]>
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" | |
265be2d0 | 35 | #include "drbd_req.h" |
b411b363 PR |
36 | #include "drbd_wrappers.h" |
37 | #include <asm/unaligned.h> | |
b411b363 | 38 | #include <linux/drbd_limits.h> |
87f7be4c | 39 | #include <linux/kthread.h> |
b411b363 | 40 | |
3b98c0c2 LE |
41 | #include <net/genetlink.h> |
42 | ||
43 | /* .doit */ | |
44 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
45 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
46 | ||
47 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info); | |
48 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info); | |
49 | ||
789c1b62 AG |
50 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
51 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 52 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
53 | |
54 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
55 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 56 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
57 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
58 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
62 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
63 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
64 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 71 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
72 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
73 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
74 | /* .dumpit */ | |
75 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
76 | ||
77 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 78 | #include "drbd_nla.h" |
3b98c0c2 LE |
79 | #include <linux/genl_magic_func.h> |
80 | ||
81 | /* used blkdev_get_by_path, to claim our meta data device(s) */ | |
b411b363 PR |
82 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
83 | ||
3b98c0c2 LE |
84 | /* Configuration is strictly serialized, because generic netlink message |
85 | * processing is strictly serialized by the genl_lock(). | |
86 | * Which means we can use one static global drbd_config_context struct. | |
87 | */ | |
88 | static struct drbd_config_context { | |
89 | /* assigned from drbd_genlmsghdr */ | |
90 | unsigned int minor; | |
91 | /* assigned from request attributes, if present */ | |
92 | unsigned int volume; | |
93 | #define VOLUME_UNSPECIFIED (-1U) | |
94 | /* pointer into the request skb, | |
95 | * limited lifetime! */ | |
7c3063cc | 96 | char *resource_name; |
089c075d AG |
97 | struct nlattr *my_addr; |
98 | struct nlattr *peer_addr; | |
3b98c0c2 LE |
99 | |
100 | /* reply buffer */ | |
101 | struct sk_buff *reply_skb; | |
102 | /* pointer into reply buffer */ | |
103 | struct drbd_genlmsghdr *reply_dh; | |
104 | /* resolved from attributes, if possible */ | |
105 | struct drbd_conf *mdev; | |
106 | struct drbd_tconn *tconn; | |
107 | } adm_ctx; | |
108 | ||
109 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) | |
110 | { | |
111 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
112 | if (genlmsg_reply(skb, info)) | |
113 | printk(KERN_ERR "drbd: error sending genl reply\n"); | |
b411b363 | 114 | } |
3b98c0c2 LE |
115 | |
116 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
117 | * reason it could fail was no space in skb, and there are 4k available. */ | |
8432b314 | 118 | int drbd_msg_put_info(const char *info) |
3b98c0c2 LE |
119 | { |
120 | struct sk_buff *skb = adm_ctx.reply_skb; | |
121 | struct nlattr *nla; | |
122 | int err = -EMSGSIZE; | |
123 | ||
124 | if (!info || !info[0]) | |
125 | return 0; | |
126 | ||
127 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
128 | if (!nla) | |
129 | return err; | |
130 | ||
131 | err = nla_put_string(skb, T_info_text, info); | |
132 | if (err) { | |
133 | nla_nest_cancel(skb, nla); | |
134 | return err; | |
135 | } else | |
136 | nla_nest_end(skb, nla); | |
137 | return 0; | |
b411b363 PR |
138 | } |
139 | ||
3b98c0c2 LE |
140 | /* This would be a good candidate for a "pre_doit" hook, |
141 | * and per-family private info->pointers. | |
142 | * But we need to stay compatible with older kernels. | |
143 | * If it returns successfully, adm_ctx members are valid. | |
144 | */ | |
145 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 146 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 147 | #define DRBD_ADM_NEED_CONNECTION 4 |
3b98c0c2 LE |
148 | static int drbd_adm_prepare(struct sk_buff *skb, struct genl_info *info, |
149 | unsigned flags) | |
150 | { | |
151 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
152 | const u8 cmd = info->genlhdr->cmd; | |
153 | int err; | |
154 | ||
155 | memset(&adm_ctx, 0, sizeof(adm_ctx)); | |
156 | ||
157 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
158 | if (cmd != DRBD_ADM_GET_STATUS | |
159 | && security_netlink_recv(skb, CAP_SYS_ADMIN)) | |
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 && | |
206 | nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.tconn->my_addr)) || | |
207 | (adm_ctx.peer_addr && | |
208 | nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.tconn->peer_addr))) { | |
209 | err = -EINVAL; | |
210 | goto fail; | |
211 | } | |
212 | } | |
3b98c0c2 LE |
213 | |
214 | adm_ctx.minor = d_in->minor; | |
215 | adm_ctx.mdev = minor_to_mdev(d_in->minor); | |
7c3063cc | 216 | adm_ctx.tconn = conn_get_by_name(adm_ctx.resource_name); |
3b98c0c2 LE |
217 | |
218 | if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) { | |
219 | drbd_msg_put_info("unknown minor"); | |
220 | return ERR_MINOR_INVALID; | |
221 | } | |
44e52cfa AG |
222 | if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_RESOURCE)) { |
223 | drbd_msg_put_info("unknown resource"); | |
3b98c0c2 LE |
224 | return ERR_INVALID_REQUEST; |
225 | } | |
226 | ||
089c075d AG |
227 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
228 | if (adm_ctx.tconn && !(flags & DRBD_ADM_NEED_RESOURCE)) { | |
229 | drbd_msg_put_info("no resource name expected"); | |
230 | return ERR_INVALID_REQUEST; | |
231 | } | |
232 | if (adm_ctx.mdev) { | |
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) | |
237 | adm_ctx.tconn = conn_get_by_addrs(nla_data(adm_ctx.my_addr), | |
238 | nla_len(adm_ctx.my_addr), | |
239 | nla_data(adm_ctx.peer_addr), | |
240 | nla_len(adm_ctx.peer_addr)); | |
241 | if (!adm_ctx.tconn) { | |
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 */ |
527f4b24 LE |
248 | if (adm_ctx.mdev && adm_ctx.tconn && |
249 | adm_ctx.mdev->tconn != adm_ctx.tconn) { | |
44e52cfa | 250 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to connection %s\n", |
7c3063cc AG |
251 | adm_ctx.minor, adm_ctx.resource_name, |
252 | adm_ctx.mdev->tconn->name); | |
44e52cfa | 253 | drbd_msg_put_info("minor exists in different resource"); |
527f4b24 LE |
254 | return ERR_INVALID_REQUEST; |
255 | } | |
3b98c0c2 LE |
256 | if (adm_ctx.mdev && |
257 | adm_ctx.volume != VOLUME_UNSPECIFIED && | |
258 | adm_ctx.volume != adm_ctx.mdev->vnr) { | |
259 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", | |
260 | adm_ctx.minor, adm_ctx.volume, | |
261 | adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->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 | { | |
0ace9dfa PR |
276 | if (adm_ctx.tconn) { |
277 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
278 | adm_ctx.tconn = NULL; | |
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 | |
6b75dced | 289 | static void setup_khelper_env(struct drbd_tconn *tconn, char **envp) |
b411b363 | 290 | { |
6b75dced | 291 | char *afs; |
b411b363 | 292 | |
089c075d AG |
293 | /* FIXME: A future version will not allow this case. */ |
294 | if (tconn->my_addr_len == 0 || tconn->peer_addr_len == 0) | |
295 | return; | |
296 | ||
297 | switch (((struct sockaddr *)&tconn->peer_addr)->sa_family) { | |
298 | case AF_INET6: | |
299 | afs = "ipv6"; | |
300 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
301 | &((struct sockaddr_in6 *)&tconn->peer_addr)->sin6_addr); | |
302 | break; | |
303 | case AF_INET: | |
304 | afs = "ipv4"; | |
305 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
306 | &((struct sockaddr_in *)&tconn->peer_addr)->sin_addr); | |
307 | break; | |
308 | default: | |
309 | afs = "ssocks"; | |
310 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
311 | &((struct sockaddr_in *)&tconn->peer_addr)->sin_addr); | |
b411b363 | 312 | } |
089c075d | 313 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced PR |
314 | } |
315 | ||
316 | int drbd_khelper(struct drbd_conf *mdev, char *cmd) | |
317 | { | |
318 | char *envp[] = { "HOME=/", | |
319 | "TERM=linux", | |
320 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
321 | (char[20]) { }, /* address family */ | |
322 | (char[60]) { }, /* address */ | |
323 | NULL }; | |
324 | char mb[12]; | |
325 | char *argv[] = {usermode_helper, cmd, mb, NULL }; | |
326 | struct sib_info sib; | |
327 | int ret; | |
328 | ||
329 | snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev)); | |
330 | setup_khelper_env(mdev->tconn, envp); | |
b411b363 | 331 | |
1090c056 LE |
332 | /* The helper may take some time. |
333 | * write out any unsynced meta data changes now */ | |
334 | drbd_md_sync(mdev); | |
335 | ||
b411b363 | 336 | dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
337 | sib.sib_reason = SIB_HELPER_PRE; |
338 | sib.helper_name = cmd; | |
339 | drbd_bcast_event(mdev, &sib); | |
b411b363 PR |
340 | ret = call_usermodehelper(usermode_helper, argv, envp, 1); |
341 | if (ret) | |
342 | dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
343 | usermode_helper, cmd, mb, | |
344 | (ret >> 8) & 0xff, ret); | |
345 | else | |
346 | dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
347 | usermode_helper, cmd, mb, | |
348 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
349 | sib.sib_reason = SIB_HELPER_POST; |
350 | sib.helper_exit_code = ret; | |
351 | drbd_bcast_event(mdev, &sib); | |
b411b363 PR |
352 | |
353 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
354 | ret = 0; | |
355 | ||
356 | return ret; | |
357 | } | |
358 | ||
6b75dced PR |
359 | static void conn_md_sync(struct drbd_tconn *tconn) |
360 | { | |
361 | struct drbd_conf *mdev; | |
e90285e0 | 362 | int vnr; |
6b75dced | 363 | |
c141ebda PR |
364 | rcu_read_lock(); |
365 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { | |
366 | kref_get(&mdev->kref); | |
367 | rcu_read_unlock(); | |
6b75dced | 368 | drbd_md_sync(mdev); |
c141ebda PR |
369 | kref_put(&mdev->kref, &drbd_minor_destroy); |
370 | rcu_read_lock(); | |
371 | } | |
372 | rcu_read_unlock(); | |
6b75dced PR |
373 | } |
374 | ||
375 | int conn_khelper(struct drbd_tconn *tconn, char *cmd) | |
376 | { | |
377 | char *envp[] = { "HOME=/", | |
378 | "TERM=linux", | |
379 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
380 | (char[20]) { }, /* address family */ | |
381 | (char[60]) { }, /* address */ | |
382 | NULL }; | |
383 | char *argv[] = {usermode_helper, cmd, tconn->name, NULL }; | |
384 | int ret; | |
385 | ||
386 | setup_khelper_env(tconn, envp); | |
387 | conn_md_sync(tconn); | |
388 | ||
389 | conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name); | |
390 | /* TODO: conn_bcast_event() ?? */ | |
391 | ||
392 | ret = call_usermodehelper(usermode_helper, argv, envp, 1); | |
393 | if (ret) | |
394 | conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
395 | usermode_helper, cmd, tconn->name, | |
396 | (ret >> 8) & 0xff, ret); | |
397 | else | |
398 | conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
399 | usermode_helper, cmd, tconn->name, | |
400 | (ret >> 8) & 0xff, ret); | |
401 | /* TODO: conn_bcast_event() ?? */ | |
402 | ||
403 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
404 | ret = 0; | |
405 | ||
406 | return ret; | |
407 | } | |
408 | ||
cb703454 | 409 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn) |
b411b363 | 410 | { |
cb703454 PR |
411 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
412 | struct drbd_conf *mdev; | |
413 | int vnr; | |
414 | ||
695d08fa | 415 | rcu_read_lock(); |
cb703454 PR |
416 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
417 | if (get_ldev_if_state(mdev, D_CONSISTENT)) { | |
daeda1cc PR |
418 | fp = max_t(enum drbd_fencing_p, fp, |
419 | rcu_dereference(mdev->ldev->disk_conf)->fencing); | |
cb703454 PR |
420 | put_ldev(mdev); |
421 | } | |
422 | } | |
695d08fa | 423 | rcu_read_unlock(); |
cb703454 PR |
424 | |
425 | return fp; | |
426 | } | |
427 | ||
428 | bool conn_try_outdate_peer(struct drbd_tconn *tconn) | |
429 | { | |
430 | union drbd_state mask = { }; | |
431 | union drbd_state val = { }; | |
432 | enum drbd_fencing_p fp; | |
b411b363 PR |
433 | char *ex_to_string; |
434 | int r; | |
b411b363 | 435 | |
cb703454 PR |
436 | if (tconn->cstate >= C_WF_REPORT_PARAMS) { |
437 | conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n"); | |
438 | return false; | |
439 | } | |
b411b363 | 440 | |
cb703454 PR |
441 | fp = highest_fencing_policy(tconn); |
442 | switch (fp) { | |
443 | case FP_NOT_AVAIL: | |
444 | conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n"); | |
fb22c402 | 445 | goto out; |
cb703454 PR |
446 | case FP_DONT_CARE: |
447 | return true; | |
448 | default: ; | |
b411b363 PR |
449 | } |
450 | ||
cb703454 | 451 | r = conn_khelper(tconn, "fence-peer"); |
b411b363 PR |
452 | |
453 | switch ((r>>8) & 0xff) { | |
454 | case 3: /* peer is inconsistent */ | |
455 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
456 | mask.pdsk = D_MASK; |
457 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
458 | break; |
459 | case 4: /* peer got outdated, or was already outdated */ | |
460 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
461 | mask.pdsk = D_MASK; |
462 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
463 | break; |
464 | case 5: /* peer was down */ | |
cb703454 | 465 | if (conn_highest_disk(tconn) == D_UP_TO_DATE) { |
b411b363 PR |
466 | /* we will(have) create(d) a new UUID anyways... */ |
467 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
468 | mask.pdsk = D_MASK; |
469 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
470 | } else { |
471 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
472 | } |
473 | break; | |
474 | case 6: /* Peer is primary, voluntarily outdate myself. | |
475 | * This is useful when an unconnected R_SECONDARY is asked to | |
476 | * become R_PRIMARY, but finds the other peer being active. */ | |
477 | ex_to_string = "peer is active"; | |
cb703454 PR |
478 | conn_warn(tconn, "Peer is primary, outdating myself.\n"); |
479 | mask.disk = D_MASK; | |
480 | val.disk = D_OUTDATED; | |
b411b363 PR |
481 | break; |
482 | case 7: | |
483 | if (fp != FP_STONITH) | |
cb703454 | 484 | conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 485 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
486 | mask.pdsk = D_MASK; |
487 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
488 | break; |
489 | default: | |
490 | /* The script is broken ... */ | |
cb703454 PR |
491 | conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
492 | return false; /* Eventually leave IO frozen */ | |
b411b363 PR |
493 | } |
494 | ||
cb703454 PR |
495 | conn_info(tconn, "fence-peer helper returned %d (%s)\n", |
496 | (r>>8) & 0xff, ex_to_string); | |
fb22c402 | 497 | |
cb703454 | 498 | out: |
fb22c402 | 499 | |
cb703454 PR |
500 | /* Not using |
501 | conn_request_state(tconn, mask, val, CS_VERBOSE); | |
