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" :( */ | |
98683650 | 158 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
159 | return -EPERM; |
160 | ||
161 | adm_ctx.reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); | |
1e2a2551 AG |
162 | if (!adm_ctx.reply_skb) { |
163 | err = -ENOMEM; | |
3b98c0c2 | 164 | goto fail; |
1e2a2551 | 165 | } |
3b98c0c2 LE |
166 | |
167 | adm_ctx.reply_dh = genlmsg_put_reply(adm_ctx.reply_skb, | |
168 | info, &drbd_genl_family, 0, cmd); | |
169 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
170 | * but anyways */ | |
1e2a2551 AG |
171 | if (!adm_ctx.reply_dh) { |
172 | err = -ENOMEM; | |
3b98c0c2 | 173 | goto fail; |
1e2a2551 | 174 | } |
3b98c0c2 LE |
175 | |
176 | adm_ctx.reply_dh->minor = d_in->minor; | |
177 | adm_ctx.reply_dh->ret_code = NO_ERROR; | |
178 | ||
089c075d | 179 | adm_ctx.volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
180 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
181 | struct nlattr *nla; | |
182 | /* parse and validate only */ | |
f399002e | 183 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
184 | if (err) |
185 | goto fail; | |
186 | ||
187 | /* It was present, and valid, | |
188 | * copy it over to the reply skb. */ | |
189 | err = nla_put_nohdr(adm_ctx.reply_skb, | |
190 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, | |
191 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
192 | if (err) | |
193 | goto fail; | |
194 | ||
195 | /* and assign stuff to the global adm_ctx */ | |
196 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; | |
089c075d AG |
197 | if (nla) |
198 | adm_ctx.volume = nla_get_u32(nla); | |
7c3063cc | 199 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 200 | if (nla) |
7c3063cc | 201 | adm_ctx.resource_name = nla_data(nla); |
089c075d AG |
202 | adm_ctx.my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; |
203 | adm_ctx.peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
204 | if ((adm_ctx.my_addr && | |
205 | nla_len(adm_ctx.my_addr) > sizeof(adm_ctx.tconn->my_addr)) || | |
206 | (adm_ctx.peer_addr && | |
207 | nla_len(adm_ctx.peer_addr) > sizeof(adm_ctx.tconn->peer_addr))) { | |
208 | err = -EINVAL; | |
209 | goto fail; | |
210 | } | |
211 | } | |
3b98c0c2 LE |
212 | |
213 | adm_ctx.minor = d_in->minor; | |
214 | adm_ctx.mdev = minor_to_mdev(d_in->minor); | |
7c3063cc | 215 | adm_ctx.tconn = conn_get_by_name(adm_ctx.resource_name); |
3b98c0c2 LE |
216 | |
217 | if (!adm_ctx.mdev && (flags & DRBD_ADM_NEED_MINOR)) { | |
218 | drbd_msg_put_info("unknown minor"); | |
219 | return ERR_MINOR_INVALID; | |
220 | } | |
44e52cfa AG |
221 | if (!adm_ctx.tconn && (flags & DRBD_ADM_NEED_RESOURCE)) { |
222 | drbd_msg_put_info("unknown resource"); | |
3b98c0c2 LE |
223 | return ERR_INVALID_REQUEST; |
224 | } | |
225 | ||
089c075d AG |
226 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
227 | if (adm_ctx.tconn && !(flags & DRBD_ADM_NEED_RESOURCE)) { | |
228 | drbd_msg_put_info("no resource name expected"); | |
229 | return ERR_INVALID_REQUEST; | |
230 | } | |
231 | if (adm_ctx.mdev) { | |
232 | drbd_msg_put_info("no minor number expected"); | |
233 | return ERR_INVALID_REQUEST; | |
234 | } | |
235 | if (adm_ctx.my_addr && adm_ctx.peer_addr) | |
236 | adm_ctx.tconn = conn_get_by_addrs(nla_data(adm_ctx.my_addr), | |
237 | nla_len(adm_ctx.my_addr), | |
238 | nla_data(adm_ctx.peer_addr), | |
239 | nla_len(adm_ctx.peer_addr)); | |
240 | if (!adm_ctx.tconn) { | |
241 | drbd_msg_put_info("unknown connection"); | |
242 | return ERR_INVALID_REQUEST; | |
243 | } | |
244 | } | |
245 | ||
3b98c0c2 | 246 | /* some more paranoia, if the request was over-determined */ |
527f4b24 LE |
247 | if (adm_ctx.mdev && adm_ctx.tconn && |
248 | adm_ctx.mdev->tconn != adm_ctx.tconn) { | |
44e52cfa | 249 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to connection %s\n", |
