]>
Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
b411b363 PR |
26 | #include <linux/module.h> |
27 | #include <linux/drbd.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/fs.h> | |
30 | #include <linux/file.h> | |
31 | #include <linux/slab.h> | |
b411b363 PR |
32 | #include <linux/blkpg.h> |
33 | #include <linux/cpumask.h> | |
34 | #include "drbd_int.h" | |
a3603a6e | 35 | #include "drbd_protocol.h" |
265be2d0 | 36 | #include "drbd_req.h" |
b411b363 PR |
37 | #include "drbd_wrappers.h" |
38 | #include <asm/unaligned.h> | |
b411b363 | 39 | #include <linux/drbd_limits.h> |
87f7be4c | 40 | #include <linux/kthread.h> |
b411b363 | 41 | |
3b98c0c2 LE |
42 | #include <net/genetlink.h> |
43 | ||
44 | /* .doit */ | |
45 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
47 | ||
05a10ec7 AG |
48 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
49 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 50 | |
789c1b62 AG |
51 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
52 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 53 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
54 | |
55 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
56 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 57 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
58 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
59 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 60 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
61 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
62 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 72 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
73 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
74 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
75 | /* .dumpit */ | |
76 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
77 | ||
78 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 79 | #include "drbd_nla.h" |
3b98c0c2 LE |
80 | #include <linux/genl_magic_func.h> |
81 | ||
82 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
83 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
84 | ||
3b98c0c2 LE |
85 | /* Configuration is strictly serialized, because generic netlink message |
86 | * processing is strictly serialized by the genl_lock(). | |
87 | * Which means we can use one static global drbd_config_context struct. | |
88 | */ | |
89 | static struct drbd_config_context { | |
90 | /* assigned from drbd_genlmsghdr */ | |
91 | unsigned int minor; | |
92 | /* assigned from request attributes, if present */ | |
93 | unsigned int volume; | |
94 | #define VOLUME_UNSPECIFIED (-1U) | |
95 | /* pointer into the request skb, | |
96 | * limited lifetime! */ | |
7c3063cc | 97 | char *resource_name; |
089c075d AG |
98 | struct nlattr *my_addr; |
99 | struct nlattr *peer_addr; | |
3b98c0c2 LE |
100 | |
101 | /* reply buffer */ | |
102 | struct sk_buff *reply_skb; | |
103 | /* pointer into reply buffer */ | |
104 | struct drbd_genlmsghdr *reply_dh; | |
105 | /* resolved from attributes, if possible */ | |
b30ab791 | 106 | struct drbd_device *device; |
bde89a9e | 107 | struct drbd_connection *connection; |
3b98c0c2 LE |
108 | } adm_ctx; |
109 | ||
110 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
111 | { | |
112 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
113 | if (genlmsg_reply(skb, info)) | |
114 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 115 | } |
3b98c0c2 LE |
116 | |
117 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
118 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 119 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
120 | { |
121 | struct sk_buff *skb = adm_ctx.reply_skb; | |
122 | struct nlattr *nla; | |
123 | int err = -EMSGSIZE; | |
124 | ||
125 | if (!info || !info[0]) | |
126 | return 0; | |
127 | ||
128 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
129 | if (!nla) | |
130 | return err; | |
131 | ||
132 | err = nla_put_string(skb, T_info_text, info); | |
133 | if (err) { | |
134 | nla_nest_cancel(skb, nla); | |
135 | return err; | |
136 | } else | |
137 | nla_nest_end(skb, nla); | |
138 | return 0; | |
b411b363 PR |
139 | } |
140 | ||
3b98c0c2 LE |
141 | /* This would be a good candidate for a "pre_doit" hook, |
142 | * and per-family private info->pointers. | |
143 | * But we need to stay compatible with older kernels. | |
144 | * If it returns successfully, adm_ctx members are valid. | |
145 | */ | |
146 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 147 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 148 | #define DRBD_ADM_NEED_CONNECTION 4 |
3b98c0c2 LE |
149 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, |
150 | unsigned flags) | |
151 | { | |
152 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
153 | const u8 cmd = info->genlhdr->cmd; | |
154 | int err; | |
155 | ||
156 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
157 | ||
158 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 159 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
160 | return -EPERM; |
161 | ||
162 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
1e2a2551 AG |
163 | if (!adm_ctx.reply_skb) { |
164 | err = -ENOMEM; | |
3b98c0c2 | 165 | goto fail; |
1e2a2551 | 166 | } |
3b98c0c2 LE |
167 | |
168 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
169 | info, &drbd_genl_family, 0, cmd); | |
170 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
171 | * but anyways */ | |
1e2a2551 AG |
172 | if (!adm_ctx.reply_dh) { |
173 | err = -ENOMEM; | |
3b98c0c2 | 174 | goto fail; |
1e2a2551 | 175 | } |
3b98c0c2 LE |
176 | |
177 | adm_ctx.reply_dh->minor = d_in->minor; | |
178 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
179 | ||
089c075d | 180 | adm_ctx.volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
181 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
182 | struct nlattr *nla; | |
183 | /* parse and validate only */ | |
f399002e | 184 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
185 | if (err) |
186 | goto fail; | |
187 | ||
188 | /* It was present, and valid, | |
189 | * copy it over to the reply skb. */ | |
190 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
191 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
192 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
193 | if (err) | |
194 | goto fail; | |
195 | ||
196 | /* and assign stuff to the global adm_ctx */ | |
197 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
089c075d AG |
198 | if (nla) |
199 | adm_ctx.volume = nla_get_u32(nla); | |
7c3063cc | 200 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 201 | if (nla) |
7c3063cc | 202 | adm_ctx.resource_name = nla_data(nla); |
089c075d AG |
203 | adm_ctx.my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; |
204 | adm_ctx.peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
205 | if ((adm_ctx.my_addr && | |
bde89a9e | 206 | nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.connection->my_addr)) || |
089c075d | 207 | (adm_ctx.peer_addr && |
bde89a9e | 208 | nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.connection->peer_addr))) { |
089c075d AG |
209 | err = -EINVAL; |
210 | goto fail; | |
211 | } | |
212 | } | |
3b98c0c2 LE |
213 | |
214 | adm_ctx.minor = d_in->minor; | |
b30ab791 | 215 | adm_ctx.device = minor_to_device(d_in->minor); |
bde89a9e | 216 | adm_ctx.connection = conn_get_by_name(adm_ctx.resource_name); |
3b98c0c2 | 217 | |
b30ab791 | 218 | if (!adm_ctx.device && (flags & DRBD_ADM_NEED_MINOR)) { |
3b98c0c2 LE |
219 | drbd_msg_put_info("unknown minor"); |
220 | return ERR_MINOR_INVALID; | |
221 | } | |
bde89a9e | 222 | if (!adm_ctx.connection && (flags & DRBD_ADM_NEED_RESOURCE)) { |
44e52cfa | 223 | drbd_msg_put_info("unknown resource"); |
a10f6b8a AG |
224 | if (adm_ctx.resource_name) |
225 | return ERR_RES_NOT_KNOWN; | |
3b98c0c2 LE |
226 | return ERR_INVALID_REQUEST; |
227 | } | |
228 | ||
089c075d | 229 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
bde89a9e | 230 | if (adm_ctx.connection && !(flags & DRBD_ADM_NEED_RESOURCE)) { |
089c075d AG |
231 | drbd_msg_put_info("no resource name expected"); |
232 | return ERR_INVALID_REQUEST; | |
233 | } | |
b30ab791 | 234 | if (adm_ctx.device) { |
089c075d AG |
235 | drbd_msg_put_info("no minor number expected"); |
236 | return ERR_INVALID_REQUEST; | |
237 | } | |
238 | if (adm_ctx.my_addr && adm_ctx.peer_addr) | |
bde89a9e | 239 | adm_ctx.connection = conn_get_by_addrs(nla_data(adm_ctx.my_addr), |
089c075d AG |
240 | nla_len(adm_ctx.my_addr), |
241 | nla_data(adm_ctx.peer_addr), | |
242 | nla_len(adm_ctx.peer_addr)); | |
bde89a9e | 243 | if (!adm_ctx.connection) { |
089c075d AG |
244 | drbd_msg_put_info("unknown connection"); |
245 | return ERR_INVALID_REQUEST; | |
246 | } | |
247 | } | |
248 | ||
3b98c0c2 | 249 | /* some more paranoia, if the request was over-determined */ |
bde89a9e | 250 | if (adm_ctx.device && adm_ctx.connection && |
a6b32bc3 | 251 | first_peer_device(adm_ctx.device)->connection != adm_ctx.connection) { |
44e52cfa | 252 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to connection %s\n", |
7c3063cc | 253 | adm_ctx.minor, adm_ctx.resource_name, |
d8628a86 | 254 | adm_ctx.device->resource->name); |
44e52cfa | 255 | drbd_msg_put_info("minor exists in different resource"); |
527f4b24 LE |
256 | return ERR_INVALID_REQUEST; |
257 | } | |
b30ab791 | 258 | if (adm_ctx.device && |
3b98c0c2 | 259 | adm_ctx.volume != VOLUME_UNSPECIFIED && |
b30ab791 | 260 | adm_ctx.volume != adm_ctx.device->vnr) { |
3b98c0c2 LE |
261 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
262 | adm_ctx.minor, adm_ctx.volume, | |
77c556f6 | 263 | adm_ctx.device->vnr, |
d8628a86 | 264 | adm_ctx.device->resource->name); |
527f4b24 | 265 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
266 | return ERR_INVALID_REQUEST; |
267 | } | |
0ace9dfa | 268 | |
3b98c0c2 LE |
269 | return NO_ERROR; |
270 | ||
271 | fail: | |
272 | nlmsg_free(adm_ctx.reply_skb); | |
273 | adm_ctx.reply_skb = NULL; | |
1e2a2551 | 274 | return err; |
3b98c0c2 LE |
275 | } |
276 | ||
277 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
278 | { | |
bde89a9e | 279 | if (adm_ctx.connection) { |
05a10ec7 | 280 | kref_put(&adm_ctx.connection->kref, drbd_destroy_connection); |
bde89a9e | 281 | adm_ctx.connection = NULL; |
0ace9dfa PR |
282 | } |
283 | ||
3b98c0c2 LE |
284 | if (!adm_ctx.reply_skb) |
285 | return -ENOMEM; | |
286 | ||
287 | adm_ctx.reply_dh->ret_code = retcode; | |
3b98c0c2 LE |
288 | drbd_adm_send_reply(adm_ctx.reply_skb, info); |
289 | return 0; | |
290 | } | |
b411b363 | 291 | |
bde89a9e | 292 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 293 | { |
6b75dced | 294 | char *afs; |
b411b363 | 295 | |
089c075d | 296 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 297 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
298 | return; |
299 | ||
bde89a9e | 300 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
301 | case AF_INET6: |
302 | afs = "ipv6"; | |
303 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 304 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 305 | break; |
089c075d AG |
306 | case AF_INET: |
307 | afs = "ipv4"; | |
308 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 309 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 310 | break; |
089c075d AG |
311 | default: |
312 | afs = "ssocks"; | |
313 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 314 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 315 | } |
089c075d | 316 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 317 | } |
b411b363 | 318 | |
b30ab791 | 319 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
320 | { |
321 | char *envp[] = { "HOME=/", | |
322 | "TERM=linux", | |
323 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
324 | (char[20]) { }, /* address family */ |
325 | (char[60]) { }, /* address */ | |
b411b363 | 326 | NULL }; |
6b75dced | 327 | char mb[12]; |
b411b363 | 328 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 329 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 330 | struct sib_info sib; |
b411b363 PR |
331 | int ret; |
332 | ||
bde89a9e AG |
333 | if (current == connection->worker.task) |
334 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 335 | |
b30ab791 | 336 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 337 | setup_khelper_env(connection, envp); |
b411b363 | 338 | |
1090c056 LE |
339 | /* The helper may take some time. |
340 | * write out any unsynced meta data changes now */ | |
b30ab791 | 341 | drbd_md_sync(device); |
1090c056 | 342 | |
b411b363 | 343 | dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
344 | sib.sib_reason = SIB_HELPER_PRE; |
345 | sib.helper_name = cmd; | |
b30ab791 | 346 | drbd_bcast_event(device, &sib); |
70834d30 | 347 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 PR |
348 | if (ret) |
349 | dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
350 | usermode_helper, cmd, mb, | |
351 | (ret >> 8) & 0xff, ret); | |
352 | else | |
353 | dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
354 | usermode_helper, cmd, mb, | |
355 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
356 | sib.sib_reason = SIB_HELPER_POST; |
357 | sib.helper_exit_code = ret; | |
b30ab791 | 358 | drbd_bcast_event(device, &sib); |
b411b363 | 359 | |
bde89a9e AG |
360 | if (current == connection->worker.task) |
361 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
362 | |
363 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
364 | ret = 0; | |
365 | ||
366 | return ret; | |
367 | } | |
368 | ||
bde89a9e | 369 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
370 | { |
371 | char *envp[] = { "HOME=/", | |
372 | "TERM=linux", | |
373 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
374 | (char[20]) { }, /* address family */ | |
375 | (char[60]) { }, /* address */ | |
376 | NULL }; | |
77c556f6 AG |
377 | char *resource_name = connection->resource->name; |
378 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
379 | int ret; |
380 | ||
bde89a9e AG |
381 | setup_khelper_env(connection, envp); |
382 | conn_md_sync(connection); | |
6b75dced | 383 | |
77c556f6 | 384 | conn_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced PR |
385 | /* TODO: conn_bcast_event() ?? */ |
386 | ||
98683650 | 387 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 388 | if (ret) |
bde89a9e | 389 | conn_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 390 | usermode_helper, cmd, resource_name, |
6b75dced PR |
391 | (ret >> 8) & 0xff, ret); |
392 | else | |
bde89a9e | 393 | conn_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 394 | usermode_helper, cmd, resource_name, |
6b75dced PR |
395 | (ret >> 8) & 0xff, ret); |
396 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 397 | |
b411b363 PR |
398 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
399 | ret = 0; | |
400 | ||
401 | return ret; | |
402 | } | |
403 | ||
bde89a9e | 404 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 405 | { |
cb703454 | 406 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
b30ab791 | 407 | struct drbd_device *device; |
cb703454 PR |
408 | int vnr; |
409 | ||
695d08fa | 410 | rcu_read_lock(); |
bde89a9e | 411 | idr_for_each_entry(&connection->volumes, device, vnr) { |
b30ab791 | 412 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
daeda1cc | 413 | fp = max_t(enum drbd_fencing_p, fp, |
b30ab791 AG |
414 | rcu_dereference(device->ldev->disk_conf)->fencing); |
415 | put_ldev(device); | |
cb703454 PR |
416 | } |
417 | } | |
695d08fa | 418 | rcu_read_unlock(); |
cb703454 PR |
419 | |
420 | return fp; | |
421 | } | |
422 | ||
bde89a9e | 423 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 424 | { |
28e448bb | 425 | unsigned int connect_cnt; |
cb703454 PR |
426 | union drbd_state mask = { }; |
427 | union drbd_state val = { }; | |
428 | enum drbd_fencing_p fp; | |
b411b363 PR |
429 | char *ex_to_string; |
430 | int r; | |
b411b363 | 431 | |
bde89a9e AG |
432 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
433 | conn_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); | |
cb703454 PR |
434 | return false; |
435 | } | |
b411b363 | 436 | |
bde89a9e AG |
437 | spin_lock_irq(&connection->req_lock); |
438 | connect_cnt = connection->connect_cnt; | |
439 | spin_unlock_irq(&connection->req_lock); | |
28e448bb | 440 | |
bde89a9e | 441 | fp = highest_fencing_policy(connection); |
cb703454 PR |
442 | switch (fp) { |
443 | case FP_NOT_AVAIL: | |
bde89a9e | 444 | conn_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 445 | goto out; |
cb703454 PR |
446 | case FP_DONT_CARE: |
447 | return true; | |
448 | default: ; | |
b411b363 PR |
449 | } |
450 | ||
bde89a9e | 451 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
452 | |
453 | switch ((r>>8) & 0xff) { | |
454 | case 3: /* peer is inconsistent */ | |
455 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
456 | mask.pdsk = D_MASK; |
457 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
458 | break; |
459 | case 4: /* peer got outdated, or was already outdated */ | |
460 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
461 | mask.pdsk = D_MASK; |
462 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
463 | break; |
464 | case 5: /* peer was down */ | |
bde89a9e | 465 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
466 | /* we will(have) create(d) a new UUID anyways... */ |