502 | here, because we might were able to re-establish the connection in the | |
503 | meantime. */ | |
504 | spin_lock_irq(&tconn->req_lock); | |
a1096a6e | 505 | if (tconn->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &tconn->flags)) |
cb703454 PR |
506 | _conn_request_state(tconn, mask, val, CS_VERBOSE); |
507 | spin_unlock_irq(&tconn->req_lock); | |
508 | ||
509 | return conn_highest_pdsk(tconn) <= D_OUTDATED; | |
b411b363 PR |
510 | } |
511 | ||
87f7be4c PR |
512 | static int _try_outdate_peer_async(void *data) |
513 | { | |
cb703454 | 514 | struct drbd_tconn *tconn = (struct drbd_tconn *)data; |
87f7be4c | 515 | |
cb703454 | 516 | conn_try_outdate_peer(tconn); |
87f7be4c | 517 | |
9dc9fbb3 | 518 | kref_put(&tconn->kref, &conn_destroy); |
87f7be4c PR |
519 | return 0; |
520 | } | |
521 | ||
cb703454 | 522 | void conn_try_outdate_peer_async(struct drbd_tconn *tconn) |
87f7be4c PR |
523 | { |
524 | struct task_struct *opa; | |
525 | ||
9dc9fbb3 | 526 | kref_get(&tconn->kref); |
cb703454 | 527 | opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h"); |
9dc9fbb3 | 528 | if (IS_ERR(opa)) { |
cb703454 | 529 | conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n"); |
9dc9fbb3 PR |
530 | kref_put(&tconn->kref, &conn_destroy); |
531 | } | |
87f7be4c | 532 | } |
b411b363 | 533 | |
bf885f8a AG |
534 | enum drbd_state_rv |
535 | drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) | |
b411b363 PR |
536 | { |
537 | const int max_tries = 4; | |
bf885f8a | 538 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 539 | struct net_conf *nc; |
b411b363 PR |
540 | int try = 0; |
541 | int forced = 0; | |
542 | union drbd_state mask, val; | |
b411b363 PR |
543 | |
544 | if (new_role == R_PRIMARY) | |
0625ac19 | 545 | request_ping(mdev->tconn); /* Detect a dead peer ASAP */ |
b411b363 | 546 | |
8410da8f | 547 | mutex_lock(mdev->state_mutex); |
b411b363 PR |
548 | |
549 | mask.i = 0; mask.role = R_MASK; | |
550 | val.i = 0; val.role = new_role; | |
551 | ||
552 | while (try++ < max_tries) { | |
bf885f8a | 553 | rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
554 | |
555 | /* in case we first succeeded to outdate, | |
556 | * but now suddenly could establish a connection */ | |
bf885f8a | 557 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
558 | val.pdsk = 0; |
559 | mask.pdsk = 0; | |
560 | continue; | |
561 | } | |
562 | ||
bf885f8a | 563 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
d10a33c6 PR |
564 | (mdev->state.disk < D_UP_TO_DATE && |
565 | mdev->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
566 | mask.disk = D_MASK; |
567 | val.disk = D_UP_TO_DATE; | |
568 | forced = 1; | |
569 | continue; | |
570 | } | |
571 | ||
bf885f8a | 572 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b411b363 PR |
573 | mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
574 | D_ASSERT(mdev->state.pdsk == D_UNKNOWN); | |
b411b363 | 575 | |
cb703454 | 576 | if (conn_try_outdate_peer(mdev->tconn)) { |
b411b363 PR |
577 | val.disk = D_UP_TO_DATE; |
578 | mask.disk = D_MASK; | |
579 | } | |
b411b363 PR |
580 | continue; |
581 | } | |
582 | ||
bf885f8a | 583 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 584 | goto out; |
bf885f8a | 585 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
cb703454 | 586 | if (!conn_try_outdate_peer(mdev->tconn) && force) { |
b411b363 | 587 | dev_warn(DEV, "Forced into split brain situation!\n"); |
cb703454 PR |
588 | mask.pdsk = D_MASK; |
589 | val.pdsk = D_OUTDATED; | |
b411b363 | 590 | |
cb703454 | 591 | } |
b411b363 PR |
592 | continue; |
593 | } | |
bf885f8a | 594 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
595 | /* Maybe the peer is detected as dead very soon... |
596 | retry at most once more in this case. */ | |
44ed167d PR |
597 | int timeo; |
598 | rcu_read_lock(); | |
599 | nc = rcu_dereference(mdev->tconn->net_conf); | |
600 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; | |
601 | rcu_read_unlock(); | |
602 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
603 | if (try < max_tries) |
604 | try = max_tries - 1; | |
605 | continue; | |
606 | } | |
bf885f8a AG |
607 | if (rv < SS_SUCCESS) { |
608 | rv = _drbd_request_state(mdev, mask, val, | |
b411b363 | 609 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 610 | if (rv < SS_SUCCESS) |
3b98c0c2 | 611 | goto out; |
b411b363 PR |
612 | } |
613 | break; | |
614 | } | |
615 | ||
bf885f8a | 616 | if (rv < SS_SUCCESS) |
3b98c0c2 | 617 | goto out; |
b411b363 PR |
618 | |
619 | if (forced) | |
620 | dev_warn(DEV, "Forced to consider local data as UpToDate!\n"); | |
621 | ||
622 | /* Wait until nothing is on the fly :) */ | |
623 | wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0); | |
624 | ||
625 | if (new_role == R_SECONDARY) { | |
81e84650 | 626 | set_disk_ro(mdev->vdisk, true); |
b411b363 PR |
627 | if (get_ldev(mdev)) { |
628 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
629 | put_ldev(mdev); | |
630 | } | |
631 | } else { | |
a0095508 | 632 | mutex_lock(&mdev->tconn->conf_update); |
91fd4dad | 633 | nc = mdev->tconn->net_conf; |
44ed167d | 634 | if (nc) |
6139f60d | 635 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
a0095508 | 636 | mutex_unlock(&mdev->tconn->conf_update); |
91fd4dad | 637 | |
81e84650 | 638 | set_disk_ro(mdev->vdisk, false); |
b411b363 PR |
639 | if (get_ldev(mdev)) { |
640 | if (((mdev->state.conn < C_CONNECTED || | |
641 | mdev->state.pdsk <= D_FAILED) | |
642 | && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
643 | drbd_uuid_new_current(mdev); | |
644 | ||
645 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
646 | put_ldev(mdev); | |
647 | } | |
648 | } | |
649 | ||
19f843aa LE |
650 | /* writeout of activity log covered areas of the bitmap |
651 | * to stable storage done in after state change already */ | |
b411b363 PR |
652 | |
653 | if (mdev->state.conn >= C_WF_REPORT_PARAMS) { | |
654 | /* if this was forced, we should consider sync */ | |
655 | if (forced) | |
656 | drbd_send_uuids(mdev); | |
43de7c85 | 657 | drbd_send_current_state(mdev); |
b411b363 PR |
658 | } |
659 | ||
660 | drbd_md_sync(mdev); | |
661 | ||
662 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
3b98c0c2 | 663 | out: |
8410da8f | 664 | mutex_unlock(mdev->state_mutex); |
bf885f8a | 665 | return rv; |
b411b363 PR |
666 | } |
667 | ||
3b98c0c2 | 668 | static const char *from_attrs_err_to_txt(int err) |
b411b363 | 669 | { |
3b98c0c2 LE |
670 | return err == -ENOMSG ? "required attribute missing" : |
671 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 672 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 673 | "invalid attribute value"; |
b411b363 PR |
674 | } |
675 | ||
3b98c0c2 | 676 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 677 | { |
3b98c0c2 LE |
678 | struct set_role_parms parms; |
679 | int err; | |
680 | enum drbd_ret_code retcode; | |
b411b363 | 681 | |
3b98c0c2 LE |
682 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
683 | if (!adm_ctx.reply_skb) | |
684 | return retcode; | |
685 | if (retcode != NO_ERROR) | |
686 | goto out; | |
687 | ||
688 | memset(&parms, 0, sizeof(parms)); | |
689 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 690 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
691 | if (err) { |
692 | retcode = ERR_MANDATORY_TAG; | |
693 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
694 | goto out; | |
695 | } | |
696 | } | |
697 | ||
698 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) | |
699 | retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate); | |
700 | else | |
701 | retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0); | |
702 | out: | |
703 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
704 | return 0; |
705 | } | |
706 | ||
707 | /* initializes the md.*_offset members, so we are able to find | |
708 | * the on disk meta data */ | |
709 | static void drbd_md_set_sector_offsets(struct drbd_conf *mdev, | |
710 | struct drbd_backing_dev *bdev) | |
711 | { | |
712 | sector_t md_size_sect = 0; | |
daeda1cc PR |
713 | int meta_dev_idx; |
714 | ||
715 | rcu_read_lock(); | |
716 | meta_dev_idx = rcu_dereference(bdev->disk_conf)->meta_dev_idx; | |
717 | ||
718 | switch (meta_dev_idx) { | |
b411b363 PR |
719 | default: |
720 | /* v07 style fixed size indexed meta data */ | |
721 | bdev->md.md_size_sect = MD_RESERVED_SECT; | |
722 | bdev->md.md_offset = drbd_md_ss__(mdev, bdev); | |
723 | bdev->md.al_offset = MD_AL_OFFSET; | |
724 | bdev->md.bm_offset = MD_BM_OFFSET; | |
725 | break; | |
726 | case DRBD_MD_INDEX_FLEX_EXT: | |
727 | /* just occupy the full device; unit: sectors */ | |
728 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
729 | bdev->md.md_offset = 0; | |
730 | bdev->md.al_offset = MD_AL_OFFSET; | |
731 | bdev->md.bm_offset = MD_BM_OFFSET; | |
732 | break; | |
733 | case DRBD_MD_INDEX_INTERNAL: | |
734 | case DRBD_MD_INDEX_FLEX_INT: | |
735 | bdev->md.md_offset = drbd_md_ss__(mdev, bdev); | |
736 | /* al size is still fixed */ | |
7ad651b5 | 737 | bdev->md.al_offset = -MD_AL_SECTORS; |
b411b363 PR |
738 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
739 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
740 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
741 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
742 | md_size_sect = ALIGN(md_size_sect, 8); | |
743 | ||
744 | /* plus the "drbd meta data super block", | |
745 | * and the activity log; */ | |
746 | md_size_sect += MD_BM_OFFSET; | |
747 | ||
748 | bdev->md.md_size_sect = md_size_sect; | |
749 | /* bitmap offset is adjusted by 'super' block size */ | |
750 | bdev->md.bm_offset = -md_size_sect + MD_AL_OFFSET; | |
751 | break; | |
752 | } | |
daeda1cc | 753 | rcu_read_unlock(); |
b411b363 PR |
754 | } |
755 | ||
4b0715f0 | 756 | /* input size is expected to be in KB */ |
b411b363 PR |
757 | char *ppsize(char *buf, unsigned long long size) |
758 | { | |
4b0715f0 LE |
759 | /* Needs 9 bytes at max including trailing NUL: |
760 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
761 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
762 | int base = 0; | |
4b0715f0 | 763 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
764 | /* shift + round */ |
765 | size = (size >> 10) + !!(size & (1<<9)); | |
766 | base++; | |
767 | } | |
4b0715f0 | 768 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
769 | |
770 | return buf; | |
771 | } | |
772 | ||
773 | /* there is still a theoretical deadlock when called from receiver | |
774 | * on an D_INCONSISTENT R_PRIMARY: | |
775 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
776 | * packet from remote to dec_ap_bio again. | |
777 | * receiver receive_sizes(), comes here, | |
778 | * waits for ap_bio_cnt == 0. -> deadlock. | |
779 | * but this cannot happen, actually, because: | |
780 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
781 | * (not connected, or bad/no disk on peer): | |
782 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
783 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
784 | * peer may not initiate a resize. | |
785 | */ | |
3b98c0c2 LE |
786 | /* Note these are not to be confused with |
787 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
788 | * which are (sub) state changes triggered by admin (drbdsetup), | |
789 | * and can be long lived. | |
790 | * This changes an mdev->flag, is triggered by drbd internals, | |
791 | * and should be short-lived. */ | |
b411b363 PR |
792 | void drbd_suspend_io(struct drbd_conf *mdev) |
793 | { | |
794 | set_bit(SUSPEND_IO, &mdev->flags); | |
2aebfabb | 795 | if (drbd_suspended(mdev)) |
265be2d0 | 796 | return; |
b411b363 PR |
797 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); |
798 | } | |
799 | ||
800 | void drbd_resume_io(struct drbd_conf *mdev) | |
801 | { | |
802 | clear_bit(SUSPEND_IO, &mdev->flags); | |
803 | wake_up(&mdev->misc_wait); | |
804 | } | |
805 | ||
806 | /** | |
807 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
808 | * @mdev: DRBD device. | |
809 | * | |
810 | * Returns 0 on success, negative return values indicate errors. | |
811 | * You should call drbd_md_sync() after calling this function. | |
812 | */ | |
24c4830c | 813 | enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local) |
b411b363 PR |
814 | { |
815 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
ef5e44a6 | 816 | sector_t la_size, u_size; |
b411b363 PR |
817 | sector_t size; |
818 | char ppb[10]; | |
819 | ||
820 | int md_moved, la_size_changed; | |
821 | enum determine_dev_size rv = unchanged; | |
822 | ||
823 | /* race: | |
824 | * application request passes inc_ap_bio, | |
825 | * but then cannot get an AL-reference. | |
826 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
827 | * | |
828 | * to avoid that: | |
829 | * Suspend IO right here. | |
830 | * still lock the act_log to not trigger ASSERTs there. | |
831 | */ | |
832 | drbd_suspend_io(mdev); | |
833 | ||
834 | /* no wait necessary anymore, actually we could assert that */ | |
835 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); | |
836 | ||
837 | prev_first_sect = drbd_md_first_sector(mdev->ldev); | |
838 | prev_size = mdev->ldev->md.md_size_sect; | |
839 | la_size = mdev->ldev->md.la_size_sect; | |
840 | ||
841 | /* TODO: should only be some assert here, not (re)init... */ | |
842 | drbd_md_set_sector_offsets(mdev, mdev->ldev); | |
843 | ||
daeda1cc PR |
844 | rcu_read_lock(); |
845 | u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size; | |
846 | rcu_read_unlock(); | |
ef5e44a6 | 847 | size = drbd_new_dev_size(mdev, mdev->ldev, u_size, flags & DDSF_FORCED); |
b411b363 PR |
848 | |
849 | if (drbd_get_capacity(mdev->this_bdev) != size || | |
850 | drbd_bm_capacity(mdev) != size) { | |
851 | int err; | |
02d9a94b | 852 | err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
853 | if (unlikely(err)) { |
854 | /* currently there is only one error: ENOMEM! */ | |
855 | size = drbd_bm_capacity(mdev)>>1; | |
856 | if (size == 0) { | |
857 | dev_err(DEV, "OUT OF MEMORY! " | |
858 | "Could not allocate bitmap!\n"); | |
859 | } else { | |
860 | dev_err(DEV, "BM resizing failed. " | |
861 | "Leaving size unchanged at size = %lu KB\n", | |
862 | (unsigned long)size); | |
863 | } | |
864 | rv = dev_size_error; | |
865 | } | |
866 | /* racy, see comments above. */ | |
867 | drbd_set_my_capacity(mdev, size); | |
868 | mdev->ldev->md.la_size_sect = size; | |
869 | dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), | |
870 | (unsigned long long)size>>1); | |
871 | } | |
872 | if (rv == dev_size_error) | |
873 | goto out; | |
874 | ||
875 | la_size_changed = (la_size != mdev->ldev->md.la_size_sect); | |
876 | ||
877 | md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev) | |
878 | || prev_size != mdev->ldev->md.md_size_sect; | |
879 | ||
880 | if (la_size_changed || md_moved) { | |
24dccabb AG |
881 | int err; |
882 | ||
b411b363 PR |
883 | drbd_al_shrink(mdev); /* All extents inactive. */ |
884 | dev_info(DEV, "Writing the whole bitmap, %s\n", | |
885 | la_size_changed && md_moved ? "size changed and md moved" : | |
886 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 LE |
887 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
888 | err = drbd_bitmap_io(mdev, &drbd_bm_write, | |
889 | "size changed", BM_LOCKED_MASK); | |
24dccabb AG |
890 | if (err) { |
