conf_update); |
44a4d551 | 683 | nc = connection->net_conf; |
44ed167d | 684 | if (nc) |
6139f60d | 685 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
66ce6dbc | 686 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 687 | |
b30ab791 AG |
688 | if (get_ldev(device)) { |
689 | if (((device->state.conn < C_CONNECTED || | |
690 | device->state.pdsk <= D_FAILED) | |
691 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
692 | drbd_uuid_new_current(device); | |
b411b363 | 693 | |
b30ab791 AG |
694 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
695 | put_ldev(device); | |
b411b363 PR |
696 | } |
697 | } | |
698 | ||
19f843aa LE |
699 | /* writeout of activity log covered areas of the bitmap |
700 | * to stable storage done in after state change already */ | |
b411b363 | 701 | |
b30ab791 | 702 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
703 | /* if this was forced, we should consider sync */ |
704 | if (forced) | |
44a4d551 LE |
705 | drbd_send_uuids(peer_device); |
706 | drbd_send_current_state(peer_device); | |
b411b363 PR |
707 | } |
708 | ||
b30ab791 | 709 | drbd_md_sync(device); |
720979fb | 710 | set_disk_ro(device->vdisk, new_role == R_SECONDARY); |
b30ab791 | 711 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 712 | out: |
b30ab791 | 713 | mutex_unlock(device->state_mutex); |
bf885f8a | 714 | return rv; |
b411b363 PR |
715 | } |
716 | ||
3b98c0c2 | 717 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 718 | { |
3b98c0c2 LE |
719 | return err == -ENOMSG ? "required attribute missing" : |
720 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 721 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 722 | "invalid attribute value"; |
ef50a3e3 | 723 | } |
b411b363 | 724 | |
3b98c0c2 | 725 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 726 | { |
a910b123 | 727 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
728 | struct set_role_parms parms; |
729 | int err; | |
730 | enum drbd_ret_code retcode; | |
b411b363 | 731 | |
a910b123 | 732 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
733 | if (!adm_ctx.reply_skb) |
734 | return retcode; | |
735 | if (retcode != NO_ERROR) | |
736 | goto out; | |
b411b363 | 737 | |
3b98c0c2 LE |
738 | memset(&parms, 0, sizeof(parms)); |
739 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 740 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
741 | if (err) { |
742 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 743 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
744 | goto out; |
745 | } | |
746 | } | |
9e276872 LE |
747 | genl_unlock(); |
748 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 749 | |
3b98c0c2 | 750 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 751 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 752 | else |
b30ab791 | 753 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
754 | |
755 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
756 | genl_lock(); | |
3b98c0c2 | 757 | out: |
a910b123 | 758 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
759 | return 0; |
760 | } | |
761 | ||
ae8bf312 LE |
762 | /* Initializes the md.*_offset members, so we are able to find |
763 | * the on disk meta data. | |
764 | * | |
765 | * We currently have two possible layouts: | |
766 | * external: | |
767 | * |----------- md_size_sect ------------------| | |
768 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
769 | * | al_offset == 8 | | |
770 | * | bm_offset = al_offset + X | | |
771 | * ==> bitmap sectors = md_size_sect - bm_offset | |
772 | * | |
773 | * internal: | |
774 | * |----------- md_size_sect ------------------| | |
775 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
776 | * | al_offset < 0 | | |
777 | * | bm_offset = al_offset - Y | | |
778 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
779 | * | |
780 | * Activity log size used to be fixed 32kB, | |
781 | * but is about to become configurable. | |
782 | */ | |
b30ab791 | 783 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
784 | struct drbd_backing_dev *bdev) |
785 | { | |
786 | sector_t md_size_sect = 0; | |
c04ccaa6 | 787 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 788 | |
3a4d4eb3 LE |
789 | bdev->md.md_offset = drbd_md_ss(bdev); |
790 | ||
68e41a43 | 791 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
792 | default: |
793 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 794 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
795 | bdev->md.al_offset = MD_4kB_SECT; |
796 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
797 | break; |
798 | case DRBD_MD_INDEX_FLEX_EXT: | |
799 | /* just occupy the full device; unit: sectors */ | |
800 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
801 | bdev->md.al_offset = MD_4kB_SECT; |
802 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
803 | break; |
804 | case DRBD_MD_INDEX_INTERNAL: | |
805 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 806 | /* al size is still fixed */ |
ae8bf312 | 807 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
808 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
809 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
810 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
811 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
812 | md_size_sect = ALIGN(md_size_sect, 8); | |
813 | ||
814 | /* plus the "drbd meta data super block", | |
815 | * and the activity log; */ | |
ae8bf312 | 816 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
817 | |
818 | bdev->md.md_size_sect = md_size_sect; | |
819 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 820 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
821 | break; |
822 | } | |
823 | } | |
824 | ||
4b0715f0 | 825 | /* input size is expected to be in KB */ |
b411b363 PR |
826 | char *ppsize(char *buf, unsigned long long size) |
827 | { | |
4b0715f0 LE |
828 | /* Needs 9 bytes at max including trailing NUL: |
829 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
830 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
831 | int base = 0; | |
4b0715f0 | 832 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
833 | /* shift + round */ |
834 | size = (size >> 10) + !!(size & (1<<9)); | |
835 | base++; | |
836 | } | |
4b0715f0 | 837 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
838 | |
839 | return buf; | |
840 | } | |
841 | ||
842 | /* there is still a theoretical deadlock when called from receiver | |
843 | * on an D_INCONSISTENT R_PRIMARY: | |
844 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
845 | * packet from remote to dec_ap_bio again. | |
846 | * receiver receive_sizes(), comes here, | |
847 | * waits for ap_bio_cnt == 0. -> deadlock. | |
848 | * but this cannot happen, actually, because: | |
849 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
850 | * (not connected, or bad/no disk on peer): | |
851 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
852 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
853 | * peer may not initiate a resize. | |
854 | */ | |
3b98c0c2 LE |
855 | /* Note these are not to be confused with |
856 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
857 | * which are (sub) state changes triggered by admin (drbdsetup), | |
858 | * and can be long lived. | |
b30ab791 | 859 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 860 | * and should be short-lived. */ |
b30ab791 | 861 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 862 | { |
b30ab791 AG |
863 | set_bit(SUSPEND_IO, &device->flags); |
864 | if (drbd_suspended(device)) | |
265be2d0 | 865 | return; |
b30ab791 | 866 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
867 | } |
868 | ||
b30ab791 | 869 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 870 | { |
b30ab791 AG |
871 | clear_bit(SUSPEND_IO, &device->flags); |
872 | wake_up(&device->misc_wait); | |
b411b363 PR |
873 | } |
874 | ||
875 | /** | |
876 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 877 | * @device: DRBD device. |
b411b363 PR |
878 | * |
879 | * Returns 0 on success, negative return values indicate errors. | |
880 | * You should call drbd_md_sync() after calling this function. | |
881 | */ | |
d752b269 | 882 | enum determine_dev_size |
b30ab791 | 883 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
884 | { |
885 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 886 | sector_t la_size_sect, u_size; |
b30ab791 | 887 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
888 | u32 prev_al_stripe_size_4k; |
889 | u32 prev_al_stripes; | |
b411b363 PR |
890 | sector_t size; |
891 | char ppb[10]; | |
d752b269 | 892 | void *buffer; |
b411b363 PR |
893 | |
894 | int md_moved, la_size_changed; | |
e96c9633 | 895 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
896 | |
897 | /* race: | |
898 | * application request passes inc_ap_bio, | |
899 | * but then cannot get an AL-reference. | |
900 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
901 | * | |
902 | * to avoid that: | |
903 | * Suspend IO right here. | |
904 | * still lock the act_log to not trigger ASSERTs there. | |
905 | */ | |
b30ab791 | 906 | drbd_suspend_io(device); |
e37d2438 | 907 | buffer = drbd_md_get_buffer(device, __func__); /* Lock meta-data IO */ |
d752b269 | 908 | if (!buffer) { |
b30ab791 | 909 | drbd_resume_io(device); |
d752b269 PR |
910 | return DS_ERROR; |
911 | } | |
b411b363 PR |
912 | |
913 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 914 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 915 | |
b30ab791 AG |
916 | prev_first_sect = drbd_md_first_sector(device->ldev); |
917 | prev_size = device->ldev->md.md_size_sect; | |
918 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 919 | |
d752b269 PR |
920 | if (rs) { |
921 | /* rs is non NULL if we should change the AL layout only */ | |
922 | ||
923 | prev_al_stripes = md->al_stripes; | |
924 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
925 | ||
926 | md->al_stripes = rs->al_stripes; | |
927 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
928 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
929 | } | |
930 | ||
b30ab791 | 931 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 932 | |
daeda1cc | 933 | rcu_read_lock(); |
b30ab791 | 934 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 935 | rcu_read_unlock(); |
b30ab791 | 936 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 937 | |
d752b269 PR |
938 | if (size < la_size_sect) { |
939 | if (rs && u_size == 0) { | |
940 | /* Remove "rs &&" later. This check should always be active, but | |
941 | right now the receiver expects the permissive behavior */ | |
d0180171 | 942 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
943 | "Use --size=%llus for explicit shrink.\n", |
944 | (unsigned long long)size); | |
945 | rv = DS_ERROR_SHRINK; | |
946 | } | |
947 | if (u_size > size) | |
948 | rv = DS_ERROR_SPACE_MD; | |
949 | if (rv != DS_UNCHANGED) | |
950 | goto err_out; | |
951 | } | |
952 | ||
b30ab791 AG |
953 | if (drbd_get_capacity(device->this_bdev) != size || |
954 | drbd_bm_capacity(device) != size) { | |
b411b363 | 955 | int err; |
b30ab791 | 956 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
957 | if (unlikely(err)) { |
958 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 959 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 960 | if (size == 0) { |
d0180171 | 961 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
962 | "Could not allocate bitmap!\n"); |
963 | } else { | |
d0180171 | 964 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
965 | "Leaving size unchanged at size = %lu KB\n", |
966 | (unsigned long)size); | |
967 | } | |
e96c9633 | 968 | rv = DS_ERROR; |
b411b363 PR |
969 | } |
970 | /* racy, see comments above. */ | |
b30ab791 AG |
971 | drbd_set_my_capacity(device, size); |
972 | device->ldev->md.la_size_sect = size; | |
d0180171 | 973 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
974 | (unsigned long long)size>>1); |
975 | } | |
d752b269 PR |
976 | if (rv <= DS_ERROR) |
977 | goto err_out; | |
b411b363 | 978 | |
b30ab791 | 979 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 980 | |
b30ab791 AG |
981 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
982 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 983 | |
d752b269 PR |
984 | if (la_size_changed || md_moved || rs) { |
985 | u32 prev_flags; | |
24dccabb | 986 | |
fcb09674 LE |
987 | /* We do some synchronous IO below, which may take some time. |
988 | * Clear the timer, to avoid scary "timer expired!" messages, | |
989 | * "Superblock" is written out at least twice below, anyways. */ | |
990 | del_timer(&device->md_sync_timer); | |
b30ab791 | 991 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
992 | |
993 | prev_flags = md->flags; | |
994 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 995 | drbd_md_write(device, buffer); |
d752b269 | 996 | |
d0180171 | 997 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
998 | la_size_changed && md_moved ? "size changed and md moved" : |
999 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 1000 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 1001 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 1002 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 1003 | drbd_initialize_al(device, buffer); |
d752b269 PR |
1004 | |
1005 | md->flags = prev_flags; | |
b30ab791 | 1006 | drbd_md_write(device, buffer); |
d752b269 PR |
1007 | |
1008 | if (rs) | |
d0180171 AG |
1009 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
1010 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
1011 | } |
1012 | ||
cccac985 | 1013 | if (size > la_size_sect) |
57737adc | 1014 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 1015 | if (size < la_size_sect) |
e96c9633 | 1016 | rv = DS_SHRUNK; |
d752b269 PR |
1017 | |
1018 | if (0) { | |
1019 | err_out: | |
1020 | if (rs) { | |
1021 | md->al_stripes = prev_al_stripes; | |
1022 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
1023 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
1024 | ||
b30ab791 | 1025 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
1026 | } |
1027 | } | |
b30ab791 AG |
1028 | lc_unlock(device->act_log); |
1029 | wake_up(&device->al_wait); | |
1030 | drbd_md_put_buffer(device); | |
1031 | drbd_resume_io(device); | |
b411b363 PR |
1032 | |
1033 | return rv; | |
1034 | } | |
1035 | ||
1036 | sector_t | |
b30ab791 | 1037 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1038 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1039 | { |
b30ab791 | 1040 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1041 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1042 | sector_t m_size; /* my size */ |
b411b363 PR |
1043 | sector_t size = 0; |
1044 | ||
1045 | m_size = drbd_get_max_capacity(bdev); | |
1046 | ||
b30ab791 | 1047 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1048 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1049 | p_size = m_size; |
1050 | } | |
1051 | ||
b411b363 PR |
1052 | if (p_size && m_size) { |
1053 | size = min_t(sector_t, p_size, m_size); | |
1054 | } else { | |
cccac985 LE |
1055 | if (la_size_sect) { |
1056 | size = la_size_sect; | |
b411b363 PR |
1057 | if (m_size && m_size < size) |
1058 | size = m_size; | |
1059 | if (p_size && p_size < size) | |
1060 | size = p_size; | |
1061 | } else { | |
1062 | if (m_size) | |
1063 | size = m_size; | |
1064 | if (p_size) | |
1065 | size = p_size; | |
1066 | } | |
1067 | } | |
1068 | ||
1069 | if (size == 0) | |
d0180171 | 1070 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1071 | |
1072 | if (u_size) { | |
1073 | if (u_size > size) | |
d0180171 | 1074 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1075 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1076 | else | |
1077 | size = u_size; | |
1078 | } | |
1079 | ||
1080 | return size; | |
1081 | } | |
1082 | ||
1083 | /** | |
1084 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1085 | * @device: DRBD device. |
b411b363 PR |
1086 | * |
1087 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1088 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1089 | * this function. | |
1090 | */ | |
b30ab791 | 1091 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1092 | { |
1093 | struct lru_cache *n, *t; | |
1094 | struct lc_element *e; | |
1095 | unsigned int in_use; | |
1096 | int i; | |
1097 | ||
b30ab791 AG |
1098 | if (device->act_log && |
1099 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1100 | return 0; |
1101 | ||
1102 | in_use = 0; | |
b30ab791 | 1103 | t = device->act_log; |
7ad651b5 | 1104 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1105 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1106 | |
1107 | if (n == NULL) { | |
d0180171 | 1108 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1109 | return -ENOMEM; |
1110 | } | |
b30ab791 | 1111 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1112 | if (t) { |
1113 | for (i = 0; i < t->nr_elements; i++) { | |
1114 | e = lc_element_by_index(t, i); | |
1115 | if (e->refcnt) | |
d0180171 | 1116 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1117 | e->lc_number, e->refcnt); |
1118 | in_use += e->refcnt; | |
1119 | } | |
1120 | } | |
1121 | if (!in_use) | |
b30ab791 AG |
1122 | device->act_log = n; |
1123 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1124 | if (in_use) { |
d0180171 | 1125 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1126 | lc_destroy(n); |
1127 | return -EBUSY; | |
1128 | } else { | |
1129 | if (t) | |
1130 | lc_destroy(t); | |
1131 | } | |
b30ab791 | 1132 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1133 | return 0; |
1134 | } | |
1135 | ||
8fe39aac PR |
1136 | static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev, |
1137 | unsigned int max_bio_size) | |
b411b363 | 1138 | { |
b30ab791 | 1139 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1140 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1141 | unsigned int max_segments = 0; | |
c1b3156f | 1142 | struct request_queue *b = NULL; |
99432fcc | 1143 | |
8fe39aac PR |
1144 | if (bdev) { |