467 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
468 | mask.pdsk = D_MASK; |
469 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
470 | } else { |
471 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
472 | } |
473 | break; | |
474 | case 6: /* Peer is primary, voluntarily outdate myself. | |
475 | * This is useful when an unconnected R_SECONDARY is asked to | |
476 | * become R_PRIMARY, but finds the other peer being active. */ | |
477 | ex_to_string = "peer is active"; | |
bde89a9e | 478 | conn_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
479 | mask.disk = D_MASK; |
480 | val.disk = D_OUTDATED; | |
b411b363 PR |
481 | break; |
482 | case 7: | |
483 | if (fp != FP_STONITH) | |
bde89a9e | 484 | conn_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 485 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
486 | mask.pdsk = D_MASK; |
487 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
488 | break; |
489 | default: | |
490 | /* The script is broken ... */ | |
bde89a9e | 491 | conn_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 492 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
493 | } |
494 | ||
bde89a9e | 495 | conn_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 496 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 497 | |
cb703454 | 498 | out: |
fb22c402 | 499 | |
cb703454 | 500 | /* Not using |
bde89a9e | 501 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
502 | here, because we might were able to re-establish the connection in the |
503 | meantime. */ | |
bde89a9e AG |
504 | spin_lock_irq(&connection->req_lock); |
505 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { | |
506 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
507 | /* In case the connection was established and droped |
508 | while the fence-peer handler was running, ignore it */ | |
bde89a9e | 509 | conn_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 510 | else |
bde89a9e | 511 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 512 | } |
bde89a9e | 513 | spin_unlock_irq(&connection->req_lock); |
cb703454 | 514 | |
bde89a9e | 515 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
516 | } |
517 | ||
87f7be4c PR |
518 | static int _try_outdate_peer_async(void *data) |
519 | { | |
bde89a9e | 520 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 521 | |
bde89a9e | 522 | conn_try_outdate_peer(connection); |
87f7be4c | 523 | |
05a10ec7 | 524 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
525 | return 0; |
526 | } | |
527 | ||
bde89a9e | 528 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
529 | { |
530 | struct task_struct *opa; | |
531 | ||
bde89a9e AG |
532 | kref_get(&connection->kref); |
533 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); | |
9dc9fbb3 | 534 | if (IS_ERR(opa)) { |
bde89a9e | 535 | conn_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 536 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 537 | } |
87f7be4c | 538 | } |
b411b363 | 539 | |
bf885f8a | 540 | enum drbd_state_rv |
b30ab791 | 541 | drbd_set_role(struct drbd_device *device, enum drbd_role new_role, int force) |
b411b363 PR |
542 | { |
543 | const int max_tries = 4; | |
bf885f8a | 544 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 545 | struct net_conf *nc; |
b411b363 PR |
546 | int try = 0; |
547 | int forced = 0; | |
548 | union drbd_state mask, val; | |
b411b363 PR |
549 | |
550 | if (new_role == R_PRIMARY) | |
a6b32bc3 | 551 | request_ping(first_peer_device(device)->connection); /* Detect a dead peer ASAP */ |
b411b363 | 552 | |
b30ab791 | 553 | mutex_lock(device->state_mutex); |
b411b363 PR |
554 | |
555 | mask.i = 0; mask.role = R_MASK; | |
556 | val.i = 0; val.role = new_role; | |
557 | ||
558 | while (try++ < max_tries) { | |
b30ab791 | 559 | rv = _drbd_request_state(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
560 | |
561 | /* in case we first succeeded to outdate, | |
562 | * but now suddenly could establish a connection */ | |
bf885f8a | 563 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
564 | val.pdsk = 0; |
565 | mask.pdsk = 0; | |
566 | continue; | |
567 | } | |
568 | ||
bf885f8a | 569 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
570 | (device->state.disk < D_UP_TO_DATE && |
571 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
572 | mask.disk = D_MASK; |
573 | val.disk = D_UP_TO_DATE; | |
574 | forced = 1; | |
575 | continue; | |
576 | } | |
577 | ||
bf885f8a | 578 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 AG |
579 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
580 | D_ASSERT(device->state.pdsk == D_UNKNOWN); | |
b411b363 | 581 | |
a6b32bc3 | 582 | if (conn_try_outdate_peer(first_peer_device(device)->connection)) { |
b411b363 PR |
583 | val.disk = D_UP_TO_DATE; |
584 | mask.disk = D_MASK; | |
585 | } | |
b411b363 PR |
586 | continue; |
587 | } | |
588 | ||
bf885f8a | 589 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 590 | goto out; |
bf885f8a | 591 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
a6b32bc3 | 592 | if (!conn_try_outdate_peer(first_peer_device(device)->connection) && force) { |
b411b363 | 593 | dev_warn(DEV, "Forced into split brain situation!\n"); |
cb703454 PR |
594 | mask.pdsk = D_MASK; |
595 | val.pdsk = D_OUTDATED; | |
b411b363 | 596 | |
cb703454 | 597 | } |
b411b363 PR |
598 | continue; |
599 | } | |
bf885f8a | 600 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
601 | /* Maybe the peer is detected as dead very soon... |
602 | retry at most once more in this case. */ | |
44ed167d PR |
603 | int timeo; |
604 | rcu_read_lock(); | |
a6b32bc3 | 605 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d PR |
606 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
607 | rcu_read_unlock(); | |
608 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
609 | if (try < max_tries) |
610 | try = max_tries - 1; | |
611 | continue; | |
612 | } | |
bf885f8a | 613 | if (rv < SS_SUCCESS) { |
b30ab791 | 614 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 615 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 616 | if (rv < SS_SUCCESS) |
3b98c0c2 | 617 | goto out; |
b411b363 PR |
618 | } |
619 | break; | |
620 | } | |
621 | ||
bf885f8a | 622 | if (rv < SS_SUCCESS) |
3b98c0c2 | 623 | goto out; |
b411b363 PR |
624 | |
625 | if (forced) | |
626 | dev_warn(DEV, "Forced to consider local data as UpToDate!\n"); | |
627 | ||
628 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 629 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 630 | |
b6dd1a89 LE |
631 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
632 | ||
b411b363 | 633 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
634 | set_disk_ro(device->vdisk, true); |
635 | if (get_ldev(device)) { | |
636 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
637 | put_ldev(device); | |
b411b363 PR |
638 | } |
639 | } else { | |
a6b32bc3 AG |
640 | mutex_lock(&first_peer_device(device)->connection->conf_update); |
641 | nc = first_peer_device(device)->connection->net_conf; | |
44ed167d | 642 | if (nc) |
6139f60d | 643 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
a6b32bc3 | 644 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
91fd4dad | 645 | |
b30ab791 AG |
646 | set_disk_ro(device->vdisk, false); |
647 | if (get_ldev(device)) { | |
648 | if (((device->state.conn < C_CONNECTED || | |
649 | device->state.pdsk <= D_FAILED) | |
650 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
651 | drbd_uuid_new_current(device); | |
b411b363 | 652 | |
b30ab791 AG |
653 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
654 | put_ldev(device); | |
b411b363 PR |
655 | } |
656 | } | |
657 | ||
19f843aa LE |
658 | /* writeout of activity log covered areas of the bitmap |
659 | * to stable storage done in after state change already */ | |
b411b363 | 660 | |
b30ab791 | 661 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
662 | /* if this was forced, we should consider sync */ |
663 | if (forced) | |
b30ab791 AG |
664 | drbd_send_uuids(device); |
665 | drbd_send_current_state(device); | |
b411b363 PR |
666 | } |
667 | ||
b30ab791 | 668 | drbd_md_sync(device); |
b411b363 | 669 | |
b30ab791 | 670 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 671 | out: |
b30ab791 | 672 | mutex_unlock(device->state_mutex); |
bf885f8a | 673 | return rv; |
b411b363 PR |
674 | } |
675 | ||
3b98c0c2 | 676 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 677 | { |
3b98c0c2 LE |
678 | return err == -ENOMSG ? "required attribute missing" : |
679 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 680 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 681 | "invalid attribute value"; |
ef50a3e3 | 682 | } |
b411b363 | 683 | |
3b98c0c2 | 684 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 685 | { |
3b98c0c2 LE |
686 | struct set_role_parms parms; |
687 | int err; | |
688 | enum drbd_ret_code retcode; | |
b411b363 | 689 | |
3b98c0c2 LE |
690 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
691 | if (!adm_ctx.reply_skb) | |
692 | return retcode; | |
693 | if (retcode != NO_ERROR) | |
694 | goto out; | |
b411b363 | 695 | |
3b98c0c2 LE |
696 | memset(&parms, 0, sizeof(parms)); |
697 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 698 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
699 | if (err) { |
700 | retcode = ERR_MANDATORY_TAG; | |
701 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
702 | goto out; | |
703 | } | |
704 | } | |
b411b363 | 705 | |
3b98c0c2 | 706 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 707 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 708 | else |
b30ab791 | 709 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
3b98c0c2 LE |
710 | out: |
711 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
712 | return 0; |
713 | } | |
714 | ||
ae8bf312 LE |
715 | /* Initializes the md.*_offset members, so we are able to find |
716 | * the on disk meta data. | |
717 | * | |
718 | * We currently have two possible layouts: | |
719 | * external: | |
720 | * |----------- md_size_sect ------------------| | |
721 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
722 | * | al_offset == 8 | | |
723 | * | bm_offset = al_offset + X | | |
724 | * ==> bitmap sectors = md_size_sect - bm_offset | |
725 | * | |
726 | * internal: | |
727 | * |----------- md_size_sect ------------------| | |
728 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
729 | * | al_offset < 0 | | |
730 | * | bm_offset = al_offset - Y | | |
731 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
732 | * | |
733 | * Activity log size used to be fixed 32kB, | |
734 | * but is about to become configurable. | |
735 | */ | |
b30ab791 | 736 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
737 | struct drbd_backing_dev *bdev) |
738 | { | |
739 | sector_t md_size_sect = 0; | |
c04ccaa6 | 740 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 741 | |
3a4d4eb3 LE |
742 | bdev->md.md_offset = drbd_md_ss(bdev); |
743 | ||
68e41a43 | 744 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
745 | default: |
746 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 747 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
748 | bdev->md.al_offset = MD_4kB_SECT; |
749 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
750 | break; |
751 | case DRBD_MD_INDEX_FLEX_EXT: | |
752 | /* just occupy the full device; unit: sectors */ | |
753 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
754 | bdev->md.al_offset = MD_4kB_SECT; |
755 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
756 | break; |
757 | case DRBD_MD_INDEX_INTERNAL: | |
758 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 759 | /* al size is still fixed */ |
ae8bf312 | 760 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
761 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
762 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
763 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
764 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
765 | md_size_sect = ALIGN(md_size_sect, 8); | |
766 | ||
767 | /* plus the "drbd meta data super block", | |
768 | * and the activity log; */ | |
ae8bf312 | 769 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
770 | |
771 | bdev->md.md_size_sect = md_size_sect; | |
772 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 773 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
774 | break; |
775 | } | |
776 | } | |
777 | ||
4b0715f0 | 778 | /* input size is expected to be in KB */ |
b411b363 PR |
779 | char *ppsize(char *buf, unsigned long long size) |
780 | { | |
4b0715f0 LE |
781 | /* Needs 9 bytes at max including trailing NUL: |
782 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
783 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
784 | int base = 0; | |
4b0715f0 | 785 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
786 | /* shift + round */ |
787 | size = (size >> 10) + !!(size & (1<<9)); | |
788 | base++; | |
789 | } | |
4b0715f0 | 790 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
791 | |
792 | return buf; | |
793 | } | |
794 | ||
795 | /* there is still a theoretical deadlock when called from receiver | |
796 | * on an D_INCONSISTENT R_PRIMARY: | |
797 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
798 | * packet from remote to dec_ap_bio again. | |
799 | * receiver receive_sizes(), comes here, | |
800 | * waits for ap_bio_cnt == 0. -> deadlock. | |
801 | * but this cannot happen, actually, because: | |
802 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
803 | * (not connected, or bad/no disk on peer): | |
804 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
805 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
806 | * peer may not initiate a resize. | |
807 | */ | |
3b98c0c2 LE |
808 | /* Note these are not to be confused with |
809 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
810 | * which are (sub) state changes triggered by admin (drbdsetup), | |
811 | * and can be long lived. | |
b30ab791 | 812 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 813 | * and should be short-lived. */ |
b30ab791 | 814 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 815 | { |
b30ab791 AG |
816 | set_bit(SUSPEND_IO, &device->flags); |
817 | if (drbd_suspended(device)) | |
265be2d0 | 818 | return; |
b30ab791 | 819 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
820 | } |
821 | ||
b30ab791 | 822 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 823 | { |
b30ab791 AG |
824 | clear_bit(SUSPEND_IO, &device->flags); |
825 | wake_up(&device->misc_wait); | |
b411b363 PR |
826 | } |
827 | ||
828 | /** | |
829 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 830 | * @device: DRBD device. |
b411b363 PR |
831 | * |
832 | * Returns 0 on success, negative return values indicate errors. | |
833 | * You should call drbd_md_sync() after calling this function. | |
834 | */ | |
d752b269 | 835 | enum determine_dev_size |
b30ab791 | 836 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
837 | { |
838 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 839 | sector_t la_size_sect, u_size; |
b30ab791 | 840 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
841 | u32 prev_al_stripe_size_4k; |
842 | u32 prev_al_stripes; | |
b411b363 PR |
843 | sector_t size; |
844 | char ppb[10]; | |
d752b269 | 845 | void *buffer; |
b411b363 PR |
846 | |
847 | int md_moved, la_size_changed; | |
e96c9633 | 848 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
849 | |
850 | /* race: | |
851 | * application request passes inc_ap_bio, | |
852 | * but then cannot get an AL-reference. | |
853 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
854 | * | |
855 | * to avoid that: | |
856 | * Suspend IO right here. | |
857 | * still lock the act_log to not trigger ASSERTs there. | |
858 | */ | |
b30ab791 AG |
859 | drbd_suspend_io(device); |
860 | buffer = drbd_md_get_buffer(device); /* Lock meta-data IO */ | |
d752b269 | 861 | if (!buffer) { |
b30ab791 | 862 | drbd_resume_io(device); |
d752b269 PR |
863 | return DS_ERROR; |
864 | } | |
b411b363 PR |
865 | |
866 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 867 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 868 | |
b30ab791 AG |
869 | prev_first_sect = drbd_md_first_sector(device->ldev); |
870 | prev_size = device->ldev->md.md_size_sect; | |
871 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 872 | |
d752b269 PR |
873 | if (rs) { |
874 | /* rs is non NULL if we should change the AL layout only */ | |
875 | ||
876 | prev_al_stripes = md->al_stripes; | |
877 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
878 | ||
879 | md->al_stripes = rs->al_stripes; | |
880 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
881 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
882 | } | |
883 | ||
b30ab791 | 884 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 885 | |
daeda1cc | 886 | rcu_read_lock(); |
b30ab791 | 887 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 888 | rcu_read_unlock(); |
b30ab791 | 889 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 890 | |
d752b269 PR |
891 | if (size < la_size_sect) { |
892 | if (rs && u_size == 0) { | |
893 | /* Remove "rs &&" later. This check should always be active, but | |
894 | right now the receiver expects the permissive behavior */ | |