891 | rv = dev_size_error; | |
892 | goto out; | |
893 | } | |
b411b363 PR |
894 | drbd_md_mark_dirty(mdev); |
895 | } | |
896 | ||
897 | if (size > la_size) | |
898 | rv = grew; | |
899 | if (size < la_size) | |
900 | rv = shrunk; | |
901 | out: | |
902 | lc_unlock(mdev->act_log); | |
903 | wake_up(&mdev->al_wait); | |
904 | drbd_resume_io(mdev); | |
905 | ||
906 | return rv; | |
907 | } | |
908 | ||
909 | sector_t | |
ef5e44a6 PR |
910 | drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, |
911 | sector_t u_size, int assume_peer_has_space) | |
b411b363 PR |
912 | { |
913 | sector_t p_size = mdev->p_size; /* partner's disk size. */ | |
914 | sector_t la_size = bdev->md.la_size_sect; /* last agreed size. */ | |
915 | sector_t m_size; /* my size */ | |
b411b363 PR |
916 | sector_t size = 0; |
917 | ||
918 | m_size = drbd_get_max_capacity(bdev); | |
919 | ||
a393db6f PR |
920 | if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) { |
921 | dev_warn(DEV, "Resize while not connected was forced by the user!\n"); | |
922 | p_size = m_size; | |
923 | } | |
924 | ||
b411b363 PR |
925 | if (p_size && m_size) { |
926 | size = min_t(sector_t, p_size, m_size); | |
927 | } else { | |
928 | if (la_size) { | |
929 | size = la_size; | |
930 | if (m_size && m_size < size) | |
931 | size = m_size; | |
932 | if (p_size && p_size < size) | |
933 | size = p_size; | |
934 | } else { | |
935 | if (m_size) | |
936 | size = m_size; | |
937 | if (p_size) | |
938 | size = p_size; | |
939 | } | |
940 | } | |
941 | ||
942 | if (size == 0) | |
943 | dev_err(DEV, "Both nodes diskless!\n"); | |
944 | ||
945 | if (u_size) { | |
946 | if (u_size > size) | |
947 | dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n", | |
948 | (unsigned long)u_size>>1, (unsigned long)size>>1); | |
949 | else | |
950 | size = u_size; | |
951 | } | |
952 | ||
953 | return size; | |
954 | } | |
955 | ||
956 | /** | |
957 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
958 | * @mdev: DRBD device. | |
959 | * | |
960 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
961 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
962 | * this function. | |
963 | */ | |
f399002e | 964 | static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc) |
b411b363 PR |
965 | { |
966 | struct lru_cache *n, *t; | |
967 | struct lc_element *e; | |
968 | unsigned int in_use; | |
969 | int i; | |
970 | ||
b411b363 | 971 | if (mdev->act_log && |
f399002e | 972 | mdev->act_log->nr_elements == dc->al_extents) |
b411b363 PR |
973 | return 0; |
974 | ||
975 | in_use = 0; | |
976 | t = mdev->act_log; | |
7ad651b5 | 977 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 978 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
979 | |
980 | if (n == NULL) { | |
981 | dev_err(DEV, "Cannot allocate act_log lru!\n"); | |
982 | return -ENOMEM; | |
983 | } | |
984 | spin_lock_irq(&mdev->al_lock); | |
985 | if (t) { | |
986 | for (i = 0; i < t->nr_elements; i++) { | |
987 | e = lc_element_by_index(t, i); | |
988 | if (e->refcnt) | |
989 | dev_err(DEV, "refcnt(%d)==%d\n", | |
990 | e->lc_number, e->refcnt); | |
991 | in_use += e->refcnt; | |
992 | } | |
993 | } | |
994 | if (!in_use) | |
995 | mdev->act_log = n; | |
996 | spin_unlock_irq(&mdev->al_lock); | |
997 | if (in_use) { | |
998 | dev_err(DEV, "Activity log still in use!\n"); | |
999 | lc_destroy(n); | |
1000 | return -EBUSY; | |
1001 | } else { | |
1002 | if (t) | |
1003 | lc_destroy(t); | |
1004 | } | |
1005 | drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */ | |
1006 | return 0; | |
1007 | } | |
1008 | ||
99432fcc | 1009 | static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size) |
b411b363 PR |
1010 | { |
1011 | struct request_queue * const q = mdev->rq_queue; | |
99432fcc PR |
1012 | int max_hw_sectors = max_bio_size >> 9; |
1013 | int max_segments = 0; | |
1014 | ||
1015 | if (get_ldev_if_state(mdev, D_ATTACHING)) { | |
1016 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
1017 | ||
1018 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc PR |
1019 | rcu_read_lock(); |
1020 | max_segments = rcu_dereference(mdev->ldev->disk_conf)->max_bio_bvecs; | |
1021 | rcu_read_unlock(); | |
99432fcc PR |
1022 | put_ldev(mdev); |
1023 | } | |
b411b363 | 1024 | |
b411b363 | 1025 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1026 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1027 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1028 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1029 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1030 | |
99432fcc PR |
1031 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1032 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
1033 | ||
1034 | blk_queue_stack_limits(q, b); | |
1035 | ||
1036 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
1037 | dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", | |
1038 | q->backing_dev_info.ra_pages, | |
1039 | b->backing_dev_info.ra_pages); | |
1040 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1041 | } | |
1042 | put_ldev(mdev); | |
1043 | } | |
1044 | } | |
1045 | ||
1046 | void drbd_reconsider_max_bio_size(struct drbd_conf *mdev) | |
1047 | { | |
1048 | int now, new, local, peer; | |
1049 | ||
1050 | now = queue_max_hw_sectors(mdev->rq_queue) << 9; | |
1051 | local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1052 | peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1053 | |
99432fcc PR |
1054 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1055 | local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9; | |
1056 | mdev->local_max_bio_size = local; | |
1057 | put_ldev(mdev); | |
b411b363 | 1058 | } |
99432fcc PR |
1059 | |
1060 | /* We may ignore peer limits if the peer is modern enough. | |
1061 | Because new from 8.3.8 onwards the peer can use multiple | |
1062 | BIOs for a single peer_request */ | |
1063 | if (mdev->state.conn >= C_CONNECTED) { | |
31890f4a | 1064 | if (mdev->tconn->agreed_pro_version < 94) |
2ffca4f3 PR |
1065 | peer = min_t(int, mdev->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
1066 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ | |
31890f4a | 1067 | else if (mdev->tconn->agreed_pro_version == 94) |
99432fcc | 1068 | peer = DRBD_MAX_SIZE_H80_PACKET; |
2ffca4f3 PR |
1069 | else if (mdev->tconn->agreed_pro_version < 100) |
1070 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ | |
1071 | else | |
99432fcc PR |
1072 | peer = DRBD_MAX_BIO_SIZE; |
1073 | } | |
1074 | ||
1075 | new = min_t(int, local, peer); | |
1076 | ||
1077 | if (mdev->state.role == R_PRIMARY && new < now) | |
1078 | dev_err(DEV, "ASSERT FAILED new < now; (%d < %d)\n", new, now); | |
1079 | ||
1080 | if (new != now) | |
1081 | dev_info(DEV, "max BIO size = %u\n", new); | |
1082 | ||
1083 | drbd_setup_queue_param(mdev, new); | |
b411b363 PR |
1084 | } |
1085 | ||
a18e9d1e | 1086 | /* Starts the worker thread */ |
0e29d163 | 1087 | static void conn_reconfig_start(struct drbd_tconn *tconn) |
b411b363 | 1088 | { |
0e29d163 PR |
1089 | drbd_thread_start(&tconn->worker); |
1090 | conn_flush_workqueue(tconn); | |
b411b363 PR |
1091 | } |
1092 | ||
a18e9d1e | 1093 | /* if still unconfigured, stops worker again. */ |
0e29d163 | 1094 | static void conn_reconfig_done(struct drbd_tconn *tconn) |
b411b363 | 1095 | { |
992d6e91 | 1096 | bool stop_threads; |
0e29d163 | 1097 | spin_lock_irq(&tconn->req_lock); |
e0e16653 PR |
1098 | stop_threads = conn_all_vols_unconf(tconn) && |
1099 | tconn->cstate == C_STANDALONE; | |
0e29d163 | 1100 | spin_unlock_irq(&tconn->req_lock); |
992d6e91 LE |
1101 | if (stop_threads) { |
1102 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1103 | * in conn_disconnect() */ |
992d6e91 LE |
1104 | drbd_thread_stop(&tconn->receiver); |
1105 | drbd_thread_stop(&tconn->worker); | |
1106 | } | |
b411b363 PR |
1107 | } |
1108 | ||
0778286a PR |
1109 | /* Make sure IO is suspended before calling this function(). */ |
1110 | static void drbd_suspend_al(struct drbd_conf *mdev) | |
1111 | { | |
1112 | int s = 0; | |
1113 | ||
61610420 | 1114 | if (!lc_try_lock(mdev->act_log)) { |
0778286a PR |
1115 | dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n"); |
1116 | return; | |
1117 | } | |
1118 | ||
61610420 | 1119 | drbd_al_shrink(mdev); |
87eeee41 | 1120 | spin_lock_irq(&mdev->tconn->req_lock); |
0778286a PR |
1121 | if (mdev->state.conn < C_CONNECTED) |
1122 | s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags); | |
87eeee41 | 1123 | spin_unlock_irq(&mdev->tconn->req_lock); |
61610420 | 1124 | lc_unlock(mdev->act_log); |
0778286a PR |
1125 | |
1126 | if (s) | |
1127 | dev_info(DEV, "Suspended AL updates\n"); | |
1128 | } | |
1129 | ||
5979e361 LE |
1130 | |
1131 | static bool should_set_defaults(struct genl_info *info) | |
1132 | { | |
1133 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1134 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1135 | } | |
1136 | ||
d589a21e PR |
1137 | static void enforce_disk_conf_limits(struct disk_conf *dc) |
1138 | { | |
1139 | if (dc->al_extents < DRBD_AL_EXTENTS_MIN) | |
1140 | dc->al_extents = DRBD_AL_EXTENTS_MIN; | |
1141 | if (dc->al_extents > DRBD_AL_EXTENTS_MAX) | |
1142 | dc->al_extents = DRBD_AL_EXTENTS_MAX; | |
1143 | ||
1144 | if (dc->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1145 | dc->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
1146 | } | |
1147 | ||
f399002e LE |
1148 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1149 | { | |
1150 | enum drbd_ret_code retcode; | |
1151 | struct drbd_conf *mdev; | |
daeda1cc | 1152 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1153 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1154 | int err, fifo_size; |
f399002e LE |
1155 | |
1156 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1157 | if (!adm_ctx.reply_skb) | |
1158 | return retcode; | |
1159 | if (retcode != NO_ERROR) | |
1160 | goto out; | |
1161 | ||
1162 | mdev = adm_ctx.mdev; | |
1163 | ||
1164 | /* we also need a disk | |
1165 | * to change the options on */ | |
1166 | if (!get_ldev(mdev)) { | |
1167 | retcode = ERR_NO_DISK; | |
1168 | goto out; | |
1169 | } | |
1170 | ||
daeda1cc | 1171 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1172 | if (!new_disk_conf) { |
f399002e LE |
1173 | retcode = ERR_NOMEM; |
1174 | goto fail; | |
1175 | } | |
1176 | ||
daeda1cc PR |
1177 | mutex_lock(&mdev->tconn->conf_update); |
1178 | old_disk_conf = mdev->ldev->disk_conf; | |
1179 | *new_disk_conf = *old_disk_conf; | |
5979e361 | 1180 | if (should_set_defaults(info)) |
b966b5dd | 1181 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1182 | |
5ecc72c3 | 1183 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1184 | if (err && err != -ENOMSG) { |
f399002e LE |
1185 | retcode = ERR_MANDATORY_TAG; |
1186 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1187 | } | |
1188 | ||
5ecc72c3 LE |
1189 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1190 | new_disk_conf->resync_rate = 1; | |
f399002e | 1191 | |
d589a21e | 1192 | enforce_disk_conf_limits(new_disk_conf); |
f399002e | 1193 | |
5ecc72c3 | 1194 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
9958c857 | 1195 | if (fifo_size != mdev->rs_plan_s->size) { |
813472ce PR |
1196 | new_plan = fifo_alloc(fifo_size); |
1197 | if (!new_plan) { | |
f399002e LE |
1198 | dev_err(DEV, "kmalloc of fifo_buffer failed"); |
1199 | retcode = ERR_NOMEM; | |
daeda1cc | 1200 | goto fail_unlock; |
f399002e LE |
1201 | } |
1202 | } | |
1203 | ||
f399002e LE |
1204 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); |
1205 | drbd_al_shrink(mdev); | |
5ecc72c3 | 1206 | err = drbd_check_al_size(mdev, new_disk_conf); |
f399002e LE |
1207 | lc_unlock(mdev->act_log); |
1208 | wake_up(&mdev->al_wait); | |
1209 | ||
1210 | if (err) { | |
1211 | retcode = ERR_NOMEM; | |
daeda1cc | 1212 | goto fail_unlock; |
f399002e LE |
1213 | } |
1214 | ||
dc97b708 | 1215 | write_lock_irq(&global_state_lock); |
95f8efd0 | 1216 | retcode = drbd_resync_after_valid(mdev, new_disk_conf->resync_after); |
dc97b708 | 1217 | if (retcode == NO_ERROR) { |
daeda1cc | 1218 | rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf); |
95f8efd0 | 1219 | drbd_resync_after_changed(mdev); |
dc97b708 PR |
1220 | } |
1221 | write_unlock_irq(&global_state_lock); | |
f399002e | 1222 | |
daeda1cc PR |
1223 | if (retcode != NO_ERROR) |
1224 | goto fail_unlock; | |
f399002e | 1225 | |
813472ce PR |
1226 | if (new_plan) { |
1227 | old_plan = mdev->rs_plan_s; | |
1228 | rcu_assign_pointer(mdev->rs_plan_s, new_plan); | |
9958c857 | 1229 | } |
9958c857 | 1230 | |
c141ebda | 1231 | mutex_unlock(&mdev->tconn->conf_update); |
27eb13e9 PR |
1232 | |
1233 | drbd_bump_write_ordering(mdev->tconn, WO_bdev_flush); | |
1234 | ||
daeda1cc | 1235 | drbd_md_sync(mdev); |
f399002e LE |
1236 | |
1237 | if (mdev->state.conn >= C_CONNECTED) | |
1238 | drbd_send_sync_param(mdev); | |
1239 | ||
daeda1cc PR |
1240 | synchronize_rcu(); |
1241 | kfree(old_disk_conf); | |
813472ce | 1242 | kfree(old_plan); |
cdfda633 | 1243 | mod_timer(&mdev->request_timer, jiffies + HZ); |
daeda1cc PR |
1244 | goto success; |
1245 | ||
1246 | fail_unlock: | |
1247 | mutex_unlock(&mdev->tconn->conf_update); | |
f399002e | 1248 | fail: |
5ecc72c3 | 1249 | kfree(new_disk_conf); |
813472ce | 1250 | kfree(new_plan); |
daeda1cc PR |
1251 | success: |
1252 | put_ldev(mdev); | |
f399002e LE |
1253 | out: |
1254 | drbd_adm_finish(info, retcode); | |
1255 | return 0; | |
1256 | } | |
1257 | ||
3b98c0c2 | 1258 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1259 | { |
3b98c0c2 LE |
1260 | struct drbd_conf *mdev; |
1261 | int err; | |
116676ca | 1262 | enum drbd_ret_code retcode; |
b411b363 PR |
1263 | enum determine_dev_size dd; |
1264 | sector_t max_possible_sectors; | |
1265 | sector_t min_md_device_sectors; | |
1266 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1267 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1268 | struct block_device *bdev; |
b411b363 | 1269 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1270 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1271 | union drbd_state ns, os; |
f2024e7c | 1272 | enum drbd_state_rv rv; |
44ed167d | 1273 | struct net_conf *nc; |
b411b363 | 1274 | |
3b98c0c2 LE |
1275 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1276 | if (!adm_ctx.reply_skb) | |
1277 | return retcode; | |
1278 | if (retcode != NO_ERROR) | |
40cbf085 | 1279 | goto finish; |
3b98c0c2 LE |
1280 | |
1281 | mdev = adm_ctx.mdev; | |
0e29d163 | 1282 | conn_reconfig_start(mdev->tconn); |
b411b363 PR |
1283 | |
1284 | /* if you want to reconfigure, please tear down first */ | |
1285 | if (mdev->state.disk > D_DISKLESS) { | |
1286 | retcode = ERR_DISK_CONFIGURED; | |
1287 | goto fail; | |
1288 | } | |
82f59cc6 LE |
1289 | /* It may just now have detached because of IO error. Make sure |
1290 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1291 | * e.g. if someone calls attach from the on-io-error handler, | |
1292 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
1293 | wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); | |
b411b363 | 1294 | |
3b98c0c2 | 1295 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1296 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1297 | if (!nbc) { | |
1298 | retcode = ERR_NOMEM; | |
1299 | goto fail; | |
1300 | } | |
daeda1cc PR |