1145 | b = bdev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1146 | |
1147 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1148 | rcu_read_lock(); |
b30ab791 | 1149 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1150 | rcu_read_unlock(); |
c1b3156f PR |
1151 | |
1152 | blk_set_stacking_limits(&q->limits); | |
1153 | blk_queue_max_write_same_sectors(q, 0); | |
99432fcc | 1154 | } |
b411b363 | 1155 | |
b411b363 | 1156 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1157 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1158 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1159 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1160 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1161 | |
c1b3156f | 1162 | if (b) { |
20c68fde LE |
1163 | struct drbd_connection *connection = first_peer_device(device)->connection; |
1164 | ||
1165 | if (blk_queue_discard(b) && | |
1166 | (connection->cstate < C_CONNECTED || connection->agreed_features & FF_TRIM)) { | |
20c68fde LE |
1167 | /* For now, don't allow more than one activity log extent worth of data |
1168 | * to be discarded in one go. We may need to rework drbd_al_begin_io() | |
1169 | * to allow for even larger discard ranges */ | |
2bb4cd5c | 1170 | blk_queue_max_discard_sectors(q, DRBD_MAX_DISCARD_SECTORS); |
20c68fde LE |
1171 | |
1172 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1173 | /* REALLY? Is stacking secdiscard "legal"? */ | |
1174 | if (blk_queue_secdiscard(b)) | |
1175 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1176 | } else { | |
2bb4cd5c | 1177 | blk_queue_max_discard_sectors(q, 0); |
20c68fde LE |
1178 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); |
1179 | queue_flag_clear_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1180 | } | |
99432fcc PR |
1181 | |
1182 | blk_queue_stack_limits(q, b); | |
1183 | ||
1184 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1185 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1186 | q->backing_dev_info.ra_pages, |
1187 | b->backing_dev_info.ra_pages); | |
1188 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1189 | } | |
99432fcc PR |
1190 | } |
1191 | } | |
1192 | ||
8fe39aac | 1193 | void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev) |
99432fcc | 1194 | { |
db141b2f | 1195 | unsigned int now, new, local, peer; |
99432fcc | 1196 | |
b30ab791 AG |
1197 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1198 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1199 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1200 | |
8fe39aac PR |
1201 | if (bdev) { |
1202 | local = queue_max_hw_sectors(bdev->backing_bdev->bd_disk->queue) << 9; | |
b30ab791 | 1203 | device->local_max_bio_size = local; |
b411b363 | 1204 | } |
db141b2f | 1205 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1206 | |
1207 | /* We may ignore peer limits if the peer is modern enough. | |
1208 | Because new from 8.3.8 onwards the peer can use multiple | |
1209 | BIOs for a single peer_request */ | |
b30ab791 | 1210 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1211 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1212 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1213 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1214 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1215 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1216 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1217 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1218 | else | |
99432fcc | 1219 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1220 | |
fa090e70 LE |
1221 | /* We may later detach and re-attach on a disconnected Primary. |
1222 | * Avoid this setting to jump back in that case. | |
1223 | * We want to store what we know the peer DRBD can handle, | |
1224 | * not what the peer IO backend can handle. */ | |
1225 | if (peer > device->peer_max_bio_size) | |
1226 | device->peer_max_bio_size = peer; | |
1227 | } | |
db141b2f | 1228 | new = min(local, peer); |
99432fcc | 1229 | |
b30ab791 | 1230 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1231 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1232 | |
1233 | if (new != now) | |
d0180171 | 1234 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1235 | |
8fe39aac | 1236 | drbd_setup_queue_param(device, bdev, new); |
b411b363 PR |
1237 | } |
1238 | ||
a18e9d1e | 1239 | /* Starts the worker thread */ |
bde89a9e | 1240 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1241 | { |
bde89a9e | 1242 | drbd_thread_start(&connection->worker); |
b5043c5e | 1243 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1244 | } |
1245 | ||
a18e9d1e | 1246 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1247 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1248 | { |
992d6e91 | 1249 | bool stop_threads; |
0500813f | 1250 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1251 | stop_threads = conn_all_vols_unconf(connection) && |
1252 | connection->cstate == C_STANDALONE; | |
0500813f | 1253 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1254 | if (stop_threads) { |
1255 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1256 | * in conn_disconnect() */ |
bde89a9e AG |
1257 | drbd_thread_stop(&connection->receiver); |
1258 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1259 | } |
b411b363 PR |
1260 | } |
1261 | ||
0778286a | 1262 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1263 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1264 | { |
1265 | int s = 0; | |
1266 | ||
b30ab791 | 1267 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1268 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1269 | return; |
1270 | } | |
1271 | ||
b30ab791 | 1272 | drbd_al_shrink(device); |
0500813f | 1273 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1274 | if (device->state.conn < C_CONNECTED) |
1275 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1276 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1277 | lc_unlock(device->act_log); |
0778286a PR |
1278 | |
1279 | if (s) | |
d0180171 | 1280 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1281 | } |
1282 | ||
5979e361 LE |
1283 | |
1284 | static bool should_set_defaults(struct genl_info *info) | |
1285 | { | |
1286 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1287 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1288 | } | |
1289 | ||
5bbcf5e6 | 1290 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1291 | { |
5bbcf5e6 LE |
1292 | /* This is limited by 16 bit "slot" numbers, |
1293 | * and by available on-disk context storage. | |
1294 | * | |
1295 | * Also (u16)~0 is special (denotes a "free" extent). | |
1296 | * | |
1297 | * One transaction occupies one 4kB on-disk block, | |
1298 | * we have n such blocks in the on disk ring buffer, | |
1299 | * the "current" transaction may fail (n-1), | |
1300 | * and there is 919 slot numbers context information per transaction. | |
1301 | * | |
1302 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1303 | * so cap there first. | |
1304 | */ | |
1305 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1306 | const unsigned int sufficient_on_disk = | |
1307 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1308 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1309 | |
5bbcf5e6 LE |
1310 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1311 | ||
1312 | if (al_size_4k > sufficient_on_disk) | |
1313 | return max_al_nr; | |
1314 | ||
1315 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1316 | } |
1317 | ||
70df7092 LE |
1318 | static bool write_ordering_changed(struct disk_conf *a, struct disk_conf *b) |
1319 | { | |
1320 | return a->disk_barrier != b->disk_barrier || | |
1321 | a->disk_flushes != b->disk_flushes || | |
1322 | a->disk_drain != b->disk_drain; | |
1323 | } | |
1324 | ||
f399002e LE |
1325 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1326 | { | |
a910b123 | 1327 | struct drbd_config_context adm_ctx; |
f399002e | 1328 | enum drbd_ret_code retcode; |
b30ab791 | 1329 | struct drbd_device *device; |
daeda1cc | 1330 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1331 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1332 | int err, fifo_size; |
f399002e | 1333 | |
a910b123 | 1334 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1335 | if (!adm_ctx.reply_skb) |
1336 | return retcode; | |
1337 | if (retcode != NO_ERROR) | |
9e276872 | 1338 | goto finish; |
f399002e | 1339 | |
b30ab791 | 1340 | device = adm_ctx.device; |
9e276872 | 1341 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1342 | |
1343 | /* we also need a disk | |
1344 | * to change the options on */ | |
b30ab791 | 1345 | if (!get_ldev(device)) { |
f399002e LE |
1346 | retcode = ERR_NO_DISK; |
1347 | goto out; | |
1348 | } | |
1349 | ||
daeda1cc | 1350 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1351 | if (!new_disk_conf) { |
f399002e LE |
1352 | retcode = ERR_NOMEM; |
1353 | goto fail; | |
1354 | } | |
1355 | ||
0500813f | 1356 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1357 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1358 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1359 | if (should_set_defaults(info)) |
b966b5dd | 1360 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1361 | |
5ecc72c3 | 1362 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1363 | if (err && err != -ENOMSG) { |
f399002e | 1364 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1365 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1366 | goto fail_unlock; |
f399002e LE |
1367 | } |
1368 | ||
5ecc72c3 LE |
1369 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1370 | new_disk_conf->resync_rate = 1; | |
f399002e | 1371 | |
5bbcf5e6 LE |
1372 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1373 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1374 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1375 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1376 | |
1377 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1378 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1379 | |
5ecc72c3 | 1380 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1381 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1382 | new_plan = fifo_alloc(fifo_size); |
1383 | if (!new_plan) { | |
d0180171 | 1384 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1385 | retcode = ERR_NOMEM; |
daeda1cc | 1386 | goto fail_unlock; |
f399002e LE |
1387 | } |
1388 | } | |
1389 | ||
b30ab791 AG |
1390 | drbd_suspend_io(device); |
1391 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1392 | drbd_al_shrink(device); | |
1393 | err = drbd_check_al_size(device, new_disk_conf); | |
1394 | lc_unlock(device->act_log); | |
1395 | wake_up(&device->al_wait); | |
1396 | drbd_resume_io(device); | |
f399002e LE |
1397 | |
1398 | if (err) { | |
1399 | retcode = ERR_NOMEM; | |
daeda1cc | 1400 | goto fail_unlock; |
f399002e LE |
1401 | } |
1402 | ||
28bc3b8c | 1403 | lock_all_resources(); |
b30ab791 | 1404 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1405 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1406 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1407 | drbd_resync_after_changed(device); | |
dc97b708 | 1408 | } |
28bc3b8c | 1409 | unlock_all_resources(); |
f399002e | 1410 | |
daeda1cc PR |
1411 | if (retcode != NO_ERROR) |
1412 | goto fail_unlock; | |
f399002e | 1413 | |
813472ce | 1414 | if (new_plan) { |
b30ab791 AG |
1415 | old_plan = device->rs_plan_s; |
1416 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1417 | } |
9958c857 | 1418 | |
0500813f | 1419 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1420 | |
9a51ab1c | 1421 | if (new_disk_conf->al_updates) |
b30ab791 | 1422 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1423 | else |
b30ab791 | 1424 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1425 | |
691631c0 | 1426 | if (new_disk_conf->md_flushes) |
b30ab791 | 1427 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1428 | else |
b30ab791 | 1429 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1430 | |
70df7092 | 1431 | if (write_ordering_changed(old_disk_conf, new_disk_conf)) |
f6ba8636 | 1432 | drbd_bump_write_ordering(device->resource, NULL, WO_BDEV_FLUSH); |
27eb13e9 | 1433 | |
b30ab791 | 1434 | drbd_md_sync(device); |
f399002e | 1435 | |
69a22773 AG |
1436 | if (device->state.conn >= C_CONNECTED) { |
1437 | struct drbd_peer_device *peer_device; | |
1438 | ||
1439 | for_each_peer_device(peer_device, device) | |
1440 | drbd_send_sync_param(peer_device); | |
1441 | } | |
f399002e | 1442 | |
daeda1cc PR |
1443 | synchronize_rcu(); |
1444 | kfree(old_disk_conf); | |
813472ce | 1445 | kfree(old_plan); |
b30ab791 | 1446 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1447 | goto success; |
1448 | ||
1449 | fail_unlock: | |
0500813f | 1450 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1451 | fail: |
5ecc72c3 | 1452 | kfree(new_disk_conf); |
813472ce | 1453 | kfree(new_plan); |
daeda1cc | 1454 | success: |
b30ab791 | 1455 | put_ldev(device); |
f399002e | 1456 | out: |
9e276872 LE |
1457 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1458 | finish: | |
a910b123 | 1459 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1460 | return 0; |
1461 | } | |
1462 | ||
3b98c0c2 | 1463 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1464 | { |
a910b123 | 1465 | struct drbd_config_context adm_ctx; |
b30ab791 | 1466 | struct drbd_device *device; |
44a4d551 LE |
1467 | struct drbd_peer_device *peer_device; |
1468 | struct drbd_connection *connection; | |
3b98c0c2 | 1469 | int err; |
116676ca | 1470 | enum drbd_ret_code retcode; |
b411b363 PR |
1471 | enum determine_dev_size dd; |
1472 | sector_t max_possible_sectors; | |
1473 | sector_t min_md_device_sectors; | |
1474 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1475 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1476 | struct block_device *bdev; |
b411b363 | 1477 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1478 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1479 | union drbd_state ns, os; |
f2024e7c | 1480 | enum drbd_state_rv rv; |
44ed167d | 1481 | struct net_conf *nc; |
b411b363 | 1482 | |
a910b123 | 1483 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1484 | if (!adm_ctx.reply_skb) |
1485 | return retcode; | |
1486 | if (retcode != NO_ERROR) | |
40cbf085 | 1487 | goto finish; |
b411b363 | 1488 | |
b30ab791 | 1489 | device = adm_ctx.device; |
9e276872 | 1490 | mutex_lock(&adm_ctx.resource->adm_mutex); |
44a4d551 | 1491 | peer_device = first_peer_device(device); |
3b8a44f8 | 1492 | connection = peer_device->connection; |
44a4d551 | 1493 | conn_reconfig_start(connection); |
b411b363 PR |
1494 | |
1495 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1496 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1497 | retcode = ERR_DISK_CONFIGURED; |
1498 | goto fail; | |
1499 | } | |
82f59cc6 LE |
1500 | /* It may just now have detached because of IO error. Make sure |
1501 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1502 | * e.g. if someone calls attach from the on-io-error handler, | |
1503 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
e334f550 | 1504 | wait_event(device->misc_wait, !test_bit(GOING_DISKLESS, &device->flags)); |
b411b363 | 1505 | |
383606e0 | 1506 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1507 | clear_bit(FORCE_DETACH, &device->flags); |
1508 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1509 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1510 | |
0029d624 | 1511 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1512 | device->rs_total = 0; |
1513 | device->rs_failed = 0; | |
1514 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1515 | |
3b98c0c2 | 1516 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1517 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1518 | if (!nbc) { | |
1519 | retcode = ERR_NOMEM; | |
1520 | goto fail; | |
1521 | } | |
9f2247bb PR |
1522 | spin_lock_init(&nbc->md.uuid_lock); |
1523 | ||
daeda1cc PR |
1524 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1525 | if (!new_disk_conf) { | |
1526 | retcode = ERR_NOMEM; | |
b411b363 PR |
1527 | goto fail; |
1528 | } | |
daeda1cc | 1529 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1530 | |
daeda1cc PR |
1531 | set_disk_conf_defaults(new_disk_conf); |
1532 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1533 | if (err) { |
b411b363 | 1534 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1535 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1536 | goto fail; |
1537 | } | |
1538 | ||
5bbcf5e6 LE |
1539 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1540 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1541 | |
9958c857 PR |
1542 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1543 | if (!new_plan) { | |
1544 | retcode = ERR_NOMEM; | |
1545 | goto fail; | |
1546 | } | |
1547 | ||
daeda1cc | 1548 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1549 | retcode = ERR_MD_IDX_INVALID; |
1550 | goto fail; | |
1551 | } | |
1552 | ||
44ed167d | 1553 | rcu_read_lock(); |
44a4d551 | 1554 | nc = rcu_dereference(connection->net_conf); |
44ed167d | 1555 | if (nc) { |
daeda1cc | 1556 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1557 | rcu_read_unlock(); |
47ff2d0a | 1558 | retcode = ERR_STONITH_AND_PROT_A; |
28bc3b8c | 1559 | goto fail; |
47ff2d0a PR |
1560 | } |
1561 | } | |
44ed167d | 1562 | rcu_read_unlock(); |
47ff2d0a | 1563 | |
daeda1cc | 1564 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1565 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1566 | if (IS_ERR(bdev)) { |
d0180171 | 1567 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1568 | PTR_ERR(bdev)); |
b411b363 | 1569 | retcode = ERR_OPEN_DISK; |
28bc3b8c | 1570 | goto fail; |
b411b363 | 1571 | } |
e525fd89 TH |
1572 | nbc->backing_bdev = bdev; |
1573 | ||
1574 | /* | |
1575 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1576 | * drbd sharing one meta device. TODO in that case, paranoia | |