895 | dev_warn(DEV, "Implicit shrink not allowed. " | |
896 | "Use --size=%llus for explicit shrink.\n", | |
897 | (unsigned long long)size); | |
898 | rv = DS_ERROR_SHRINK; | |
899 | } | |
900 | if (u_size > size) | |
901 | rv = DS_ERROR_SPACE_MD; | |
902 | if (rv != DS_UNCHANGED) | |
903 | goto err_out; | |
904 | } | |
905 | ||
b30ab791 AG |
906 | if (drbd_get_capacity(device->this_bdev) != size || |
907 | drbd_bm_capacity(device) != size) { | |
b411b363 | 908 | int err; |
b30ab791 | 909 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
910 | if (unlikely(err)) { |
911 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 912 | size = drbd_bm_capacity(device)>>1; |
b411b363 PR |
913 | if (size == 0) { |
914 | dev_err(DEV, "OUT OF MEMORY! " | |
915 | "Could not allocate bitmap!\n"); | |
916 | } else { | |
917 | dev_err(DEV, "BM resizing failed. " | |
918 | "Leaving size unchanged at size = %lu KB\n", | |
919 | (unsigned long)size); | |
920 | } | |
e96c9633 | 921 | rv = DS_ERROR; |
b411b363 PR |
922 | } |
923 | /* racy, see comments above. */ | |
b30ab791 AG |
924 | drbd_set_my_capacity(device, size); |
925 | device->ldev->md.la_size_sect = size; | |
b411b363 PR |
926 | dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
927 | (unsigned long long)size>>1); | |
928 | } | |
d752b269 PR |
929 | if (rv <= DS_ERROR) |
930 | goto err_out; | |
b411b363 | 931 | |
b30ab791 | 932 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 933 | |
b30ab791 AG |
934 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
935 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 936 | |
d752b269 PR |
937 | if (la_size_changed || md_moved || rs) { |
938 | u32 prev_flags; | |
24dccabb | 939 | |
b30ab791 | 940 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
941 | |
942 | prev_flags = md->flags; | |
943 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 944 | drbd_md_write(device, buffer); |
d752b269 | 945 | |
b411b363 PR |
946 | dev_info(DEV, "Writing the whole bitmap, %s\n", |
947 | la_size_changed && md_moved ? "size changed and md moved" : | |
948 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 949 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 950 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 951 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 952 | drbd_initialize_al(device, buffer); |
d752b269 PR |
953 | |
954 | md->flags = prev_flags; | |
b30ab791 | 955 | drbd_md_write(device, buffer); |
d752b269 PR |
956 | |
957 | if (rs) | |
958 | dev_info(DEV, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", | |
959 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
960 | } |
961 | ||
cccac985 | 962 | if (size > la_size_sect) |
57737adc | 963 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 964 | if (size < la_size_sect) |
e96c9633 | 965 | rv = DS_SHRUNK; |
d752b269 PR |
966 | |
967 | if (0) { | |
968 | err_out: | |
969 | if (rs) { | |
970 | md->al_stripes = prev_al_stripes; | |
971 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
972 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
973 | ||
b30ab791 | 974 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
975 | } |
976 | } | |
b30ab791 AG |
977 | lc_unlock(device->act_log); |
978 | wake_up(&device->al_wait); | |
979 | drbd_md_put_buffer(device); | |
980 | drbd_resume_io(device); | |
b411b363 PR |
981 | |
982 | return rv; | |
983 | } | |
984 | ||
985 | sector_t | |
b30ab791 | 986 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 987 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 988 | { |
b30ab791 | 989 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 990 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 991 | sector_t m_size; /* my size */ |
b411b363 PR |
992 | sector_t size = 0; |
993 | ||
994 | m_size = drbd_get_max_capacity(bdev); | |
995 | ||
b30ab791 | 996 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
a393db6f PR |
997 | dev_warn(DEV, "Resize while not connected was forced by the user!\n"); |
998 | p_size = m_size; | |
999 | } | |
1000 | ||
b411b363 PR |
1001 | if (p_size && m_size) { |
1002 | size = min_t(sector_t, p_size, m_size); | |
1003 | } else { | |
cccac985 LE |
1004 | if (la_size_sect) { |
1005 | size = la_size_sect; | |
b411b363 PR |
1006 | if (m_size && m_size < size) |
1007 | size = m_size; | |
1008 | if (p_size && p_size < size) | |
1009 | size = p_size; | |
1010 | } else { | |
1011 | if (m_size) | |
1012 | size = m_size; | |
1013 | if (p_size) | |
1014 | size = p_size; | |
1015 | } | |
1016 | } | |
1017 | ||
1018 | if (size == 0) | |
1019 | dev_err(DEV, "Both nodes diskless!\n"); | |
1020 | ||
1021 | if (u_size) { | |
1022 | if (u_size > size) | |
1023 | dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n", | |
1024 | (unsigned long)u_size>>1, (unsigned long)size>>1); | |
1025 | else | |
1026 | size = u_size; | |
1027 | } | |
1028 | ||
1029 | return size; | |
1030 | } | |
1031 | ||
1032 | /** | |
1033 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1034 | * @device: DRBD device. |
b411b363 PR |
1035 | * |
1036 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1037 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1038 | * this function. | |
1039 | */ | |
b30ab791 | 1040 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1041 | { |
1042 | struct lru_cache *n, *t; | |
1043 | struct lc_element *e; | |
1044 | unsigned int in_use; | |
1045 | int i; | |
1046 | ||
b30ab791 AG |
1047 | if (device->act_log && |
1048 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1049 | return 0; |
1050 | ||
1051 | in_use = 0; | |
b30ab791 | 1052 | t = device->act_log; |
7ad651b5 | 1053 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1054 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1055 | |
1056 | if (n == NULL) { | |
1057 | dev_err(DEV, "Cannot allocate act_log lru!\n"); | |
1058 | return -ENOMEM; | |
1059 | } | |
b30ab791 | 1060 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1061 | if (t) { |
1062 | for (i = 0; i < t->nr_elements; i++) { | |
1063 | e = lc_element_by_index(t, i); | |
1064 | if (e->refcnt) | |
1065 | dev_err(DEV, "refcnt(%d)==%d\n", | |
1066 | e->lc_number, e->refcnt); | |
1067 | in_use += e->refcnt; | |
1068 | } | |
1069 | } | |
1070 | if (!in_use) | |
b30ab791 AG |
1071 | device->act_log = n; |
1072 | spin_unlock_irq(&device->al_lock); | |
b411b363 PR |
1073 | if (in_use) { |
1074 | dev_err(DEV, "Activity log still in use!\n"); | |
1075 | lc_destroy(n); | |
1076 | return -EBUSY; | |
1077 | } else { | |
1078 | if (t) | |
1079 | lc_destroy(t); | |
1080 | } | |
b30ab791 | 1081 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1082 | return 0; |
1083 | } | |
1084 | ||
b30ab791 | 1085 | static void drbd_setup_queue_param(struct drbd_device *device, unsigned int max_bio_size) |
b411b363 | 1086 | { |
b30ab791 | 1087 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1088 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1089 | unsigned int max_segments = 0; | |
99432fcc | 1090 | |
b30ab791 AG |
1091 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1092 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1093 | |
1094 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1095 | rcu_read_lock(); |
b30ab791 | 1096 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1097 | rcu_read_unlock(); |
b30ab791 | 1098 | put_ldev(device); |
99432fcc | 1099 | } |
b411b363 | 1100 | |
b411b363 | 1101 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1102 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1103 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1104 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1105 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1106 | |
b30ab791 AG |
1107 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1108 | struct request_queue * const b = device->ldev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1109 | |
1110 | blk_queue_stack_limits(q, b); | |
1111 | ||
1112 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
1113 | dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", | |
1114 | q->backing_dev_info.ra_pages, | |
1115 | b->backing_dev_info.ra_pages); | |
1116 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1117 | } | |
b30ab791 | 1118 | put_ldev(device); |
99432fcc PR |
1119 | } |
1120 | } | |
1121 | ||
b30ab791 | 1122 | void drbd_reconsider_max_bio_size(struct drbd_device *device) |
99432fcc | 1123 | { |
db141b2f | 1124 | unsigned int now, new, local, peer; |
99432fcc | 1125 | |
b30ab791 AG |
1126 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1127 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1128 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1129 | |
b30ab791 AG |
1130 | if (get_ldev_if_state(device, D_ATTACHING)) { |
1131 | local = queue_max_hw_sectors(device->ldev->backing_bdev->bd_disk->queue) << 9; | |
1132 | device->local_max_bio_size = local; | |
1133 | put_ldev(device); | |
b411b363 | 1134 | } |
db141b2f | 1135 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1136 | |
1137 | /* We may ignore peer limits if the peer is modern enough. | |
1138 | Because new from 8.3.8 onwards the peer can use multiple | |
1139 | BIOs for a single peer_request */ | |
b30ab791 | 1140 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1141 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1142 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1143 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1144 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1145 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1146 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1147 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1148 | else | |
99432fcc PR |
1149 | peer = DRBD_MAX_BIO_SIZE; |
1150 | } | |
1151 | ||
db141b2f | 1152 | new = min(local, peer); |
99432fcc | 1153 | |
b30ab791 | 1154 | if (device->state.role == R_PRIMARY && new < now) |
db141b2f | 1155 | dev_err(DEV, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1156 | |
1157 | if (new != now) | |
1158 | dev_info(DEV, "max BIO size = %u\n", new); | |
1159 | ||
b30ab791 | 1160 | drbd_setup_queue_param(device, new); |
b411b363 PR |
1161 | } |
1162 | ||
a18e9d1e | 1163 | /* Starts the worker thread */ |
bde89a9e | 1164 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1165 | { |
bde89a9e AG |
1166 | drbd_thread_start(&connection->worker); |
1167 | conn_flush_workqueue(connection); | |
b411b363 PR |
1168 | } |
1169 | ||
a18e9d1e | 1170 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1171 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1172 | { |
992d6e91 | 1173 | bool stop_threads; |
bde89a9e AG |
1174 | spin_lock_irq(&connection->req_lock); |
1175 | stop_threads = conn_all_vols_unconf(connection) && | |
1176 | connection->cstate == C_STANDALONE; | |
1177 | spin_unlock_irq(&connection->req_lock); | |
992d6e91 LE |
1178 | if (stop_threads) { |
1179 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1180 | * in conn_disconnect() */ |
bde89a9e AG |
1181 | drbd_thread_stop(&connection->receiver); |
1182 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1183 | } |
b411b363 PR |
1184 | } |
1185 | ||
0778286a | 1186 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1187 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1188 | { |
1189 | int s = 0; | |
1190 | ||
b30ab791 | 1191 | if (!lc_try_lock(device->act_log)) { |
0778286a PR |
1192 | dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n"); |
1193 | return; | |
1194 | } | |
1195 | ||
b30ab791 | 1196 | drbd_al_shrink(device); |
a6b32bc3 | 1197 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 AG |
1198 | if (device->state.conn < C_CONNECTED) |
1199 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
a6b32bc3 | 1200 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 1201 | lc_unlock(device->act_log); |
0778286a PR |
1202 | |
1203 | if (s) | |
1204 | dev_info(DEV, "Suspended AL updates\n"); | |
1205 | } | |
1206 | ||
5979e361 LE |
1207 | |
1208 | static bool should_set_defaults(struct genl_info *info) | |
1209 | { | |
1210 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1211 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1212 | } | |
1213 | ||
5bbcf5e6 | 1214 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1215 | { |
5bbcf5e6 LE |
1216 | /* This is limited by 16 bit "slot" numbers, |
1217 | * and by available on-disk context storage. | |
1218 | * | |
1219 | * Also (u16)~0 is special (denotes a "free" extent). | |
1220 | * | |
1221 | * One transaction occupies one 4kB on-disk block, | |
1222 | * we have n such blocks in the on disk ring buffer, | |
1223 | * the "current" transaction may fail (n-1), | |
1224 | * and there is 919 slot numbers context information per transaction. | |
1225 | * | |
1226 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1227 | * so cap there first. | |
1228 | */ | |
1229 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1230 | const unsigned int sufficient_on_disk = | |
1231 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1232 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1233 | |
5bbcf5e6 LE |
1234 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1235 | ||
1236 | if (al_size_4k > sufficient_on_disk) | |
1237 | return max_al_nr; | |
1238 | ||
1239 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1240 | } |
1241 | ||
f399002e LE |
1242 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1243 | { | |
1244 | enum drbd_ret_code retcode; | |
b30ab791 | 1245 | struct drbd_device *device; |
daeda1cc | 1246 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1247 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1248 | int err, fifo_size; |
f399002e LE |
1249 | |
1250 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1251 | if (!adm_ctx.reply_skb) | |
1252 | return retcode; | |
1253 | if (retcode != NO_ERROR) | |
1254 | goto out; | |
1255 | ||
b30ab791 | 1256 | device = adm_ctx.device; |
f399002e LE |
1257 | |
1258 | /* we also need a disk | |
1259 | * to change the options on */ | |
b30ab791 | 1260 | if (!get_ldev(device)) { |
f399002e LE |
1261 | retcode = ERR_NO_DISK; |
1262 | goto out; | |
1263 | } | |
1264 | ||
daeda1cc | 1265 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1266 | if (!new_disk_conf) { |
f399002e LE |
1267 | retcode = ERR_NOMEM; |
1268 | goto fail; | |
1269 | } | |
1270 | ||
a6b32bc3 | 1271 | mutex_lock(&first_peer_device(device)->connection->conf_update); |
b30ab791 | 1272 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1273 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1274 | if (should_set_defaults(info)) |
b966b5dd | 1275 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1276 | |
5ecc72c3 | 1277 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1278 | if (err && err != -ENOMSG) { |
f399002e LE |
1279 | retcode = ERR_MANDATORY_TAG; |
1280 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
8e229434 | 1281 | goto fail_unlock; |
f399002e LE |
1282 | } |
1283 | ||
5ecc72c3 LE |
1284 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1285 | new_disk_conf->resync_rate = 1; | |
f399002e | 1286 | |
5bbcf5e6 LE |
1287 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1288 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1289 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1290 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1291 | |
1292 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1293 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1294 | |
5ecc72c3 | 1295 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1296 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1297 | new_plan = fifo_alloc(fifo_size); |
1298 | if (!new_plan) { | |
f399002e LE |
1299 | dev_err(DEV, "kmalloc of fifo_buffer failed"); |
1300 | retcode = ERR_NOMEM; | |
daeda1cc | 1301 | goto fail_unlock; |
f399002e LE |
1302 | } |
1303 | } | |
1304 | ||
b30ab791 AG |
1305 | drbd_suspend_io(device); |
1306 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1307 | drbd_al_shrink(device); | |
1308 | err = drbd_check_al_size(device, new_disk_conf); | |
1309 | lc_unlock(device->act_log); | |
1310 | wake_up(&device->al_wait); | |
1311 | drbd_resume_io(device); | |
f399002e LE |
1312 | |
1313 | if (err) { | |
1314 | retcode = ERR_NOMEM; | |
daeda1cc | 1315 | goto fail_unlock; |
f399002e LE |
1316 | } |
1317 | ||
dc97b708 | 1318 | write_lock_irq(&global_state_lock); |
b30ab791 | 1319 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1320 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1321 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1322 | drbd_resync_after_changed(device); | |
dc97b708 PR |
1323 | } |
1324 | write_unlock_irq(&global_state_lock); | |
f399002e | 1325 | |
daeda1cc PR |
1326 | if (retcode != NO_ERROR) |
1327 | goto fail_unlock; | |
f399002e | 1328 | |
813472ce | 1329 | if (new_plan) { |
b30ab791 AG |
1330 | old_plan = device->rs_plan_s; |
1331 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1332 | } |
9958c857 | 1333 | |
a6b32bc3 | 1334 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
27eb13e9 | 1335 | |
9a51ab1c | 1336 | if (new_disk_conf->al_updates) |
b30ab791 | 1337 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1338 | else |
b30ab791 | 1339 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1340 | |
691631c0 | 1341 | if (new_disk_conf->md_flushes) |