1301 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1302 | if (!new_disk_conf) { | |
1303 | retcode = ERR_NOMEM; | |
1304 | goto fail; | |
1305 | } | |
1306 | nbc->disk_conf = new_disk_conf; | |
b411b363 | 1307 | |
daeda1cc PR |
1308 | set_disk_conf_defaults(new_disk_conf); |
1309 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1310 | if (err) { |
b411b363 | 1311 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1312 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1313 | goto fail; |
1314 | } | |
1315 | ||
d589a21e PR |
1316 | enforce_disk_conf_limits(new_disk_conf); |
1317 | ||
9958c857 PR |
1318 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1319 | if (!new_plan) { | |
1320 | retcode = ERR_NOMEM; | |
1321 | goto fail; | |
1322 | } | |
1323 | ||
daeda1cc | 1324 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1325 | retcode = ERR_MD_IDX_INVALID; |
1326 | goto fail; | |
1327 | } | |
1328 | ||
44ed167d PR |
1329 | rcu_read_lock(); |
1330 | nc = rcu_dereference(mdev->tconn->net_conf); | |
1331 | if (nc) { | |
daeda1cc | 1332 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1333 | rcu_read_unlock(); |
47ff2d0a PR |
1334 | retcode = ERR_STONITH_AND_PROT_A; |
1335 | goto fail; | |
1336 | } | |
1337 | } | |
44ed167d | 1338 | rcu_read_unlock(); |
47ff2d0a | 1339 | |
daeda1cc | 1340 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
d4d77629 | 1341 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev); |
e525fd89 | 1342 | if (IS_ERR(bdev)) { |
daeda1cc | 1343 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1344 | PTR_ERR(bdev)); |
b411b363 PR |
1345 | retcode = ERR_OPEN_DISK; |
1346 | goto fail; | |
1347 | } | |
e525fd89 TH |
1348 | nbc->backing_bdev = bdev; |
1349 | ||
1350 | /* | |
1351 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1352 | * drbd sharing one meta device. TODO in that case, paranoia | |
1353 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1354 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1355 | * should check it for you already; but if you don't, or | |
1356 | * someone fooled it, we need to double check here) | |
1357 | */ | |
daeda1cc | 1358 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1359 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1360 | (new_disk_conf->meta_dev_idx < 0) ? |
d4d77629 | 1361 | (void *)mdev : (void *)drbd_m_holder); |
e525fd89 | 1362 | if (IS_ERR(bdev)) { |
daeda1cc | 1363 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1364 | PTR_ERR(bdev)); |
b411b363 PR |
1365 | retcode = ERR_OPEN_MD_DISK; |
1366 | goto fail; | |
1367 | } | |
e525fd89 | 1368 | nbc->md_bdev = bdev; |
b411b363 | 1369 | |
e525fd89 | 1370 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1371 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1372 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1373 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1374 | goto fail; |
1375 | } | |
1376 | ||
1377 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1378 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1379 | offsetof(struct bm_extent, lce)); |
1380 | if (!resync_lru) { | |
1381 | retcode = ERR_NOMEM; | |
e525fd89 | 1382 | goto fail; |
b411b363 PR |
1383 | } |
1384 | ||
1385 | /* RT - for drbd_get_max_capacity() DRBD_MD_INDEX_FLEX_INT */ | |
1386 | drbd_md_set_sector_offsets(mdev, nbc); | |
1387 | ||
daeda1cc | 1388 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
b411b363 PR |
1389 | dev_err(DEV, "max capacity %llu smaller than disk size %llu\n", |
1390 | (unsigned long long) drbd_get_max_capacity(nbc), | |
daeda1cc | 1391 | (unsigned long long) new_disk_conf->disk_size); |
67b58bf7 | 1392 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1393 | goto fail; |
b411b363 PR |
1394 | } |
1395 | ||
daeda1cc | 1396 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1397 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1398 | /* at least one MB, otherwise it does not make sense */ | |
1399 | min_md_device_sectors = (2<<10); | |
1400 | } else { | |
1401 | max_possible_sectors = DRBD_MAX_SECTORS; | |
daeda1cc | 1402 | min_md_device_sectors = MD_RESERVED_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1403 | } |
1404 | ||
b411b363 | 1405 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
67b58bf7 | 1406 | retcode = ERR_MD_DISK_TOO_SMALL; |
b411b363 PR |
1407 | dev_warn(DEV, "refusing attach: md-device too small, " |
1408 | "at least %llu sectors needed for this meta-disk type\n", | |
1409 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1410 | goto fail; |
b411b363 PR |
1411 | } |
1412 | ||
1413 | /* Make sure the new disk is big enough | |
1414 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1415 | if (drbd_get_max_capacity(nbc) < | |
1416 | drbd_get_capacity(mdev->this_bdev)) { | |
67b58bf7 | 1417 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1418 | goto fail; |
b411b363 PR |
1419 | } |
1420 | ||
1421 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1422 | ||
1352994b LE |
1423 | if (nbc->known_size > max_possible_sectors) { |
1424 | dev_warn(DEV, "==> truncating very big lower level device " | |
1425 | "to currently maximum possible %llu sectors <==\n", | |
1426 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1427 | if (new_disk_conf->meta_dev_idx >= 0) |
1352994b LE |
1428 | dev_warn(DEV, "==>> using internal or flexible " |
1429 | "meta data may help <<==\n"); | |
1430 | } | |
1431 | ||
b411b363 PR |
1432 | drbd_suspend_io(mdev); |
1433 | /* also wait for the last barrier ack. */ | |
2aebfabb | 1434 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev)); |
b411b363 | 1435 | /* and for any other previously queued work */ |
a21e9298 | 1436 | drbd_flush_workqueue(mdev); |
b411b363 | 1437 | |
f2024e7c AG |
1438 | rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE); |
1439 | retcode = rv; /* FIXME: Type mismatch. */ | |
b411b363 | 1440 | drbd_resume_io(mdev); |
f2024e7c | 1441 | if (rv < SS_SUCCESS) |
e525fd89 | 1442 | goto fail; |
b411b363 PR |
1443 | |
1444 | if (!get_ldev_if_state(mdev, D_ATTACHING)) | |
1445 | goto force_diskless; | |
1446 | ||
1447 | drbd_md_set_sector_offsets(mdev, nbc); | |
1448 | ||
1449 | if (!mdev->bitmap) { | |
1450 | if (drbd_bm_init(mdev)) { | |
1451 | retcode = ERR_NOMEM; | |
1452 | goto force_diskless_dec; | |
1453 | } | |
1454 | } | |
1455 | ||
1456 | retcode = drbd_md_read(mdev, nbc); | |
1457 | if (retcode != NO_ERROR) | |
1458 | goto force_diskless_dec; | |
1459 | ||
1460 | if (mdev->state.conn < C_CONNECTED && | |
1461 | mdev->state.role == R_PRIMARY && | |
1462 | (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
1463 | dev_err(DEV, "Can only attach to data with current UUID=%016llX\n", | |
1464 | (unsigned long long)mdev->ed_uuid); | |
1465 | retcode = ERR_DATA_NOT_CURRENT; | |
1466 | goto force_diskless_dec; | |
1467 | } | |
1468 | ||
1469 | /* Since we are diskless, fix the activity log first... */ | |
daeda1cc | 1470 | if (drbd_check_al_size(mdev, new_disk_conf)) { |
b411b363 PR |
1471 | retcode = ERR_NOMEM; |
1472 | goto force_diskless_dec; | |
1473 | } | |
1474 | ||
1475 | /* Prevent shrinking of consistent devices ! */ | |
1476 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
daeda1cc | 1477 | drbd_new_dev_size(mdev, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
b411b363 | 1478 | dev_warn(DEV, "refusing to truncate a consistent device\n"); |
67b58bf7 | 1479 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1480 | goto force_diskless_dec; |
1481 | } | |
1482 | ||
b411b363 PR |
1483 | /* Reset the "barriers don't work" bits here, then force meta data to |
1484 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1485 | if (new_disk_conf->md_flushes) |
a8a4e51e | 1486 | clear_bit(MD_NO_FUA, &mdev->flags); |
e544046a AG |
1487 | else |
1488 | set_bit(MD_NO_FUA, &mdev->flags); | |
b411b363 PR |
1489 | |
1490 | /* Point of no return reached. | |
1491 | * Devices and memory are no longer released by error cleanup below. | |
1492 | * now mdev takes over responsibility, and the state engine should | |
1493 | * clean it up somewhere. */ | |
1494 | D_ASSERT(mdev->ldev == NULL); | |
1495 | mdev->ldev = nbc; | |
1496 | mdev->resync = resync_lru; | |
9958c857 | 1497 | mdev->rs_plan_s = new_plan; |
b411b363 PR |
1498 | nbc = NULL; |
1499 | resync_lru = NULL; | |
daeda1cc | 1500 | new_disk_conf = NULL; |
9958c857 | 1501 | new_plan = NULL; |
b411b363 | 1502 | |
4b0007c0 | 1503 | drbd_bump_write_ordering(mdev->tconn, WO_bdev_flush); |
b411b363 PR |
1504 | |
1505 | if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY)) | |
1506 | set_bit(CRASHED_PRIMARY, &mdev->flags); | |
1507 | else | |
1508 | clear_bit(CRASHED_PRIMARY, &mdev->flags); | |
1509 | ||
894c6a94 | 1510 | if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && |
d5d7ebd4 | 1511 | !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod)) |
b411b363 | 1512 | set_bit(CRASHED_PRIMARY, &mdev->flags); |
b411b363 PR |
1513 | |
1514 | mdev->send_cnt = 0; | |
1515 | mdev->recv_cnt = 0; | |
1516 | mdev->read_cnt = 0; | |
1517 | mdev->writ_cnt = 0; | |
1518 | ||
99432fcc | 1519 | drbd_reconsider_max_bio_size(mdev); |
b411b363 PR |
1520 | |
1521 | /* If I am currently not R_PRIMARY, | |
1522 | * but meta data primary indicator is set, | |
1523 | * I just now recover from a hard crash, | |
1524 | * and have been R_PRIMARY before that crash. | |
1525 | * | |
1526 | * Now, if I had no connection before that crash | |
1527 | * (have been degraded R_PRIMARY), chances are that | |
1528 | * I won't find my peer now either. | |
1529 | * | |
1530 | * In that case, and _only_ in that case, | |
1531 | * we use the degr-wfc-timeout instead of the default, | |
1532 | * so we can automatically recover from a crash of a | |
1533 | * degraded but active "cluster" after a certain timeout. | |
1534 | */ | |
1535 | clear_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1536 | if (mdev->state.role != R_PRIMARY && | |
1537 | drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && | |
1538 | !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND)) | |
1539 | set_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1540 | ||
24c4830c | 1541 | dd = drbd_determine_dev_size(mdev, 0); |
b411b363 PR |
1542 | if (dd == dev_size_error) { |
1543 | retcode = ERR_NOMEM_BITMAP; | |
1544 | goto force_diskless_dec; | |
1545 | } else if (dd == grew) | |
1546 | set_bit(RESYNC_AFTER_NEG, &mdev->flags); | |
1547 | ||
1548 | if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC)) { | |
1549 | dev_info(DEV, "Assuming that all blocks are out of sync " | |
1550 | "(aka FullSync)\n"); | |
20ceb2b2 LE |
1551 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, |
1552 | "set_n_write from attaching", BM_LOCKED_MASK)) { | |
b411b363 PR |
1553 | retcode = ERR_IO_MD_DISK; |
1554 | goto force_diskless_dec; | |
1555 | } | |
1556 | } else { | |
20ceb2b2 | 1557 | if (drbd_bitmap_io(mdev, &drbd_bm_read, |
22ab6a30 | 1558 | "read from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1559 | retcode = ERR_IO_MD_DISK; |
1560 | goto force_diskless_dec; | |
1561 | } | |
1562 | } | |
1563 | ||
0778286a PR |
1564 | if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev)) |
1565 | drbd_suspend_al(mdev); /* IO is still suspended here... */ | |
1566 | ||
87eeee41 | 1567 | spin_lock_irq(&mdev->tconn->req_lock); |
78bae59b PR |
1568 | os = drbd_read_state(mdev); |
1569 | ns = os; | |
b411b363 PR |
1570 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1571 | otherwise investigate MDF_WasUpToDate... | |
1572 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1573 | otherwise into D_CONSISTENT state. | |
1574 | */ | |
1575 | if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) { | |
1576 | if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE)) | |
1577 | ns.disk = D_CONSISTENT; | |
1578 | else | |
1579 | ns.disk = D_OUTDATED; | |
1580 | } else { | |
1581 | ns.disk = D_INCONSISTENT; | |
1582 | } | |
1583 | ||
1584 | if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED)) | |
1585 | ns.pdsk = D_OUTDATED; | |
1586 | ||
daeda1cc PR |
1587 | rcu_read_lock(); |
1588 | if (ns.disk == D_CONSISTENT && | |
1589 | (ns.pdsk == D_OUTDATED || rcu_dereference(mdev->ldev->disk_conf)->fencing == FP_DONT_CARE)) | |
b411b363 | 1590 | ns.disk = D_UP_TO_DATE; |
daeda1cc | 1591 | rcu_read_unlock(); |
b411b363 PR |
1592 | |
1593 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1594 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1595 | this point, because drbd_request_state() modifies these | |
1596 | flags. */ | |
1597 | ||
1598 | /* In case we are C_CONNECTED postpone any decision on the new disk | |
1599 | state after the negotiation phase. */ | |
1600 | if (mdev->state.conn == C_CONNECTED) { | |
1601 | mdev->new_state_tmp.i = ns.i; | |
1602 | ns.i = os.i; | |
1603 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1604 | |
1605 | /* We expect to receive up-to-date UUIDs soon. | |
1606 | To avoid a race in receive_state, free p_uuid while | |
1607 | holding req_lock. I.e. atomic with the state change */ | |
1608 | kfree(mdev->p_uuid); | |
1609 | mdev->p_uuid = NULL; | |
b411b363 PR |
1610 | } |
1611 | ||
1612 | rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL); | |
87eeee41 | 1613 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
1614 | |
1615 | if (rv < SS_SUCCESS) | |
1616 | goto force_diskless_dec; | |
1617 | ||
cdfda633 PR |
1618 | mod_timer(&mdev->request_timer, jiffies + HZ); |
1619 | ||
b411b363 PR |
1620 | if (mdev->state.role == R_PRIMARY) |
1621 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
1622 | else | |
1623 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
1624 | ||
1625 | drbd_md_mark_dirty(mdev); | |
1626 | drbd_md_sync(mdev); | |
1627 | ||
1628 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
1629 | put_ldev(mdev); | |
0e29d163 | 1630 | conn_reconfig_done(mdev->tconn); |
3b98c0c2 | 1631 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1632 | return 0; |
1633 | ||
1634 | force_diskless_dec: | |
1635 | put_ldev(mdev); | |
1636 | force_diskless: | |
9510b241 | 1637 | drbd_force_state(mdev, NS(disk, D_DISKLESS)); |
b411b363 | 1638 | drbd_md_sync(mdev); |
b411b363 | 1639 | fail: |
40cbf085 | 1640 | conn_reconfig_done(mdev->tconn); |
b411b363 | 1641 | if (nbc) { |
e525fd89 TH |
1642 | if (nbc->backing_bdev) |
1643 | blkdev_put(nbc->backing_bdev, | |
1644 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1645 | if (nbc->md_bdev) | |
1646 | blkdev_put(nbc->md_bdev, | |
1647 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1648 | kfree(nbc); |
1649 | } | |
daeda1cc | 1650 | kfree(new_disk_conf); |
b411b363 | 1651 | lc_destroy(resync_lru); |
9958c857 | 1652 | kfree(new_plan); |
b411b363 | 1653 | |
40cbf085 | 1654 | finish: |
3b98c0c2 | 1655 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1656 | return 0; |
1657 | } | |
1658 | ||
cdfda633 | 1659 | static int adm_detach(struct drbd_conf *mdev, int force) |
85f75dd7 | 1660 | { |
19f83c76 | 1661 | enum drbd_state_rv retcode; |
009ba89d | 1662 | int ret; |
cdfda633 PR |
1663 | |
1664 | if (force) { | |
1665 | drbd_force_state(mdev, NS(disk, D_FAILED)); | |
1666 | retcode = SS_SUCCESS; | |
1667 | goto out; | |
1668 | } | |
1669 | ||
85f75dd7 | 1670 | drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */ |
0cfac5dd | 1671 | drbd_md_get_buffer(mdev); /* make sure there is no in-flight meta-data IO */ |