1577 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1578 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1579 | * should check it for you already; but if you don't, or | |
1580 | * someone fooled it, we need to double check here) | |
1581 | */ | |
daeda1cc | 1582 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1583 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1584 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1585 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1586 | if (IS_ERR(bdev)) { |
d0180171 | 1587 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1588 | PTR_ERR(bdev)); |
b411b363 | 1589 | retcode = ERR_OPEN_MD_DISK; |
28bc3b8c | 1590 | goto fail; |
b411b363 | 1591 | } |
e525fd89 | 1592 | nbc->md_bdev = bdev; |
b411b363 | 1593 | |
e525fd89 | 1594 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1595 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1596 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1597 | retcode = ERR_MD_IDX_INVALID; |
28bc3b8c | 1598 | goto fail; |
b411b363 PR |
1599 | } |
1600 | ||
1601 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1602 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1603 | offsetof(struct bm_extent, lce)); |
1604 | if (!resync_lru) { | |
1605 | retcode = ERR_NOMEM; | |
28bc3b8c | 1606 | goto fail; |
b411b363 PR |
1607 | } |
1608 | ||
c04ccaa6 LE |
1609 | /* Read our meta data super block early. |
1610 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1611 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 | 1612 | if (retcode != NO_ERROR) |
28bc3b8c | 1613 | goto fail; |
b411b363 | 1614 | |
5bbcf5e6 LE |
1615 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1616 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1617 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1618 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1619 | ||
daeda1cc | 1620 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1621 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1622 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1623 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1624 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1625 | goto fail; |
b411b363 PR |
1626 | } |
1627 | ||
daeda1cc | 1628 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1629 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1630 | /* at least one MB, otherwise it does not make sense */ | |
1631 | min_md_device_sectors = (2<<10); | |
1632 | } else { | |
1633 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1634 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1635 | } |
1636 | ||
b411b363 | 1637 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1638 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1639 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1640 | "at least %llu sectors needed for this meta-disk type\n", |
1641 | (unsigned long long) min_md_device_sectors); | |
28bc3b8c | 1642 | goto fail; |
b411b363 PR |
1643 | } |
1644 | ||
1645 | /* Make sure the new disk is big enough | |
1646 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1647 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1648 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1649 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1650 | goto fail; |
b411b363 PR |
1651 | } |
1652 | ||
1653 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1654 | ||
1352994b | 1655 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1656 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1657 | "to currently maximum possible %llu sectors <==\n", |
1658 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1659 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1660 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1661 | "meta data may help <<==\n"); |
1662 | } | |
1663 | ||
b30ab791 | 1664 | drbd_suspend_io(device); |
b411b363 | 1665 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1666 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1667 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1668 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1669 | * As barriers are counted per resource, | |
1670 | * we'd need to suspend io on all devices of a resource. | |
1671 | */ | |
b30ab791 | 1672 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1673 | /* and for any other previously queued work */ |
44a4d551 | 1674 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 1675 | |
b30ab791 | 1676 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1677 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1678 | drbd_resume_io(device); |
f2024e7c | 1679 | if (rv < SS_SUCCESS) |
28bc3b8c | 1680 | goto fail; |
b411b363 | 1681 | |
b30ab791 | 1682 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1683 | goto force_diskless; |
1684 | ||
b30ab791 AG |
1685 | if (!device->bitmap) { |
1686 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1687 | retcode = ERR_NOMEM; |
1688 | goto force_diskless_dec; | |
1689 | } | |
1690 | } | |
1691 | ||
b30ab791 | 1692 | if (device->state.conn < C_CONNECTED && |
babea49e | 1693 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1694 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1695 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1696 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1697 | retcode = ERR_DATA_NOT_CURRENT; |
1698 | goto force_diskless_dec; | |
1699 | } | |
1700 | ||
1701 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1702 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1703 | retcode = ERR_NOMEM; |
1704 | goto force_diskless_dec; | |
1705 | } | |
1706 | ||
1707 | /* Prevent shrinking of consistent devices ! */ | |
1708 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1709 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1710 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1711 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1712 | goto force_diskless_dec; |
1713 | } | |
1714 | ||
28bc3b8c AG |
1715 | lock_all_resources(); |
1716 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); | |
1717 | if (retcode != NO_ERROR) { | |
1718 | unlock_all_resources(); | |
1719 | goto force_diskless_dec; | |
1720 | } | |
1721 | ||
b411b363 PR |
1722 | /* Reset the "barriers don't work" bits here, then force meta data to |
1723 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1724 | if (new_disk_conf->md_flushes) |
b30ab791 | 1725 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1726 | else |
b30ab791 | 1727 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1728 | |
1729 | /* Point of no return reached. | |
1730 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1731 | * now device takes over responsibility, and the state engine should |
b411b363 | 1732 | * clean it up somewhere. */ |
0b0ba1ef | 1733 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1734 | device->ldev = nbc; |
1735 | device->resync = resync_lru; | |
1736 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1737 | nbc = NULL; |
1738 | resync_lru = NULL; | |
daeda1cc | 1739 | new_disk_conf = NULL; |
9958c857 | 1740 | new_plan = NULL; |
b411b363 | 1741 | |
1ec317d3 | 1742 | drbd_resync_after_changed(device); |
f6ba8636 | 1743 | drbd_bump_write_ordering(device->resource, device->ldev, WO_BDEV_FLUSH); |
28bc3b8c | 1744 | unlock_all_resources(); |
b411b363 | 1745 | |
b30ab791 AG |
1746 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1747 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1748 | else |
b30ab791 | 1749 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1750 | |
b30ab791 | 1751 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1752 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1753 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1754 | |
b30ab791 AG |
1755 | device->send_cnt = 0; |
1756 | device->recv_cnt = 0; | |
1757 | device->read_cnt = 0; | |
1758 | device->writ_cnt = 0; | |
b411b363 | 1759 | |
8fe39aac | 1760 | drbd_reconsider_max_bio_size(device, device->ldev); |
b411b363 PR |
1761 | |
1762 | /* If I am currently not R_PRIMARY, | |
1763 | * but meta data primary indicator is set, | |
1764 | * I just now recover from a hard crash, | |
1765 | * and have been R_PRIMARY before that crash. | |
1766 | * | |
1767 | * Now, if I had no connection before that crash | |
1768 | * (have been degraded R_PRIMARY), chances are that | |
1769 | * I won't find my peer now either. | |
1770 | * | |
1771 | * In that case, and _only_ in that case, | |
1772 | * we use the degr-wfc-timeout instead of the default, | |
1773 | * so we can automatically recover from a crash of a | |
1774 | * degraded but active "cluster" after a certain timeout. | |
1775 | */ | |
b30ab791 AG |
1776 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1777 | if (device->state.role != R_PRIMARY && | |
1778 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1779 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1780 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1781 | |
b30ab791 | 1782 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1783 | if (dd <= DS_ERROR) { |
b411b363 PR |
1784 | retcode = ERR_NOMEM_BITMAP; |
1785 | goto force_diskless_dec; | |
e96c9633 | 1786 | } else if (dd == DS_GREW) |
b30ab791 | 1787 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1788 | |
b30ab791 AG |
1789 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1790 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1791 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1792 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1793 | "(aka FullSync)\n"); |
b30ab791 | 1794 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1795 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1796 | retcode = ERR_IO_MD_DISK; |
1797 | goto force_diskless_dec; | |
1798 | } | |
1799 | } else { | |
b30ab791 | 1800 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1801 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1802 | retcode = ERR_IO_MD_DISK; |
1803 | goto force_diskless_dec; | |
1804 | } | |
b411b363 PR |
1805 | } |
1806 | ||
b30ab791 AG |
1807 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1808 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1809 | |
0500813f | 1810 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1811 | os = drbd_read_state(device); |
78bae59b | 1812 | ns = os; |
b411b363 PR |
1813 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1814 | otherwise investigate MDF_WasUpToDate... | |
1815 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1816 | otherwise into D_CONSISTENT state. | |
1817 | */ | |
b30ab791 AG |
1818 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1819 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1820 | ns.disk = D_CONSISTENT; |
1821 | else | |
1822 | ns.disk = D_OUTDATED; | |
1823 | } else { | |
1824 | ns.disk = D_INCONSISTENT; | |
1825 | } | |
1826 | ||
b30ab791 | 1827 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1828 | ns.pdsk = D_OUTDATED; |
1829 | ||
daeda1cc PR |
1830 | rcu_read_lock(); |
1831 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1832 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1833 | ns.disk = D_UP_TO_DATE; |
1834 | ||
1835 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1836 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1837 | this point, because drbd_request_state() modifies these | |
1838 | flags. */ | |
1839 | ||
b30ab791 AG |
1840 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1841 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1842 | else |
b30ab791 | 1843 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1844 | |
1845 | rcu_read_unlock(); | |
1846 | ||
b411b363 PR |
1847 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1848 | state after the negotiation phase. */ | |
b30ab791 AG |
1849 | if (device->state.conn == C_CONNECTED) { |
1850 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1851 | ns.i = os.i; |
1852 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1853 | |
1854 | /* We expect to receive up-to-date UUIDs soon. | |
1855 | To avoid a race in receive_state, free p_uuid while | |
1856 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1857 | kfree(device->p_uuid); |
1858 | device->p_uuid = NULL; | |
b411b363 PR |
1859 | } |
1860 | ||
b30ab791 | 1861 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1862 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1863 | |
1864 | if (rv < SS_SUCCESS) | |
1865 | goto force_diskless_dec; | |
1866 | ||
b30ab791 | 1867 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1868 | |
b30ab791 AG |
1869 | if (device->state.role == R_PRIMARY) |
1870 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1871 | else |
b30ab791 | 1872 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1873 | |
b30ab791 AG |
1874 | drbd_md_mark_dirty(device); |
1875 | drbd_md_sync(device); | |
b411b363 | 1876 | |
b30ab791 AG |
1877 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1878 | put_ldev(device); | |
44a4d551 | 1879 | conn_reconfig_done(connection); |
9e276872 | 1880 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 1881 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1882 | return 0; |
1883 | ||
1884 | force_diskless_dec: | |
b30ab791 | 1885 | put_ldev(device); |
b411b363 | 1886 | force_diskless: |
b30ab791 AG |
1887 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1888 | drbd_md_sync(device); | |
b411b363 | 1889 | fail: |
44a4d551 | 1890 | conn_reconfig_done(connection); |
b411b363 | 1891 | if (nbc) { |
e525fd89 TH |
1892 | if (nbc->backing_bdev) |
1893 | blkdev_put(nbc->backing_bdev, | |
1894 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1895 | if (nbc->md_bdev) | |
1896 | blkdev_put(nbc->md_bdev, | |
1897 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1898 | kfree(nbc); |
1899 | } | |
daeda1cc | 1900 | kfree(new_disk_conf); |
b411b363 | 1901 | lc_destroy(resync_lru); |
9958c857 | 1902 | kfree(new_plan); |
9e276872 | 1903 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 1904 | finish: |
a910b123 | 1905 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1906 | return 0; |
1907 | } | |
1908 | ||
b30ab791 | 1909 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1910 | { |
19f83c76 | 1911 | enum drbd_state_rv retcode; |
9a0d9d03 | 1912 | int ret; |
02ee8f95 | 1913 | |
cdfda633 | 1914 | if (force) { |
b30ab791 AG |
1915 | set_bit(FORCE_DETACH, &device->flags); |
1916 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1917 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1918 | goto out; |
1919 | } | |
1920 | ||
b30ab791 | 1921 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
e37d2438 | 1922 | drbd_md_get_buffer(device, __func__); /* make sure there is no in-flight meta-data IO */ |
b30ab791 AG |
1923 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); |
1924 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1925 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1926 | ret = wait_event_interruptible(device->misc_wait, |
1927 | device->state.disk != D_FAILED); | |
1928 | drbd_resume_io(device); | |
9b2f61ae | 1929 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1930 | retcode = SS_NOTHING_TO_DO; |
1931 | if (ret) | |
1932 | retcode = ERR_INTR; | |
02ee8f95 | 1933 | out: |
85f75dd7 | 1934 | return retcode; |
b411b363 PR |
1935 | } |
1936 | ||
82f59cc6 LE |
1937 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1938 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1939 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1940 | * internal references as well. | |
1941 | * Only then we have finally detached. */ | |
3b98c0c2 | 1942 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1943 | { |
a910b123 | 1944 | struct drbd_config_context adm_ctx; |
116676ca | 1945 | enum drbd_ret_code retcode; |
cdfda633 PR |
1946 | struct detach_parms parms = { }; |
1947 | int err; | |
b411b363 | 1948 | |
a910b123 | 1949 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1950 | if (!adm_ctx.reply_skb) |
1951 | return retcode; | |
1952 | if (retcode != NO_ERROR) | |
1953 | goto out; | |
b411b363 | 1954 | |
cdfda633 PR |
1955 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1956 | err = detach_parms_from_attrs(&parms, info); | |
1957 | if (err) { | |
1958 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 1959 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
1960 | goto out; |
1961 | } | |
b411b363 PR |
1962 | } |
1963 | ||
9e276872 | 1964 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 1965 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 1966 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 1967 | out: |
a910b123 | 1968 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1969 | return 0; |
1970 | } | |
b411b363 | 1971 | |
bde89a9e | 1972 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1973 | { |
c06ece6b | 1974 | struct drbd_peer_device *peer_device; |
695d08fa | 1975 | bool rv = false; |
f399002e LE |
1976 | int vnr; |
1977 | ||
695d08fa | 1978 | rcu_read_lock(); |
c06ece6b AG |
1979 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1980 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1981 | if (device->state.conn == C_SYNC_SOURCE || |
1982 | device->state.conn == C_SYNC_TARGET || | |
1983 | device->state.conn == C_PAUSED_SYNC_S || | |
1984 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1985 | rv = true; |
1986 | break; | |
1987 | } | |
b411b363 | 1988 | } |
695d08fa | 1989 | rcu_read_unlock(); |
b411b363 | 1990 | |
695d08fa | 1991 | return rv; |
f399002e | 1992 | } |
47ff2d0a | 1993 | |
bde89a9e | 1994 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1995 | { |
c06ece6b | 1996 | struct drbd_peer_device *peer_device; |
695d08fa | 1997 | bool rv = false; |
f399002e LE |
1998 | int vnr; |
1999 | ||
695d08fa | 2000 | rcu_read_lock(); |
c06ece6b AG |
2001 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2002 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2003 | if (device->state.conn == C_VERIFY_S || |
2004 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
2005 | rv = true; |
2006 | break; | |
47ff2d0a PR |
2007 | } |
2008 | } | |
695d08fa | 2009 | rcu_read_unlock(); |