b30ab791 | 1342 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1343 | else |
b30ab791 | 1344 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1345 | |
a6b32bc3 | 1346 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
27eb13e9 | 1347 | |
b30ab791 | 1348 | drbd_md_sync(device); |
f399002e | 1349 | |
b30ab791 AG |
1350 | if (device->state.conn >= C_CONNECTED) |
1351 | drbd_send_sync_param(device); | |
f399002e | 1352 | |
daeda1cc PR |
1353 | synchronize_rcu(); |
1354 | kfree(old_disk_conf); | |
813472ce | 1355 | kfree(old_plan); |
b30ab791 | 1356 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1357 | goto success; |
1358 | ||
1359 | fail_unlock: | |
a6b32bc3 | 1360 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
f399002e | 1361 | fail: |
5ecc72c3 | 1362 | kfree(new_disk_conf); |
813472ce | 1363 | kfree(new_plan); |
daeda1cc | 1364 | success: |
b30ab791 | 1365 | put_ldev(device); |
f399002e LE |
1366 | out: |
1367 | drbd_adm_finish(info, retcode); | |
1368 | return 0; | |
1369 | } | |
1370 | ||
3b98c0c2 | 1371 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1372 | { |
b30ab791 | 1373 | struct drbd_device *device; |
3b98c0c2 | 1374 | int err; |
116676ca | 1375 | enum drbd_ret_code retcode; |
b411b363 PR |
1376 | enum determine_dev_size dd; |
1377 | sector_t max_possible_sectors; | |
1378 | sector_t min_md_device_sectors; | |
1379 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1380 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1381 | struct block_device *bdev; |
b411b363 | 1382 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1383 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1384 | union drbd_state ns, os; |
f2024e7c | 1385 | enum drbd_state_rv rv; |
44ed167d | 1386 | struct net_conf *nc; |
b411b363 | 1387 | |
3b98c0c2 LE |
1388 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1389 | if (!adm_ctx.reply_skb) | |
1390 | return retcode; | |
1391 | if (retcode != NO_ERROR) | |
40cbf085 | 1392 | goto finish; |
b411b363 | 1393 | |
b30ab791 | 1394 | device = adm_ctx.device; |
a6b32bc3 | 1395 | conn_reconfig_start(first_peer_device(device)->connection); |
b411b363 PR |
1396 | |
1397 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1398 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1399 | retcode = ERR_DISK_CONFIGURED; |
1400 | goto fail; | |
1401 | } | |
82f59cc6 LE |
1402 | /* It may just now have detached because of IO error. Make sure |
1403 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1404 | * e.g. if someone calls attach from the on-io-error handler, | |
1405 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
b30ab791 | 1406 | wait_event(device->misc_wait, !atomic_read(&device->local_cnt)); |
b411b363 | 1407 | |
383606e0 | 1408 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1409 | clear_bit(FORCE_DETACH, &device->flags); |
1410 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1411 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1412 | |
0029d624 | 1413 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1414 | device->rs_total = 0; |
1415 | device->rs_failed = 0; | |
1416 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1417 | |
3b98c0c2 | 1418 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1419 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1420 | if (!nbc) { | |
1421 | retcode = ERR_NOMEM; | |
1422 | goto fail; | |
1423 | } | |
9f2247bb PR |
1424 | spin_lock_init(&nbc->md.uuid_lock); |
1425 | ||
daeda1cc PR |
1426 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1427 | if (!new_disk_conf) { | |
1428 | retcode = ERR_NOMEM; | |
b411b363 PR |
1429 | goto fail; |
1430 | } | |
daeda1cc | 1431 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1432 | |
daeda1cc PR |
1433 | set_disk_conf_defaults(new_disk_conf); |
1434 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1435 | if (err) { |
b411b363 | 1436 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1437 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1438 | goto fail; |
1439 | } | |
1440 | ||
5bbcf5e6 LE |
1441 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1442 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1443 | |
9958c857 PR |
1444 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1445 | if (!new_plan) { | |
1446 | retcode = ERR_NOMEM; | |
1447 | goto fail; | |
1448 | } | |
1449 | ||
daeda1cc | 1450 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1451 | retcode = ERR_MD_IDX_INVALID; |
1452 | goto fail; | |
1453 | } | |
1454 | ||
a3f8f7dc | 1455 | write_lock_irq(&global_state_lock); |
b30ab791 | 1456 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
a3f8f7dc LE |
1457 | write_unlock_irq(&global_state_lock); |
1458 | if (retcode != NO_ERROR) | |
1459 | goto fail; | |
1460 | ||
44ed167d | 1461 | rcu_read_lock(); |
a6b32bc3 | 1462 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
44ed167d | 1463 | if (nc) { |
daeda1cc | 1464 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1465 | rcu_read_unlock(); |
47ff2d0a PR |
1466 | retcode = ERR_STONITH_AND_PROT_A; |
1467 | goto fail; | |
1468 | } | |
1469 | } | |
44ed167d | 1470 | rcu_read_unlock(); |
47ff2d0a | 1471 | |
daeda1cc | 1472 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1473 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1474 | if (IS_ERR(bdev)) { |
daeda1cc | 1475 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1476 | PTR_ERR(bdev)); |
b411b363 PR |
1477 | retcode = ERR_OPEN_DISK; |
1478 | goto fail; | |
1479 | } | |
e525fd89 TH |
1480 | nbc->backing_bdev = bdev; |
1481 | ||
1482 | /* | |
1483 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1484 | * drbd sharing one meta device. TODO in that case, paranoia | |
1485 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1486 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1487 | * should check it for you already; but if you don't, or | |
1488 | * someone fooled it, we need to double check here) | |
1489 | */ | |
daeda1cc | 1490 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1491 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1492 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1493 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1494 | if (IS_ERR(bdev)) { |
daeda1cc | 1495 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1496 | PTR_ERR(bdev)); |
b411b363 PR |
1497 | retcode = ERR_OPEN_MD_DISK; |
1498 | goto fail; | |
1499 | } | |
e525fd89 | 1500 | nbc->md_bdev = bdev; |
b411b363 | 1501 | |
e525fd89 | 1502 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1503 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1504 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1505 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1506 | goto fail; |
1507 | } | |
1508 | ||
1509 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1510 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1511 | offsetof(struct bm_extent, lce)); |
1512 | if (!resync_lru) { | |
1513 | retcode = ERR_NOMEM; | |
e525fd89 | 1514 | goto fail; |
b411b363 PR |
1515 | } |
1516 | ||
c04ccaa6 LE |
1517 | /* Read our meta data super block early. |
1518 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1519 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 LE |
1520 | if (retcode != NO_ERROR) |
1521 | goto fail; | |
b411b363 | 1522 | |
5bbcf5e6 LE |
1523 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1524 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1525 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1526 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1527 | ||
daeda1cc | 1528 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
b411b363 PR |
1529 | dev_err(DEV, "max capacity %llu smaller than disk size %llu\n", |
1530 | (unsigned long long) drbd_get_max_capacity(nbc), | |
daeda1cc | 1531 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1532 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1533 | goto fail; |
b411b363 PR |
1534 | } |
1535 | ||
daeda1cc | 1536 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1537 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1538 | /* at least one MB, otherwise it does not make sense */ | |
1539 | min_md_device_sectors = (2<<10); | |
1540 | } else { | |
1541 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1542 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1543 | } |
1544 | ||
b411b363 | 1545 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1546 | retcode = ERR_MD_DISK_TOO_SMALL; |
b411b363 PR |
1547 | dev_warn(DEV, "refusing attach: md-device too small, " |
1548 | "at least %llu sectors needed for this meta-disk type\n", | |
1549 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1550 | goto fail; |
b411b363 PR |
1551 | } |
1552 | ||
1553 | /* Make sure the new disk is big enough | |
1554 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1555 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1556 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1557 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1558 | goto fail; |
b411b363 PR |
1559 | } |
1560 | ||
1561 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1562 | ||
1352994b LE |
1563 | if (nbc->known_size > max_possible_sectors) { |
1564 | dev_warn(DEV, "==> truncating very big lower level device " | |
1565 | "to currently maximum possible %llu sectors <==\n", | |
1566 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1567 | if (new_disk_conf->meta_dev_idx >= 0) |
1352994b LE |
1568 | dev_warn(DEV, "==>> using internal or flexible " |
1569 | "meta data may help <<==\n"); | |
1570 | } | |
1571 | ||
b30ab791 | 1572 | drbd_suspend_io(device); |
b411b363 | 1573 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1574 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1575 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1576 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1577 | * As barriers are counted per resource, | |
1578 | * we'd need to suspend io on all devices of a resource. | |
1579 | */ | |
b30ab791 | 1580 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1581 | /* and for any other previously queued work */ |
b30ab791 | 1582 | drbd_flush_workqueue(device); |
b411b363 | 1583 | |
b30ab791 | 1584 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1585 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1586 | drbd_resume_io(device); |
f2024e7c | 1587 | if (rv < SS_SUCCESS) |
e525fd89 | 1588 | goto fail; |
b411b363 | 1589 | |
b30ab791 | 1590 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1591 | goto force_diskless; |
1592 | ||
b30ab791 AG |
1593 | if (!device->bitmap) { |
1594 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1595 | retcode = ERR_NOMEM; |
1596 | goto force_diskless_dec; | |
1597 | } | |
1598 | } | |
1599 | ||
b30ab791 AG |
1600 | if (device->state.conn < C_CONNECTED && |
1601 | device->state.role == R_PRIMARY && | |
1602 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
b411b363 | 1603 | dev_err(DEV, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1604 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1605 | retcode = ERR_DATA_NOT_CURRENT; |
1606 | goto force_diskless_dec; | |
1607 | } | |
1608 | ||
1609 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1610 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1611 | retcode = ERR_NOMEM; |
1612 | goto force_diskless_dec; | |
1613 | } | |
1614 | ||
1615 | /* Prevent shrinking of consistent devices ! */ | |
1616 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1617 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
b411b363 | 1618 | dev_warn(DEV, "refusing to truncate a consistent device\n"); |
7948bcdc | 1619 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1620 | goto force_diskless_dec; |
1621 | } | |
1622 | ||
b411b363 PR |
1623 | /* Reset the "barriers don't work" bits here, then force meta data to |
1624 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1625 | if (new_disk_conf->md_flushes) |
b30ab791 | 1626 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1627 | else |
b30ab791 | 1628 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1629 | |
1630 | /* Point of no return reached. | |
1631 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1632 | * now device takes over responsibility, and the state engine should |
b411b363 | 1633 | * clean it up somewhere. */ |
b30ab791 AG |
1634 | D_ASSERT(device->ldev == NULL); |
1635 | device->ldev = nbc; | |
1636 | device->resync = resync_lru; | |
1637 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1638 | nbc = NULL; |
1639 | resync_lru = NULL; | |
daeda1cc | 1640 | new_disk_conf = NULL; |
9958c857 | 1641 | new_plan = NULL; |
b411b363 | 1642 | |
a6b32bc3 | 1643 | drbd_bump_write_ordering(first_peer_device(device)->connection, WO_bdev_flush); |
b411b363 | 1644 | |
b30ab791 AG |
1645 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1646 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1647 | else |
b30ab791 | 1648 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1649 | |
b30ab791 | 1650 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
a6b32bc3 AG |
1651 | !(device->state.role == R_PRIMARY && |
1652 | first_peer_device(device)->connection->susp_nod)) | |
b30ab791 | 1653 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1654 | |
b30ab791 AG |
1655 | device->send_cnt = 0; |
1656 | device->recv_cnt = 0; | |
1657 | device->read_cnt = 0; | |
1658 | device->writ_cnt = 0; | |
b411b363 | 1659 | |
b30ab791 | 1660 | drbd_reconsider_max_bio_size(device); |
b411b363 PR |
1661 | |
1662 | /* If I am currently not R_PRIMARY, | |
1663 | * but meta data primary indicator is set, | |
1664 | * I just now recover from a hard crash, | |
1665 | * and have been R_PRIMARY before that crash. | |
1666 | * | |
1667 | * Now, if I had no connection before that crash | |
1668 | * (have been degraded R_PRIMARY), chances are that | |
1669 | * I won't find my peer now either. | |
1670 | * | |
1671 | * In that case, and _only_ in that case, | |
1672 | * we use the degr-wfc-timeout instead of the default, | |
1673 | * so we can automatically recover from a crash of a | |
1674 | * degraded but active "cluster" after a certain timeout. | |
1675 | */ | |
b30ab791 AG |
1676 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1677 | if (device->state.role != R_PRIMARY && | |
1678 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1679 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1680 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1681 | |
b30ab791 | 1682 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1683 | if (dd <= DS_ERROR) { |
b411b363 PR |
1684 | retcode = ERR_NOMEM_BITMAP; |
1685 | goto force_diskless_dec; | |
e96c9633 | 1686 | } else if (dd == DS_GREW) |
b30ab791 | 1687 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1688 | |
b30ab791 AG |
1689 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1690 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1691 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
b411b363 PR |
1692 | dev_info(DEV, "Assuming that all blocks are out of sync " |
1693 | "(aka FullSync)\n"); | |
b30ab791 | 1694 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1695 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1696 | retcode = ERR_IO_MD_DISK; |
1697 | goto force_diskless_dec; | |
1698 | } | |
1699 | } else { | |
b30ab791 | 1700 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1701 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1702 | retcode = ERR_IO_MD_DISK; |
1703 | goto force_diskless_dec; | |
1704 | } | |
b411b363 PR |
1705 | } |
1706 | ||
b30ab791 AG |
1707 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1708 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1709 | |
a6b32bc3 | 1710 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 1711 | os = drbd_read_state(device); |
78bae59b | 1712 | ns = os; |
b411b363 PR |
1713 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1714 | otherwise investigate MDF_WasUpToDate... | |
1715 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1716 | otherwise into D_CONSISTENT state. | |
1717 | */ | |
b30ab791 AG |
1718 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1719 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1720 | ns.disk = D_CONSISTENT; |
1721 | else | |
1722 | ns.disk = D_OUTDATED; | |
1723 | } else { | |
1724 | ns.disk = D_INCONSISTENT; | |
1725 | } | |
1726 | ||
b30ab791 | 1727 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1728 | ns.pdsk = D_OUTDATED; |
1729 | ||
daeda1cc PR |
1730 | rcu_read_lock(); |
1731 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1732 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1733 | ns.disk = D_UP_TO_DATE; |
1734 | ||
1735 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1736 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1737 | this point, because drbd_request_state() modifies these | |
1738 | flags. */ | |
1739 | ||
b30ab791 AG |
1740 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1741 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1742 | else |
b30ab791 | 1743 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1744 | |
1745 | rcu_read_unlock(); | |
1746 | ||
b411b363 PR |
1747 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1748 | state after the negotiation phase. */ | |
b30ab791 AG |
1749 | if (device->state.conn == C_CONNECTED) { |