009ba89d | 1672 | retcode = drbd_request_state(mdev, NS(disk, D_FAILED)); |
0cfac5dd | 1673 | drbd_md_put_buffer(mdev); |
009ba89d LE |
1674 | /* D_FAILED will transition to DISKLESS. */ |
1675 | ret = wait_event_interruptible(mdev->misc_wait, | |
1676 | mdev->state.disk != D_FAILED); | |
85f75dd7 | 1677 | drbd_resume_io(mdev); |
009ba89d LE |
1678 | if ((int)retcode == (int)SS_IS_DISKLESS) |
1679 | retcode = SS_NOTHING_TO_DO; | |
1680 | if (ret) | |
1681 | retcode = ERR_INTR; | |
cdfda633 | 1682 | out: |
85f75dd7 LE |
1683 | return retcode; |
1684 | } | |
1685 | ||
82f59cc6 LE |
1686 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1687 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1688 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1689 | * internal references as well. | |
1690 | * Only then we have finally detached. */ | |
3b98c0c2 | 1691 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1692 | { |
9a0d9d03 | 1693 | enum drbd_ret_code retcode; |
cdfda633 PR |
1694 | struct detach_parms parms = { }; |
1695 | int err; | |
3b98c0c2 LE |
1696 | |
1697 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1698 | if (!adm_ctx.reply_skb) | |
1699 | return retcode; | |
1700 | if (retcode != NO_ERROR) | |
1701 | goto out; | |
1702 | ||
cdfda633 PR |
1703 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1704 | err = detach_parms_from_attrs(&parms, info); | |
1705 | if (err) { | |
1706 | retcode = ERR_MANDATORY_TAG; | |
1707 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1708 | goto out; | |
1709 | } | |
1710 | } | |
1711 | ||
1712 | retcode = adm_detach(adm_ctx.mdev, parms.force_detach); | |
3b98c0c2 LE |
1713 | out: |
1714 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1715 | return 0; |
1716 | } | |
1717 | ||
f399002e LE |
1718 | static bool conn_resync_running(struct drbd_tconn *tconn) |
1719 | { | |
1720 | struct drbd_conf *mdev; | |
695d08fa | 1721 | bool rv = false; |
f399002e LE |
1722 | int vnr; |
1723 | ||
695d08fa | 1724 | rcu_read_lock(); |
f399002e LE |
1725 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1726 | if (mdev->state.conn == C_SYNC_SOURCE || | |
1727 | mdev->state.conn == C_SYNC_TARGET || | |
1728 | mdev->state.conn == C_PAUSED_SYNC_S || | |
695d08fa PR |
1729 | mdev->state.conn == C_PAUSED_SYNC_T) { |
1730 | rv = true; | |
1731 | break; | |
1732 | } | |
f399002e | 1733 | } |
695d08fa PR |
1734 | rcu_read_unlock(); |
1735 | ||
1736 | return rv; | |
f399002e LE |
1737 | } |
1738 | ||
1739 | static bool conn_ov_running(struct drbd_tconn *tconn) | |
1740 | { | |
1741 | struct drbd_conf *mdev; | |
695d08fa | 1742 | bool rv = false; |
f399002e LE |
1743 | int vnr; |
1744 | ||
695d08fa | 1745 | rcu_read_lock(); |
f399002e LE |
1746 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1747 | if (mdev->state.conn == C_VERIFY_S || | |
695d08fa PR |
1748 | mdev->state.conn == C_VERIFY_T) { |
1749 | rv = true; | |
1750 | break; | |
1751 | } | |
f399002e | 1752 | } |
695d08fa PR |
1753 | rcu_read_unlock(); |
1754 | ||
1755 | return rv; | |
f399002e LE |
1756 | } |
1757 | ||
cd64397c | 1758 | static enum drbd_ret_code |
44ed167d | 1759 | _check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf) |
cd64397c PR |
1760 | { |
1761 | struct drbd_conf *mdev; | |
1762 | int i; | |
1763 | ||
dcb20d1a PR |
1764 | if (old_conf && tconn->cstate == C_WF_REPORT_PARAMS && tconn->agreed_pro_version < 100) { |
1765 | if (new_conf->wire_protocol != old_conf->wire_protocol) | |
1766 | return ERR_NEED_APV_100; | |
1767 | ||
1768 | if (new_conf->two_primaries != old_conf->two_primaries) | |
1769 | return ERR_NEED_APV_100; | |
1770 | ||
1771 | if (!new_conf->integrity_alg != !old_conf->integrity_alg) | |
1772 | return ERR_NEED_APV_100; | |
1773 | ||
1774 | if (strcmp(new_conf->integrity_alg, old_conf->integrity_alg)) | |
1775 | return ERR_NEED_APV_100; | |
1776 | } | |
1777 | ||
1778 | if (!new_conf->two_primaries && | |
1779 | conn_highest_role(tconn) == R_PRIMARY && | |
1780 | conn_highest_peer(tconn) == R_PRIMARY) | |
1781 | return ERR_NEED_ALLOW_TWO_PRI; | |
b032b6fa | 1782 | |
cd64397c PR |
1783 | if (new_conf->two_primaries && |
1784 | (new_conf->wire_protocol != DRBD_PROT_C)) | |
1785 | return ERR_NOT_PROTO_C; | |
1786 | ||
cd64397c PR |
1787 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
1788 | if (get_ldev(mdev)) { | |
daeda1cc | 1789 | enum drbd_fencing_p fp = rcu_dereference(mdev->ldev->disk_conf)->fencing; |
cd64397c | 1790 | put_ldev(mdev); |
44ed167d | 1791 | if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1792 | return ERR_STONITH_AND_PROT_A; |
cd64397c | 1793 | } |
6139f60d | 1794 | if (mdev->state.role == R_PRIMARY && new_conf->discard_my_data) |
cd64397c | 1795 | return ERR_DISCARD; |
cd64397c | 1796 | } |
cd64397c PR |
1797 | |
1798 | if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A) | |
1799 | return ERR_CONG_NOT_PROTO_A; | |
1800 | ||
1801 | return NO_ERROR; | |
1802 | } | |
1803 | ||
44ed167d PR |
1804 | static enum drbd_ret_code |
1805 | check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf) | |
1806 | { | |
1807 | static enum drbd_ret_code rv; | |
1808 | struct drbd_conf *mdev; | |
1809 | int i; | |
1810 | ||
1811 | rcu_read_lock(); | |
1812 | rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf); | |
1813 | rcu_read_unlock(); | |
1814 | ||
1815 | /* tconn->volumes protected by genl_lock() here */ | |
1816 | idr_for_each_entry(&tconn->volumes, mdev, i) { | |
1817 | if (!mdev->bitmap) { | |
1818 | if(drbd_bm_init(mdev)) | |
1819 | return ERR_NOMEM; | |
1820 | } | |
1821 | } | |
1822 | ||
1823 | return rv; | |
1824 | } | |
1825 | ||
0fd0ea06 PR |
1826 | struct crypto { |
1827 | struct crypto_hash *verify_tfm; | |
1828 | struct crypto_hash *csums_tfm; | |
1829 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 1830 | struct crypto_hash *integrity_tfm; |
0fd0ea06 PR |
1831 | }; |
1832 | ||
1833 | static int | |
4b6ad6d4 | 1834 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
1835 | { |
1836 | if (!tfm_name[0]) | |
1837 | return NO_ERROR; | |
1838 | ||
1839 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); | |
1840 | if (IS_ERR(*tfm)) { | |
1841 | *tfm = NULL; | |
1842 | return err_alg; | |
1843 | } | |
1844 | ||
0fd0ea06 PR |
1845 | return NO_ERROR; |
1846 | } | |
1847 | ||
1848 | static enum drbd_ret_code | |
1849 | alloc_crypto(struct crypto *crypto, struct net_conf *new_conf) | |
1850 | { | |
1851 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
1852 | enum drbd_ret_code rv; | |
0fd0ea06 | 1853 | |
4b6ad6d4 AG |
1854 | rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg, |
1855 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
1856 | if (rv != NO_ERROR) |
1857 | return rv; | |
4b6ad6d4 AG |
1858 | rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg, |
1859 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
1860 | if (rv != NO_ERROR) |
1861 | return rv; | |
4b6ad6d4 AG |
1862 | rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg, |
1863 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
1864 | if (rv != NO_ERROR) |
1865 | return rv; | |
0fd0ea06 PR |
1866 | if (new_conf->cram_hmac_alg[0] != 0) { |
1867 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", | |
1868 | new_conf->cram_hmac_alg); | |
1869 | ||
4b6ad6d4 AG |
1870 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
1871 | ERR_AUTH_ALG); | |
0fd0ea06 | 1872 | } |
0fd0ea06 PR |
1873 | |
1874 | return rv; | |
1875 | } | |
1876 | ||
1877 | static void free_crypto(struct crypto *crypto) | |
1878 | { | |
0fd0ea06 | 1879 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 1880 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
1881 | crypto_free_hash(crypto->csums_tfm); |
1882 | crypto_free_hash(crypto->verify_tfm); | |
1883 | } | |
1884 | ||
f399002e LE |
1885 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
1886 | { | |
1887 | enum drbd_ret_code retcode; | |
1888 | struct drbd_tconn *tconn; | |
44ed167d | 1889 | struct net_conf *old_conf, *new_conf = NULL; |
f399002e LE |
1890 | int err; |
1891 | int ovr; /* online verify running */ | |
1892 | int rsr; /* re-sync running */ | |
0fd0ea06 | 1893 | struct crypto crypto = { }; |
f399002e | 1894 | |
089c075d | 1895 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
1896 | if (!adm_ctx.reply_skb) |
1897 | return retcode; | |
1898 | if (retcode != NO_ERROR) | |
1899 | goto out; | |
1900 | ||
1901 | tconn = adm_ctx.tconn; | |
1902 | ||
1903 | new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); | |
1904 | if (!new_conf) { | |
1905 | retcode = ERR_NOMEM; | |
1906 | goto out; | |
1907 | } | |
1908 | ||
44ed167d PR |
1909 | conn_reconfig_start(tconn); |
1910 | ||
88104ca4 | 1911 | mutex_lock(&tconn->data.mutex); |
a0095508 | 1912 | mutex_lock(&tconn->conf_update); |
91fd4dad | 1913 | old_conf = tconn->net_conf; |
44ed167d PR |
1914 | |
1915 | if (!old_conf) { | |
f399002e LE |
1916 | drbd_msg_put_info("net conf missing, try connect"); |
1917 | retcode = ERR_INVALID_REQUEST; | |
91fd4dad | 1918 | goto fail; |
f399002e LE |
1919 | } |
1920 | ||
44ed167d | 1921 | *new_conf = *old_conf; |
5979e361 | 1922 | if (should_set_defaults(info)) |
b966b5dd | 1923 | set_net_conf_defaults(new_conf); |
f399002e | 1924 | |
f399002e | 1925 | err = net_conf_from_attrs_for_change(new_conf, info); |
c75b9b10 | 1926 | if (err && err != -ENOMSG) { |
f399002e LE |
1927 | retcode = ERR_MANDATORY_TAG; |
1928 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1929 | goto fail; | |
1930 | } | |
1931 | ||
cd64397c PR |
1932 | retcode = check_net_options(tconn, new_conf); |
1933 | if (retcode != NO_ERROR) | |
1934 | goto fail; | |
1935 | ||
f399002e LE |
1936 | /* re-sync running */ |
1937 | rsr = conn_resync_running(tconn); | |
0fd0ea06 | 1938 | if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) { |
f399002e | 1939 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 1940 | goto fail; |
f399002e LE |
1941 | } |
1942 | ||
f399002e LE |
1943 | /* online verify running */ |
1944 | ovr = conn_ov_running(tconn); | |
0fd0ea06 PR |
1945 | if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) { |
1946 | retcode = ERR_VERIFY_RUNNING; | |
1947 | goto fail; | |
f399002e LE |
1948 | } |
1949 | ||
0fd0ea06 PR |
1950 | retcode = alloc_crypto(&crypto, new_conf); |
1951 | if (retcode != NO_ERROR) | |
1952 | goto fail; | |
f399002e | 1953 | |
44ed167d | 1954 | rcu_assign_pointer(tconn->net_conf, new_conf); |
f399002e LE |
1955 | |
1956 | if (!rsr) { | |
1957 | crypto_free_hash(tconn->csums_tfm); | |
0fd0ea06 PR |
1958 | tconn->csums_tfm = crypto.csums_tfm; |
1959 | crypto.csums_tfm = NULL; | |
f399002e LE |
1960 | } |
1961 | if (!ovr) { | |
1962 | crypto_free_hash(tconn->verify_tfm); | |
0fd0ea06 PR |
1963 | tconn->verify_tfm = crypto.verify_tfm; |
1964 | crypto.verify_tfm = NULL; | |
f399002e LE |
1965 | } |
1966 | ||
8d412fc6 AG |
1967 | crypto_free_hash(tconn->integrity_tfm); |
1968 | tconn->integrity_tfm = crypto.integrity_tfm; | |
d659f2aa | 1969 | if (tconn->cstate >= C_WF_REPORT_PARAMS && tconn->agreed_pro_version >= 100) |
88104ca4 | 1970 | /* Do this without trying to take tconn->data.mutex again. */ |
d659f2aa | 1971 | __drbd_send_protocol(tconn, P_PROTOCOL_UPDATE); |
0fd0ea06 | 1972 | |
0fd0ea06 PR |
1973 | crypto_free_hash(tconn->cram_hmac_tfm); |
1974 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
1975 | ||
a0095508 | 1976 | mutex_unlock(&tconn->conf_update); |
88104ca4 | 1977 | mutex_unlock(&tconn->data.mutex); |
91fd4dad PR |
1978 | synchronize_rcu(); |
1979 | kfree(old_conf); | |
1980 | ||
f399002e LE |
1981 | if (tconn->cstate >= C_WF_REPORT_PARAMS) |
1982 | drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn))); | |
1983 | ||
91fd4dad PR |
1984 | goto done; |
1985 | ||
f399002e | 1986 | fail: |
a0095508 | 1987 | mutex_unlock(&tconn->conf_update); |
88104ca4 | 1988 | mutex_unlock(&tconn->data.mutex); |
0fd0ea06 | 1989 | free_crypto(&crypto); |
f399002e | 1990 | kfree(new_conf); |
91fd4dad | 1991 | done: |
f399002e LE |
1992 | conn_reconfig_done(tconn); |
1993 | out: | |
1994 | drbd_adm_finish(info, retcode); | |
1995 | return 0; | |
1996 | } | |
1997 | ||
3b98c0c2 | 1998 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1999 | { |
3b98c0c2 | 2000 | struct drbd_conf *mdev; |
44ed167d | 2001 | struct net_conf *old_conf, *new_conf = NULL; |
0fd0ea06 | 2002 | struct crypto crypto = { }; |
3b98c0c2 | 2003 | struct drbd_tconn *tconn; |
3b98c0c2 LE |
2004 | enum drbd_ret_code retcode; |
2005 | int i; | |
2006 | int err; | |
b411b363 | 2007 | |
44e52cfa | 2008 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
089c075d | 2009 | |
3b98c0c2 LE |
2010 | if (!adm_ctx.reply_skb) |
2011 | return retcode; | |
2012 | if (retcode != NO_ERROR) | |
2013 | goto out; | |
089c075d AG |
2014 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
2015 | drbd_msg_put_info("connection endpoint(s) missing"); | |
2016 | retcode = ERR_INVALID_REQUEST; | |
2017 | goto out; | |
2018 | } | |
2019 | ||
2020 | /* No need for _rcu here. All reconfiguration is | |
2021 | * strictly serialized on genl_lock(). We are protected against | |
2022 | * concurrent reconfiguration/addition/deletion */ | |
2023 | list_for_each_entry(tconn, &drbd_tconns, all_tconn) { | |
2024 | if (nla_len(adm_ctx.my_addr) == tconn->my_addr_len && | |
2025 | !memcmp(nla_data(adm_ctx.my_addr), &tconn->my_addr, tconn->my_addr_len)) { | |
2026 | retcode = ERR_LOCAL_ADDR; | |
2027 | goto out; | |
2028 | } | |
2029 | ||
2030 | if (nla_len(adm_ctx.peer_addr) == tconn->peer_addr_len && | |
2031 | !memcmp(nla_data(adm_ctx.peer_addr), &tconn->peer_addr, tconn->peer_addr_len)) { | |
2032 | retcode = ERR_PEER_ADDR; | |
2033 | goto out; | |
2034 | } | |
2035 | } | |
3b98c0c2 LE |
2036 | |
2037 | tconn = adm_ctx.tconn; | |
80883197 | 2038 | conn_reconfig_start(tconn); |
b411b363 | 2039 | |
80883197 | 2040 | if (tconn->cstate > C_STANDALONE) { |
b411b363 PR |
2041 | retcode = ERR_NET_CONFIGURED; |
2042 | goto fail; | |
2043 | } | |
2044 | ||
a209b4ae | 2045 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
5979e361 | 2046 | new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL); |
b411b363 PR |
2047 | if (!new_conf) { |
2048 | retcode = ERR_NOMEM; | |
2049 | goto fail; | |
2050 | } | |
2051 | ||
b966b5dd | 2052 | set_net_conf_defaults(new_conf); |
f399002e LE |
2053 | |
2054 | err = net_conf_from_attrs(new_conf, info); | |
25e40932 | 2055 | if (err && err != -ENOMSG) { |
b411b363 | 2056 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2057 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2058 | goto fail; |
2059 | } | |
2060 | ||
cd64397c PR |
2061 | retcode = check_net_options(tconn, new_conf); |
2062 | if (retcode != NO_ERROR) | |
422028b1 | 2063 | goto fail; |
422028b1 | 2064 | |
0fd0ea06 PR |
2065 | retcode = alloc_crypto(&crypto, new_conf); |
2066 | if (retcode != NO_ERROR) | |
2067 | goto fail; | |
b411b363 | 2068 | |
b411b363 PR |
2069 | ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
2070 | ||
80883197 | 2071 | conn_flush_workqueue(tconn); |
91fd4dad | 2072 | |
a0095508 | 2073 | mutex_lock(&tconn->conf_update); |
91fd4dad PR |
2074 | old_conf = tconn->net_conf; |
2075 | if (old_conf) { | |
b411b363 | 2076 | retcode = ERR_NET_CONFIGURED; |
a0095508 | 2077 | mutex_unlock(&tconn->conf_update); |
b411b363 PR |
2078 | goto fail; |