b411b363 | 2010 | |
695d08fa | 2011 | return rv; |
f399002e | 2012 | } |
422028b1 | 2013 | |
cd64397c | 2014 | static enum drbd_ret_code |
270eb5c9 | 2015 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 2016 | { |
c06ece6b | 2017 | struct drbd_peer_device *peer_device; |
cd64397c | 2018 | int i; |
b411b363 | 2019 | |
270eb5c9 AG |
2020 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
2021 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 2022 | return ERR_NEED_APV_100; |
b411b363 | 2023 | |
270eb5c9 | 2024 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
2025 | return ERR_NEED_APV_100; |
2026 | ||
270eb5c9 | 2027 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 2028 | return ERR_NEED_APV_100; |
b411b363 PR |
2029 | } |
2030 | ||
270eb5c9 | 2031 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
2032 | conn_highest_role(connection) == R_PRIMARY && |
2033 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 2034 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 2035 | |
270eb5c9 AG |
2036 | if (new_net_conf->two_primaries && |
2037 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2038 | return ERR_NOT_PROTO_C; |
2039 | ||
c06ece6b AG |
2040 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2041 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2042 | if (get_ldev(device)) { |
2043 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2044 | put_ldev(device); | |
270eb5c9 | 2045 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2046 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2047 | } |
270eb5c9 | 2048 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2049 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2050 | } |
2051 | ||
270eb5c9 | 2052 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2053 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2054 | |
cd64397c PR |
2055 | return NO_ERROR; |
2056 | } | |
b411b363 | 2057 | |
44ed167d | 2058 | static enum drbd_ret_code |
270eb5c9 | 2059 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
2060 | { |
2061 | static enum drbd_ret_code rv; | |
c06ece6b | 2062 | struct drbd_peer_device *peer_device; |
44ed167d | 2063 | int i; |
b411b363 | 2064 | |
44ed167d | 2065 | rcu_read_lock(); |
270eb5c9 | 2066 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2067 | rcu_read_unlock(); |
b411b363 | 2068 | |
179e20b8 | 2069 | /* connection->peer_devices protected by genl_lock() here */ |
c06ece6b AG |
2070 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2071 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2072 | if (!device->bitmap) { |
2073 | if (drbd_bm_init(device)) | |
44ed167d | 2074 | return ERR_NOMEM; |
b411b363 PR |
2075 | } |
2076 | } | |
2077 | ||
44ed167d PR |
2078 | return rv; |
2079 | } | |
b411b363 | 2080 | |
0fd0ea06 PR |
2081 | struct crypto { |
2082 | struct crypto_hash *verify_tfm; | |
2083 | struct crypto_hash *csums_tfm; | |
2084 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2085 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2086 | }; |
b411b363 | 2087 | |
0fd0ea06 | 2088 | static int |
4b6ad6d4 | 2089 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2090 | { |
2091 | if (!tfm_name[0]) | |
2092 | return NO_ERROR; | |
b411b363 | 2093 | |
0fd0ea06 PR |
2094 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2095 | if (IS_ERR(*tfm)) { | |
2096 | *tfm = NULL; | |
2097 | return err_alg; | |
b411b363 | 2098 | } |
b411b363 | 2099 | |
0fd0ea06 PR |
2100 | return NO_ERROR; |
2101 | } | |
b411b363 | 2102 | |
0fd0ea06 | 2103 | static enum drbd_ret_code |
270eb5c9 | 2104 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2105 | { |
2106 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2107 | enum drbd_ret_code rv; | |
0fd0ea06 | 2108 | |
270eb5c9 | 2109 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2110 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2111 | if (rv != NO_ERROR) |
2112 | return rv; | |
270eb5c9 | 2113 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2114 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2115 | if (rv != NO_ERROR) |
2116 | return rv; | |
270eb5c9 | 2117 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2118 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2119 | if (rv != NO_ERROR) |
2120 | return rv; | |
270eb5c9 | 2121 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2122 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2123 | new_net_conf->cram_hmac_alg); |
b411b363 | 2124 | |
4b6ad6d4 AG |
2125 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2126 | ERR_AUTH_ALG); | |
b411b363 PR |
2127 | } |
2128 | ||
0fd0ea06 PR |
2129 | return rv; |
2130 | } | |
b411b363 | 2131 | |
0fd0ea06 PR |
2132 | static void free_crypto(struct crypto *crypto) |
2133 | { | |
0fd0ea06 | 2134 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2135 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2136 | crypto_free_hash(crypto->csums_tfm); |
2137 | crypto_free_hash(crypto->verify_tfm); | |
2138 | } | |
b411b363 | 2139 | |
f399002e LE |
2140 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2141 | { | |
a910b123 | 2142 | struct drbd_config_context adm_ctx; |
f399002e | 2143 | enum drbd_ret_code retcode; |
bde89a9e | 2144 | struct drbd_connection *connection; |
270eb5c9 | 2145 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2146 | int err; |
2147 | int ovr; /* online verify running */ | |
2148 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2149 | struct crypto crypto = { }; |
b411b363 | 2150 | |
a910b123 | 2151 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2152 | if (!adm_ctx.reply_skb) |
2153 | return retcode; | |
2154 | if (retcode != NO_ERROR) | |
9e276872 | 2155 | goto finish; |
b411b363 | 2156 | |
bde89a9e | 2157 | connection = adm_ctx.connection; |
9e276872 | 2158 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2159 | |
270eb5c9 AG |
2160 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2161 | if (!new_net_conf) { | |
f399002e LE |
2162 | retcode = ERR_NOMEM; |
2163 | goto out; | |
2164 | } | |
b411b363 | 2165 | |
bde89a9e | 2166 | conn_reconfig_start(connection); |
b411b363 | 2167 | |
bde89a9e | 2168 | mutex_lock(&connection->data.mutex); |
0500813f | 2169 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2170 | old_net_conf = connection->net_conf; |
2561b9c1 | 2171 | |
270eb5c9 | 2172 | if (!old_net_conf) { |
a910b123 | 2173 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2174 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2175 | goto fail; |
2176 | } | |
2177 | ||
270eb5c9 | 2178 | *new_net_conf = *old_net_conf; |
5979e361 | 2179 | if (should_set_defaults(info)) |
270eb5c9 | 2180 | set_net_conf_defaults(new_net_conf); |
f399002e | 2181 | |
270eb5c9 | 2182 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2183 | if (err && err != -ENOMSG) { |
f399002e | 2184 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2185 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2186 | goto fail; |
2561b9c1 | 2187 | } |
b411b363 | 2188 | |
270eb5c9 | 2189 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2190 | if (retcode != NO_ERROR) |
2191 | goto fail; | |
b411b363 | 2192 | |
f399002e | 2193 | /* re-sync running */ |
bde89a9e | 2194 | rsr = conn_resync_running(connection); |
270eb5c9 | 2195 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2196 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2197 | goto fail; |
b411b363 PR |
2198 | } |
2199 | ||
f399002e | 2200 | /* online verify running */ |
bde89a9e | 2201 | ovr = conn_ov_running(connection); |
270eb5c9 | 2202 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2203 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2204 | goto fail; |
f399002e | 2205 | } |
b411b363 | 2206 | |
270eb5c9 | 2207 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2208 | if (retcode != NO_ERROR) |
b411b363 | 2209 | goto fail; |
f399002e | 2210 | |
270eb5c9 | 2211 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2212 | |
2213 | if (!rsr) { | |
bde89a9e AG |
2214 | crypto_free_hash(connection->csums_tfm); |
2215 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2216 | crypto.csums_tfm = NULL; |
f399002e LE |
2217 | } |
2218 | if (!ovr) { | |
bde89a9e AG |
2219 | crypto_free_hash(connection->verify_tfm); |
2220 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2221 | crypto.verify_tfm = NULL; |
b411b363 PR |
2222 | } |
2223 | ||
bde89a9e AG |
2224 | crypto_free_hash(connection->integrity_tfm); |
2225 | connection->integrity_tfm = crypto.integrity_tfm; | |
2226 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2227 | /* Do this without trying to take connection->data.mutex again. */ | |
2228 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2229 | |
bde89a9e AG |
2230 | crypto_free_hash(connection->cram_hmac_tfm); |
2231 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2232 | |
0500813f | 2233 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2234 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2235 | synchronize_rcu(); |
270eb5c9 | 2236 | kfree(old_net_conf); |
91fd4dad | 2237 | |
69a22773 AG |
2238 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2239 | struct drbd_peer_device *peer_device; | |
2240 | int vnr; | |
2241 | ||
2242 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2243 | drbd_send_sync_param(peer_device); | |
2244 | } | |
f399002e | 2245 | |
91fd4dad PR |
2246 | goto done; |
2247 | ||
b411b363 | 2248 | fail: |
0500813f | 2249 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2250 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2251 | free_crypto(&crypto); |
270eb5c9 | 2252 | kfree(new_net_conf); |
91fd4dad | 2253 | done: |
bde89a9e | 2254 | conn_reconfig_done(connection); |
f399002e | 2255 | out: |
9e276872 LE |
2256 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2257 | finish: | |
a910b123 | 2258 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2259 | return 0; |
2260 | } | |
2261 | ||
a2972846 AG |
2262 | static void connection_to_info(struct connection_info *info, |
2263 | struct drbd_connection *connection) | |
2264 | { | |
2265 | info->conn_connection_state = connection->cstate; | |
2266 | info->conn_role = conn_highest_peer(connection); | |
2267 | } | |
2268 | ||
2269 | static void peer_device_to_info(struct peer_device_info *info, | |
2270 | struct drbd_peer_device *peer_device) | |
2271 | { | |
2272 | struct drbd_device *device = peer_device->device; | |
2273 | ||
2274 | info->peer_repl_state = | |
2275 | max_t(enum drbd_conns, C_WF_REPORT_PARAMS, device->state.conn); | |
2276 | info->peer_disk_state = device->state.pdsk; | |
2277 | info->peer_resync_susp_user = device->state.user_isp; | |
2278 | info->peer_resync_susp_peer = device->state.peer_isp; | |
2279 | info->peer_resync_susp_dependency = device->state.aftr_isp; | |
2280 | } | |
2281 | ||
3b98c0c2 | 2282 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2283 | { |
a2972846 AG |
2284 | struct connection_info connection_info; |
2285 | enum drbd_notification_type flags; | |
2286 | unsigned int peer_devices = 0; | |
a910b123 | 2287 | struct drbd_config_context adm_ctx; |
c06ece6b | 2288 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2289 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2290 | struct crypto crypto = { }; |
77c556f6 | 2291 | struct drbd_resource *resource; |
bde89a9e | 2292 | struct drbd_connection *connection; |
3b98c0c2 LE |
2293 | enum drbd_ret_code retcode; |
2294 | int i; | |
2295 | int err; | |
b411b363 | 2296 | |
a910b123 | 2297 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2298 | |
3b98c0c2 LE |
2299 | if (!adm_ctx.reply_skb) |
2300 | return retcode; | |
2301 | if (retcode != NO_ERROR) | |
2302 | goto out; | |
089c075d | 2303 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2304 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2305 | retcode = ERR_INVALID_REQUEST; |
2306 | goto out; | |
2307 | } | |
b411b363 | 2308 | |
089c075d AG |
2309 | /* No need for _rcu here. All reconfiguration is |
2310 | * strictly serialized on genl_lock(). We are protected against | |
2311 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2312 | for_each_resource(resource, &drbd_resources) { |
2313 | for_each_connection(connection, resource) { | |
2314 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2315 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2316 | connection->my_addr_len)) { | |
2317 | retcode = ERR_LOCAL_ADDR; | |
2318 | goto out; | |
2319 | } | |
b411b363 | 2320 | |
77c556f6 AG |
2321 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2322 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2323 | connection->peer_addr_len)) { | |
2324 | retcode = ERR_PEER_ADDR; | |
2325 | goto out; | |
2326 | } | |
089c075d | 2327 | } |
b411b363 PR |
2328 | } |
2329 | ||
9e276872 | 2330 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2331 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2332 | conn_reconfig_start(connection); |
b411b363 | 2333 | |
bde89a9e | 2334 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2335 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2336 | goto fail; |
2337 | } | |
2338 | ||
a209b4ae | 2339 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2340 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2341 | if (!new_net_conf) { | |
b411b363 | 2342 | retcode = ERR_NOMEM; |
b411b363 PR |
2343 | goto fail; |
2344 | } | |
2345 | ||
270eb5c9 | 2346 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2347 | |
270eb5c9 | 2348 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2349 | if (err && err != -ENOMSG) { |
b411b363 | 2350 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2351 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2352 | goto fail; |
2353 | } | |
2354 | ||
270eb5c9 | 2355 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2356 | if (retcode != NO_ERROR) |
422028b1 | 2357 | goto fail; |
b411b363 | 2358 | |
270eb5c9 | 2359 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2360 | if (retcode != NO_ERROR) |
2361 | goto fail; | |
b411b363 | 2362 | |
270eb5c9 | 2363 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2364 | |
b5043c5e | 2365 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2366 | |
0500813f | 2367 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2368 | old_net_conf = connection->net_conf; |
2369 | if (old_net_conf) { | |
b411b363 | 2370 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2371 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2372 | goto fail; |
2373 | } | |
270eb5c9 | 2374 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2375 | |
bde89a9e AG |
2376 | conn_free_crypto(connection); |
2377 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2378 | connection->integrity_tfm = crypto.integrity_tfm; | |
2379 | connection->csums_tfm = crypto.csums_tfm; | |
2380 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2381 | |
bde89a9e AG |
2382 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2383 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2384 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2385 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2386 | |
a2972846 AG |
2387 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2388 | peer_devices++; | |
2389 | } | |
2390 | ||
2391 | connection_to_info(&connection_info, connection); | |
2392 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2393 | mutex_lock(¬ification_mutex); | |
2394 | notify_connection_state(NULL, 0, connection, &connection_info, NOTIFY_CREATE | flags); | |
2395 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
2396 | struct peer_device_info peer_device_info; | |
2397 | ||
2398 | peer_device_to_info(&peer_device_info, peer_device); | |
2399 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2400 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, NOTIFY_CREATE | flags); | |
2401 | } | |
2402 | mutex_unlock(¬ification_mutex); | |
0500813f | 2403 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2404 | |
695d08fa | 2405 | rcu_read_lock(); |
c06ece6b AG |
2406 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2407 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2408 | device->send_cnt = 0; |
2409 | device->recv_cnt = 0; | |
b411b363 | 2410 | } |
695d08fa | 2411 | rcu_read_unlock(); |
b411b363 | 2412 | |
bde89a9e | 2413 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2414 | |
bde89a9e | 2415 | conn_reconfig_done(connection); |
9e276872 | 2416 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2417 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2418 | return 0; |
b411b363 | 2419 | |
b411b363 | 2420 | fail: |
0fd0ea06 | 2421 | free_crypto(&crypto); |
270eb5c9 | 2422 | kfree(new_net_conf); |
b411b363 | 2423 | |
bde89a9e | 2424 | conn_reconfig_done(connection); |
9e276872 | 2425 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2426 | out: |
a910b123 | 2427 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2428 | return 0; |
2429 | } | |
2430 | ||
bde89a9e | 2431 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2432 | { |
2433 | enum drbd_state_rv rv; | |
85f75dd7 | 2434 | |
bde89a9e | 2435 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2436 | force ? CS_HARD : 0); |
85f75dd7 LE |
2437 | |
2438 | switch (rv) { | |
2439 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2440 | break; |
85f75dd7 LE |
2441 | case SS_ALREADY_STANDALONE: |
2442 | return SS_SUCCESS; | |
2443 | case SS_PRIMARY_NOP: | |
2444 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2445 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2446 | |
2447 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2448 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2449 | |
85f75dd7 LE |
2450 | break; |
2451 | case SS_CW_FAILED_BY_PEER: | |
2452 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2453 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2454 | disk, D_OUTDATED), 0); |
2455 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2456 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2457 | CS_HARD); |
b411b363 | 2458 | } |
85f75dd7 LE |
2459 | break; |
2460 | default:; | |