7c3063cc AG |
250 | adm_ctx.minor, adm_ctx.resource_name, |
251 | adm_ctx.mdev->tconn->name); | |
44e52cfa | 252 | drbd_msg_put_info("minor exists in different resource"); |
527f4b24 LE |
253 | return ERR_INVALID_REQUEST; |
254 | } | |
3b98c0c2 LE |
255 | if (adm_ctx.mdev && |
256 | adm_ctx.volume != VOLUME_UNSPECIFIED && | |
257 | adm_ctx.volume != adm_ctx.mdev->vnr) { | |
258 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", | |
259 | adm_ctx.minor, adm_ctx.volume, | |
260 | adm_ctx.mdev->vnr, adm_ctx.mdev->tconn->name); | |
527f4b24 | 261 | drbd_msg_put_info("minor exists as different volume"); |
3b98c0c2 LE |
262 | return ERR_INVALID_REQUEST; |
263 | } | |
0ace9dfa | 264 | |
3b98c0c2 LE |
265 | return NO_ERROR; |
266 | ||
267 | fail: | |
268 | nlmsg_free(adm_ctx.reply_skb); | |
269 | adm_ctx.reply_skb = NULL; | |
1e2a2551 | 270 | return err; |
3b98c0c2 LE |
271 | } |
272 | ||
273 | static int drbd_adm_finish(struct genl_info *info, int retcode) | |
274 | { | |
0ace9dfa PR |
275 | if (adm_ctx.tconn) { |
276 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); | |
277 | adm_ctx.tconn = NULL; | |
278 | } | |
279 | ||
3b98c0c2 LE |
280 | if (!adm_ctx.reply_skb) |
281 | return -ENOMEM; | |
282 | ||
283 | adm_ctx.reply_dh->ret_code = retcode; | |
3b98c0c2 LE |
284 | drbd_adm_send_reply(adm_ctx.reply_skb, info); |
285 | return 0; | |
286 | } | |
b411b363 | 287 | |
6b75dced | 288 | static void setup_khelper_env(struct drbd_tconn *tconn, char **envp) |
b411b363 | 289 | { |
6b75dced | 290 | char *afs; |
b411b363 | 291 | |
089c075d AG |
292 | /* FIXME: A future version will not allow this case. */ |
293 | if (tconn->my_addr_len == 0 || tconn->peer_addr_len == 0) | |
294 | return; | |
295 | ||
296 | switch (((struct sockaddr *)&tconn->peer_addr)->sa_family) { | |
297 | case AF_INET6: | |
298 | afs = "ipv6"; | |
299 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
300 | &((struct sockaddr_in6 *)&tconn->peer_addr)->sin6_addr); | |
b411b363 | 301 | break; |
089c075d AG |
302 | case AF_INET: |
303 | afs = "ipv4"; | |
304 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
305 | &((struct sockaddr_in *)&tconn->peer_addr)->sin_addr); | |
b411b363 | 306 | break; |
089c075d AG |
307 | default: |
308 | afs = "ssocks"; | |
309 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
310 | &((struct sockaddr_in *)&tconn->peer_addr)->sin_addr); | |
b411b363 | 311 | } |
089c075d | 312 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 313 | } |
b411b363 PR |
314 | |
315 | int drbd_khelper(struct drbd_conf *mdev, char *cmd) | |
316 | { | |
317 | char *envp[] = { "HOME=/", | |
318 | "TERM=linux", | |
319 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
320 | (char[20]) { }, /* address family */ |
321 | (char[60]) { }, /* address */ | |
b411b363 | 322 | NULL }; |
6b75dced | 323 | char mb[12]; |
b411b363 | 324 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
6f3465ed | 325 | struct drbd_tconn *tconn = mdev->tconn; |
6b75dced | 326 | struct sib_info sib; |
b411b363 PR |
327 | int ret; |
328 | ||
6f3465ed LE |
329 | if (current == tconn->worker.task) |
330 | set_bit(CALLBACK_PENDING, &tconn->flags); | |
c2ba686f | 331 | |
b411b363 | 332 | snprintf(mb, 12, "minor-%d", mdev_to_minor(mdev)); |
6f3465ed | 333 | setup_khelper_env(tconn, envp); |
b411b363 | 334 | |
1090c056 LE |
335 | /* The helper may take some time. |
336 | * write out any unsynced meta data changes now */ | |
337 | drbd_md_sync(mdev); | |
338 | ||
b411b363 | 339 | dev_info(DEV, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
340 | sib.sib_reason = SIB_HELPER_PRE; |
341 | sib.helper_name = cmd; | |
342 | drbd_bcast_event(mdev, &sib); | |
70834d30 | 343 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 PR |
344 | if (ret) |
345 | dev_warn(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
346 | usermode_helper, cmd, mb, | |