2079 | } | |
44ed167d | 2080 | rcu_assign_pointer(tconn->net_conf, new_conf); |
b411b363 | 2081 | |
91fd4dad | 2082 | conn_free_crypto(tconn); |
0fd0ea06 | 2083 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; |
8d412fc6 | 2084 | tconn->integrity_tfm = crypto.integrity_tfm; |
0fd0ea06 PR |
2085 | tconn->csums_tfm = crypto.csums_tfm; |
2086 | tconn->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2087 | |
089c075d AG |
2088 | tconn->my_addr_len = nla_len(adm_ctx.my_addr); |
2089 | memcpy(&tconn->my_addr, nla_data(adm_ctx.my_addr), tconn->my_addr_len); | |
2090 | tconn->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2091 | memcpy(&tconn->peer_addr, nla_data(adm_ctx.peer_addr), tconn->peer_addr_len); | |
2092 | ||
a0095508 | 2093 | mutex_unlock(&tconn->conf_update); |
91fd4dad | 2094 | |
695d08fa | 2095 | rcu_read_lock(); |
80883197 PR |
2096 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
2097 | mdev->send_cnt = 0; | |
2098 | mdev->recv_cnt = 0; | |
80883197 | 2099 | } |
695d08fa | 2100 | rcu_read_unlock(); |
5ee743e9 LE |
2101 | |
2102 | retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE); | |
2103 | ||
80883197 | 2104 | conn_reconfig_done(tconn); |
3b98c0c2 | 2105 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2106 | return 0; |
2107 | ||
2108 | fail: | |
0fd0ea06 | 2109 | free_crypto(&crypto); |
b411b363 PR |
2110 | kfree(new_conf); |
2111 | ||
80883197 | 2112 | conn_reconfig_done(tconn); |
3b98c0c2 LE |
2113 | out: |
2114 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2115 | return 0; |
2116 | } | |
2117 | ||
85f75dd7 LE |
2118 | static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force) |
2119 | { | |
2120 | enum drbd_state_rv rv; | |
85f75dd7 | 2121 | |
f3dfa40a LE |
2122 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2123 | force ? CS_HARD : 0); | |
85f75dd7 LE |
2124 | |
2125 | switch (rv) { | |
2126 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2127 | break; |
85f75dd7 LE |
2128 | case SS_ALREADY_STANDALONE: |
2129 | return SS_SUCCESS; | |
2130 | case SS_PRIMARY_NOP: | |
2131 | /* Our state checking code wants to see the peer outdated. */ | |
2132 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
f3dfa40a | 2133 | pdsk, D_OUTDATED), CS_VERBOSE); |
85f75dd7 LE |
2134 | break; |
2135 | case SS_CW_FAILED_BY_PEER: | |
2136 | /* The peer probably wants to see us outdated. */ | |
2137 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
2138 | disk, D_OUTDATED), 0); | |
2139 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
f3dfa40a LE |
2140 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2141 | CS_HARD); | |
85f75dd7 LE |
2142 | } |
2143 | break; | |
2144 | default:; | |
2145 | /* no special handling necessary */ | |
2146 | } | |
2147 | ||
f3dfa40a LE |
2148 | if (rv >= SS_SUCCESS) { |
2149 | enum drbd_state_rv rv2; | |
2150 | /* No one else can reconfigure the network while I am here. | |
2151 | * The state handling only uses drbd_thread_stop_nowait(), | |
2152 | * we want to really wait here until the receiver is no more. | |
2153 | */ | |
2154 | drbd_thread_stop(&adm_ctx.tconn->receiver); | |
2155 | ||
2156 | /* Race breaker. This additional state change request may be | |
2157 | * necessary, if this was a forced disconnect during a receiver | |
2158 | * restart. We may have "killed" the receiver thread just | |
2159 | * after drbdd_init() returned. Typically, we should be | |
2160 | * C_STANDALONE already, now, and this becomes a no-op. | |
2161 | */ | |
2162 | rv2 = conn_request_state(tconn, NS(conn, C_STANDALONE), | |
2163 | CS_VERBOSE | CS_HARD); | |
2164 | if (rv2 < SS_SUCCESS) | |
2165 | conn_err(tconn, | |
2166 | "unexpected rv2=%d in conn_try_disconnect()\n", | |
2167 | rv2); | |
2168 | } | |
85f75dd7 LE |
2169 | return rv; |
2170 | } | |
2171 | ||
3b98c0c2 | 2172 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2173 | { |
3b98c0c2 LE |
2174 | struct disconnect_parms parms; |
2175 | struct drbd_tconn *tconn; | |
85f75dd7 | 2176 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2177 | enum drbd_ret_code retcode; |
2178 | int err; | |
2561b9c1 | 2179 | |
089c075d | 2180 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2181 | if (!adm_ctx.reply_skb) |
2182 | return retcode; | |
2183 | if (retcode != NO_ERROR) | |
2561b9c1 | 2184 | goto fail; |
3b98c0c2 LE |
2185 | |
2186 | tconn = adm_ctx.tconn; | |
2187 | memset(&parms, 0, sizeof(parms)); | |
2188 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2189 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2190 | if (err) { |
2191 | retcode = ERR_MANDATORY_TAG; | |
2192 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2193 | goto fail; | |
2194 | } | |
2561b9c1 PR |
2195 | } |
2196 | ||
85f75dd7 LE |
2197 | rv = conn_try_disconnect(tconn, parms.force_disconnect); |
2198 | if (rv < SS_SUCCESS) | |
f3dfa40a LE |
2199 | retcode = rv; /* FIXME: Type mismatch. */ |
2200 | else | |
2201 | retcode = NO_ERROR; | |
b411b363 | 2202 | fail: |
3b98c0c2 | 2203 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2204 | return 0; |
2205 | } | |
2206 | ||
2207 | void resync_after_online_grow(struct drbd_conf *mdev) | |
2208 | { | |
2209 | int iass; /* I am sync source */ | |
2210 | ||
2211 | dev_info(DEV, "Resync of new storage after online grow\n"); | |
2212 | if (mdev->state.role != mdev->state.peer) | |
2213 | iass = (mdev->state.role == R_PRIMARY); | |
2214 | else | |
25703f83 | 2215 | iass = test_bit(DISCARD_CONCURRENT, &mdev->tconn->flags); |
b411b363 PR |
2216 | |
2217 | if (iass) | |
2218 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
2219 | else | |
2220 | _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); | |
2221 | } | |
2222 | ||
3b98c0c2 | 2223 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2224 | { |
daeda1cc | 2225 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 LE |
2226 | struct resize_parms rs; |
2227 | struct drbd_conf *mdev; | |
2228 | enum drbd_ret_code retcode; | |
b411b363 | 2229 | enum determine_dev_size dd; |
6495d2c6 | 2230 | enum dds_flags ddsf; |
daeda1cc | 2231 | sector_t u_size; |
3b98c0c2 | 2232 | int err; |
b411b363 | 2233 | |
3b98c0c2 LE |
2234 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2235 | if (!adm_ctx.reply_skb) | |
2236 | return retcode; | |
2237 | if (retcode != NO_ERROR) | |
b411b363 | 2238 | goto fail; |
3b98c0c2 LE |
2239 | |
2240 | memset(&rs, 0, sizeof(struct resize_parms)); | |
2241 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { | |
f399002e | 2242 | err = resize_parms_from_attrs(&rs, info); |
3b98c0c2 LE |
2243 | if (err) { |
2244 | retcode = ERR_MANDATORY_TAG; | |
2245 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2246 | goto fail; | |
2247 | } | |
b411b363 PR |
2248 | } |
2249 | ||
3b98c0c2 | 2250 | mdev = adm_ctx.mdev; |
b411b363 PR |
2251 | if (mdev->state.conn > C_CONNECTED) { |
2252 | retcode = ERR_RESIZE_RESYNC; | |
2253 | goto fail; | |
2254 | } | |
2255 | ||
2256 | if (mdev->state.role == R_SECONDARY && | |
2257 | mdev->state.peer == R_SECONDARY) { | |
2258 | retcode = ERR_NO_PRIMARY; | |
2259 | goto fail; | |
2260 | } | |
2261 | ||
2262 | if (!get_ldev(mdev)) { | |
2263 | retcode = ERR_NO_DISK; | |
2264 | goto fail; | |
2265 | } | |
2266 | ||
31890f4a | 2267 | if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) { |
6495d2c6 | 2268 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2269 | goto fail_ldev; |
6495d2c6 PR |
2270 | } |
2271 | ||
daeda1cc PR |
2272 | rcu_read_lock(); |
2273 | u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size; | |
2274 | rcu_read_unlock(); | |
2275 | if (u_size != (sector_t)rs.resize_size) { | |
2276 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2277 | if (!new_disk_conf) { | |
2278 | retcode = ERR_NOMEM; | |
9bcd2521 | 2279 | goto fail_ldev; |
daeda1cc PR |
2280 | } |
2281 | } | |
2282 | ||
087c2492 | 2283 | if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) |
b411b363 | 2284 | mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); |
b411b363 | 2285 | |
daeda1cc PR |
2286 | if (new_disk_conf) { |
2287 | mutex_lock(&mdev->tconn->conf_update); | |
2288 | old_disk_conf = mdev->ldev->disk_conf; | |
2289 | *new_disk_conf = *old_disk_conf; | |
2290 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
2291 | rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf); | |
2292 | mutex_unlock(&mdev->tconn->conf_update); | |
2293 | synchronize_rcu(); | |
2294 | kfree(old_disk_conf); | |
2295 | } | |
2296 | ||
6495d2c6 | 2297 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
24c4830c | 2298 | dd = drbd_determine_dev_size(mdev, ddsf); |
b411b363 PR |
2299 | drbd_md_sync(mdev); |
2300 | put_ldev(mdev); | |
2301 | if (dd == dev_size_error) { | |
2302 | retcode = ERR_NOMEM_BITMAP; | |
2303 | goto fail; | |
2304 | } | |
2305 | ||
087c2492 | 2306 | if (mdev->state.conn == C_CONNECTED) { |
b411b363 PR |
2307 | if (dd == grew) |
2308 | set_bit(RESIZE_PENDING, &mdev->flags); | |
2309 | ||
2310 | drbd_send_uuids(mdev); | |
6495d2c6 | 2311 | drbd_send_sizes(mdev, 1, ddsf); |
b411b363 PR |
2312 | } |
2313 | ||
2314 | fail: | |
3b98c0c2 | 2315 | drbd_adm_finish(info, retcode); |
b411b363 | 2316 | return 0; |
9bcd2521 PR |
2317 | |
2318 | fail_ldev: | |
2319 | put_ldev(mdev); | |
2320 | goto fail; | |
b411b363 PR |
2321 | } |
2322 | ||
f399002e | 2323 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2324 | { |
3b98c0c2 | 2325 | enum drbd_ret_code retcode; |
f399002e | 2326 | struct drbd_tconn *tconn; |
b57a1e27 | 2327 | struct res_opts res_opts; |
f399002e | 2328 | int err; |
b411b363 | 2329 | |
44e52cfa | 2330 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2331 | if (!adm_ctx.reply_skb) |
2332 | return retcode; | |
2333 | if (retcode != NO_ERROR) | |
2334 | goto fail; | |
f399002e | 2335 | tconn = adm_ctx.tconn; |
3b98c0c2 | 2336 | |
b57a1e27 | 2337 | res_opts = tconn->res_opts; |
5979e361 | 2338 | if (should_set_defaults(info)) |
b966b5dd | 2339 | set_res_opts_defaults(&res_opts); |
b411b363 | 2340 | |
b57a1e27 | 2341 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2342 | if (err && err != -ENOMSG) { |
b411b363 | 2343 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2344 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2345 | goto fail; |
2346 | } | |
2347 | ||
afbbfa88 AG |
2348 | err = set_resource_options(tconn, &res_opts); |
2349 | if (err) { | |
2350 | retcode = ERR_INVALID_REQUEST; | |
2351 | if (err == -ENOMEM) | |
2352 | retcode = ERR_NOMEM; | |
b411b363 PR |
2353 | } |
2354 | ||
b411b363 | 2355 | fail: |
3b98c0c2 | 2356 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2357 | return 0; |
2358 | } | |
2359 | ||
3b98c0c2 | 2360 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2361 | { |
3b98c0c2 LE |
2362 | struct drbd_conf *mdev; |
2363 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2364 | ||
2365 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2366 | if (!adm_ctx.reply_skb) | |
2367 | return retcode; | |
2368 | if (retcode != NO_ERROR) | |
2369 | goto out; | |
2370 | ||
2371 | mdev = adm_ctx.mdev; | |
b411b363 | 2372 | |
194bfb32 LE |
2373 | /* If there is still bitmap IO pending, probably because of a previous |
2374 | * resync just being finished, wait for it before requesting a new resync. */ | |
5016b82a | 2375 | drbd_suspend_io(mdev); |
194bfb32 LE |
2376 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); |
2377 | ||
b411b363 PR |
2378 | retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED); |
2379 | ||
2380 | if (retcode < SS_SUCCESS && retcode != SS_NEED_CONNECTION) | |
2381 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
2382 | ||
2383 | while (retcode == SS_NEED_CONNECTION) { | |
87eeee41 | 2384 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2385 | if (mdev->state.conn < C_CONNECTED) |
2386 | retcode = _drbd_set_state(_NS(mdev, disk, D_INCONSISTENT), CS_VERBOSE, NULL); | |
87eeee41 | 2387 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2388 | |
2389 | if (retcode != SS_NEED_CONNECTION) | |
2390 | break; | |
2391 | ||
2392 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
2393 | } | |
5016b82a | 2394 | drbd_resume_io(mdev); |
b411b363 | 2395 | |
3b98c0c2 LE |
2396 | out: |
2397 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2398 | return 0; |
2399 | } | |
2400 | ||
3b98c0c2 LE |
2401 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2402 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2403 | { |
3b98c0c2 | 2404 | enum drbd_ret_code retcode; |
194bfb32 | 2405 | |
3b98c0c2 LE |
2406 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2407 | if (!adm_ctx.reply_skb) | |
2408 | return retcode; | |
2409 | if (retcode != NO_ERROR) | |
2410 | goto out; | |
b411b363 | 2411 | |
3b98c0c2 LE |
2412 | retcode = drbd_request_state(adm_ctx.mdev, mask, val); |
2413 | out: | |
2414 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2415 | return 0; |
2416 | } | |
2417 | ||
25b0d6c8 PR |
2418 | static int drbd_bmio_set_susp_al(struct drbd_conf *mdev) |
2419 | { | |
2420 | int rv; | |
2421 | ||
2422 | rv = drbd_bmio_set_n_write(mdev); | |
2423 | drbd_suspend_al(mdev); | |
2424 | return rv; | |
2425 | } | |
2426 | ||
3b98c0c2 | 2427 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2428 | { |
25b0d6c8 PR |
2429 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
2430 | struct drbd_conf *mdev; | |
2431 | ||
2432 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2433 | if (!adm_ctx.reply_skb) | |
2434 | return retcode; | |
2435 | if (retcode != NO_ERROR) | |
2436 | goto out; | |
2437 | ||
2438 | mdev = adm_ctx.mdev; | |
2439 | ||
5016b82a LE |
2440 | /* If there is still bitmap IO pending, probably because of a previous |
2441 | * resync just being finished, wait for it before requesting a new resync. */ | |
2442 | drbd_suspend_io(mdev); | |
2443 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); | |
2444 | ||
25b0d6c8 PR |
2445 | retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED); |
2446 | if (retcode < SS_SUCCESS) { | |
2447 | if (retcode == SS_NEED_CONNECTION && mdev->state.role == R_PRIMARY) { | |
2448 | /* The peer will get a resync upon connect anyways. | |
2449 | * Just make that into a full resync. */ | |
2450 | retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT)); | |
2451 | if (retcode >= SS_SUCCESS) { | |
2452 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al, | |
2453 | "set_n_write from invalidate_peer", | |
2454 | BM_LOCKED_SET_ALLOWED)) | |
2455 | retcode = ERR_IO_MD_DISK; | |
2456 | } | |
2457 | } else | |
2458 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S)); | |
2459 | } | |
5016b82a | 2460 | drbd_resume_io(mdev); |
25b0d6c8 PR |
2461 | |
2462 | out: | |
2463 | drbd_adm_finish(info, retcode); | |
2464 | return 0; | |
3b98c0c2 | 2465 | } |
b411b363 | 2466 | |
3b98c0c2 LE |
2467 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
2468 | { | |
2469 | enum drbd_ret_code retcode; | |
2470 | ||
2471 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2472 | if (!adm_ctx.reply_skb) | |
2473 | return retcode; | |
2474 | if (retcode != NO_ERROR) | |
2475 | goto out; | |
b411b363 | 2476 | |
3b98c0c2 LE |
2477 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
2478 | retcode = ERR_PAUSE_IS_SET; | |
2479 | out: | |
2480 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2481 | return 0; |
2482 | } | |
2483 | ||