2461 | /* no special handling necessary */ | |
2462 | } | |
2463 | ||
f3dfa40a LE |
2464 | if (rv >= SS_SUCCESS) { |
2465 | enum drbd_state_rv rv2; | |
2466 | /* No one else can reconfigure the network while I am here. | |
2467 | * The state handling only uses drbd_thread_stop_nowait(), | |
2468 | * we want to really wait here until the receiver is no more. | |
2469 | */ | |
9693da23 | 2470 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2471 | |
2472 | /* Race breaker. This additional state change request may be | |
2473 | * necessary, if this was a forced disconnect during a receiver | |
2474 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2475 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2476 | * C_STANDALONE already, now, and this becomes a no-op. |
2477 | */ | |
bde89a9e | 2478 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2479 | CS_VERBOSE | CS_HARD); |
2480 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2481 | drbd_err(connection, |
f3dfa40a LE |
2482 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2483 | rv2); | |
a2972846 AG |
2484 | /* Unlike in DRBD 9, the state engine has generated |
2485 | * NOTIFY_DESTROY events before clearing connection->net_conf. */ | |
b411b363 | 2486 | } |
85f75dd7 LE |
2487 | return rv; |
2488 | } | |
b411b363 | 2489 | |
3b98c0c2 | 2490 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2491 | { |
a910b123 | 2492 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2493 | struct disconnect_parms parms; |
bde89a9e | 2494 | struct drbd_connection *connection; |
85f75dd7 | 2495 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2496 | enum drbd_ret_code retcode; |
2497 | int err; | |
2561b9c1 | 2498 | |
a910b123 | 2499 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2500 | if (!adm_ctx.reply_skb) |
2501 | return retcode; | |
2502 | if (retcode != NO_ERROR) | |
2561b9c1 | 2503 | goto fail; |
b411b363 | 2504 | |
bde89a9e | 2505 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2506 | memset(&parms, 0, sizeof(parms)); |
2507 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2508 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2509 | if (err) { |
2510 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2511 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2512 | goto fail; |
2513 | } | |
2514 | } | |
2515 | ||
9e276872 | 2516 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2517 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2518 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2519 | retcode = rv; /* FIXME: Type mismatch. */ |
2520 | else | |
2521 | retcode = NO_ERROR; | |
9e276872 | 2522 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2523 | fail: |
a910b123 | 2524 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2525 | return 0; |
2526 | } | |
2527 | ||
b30ab791 | 2528 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2529 | { |
2530 | int iass; /* I am sync source */ | |
2531 | ||
d0180171 | 2532 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2533 | if (device->state.role != device->state.peer) |
2534 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2535 | else |
a6b32bc3 | 2536 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2537 | |
2538 | if (iass) | |
b30ab791 | 2539 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2540 | else |
b30ab791 | 2541 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2542 | } |
2543 | ||
3b98c0c2 | 2544 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2545 | { |
a910b123 | 2546 | struct drbd_config_context adm_ctx; |
daeda1cc | 2547 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2548 | struct resize_parms rs; |
b30ab791 | 2549 | struct drbd_device *device; |
3b98c0c2 | 2550 | enum drbd_ret_code retcode; |
b411b363 | 2551 | enum determine_dev_size dd; |
d752b269 | 2552 | bool change_al_layout = false; |
6495d2c6 | 2553 | enum dds_flags ddsf; |
daeda1cc | 2554 | sector_t u_size; |
3b98c0c2 | 2555 | int err; |
b411b363 | 2556 | |
a910b123 | 2557 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2558 | if (!adm_ctx.reply_skb) |
2559 | return retcode; | |
2560 | if (retcode != NO_ERROR) | |
9e276872 | 2561 | goto finish; |
3b98c0c2 | 2562 | |
9e276872 | 2563 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2564 | device = adm_ctx.device; |
2565 | if (!get_ldev(device)) { | |
d752b269 PR |
2566 | retcode = ERR_NO_DISK; |
2567 | goto fail; | |
2568 | } | |
2569 | ||
3b98c0c2 | 2570 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2571 | rs.al_stripes = device->ldev->md.al_stripes; |
2572 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2573 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2574 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2575 | if (err) { |
3b98c0c2 | 2576 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2577 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2578 | goto fail_ldev; |
b411b363 PR |
2579 | } |
2580 | } | |
2581 | ||
b30ab791 | 2582 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2583 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2584 | goto fail_ldev; |
b411b363 | 2585 | } |
b411b363 | 2586 | |
b30ab791 AG |
2587 | if (device->state.role == R_SECONDARY && |
2588 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2589 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2590 | goto fail_ldev; |
b411b363 | 2591 | } |
b411b363 | 2592 | |
a6b32bc3 | 2593 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2594 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2595 | goto fail_ldev; |
6495d2c6 PR |
2596 | } |
2597 | ||
daeda1cc | 2598 | rcu_read_lock(); |
b30ab791 | 2599 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2600 | rcu_read_unlock(); |
2601 | if (u_size != (sector_t)rs.resize_size) { | |
2602 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2603 | if (!new_disk_conf) { | |
778f271d | 2604 | retcode = ERR_NOMEM; |
9bcd2521 | 2605 | goto fail_ldev; |
778f271d PR |
2606 | } |
2607 | } | |
2608 | ||
b30ab791 AG |
2609 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2610 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2611 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2612 | ||
2613 | if (al_size_k > (16 * 1024 * 1024)) { | |
2614 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2615 | goto fail_ldev; | |
2616 | } | |
2617 | ||
2618 | if (al_size_k < MD_32kB_SECT/2) { | |
2619 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2620 | goto fail_ldev; | |
2621 | } | |
2622 | ||
cdc6af8d | 2623 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2624 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2625 | goto fail_ldev; | |
2626 | } | |
2627 | ||
2628 | change_al_layout = true; | |
2629 | } | |
2630 | ||
b30ab791 AG |
2631 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2632 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2633 | |
daeda1cc | 2634 | if (new_disk_conf) { |
0500813f | 2635 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2636 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2637 | *new_disk_conf = *old_disk_conf; |
2638 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2639 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2640 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2641 | synchronize_rcu(); |
2642 | kfree(old_disk_conf); | |
b411b363 PR |
2643 | } |
2644 | ||
6495d2c6 | 2645 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2646 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2647 | drbd_md_sync(device); | |
2648 | put_ldev(device); | |
e96c9633 | 2649 | if (dd == DS_ERROR) { |
b411b363 PR |
2650 | retcode = ERR_NOMEM_BITMAP; |
2651 | goto fail; | |
d752b269 PR |
2652 | } else if (dd == DS_ERROR_SPACE_MD) { |
2653 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2654 | goto fail; | |
2655 | } else if (dd == DS_ERROR_SHRINK) { | |
2656 | retcode = ERR_IMPLICIT_SHRINK; | |
2657 | goto fail; | |
b411b363 | 2658 | } |
778f271d | 2659 | |
b30ab791 | 2660 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2661 | if (dd == DS_GREW) |
b30ab791 | 2662 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2663 | |
69a22773 AG |
2664 | drbd_send_uuids(first_peer_device(device)); |
2665 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2666 | } |
2667 | ||
b411b363 | 2668 | fail: |
9e276872 LE |
2669 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2670 | finish: | |
a910b123 | 2671 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2672 | return 0; |
b411b363 | 2673 | |
9bcd2521 | 2674 | fail_ldev: |
b30ab791 | 2675 | put_ldev(device); |
9bcd2521 | 2676 | goto fail; |
b411b363 | 2677 | } |
b411b363 | 2678 | |
f399002e | 2679 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2680 | { |
a910b123 | 2681 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2682 | enum drbd_ret_code retcode; |
b57a1e27 | 2683 | struct res_opts res_opts; |
f399002e | 2684 | int err; |
b411b363 | 2685 | |
a910b123 | 2686 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2687 | if (!adm_ctx.reply_skb) |
2688 | return retcode; | |
2689 | if (retcode != NO_ERROR) | |
2690 | goto fail; | |
b411b363 | 2691 | |
eb6bea67 | 2692 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2693 | if (should_set_defaults(info)) |
b966b5dd | 2694 | set_res_opts_defaults(&res_opts); |
b411b363 | 2695 | |
b57a1e27 | 2696 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2697 | if (err && err != -ENOMSG) { |
b411b363 | 2698 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2699 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2700 | goto fail; |
2701 | } | |
2702 | ||
9e276872 | 2703 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2704 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2705 | if (err) { |
2706 | retcode = ERR_INVALID_REQUEST; | |
2707 | if (err == -ENOMEM) | |
2708 | retcode = ERR_NOMEM; | |
b411b363 | 2709 | } |
9e276872 | 2710 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2711 | |
b411b363 | 2712 | fail: |
a910b123 | 2713 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2714 | return 0; |
2715 | } | |
2716 | ||
3b98c0c2 | 2717 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2718 | { |
a910b123 | 2719 | struct drbd_config_context adm_ctx; |
b30ab791 | 2720 | struct drbd_device *device; |
3b98c0c2 LE |
2721 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2722 | ||
a910b123 | 2723 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2724 | if (!adm_ctx.reply_skb) |
2725 | return retcode; | |
2726 | if (retcode != NO_ERROR) | |
2727 | goto out; | |
2728 | ||
b30ab791 | 2729 | device = adm_ctx.device; |
8fe39aac PR |
2730 | if (!get_ldev(device)) { |
2731 | retcode = ERR_NO_DISK; | |
2732 | goto out; | |
2733 | } | |
2734 | ||
2735 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2736 | |
194bfb32 | 2737 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2738 | * resync just being finished, wait for it before requesting a new resync. |
2739 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2740 | drbd_suspend_io(device); |
2741 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2742 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2743 | |
0b2dafcd PR |
2744 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2745 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2746 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2747 | */ |
b30ab791 AG |
2748 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2749 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2750 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2751 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2752 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2753 | retcode = ERR_IO_MD_DISK; | |
2754 | } | |
2755 | } else | |
b30ab791 AG |
2756 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2757 | drbd_resume_io(device); | |
9e276872 | 2758 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2759 | put_ldev(device); |
3b98c0c2 | 2760 | out: |
a910b123 | 2761 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2762 | return 0; |
2763 | } | |
2764 | ||
3b98c0c2 LE |
2765 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2766 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2767 | { |
a910b123 | 2768 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2769 | enum drbd_ret_code retcode; |
194bfb32 | 2770 | |
a910b123 | 2771 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2772 | if (!adm_ctx.reply_skb) |
2773 | return retcode; | |
2774 | if (retcode != NO_ERROR) | |
2775 | goto out; | |
b411b363 | 2776 | |
9e276872 | 2777 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2778 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 2779 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2780 | out: |
a910b123 | 2781 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2782 | return 0; |
2783 | } | |
2784 | ||
8fe39aac | 2785 | static int drbd_bmio_set_susp_al(struct drbd_device *device) __must_hold(local) |
0778286a PR |
2786 | { |
2787 | int rv; | |
2788 | ||
b30ab791 AG |
2789 | rv = drbd_bmio_set_n_write(device); |
2790 | drbd_suspend_al(device); | |
0778286a PR |
2791 | return rv; |
2792 | } | |
2793 | ||
3b98c0c2 | 2794 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2795 | { |
a910b123 | 2796 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 2797 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2798 | struct drbd_device *device; |
25b0d6c8 | 2799 | |
a910b123 | 2800 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
2801 | if (!adm_ctx.reply_skb) |
2802 | return retcode; | |
2803 | if (retcode != NO_ERROR) | |
2804 | goto out; | |
2805 | ||
b30ab791 | 2806 | device = adm_ctx.device; |
8fe39aac PR |
2807 | if (!get_ldev(device)) { |
2808 | retcode = ERR_NO_DISK; | |
2809 | goto out; | |
2810 | } | |
2811 | ||
2812 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2813 | |
194bfb32 | 2814 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2815 | * resync just being finished, wait for it before requesting a new resync. |
2816 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2817 | drbd_suspend_io(device); |
2818 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2819 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2820 | |
0b2dafcd PR |
2821 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2822 | * in the bitmap. Otherwise, try to start a resync handshake | |
2823 | * as sync source for full sync. | |
2824 | */ | |
b30ab791 | 2825 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2826 | /* The peer will get a resync upon connect anyways. Just make that |
2827 | into a full resync. */ | |
b30ab791 | 2828 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2829 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2830 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2831 | "set_n_write from invalidate_peer", |
2832 | BM_LOCKED_SET_ALLOWED)) | |
2833 | retcode = ERR_IO_MD_DISK; | |
2834 | } | |
2835 | } else | |
b30ab791 AG |
2836 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2837 | drbd_resume_io(device); | |
9e276872 | 2838 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2839 | put_ldev(device); |
25b0d6c8 | 2840 | out: |
a910b123 | 2841 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2842 | return 0; |
2843 | } | |
2844 | ||
3b98c0c2 | 2845 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2846 | { |
a910b123 | 2847 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2848 | enum drbd_ret_code retcode; |
b411b363 | 2849 | |
a910b123 | 2850 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2851 | if (!adm_ctx.reply_skb) |
2852 | return retcode; | |
2853 | if (retcode != NO_ERROR) | |
2854 | goto out; | |
b411b363 | 2855 | |
9e276872 | 2856 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2857 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 2858 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 2859 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2860 | out: |
a910b123 | 2861 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2862 | return 0; |
2863 | } | |
2864 | ||
3b98c0c2 | 2865 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2866 | { |
a910b123 | 2867 | struct drbd_config_context adm_ctx; |
da9fbc27 | 2868 | union drbd_dev_state s; |
3b98c0c2 LE |
2869 | enum drbd_ret_code retcode; |
2870 | ||
a910b123 | 2871 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2872 | if (!adm_ctx.reply_skb) |
2873 | return retcode; | |
2874 | if (retcode != NO_ERROR) | |
2875 | goto out; | |
b411b363 | 2876 | |
9e276872 | 2877 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2878 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2879 | s = adm_ctx.device->state; | |
cd88d030 PR |
2880 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2881 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2882 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2883 | } else { | |
2884 | retcode = ERR_PAUSE_IS_CLEAR; | |
2885 | } | |
2886 | } | |
9e276872 | 2887 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2888 | out: |
a910b123 | 2889 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2890 | return 0; |
2891 | } | |
2892 | ||
3b98c0c2 | 2893 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2894 | { |
3b98c0c2 | 2895 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2896 | } |
2897 | ||
3b98c0c2 | 2898 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2899 | { |
a910b123 | 2900 | struct drbd_config_context adm_ctx; |
b30ab791 | 2901 | struct drbd_device *device; |
3b98c0c2 LE |
2902 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2903 | ||
a910b123 | 2904 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2905 | if (!adm_ctx.reply_skb) |
2906 | return retcode; | |
2907 | if (retcode != NO_ERROR) | |
2908 | goto out; | |
2909 | ||
9e276872 | 2910 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2911 | device = adm_ctx.device; |
2912 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2913 | drbd_uuid_new_current(device); | |
2914 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2915 | } |
b30ab791 AG |
2916 | drbd_suspend_io(device); |