347 | (ret >> 8) & 0xff, ret); | |
348 | else | |
349 | dev_info(DEV, "helper command: %s %s %s exit code %u (0x%x)\n", | |
350 | usermode_helper, cmd, mb, | |
351 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
352 | sib.sib_reason = SIB_HELPER_POST; |
353 | sib.helper_exit_code = ret; | |
354 | drbd_bcast_event(mdev, &sib); | |
b411b363 | 355 | |
6f3465ed LE |
356 | if (current == tconn->worker.task) |
357 | clear_bit(CALLBACK_PENDING, &tconn->flags); | |
b411b363 PR |
358 | |
359 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
360 | ret = 0; | |
361 | ||
362 | return ret; | |
363 | } | |
364 | ||
6b75dced PR |
365 | int conn_khelper(struct drbd_tconn *tconn, char *cmd) |
366 | { | |
367 | char *envp[] = { "HOME=/", | |
368 | "TERM=linux", | |
369 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
370 | (char[20]) { }, /* address family */ | |
371 | (char[60]) { }, /* address */ | |
372 | NULL }; | |
373 | char *argv[] = {usermode_helper, cmd, tconn->name, NULL }; | |
374 | int ret; | |
375 | ||
376 | setup_khelper_env(tconn, envp); | |
377 | conn_md_sync(tconn); | |
378 | ||
379 | conn_info(tconn, "helper command: %s %s %s\n", usermode_helper, cmd, tconn->name); | |
380 | /* TODO: conn_bcast_event() ?? */ | |
381 | ||
98683650 | 382 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced PR |
383 | if (ret) |
384 | conn_warn(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
385 | usermode_helper, cmd, tconn->name, | |
386 | (ret >> 8) & 0xff, ret); | |
387 | else | |
388 | conn_info(tconn, "helper command: %s %s %s exit code %u (0x%x)\n", | |
389 | usermode_helper, cmd, tconn->name, | |
390 | (ret >> 8) & 0xff, ret); | |
391 | /* TODO: conn_bcast_event() ?? */ | |
c2ba686f | 392 | |
b411b363 PR |
393 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
394 | ret = 0; | |
395 | ||
396 | return ret; | |
397 | } | |
398 | ||
cb703454 | 399 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_tconn *tconn) |
b411b363 | 400 | { |
cb703454 PR |
401 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
402 | struct drbd_conf *mdev; | |
403 | int vnr; | |
404 | ||
695d08fa | 405 | rcu_read_lock(); |
cb703454 PR |
406 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
407 | if (get_ldev_if_state(mdev, D_CONSISTENT)) { | |
daeda1cc PR |
408 | fp = max_t(enum drbd_fencing_p, fp, |
409 | rcu_dereference(mdev->ldev->disk_conf)->fencing); | |
cb703454 PR |
410 | put_ldev(mdev); |
411 | } | |
412 | } | |
695d08fa | 413 | rcu_read_unlock(); |
cb703454 PR |
414 | |
415 | return fp; | |
416 | } | |
417 | ||
418 | bool conn_try_outdate_peer(struct drbd_tconn *tconn) | |
b411b363 | 419 | { |
cb703454 PR |
420 | union drbd_state mask = { }; |
421 | union drbd_state val = { }; | |
422 | enum drbd_fencing_p fp; | |
b411b363 PR |
423 | char *ex_to_string; |
424 | int r; | |
b411b363 | 425 | |
cb703454 PR |
426 | if (tconn->cstate >= C_WF_REPORT_PARAMS) { |
427 | conn_err(tconn, "Expected cstate < C_WF_REPORT_PARAMS\n"); | |
428 | return false; | |
429 | } | |
b411b363 | 430 | |
cb703454 PR |
431 | fp = highest_fencing_policy(tconn); |
432 | switch (fp) { | |
433 | case FP_NOT_AVAIL: | |
434 | conn_warn(tconn, "Not fencing peer, I'm not even Consistent myself.\n"); | |
fb22c402 | 435 | goto out; |
cb703454 PR |
436 | case FP_DONT_CARE: |
437 | return true; | |
438 | default: ; | |
b411b363 PR |
439 | } |
440 | ||
cb703454 | 441 | r = conn_khelper(tconn, "fence-peer"); |
b411b363 PR |
442 | |
443 | switch ((r>>8) & 0xff) { | |
444 | case 3: /* peer is inconsistent */ | |
445 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
446 | mask.pdsk = D_MASK; |
447 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
448 | break; |
449 | case 4: /* peer got outdated, or was already outdated */ | |
450 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
451 | mask.pdsk = D_MASK; |
452 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