3b98c0c2 | 2484 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2485 | { |
da9fbc27 | 2486 | union drbd_dev_state s; |
3b98c0c2 LE |
2487 | enum drbd_ret_code retcode; |
2488 | ||
2489 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2490 | if (!adm_ctx.reply_skb) | |
2491 | return retcode; | |
2492 | if (retcode != NO_ERROR) | |
2493 | goto out; | |
b411b363 | 2494 | |
3b98c0c2 LE |
2495 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2496 | s = adm_ctx.mdev->state; | |
cd88d030 PR |
2497 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2498 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2499 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2500 | } else { | |
2501 | retcode = ERR_PAUSE_IS_CLEAR; | |
2502 | } | |
2503 | } | |
b411b363 | 2504 | |
3b98c0c2 LE |
2505 | out: |
2506 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2507 | return 0; |
2508 | } | |
2509 | ||
3b98c0c2 | 2510 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2511 | { |
3b98c0c2 | 2512 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2513 | } |
2514 | ||
3b98c0c2 | 2515 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2516 | { |
3b98c0c2 LE |
2517 | struct drbd_conf *mdev; |
2518 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2519 | ||
2520 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2521 | if (!adm_ctx.reply_skb) | |
2522 | return retcode; | |
2523 | if (retcode != NO_ERROR) | |
2524 | goto out; | |
2525 | ||
2526 | mdev = adm_ctx.mdev; | |
43a5182c PR |
2527 | if (test_bit(NEW_CUR_UUID, &mdev->flags)) { |
2528 | drbd_uuid_new_current(mdev); | |
2529 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
43a5182c | 2530 | } |
265be2d0 | 2531 | drbd_suspend_io(mdev); |
3b98c0c2 LE |
2532 | retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); |
2533 | if (retcode == SS_SUCCESS) { | |
265be2d0 | 2534 | if (mdev->state.conn < C_CONNECTED) |
2f5cdd0b | 2535 | tl_clear(mdev->tconn); |
265be2d0 | 2536 | if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED) |
2f5cdd0b | 2537 | tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO); |
265be2d0 PR |
2538 | } |
2539 | drbd_resume_io(mdev); | |
2540 | ||
3b98c0c2 LE |
2541 | out: |
2542 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2543 | return 0; |
2544 | } | |
2545 | ||
3b98c0c2 | 2546 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2547 | { |
3b98c0c2 | 2548 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2549 | } |
2550 | ||
089c075d | 2551 | int nla_put_drbd_cfg_context(struct sk_buff *skb, struct drbd_tconn *tconn, unsigned vnr) |
543cc10b LE |
2552 | { |
2553 | struct nlattr *nla; | |
2554 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2555 | if (!nla) | |
2556 | goto nla_put_failure; | |
26ec9287 AG |
2557 | if (vnr != VOLUME_UNSPECIFIED && |
2558 | nla_put_u32(skb, T_ctx_volume, vnr)) | |
2559 | goto nla_put_failure; | |
2560 | if (nla_put_string(skb, T_ctx_resource_name, tconn->name)) | |
2561 | goto nla_put_failure; | |
2562 | if (tconn->my_addr_len && | |
2563 | nla_put(skb, T_ctx_my_addr, tconn->my_addr_len, &tconn->my_addr)) | |
2564 | goto nla_put_failure; | |
2565 | if (tconn->peer_addr_len && | |
2566 | nla_put(skb, T_ctx_peer_addr, tconn->peer_addr_len, &tconn->peer_addr)) | |
2567 | goto nla_put_failure; | |
543cc10b LE |
2568 | nla_nest_end(skb, nla); |
2569 | return 0; | |
2570 | ||
2571 | nla_put_failure: | |
2572 | if (nla) | |
2573 | nla_nest_cancel(skb, nla); | |
2574 | return -EMSGSIZE; | |
2575 | } | |
2576 | ||
3b98c0c2 LE |
2577 | int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev, |
2578 | const struct sib_info *sib) | |
b411b363 | 2579 | { |
3b98c0c2 | 2580 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
44ed167d | 2581 | struct net_conf *nc; |
3b98c0c2 LE |
2582 | struct nlattr *nla; |
2583 | int got_ldev; | |
3b98c0c2 LE |
2584 | int err = 0; |
2585 | int exclude_sensitive; | |
2586 | ||
2587 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2588 | * to. So we better exclude_sensitive information. | |
2589 | * | |
2590 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2591 | * in the context of the requesting user process. Exclude sensitive | |
2592 | * information, unless current has superuser. | |
2593 | * | |
2594 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2595 | * relies on the current implementation of netlink_dump(), which | |
2596 | * executes the dump callback successively from netlink_recvmsg(), | |
2597 | * always in the context of the receiving process */ | |
2598 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2599 | ||
2600 | got_ldev = get_ldev(mdev); | |
3b98c0c2 LE |
2601 | |
2602 | /* We need to add connection name and volume number information still. | |
2603 | * Minor number is in drbd_genlmsghdr. */ | |
089c075d | 2604 | if (nla_put_drbd_cfg_context(skb, mdev->tconn, mdev->vnr)) |
3b98c0c2 | 2605 | goto nla_put_failure; |
3b98c0c2 | 2606 | |
f399002e LE |
2607 | if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive)) |
2608 | goto nla_put_failure; | |
2609 | ||
daeda1cc | 2610 | rcu_read_lock(); |
3b98c0c2 | 2611 | if (got_ldev) |
daeda1cc | 2612 | if (disk_conf_to_skb(skb, rcu_dereference(mdev->ldev->disk_conf), exclude_sensitive)) |
3b98c0c2 | 2613 | goto nla_put_failure; |
44ed167d | 2614 | |
44ed167d PR |
2615 | nc = rcu_dereference(mdev->tconn->net_conf); |
2616 | if (nc) | |
2617 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2618 | rcu_read_unlock(); | |
2619 | if (err) | |
2620 | goto nla_put_failure; | |
3b98c0c2 | 2621 | |
3b98c0c2 LE |
2622 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2623 | if (!nla) | |
2624 | goto nla_put_failure; | |
26ec9287 AG |
2625 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
2626 | nla_put_u32(skb, T_current_state, mdev->state.i) || | |
2627 | nla_put_u64(skb, T_ed_uuid, mdev->ed_uuid) || | |
2628 | nla_put_u64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev))) | |
2629 | goto nla_put_failure; | |
3b98c0c2 LE |
2630 | |
2631 | if (got_ldev) { | |
26ec9287 AG |
2632 | if (nla_put_u32(skb, T_disk_flags, mdev->ldev->md.flags) || |
2633 | nla_put(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid) || | |
2634 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(mdev)) || | |
2635 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(mdev))) | |
2636 | goto nla_put_failure; | |
3b98c0c2 LE |
2637 | if (C_SYNC_SOURCE <= mdev->state.conn && |
2638 | C_PAUSED_SYNC_T >= mdev->state.conn) { | |
26ec9287 AG |
2639 | if (nla_put_u64(skb, T_bits_rs_total, mdev->rs_total) || |
2640 | nla_put_u64(skb, T_bits_rs_failed, mdev->rs_failed)) | |
2641 | goto nla_put_failure; | |
3b98c0c2 | 2642 | } |
b411b363 PR |
2643 | } |
2644 | ||
3b98c0c2 LE |
2645 | if (sib) { |
2646 | switch(sib->sib_reason) { | |
2647 | case SIB_SYNC_PROGRESS: | |
2648 | case SIB_GET_STATUS_REPLY: | |
2649 | break; | |
2650 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2651 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2652 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2653 | goto nla_put_failure; | |
3b98c0c2 LE |
2654 | break; |
2655 | case SIB_HELPER_POST: | |
26ec9287 AG |
2656 | if (nla_put_u32(skb, T_helper_exit_code, |
2657 | sib->helper_exit_code)) | |
2658 | goto nla_put_failure; | |
3b98c0c2 LE |
2659 | /* fall through */ |
2660 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2661 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2662 | goto nla_put_failure; | |
3b98c0c2 LE |
2663 | break; |
2664 | } | |
b411b363 | 2665 | } |
3b98c0c2 | 2666 | nla_nest_end(skb, nla); |
b411b363 | 2667 | |
3b98c0c2 LE |
2668 | if (0) |
2669 | nla_put_failure: | |
2670 | err = -EMSGSIZE; | |
2671 | if (got_ldev) | |
2672 | put_ldev(mdev); | |
3b98c0c2 | 2673 | return err; |
b411b363 PR |
2674 | } |
2675 | ||
3b98c0c2 | 2676 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2677 | { |
3b98c0c2 LE |
2678 | enum drbd_ret_code retcode; |
2679 | int err; | |
b411b363 | 2680 | |
3b98c0c2 LE |
2681 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2682 | if (!adm_ctx.reply_skb) | |
2683 | return retcode; | |
2684 | if (retcode != NO_ERROR) | |
2685 | goto out; | |
b411b363 | 2686 | |
3b98c0c2 LE |
2687 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL); |
2688 | if (err) { | |
2689 | nlmsg_free(adm_ctx.reply_skb); | |
2690 | return err; | |
b411b363 | 2691 | } |
3b98c0c2 LE |
2692 | out: |
2693 | drbd_adm_finish(info, retcode); | |
2694 | return 0; | |
b411b363 PR |
2695 | } |
2696 | ||
71932efc | 2697 | int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2698 | { |
3b98c0c2 LE |
2699 | struct drbd_conf *mdev; |
2700 | struct drbd_genlmsghdr *dh; | |
543cc10b LE |
2701 | struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0]; |
2702 | struct drbd_tconn *tconn = NULL; | |
2703 | struct drbd_tconn *tmp; | |
2704 | unsigned volume = cb->args[1]; | |
2705 | ||
2706 | /* Open coded, deferred, iteration: | |
2707 | * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) { | |
2708 | * idr_for_each_entry(&tconn->volumes, mdev, i) { | |
2709 | * ... | |
2710 | * } | |
2711 | * } | |
2712 | * where tconn is cb->args[0]; | |
2713 | * and i is cb->args[1]; | |
2714 | * | |
71932efc LE |
2715 | * cb->args[2] indicates if we shall loop over all resources, |
2716 | * or just dump all volumes of a single resource. | |
2717 | * | |
3b98c0c2 LE |
2718 | * This may miss entries inserted after this dump started, |
2719 | * or entries deleted before they are reached. | |
543cc10b LE |
2720 | * |
2721 | * We need to make sure the mdev won't disappear while | |
2722 | * we are looking at it, and revalidate our iterators | |
2723 | * on each iteration. | |
2724 | */ | |
3b98c0c2 | 2725 | |
9dc9fbb3 | 2726 | /* synchronize with conn_create()/conn_destroy() */ |
c141ebda | 2727 | rcu_read_lock(); |
543cc10b | 2728 | /* revalidate iterator position */ |
ec0bddbc | 2729 | list_for_each_entry_rcu(tmp, &drbd_tconns, all_tconn) { |
543cc10b LE |
2730 | if (pos == NULL) { |
2731 | /* first iteration */ | |
2732 | pos = tmp; | |
2733 | tconn = pos; | |
2734 | break; | |
2735 | } | |
2736 | if (tmp == pos) { | |
2737 | tconn = pos; | |
2738 | break; | |
2739 | } | |
2740 | } | |
2741 | if (tconn) { | |
71932efc | 2742 | next_tconn: |
543cc10b LE |
2743 | mdev = idr_get_next(&tconn->volumes, &volume); |
2744 | if (!mdev) { | |
2745 | /* No more volumes to dump on this tconn. | |
2746 | * Advance tconn iterator. */ | |
ec0bddbc PR |
2747 | pos = list_entry_rcu(tconn->all_tconn.next, |
2748 | struct drbd_tconn, all_tconn); | |
71932efc | 2749 | /* Did we dump any volume on this tconn yet? */ |
543cc10b | 2750 | if (volume != 0) { |
71932efc LE |
2751 | /* If we reached the end of the list, |
2752 | * or only a single resource dump was requested, | |
2753 | * we are done. */ | |
2754 | if (&pos->all_tconn == &drbd_tconns || cb->args[2]) | |
2755 | goto out; | |
543cc10b | 2756 | volume = 0; |
71932efc | 2757 | tconn = pos; |
543cc10b LE |
2758 | goto next_tconn; |
2759 | } | |
2760 | } | |
2761 | ||
3b98c0c2 LE |
2762 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).pid, |
2763 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
2764 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
2765 | if (!dh) | |
543cc10b LE |
2766 | goto out; |
2767 | ||
2768 | if (!mdev) { | |
367d675d LE |
2769 | /* This is a tconn without a single volume. |
2770 | * Suprisingly enough, it may have a network | |
2771 | * configuration. */ | |
2772 | struct net_conf *nc; | |
543cc10b LE |
2773 | dh->minor = -1U; |
2774 | dh->ret_code = NO_ERROR; | |
089c075d | 2775 | if (nla_put_drbd_cfg_context(skb, tconn, VOLUME_UNSPECIFIED)) |
367d675d LE |
2776 | goto cancel; |
2777 | nc = rcu_dereference(tconn->net_conf); | |
2778 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
2779 | goto cancel; | |
2780 | goto done; | |
543cc10b | 2781 | } |
3b98c0c2 | 2782 | |
543cc10b LE |
2783 | D_ASSERT(mdev->vnr == volume); |
2784 | D_ASSERT(mdev->tconn == tconn); | |
3b98c0c2 | 2785 | |
543cc10b | 2786 | dh->minor = mdev_to_minor(mdev); |
3b98c0c2 LE |
2787 | dh->ret_code = NO_ERROR; |
2788 | ||
2789 | if (nla_put_status_info(skb, mdev, NULL)) { | |
367d675d | 2790 | cancel: |
3b98c0c2 | 2791 | genlmsg_cancel(skb, dh); |
543cc10b | 2792 | goto out; |
3b98c0c2 | 2793 | } |
367d675d | 2794 | done: |
3b98c0c2 LE |
2795 | genlmsg_end(skb, dh); |
2796 | } | |
b411b363 | 2797 | |
543cc10b | 2798 | out: |
c141ebda | 2799 | rcu_read_unlock(); |
543cc10b LE |
2800 | /* where to start the next iteration */ |
2801 | cb->args[0] = (long)pos; | |
2802 | cb->args[1] = (pos == tconn) ? volume + 1 : 0; | |
b411b363 | 2803 | |
543cc10b LE |
2804 | /* No more tconns/volumes/minors found results in an empty skb. |
2805 | * Which will terminate the dump. */ | |
3b98c0c2 | 2806 | return skb->len; |
b411b363 PR |
2807 | } |
2808 | ||
71932efc LE |
2809 | /* |
2810 | * Request status of all resources, or of all volumes within a single resource. | |
2811 | * | |
2812 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
2813 | * Which means we cannot use the family->attrbuf or other such members, because | |
2814 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
2815 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
2816 | * | |
2817 | * Once things are setup properly, we call into get_one_status(). | |
2818 | */ | |
2819 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) | |
2820 | { | |
2821 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; | |
2822 | struct nlattr *nla; | |
7c3063cc | 2823 | const char *resource_name; |
71932efc | 2824 | struct drbd_tconn *tconn; |
7c3063cc | 2825 | int maxtype; |
71932efc LE |
2826 | |
2827 | /* Is this a followup call? */ | |
2828 | if (cb->args[0]) { | |
2829 | /* ... of a single resource dump, | |
2830 | * and the resource iterator has been advanced already? */ | |
2831 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
2832 | return 0; /* DONE. */ | |
2833 | goto dump; | |
2834 | } | |
2835 | ||
2836 | /* First call (from netlink_dump_start). We need to figure out | |
2837 | * which resource(s) the user wants us to dump. */ | |
2838 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
2839 | nlmsg_attrlen(cb->nlh, hdrlen), | |
2840 | DRBD_NLA_CFG_CONTEXT); | |
2841 | ||
2842 | /* No explicit context given. Dump all. */ | |
2843 | if (!nla) | |
2844 | goto dump; | |
7c3063cc AG |
2845 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
2846 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
2847 | if (IS_ERR(nla)) | |
2848 | return PTR_ERR(nla); | |
71932efc LE |
2849 | /* context given, but no name present? */ |
2850 | if (!nla) | |
2851 | return -EINVAL; | |
7c3063cc AG |
2852 | resource_name = nla_data(nla); |
2853 | tconn = conn_get_by_name(resource_name); | |
0ace9dfa | 2854 | |
71932efc LE |
2855 | if (!tconn) |
2856 | return -ENODEV; | |
2857 | ||
0ace9dfa PR |
2858 | kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */ |
2859 | ||
71932efc LE |
2860 | /* prime iterators, and set "filter" mode mark: |
2861 | * only dump this tconn. */ | |
2862 | cb->args[0] = (long)tconn; | |
2863 | /* cb->args[1] = 0; passed in this way. */ | |
2864 | cb->args[2] = (long)tconn; | |
2865 | ||
2866 | dump: | |
2867 | return get_one_status(skb, cb); | |
2868 | } | |
2869 | ||
3b98c0c2 | 2870 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2871 | { |