2917 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2918 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2919 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2920 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2921 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2922 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2923 | } |
b30ab791 | 2924 | drbd_resume_io(device); |
9e276872 | 2925 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2926 | out: |
a910b123 | 2927 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2928 | return 0; |
2929 | } | |
2930 | ||
3b98c0c2 | 2931 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2932 | { |
3b98c0c2 | 2933 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2934 | } |
2935 | ||
251b8f8e AG |
2936 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2937 | struct drbd_resource *resource, | |
2938 | struct drbd_connection *connection, | |
2939 | struct drbd_device *device) | |
b411b363 | 2940 | { |
543cc10b LE |
2941 | struct nlattr *nla; |
2942 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2943 | if (!nla) | |
2944 | goto nla_put_failure; | |
251b8f8e AG |
2945 | if (device && |
2946 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2947 | goto nla_put_failure; |
f597f6b8 | 2948 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 2949 | goto nla_put_failure; |
251b8f8e AG |
2950 | if (connection) { |
2951 | if (connection->my_addr_len && | |
2952 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2953 | goto nla_put_failure; | |
2954 | if (connection->peer_addr_len && | |
2955 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2956 | goto nla_put_failure; | |
2957 | } | |
543cc10b LE |
2958 | nla_nest_end(skb, nla); |
2959 | return 0; | |
b411b363 | 2960 | |
543cc10b LE |
2961 | nla_put_failure: |
2962 | if (nla) | |
2963 | nla_nest_cancel(skb, nla); | |
2964 | return -EMSGSIZE; | |
2965 | } | |
b411b363 | 2966 | |
251b8f8e AG |
2967 | /* |
2968 | * Return the connection of @resource if @resource has exactly one connection. | |
2969 | */ | |
2970 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
2971 | { | |
2972 | struct list_head *connections = &resource->connections; | |
2973 | ||
2974 | if (list_empty(connections) || connections->next->next != connections) | |
2975 | return NULL; | |
2976 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
2977 | } | |
2978 | ||
8ce953aa | 2979 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 2980 | const struct sib_info *sib) |
b411b363 | 2981 | { |
251b8f8e | 2982 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
2983 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2984 | struct nlattr *nla; | |
2985 | int got_ldev; | |
3b98c0c2 LE |
2986 | int err = 0; |
2987 | int exclude_sensitive; | |
2988 | ||
2989 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2990 | * to. So we better exclude_sensitive information. | |
2991 | * | |
2992 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2993 | * in the context of the requesting user process. Exclude sensitive | |
2994 | * information, unless current has superuser. | |
2995 | * | |
2996 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2997 | * relies on the current implementation of netlink_dump(), which | |
2998 | * executes the dump callback successively from netlink_recvmsg(), | |
2999 | * always in the context of the receiving process */ | |
3000 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
3001 | ||
b30ab791 | 3002 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
3003 | |
3004 | /* We need to add connection name and volume number information still. | |
3005 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 3006 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 3007 | goto nla_put_failure; |
3b98c0c2 | 3008 | |
eb6bea67 | 3009 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
3010 | goto nla_put_failure; |
3011 | ||
daeda1cc | 3012 | rcu_read_lock(); |
f9eb7bf4 AG |
3013 | if (got_ldev) { |
3014 | struct disk_conf *disk_conf; | |
44ed167d | 3015 | |
b30ab791 | 3016 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
3017 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
3018 | } | |
3019 | if (!err) { | |
3020 | struct net_conf *nc; | |
3021 | ||
a6b32bc3 | 3022 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
3023 | if (nc) |
3024 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
3025 | } | |
44ed167d PR |
3026 | rcu_read_unlock(); |
3027 | if (err) | |
3028 | goto nla_put_failure; | |
3b98c0c2 | 3029 | |
3b98c0c2 LE |
3030 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
3031 | if (!nla) | |
3032 | goto nla_put_failure; | |
26ec9287 | 3033 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
3034 | nla_put_u32(skb, T_current_state, device->state.i) || |
3035 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
3036 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
3037 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
3038 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
3039 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
3040 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
3041 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
3042 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
3043 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
3044 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
3045 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 3046 | goto nla_put_failure; |
3b98c0c2 LE |
3047 | |
3048 | if (got_ldev) { | |
39a1aa7f | 3049 | int err; |
b411b363 | 3050 | |
b30ab791 AG |
3051 | spin_lock_irq(&device->ldev->md.uuid_lock); |
3052 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
3053 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
3054 | |
3055 | if (err) | |
3056 | goto nla_put_failure; | |
3057 | ||
b30ab791 AG |
3058 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
3059 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
3060 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 3061 | goto nla_put_failure; |
b30ab791 AG |
3062 | if (C_SYNC_SOURCE <= device->state.conn && |
3063 | C_PAUSED_SYNC_T >= device->state.conn) { | |
3064 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
3065 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 3066 | goto nla_put_failure; |
3b98c0c2 | 3067 | } |
b411b363 | 3068 | } |
b411b363 | 3069 | |
3b98c0c2 LE |
3070 | if (sib) { |
3071 | switch(sib->sib_reason) { | |
3072 | case SIB_SYNC_PROGRESS: | |
3073 | case SIB_GET_STATUS_REPLY: | |
3074 | break; | |
3075 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
3076 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
3077 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
3078 | goto nla_put_failure; | |
3b98c0c2 LE |
3079 | break; |
3080 | case SIB_HELPER_POST: | |
26ec9287 AG |
3081 | if (nla_put_u32(skb, T_helper_exit_code, |
3082 | sib->helper_exit_code)) | |
3083 | goto nla_put_failure; | |
3b98c0c2 LE |
3084 | /* fall through */ |
3085 | case SIB_HELPER_PRE: | |
26ec9287 AG |
3086 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
3087 | goto nla_put_failure; | |
3b98c0c2 LE |
3088 | break; |
3089 | } | |
b411b363 | 3090 | } |
3b98c0c2 | 3091 | nla_nest_end(skb, nla); |
b411b363 | 3092 | |
3b98c0c2 LE |
3093 | if (0) |
3094 | nla_put_failure: | |
3095 | err = -EMSGSIZE; | |
3096 | if (got_ldev) | |
b30ab791 | 3097 | put_ldev(device); |
3b98c0c2 | 3098 | return err; |
b411b363 PR |
3099 | } |
3100 | ||
3b98c0c2 | 3101 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3102 | { |
a910b123 | 3103 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3104 | enum drbd_ret_code retcode; |
3105 | int err; | |
b411b363 | 3106 | |
a910b123 | 3107 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3108 | if (!adm_ctx.reply_skb) |
3109 | return retcode; | |
3110 | if (retcode != NO_ERROR) | |
3111 | goto out; | |
b411b363 | 3112 | |
b30ab791 | 3113 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3114 | if (err) { |
3115 | nlmsg_free(adm_ctx.reply_skb); | |
3116 | return err; | |
b411b363 | 3117 | } |
3b98c0c2 | 3118 | out: |
a910b123 | 3119 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3120 | return 0; |
b411b363 PR |
3121 | } |
3122 | ||
4b7a530f | 3123 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3124 | { |
b30ab791 | 3125 | struct drbd_device *device; |
3b98c0c2 | 3126 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3127 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3128 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3129 | struct drbd_resource *tmp; |
543cc10b LE |
3130 | unsigned volume = cb->args[1]; |
3131 | ||
3132 | /* Open coded, deferred, iteration: | |
77c556f6 | 3133 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3134 | * connection = "first connection of resource or undefined"; |
3135 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3136 | * ... |
3137 | * } | |
3138 | * } | |
77c556f6 | 3139 | * where resource is cb->args[0]; |
543cc10b LE |
3140 | * and i is cb->args[1]; |
3141 | * | |
71932efc LE |
3142 | * cb->args[2] indicates if we shall loop over all resources, |
3143 | * or just dump all volumes of a single resource. | |
3144 | * | |
3b98c0c2 LE |
3145 | * This may miss entries inserted after this dump started, |
3146 | * or entries deleted before they are reached. | |
543cc10b | 3147 | * |
b30ab791 | 3148 | * We need to make sure the device won't disappear while |
543cc10b LE |
3149 | * we are looking at it, and revalidate our iterators |
3150 | * on each iteration. | |
3151 | */ | |
b411b363 | 3152 | |
05a10ec7 | 3153 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3154 | rcu_read_lock(); |
543cc10b | 3155 | /* revalidate iterator position */ |
77c556f6 | 3156 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3157 | if (pos == NULL) { |
3158 | /* first iteration */ | |
3159 | pos = tmp; | |
77c556f6 | 3160 | resource = pos; |
543cc10b LE |
3161 | break; |
3162 | } | |
3163 | if (tmp == pos) { | |
77c556f6 | 3164 | resource = pos; |
543cc10b LE |
3165 | break; |
3166 | } | |
b411b363 | 3167 | } |
77c556f6 AG |
3168 | if (resource) { |
3169 | next_resource: | |
251b8f8e AG |
3170 | device = idr_get_next(&resource->devices, &volume); |
3171 | if (!device) { | |
77c556f6 AG |
3172 | /* No more volumes to dump on this resource. |
3173 | * Advance resource iterator. */ | |
3174 | pos = list_entry_rcu(resource->resources.next, | |
3175 | struct drbd_resource, resources); | |
3176 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3177 | if (volume != 0) { |
71932efc LE |
3178 | /* If we reached the end of the list, |
3179 | * or only a single resource dump was requested, | |
3180 | * we are done. */ | |
77c556f6 | 3181 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3182 | goto out; |
543cc10b | 3183 | volume = 0; |
77c556f6 AG |
3184 | resource = pos; |
3185 | goto next_resource; | |
543cc10b LE |
3186 | } |
3187 | } | |
3188 | ||
98683650 | 3189 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3190 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3191 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3192 | if (!dh) | |
543cc10b LE |
3193 | goto out; |
3194 | ||
251b8f8e | 3195 | if (!device) { |
bde89a9e | 3196 | /* This is a connection without a single volume. |
367d675d LE |
3197 | * Suprisingly enough, it may have a network |
3198 | * configuration. */ | |
251b8f8e AG |
3199 | struct drbd_connection *connection; |
3200 | ||
543cc10b LE |
3201 | dh->minor = -1U; |
3202 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3203 | connection = the_only_connection(resource); |
3204 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3205 | goto cancel; |
251b8f8e AG |
3206 | if (connection) { |
3207 | struct net_conf *nc; | |
3208 | ||
3209 | nc = rcu_dereference(connection->net_conf); | |
3210 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3211 | goto cancel; | |
3212 | } | |
367d675d | 3213 | goto done; |
543cc10b | 3214 | } |
b411b363 | 3215 | |
0b0ba1ef | 3216 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3217 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3218 | |
b30ab791 | 3219 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3220 | dh->ret_code = NO_ERROR; |
3221 | ||
b30ab791 | 3222 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3223 | cancel: |
3b98c0c2 | 3224 | genlmsg_cancel(skb, dh); |
543cc10b | 3225 | goto out; |
3b98c0c2 | 3226 | } |
367d675d | 3227 | done: |
3b98c0c2 | 3228 | genlmsg_end(skb, dh); |
bde89a9e | 3229 | } |
b411b363 | 3230 | |
543cc10b | 3231 | out: |
c141ebda | 3232 | rcu_read_unlock(); |
543cc10b | 3233 | /* where to start the next iteration */ |
bde89a9e | 3234 | cb->args[0] = (long)pos; |
77c556f6 | 3235 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3236 | |
77c556f6 | 3237 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3238 | * Which will terminate the dump. */ |
3b98c0c2 | 3239 | return skb->len; |
b411b363 PR |
3240 | } |
3241 | ||
71932efc LE |
3242 | /* |
3243 | * Request status of all resources, or of all volumes within a single resource. | |
3244 | * | |
3245 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3246 | * Which means we cannot use the family->attrbuf or other such members, because | |
3247 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3248 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3249 | * | |
3250 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3251 | */ |
71932efc | 3252 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3253 | { |
71932efc LE |
3254 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3255 | struct nlattr *nla; | |
7c3063cc | 3256 | const char *resource_name; |
4bc76048 | 3257 | struct drbd_resource *resource; |
7c3063cc | 3258 | int maxtype; |
71932efc LE |
3259 | |
3260 | /* Is this a followup call? */ | |
3261 | if (cb->args[0]) { | |
3262 | /* ... of a single resource dump, | |
3263 | * and the resource iterator has been advanced already? */ | |
3264 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3265 | return 0; /* DONE. */ | |
3266 | goto dump; | |
3267 | } | |
3268 | ||
3269 | /* First call (from netlink_dump_start). We need to figure out | |
3270 | * which resource(s) the user wants us to dump. */ | |
3271 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3272 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3273 | DRBD_NLA_CFG_CONTEXT); | |
3274 | ||
3275 | /* No explicit context given. Dump all. */ | |
3276 | if (!nla) | |
3277 | goto dump; | |
7c3063cc AG |
3278 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3279 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3280 | if (IS_ERR(nla)) | |
3281 | return PTR_ERR(nla); | |
71932efc LE |
3282 | /* context given, but no name present? */ |
3283 | if (!nla) | |
3284 | return -EINVAL; | |
7c3063cc | 3285 | resource_name = nla_data(nla); |
4bc76048 AG |
3286 | if (!*resource_name) |
3287 | return -ENODEV; | |
3288 | resource = drbd_find_resource(resource_name); | |
3289 | if (!resource) | |
71932efc LE |
3290 | return -ENODEV; |
3291 | ||
4bc76048 | 3292 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3293 | |
71932efc | 3294 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3295 | * only dump this connection. */ |
4bc76048 | 3296 | cb->args[0] = (long)resource; |
71932efc | 3297 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3298 | cb->args[2] = (long)resource; |
71932efc LE |
3299 | |
3300 | dump: | |
3301 | return get_one_status(skb, cb); | |
3302 | } | |
b411b363 | 3303 | |
3b98c0c2 | 3304 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3305 | { |
a910b123 | 3306 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3307 | enum drbd_ret_code retcode; |
3308 | struct timeout_parms tp; | |
3309 | int err; | |
b411b363 | 3310 | |
a910b123 | 3311 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3312 | if (!adm_ctx.reply_skb) |
3313 | return retcode; | |
3314 | if (retcode != NO_ERROR) | |
3315 | goto out; | |
b411b363 | 3316 | |
3b98c0c2 | 3317 | tp.timeout_type = |
b30ab791 AG |
3318 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3319 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3320 | UT_DEFAULT; |
b411b363 | 3321 | |
3b98c0c2 LE |
3322 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3323 | if (err) { | |
3324 | nlmsg_free(adm_ctx.reply_skb); | |
3325 | return err; | |
3326 | } | |
3327 | out: | |
a910b123 | 3328 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3329 | return 0; |
b411b363 PR |
3330 | } |
3331 | ||
3b98c0c2 | 3332 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3333 | { |
a910b123 | 3334 | struct drbd_config_context adm_ctx; |
b30ab791 | 3335 | struct drbd_device *device; |
3b98c0c2 | 3336 | enum drbd_ret_code retcode; |
58ffa580 | 3337 | struct start_ov_parms parms; |
b411b363 | 3338 | |
a910b123 | 3339 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3340 | if (!adm_ctx.reply_skb) |
3341 | return retcode; | |
3342 | if (retcode != NO_ERROR) | |
3343 | goto out; | |
873b0d5f | 3344 | |
b30ab791 | 3345 | device = adm_ctx.device; |
58ffa580 LE |
3346 | |
3347 | /* resume from last known position, if possible */ | |
b30ab791 | 3348 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3349 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3350 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3351 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3352 | if (err) { |
3353 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3354 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3355 | goto out; |
3356 | } | |
b411b363 | 3357 | } |
9e276872 LE |
3358 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3359 | ||
58ffa580 | 3360 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3361 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3362 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3363 | |
3364 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3365 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3366 | drbd_suspend_io(device); |
3367 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3368 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3369 | drbd_resume_io(device); | |