453 | break; |
454 | case 5: /* peer was down */ | |
cb703454 | 455 | if (conn_highest_disk(tconn) == D_UP_TO_DATE) { |
b411b363 PR |
456 | /* we will(have) create(d) a new UUID anyways... */ |
457 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
458 | mask.pdsk = D_MASK; |
459 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
460 | } else { |
461 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
462 | } |
463 | break; | |
464 | case 6: /* Peer is primary, voluntarily outdate myself. | |
465 | * This is useful when an unconnected R_SECONDARY is asked to | |
466 | * become R_PRIMARY, but finds the other peer being active. */ | |
467 | ex_to_string = "peer is active"; | |
cb703454 PR |
468 | conn_warn(tconn, "Peer is primary, outdating myself.\n"); |
469 | mask.disk = D_MASK; | |
470 | val.disk = D_OUTDATED; | |
b411b363 PR |
471 | break; |
472 | case 7: | |
473 | if (fp != FP_STONITH) | |
cb703454 | 474 | conn_err(tconn, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 475 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
476 | mask.pdsk = D_MASK; |
477 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
478 | break; |
479 | default: | |
480 | /* The script is broken ... */ | |
cb703454 PR |
481 | conn_err(tconn, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
482 | return false; /* Eventually leave IO frozen */ | |
b411b363 PR |
483 | } |
484 | ||
cb703454 PR |
485 | conn_info(tconn, "fence-peer helper returned %d (%s)\n", |
486 | (r>>8) & 0xff, ex_to_string); | |
fb22c402 | 487 | |
cb703454 | 488 | out: |
fb22c402 | 489 | |
cb703454 PR |
490 | /* Not using |
491 | conn_request_state(tconn, mask, val, CS_VERBOSE); | |
492 | here, because we might were able to re-establish the connection in the | |
493 | meantime. */ | |
494 | spin_lock_irq(&tconn->req_lock); | |
a1096a6e | 495 | if (tconn->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &tconn->flags)) |
cb703454 PR |
496 | _conn_request_state(tconn, mask, val, CS_VERBOSE); |
497 | spin_unlock_irq(&tconn->req_lock); | |
498 | ||
499 | return conn_highest_pdsk(tconn) <= D_OUTDATED; | |
b411b363 PR |
500 | } |
501 | ||
87f7be4c PR |
502 | static int _try_outdate_peer_async(void *data) |
503 | { | |
cb703454 | 504 | struct drbd_tconn *tconn = (struct drbd_tconn *)data; |
87f7be4c | 505 | |
cb703454 | 506 | conn_try_outdate_peer(tconn); |
87f7be4c | 507 | |
9dc9fbb3 | 508 | kref_put(&tconn->kref, &conn_destroy); |
87f7be4c PR |
509 | return 0; |
510 | } | |
511 | ||
cb703454 | 512 | void conn_try_outdate_peer_async(struct drbd_tconn *tconn) |
87f7be4c PR |
513 | { |
514 | struct task_struct *opa; | |
515 | ||
9dc9fbb3 | 516 | kref_get(&tconn->kref); |
cb703454 | 517 | opa = kthread_run(_try_outdate_peer_async, tconn, "drbd_async_h"); |
9dc9fbb3 | 518 | if (IS_ERR(opa)) { |
cb703454 | 519 | conn_err(tconn, "out of mem, failed to invoke fence-peer helper\n"); |
9dc9fbb3 PR |
520 | kref_put(&tconn->kref, &conn_destroy); |
521 | } | |
87f7be4c | 522 | } |
b411b363 | 523 | |
bf885f8a AG |
524 | enum drbd_state_rv |
525 | drbd_set_role(struct drbd_conf *mdev, enum drbd_role new_role, int force) | |
b411b363 PR |
526 | { |
527 | const int max_tries = 4; | |
bf885f8a | 528 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 529 | struct net_conf *nc; |
b411b363 PR |
530 | int try = 0; |
531 | int forced = 0; | |
532 | union drbd_state mask, val; | |
b411b363 PR |
533 | |
534 | if (new_role == R_PRIMARY) | |
0625ac19 | 535 | request_ping(mdev->tconn); /* Detect a dead peer ASAP */ |
b411b363 | 536 | |
8410da8f | 537 | mutex_lock(mdev->state_mutex); |
b411b363 PR |
538 | |
539 | mask.i = 0; mask.role = R_MASK; | |
540 | val.i = 0; val.role = new_role; | |
541 | ||
542 | while (try++ < max_tries) { | |
bf885f8a | 543 | rv = _drbd_request_state(mdev, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
544 | |