1750 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1751 | ns.i = os.i; |
1752 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1753 | |
1754 | /* We expect to receive up-to-date UUIDs soon. | |
1755 | To avoid a race in receive_state, free p_uuid while | |
1756 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1757 | kfree(device->p_uuid); |
1758 | device->p_uuid = NULL; | |
b411b363 PR |
1759 | } |
1760 | ||
b30ab791 | 1761 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
a6b32bc3 | 1762 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
b411b363 PR |
1763 | |
1764 | if (rv < SS_SUCCESS) | |
1765 | goto force_diskless_dec; | |
1766 | ||
b30ab791 | 1767 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1768 | |
b30ab791 AG |
1769 | if (device->state.role == R_PRIMARY) |
1770 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1771 | else |
b30ab791 | 1772 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1773 | |
b30ab791 AG |
1774 | drbd_md_mark_dirty(device); |
1775 | drbd_md_sync(device); | |
b411b363 | 1776 | |
b30ab791 AG |
1777 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1778 | put_ldev(device); | |
a6b32bc3 | 1779 | conn_reconfig_done(first_peer_device(device)->connection); |
3b98c0c2 | 1780 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1781 | return 0; |
1782 | ||
1783 | force_diskless_dec: | |
b30ab791 | 1784 | put_ldev(device); |
b411b363 | 1785 | force_diskless: |
b30ab791 AG |
1786 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1787 | drbd_md_sync(device); | |
b411b363 | 1788 | fail: |
a6b32bc3 | 1789 | conn_reconfig_done(first_peer_device(device)->connection); |
b411b363 | 1790 | if (nbc) { |
e525fd89 TH |
1791 | if (nbc->backing_bdev) |
1792 | blkdev_put(nbc->backing_bdev, | |
1793 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1794 | if (nbc->md_bdev) | |
1795 | blkdev_put(nbc->md_bdev, | |
1796 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1797 | kfree(nbc); |
1798 | } | |
daeda1cc | 1799 | kfree(new_disk_conf); |
b411b363 | 1800 | lc_destroy(resync_lru); |
9958c857 | 1801 | kfree(new_plan); |
b411b363 | 1802 | |
40cbf085 | 1803 | finish: |
3b98c0c2 | 1804 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1805 | return 0; |
1806 | } | |
1807 | ||
b30ab791 | 1808 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1809 | { |
19f83c76 | 1810 | enum drbd_state_rv retcode; |
9a0d9d03 | 1811 | int ret; |
02ee8f95 | 1812 | |
cdfda633 | 1813 | if (force) { |
b30ab791 AG |
1814 | set_bit(FORCE_DETACH, &device->flags); |
1815 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1816 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1817 | goto out; |
1818 | } | |
1819 | ||
b30ab791 AG |
1820 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
1821 | drbd_md_get_buffer(device); /* make sure there is no in-flight meta-data IO */ | |
1822 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
1823 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1824 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1825 | ret = wait_event_interruptible(device->misc_wait, |
1826 | device->state.disk != D_FAILED); | |
1827 | drbd_resume_io(device); | |
9b2f61ae | 1828 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1829 | retcode = SS_NOTHING_TO_DO; |
1830 | if (ret) | |
1831 | retcode = ERR_INTR; | |
02ee8f95 | 1832 | out: |
85f75dd7 | 1833 | return retcode; |
b411b363 PR |
1834 | } |
1835 | ||
82f59cc6 LE |
1836 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1837 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1838 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1839 | * internal references as well. | |
1840 | * Only then we have finally detached. */ | |
3b98c0c2 | 1841 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1842 | { |
116676ca | 1843 | enum drbd_ret_code retcode; |
cdfda633 PR |
1844 | struct detach_parms parms = { }; |
1845 | int err; | |
b411b363 | 1846 | |
3b98c0c2 LE |
1847 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1848 | if (!adm_ctx.reply_skb) | |
1849 | return retcode; | |
1850 | if (retcode != NO_ERROR) | |
1851 | goto out; | |
b411b363 | 1852 | |
cdfda633 PR |
1853 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1854 | err = detach_parms_from_attrs(&parms, info); | |
1855 | if (err) { | |
1856 | retcode = ERR_MANDATORY_TAG; | |
1857 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1858 | goto out; | |
1859 | } | |
b411b363 PR |
1860 | } |
1861 | ||
b30ab791 | 1862 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
3b98c0c2 LE |
1863 | out: |
1864 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1865 | return 0; |
1866 | } | |
b411b363 | 1867 | |
bde89a9e | 1868 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1869 | { |
b30ab791 | 1870 | struct drbd_device *device; |
695d08fa | 1871 | bool rv = false; |
f399002e LE |
1872 | int vnr; |
1873 | ||
695d08fa | 1874 | rcu_read_lock(); |
bde89a9e | 1875 | idr_for_each_entry(&connection->volumes, device, vnr) { |
b30ab791 AG |
1876 | if (device->state.conn == C_SYNC_SOURCE || |
1877 | device->state.conn == C_SYNC_TARGET || | |
1878 | device->state.conn == C_PAUSED_SYNC_S || | |
1879 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1880 | rv = true; |
1881 | break; | |
1882 | } | |
b411b363 | 1883 | } |
695d08fa | 1884 | rcu_read_unlock(); |
b411b363 | 1885 | |
695d08fa | 1886 | return rv; |
f399002e | 1887 | } |
47ff2d0a | 1888 | |
bde89a9e | 1889 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 1890 | { |
b30ab791 | 1891 | struct drbd_device *device; |
695d08fa | 1892 | bool rv = false; |
f399002e LE |
1893 | int vnr; |
1894 | ||
695d08fa | 1895 | rcu_read_lock(); |
bde89a9e | 1896 | idr_for_each_entry(&connection->volumes, device, vnr) { |
b30ab791 AG |
1897 | if (device->state.conn == C_VERIFY_S || |
1898 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
1899 | rv = true; |
1900 | break; | |
47ff2d0a PR |
1901 | } |
1902 | } | |
695d08fa | 1903 | rcu_read_unlock(); |
b411b363 | 1904 | |
695d08fa | 1905 | return rv; |
f399002e | 1906 | } |
422028b1 | 1907 | |
cd64397c | 1908 | static enum drbd_ret_code |
bde89a9e | 1909 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_conf, struct net_conf *new_conf) |
cd64397c | 1910 | { |
b30ab791 | 1911 | struct drbd_device *device; |
cd64397c | 1912 | int i; |
b411b363 | 1913 | |
bde89a9e | 1914 | if (old_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
dcb20d1a PR |
1915 | if (new_conf->wire_protocol != old_conf->wire_protocol) |
1916 | return ERR_NEED_APV_100; | |
b411b363 | 1917 | |
dcb20d1a PR |
1918 | if (new_conf->two_primaries != old_conf->two_primaries) |
1919 | return ERR_NEED_APV_100; | |
1920 | ||
dcb20d1a PR |
1921 | if (strcmp(new_conf->integrity_alg, old_conf->integrity_alg)) |
1922 | return ERR_NEED_APV_100; | |
b411b363 PR |
1923 | } |
1924 | ||
dcb20d1a | 1925 | if (!new_conf->two_primaries && |
bde89a9e AG |
1926 | conn_highest_role(connection) == R_PRIMARY && |
1927 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 1928 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 1929 | |
cd64397c PR |
1930 | if (new_conf->two_primaries && |
1931 | (new_conf->wire_protocol != DRBD_PROT_C)) | |
1932 | return ERR_NOT_PROTO_C; | |
1933 | ||
bde89a9e | 1934 | idr_for_each_entry(&connection->volumes, device, i) { |
b30ab791 AG |
1935 | if (get_ldev(device)) { |
1936 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
1937 | put_ldev(device); | |
44ed167d | 1938 | if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1939 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 1940 | } |
b30ab791 | 1941 | if (device->state.role == R_PRIMARY && new_conf->discard_my_data) |
eb12010e | 1942 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
1943 | } |
1944 | ||
cd64397c PR |
1945 | if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A) |
1946 | return ERR_CONG_NOT_PROTO_A; | |
b411b363 | 1947 | |
cd64397c PR |
1948 | return NO_ERROR; |
1949 | } | |
b411b363 | 1950 | |
44ed167d | 1951 | static enum drbd_ret_code |
bde89a9e | 1952 | check_net_options(struct drbd_connection *connection, struct net_conf *new_conf) |
44ed167d PR |
1953 | { |
1954 | static enum drbd_ret_code rv; | |
b30ab791 | 1955 | struct drbd_device *device; |
44ed167d | 1956 | int i; |
b411b363 | 1957 | |
44ed167d | 1958 | rcu_read_lock(); |
bde89a9e | 1959 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_conf); |
44ed167d | 1960 | rcu_read_unlock(); |
b411b363 | 1961 | |
bde89a9e AG |
1962 | /* connection->volumes protected by genl_lock() here */ |
1963 | idr_for_each_entry(&connection->volumes, device, i) { | |
b30ab791 AG |
1964 | if (!device->bitmap) { |
1965 | if (drbd_bm_init(device)) | |
44ed167d | 1966 | return ERR_NOMEM; |
b411b363 PR |
1967 | } |
1968 | } | |
1969 | ||
44ed167d PR |
1970 | return rv; |
1971 | } | |
b411b363 | 1972 | |
0fd0ea06 PR |
1973 | struct crypto { |
1974 | struct crypto_hash *verify_tfm; | |
1975 | struct crypto_hash *csums_tfm; | |
1976 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 1977 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 1978 | }; |
b411b363 | 1979 | |
0fd0ea06 | 1980 | static int |
4b6ad6d4 | 1981 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
1982 | { |
1983 | if (!tfm_name[0]) | |
1984 | return NO_ERROR; | |
b411b363 | 1985 | |
0fd0ea06 PR |
1986 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
1987 | if (IS_ERR(*tfm)) { | |
1988 | *tfm = NULL; | |
1989 | return err_alg; | |
b411b363 | 1990 | } |
b411b363 | 1991 | |
0fd0ea06 PR |
1992 | return NO_ERROR; |
1993 | } | |
b411b363 | 1994 | |
0fd0ea06 PR |
1995 | static enum drbd_ret_code |
1996 | alloc_crypto(struct crypto *crypto, struct net_conf *new_conf) | |
1997 | { | |
1998 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
1999 | enum drbd_ret_code rv; | |
0fd0ea06 | 2000 | |
4b6ad6d4 AG |
2001 | rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg, |
2002 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
2003 | if (rv != NO_ERROR) |
2004 | return rv; | |
4b6ad6d4 AG |
2005 | rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg, |
2006 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
2007 | if (rv != NO_ERROR) |
2008 | return rv; | |
4b6ad6d4 AG |
2009 | rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg, |
2010 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
2011 | if (rv != NO_ERROR) |
2012 | return rv; | |
0fd0ea06 PR |
2013 | if (new_conf->cram_hmac_alg[0] != 0) { |
2014 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", | |
2015 | new_conf->cram_hmac_alg); | |
b411b363 | 2016 | |
4b6ad6d4 AG |
2017 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2018 | ERR_AUTH_ALG); | |
b411b363 PR |
2019 | } |
2020 | ||
0fd0ea06 PR |
2021 | return rv; |
2022 | } | |
b411b363 | 2023 | |
0fd0ea06 PR |
2024 | static void free_crypto(struct crypto *crypto) |
2025 | { | |
0fd0ea06 | 2026 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2027 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2028 | crypto_free_hash(crypto->csums_tfm); |
2029 | crypto_free_hash(crypto->verify_tfm); | |
2030 | } | |
b411b363 | 2031 | |
f399002e LE |
2032 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2033 | { | |
2034 | enum drbd_ret_code retcode; | |
bde89a9e | 2035 | struct drbd_connection *connection; |
44ed167d | 2036 | struct net_conf *old_conf, *new_conf = NULL; |
f399002e LE |
2037 | int err; |
2038 | int ovr; /* online verify running */ | |
2039 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2040 | struct crypto crypto = { }; |
b411b363 | 2041 | |
089c075d | 2042 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2043 | if (!adm_ctx.reply_skb) |
2044 | return retcode; | |
2045 | if (retcode != NO_ERROR) | |
2046 | goto out; | |
b411b363 | 2047 | |
bde89a9e | 2048 | connection = adm_ctx.connection; |
b411b363 | 2049 | |
f399002e LE |
2050 | new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2051 | if (!new_conf) { | |
2052 | retcode = ERR_NOMEM; | |
2053 | goto out; | |
2054 | } | |
b411b363 | 2055 | |
bde89a9e | 2056 | conn_reconfig_start(connection); |
b411b363 | 2057 | |
bde89a9e AG |
2058 | mutex_lock(&connection->data.mutex); |
2059 | mutex_lock(&connection->conf_update); | |
2060 | old_conf = connection->net_conf; | |
2561b9c1 | 2061 | |
44ed167d | 2062 | if (!old_conf) { |
f399002e LE |
2063 | drbd_msg_put_info("net conf missing, try connect"); |
2064 | retcode = ERR_INVALID_REQUEST; | |
2561b9c1 PR |
2065 | goto fail; |
2066 | } | |
2067 | ||
44ed167d | 2068 | *new_conf = *old_conf; |
5979e361 | 2069 | if (should_set_defaults(info)) |
b966b5dd | 2070 | set_net_conf_defaults(new_conf); |
f399002e | 2071 | |
f399002e | 2072 | err = net_conf_from_attrs_for_change(new_conf, info); |
c75b9b10 | 2073 | if (err && err != -ENOMSG) { |
f399002e LE |
2074 | retcode = ERR_MANDATORY_TAG; |
2075 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2076 | goto fail; | |
2561b9c1 | 2077 | } |
b411b363 | 2078 | |
bde89a9e | 2079 | retcode = check_net_options(connection, new_conf); |
cd64397c PR |
2080 | if (retcode != NO_ERROR) |
2081 | goto fail; | |
b411b363 | 2082 | |
f399002e | 2083 | /* re-sync running */ |
bde89a9e | 2084 | rsr = conn_resync_running(connection); |
0fd0ea06 | 2085 | if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) { |
f399002e | 2086 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2087 | goto fail; |
b411b363 PR |
2088 | } |
2089 | ||
f399002e | 2090 | /* online verify running */ |
bde89a9e | 2091 | ovr = conn_ov_running(connection); |
0fd0ea06 PR |
2092 | if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) { |
2093 | retcode = ERR_VERIFY_RUNNING; | |
b411b363 | 2094 | goto fail; |
f399002e | 2095 | } |
b411b363 | 2096 | |
0fd0ea06 PR |
2097 | retcode = alloc_crypto(&crypto, new_conf); |
2098 | if (retcode != NO_ERROR) | |
b411b363 | 2099 | goto fail; |
f399002e | 2100 | |
bde89a9e | 2101 | rcu_assign_pointer(connection->net_conf, new_conf); |
f399002e LE |
2102 | |
2103 | if (!rsr) { | |
bde89a9e AG |
2104 | crypto_free_hash(connection->csums_tfm); |
2105 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2106 | crypto.csums_tfm = NULL; |
f399002e LE |
2107 | } |
2108 | if (!ovr) { | |
bde89a9e AG |
2109 | crypto_free_hash(connection->verify_tfm); |
2110 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2111 | crypto.verify_tfm = NULL; |
b411b363 PR |
2112 | } |
2113 | ||
bde89a9e AG |
2114 | crypto_free_hash(connection->integrity_tfm); |
2115 | connection->integrity_tfm = crypto.integrity_tfm; | |
2116 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2117 | /* Do this without trying to take connection->data.mutex again. */ | |
2118 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2119 | |
bde89a9e AG |
2120 | crypto_free_hash(connection->cram_hmac_tfm); |
2121 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2122 | |
bde89a9e AG |
2123 | mutex_unlock(&connection->conf_update); |
2124 | mutex_unlock(&connection->data.mutex); | |
91fd4dad PR |
2125 | synchronize_rcu(); |
2126 | kfree(old_conf); | |
2127 | ||
bde89a9e AG |
2128 | if (connection->cstate >= C_WF_REPORT_PARAMS) |
2129 | drbd_send_sync_param(minor_to_device(conn_lowest_minor(connection))); | |
f399002e | 2130 | |
91fd4dad PR |
2131 | goto done; |
2132 | ||
b411b363 | 2133 | fail: |
bde89a9e AG |
2134 | mutex_unlock(&connection->conf_update); |
2135 | mutex_unlock(&connection->data.mutex); | |
0fd0ea06 | 2136 | free_crypto(&crypto); |
f399002e | 2137 | kfree(new_conf); |
91fd4dad | 2138 | done: |
bde89a9e | 2139 | conn_reconfig_done(connection); |
f399002e LE |
2140 | out: |
2141 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2142 | return 0; |
2143 | } | |
2144 | ||
3b98c0c2 | 2145 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2146 | { |
b30ab791 | 2147 | struct drbd_device *device; |
44ed167d | 2148 | struct net_conf *old_conf, *new_conf = NULL; |
0fd0ea06 | 2149 | struct crypto crypto = { }; |
77c556f6 | 2150 | struct drbd_resource *resource; |
bde89a9e | 2151 | struct drbd_connection *connection; |
3b98c0c2 LE |
2152 | enum drbd_ret_code retcode; |
2153 | int i; | |
2154 | int err; | |
b411b363 | 2155 | |
44e52cfa | 2156 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2157 | |
3b98c0c2 LE |
2158 | if (!adm_ctx.reply_skb) |
2159 | return retcode; | |
2160 | if (retcode != NO_ERROR) | |
2161 | goto out; | |
089c075d AG |
2162 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
2163 | drbd_msg_put_info("connection endpoint(s) missing"); | |
2164 | retcode = ERR_INVALID_REQUEST; | |
2165 | goto out; | |
2166 | } | |
b411b363 | 2167 | |
089c075d AG |
2168 | /* No need for _rcu here. All reconfiguration is |
2169 | * strictly serialized on genl_lock(). We are protected against | |
2170 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2171 | for_each_resource(resource, &drbd_resources) { |
2172 | for_each_connection(connection, resource) { | |
2173 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2174 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2175 | connection->my_addr_len)) { | |
2176 | retcode = ERR_LOCAL_ADDR; | |
2177 | goto out; | |
2178 | } | |
b411b363 | 2179 | |
77c556f6 AG |
2180 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2181 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2182 | connection->peer_addr_len)) { | |
2183 | retcode = ERR_PEER_ADDR; | |