9e276872 LE |
3370 | |
3371 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 3372 | out: |
a910b123 | 3373 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3374 | return 0; |
3375 | } | |
3376 | ||
3377 | ||
3b98c0c2 | 3378 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3379 | { |
a910b123 | 3380 | struct drbd_config_context adm_ctx; |
b30ab791 | 3381 | struct drbd_device *device; |
3b98c0c2 | 3382 | enum drbd_ret_code retcode; |
b411b363 PR |
3383 | int skip_initial_sync = 0; |
3384 | int err; | |
3b98c0c2 | 3385 | struct new_c_uuid_parms args; |
b411b363 | 3386 | |
a910b123 | 3387 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3388 | if (!adm_ctx.reply_skb) |
3389 | return retcode; | |
3390 | if (retcode != NO_ERROR) | |
3391 | goto out_nolock; | |
b411b363 | 3392 | |
b30ab791 | 3393 | device = adm_ctx.device; |
3b98c0c2 LE |
3394 | memset(&args, 0, sizeof(args)); |
3395 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3396 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3397 | if (err) { |
3398 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3399 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3400 | goto out_nolock; |
3401 | } | |
b411b363 PR |
3402 | } |
3403 | ||
9e276872 | 3404 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3405 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3406 | |
b30ab791 | 3407 | if (!get_ldev(device)) { |
b411b363 PR |
3408 | retcode = ERR_NO_DISK; |
3409 | goto out; | |
3410 | } | |
3411 | ||
3412 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3413 | if (device->state.conn == C_CONNECTED && |
3414 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3415 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3416 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3417 | skip_initial_sync = 1; |
b30ab791 | 3418 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3419 | retcode = ERR_CONNECTED; |
3420 | goto out_dec; | |
3421 | } | |
3422 | ||
b30ab791 AG |
3423 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3424 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3425 | |
3426 | if (args.clear_bm) { | |
b30ab791 | 3427 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3428 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3429 | if (err) { |
d0180171 | 3430 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3431 | retcode = ERR_IO_MD_DISK; |
3432 | } | |
3433 | if (skip_initial_sync) { | |
69a22773 | 3434 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
3435 | _drbd_uuid_set(device, UI_BITMAP, 0); |
3436 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3437 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3438 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3439 | CS_VERBOSE, NULL); |
0500813f | 3440 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3441 | } |
3442 | } | |
3443 | ||
b30ab791 | 3444 | drbd_md_sync(device); |
b411b363 | 3445 | out_dec: |
b30ab791 | 3446 | put_ldev(device); |
b411b363 | 3447 | out: |
b30ab791 | 3448 | mutex_unlock(device->state_mutex); |
9e276872 | 3449 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3450 | out_nolock: |
a910b123 | 3451 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3452 | return 0; |
3453 | } | |
3454 | ||
3b98c0c2 | 3455 | static enum drbd_ret_code |
a910b123 | 3456 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 3457 | { |
a910b123 | 3458 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 3459 | if (!name || !name[0]) { |
a910b123 | 3460 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 3461 | return ERR_MANDATORY_TAG; |
b411b363 | 3462 | } |
3b98c0c2 LE |
3463 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3464 | * we must not allow slashes */ | |
3465 | if (strchr(name, '/')) { | |
a910b123 | 3466 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 3467 | return ERR_INVALID_REQUEST; |
b411b363 | 3468 | } |
3b98c0c2 | 3469 | return NO_ERROR; |
774b3055 | 3470 | } |
b411b363 | 3471 | |
a2972846 AG |
3472 | static void resource_to_info(struct resource_info *info, |
3473 | struct drbd_resource *resource) | |
3474 | { | |
3475 | info->res_role = conn_highest_role(first_connection(resource)); | |
3476 | info->res_susp = resource->susp; | |
3477 | info->res_susp_nod = resource->susp_nod; | |
3478 | info->res_susp_fen = resource->susp_fen; | |
3479 | } | |
3480 | ||
789c1b62 | 3481 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3482 | { |
a2972846 | 3483 | struct drbd_connection *connection; |
a910b123 | 3484 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3485 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3486 | struct res_opts res_opts; |
3487 | int err; | |
b411b363 | 3488 | |
a910b123 | 3489 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
3490 | if (!adm_ctx.reply_skb) |
3491 | return retcode; | |
3492 | if (retcode != NO_ERROR) | |
3493 | goto out; | |
b411b363 | 3494 | |
afbbfa88 AG |
3495 | set_res_opts_defaults(&res_opts); |
3496 | err = res_opts_from_attrs(&res_opts, info); | |
3497 | if (err && err != -ENOMSG) { | |
3498 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3499 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 3500 | goto out; |
b411b363 PR |
3501 | } |
3502 | ||
a910b123 | 3503 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
3504 | if (retcode != NO_ERROR) |
3505 | goto out; | |
b411b363 | 3506 | |
5c661042 | 3507 | if (adm_ctx.resource) { |
38f19616 LE |
3508 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3509 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 3510 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
3511 | } |
3512 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3513 | goto out; |
b411b363 | 3514 | } |
b411b363 | 3515 | |
9e276872 | 3516 | /* not yet safe for genl_family.parallel_ops */ |
28bc3b8c | 3517 | mutex_lock(&resources_mutex); |
a2972846 | 3518 | connection = conn_create(adm_ctx.resource_name, &res_opts); |
28bc3b8c | 3519 | mutex_unlock(&resources_mutex); |
a2972846 AG |
3520 | |
3521 | if (connection) { | |
3522 | struct resource_info resource_info; | |
3523 | ||
3524 | mutex_lock(¬ification_mutex); | |
3525 | resource_to_info(&resource_info, connection->resource); | |
3526 | notify_resource_state(NULL, 0, connection->resource, | |
3527 | &resource_info, NOTIFY_CREATE); | |
3528 | mutex_unlock(¬ification_mutex); | |
3529 | } else | |
3530 | retcode = ERR_NOMEM; | |
3531 | ||
3b98c0c2 | 3532 | out: |
a910b123 | 3533 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3534 | return 0; |
b411b363 PR |
3535 | } |
3536 | ||
a2972846 AG |
3537 | static void device_to_info(struct device_info *info, |
3538 | struct drbd_device *device) | |
3539 | { | |
3540 | info->dev_disk_state = device->state.disk; | |
3541 | } | |
3542 | ||
3543 | ||
05a10ec7 | 3544 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3545 | { |
a910b123 | 3546 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3547 | struct drbd_genlmsghdr *dh = info->userhdr; |
3548 | enum drbd_ret_code retcode; | |
b411b363 | 3549 | |
a910b123 | 3550 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3551 | if (!adm_ctx.reply_skb) |
3552 | return retcode; | |
3553 | if (retcode != NO_ERROR) | |
3554 | goto out; | |
b411b363 | 3555 | |
f2257a56 | 3556 | if (dh->minor > MINORMASK) { |
a910b123 | 3557 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
3558 | retcode = ERR_INVALID_REQUEST; |
3559 | goto out; | |
b411b363 | 3560 | } |
0c8e36d9 | 3561 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 3562 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
3563 | retcode = ERR_INVALID_REQUEST; |
3564 | goto out; | |
b411b363 | 3565 | } |
b411b363 | 3566 | |
38f19616 | 3567 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3568 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3569 | if (adm_ctx.device) { |
38f19616 | 3570 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
179e20b8 | 3571 | retcode = ERR_MINOR_OR_VOLUME_EXISTS; |
38f19616 LE |
3572 | /* else: still NO_ERROR */ |
3573 | goto out; | |
b411b363 | 3574 | } |
38f19616 | 3575 | |
9e276872 | 3576 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 3577 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
a2972846 AG |
3578 | if (retcode == NO_ERROR) { |
3579 | struct drbd_device *device; | |
3580 | struct drbd_peer_device *peer_device; | |
3581 | struct device_info info; | |
3582 | unsigned int peer_devices = 0; | |
3583 | enum drbd_notification_type flags; | |
3584 | ||
3585 | device = minor_to_device(dh->minor); | |
3586 | for_each_peer_device(peer_device, device) { | |
3587 | if (!has_net_conf(peer_device->connection)) | |
3588 | continue; | |
3589 | peer_devices++; | |
3590 | } | |
3591 | ||
3592 | device_to_info(&info, device); | |
3593 | mutex_lock(¬ification_mutex); | |
3594 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
3595 | notify_device_state(NULL, 0, device, &info, NOTIFY_CREATE | flags); | |
3596 | for_each_peer_device(peer_device, device) { | |
3597 | struct peer_device_info peer_device_info; | |
3598 | ||
3599 | if (!has_net_conf(peer_device->connection)) | |
3600 | continue; | |
3601 | peer_device_to_info(&peer_device_info, peer_device); | |
3602 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
3603 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, | |
3604 | NOTIFY_CREATE | flags); | |
3605 | } | |
3606 | mutex_unlock(¬ification_mutex); | |
3607 | } | |
9e276872 | 3608 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3609 | out: |
a910b123 | 3610 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3611 | return 0; |
b411b363 PR |
3612 | } |
3613 | ||
05a10ec7 | 3614 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3615 | { |
a2972846 AG |
3616 | struct drbd_peer_device *peer_device; |
3617 | ||
b30ab791 AG |
3618 | if (device->state.disk == D_DISKLESS && |
3619 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3620 | * we may want to delete a minor from a live replication group. |
3621 | */ | |
b30ab791 | 3622 | device->state.role == R_SECONDARY) { |
a2972846 AG |
3623 | struct drbd_connection *connection = |
3624 | first_connection(device->resource); | |
3625 | ||
b30ab791 | 3626 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), |
369bea63 | 3627 | CS_VERBOSE + CS_WAIT_COMPLETE); |
a2972846 AG |
3628 | |
3629 | /* If the state engine hasn't stopped the sender thread yet, we | |
3630 | * need to flush the sender work queue before generating the | |
3631 | * DESTROY events here. */ | |
3632 | if (get_t_state(&connection->worker) == RUNNING) | |
3633 | drbd_flush_workqueue(&connection->sender_work); | |
3634 | ||
3635 | mutex_lock(¬ification_mutex); | |
3636 | for_each_peer_device(peer_device, device) { | |
3637 | if (!has_net_conf(peer_device->connection)) | |
3638 | continue; | |
3639 | notify_peer_device_state(NULL, 0, peer_device, NULL, | |
3640 | NOTIFY_DESTROY | NOTIFY_CONTINUES); | |
3641 | } | |
3642 | notify_device_state(NULL, 0, device, NULL, NOTIFY_DESTROY); | |
3643 | mutex_unlock(¬ification_mutex); | |
3644 | ||
f82795d6 | 3645 | drbd_delete_device(device); |
85f75dd7 LE |
3646 | return NO_ERROR; |
3647 | } else | |
3648 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3649 | } |
3650 | ||
05a10ec7 | 3651 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3652 | { |
a910b123 | 3653 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3654 | enum drbd_ret_code retcode; |
b411b363 | 3655 | |
a910b123 | 3656 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3657 | if (!adm_ctx.reply_skb) |
3658 | return retcode; | |
3659 | if (retcode != NO_ERROR) | |
3660 | goto out; | |
b411b363 | 3661 | |
9e276872 | 3662 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 3663 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 3664 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 3665 | out: |
a910b123 | 3666 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 3667 | return 0; |
b411b363 PR |
3668 | } |
3669 | ||
179e20b8 AG |
3670 | static int adm_del_resource(struct drbd_resource *resource) |
3671 | { | |
3672 | struct drbd_connection *connection; | |
3673 | ||
3674 | for_each_connection(connection, resource) { | |
3675 | if (connection->cstate > C_STANDALONE) | |
3676 | return ERR_NET_CONFIGURED; | |
3677 | } | |
3678 | if (!idr_is_empty(&resource->devices)) | |
3679 | return ERR_RES_IN_USE; | |
3680 | ||
a2972846 AG |
3681 | /* The state engine has stopped the sender thread, so we don't |
3682 | * need to flush the sender work queue before generating the | |
3683 | * DESTROY event here. */ | |
3684 | mutex_lock(¬ification_mutex); | |
3685 | notify_resource_state(NULL, 0, resource, NULL, NOTIFY_DESTROY); | |
3686 | mutex_unlock(¬ification_mutex); | |
3687 | ||
28bc3b8c | 3688 | mutex_lock(&resources_mutex); |
179e20b8 | 3689 | list_del_rcu(&resource->resources); |
28bc3b8c | 3690 | mutex_unlock(&resources_mutex); |
179e20b8 AG |
3691 | /* Make sure all threads have actually stopped: state handling only |
3692 | * does drbd_thread_stop_nowait(). */ | |
3693 | list_for_each_entry(connection, &resource->connections, connections) | |
3694 | drbd_thread_stop(&connection->worker); | |
3695 | synchronize_rcu(); | |
3696 | drbd_free_resource(resource); | |
3697 | return NO_ERROR; | |
3698 | } | |
3699 | ||
85f75dd7 | 3700 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3701 | { |
a910b123 | 3702 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
3703 | struct drbd_resource *resource; |
3704 | struct drbd_connection *connection; | |
3705 | struct drbd_device *device; | |
f3dfa40a | 3706 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 3707 | unsigned i; |
b411b363 | 3708 | |
a910b123 | 3709 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3710 | if (!adm_ctx.reply_skb) |
3711 | return retcode; | |
3712 | if (retcode != NO_ERROR) | |
9e276872 | 3713 | goto finish; |
b411b363 | 3714 | |
b6f85ef9 | 3715 | resource = adm_ctx.resource; |
9e276872 | 3716 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 3717 | /* demote */ |
b6f85ef9 AG |
3718 | for_each_connection(connection, resource) { |
3719 | struct drbd_peer_device *peer_device; | |
3720 | ||
3721 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
3722 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
3723 | if (retcode < SS_SUCCESS) { | |
a910b123 | 3724 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
3725 | goto out; |
3726 | } | |
3727 | } | |
3728 | ||
3729 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 3730 | if (retcode < SS_SUCCESS) { |
a910b123 | 3731 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 3732 | goto out; |
85f75dd7 | 3733 | } |
b411b363 | 3734 | } |
b411b363 | 3735 | |
85f75dd7 | 3736 | /* detach */ |
b6f85ef9 AG |
3737 | idr_for_each_entry(&resource->devices, device, i) { |
3738 | retcode = adm_detach(device, 0); | |
27012382 | 3739 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 3740 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 3741 | goto out; |
85f75dd7 LE |
3742 | } |
3743 | } | |
b411b363 | 3744 | |
85f75dd7 | 3745 | /* delete volumes */ |
b6f85ef9 AG |
3746 | idr_for_each_entry(&resource->devices, device, i) { |
3747 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
3748 | if (retcode != NO_ERROR) { |
3749 | /* "can not happen" */ | |
a910b123 | 3750 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 3751 | goto out; |
85f75dd7 LE |
3752 | } |
3753 | } | |
b411b363 | 3754 | |
179e20b8 | 3755 | retcode = adm_del_resource(resource); |
3b98c0c2 | 3756 | out: |
9e276872 LE |
3757 | mutex_unlock(&resource->adm_mutex); |
3758 | finish: | |
a910b123 | 3759 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3760 | return 0; |
b411b363 PR |
3761 | } |
3762 | ||
789c1b62 | 3763 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3764 | { |
a910b123 | 3765 | struct drbd_config_context adm_ctx; |
77c556f6 | 3766 | struct drbd_resource *resource; |
3b98c0c2 | 3767 | enum drbd_ret_code retcode; |
b411b363 | 3768 | |
a910b123 | 3769 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3770 | if (!adm_ctx.reply_skb) |
3771 | return retcode; | |
3772 | if (retcode != NO_ERROR) | |
9e276872 | 3773 | goto finish; |
77c556f6 | 3774 | resource = adm_ctx.resource; |
b411b363 | 3775 | |
179e20b8 AG |
3776 | mutex_lock(&resource->adm_mutex); |
3777 | retcode = adm_del_resource(resource); | |
9e276872 LE |
3778 | mutex_unlock(&resource->adm_mutex); |
3779 | finish: | |
a910b123 | 3780 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3781 | return 0; |
3782 | } | |
3783 | ||
b30ab791 | 3784 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3785 | { |
3b98c0c2 LE |
3786 | struct sk_buff *msg; |
3787 | struct drbd_genlmsghdr *d_out; | |
3788 | unsigned seq; | |
3789 | int err = -ENOMEM; | |
3790 | ||
3791 | seq = atomic_inc_return(&drbd_genl_seq); | |
3792 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3793 | if (!msg) | |
3794 | goto failed; | |
3795 | ||
3796 | err = -EMSGSIZE; | |
3797 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3798 | if (!d_out) /* cannot happen, but anyways. */ | |
3799 | goto nla_put_failure; | |
b30ab791 | 3800 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3801 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3802 | |
b30ab791 | 3803 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3804 | goto nla_put_failure; |
3805 | genlmsg_end(msg, d_out); | |
3806 | err = drbd_genl_multicast_events(msg, 0); | |
3807 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3808 | if (err && err != -ESRCH) | |
3809 | goto failed; | |
b411b363 | 3810 | |
3b98c0c2 | 3811 | return; |
b411b363 | 3812 | |
3b98c0c2 LE |
3813 | nla_put_failure: |
3814 | nlmsg_free(msg); | |
3815 | failed: | |
d0180171 | 3816 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
3817 | "Event seq:%u sib_reason:%u\n", |
3818 | err, seq, sib->sib_reason); | |
b411b363 | 3819 | } |
a2972846 AG |
3820 | |
3821 | static void device_to_statistics(struct device_statistics *s, | |
3822 | struct drbd_device *device) | |
3823 | { | |
3824 | memset(s, 0, sizeof(*s)); | |