3b98c0c2 LE |
2872 | enum drbd_ret_code retcode; |
2873 | struct timeout_parms tp; | |
2874 | int err; | |
b411b363 | 2875 | |
3b98c0c2 LE |
2876 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2877 | if (!adm_ctx.reply_skb) | |
2878 | return retcode; | |
2879 | if (retcode != NO_ERROR) | |
2880 | goto out; | |
b411b363 | 2881 | |
3b98c0c2 LE |
2882 | tp.timeout_type = |
2883 | adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : | |
2884 | test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED : | |
2885 | UT_DEFAULT; | |
b411b363 | 2886 | |
3b98c0c2 LE |
2887 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
2888 | if (err) { | |
2889 | nlmsg_free(adm_ctx.reply_skb); | |
2890 | return err; | |
2891 | } | |
2892 | out: | |
2893 | drbd_adm_finish(info, retcode); | |
2894 | return 0; | |
b411b363 PR |
2895 | } |
2896 | ||
3b98c0c2 | 2897 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2898 | { |
3b98c0c2 LE |
2899 | struct drbd_conf *mdev; |
2900 | enum drbd_ret_code retcode; | |
b411b363 | 2901 | |
3b98c0c2 LE |
2902 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2903 | if (!adm_ctx.reply_skb) | |
2904 | return retcode; | |
2905 | if (retcode != NO_ERROR) | |
2906 | goto out; | |
873b0d5f | 2907 | |
3b98c0c2 LE |
2908 | mdev = adm_ctx.mdev; |
2909 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { | |
2910 | /* resume from last known position, if possible */ | |
2911 | struct start_ov_parms parms = | |
2912 | { .ov_start_sector = mdev->ov_start_sector }; | |
f399002e | 2913 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2914 | if (err) { |
2915 | retcode = ERR_MANDATORY_TAG; | |
2916 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2917 | goto out; | |
2918 | } | |
2919 | /* w_make_ov_request expects position to be aligned */ | |
2920 | mdev->ov_start_sector = parms.ov_start_sector & ~BM_SECT_PER_BIT; | |
2921 | } | |
873b0d5f LE |
2922 | /* If there is still bitmap IO pending, e.g. previous resync or verify |
2923 | * just being finished, wait for it before requesting a new resync. */ | |
5016b82a | 2924 | drbd_suspend_io(mdev); |
873b0d5f | 2925 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); |
3b98c0c2 | 2926 | retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S)); |
5016b82a | 2927 | drbd_resume_io(mdev); |
3b98c0c2 LE |
2928 | out: |
2929 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2930 | return 0; |
2931 | } | |
2932 | ||
2933 | ||
3b98c0c2 | 2934 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2935 | { |
3b98c0c2 LE |
2936 | struct drbd_conf *mdev; |
2937 | enum drbd_ret_code retcode; | |
b411b363 PR |
2938 | int skip_initial_sync = 0; |
2939 | int err; | |
3b98c0c2 | 2940 | struct new_c_uuid_parms args; |
b411b363 | 2941 | |
3b98c0c2 LE |
2942 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2943 | if (!adm_ctx.reply_skb) | |
2944 | return retcode; | |
2945 | if (retcode != NO_ERROR) | |
2946 | goto out_nolock; | |
b411b363 | 2947 | |
3b98c0c2 LE |
2948 | mdev = adm_ctx.mdev; |
2949 | memset(&args, 0, sizeof(args)); | |
2950 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 2951 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
2952 | if (err) { |
2953 | retcode = ERR_MANDATORY_TAG; | |
2954 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
2955 | goto out_nolock; | |
2956 | } | |
b411b363 PR |
2957 | } |
2958 | ||
8410da8f | 2959 | mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 PR |
2960 | |
2961 | if (!get_ldev(mdev)) { | |
2962 | retcode = ERR_NO_DISK; | |
2963 | goto out; | |
2964 | } | |
2965 | ||
2966 | /* this is "skip initial sync", assume to be clean */ | |
31890f4a | 2967 | if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 && |
b411b363 PR |
2968 | mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
2969 | dev_info(DEV, "Preparing to skip initial sync\n"); | |
2970 | skip_initial_sync = 1; | |
2971 | } else if (mdev->state.conn != C_STANDALONE) { | |
2972 | retcode = ERR_CONNECTED; | |
2973 | goto out_dec; | |
2974 | } | |
2975 | ||
2976 | drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ | |
2977 | drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */ | |
2978 | ||
2979 | if (args.clear_bm) { | |
20ceb2b2 LE |
2980 | err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, |
2981 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); | |
b411b363 PR |
2982 | if (err) { |
2983 | dev_err(DEV, "Writing bitmap failed with %d\n",err); | |
2984 | retcode = ERR_IO_MD_DISK; | |
2985 | } | |
2986 | if (skip_initial_sync) { | |
2987 | drbd_send_uuids_skip_initial_sync(mdev); | |
2988 | _drbd_uuid_set(mdev, UI_BITMAP, 0); | |
62b0da3a | 2989 | drbd_print_uuids(mdev, "cleared bitmap UUID"); |
87eeee41 | 2990 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2991 | _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
2992 | CS_VERBOSE, NULL); | |
87eeee41 | 2993 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
2994 | } |
2995 | } | |
2996 | ||
2997 | drbd_md_sync(mdev); | |
2998 | out_dec: | |
2999 | put_ldev(mdev); | |
3000 | out: | |
8410da8f | 3001 | mutex_unlock(mdev->state_mutex); |
3b98c0c2 LE |
3002 | out_nolock: |
3003 | drbd_adm_finish(info, retcode); | |
774b3055 PR |
3004 | return 0; |
3005 | } | |
3006 | ||
3b98c0c2 | 3007 | static enum drbd_ret_code |
7c3063cc | 3008 | drbd_check_resource_name(const char *name) |
774b3055 | 3009 | { |
3b98c0c2 | 3010 | if (!name || !name[0]) { |
7c3063cc | 3011 | drbd_msg_put_info("resource name missing"); |
3b98c0c2 | 3012 | return ERR_MANDATORY_TAG; |
774b3055 | 3013 | } |
3b98c0c2 LE |
3014 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3015 | * we must not allow slashes */ | |
3016 | if (strchr(name, '/')) { | |
7c3063cc | 3017 | drbd_msg_put_info("invalid resource name"); |
3b98c0c2 | 3018 | return ERR_INVALID_REQUEST; |
774b3055 | 3019 | } |
3b98c0c2 | 3020 | return NO_ERROR; |
774b3055 PR |
3021 | } |
3022 | ||
789c1b62 | 3023 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3024 | { |
3b98c0c2 | 3025 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3026 | struct res_opts res_opts; |
3027 | int err; | |
9f5180e5 | 3028 | |
3b98c0c2 LE |
3029 | retcode = drbd_adm_prepare(skb, info, 0); |
3030 | if (!adm_ctx.reply_skb) | |
3031 | return retcode; | |
3032 | if (retcode != NO_ERROR) | |
3033 | goto out; | |
b411b363 | 3034 | |
afbbfa88 AG |
3035 | set_res_opts_defaults(&res_opts); |
3036 | err = res_opts_from_attrs(&res_opts, info); | |
3037 | if (err && err != -ENOMSG) { | |
3038 | retcode = ERR_MANDATORY_TAG; | |
3039 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3040 | goto out; | |
3041 | } | |
3042 | ||
7c3063cc | 3043 | retcode = drbd_check_resource_name(adm_ctx.resource_name); |
3b98c0c2 LE |
3044 | if (retcode != NO_ERROR) |
3045 | goto out; | |
b411b363 | 3046 | |
3b98c0c2 | 3047 | if (adm_ctx.tconn) { |
38f19616 LE |
3048 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3049 | retcode = ERR_INVALID_REQUEST; | |
789c1b62 | 3050 | drbd_msg_put_info("resource exists"); |
38f19616 LE |
3051 | } |
3052 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3053 | goto out; |
b411b363 PR |
3054 | } |
3055 | ||
afbbfa88 | 3056 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3057 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
3058 | out: |
3059 | drbd_adm_finish(info, retcode); | |
3060 | return 0; | |
b411b363 PR |
3061 | } |
3062 | ||
3b98c0c2 | 3063 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3064 | { |
3b98c0c2 LE |
3065 | struct drbd_genlmsghdr *dh = info->userhdr; |
3066 | enum drbd_ret_code retcode; | |
b411b363 | 3067 | |
44e52cfa | 3068 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3069 | if (!adm_ctx.reply_skb) |
3070 | return retcode; | |
3071 | if (retcode != NO_ERROR) | |
3072 | goto out; | |
b411b363 | 3073 | |
f2257a56 | 3074 | if (dh->minor > MINORMASK) { |
3b98c0c2 LE |
3075 | drbd_msg_put_info("requested minor out of range"); |
3076 | retcode = ERR_INVALID_REQUEST; | |
3077 | goto out; | |
b411b363 | 3078 | } |
0c8e36d9 | 3079 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
3080 | drbd_msg_put_info("requested volume id out of range"); |
3081 | retcode = ERR_INVALID_REQUEST; | |
3082 | goto out; | |
b411b363 | 3083 | } |
b411b363 | 3084 | |
38f19616 LE |
3085 | /* drbd_adm_prepare made sure already |
3086 | * that mdev->tconn and mdev->vnr match the request. */ | |
3087 | if (adm_ctx.mdev) { | |
3088 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) | |
3089 | retcode = ERR_MINOR_EXISTS; | |
3090 | /* else: still NO_ERROR */ | |
3091 | goto out; | |
3092 | } | |
3093 | ||
3b98c0c2 LE |
3094 | retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume); |
3095 | out: | |
3096 | drbd_adm_finish(info, retcode); | |
3097 | return 0; | |
b411b363 PR |
3098 | } |
3099 | ||
85f75dd7 LE |
3100 | static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev) |
3101 | { | |
3102 | if (mdev->state.disk == D_DISKLESS && | |
3103 | /* no need to be mdev->state.conn == C_STANDALONE && | |
3104 | * we may want to delete a minor from a live replication group. | |
3105 | */ | |
3106 | mdev->state.role == R_SECONDARY) { | |
369bea63 PR |
3107 | _drbd_request_state(mdev, NS(conn, C_WF_REPORT_PARAMS), |
3108 | CS_VERBOSE + CS_WAIT_COMPLETE); | |
81fa2e67 PR |
3109 | idr_remove(&mdev->tconn->volumes, mdev->vnr); |
3110 | idr_remove(&minors, mdev_to_minor(mdev)); | |
3111 | del_gendisk(mdev->vdisk); | |
3112 | synchronize_rcu(); | |
3113 | kref_put(&mdev->kref, &drbd_minor_destroy); | |
85f75dd7 LE |
3114 | return NO_ERROR; |
3115 | } else | |
3116 | return ERR_MINOR_CONFIGURED; | |
3117 | } | |
3118 | ||
3b98c0c2 | 3119 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3120 | { |
3b98c0c2 | 3121 | enum drbd_ret_code retcode; |
b411b363 | 3122 | |
3b98c0c2 LE |
3123 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3124 | if (!adm_ctx.reply_skb) | |
3125 | return retcode; | |
3126 | if (retcode != NO_ERROR) | |
3127 | goto out; | |
b411b363 | 3128 | |
85f75dd7 | 3129 | retcode = adm_delete_minor(adm_ctx.mdev); |
85f75dd7 LE |
3130 | out: |
3131 | drbd_adm_finish(info, retcode); | |
3132 | return 0; | |
3133 | } | |
3134 | ||
3135 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) | |
3136 | { | |
f3dfa40a | 3137 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 LE |
3138 | struct drbd_conf *mdev; |
3139 | unsigned i; | |
3140 | ||
3141 | retcode = drbd_adm_prepare(skb, info, 0); | |
3142 | if (!adm_ctx.reply_skb) | |
3143 | return retcode; | |
3144 | if (retcode != NO_ERROR) | |
3145 | goto out; | |
3146 | ||
3147 | if (!adm_ctx.tconn) { | |
789c1b62 | 3148 | retcode = ERR_RES_NOT_KNOWN; |
85f75dd7 LE |
3149 | goto out; |
3150 | } | |
3151 | ||
85f75dd7 LE |
3152 | /* demote */ |
3153 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
3154 | retcode = drbd_set_role(mdev, R_SECONDARY, 0); | |
3155 | if (retcode < SS_SUCCESS) { | |
3156 | drbd_msg_put_info("failed to demote"); | |
c141ebda | 3157 | goto out; |
85f75dd7 LE |
3158 | } |
3159 | } | |
3160 | ||
f3dfa40a LE |
3161 | retcode = conn_try_disconnect(adm_ctx.tconn, 0); |
3162 | if (retcode < SS_SUCCESS) { | |
85f75dd7 | 3163 | drbd_msg_put_info("failed to disconnect"); |
f3dfa40a | 3164 | goto out; |
85f75dd7 LE |
3165 | } |
3166 | ||
3167 | /* detach */ | |
3168 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
cdfda633 | 3169 | retcode = adm_detach(mdev, 0); |
f3dfa40a | 3170 | if (retcode < SS_SUCCESS) { |
85f75dd7 | 3171 | drbd_msg_put_info("failed to detach"); |
c141ebda | 3172 | goto out; |
85f75dd7 LE |
3173 | } |
3174 | } | |
3175 | ||
f3dfa40a LE |
3176 | /* If we reach this, all volumes (of this tconn) are Secondary, |
3177 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have | |
c141ebda | 3178 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
f3dfa40a LE |
3179 | drbd_thread_stop(&adm_ctx.tconn->worker); |
3180 | ||
3181 | /* Now, nothing can fail anymore */ | |
3182 | ||
85f75dd7 LE |
3183 | /* delete volumes */ |
3184 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
3185 | retcode = adm_delete_minor(mdev); | |
3186 | if (retcode != NO_ERROR) { | |
3187 | /* "can not happen" */ | |
3188 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 | 3189 | goto out; |
85f75dd7 LE |
3190 | } |
3191 | } | |
3192 | ||
85f75dd7 LE |
3193 | /* delete connection */ |
3194 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { | |
ec0bddbc PR |
3195 | list_del_rcu(&adm_ctx.tconn->all_tconn); |
3196 | synchronize_rcu(); | |
9dc9fbb3 PR |
3197 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); |
3198 | ||
85f75dd7 LE |
3199 | retcode = NO_ERROR; |
3200 | } else { | |
3201 | /* "can not happen" */ | |
789c1b62 | 3202 | retcode = ERR_RES_IN_USE; |
85f75dd7 | 3203 | drbd_msg_put_info("failed to delete connection"); |
85f75dd7 | 3204 | } |
ef356262 | 3205 | goto out; |
3b98c0c2 LE |
3206 | out: |
3207 | drbd_adm_finish(info, retcode); | |
3208 | return 0; | |
b411b363 PR |
3209 | } |
3210 | ||
789c1b62 | 3211 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3212 | { |
3b98c0c2 | 3213 | enum drbd_ret_code retcode; |
b411b363 | 3214 | |
44e52cfa | 3215 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3216 | if (!adm_ctx.reply_skb) |
3217 | return retcode; | |
3218 | if (retcode != NO_ERROR) | |
3219 | goto out; | |
3220 | ||
3221 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { | |
ec0bddbc PR |
3222 | list_del_rcu(&adm_ctx.tconn->all_tconn); |
3223 | synchronize_rcu(); | |
9dc9fbb3 PR |
3224 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); |
3225 | ||
3b98c0c2 LE |
3226 | retcode = NO_ERROR; |
3227 | } else { | |
789c1b62 | 3228 | retcode = ERR_RES_IN_USE; |
b411b363 PR |
3229 | } |
3230 | ||
992d6e91 LE |
3231 | if (retcode == NO_ERROR) |
3232 | drbd_thread_stop(&adm_ctx.tconn->worker); | |
3b98c0c2 LE |
3233 | out: |
3234 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3235 | return 0; |
3236 | } | |
3237 | ||
3b98c0c2 | 3238 | void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib) |
b411b363 | 3239 | { |
3b98c0c2 LE |
3240 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3241 | struct sk_buff *msg; | |
3242 | struct drbd_genlmsghdr *d_out; | |
3243 | unsigned seq; | |
3244 | int err = -ENOMEM; | |
3245 | ||
3246 | seq = atomic_inc_return(&drbd_genl_seq); | |
3247 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3248 | if (!msg) | |
3249 | goto failed; | |
3250 | ||
3251 | err = -EMSGSIZE; | |
3252 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3253 | if (!d_out) /* cannot happen, but anyways. */ | |
3254 | goto nla_put_failure; | |
3255 | d_out->minor = mdev_to_minor(mdev); | |
6f9b5f84 | 3256 | d_out->ret_code = NO_ERROR; |
3b98c0c2 LE |
3257 | |
3258 | if (nla_put_status_info(msg, mdev, sib)) | |
3259 | goto nla_put_failure; | |
3260 | genlmsg_end(msg, d_out); | |
3261 | err = drbd_genl_multicast_events(msg, 0); | |
3262 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3263 | if (err && err != -ESRCH) | |
3264 | goto failed; | |
b411b363 | 3265 | |
3b98c0c2 | 3266 | return; |
b411b363 | 3267 | |
3b98c0c2 LE |
3268 | nla_put_failure: |
3269 | nlmsg_free(msg); | |
3270 | failed: | |
3271 | dev_err(DEV, "Error %d while broadcasting event. " | |
3272 | "Event seq:%u sib_reason:%u\n", | |
3273 | err, seq, sib->sib_reason); | |
b411b363 | 3274 | } |