2184 | goto out; | |
2185 | } | |
089c075d | 2186 | } |
b411b363 PR |
2187 | } |
2188 | ||
bde89a9e AG |
2189 | connection = adm_ctx.connection; |
2190 | conn_reconfig_start(connection); | |
b411b363 | 2191 | |
bde89a9e | 2192 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2193 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2194 | goto fail; |
2195 | } | |
2196 | ||
a209b4ae | 2197 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
5979e361 | 2198 | new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL); |
b411b363 PR |
2199 | if (!new_conf) { |
2200 | retcode = ERR_NOMEM; | |
b411b363 PR |
2201 | goto fail; |
2202 | } | |
2203 | ||
b966b5dd | 2204 | set_net_conf_defaults(new_conf); |
b411b363 | 2205 | |
f399002e | 2206 | err = net_conf_from_attrs(new_conf, info); |
25e40932 | 2207 | if (err && err != -ENOMSG) { |
b411b363 | 2208 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2209 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2210 | goto fail; |
2211 | } | |
2212 | ||
bde89a9e | 2213 | retcode = check_net_options(connection, new_conf); |
cd64397c | 2214 | if (retcode != NO_ERROR) |
422028b1 | 2215 | goto fail; |
b411b363 | 2216 | |
0fd0ea06 PR |
2217 | retcode = alloc_crypto(&crypto, new_conf); |
2218 | if (retcode != NO_ERROR) | |
2219 | goto fail; | |
b411b363 | 2220 | |
b411b363 | 2221 | ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2222 | |
bde89a9e | 2223 | conn_flush_workqueue(connection); |
b411b363 | 2224 | |
bde89a9e AG |
2225 | mutex_lock(&connection->conf_update); |
2226 | old_conf = connection->net_conf; | |
91fd4dad | 2227 | if (old_conf) { |
b411b363 | 2228 | retcode = ERR_NET_CONFIGURED; |
bde89a9e | 2229 | mutex_unlock(&connection->conf_update); |
b411b363 PR |
2230 | goto fail; |
2231 | } | |
bde89a9e | 2232 | rcu_assign_pointer(connection->net_conf, new_conf); |
b411b363 | 2233 | |
bde89a9e AG |
2234 | conn_free_crypto(connection); |
2235 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2236 | connection->integrity_tfm = crypto.integrity_tfm; | |
2237 | connection->csums_tfm = crypto.csums_tfm; | |
2238 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2239 | |
bde89a9e AG |
2240 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2241 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2242 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2243 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2244 | |
bde89a9e | 2245 | mutex_unlock(&connection->conf_update); |
b411b363 | 2246 | |
695d08fa | 2247 | rcu_read_lock(); |
bde89a9e | 2248 | idr_for_each_entry(&connection->volumes, device, i) { |
b30ab791 AG |
2249 | device->send_cnt = 0; |
2250 | device->recv_cnt = 0; | |
b411b363 | 2251 | } |
695d08fa | 2252 | rcu_read_unlock(); |
b411b363 | 2253 | |
bde89a9e | 2254 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2255 | |
bde89a9e | 2256 | conn_reconfig_done(connection); |
3b98c0c2 | 2257 | drbd_adm_finish(info, retcode); |
b411b363 | 2258 | return 0; |
b411b363 | 2259 | |
b411b363 | 2260 | fail: |
0fd0ea06 | 2261 | free_crypto(&crypto); |
b411b363 | 2262 | kfree(new_conf); |
b411b363 | 2263 | |
bde89a9e | 2264 | conn_reconfig_done(connection); |
3b98c0c2 LE |
2265 | out: |
2266 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2267 | return 0; |
2268 | } | |
2269 | ||
bde89a9e | 2270 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2271 | { |
2272 | enum drbd_state_rv rv; | |
85f75dd7 | 2273 | |
bde89a9e | 2274 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2275 | force ? CS_HARD : 0); |
85f75dd7 LE |
2276 | |
2277 | switch (rv) { | |
2278 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2279 | break; |
85f75dd7 LE |
2280 | case SS_ALREADY_STANDALONE: |
2281 | return SS_SUCCESS; | |
2282 | case SS_PRIMARY_NOP: | |
2283 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2284 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2285 | |
2286 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2287 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2288 | |
85f75dd7 LE |
2289 | break; |
2290 | case SS_CW_FAILED_BY_PEER: | |
2291 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2292 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2293 | disk, D_OUTDATED), 0); |
2294 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2295 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2296 | CS_HARD); |
b411b363 | 2297 | } |
85f75dd7 LE |
2298 | break; |
2299 | default:; | |
2300 | /* no special handling necessary */ | |
2301 | } | |
2302 | ||
f3dfa40a LE |
2303 | if (rv >= SS_SUCCESS) { |
2304 | enum drbd_state_rv rv2; | |
2305 | /* No one else can reconfigure the network while I am here. | |
2306 | * The state handling only uses drbd_thread_stop_nowait(), | |
2307 | * we want to really wait here until the receiver is no more. | |
2308 | */ | |
bde89a9e | 2309 | drbd_thread_stop(&adm_ctx.connection->receiver); |
f3dfa40a LE |
2310 | |
2311 | /* Race breaker. This additional state change request may be | |
2312 | * necessary, if this was a forced disconnect during a receiver | |
2313 | * restart. We may have "killed" the receiver thread just | |
2314 | * after drbdd_init() returned. Typically, we should be | |
2315 | * C_STANDALONE already, now, and this becomes a no-op. | |
2316 | */ | |
bde89a9e | 2317 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2318 | CS_VERBOSE | CS_HARD); |
2319 | if (rv2 < SS_SUCCESS) | |
bde89a9e | 2320 | conn_err(connection, |
f3dfa40a LE |
2321 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2322 | rv2); | |
b411b363 | 2323 | } |
85f75dd7 LE |
2324 | return rv; |
2325 | } | |
b411b363 | 2326 | |
3b98c0c2 | 2327 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2328 | { |
3b98c0c2 | 2329 | struct disconnect_parms parms; |
bde89a9e | 2330 | struct drbd_connection *connection; |
85f75dd7 | 2331 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2332 | enum drbd_ret_code retcode; |
2333 | int err; | |
2561b9c1 | 2334 | |
089c075d | 2335 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2336 | if (!adm_ctx.reply_skb) |
2337 | return retcode; | |
2338 | if (retcode != NO_ERROR) | |
2561b9c1 | 2339 | goto fail; |
b411b363 | 2340 | |
bde89a9e | 2341 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2342 | memset(&parms, 0, sizeof(parms)); |
2343 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2344 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2345 | if (err) { |
2346 | retcode = ERR_MANDATORY_TAG; | |
2347 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
b411b363 PR |
2348 | goto fail; |
2349 | } | |
2350 | } | |
2351 | ||
bde89a9e | 2352 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2353 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2354 | retcode = rv; /* FIXME: Type mismatch. */ |
2355 | else | |
2356 | retcode = NO_ERROR; | |
b411b363 | 2357 | fail: |
3b98c0c2 | 2358 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2359 | return 0; |
2360 | } | |
2361 | ||
b30ab791 | 2362 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2363 | { |
2364 | int iass; /* I am sync source */ | |
2365 | ||
2366 | dev_info(DEV, "Resync of new storage after online grow\n"); | |
b30ab791 AG |
2367 | if (device->state.role != device->state.peer) |
2368 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2369 | else |
a6b32bc3 | 2370 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2371 | |
2372 | if (iass) | |
b30ab791 | 2373 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2374 | else |
b30ab791 | 2375 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2376 | } |
2377 | ||
3b98c0c2 | 2378 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2379 | { |
daeda1cc | 2380 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2381 | struct resize_parms rs; |
b30ab791 | 2382 | struct drbd_device *device; |
3b98c0c2 | 2383 | enum drbd_ret_code retcode; |
b411b363 | 2384 | enum determine_dev_size dd; |
d752b269 | 2385 | bool change_al_layout = false; |
6495d2c6 | 2386 | enum dds_flags ddsf; |
daeda1cc | 2387 | sector_t u_size; |
3b98c0c2 | 2388 | int err; |
b411b363 | 2389 | |
3b98c0c2 LE |
2390 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2391 | if (!adm_ctx.reply_skb) | |
2392 | return retcode; | |
2393 | if (retcode != NO_ERROR) | |
b411b363 | 2394 | goto fail; |
3b98c0c2 | 2395 | |
b30ab791 AG |
2396 | device = adm_ctx.device; |
2397 | if (!get_ldev(device)) { | |
d752b269 PR |
2398 | retcode = ERR_NO_DISK; |
2399 | goto fail; | |
2400 | } | |
2401 | ||
3b98c0c2 | 2402 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2403 | rs.al_stripes = device->ldev->md.al_stripes; |
2404 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2405 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2406 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2407 | if (err) { |
3b98c0c2 LE |
2408 | retcode = ERR_MANDATORY_TAG; |
2409 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
d752b269 | 2410 | goto fail_ldev; |
b411b363 PR |
2411 | } |
2412 | } | |
2413 | ||
b30ab791 | 2414 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2415 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2416 | goto fail_ldev; |
b411b363 | 2417 | } |
b411b363 | 2418 | |
b30ab791 AG |
2419 | if (device->state.role == R_SECONDARY && |
2420 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2421 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2422 | goto fail_ldev; |
b411b363 | 2423 | } |
b411b363 | 2424 | |
a6b32bc3 | 2425 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2426 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2427 | goto fail_ldev; |
6495d2c6 PR |
2428 | } |
2429 | ||
daeda1cc | 2430 | rcu_read_lock(); |
b30ab791 | 2431 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2432 | rcu_read_unlock(); |
2433 | if (u_size != (sector_t)rs.resize_size) { | |
2434 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2435 | if (!new_disk_conf) { | |
778f271d | 2436 | retcode = ERR_NOMEM; |
9bcd2521 | 2437 | goto fail_ldev; |
778f271d PR |
2438 | } |
2439 | } | |
2440 | ||
b30ab791 AG |
2441 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2442 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2443 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2444 | ||
2445 | if (al_size_k > (16 * 1024 * 1024)) { | |
2446 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2447 | goto fail_ldev; | |
2448 | } | |
2449 | ||
2450 | if (al_size_k < MD_32kB_SECT/2) { | |
2451 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2452 | goto fail_ldev; | |
2453 | } | |
2454 | ||
b30ab791 | 2455 | if (device->state.conn != C_CONNECTED) { |
d752b269 PR |
2456 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2457 | goto fail_ldev; | |
2458 | } | |
2459 | ||
2460 | change_al_layout = true; | |
2461 | } | |
2462 | ||
b30ab791 AG |
2463 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2464 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2465 | |
daeda1cc | 2466 | if (new_disk_conf) { |
a6b32bc3 | 2467 | mutex_lock(&first_peer_device(device)->connection->conf_update); |
b30ab791 | 2468 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2469 | *new_disk_conf = *old_disk_conf; |
2470 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2471 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
a6b32bc3 | 2472 | mutex_unlock(&first_peer_device(device)->connection->conf_update); |
daeda1cc PR |
2473 | synchronize_rcu(); |
2474 | kfree(old_disk_conf); | |
b411b363 PR |
2475 | } |
2476 | ||
6495d2c6 | 2477 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2478 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2479 | drbd_md_sync(device); | |
2480 | put_ldev(device); | |
e96c9633 | 2481 | if (dd == DS_ERROR) { |
b411b363 PR |
2482 | retcode = ERR_NOMEM_BITMAP; |
2483 | goto fail; | |
d752b269 PR |
2484 | } else if (dd == DS_ERROR_SPACE_MD) { |
2485 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2486 | goto fail; | |
2487 | } else if (dd == DS_ERROR_SHRINK) { | |
2488 | retcode = ERR_IMPLICIT_SHRINK; | |
2489 | goto fail; | |
b411b363 | 2490 | } |
778f271d | 2491 | |
b30ab791 | 2492 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2493 | if (dd == DS_GREW) |
b30ab791 | 2494 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2495 | |
b30ab791 AG |
2496 | drbd_send_uuids(device); |
2497 | drbd_send_sizes(device, 1, ddsf); | |
778f271d PR |
2498 | } |
2499 | ||
b411b363 | 2500 | fail: |
3b98c0c2 | 2501 | drbd_adm_finish(info, retcode); |
b411b363 | 2502 | return 0; |
b411b363 | 2503 | |
9bcd2521 | 2504 | fail_ldev: |
b30ab791 | 2505 | put_ldev(device); |
9bcd2521 | 2506 | goto fail; |
b411b363 | 2507 | } |
b411b363 | 2508 | |
f399002e | 2509 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2510 | { |
3b98c0c2 | 2511 | enum drbd_ret_code retcode; |
bde89a9e | 2512 | struct drbd_connection *connection; |
b57a1e27 | 2513 | struct res_opts res_opts; |
f399002e | 2514 | int err; |
b411b363 | 2515 | |
44e52cfa | 2516 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2517 | if (!adm_ctx.reply_skb) |
2518 | return retcode; | |
2519 | if (retcode != NO_ERROR) | |
2520 | goto fail; | |
bde89a9e | 2521 | connection = adm_ctx.connection; |
b411b363 | 2522 | |
bde89a9e | 2523 | res_opts = connection->res_opts; |
5979e361 | 2524 | if (should_set_defaults(info)) |
b966b5dd | 2525 | set_res_opts_defaults(&res_opts); |
b411b363 | 2526 | |
b57a1e27 | 2527 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2528 | if (err && err != -ENOMSG) { |
b411b363 | 2529 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2530 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2531 | goto fail; |
2532 | } | |
2533 | ||
bde89a9e | 2534 | err = set_resource_options(connection, &res_opts); |
afbbfa88 AG |
2535 | if (err) { |
2536 | retcode = ERR_INVALID_REQUEST; | |
2537 | if (err == -ENOMEM) | |
2538 | retcode = ERR_NOMEM; | |
b411b363 PR |
2539 | } |
2540 | ||
b411b363 | 2541 | fail: |
3b98c0c2 | 2542 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2543 | return 0; |
2544 | } | |
2545 | ||
3b98c0c2 | 2546 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2547 | { |
b30ab791 | 2548 | struct drbd_device *device; |
3b98c0c2 LE |
2549 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2550 | ||
2551 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2552 | if (!adm_ctx.reply_skb) | |
2553 | return retcode; | |
2554 | if (retcode != NO_ERROR) | |
2555 | goto out; | |
2556 | ||
b30ab791 | 2557 | device = adm_ctx.device; |
b411b363 | 2558 | |
194bfb32 | 2559 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2560 | * resync just being finished, wait for it before requesting a new resync. |
2561 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2562 | drbd_suspend_io(device); |
2563 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
2564 | drbd_flush_workqueue(device); | |
194bfb32 | 2565 | |
0b2dafcd PR |
2566 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2567 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2568 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2569 | */ |
b30ab791 AG |
2570 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2571 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2572 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2573 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2574 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2575 | retcode = ERR_IO_MD_DISK; | |
2576 | } | |
2577 | } else | |
b30ab791 AG |
2578 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2579 | drbd_resume_io(device); | |
b411b363 | 2580 | |
3b98c0c2 LE |
2581 | out: |
2582 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2583 | return 0; |
2584 | } | |
2585 | ||
3b98c0c2 LE |
2586 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2587 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2588 | { |
3b98c0c2 | 2589 | enum drbd_ret_code retcode; |
194bfb32 | 2590 | |
3b98c0c2 LE |
2591 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2592 | if (!adm_ctx.reply_skb) | |
2593 | return retcode; | |
2594 | if (retcode != NO_ERROR) | |
2595 | goto out; | |
b411b363 | 2596 | |
b30ab791 | 2597 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
3b98c0c2 LE |
2598 | out: |
2599 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2600 | return 0; |
2601 | } | |
2602 | ||
b30ab791 | 2603 | static int drbd_bmio_set_susp_al(struct drbd_device *device) |
0778286a PR |
2604 | { |
2605 | int rv; | |
2606 | ||
b30ab791 AG |
2607 | rv = drbd_bmio_set_n_write(device); |
2608 | drbd_suspend_al(device); | |
0778286a PR |
2609 | return rv; |
2610 | } | |
2611 | ||
3b98c0c2 | 2612 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2613 | { |
25b0d6c8 | 2614 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2615 | struct drbd_device *device; |
25b0d6c8 PR |
2616 | |
2617 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2618 | if (!adm_ctx.reply_skb) | |
2619 | return retcode; | |
2620 | if (retcode != NO_ERROR) | |
2621 | goto out; | |
2622 | ||
b30ab791 | 2623 | device = adm_ctx.device; |
b411b363 | 2624 | |
194bfb32 | 2625 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2626 | * resync just being finished, wait for it before requesting a new resync. |
2627 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2628 | drbd_suspend_io(device); |
2629 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