3825 | s->dev_upper_blocked = !may_inc_ap_bio(device); | |
3826 | if (get_ldev(device)) { | |
3827 | struct drbd_md *md = &device->ldev->md; | |
3828 | u64 *history_uuids = (u64 *)s->history_uuids; | |
3829 | struct request_queue *q; | |
3830 | int n; | |
3831 | ||
3832 | spin_lock_irq(&md->uuid_lock); | |
3833 | s->dev_current_uuid = md->uuid[UI_CURRENT]; | |
3834 | BUILD_BUG_ON(sizeof(s->history_uuids) < UI_HISTORY_END - UI_HISTORY_START + 1); | |
3835 | for (n = 0; n < UI_HISTORY_END - UI_HISTORY_START + 1; n++) | |
3836 | history_uuids[n] = md->uuid[UI_HISTORY_START + n]; | |
3837 | for (; n < HISTORY_UUIDS; n++) | |
3838 | history_uuids[n] = 0; | |
3839 | s->history_uuids_len = HISTORY_UUIDS; | |
3840 | spin_unlock_irq(&md->uuid_lock); | |
3841 | ||
3842 | s->dev_disk_flags = md->flags; | |
3843 | q = bdev_get_queue(device->ldev->backing_bdev); | |
3844 | s->dev_lower_blocked = | |
3845 | bdi_congested(&q->backing_dev_info, | |
3846 | (1 << WB_async_congested) | | |
3847 | (1 << WB_sync_congested)); | |
3848 | put_ldev(device); | |
3849 | } | |
3850 | s->dev_size = drbd_get_capacity(device->this_bdev); | |
3851 | s->dev_read = device->read_cnt; | |
3852 | s->dev_write = device->writ_cnt; | |
3853 | s->dev_al_writes = device->al_writ_cnt; | |
3854 | s->dev_bm_writes = device->bm_writ_cnt; | |
3855 | s->dev_upper_pending = atomic_read(&device->ap_bio_cnt); | |
3856 | s->dev_lower_pending = atomic_read(&device->local_cnt); | |
3857 | s->dev_al_suspended = test_bit(AL_SUSPENDED, &device->flags); | |
3858 | s->dev_exposed_data_uuid = device->ed_uuid; | |
3859 | } | |
3860 | ||
3861 | enum mdf_peer_flag { | |
3862 | MDF_PEER_CONNECTED = 1 << 0, | |
3863 | MDF_PEER_OUTDATED = 1 << 1, | |
3864 | MDF_PEER_FENCING = 1 << 2, | |
3865 | MDF_PEER_FULL_SYNC = 1 << 3, | |
3866 | }; | |
3867 | ||
3868 | static void peer_device_to_statistics(struct peer_device_statistics *s, | |
3869 | struct drbd_peer_device *peer_device) | |
3870 | { | |
3871 | struct drbd_device *device = peer_device->device; | |
3872 | ||
3873 | memset(s, 0, sizeof(*s)); | |
3874 | s->peer_dev_received = device->recv_cnt; | |
3875 | s->peer_dev_sent = device->send_cnt; | |
3876 | s->peer_dev_pending = atomic_read(&device->ap_pending_cnt) + | |
3877 | atomic_read(&device->rs_pending_cnt); | |
3878 | s->peer_dev_unacked = atomic_read(&device->unacked_cnt); | |
3879 | s->peer_dev_out_of_sync = drbd_bm_total_weight(device) << (BM_BLOCK_SHIFT - 9); | |
3880 | s->peer_dev_resync_failed = device->rs_failed << (BM_BLOCK_SHIFT - 9); | |
3881 | if (get_ldev(device)) { | |
3882 | struct drbd_md *md = &device->ldev->md; | |
3883 | ||
3884 | spin_lock_irq(&md->uuid_lock); | |
3885 | s->peer_dev_bitmap_uuid = md->uuid[UI_BITMAP]; | |
3886 | spin_unlock_irq(&md->uuid_lock); | |
3887 | s->peer_dev_flags = | |
3888 | (drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND) ? | |
3889 | MDF_PEER_CONNECTED : 0) + | |
3890 | (drbd_md_test_flag(device->ldev, MDF_CONSISTENT) && | |
3891 | !drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE) ? | |
3892 | MDF_PEER_OUTDATED : 0) + | |
3893 | /* FIXME: MDF_PEER_FENCING? */ | |
3894 | (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) ? | |
3895 | MDF_PEER_FULL_SYNC : 0); | |
3896 | put_ldev(device); | |
3897 | } | |
3898 | } | |
3899 | ||
3900 | static int nla_put_notification_header(struct sk_buff *msg, | |
3901 | enum drbd_notification_type type) | |
3902 | { | |
3903 | struct drbd_notification_header nh = { | |
3904 | .nh_type = type, | |
3905 | }; | |
3906 | ||
3907 | return drbd_notification_header_to_skb(msg, &nh, true); | |
3908 | } | |
3909 | ||
3910 | void notify_resource_state(struct sk_buff *skb, | |
3911 | unsigned int seq, | |
3912 | struct drbd_resource *resource, | |
3913 | struct resource_info *resource_info, | |
3914 | enum drbd_notification_type type) | |
3915 | { | |
3916 | struct resource_statistics resource_statistics; | |
3917 | struct drbd_genlmsghdr *dh; | |
3918 | bool multicast = false; | |
3919 | int err; | |
3920 | ||
3921 | if (!skb) { | |
3922 | seq = atomic_inc_return(¬ify_genl_seq); | |
3923 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3924 | err = -ENOMEM; | |
3925 | if (!skb) | |
3926 | goto failed; | |
3927 | multicast = true; | |
3928 | } | |
3929 | ||
3930 | err = -EMSGSIZE; | |
3931 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_RESOURCE_STATE); | |
3932 | if (!dh) | |
3933 | goto nla_put_failure; | |
3934 | dh->minor = -1U; | |
3935 | dh->ret_code = NO_ERROR; | |
3936 | if (nla_put_drbd_cfg_context(skb, resource, NULL, NULL) || | |
3937 | nla_put_notification_header(skb, type) || | |
3938 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
3939 | resource_info_to_skb(skb, resource_info, true))) | |
3940 | goto nla_put_failure; | |
3941 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
3942 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
3943 | if (err) | |
3944 | goto nla_put_failure; | |
3945 | genlmsg_end(skb, dh); | |
3946 | if (multicast) { | |
3947 | err = drbd_genl_multicast_events(skb, 0); | |
3948 | /* skb has been consumed or freed in netlink_broadcast() */ | |
3949 | if (err && err != -ESRCH) | |
3950 | goto failed; | |
3951 | } | |
3952 | return; | |
3953 | ||
3954 | nla_put_failure: | |
3955 | nlmsg_free(skb); | |
3956 | failed: | |
3957 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
3958 | err, seq); | |
3959 | } | |
3960 | ||
3961 | void notify_device_state(struct sk_buff *skb, | |
3962 | unsigned int seq, | |
3963 | struct drbd_device *device, | |
3964 | struct device_info *device_info, | |
3965 | enum drbd_notification_type type) | |
3966 | { | |
3967 | struct device_statistics device_statistics; | |
3968 | struct drbd_genlmsghdr *dh; | |
3969 | bool multicast = false; | |
3970 | int err; | |
3971 | ||
3972 | if (!skb) { | |
3973 | seq = atomic_inc_return(¬ify_genl_seq); | |
3974 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3975 | err = -ENOMEM; | |
3976 | if (!skb) | |
3977 | goto failed; | |
3978 | multicast = true; | |
3979 | } | |
3980 | ||
3981 | err = -EMSGSIZE; | |
3982 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_DEVICE_STATE); | |
3983 | if (!dh) | |
3984 | goto nla_put_failure; | |
3985 | dh->minor = device->minor; | |
3986 | dh->ret_code = NO_ERROR; | |
3987 | if (nla_put_drbd_cfg_context(skb, device->resource, NULL, device) || | |
3988 | nla_put_notification_header(skb, type) || | |
3989 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
3990 | device_info_to_skb(skb, device_info, true))) | |
3991 | goto nla_put_failure; | |
3992 | device_to_statistics(&device_statistics, device); | |
3993 | device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
3994 | genlmsg_end(skb, dh); | |
3995 | if (multicast) { | |
3996 | err = drbd_genl_multicast_events(skb, 0); | |
3997 | /* skb has been consumed or freed in netlink_broadcast() */ | |
3998 | if (err && err != -ESRCH) | |
3999 | goto failed; | |
4000 | } | |
4001 | return; | |
4002 | ||
4003 | nla_put_failure: | |
4004 | nlmsg_free(skb); | |
4005 | failed: | |
4006 | drbd_err(device, "Error %d while broadcasting event. Event seq:%u\n", | |
4007 | err, seq); | |
4008 | } | |
4009 | ||
4010 | void notify_connection_state(struct sk_buff *skb, | |
4011 | unsigned int seq, | |
4012 | struct drbd_connection *connection, | |
4013 | struct connection_info *connection_info, | |
4014 | enum drbd_notification_type type) | |
4015 | { | |
4016 | struct connection_statistics connection_statistics; | |
4017 | struct drbd_genlmsghdr *dh; | |
4018 | bool multicast = false; | |
4019 | int err; | |
4020 | ||
4021 | if (!skb) { | |
4022 | seq = atomic_inc_return(¬ify_genl_seq); | |
4023 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4024 | err = -ENOMEM; | |
4025 | if (!skb) | |
4026 | goto failed; | |
4027 | multicast = true; | |
4028 | } | |
4029 | ||
4030 | err = -EMSGSIZE; | |
4031 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_CONNECTION_STATE); | |
4032 | if (!dh) | |
4033 | goto nla_put_failure; | |
4034 | dh->minor = -1U; | |
4035 | dh->ret_code = NO_ERROR; | |
4036 | if (nla_put_drbd_cfg_context(skb, connection->resource, connection, NULL) || | |
4037 | nla_put_notification_header(skb, type) || | |
4038 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4039 | connection_info_to_skb(skb, connection_info, true))) | |
4040 | goto nla_put_failure; | |
4041 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
4042 | connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
4043 | genlmsg_end(skb, dh); | |
4044 | if (multicast) { | |
4045 | err = drbd_genl_multicast_events(skb, 0); | |
4046 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4047 | if (err && err != -ESRCH) | |
4048 | goto failed; | |
4049 | } | |
4050 | return; | |
4051 | ||
4052 | nla_put_failure: | |
4053 | nlmsg_free(skb); | |
4054 | failed: | |
4055 | drbd_err(connection, "Error %d while broadcasting event. Event seq:%u\n", | |
4056 | err, seq); | |
4057 | } | |
4058 | ||
4059 | void notify_peer_device_state(struct sk_buff *skb, | |
4060 | unsigned int seq, | |
4061 | struct drbd_peer_device *peer_device, | |
4062 | struct peer_device_info *peer_device_info, | |
4063 | enum drbd_notification_type type) | |
4064 | { | |
4065 | struct peer_device_statistics peer_device_statistics; | |
4066 | struct drbd_resource *resource = peer_device->device->resource; | |
4067 | struct drbd_genlmsghdr *dh; | |
4068 | bool multicast = false; | |
4069 | int err; | |
4070 | ||
4071 | if (!skb) { | |
4072 | seq = atomic_inc_return(¬ify_genl_seq); | |
4073 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4074 | err = -ENOMEM; | |
4075 | if (!skb) | |
4076 | goto failed; | |
4077 | multicast = true; | |
4078 | } | |
4079 | ||
4080 | err = -EMSGSIZE; | |
4081 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_PEER_DEVICE_STATE); | |
4082 | if (!dh) | |
4083 | goto nla_put_failure; | |
4084 | dh->minor = -1U; | |
4085 | dh->ret_code = NO_ERROR; | |
4086 | if (nla_put_drbd_cfg_context(skb, resource, peer_device->connection, peer_device->device) || | |
4087 | nla_put_notification_header(skb, type) || | |
4088 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4089 | peer_device_info_to_skb(skb, peer_device_info, true))) | |
4090 | goto nla_put_failure; | |
4091 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
4092 | peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
4093 | genlmsg_end(skb, dh); | |
4094 | if (multicast) { | |
4095 | err = drbd_genl_multicast_events(skb, 0); | |
4096 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4097 | if (err && err != -ESRCH) | |
4098 | goto failed; | |
4099 | } | |
4100 | return; | |
4101 | ||
4102 | nla_put_failure: | |
4103 | nlmsg_free(skb); | |
4104 | failed: | |
4105 | drbd_err(peer_device, "Error %d while broadcasting event. Event seq:%u\n", | |
4106 | err, seq); | |
4107 | } | |
4108 | ||
4109 | void notify_helper(enum drbd_notification_type type, | |
4110 | struct drbd_device *device, struct drbd_connection *connection, | |
4111 | const char *name, int status) | |
4112 | { | |
4113 | struct drbd_resource *resource = device ? device->resource : connection->resource; | |
4114 | struct drbd_helper_info helper_info; | |
4115 | unsigned int seq = atomic_inc_return(¬ify_genl_seq); | |
4116 | struct sk_buff *skb = NULL; | |
4117 | struct drbd_genlmsghdr *dh; | |
4118 | int err; | |
4119 | ||
4120 | strlcpy(helper_info.helper_name, name, sizeof(helper_info.helper_name)); | |
4121 | helper_info.helper_name_len = min(strlen(name), sizeof(helper_info.helper_name)); | |
4122 | helper_info.helper_status = status; | |
4123 | ||
4124 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4125 | err = -ENOMEM; | |
4126 | if (!skb) | |
4127 | goto fail; | |
4128 | ||
4129 | err = -EMSGSIZE; | |
4130 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_HELPER); | |
4131 | if (!dh) | |
4132 | goto fail; | |
4133 | dh->minor = device ? device->minor : -1; | |
4134 | dh->ret_code = NO_ERROR; | |
4135 | mutex_lock(¬ification_mutex); | |
4136 | if (nla_put_drbd_cfg_context(skb, resource, connection, device) || | |
4137 | nla_put_notification_header(skb, type) || | |
4138 | drbd_helper_info_to_skb(skb, &helper_info, true)) | |
4139 | goto unlock_fail; | |
4140 | genlmsg_end(skb, dh); | |
4141 | err = drbd_genl_multicast_events(skb, 0); | |
4142 | skb = NULL; | |
4143 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4144 | if (err && err != -ESRCH) | |
4145 | goto unlock_fail; | |
4146 | mutex_unlock(¬ification_mutex); | |
4147 | return; | |
4148 | ||
4149 | unlock_fail: | |
4150 | mutex_unlock(¬ification_mutex); | |
4151 | fail: | |
4152 | nlmsg_free(skb); | |
4153 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4154 | err, seq); | |
4155 | } | |
4156 | ||
4157 | static void notify_initial_state_done(struct sk_buff *skb, unsigned int seq) | |
4158 | { | |
4159 | struct drbd_genlmsghdr *dh; | |
4160 | int err; | |
4161 | ||
4162 | err = -EMSGSIZE; | |
4163 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_INITIAL_STATE_DONE); | |
4164 | if (!dh) | |
4165 | goto nla_put_failure; | |
4166 | dh->minor = -1U; | |
4167 | dh->ret_code = NO_ERROR; | |
4168 | if (nla_put_notification_header(skb, NOTIFY_EXISTS)) | |
4169 | goto nla_put_failure; | |
4170 | genlmsg_end(skb, dh); | |
4171 | return; | |
4172 | ||
4173 | nla_put_failure: | |
4174 | nlmsg_free(skb); | |
4175 | pr_err("Error %d sending event. Event seq:%u\n", err, seq); | |
4176 | } | |
4177 | ||
4178 | static void free_state_changes(struct list_head *list) | |
4179 | { | |
4180 | while (!list_empty(list)) { | |
4181 | struct drbd_state_change *state_change = | |
4182 | list_first_entry(list, struct drbd_state_change, list); | |
4183 | list_del(&state_change->list); | |
4184 | forget_state_change(state_change); | |
4185 | } | |
4186 | } | |
4187 | ||
4188 | static unsigned int notifications_for_state_change(struct drbd_state_change *state_change) | |
4189 | { | |
4190 | return 1 + | |
4191 | state_change->n_connections + | |
4192 | state_change->n_devices + | |
4193 | state_change->n_devices * state_change->n_connections; | |
4194 | } | |
4195 | ||
4196 | static int get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4197 | { | |
4198 | struct drbd_state_change *state_change = (struct drbd_state_change *)cb->args[0]; | |
4199 | unsigned int seq = cb->args[2]; | |
4200 | unsigned int n; | |
4201 | enum drbd_notification_type flags = 0; | |
4202 | ||
4203 | /* There is no need for taking notification_mutex here: it doesn't | |
4204 | matter if the initial state events mix with later state chage | |
4205 | events; we can always tell the events apart by the NOTIFY_EXISTS | |
4206 | flag. */ | |
4207 | ||
4208 | cb->args[5]--; | |
4209 | if (cb->args[5] == 1) { | |
4210 | notify_initial_state_done(skb, seq); | |
4211 | goto out; | |
4212 | } | |
4213 | n = cb->args[4]++; | |
4214 | if (cb->args[4] < cb->args[3]) | |
4215 | flags |= NOTIFY_CONTINUES; | |
4216 | if (n < 1) { | |
4217 | notify_resource_state_change(skb, seq, state_change->resource, | |
4218 | NOTIFY_EXISTS | flags); | |
4219 | goto next; | |
4220 | } | |
4221 | n--; | |
4222 | if (n < state_change->n_connections) { | |
4223 | notify_connection_state_change(skb, seq, &state_change->connections[n], | |
4224 | NOTIFY_EXISTS | flags); | |
4225 | goto next; | |
4226 | } | |
4227 | n -= state_change->n_connections; | |
4228 | if (n < state_change->n_devices) { | |
4229 | notify_device_state_change(skb, seq, &state_change->devices[n], | |
4230 | NOTIFY_EXISTS | flags); | |
4231 | goto next; | |
4232 | } | |
4233 | n -= state_change->n_devices; | |
4234 | if (n < state_change->n_devices * state_change->n_connections) { | |
4235 | notify_peer_device_state_change(skb, seq, &state_change->peer_devices[n], | |
4236 | NOTIFY_EXISTS | flags); | |
4237 | goto next; | |
4238 | } | |
4239 | ||
4240 | next: | |
4241 | if (cb->args[4] == cb->args[3]) { | |
4242 | struct drbd_state_change *next_state_change = | |
4243 | list_entry(state_change->list.next, | |
4244 | struct drbd_state_change, list); | |
4245 | cb->args[0] = (long)next_state_change; | |
4246 | cb->args[3] = notifications_for_state_change(next_state_change); | |
4247 | cb->args[4] = 0; | |
4248 | } | |
4249 | out: | |
4250 | return skb->len; | |
4251 | } | |
4252 | ||
4253 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4254 | { | |
4255 | struct drbd_resource *resource; | |
4256 | LIST_HEAD(head); | |
4257 | ||
4258 | if (cb->args[5] >= 1) { | |
4259 | if (cb->args[5] > 1) | |
4260 | return get_initial_state(skb, cb); | |
4261 | if (cb->args[0]) { | |
4262 | struct drbd_state_change *state_change = | |
4263 | (struct drbd_state_change *)cb->args[0]; | |
4264 | ||
4265 | /* connect list to head */ | |
4266 | list_add(&head, &state_change->list); | |
4267 | free_state_changes(&head); | |
4268 | } | |
4269 | return 0; | |
4270 | } | |
4271 | ||
4272 | cb->args[5] = 2; /* number of iterations */ | |
4273 | mutex_lock(&resources_mutex); | |
4274 | for_each_resource(resource, &drbd_resources) { | |
4275 | struct drbd_state_change *state_change; | |
4276 | ||
4277 | state_change = remember_old_state(resource, GFP_KERNEL); | |
4278 | if (!state_change) { | |
4279 | if (!list_empty(&head)) | |
4280 | free_state_changes(&head); | |
4281 | mutex_unlock(&resources_mutex); | |
4282 | return -ENOMEM; | |
4283 | } | |
4284 | copy_old_to_new_state_change(state_change); | |
4285 | list_add_tail(&state_change->list, &head); | |
4286 | cb->args[5] += notifications_for_state_change(state_change); | |
4287 | } | |
4288 | mutex_unlock(&resources_mutex); | |
4289 | ||
4290 | if (!list_empty(&head)) { | |
4291 | struct drbd_state_change *state_change = | |
4292 | list_entry(head.next, struct drbd_state_change, list); | |
4293 | cb->args[0] = (long)state_change; | |
4294 | cb->args[3] = notifications_for_state_change(state_change); | |
4295 | list_del(&head); /* detach list from head */ | |
4296 | } | |
4297 | ||
4298 | cb->args[2] = cb->nlh->nlmsg_seq; | |
4299 | return get_initial_state(skb, cb); | |
4300 | } |