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