2630 | drbd_flush_workqueue(device); | |
194bfb32 | 2631 | |
0b2dafcd PR |
2632 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2633 | * in the bitmap. Otherwise, try to start a resync handshake | |
2634 | * as sync source for full sync. | |
2635 | */ | |
b30ab791 | 2636 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2637 | /* The peer will get a resync upon connect anyways. Just make that |
2638 | into a full resync. */ | |
b30ab791 | 2639 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2640 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2641 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2642 | "set_n_write from invalidate_peer", |
2643 | BM_LOCKED_SET_ALLOWED)) | |
2644 | retcode = ERR_IO_MD_DISK; | |
2645 | } | |
2646 | } else | |
b30ab791 AG |
2647 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2648 | drbd_resume_io(device); | |
b411b363 | 2649 | |
25b0d6c8 PR |
2650 | out: |
2651 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2652 | return 0; |
2653 | } | |
2654 | ||
3b98c0c2 | 2655 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2656 | { |
3b98c0c2 | 2657 | enum drbd_ret_code retcode; |
b411b363 | 2658 | |
3b98c0c2 LE |
2659 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2660 | if (!adm_ctx.reply_skb) | |
2661 | return retcode; | |
2662 | if (retcode != NO_ERROR) | |
2663 | goto out; | |
b411b363 | 2664 | |
b30ab791 | 2665 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 LE |
2666 | retcode = ERR_PAUSE_IS_SET; |
2667 | out: | |
2668 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2669 | return 0; |
2670 | } | |
2671 | ||
3b98c0c2 | 2672 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2673 | { |
da9fbc27 | 2674 | union drbd_dev_state s; |
3b98c0c2 LE |
2675 | enum drbd_ret_code retcode; |
2676 | ||
2677 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2678 | if (!adm_ctx.reply_skb) | |
2679 | return retcode; | |
2680 | if (retcode != NO_ERROR) | |
2681 | goto out; | |
b411b363 | 2682 | |
b30ab791 AG |
2683 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2684 | s = adm_ctx.device->state; | |
cd88d030 PR |
2685 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2686 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2687 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2688 | } else { | |
2689 | retcode = ERR_PAUSE_IS_CLEAR; | |
2690 | } | |
2691 | } | |
b411b363 | 2692 | |
3b98c0c2 LE |
2693 | out: |
2694 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2695 | return 0; |
2696 | } | |
2697 | ||
3b98c0c2 | 2698 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2699 | { |
3b98c0c2 | 2700 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2701 | } |
2702 | ||
3b98c0c2 | 2703 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2704 | { |
b30ab791 | 2705 | struct drbd_device *device; |
3b98c0c2 LE |
2706 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2707 | ||
2708 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2709 | if (!adm_ctx.reply_skb) | |
2710 | return retcode; | |
2711 | if (retcode != NO_ERROR) | |
2712 | goto out; | |
2713 | ||
b30ab791 AG |
2714 | device = adm_ctx.device; |
2715 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2716 | drbd_uuid_new_current(device); | |
2717 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2718 | } |
b30ab791 AG |
2719 | drbd_suspend_io(device); |
2720 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2721 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2722 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2723 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2724 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2725 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2726 | } |
b30ab791 | 2727 | drbd_resume_io(device); |
265be2d0 | 2728 | |
3b98c0c2 LE |
2729 | out: |
2730 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2731 | return 0; |
2732 | } | |
2733 | ||
3b98c0c2 | 2734 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2735 | { |
3b98c0c2 | 2736 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2737 | } |
2738 | ||
bde89a9e | 2739 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, struct drbd_connection *connection, unsigned vnr) |
b411b363 | 2740 | { |
543cc10b LE |
2741 | struct nlattr *nla; |
2742 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2743 | if (!nla) | |
2744 | goto nla_put_failure; | |
26ec9287 AG |
2745 | if (vnr != VOLUME_UNSPECIFIED && |
2746 | nla_put_u32(skb, T_ctx_volume, vnr)) | |
2747 | goto nla_put_failure; | |
77c556f6 | 2748 | if (nla_put_string(skb, T_ctx_resource_name, connection->resource->name)) |
26ec9287 | 2749 | goto nla_put_failure; |
bde89a9e AG |
2750 | if (connection->my_addr_len && |
2751 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
26ec9287 | 2752 | goto nla_put_failure; |
bde89a9e AG |
2753 | if (connection->peer_addr_len && |
2754 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
26ec9287 | 2755 | goto nla_put_failure; |
543cc10b LE |
2756 | nla_nest_end(skb, nla); |
2757 | return 0; | |
b411b363 | 2758 | |
543cc10b LE |
2759 | nla_put_failure: |
2760 | if (nla) | |
2761 | nla_nest_cancel(skb, nla); | |
2762 | return -EMSGSIZE; | |
2763 | } | |
b411b363 | 2764 | |
b30ab791 | 2765 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 2766 | const struct sib_info *sib) |
b411b363 | 2767 | { |
3b98c0c2 LE |
2768 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
2769 | struct nlattr *nla; | |
2770 | int got_ldev; | |
3b98c0c2 LE |
2771 | int err = 0; |
2772 | int exclude_sensitive; | |
2773 | ||
2774 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2775 | * to. So we better exclude_sensitive information. | |
2776 | * | |
2777 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2778 | * in the context of the requesting user process. Exclude sensitive | |
2779 | * information, unless current has superuser. | |
2780 | * | |
2781 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2782 | * relies on the current implementation of netlink_dump(), which | |
2783 | * executes the dump callback successively from netlink_recvmsg(), | |
2784 | * always in the context of the receiving process */ | |
2785 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2786 | ||
b30ab791 | 2787 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
2788 | |
2789 | /* We need to add connection name and volume number information still. | |
2790 | * Minor number is in drbd_genlmsghdr. */ | |
a6b32bc3 | 2791 | if (nla_put_drbd_cfg_context(skb, first_peer_device(device)->connection, device->vnr)) |
3b98c0c2 | 2792 | goto nla_put_failure; |
3b98c0c2 | 2793 | |
a6b32bc3 | 2794 | if (res_opts_to_skb(skb, &first_peer_device(device)->connection->res_opts, exclude_sensitive)) |
f399002e LE |
2795 | goto nla_put_failure; |
2796 | ||
daeda1cc | 2797 | rcu_read_lock(); |
f9eb7bf4 AG |
2798 | if (got_ldev) { |
2799 | struct disk_conf *disk_conf; | |
44ed167d | 2800 | |
b30ab791 | 2801 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
2802 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
2803 | } | |
2804 | if (!err) { | |
2805 | struct net_conf *nc; | |
2806 | ||
a6b32bc3 | 2807 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
2808 | if (nc) |
2809 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2810 | } | |
44ed167d PR |
2811 | rcu_read_unlock(); |
2812 | if (err) | |
2813 | goto nla_put_failure; | |
3b98c0c2 | 2814 | |
3b98c0c2 LE |
2815 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2816 | if (!nla) | |
2817 | goto nla_put_failure; | |
26ec9287 | 2818 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
2819 | nla_put_u32(skb, T_current_state, device->state.i) || |
2820 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
2821 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
2822 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
2823 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
2824 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
2825 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
2826 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
2827 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
2828 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
2829 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
2830 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 2831 | goto nla_put_failure; |
3b98c0c2 LE |
2832 | |
2833 | if (got_ldev) { | |
39a1aa7f | 2834 | int err; |
b411b363 | 2835 | |
b30ab791 AG |
2836 | spin_lock_irq(&device->ldev->md.uuid_lock); |
2837 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
2838 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
2839 | |
2840 | if (err) | |
2841 | goto nla_put_failure; | |
2842 | ||
b30ab791 AG |
2843 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
2844 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
2845 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 2846 | goto nla_put_failure; |
b30ab791 AG |
2847 | if (C_SYNC_SOURCE <= device->state.conn && |
2848 | C_PAUSED_SYNC_T >= device->state.conn) { | |
2849 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
2850 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 2851 | goto nla_put_failure; |
3b98c0c2 | 2852 | } |
b411b363 | 2853 | } |
b411b363 | 2854 | |
3b98c0c2 LE |
2855 | if (sib) { |
2856 | switch(sib->sib_reason) { | |
2857 | case SIB_SYNC_PROGRESS: | |
2858 | case SIB_GET_STATUS_REPLY: | |
2859 | break; | |
2860 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2861 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2862 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2863 | goto nla_put_failure; | |
3b98c0c2 LE |
2864 | break; |
2865 | case SIB_HELPER_POST: | |
26ec9287 AG |
2866 | if (nla_put_u32(skb, T_helper_exit_code, |
2867 | sib->helper_exit_code)) | |
2868 | goto nla_put_failure; | |
3b98c0c2 LE |
2869 | /* fall through */ |
2870 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2871 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2872 | goto nla_put_failure; | |
3b98c0c2 LE |
2873 | break; |
2874 | } | |
b411b363 | 2875 | } |
3b98c0c2 | 2876 | nla_nest_end(skb, nla); |
b411b363 | 2877 | |
3b98c0c2 LE |
2878 | if (0) |
2879 | nla_put_failure: | |
2880 | err = -EMSGSIZE; | |
2881 | if (got_ldev) | |
b30ab791 | 2882 | put_ldev(device); |
3b98c0c2 | 2883 | return err; |
b411b363 PR |
2884 | } |
2885 | ||
3b98c0c2 | 2886 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2887 | { |
3b98c0c2 LE |
2888 | enum drbd_ret_code retcode; |
2889 | int err; | |
b411b363 | 2890 | |
3b98c0c2 LE |
2891 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2892 | if (!adm_ctx.reply_skb) | |
2893 | return retcode; | |
2894 | if (retcode != NO_ERROR) | |
2895 | goto out; | |
b411b363 | 2896 | |
b30ab791 | 2897 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
2898 | if (err) { |
2899 | nlmsg_free(adm_ctx.reply_skb); | |
2900 | return err; | |
b411b363 | 2901 | } |
3b98c0c2 LE |
2902 | out: |
2903 | drbd_adm_finish(info, retcode); | |
2904 | return 0; | |
b411b363 PR |
2905 | } |
2906 | ||
4b7a530f | 2907 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2908 | { |
b30ab791 | 2909 | struct drbd_device *device; |
3b98c0c2 | 2910 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
2911 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
2912 | struct drbd_resource *resource = NULL; | |
2913 | struct drbd_connection *connection; | |
2914 | struct drbd_resource *tmp; | |
543cc10b LE |
2915 | unsigned volume = cb->args[1]; |
2916 | ||
2917 | /* Open coded, deferred, iteration: | |
77c556f6 AG |
2918 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
2919 | * connection = "first connection of resource"; | |
bde89a9e | 2920 | * idr_for_each_entry(&connection->volumes, device, i) { |
543cc10b LE |
2921 | * ... |
2922 | * } | |
2923 | * } | |
77c556f6 | 2924 | * where resource is cb->args[0]; |
543cc10b LE |
2925 | * and i is cb->args[1]; |
2926 | * | |
71932efc LE |
2927 | * cb->args[2] indicates if we shall loop over all resources, |
2928 | * or just dump all volumes of a single resource. | |
2929 | * | |
3b98c0c2 LE |
2930 | * This may miss entries inserted after this dump started, |
2931 | * or entries deleted before they are reached. | |
543cc10b | 2932 | * |
b30ab791 | 2933 | * We need to make sure the device won't disappear while |
543cc10b LE |
2934 | * we are looking at it, and revalidate our iterators |
2935 | * on each iteration. | |
2936 | */ | |
b411b363 | 2937 | |
05a10ec7 | 2938 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 2939 | rcu_read_lock(); |
543cc10b | 2940 | /* revalidate iterator position */ |
77c556f6 | 2941 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
2942 | if (pos == NULL) { |
2943 | /* first iteration */ | |
2944 | pos = tmp; | |
77c556f6 | 2945 | resource = pos; |
543cc10b LE |
2946 | break; |
2947 | } | |
2948 | if (tmp == pos) { | |
77c556f6 | 2949 | resource = pos; |
543cc10b LE |
2950 | break; |
2951 | } | |
b411b363 | 2952 | } |
77c556f6 AG |
2953 | if (resource) { |
2954 | next_resource: | |
2955 | connection = first_connection(resource); | |
bde89a9e | 2956 | device = idr_get_next(&connection->volumes, &volume); |
b30ab791 | 2957 | if (!device) { |
77c556f6 AG |
2958 | /* No more volumes to dump on this resource. |
2959 | * Advance resource iterator. */ | |
2960 | pos = list_entry_rcu(resource->resources.next, | |
2961 | struct drbd_resource, resources); | |
2962 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 2963 | if (volume != 0) { |
71932efc LE |
2964 | /* If we reached the end of the list, |
2965 | * or only a single resource dump was requested, | |
2966 | * we are done. */ | |
77c556f6 | 2967 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 2968 | goto out; |
543cc10b | 2969 | volume = 0; |
77c556f6 AG |
2970 | resource = pos; |
2971 | goto next_resource; | |
543cc10b LE |
2972 | } |
2973 | } | |
2974 | ||
98683650 | 2975 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
2976 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
2977 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
2978 | if (!dh) | |
543cc10b LE |
2979 | goto out; |
2980 | ||
b30ab791 | 2981 | if (!device) { |
bde89a9e | 2982 | /* This is a connection without a single volume. |
367d675d LE |
2983 | * Suprisingly enough, it may have a network |
2984 | * configuration. */ | |
2985 | struct net_conf *nc; | |
543cc10b LE |
2986 | dh->minor = -1U; |
2987 | dh->ret_code = NO_ERROR; | |
bde89a9e | 2988 | if (nla_put_drbd_cfg_context(skb, connection, VOLUME_UNSPECIFIED)) |
367d675d | 2989 | goto cancel; |
bde89a9e | 2990 | nc = rcu_dereference(connection->net_conf); |
367d675d LE |
2991 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) |
2992 | goto cancel; | |
2993 | goto done; | |
543cc10b | 2994 | } |
b411b363 | 2995 | |
b30ab791 | 2996 | D_ASSERT(device->vnr == volume); |
a6b32bc3 | 2997 | D_ASSERT(first_peer_device(device)->connection == connection); |
3b98c0c2 | 2998 | |
b30ab791 | 2999 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3000 | dh->ret_code = NO_ERROR; |
3001 | ||
b30ab791 | 3002 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3003 | cancel: |
3b98c0c2 | 3004 | genlmsg_cancel(skb, dh); |
543cc10b | 3005 | goto out; |
3b98c0c2 | 3006 | } |
367d675d | 3007 | done: |
3b98c0c2 | 3008 | genlmsg_end(skb, dh); |
bde89a9e | 3009 | } |
b411b363 | 3010 | |
543cc10b | 3011 | out: |
c141ebda | 3012 | rcu_read_unlock(); |
543cc10b | 3013 | /* where to start the next iteration */ |
bde89a9e | 3014 | cb->args[0] = (long)pos; |
77c556f6 | 3015 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3016 | |
77c556f6 | 3017 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3018 | * Which will terminate the dump. */ |
3b98c0c2 | 3019 | return skb->len; |
b411b363 PR |
3020 | } |
3021 | ||
71932efc LE |
3022 | /* |
3023 | * Request status of all resources, or of all volumes within a single resource. | |
3024 | * | |
3025 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3026 | * Which means we cannot use the family->attrbuf or other such members, because | |
3027 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3028 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3029 | * | |
3030 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3031 | */ |
71932efc | 3032 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3033 | { |
71932efc LE |
3034 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3035 | struct nlattr *nla; | |
7c3063cc | 3036 | const char *resource_name; |
bde89a9e | 3037 | struct drbd_connection *connection; |
7c3063cc | 3038 | int maxtype; |
71932efc LE |
3039 | |
3040 | /* Is this a followup call? */ | |
3041 | if (cb->args[0]) { | |
3042 | /* ... of a single resource dump, | |
3043 | * and the resource iterator has been advanced already? */ | |
3044 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3045 | return 0; /* DONE. */ | |
3046 | goto dump; | |
3047 | } | |
3048 | ||
3049 | /* First call (from netlink_dump_start). We need to figure out | |
3050 | * which resource(s) the user wants us to dump. */ | |
3051 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3052 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3053 | DRBD_NLA_CFG_CONTEXT); | |
3054 | ||
3055 | /* No explicit context given. Dump all. */ | |
3056 | if (!nla) | |
3057 | goto dump; | |
7c3063cc AG |
3058 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3059 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3060 | if (IS_ERR(nla)) | |
3061 | return PTR_ERR(nla); | |
71932efc LE |
3062 | /* context given, but no name present? */ |
3063 | if (!nla) | |
3064 | return -EINVAL; | |
7c3063cc | 3065 | resource_name = nla_data(nla); |
bde89a9e | 3066 | connection = conn_get_by_name(resource_name); |
0ace9dfa | 3067 | |
bde89a9e | 3068 | if (!connection) |
71932efc LE |
3069 | return -ENODEV; |
3070 | ||
05a10ec7 | 3071 | kref_put(&connection->kref, drbd_destroy_connection); /* get_one_status() (re)validates connection by itself */ |
0ace9dfa | 3072 | |
71932efc | 3073 | /* prime iterators, and set "filter" mode mark: |
bde89a9e AG |
3074 | * only dump this connection. */ |
3075 | cb->args[0] = (long)connection; | |
71932efc | 3076 | /* cb->args[1] = 0; passed in this way. */ |
bde89a9e | 3077 | cb->args[2] = (long)connection; |
71932efc LE |
3078 | |
3079 | dump: | |
3080 | return get_one_status(skb, cb); | |
3081 | } | |
b411b363 | 3082 | |
3b98c0c2 | 3083 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3084 | { |
3b98c0c2 LE |
3085 | enum drbd_ret_code retcode; |
3086 | struct timeout_parms tp; | |
3087 | int err; | |
b411b363 | 3088 | |
3b98c0c2 LE |
3089 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3090 | if (!adm_ctx.reply_skb) | |
3091 | return retcode; | |
3092 | if (retcode != NO_ERROR) | |
3093 | goto out; | |
b411b363 | 3094 | |
3b98c0c2 | 3095 | tp.timeout_type = |
b30ab791 AG |
3096 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3097 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3098 | UT_DEFAULT; |
b411b363 | 3099 | |
3b98c0c2 LE |
3100 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3101 | if (err) { | |
3102 | nlmsg_free(adm_ctx.reply_skb); | |
3103 | return err; | |
3104 | } | |
3105 | out: | |
3106 | drbd_adm_finish(info, retcode); | |
3107 | return 0; | |
b411b363 PR |
3108 | } |
3109 | ||
3b98c0c2 | 3110 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3111 | { |
b30ab791 | 3112 | struct drbd_device *device; |
3b98c0c2 | 3113 | enum drbd_ret_code retcode; |
58ffa580 | 3114 | struct start_ov_parms parms; |
b411b363 | 3115 | |
3b98c0c2 LE |
3116 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3117 | if (!adm_ctx.reply_skb) | |
3118 | return retcode; | |
3119 | if (retcode != NO_ERROR) | |
3120 | goto out; | |
873b0d5f | 3121 | |
b30ab791 | 3122 | device = adm_ctx.device; |
58ffa580 LE |
3123 | |
3124 | /* resume from last known position, if possible */ | |
b30ab791 | 3125 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3126 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3127 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3128 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3129 | if (err) { |
3130 | retcode = ERR_MANDATORY_TAG; | |
3131 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3132 | goto out; | |
3133 | } | |
b411b363 | 3134 | } |
58ffa580 | 3135 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3136 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3137 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3138 | |
3139 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3140 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3141 | drbd_suspend_io(device); |
3142 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3143 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3144 | drbd_resume_io(device); | |
3b98c0c2 LE |
3145 | out: |
3146 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3147 | return 0; |
3148 | } | |
3149 | ||
3150 | ||
3b98c0c2 | 3151 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3152 | { |
b30ab791 | 3153 | struct drbd_device *device; |
3b98c0c2 | 3154 | enum drbd_ret_code retcode; |
b411b363 PR |
3155 | int skip_initial_sync = 0; |
3156 | int err; | |
3b98c0c2 | 3157 | struct new_c_uuid_parms args; |
b411b363 | 3158 | |
3b98c0c2 LE |
3159 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3160 | if (!adm_ctx.reply_skb) | |
3161 | return retcode; | |
3162 | if (retcode != NO_ERROR) | |
3163 | goto out_nolock; | |
b411b363 | 3164 | |
b30ab791 | 3165 | device = adm_ctx.device; |
3b98c0c2 LE |
3166 | memset(&args, 0, sizeof(args)); |
3167 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3168 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3169 | if (err) { |
3170 | retcode = ERR_MANDATORY_TAG; | |
3171 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3172 | goto out_nolock; | |
3173 | } | |
b411b363 PR |
3174 | } |
3175 | ||
b30ab791 | 3176 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3177 | |
b30ab791 | 3178 | if (!get_ldev(device)) { |
b411b363 PR |
3179 | retcode = ERR_NO_DISK; |
3180 | goto out; | |
3181 | } | |
3182 | ||
3183 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3184 | if (device->state.conn == C_CONNECTED && |
3185 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3186 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
b411b363 PR |
3187 | dev_info(DEV, "Preparing to skip initial sync\n"); |
3188 | skip_initial_sync = 1; | |
b30ab791 | 3189 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3190 | retcode = ERR_CONNECTED; |
3191 | goto out_dec; | |
3192 | } | |
3193 | ||
b30ab791 AG |
3194 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3195 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3196 | |
3197 | if (args.clear_bm) { | |
b30ab791 | 3198 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3199 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 PR |
3200 | if (err) { |
3201 | dev_err(DEV, "Writing bitmap failed with %d\n",err); | |
3202 | retcode = ERR_IO_MD_DISK; | |
3203 | } | |
3204 | if (skip_initial_sync) { | |
b30ab791 AG |
3205 | drbd_send_uuids_skip_initial_sync(device); |
3206 | _drbd_uuid_set(device, UI_BITMAP, 0); | |
3207 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
a6b32bc3 | 3208 | spin_lock_irq(&first_peer_device(device)->connection->req_lock); |
b30ab791 | 3209 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3210 | CS_VERBOSE, NULL); |
a6b32bc3 | 3211 | spin_unlock_irq(&first_peer_device(device)->connection->req_lock); |
b411b363 PR |
3212 | } |
3213 | } | |
3214 | ||
b30ab791 | 3215 | drbd_md_sync(device); |
b411b363 | 3216 | out_dec: |
b30ab791 | 3217 | put_ldev(device); |
b411b363 | 3218 | out: |
b30ab791 | 3219 | mutex_unlock(device->state_mutex); |
3b98c0c2 LE |
3220 | out_nolock: |
3221 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3222 | return 0; |
3223 | } | |
3224 | ||
3b98c0c2 | 3225 | static enum drbd_ret_code |
7c3063cc | 3226 | drbd_check_resource_name(const char *name) |
b411b363 | 3227 | { |
3b98c0c2 | 3228 | if (!name || !name[0]) { |
7c3063cc | 3229 | drbd_msg_put_info("resource name missing"); |
3b98c0c2 | 3230 | return ERR_MANDATORY_TAG; |
b411b363 | 3231 | } |
3b98c0c2 LE |
3232 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3233 | * we must not allow slashes */ | |
3234 | if (strchr(name, '/')) { | |
7c3063cc | 3235 | drbd_msg_put_info("invalid resource name"); |
3b98c0c2 | 3236 | return ERR_INVALID_REQUEST; |
b411b363 | 3237 | } |
3b98c0c2 | 3238 | return NO_ERROR; |
774b3055 | 3239 | } |
b411b363 | 3240 | |
789c1b62 | 3241 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3242 | { |
3b98c0c2 | 3243 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3244 | struct res_opts res_opts; |
3245 | int err; | |
b411b363 | 3246 | |
3b98c0c2 LE |
3247 | retcode = drbd_adm_prepare(skb, info, 0); |
3248 | if (!adm_ctx.reply_skb) | |
3249 | return retcode; | |
3250 | if (retcode != NO_ERROR) | |
3251 | goto out; | |
b411b363 | 3252 | |
afbbfa88 AG |
3253 | set_res_opts_defaults(&res_opts); |
3254 | err = res_opts_from_attrs(&res_opts, info); | |
3255 | if (err && err != -ENOMSG) { | |
3256 | retcode = ERR_MANDATORY_TAG; | |
3257 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3258 | goto out; | |
b411b363 PR |
3259 | } |
3260 | ||
7c3063cc | 3261 | retcode = drbd_check_resource_name(adm_ctx.resource_name); |
3b98c0c2 LE |
3262 | if (retcode != NO_ERROR) |
3263 | goto out; | |
b411b363 | 3264 | |
bde89a9e | 3265 | if (adm_ctx.connection) { |
38f19616 LE |
3266 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3267 | retcode = ERR_INVALID_REQUEST; | |
789c1b62 | 3268 | drbd_msg_put_info("resource exists"); |
38f19616 LE |
3269 | } |
3270 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3271 | goto out; |
b411b363 | 3272 | } |
b411b363 | 3273 | |
afbbfa88 | 3274 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3275 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
3276 | out: |
3277 | drbd_adm_finish(info, retcode); | |
3278 | return 0; | |
b411b363 PR |
3279 | } |
3280 | ||
05a10ec7 | 3281 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3282 | { |
3b98c0c2 LE |
3283 | struct drbd_genlmsghdr *dh = info->userhdr; |
3284 | enum drbd_ret_code retcode; | |
b411b363 | 3285 | |
44e52cfa | 3286 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3287 | if (!adm_ctx.reply_skb) |
3288 | return retcode; | |
3289 | if (retcode != NO_ERROR) | |
3290 | goto out; | |
b411b363 | 3291 | |
f2257a56 | 3292 | if (dh->minor > MINORMASK) { |
3b98c0c2 LE |
3293 | drbd_msg_put_info("requested minor out of range"); |
3294 | retcode = ERR_INVALID_REQUEST; | |
3295 | goto out; | |
b411b363 | 3296 | } |
0c8e36d9 | 3297 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
3298 | drbd_msg_put_info("requested volume id out of range"); |
3299 | retcode = ERR_INVALID_REQUEST; | |
3300 | goto out; | |
b411b363 | 3301 | } |
b411b363 | 3302 | |
38f19616 | 3303 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 3304 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 3305 | if (adm_ctx.device) { |
38f19616 LE |
3306 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
3307 | retcode = ERR_MINOR_EXISTS; | |
3308 | /* else: still NO_ERROR */ | |
3309 | goto out; | |
b411b363 | 3310 | } |
38f19616 | 3311 | |
a6b32bc3 | 3312 | retcode = drbd_create_minor(adm_ctx.connection, dh->minor, adm_ctx.volume); |
3b98c0c2 LE |
3313 | out: |
3314 | drbd_adm_finish(info, retcode); | |
3315 | return 0; | |
b411b363 PR |
3316 | } |
3317 | ||
05a10ec7 | 3318 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 3319 | { |
b30ab791 AG |
3320 | if (device->state.disk == D_DISKLESS && |
3321 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
3322 | * we may want to delete a minor from a live replication group. |
3323 | */ | |
b30ab791 AG |
3324 | device->state.role == R_SECONDARY) { |
3325 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), | |
369bea63 | 3326 | CS_VERBOSE + CS_WAIT_COMPLETE); |
803ea134 | 3327 | drbd_delete_minor(device); |
85f75dd7 LE |
3328 | return NO_ERROR; |
3329 | } else | |
3330 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3331 | } |
3332 | ||
05a10ec7 | 3333 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3334 | { |
3b98c0c2 | 3335 | enum drbd_ret_code retcode; |
b411b363 | 3336 | |
3b98c0c2 LE |
3337 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3338 | if (!adm_ctx.reply_skb) | |
3339 | return retcode; | |
3340 | if (retcode != NO_ERROR) | |
3341 | goto out; | |
b411b363 | 3342 | |
05a10ec7 | 3343 | retcode = adm_del_minor(adm_ctx.device); |
85f75dd7 LE |
3344 | out: |
3345 | drbd_adm_finish(info, retcode); | |
3346 | return 0; | |
b411b363 PR |
3347 | } |
3348 | ||
85f75dd7 | 3349 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3350 | { |
f3dfa40a | 3351 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
b30ab791 | 3352 | struct drbd_device *device; |
85f75dd7 | 3353 | unsigned i; |
b411b363 | 3354 | |
a10f6b8a | 3355 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
3356 | if (!adm_ctx.reply_skb) |
3357 | return retcode; | |
3358 | if (retcode != NO_ERROR) | |
3359 | goto out; | |
b411b363 | 3360 | |
85f75dd7 | 3361 | /* demote */ |
bde89a9e | 3362 | idr_for_each_entry(&adm_ctx.connection->volumes, device, i) { |
b30ab791 | 3363 | retcode = drbd_set_role(device, R_SECONDARY, 0); |
85f75dd7 LE |
3364 | if (retcode < SS_SUCCESS) { |
3365 | drbd_msg_put_info("failed to demote"); | |
c141ebda | 3366 | goto out; |
85f75dd7 | 3367 | } |
b411b363 | 3368 | } |
b411b363 | 3369 | |
bde89a9e | 3370 | retcode = conn_try_disconnect(adm_ctx.connection, 0); |
f3dfa40a | 3371 | if (retcode < SS_SUCCESS) { |
85f75dd7 | 3372 | drbd_msg_put_info("failed to disconnect"); |
f3dfa40a | 3373 | goto out; |
85f75dd7 | 3374 | } |
b411b363 | 3375 | |
85f75dd7 | 3376 | /* detach */ |
bde89a9e | 3377 | idr_for_each_entry(&adm_ctx.connection->volumes, device, i) { |
b30ab791 | 3378 | retcode = adm_detach(device, 0); |
27012382 | 3379 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
85f75dd7 | 3380 | drbd_msg_put_info("failed to detach"); |
c141ebda | 3381 | goto out; |
85f75dd7 LE |
3382 | } |
3383 | } | |
b411b363 | 3384 | |
bde89a9e | 3385 | /* If we reach this, all volumes (of this connection) are Secondary, |
f3dfa40a | 3386 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have |
c141ebda | 3387 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
bde89a9e | 3388 | drbd_thread_stop(&adm_ctx.connection->worker); |
b411b363 | 3389 | |
f3dfa40a | 3390 | /* Now, nothing can fail anymore */ |
b411b363 | 3391 | |
85f75dd7 | 3392 | /* delete volumes */ |
bde89a9e | 3393 | idr_for_each_entry(&adm_ctx.connection->volumes, device, i) { |
05a10ec7 | 3394 | retcode = adm_del_minor(device); |
85f75dd7 LE |
3395 | if (retcode != NO_ERROR) { |
3396 | /* "can not happen" */ | |
3397 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 | 3398 | goto out; |
85f75dd7 LE |
3399 | } |
3400 | } | |
b411b363 | 3401 | |
85f75dd7 | 3402 | /* delete connection */ |
bde89a9e | 3403 | if (conn_lowest_minor(adm_ctx.connection) < 0) { |
77c556f6 AG |
3404 | struct drbd_resource *resource = adm_ctx.connection->resource; |
3405 | ||
3406 | list_del_rcu(&resource->resources); | |
ec0bddbc | 3407 | synchronize_rcu(); |
77c556f6 | 3408 | drbd_free_resource(resource); |
b411b363 | 3409 | |
85f75dd7 LE |
3410 | retcode = NO_ERROR; |
3411 | } else { | |
3412 | /* "can not happen" */ | |
789c1b62 | 3413 | retcode = ERR_RES_IN_USE; |
85f75dd7 | 3414 | drbd_msg_put_info("failed to delete connection"); |
85f75dd7 | 3415 | } |
ef356262 | 3416 | goto out; |
3b98c0c2 LE |
3417 | out: |
3418 | drbd_adm_finish(info, retcode); | |
3419 | return 0; | |
b411b363 PR |
3420 | } |
3421 | ||
789c1b62 | 3422 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3423 | { |
77c556f6 AG |
3424 | struct drbd_resource *resource; |
3425 | struct drbd_connection *connection; | |
3b98c0c2 | 3426 | enum drbd_ret_code retcode; |
b411b363 | 3427 | |
44e52cfa | 3428 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3429 | if (!adm_ctx.reply_skb) |
3430 | return retcode; | |
3431 | if (retcode != NO_ERROR) | |
3432 | goto out; | |
b411b363 | 3433 | |
77c556f6 AG |
3434 | resource = adm_ctx.resource; |
3435 | for_each_connection(connection, resource) { | |
3436 | if (connection->cstate > C_STANDALONE) { | |
3437 | retcode = ERR_NET_CONFIGURED; | |
3438 | goto out; | |
3439 | } | |
3440 | } | |
3441 | if (!idr_is_empty(&resource->devices)) { | |
789c1b62 | 3442 | retcode = ERR_RES_IN_USE; |
77c556f6 | 3443 | goto out; |
b411b363 PR |
3444 | } |
3445 | ||
77c556f6 AG |
3446 | list_del_rcu(&resource->resources); |
3447 | for_each_connection(connection, resource) | |
3448 | drbd_thread_stop(&connection->worker); | |
3449 | synchronize_rcu(); | |
3450 | drbd_free_resource(resource); | |
3451 | retcode = NO_ERROR; | |
3b98c0c2 LE |
3452 | out: |
3453 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3454 | return 0; |
3455 | } | |
3456 | ||
b30ab791 | 3457 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 3458 | { |
3b98c0c2 LE |
3459 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3460 | struct sk_buff *msg; | |
3461 | struct drbd_genlmsghdr *d_out; | |
3462 | unsigned seq; | |
3463 | int err = -ENOMEM; | |
3464 | ||
ef86b779 | 3465 | if (sib->sib_reason == SIB_SYNC_PROGRESS) { |
b30ab791 AG |
3466 | if (time_after(jiffies, device->rs_last_bcast + HZ)) |
3467 | device->rs_last_bcast = jiffies; | |
ef86b779 PR |
3468 | else |
3469 | return; | |
3470 | } | |
b411b363 | 3471 | |
3b98c0c2 LE |
3472 | seq = atomic_inc_return(&drbd_genl_seq); |
3473 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3474 | if (!msg) | |
3475 | goto failed; | |
3476 | ||
3477 | err = -EMSGSIZE; | |
3478 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3479 | if (!d_out) /* cannot happen, but anyways. */ | |
3480 | goto nla_put_failure; | |
b30ab791 | 3481 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 3482 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 3483 | |
b30ab791 | 3484 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
3485 | goto nla_put_failure; |
3486 | genlmsg_end(msg, d_out); | |
3487 | err = drbd_genl_multicast_events(msg, 0); | |
3488 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3489 | if (err && err != -ESRCH) | |
3490 | goto failed; | |
b411b363 | 3491 | |
3b98c0c2 | 3492 | return; |
b411b363 | 3493 | |
3b98c0c2 LE |
3494 | nla_put_failure: |
3495 | nlmsg_free(msg); | |
3496 | failed: | |
3497 | dev_err(DEV, "Error %d while broadcasting event. " | |
3498 | "Event seq:%u sib_reason:%u\n", | |
3499 | err, seq, sib->sib_reason); | |
b411b363 | 3500 | } |