545 | /* in case we first succeeded to outdate, | |
546 | * but now suddenly could establish a connection */ | |
bf885f8a | 547 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
548 | val.pdsk = 0; |
549 | mask.pdsk = 0; | |
550 | continue; | |
551 | } | |
552 | ||
bf885f8a | 553 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
d10a33c6 PR |
554 | (mdev->state.disk < D_UP_TO_DATE && |
555 | mdev->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
556 | mask.disk = D_MASK; |
557 | val.disk = D_UP_TO_DATE; | |
558 | forced = 1; | |
559 | continue; | |
560 | } | |
561 | ||
bf885f8a | 562 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b411b363 PR |
563 | mdev->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
564 | D_ASSERT(mdev->state.pdsk == D_UNKNOWN); | |
b411b363 | 565 | |
cb703454 | 566 | if (conn_try_outdate_peer(mdev->tconn)) { |
b411b363 PR |
567 | val.disk = D_UP_TO_DATE; |
568 | mask.disk = D_MASK; | |
569 | } | |
b411b363 PR |
570 | continue; |
571 | } | |
572 | ||
bf885f8a | 573 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 574 | goto out; |
bf885f8a | 575 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
cb703454 | 576 | if (!conn_try_outdate_peer(mdev->tconn) && force) { |
b411b363 | 577 | dev_warn(DEV, "Forced into split brain situation!\n"); |
cb703454 PR |
578 | mask.pdsk = D_MASK; |
579 | val.pdsk = D_OUTDATED; | |
b411b363 | 580 | |
cb703454 | 581 | } |
b411b363 PR |
582 | continue; |
583 | } | |
bf885f8a | 584 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
585 | /* Maybe the peer is detected as dead very soon... |
586 | retry at most once more in this case. */ | |
44ed167d PR |
587 | int timeo; |
588 | rcu_read_lock(); | |
589 | nc = rcu_dereference(mdev->tconn->net_conf); | |
590 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; | |
591 | rcu_read_unlock(); | |
592 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
593 | if (try < max_tries) |
594 | try = max_tries - 1; | |
595 | continue; | |
596 | } | |
bf885f8a AG |
597 | if (rv < SS_SUCCESS) { |
598 | rv = _drbd_request_state(mdev, mask, val, | |
b411b363 | 599 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 600 | if (rv < SS_SUCCESS) |
3b98c0c2 | 601 | goto out; |
b411b363 PR |
602 | } |
603 | break; | |
604 | } | |
605 | ||
bf885f8a | 606 | if (rv < SS_SUCCESS) |
3b98c0c2 | 607 | goto out; |
b411b363 PR |
608 | |
609 | if (forced) | |
610 | dev_warn(DEV, "Forced to consider local data as UpToDate!\n"); | |
611 | ||
612 | /* Wait until nothing is on the fly :) */ | |
613 | wait_event(mdev->misc_wait, atomic_read(&mdev->ap_pending_cnt) == 0); | |
614 | ||
b6dd1a89 LE |
615 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
616 | ||
b411b363 | 617 | if (new_role == R_SECONDARY) { |
81e84650 | 618 | set_disk_ro(mdev->vdisk, true); |
b411b363 PR |
619 | if (get_ldev(mdev)) { |
620 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
621 | put_ldev(mdev); | |
622 | } | |
623 | } else { | |
a0095508 | 624 | mutex_lock(&mdev->tconn->conf_update); |
91fd4dad | 625 | nc = mdev->tconn->net_conf; |
44ed167d | 626 | if (nc) |
6139f60d | 627 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
a0095508 | 628 | mutex_unlock(&mdev->tconn->conf_update); |
91fd4dad | 629 | |
81e84650 | 630 | set_disk_ro(mdev->vdisk, false); |
b411b363 PR |
631 | if (get_ldev(mdev)) { |
632 | if (((mdev->state.conn < C_CONNECTED || | |
633 | mdev->state.pdsk <= D_FAILED) | |
634 | && mdev->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
635 | drbd_uuid_new_current(mdev); | |
636 | ||
637 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
638 | put_ldev(mdev); | |
639 | } | |
640 | } | |
641 | ||
19f843aa LE |
642 | /* writeout of activity log covered areas of the bitmap |
643 | * to stable storage done in after state change already */ | |
b411b363 PR |
644 | |
645 | if (mdev->state.conn >= C_WF_REPORT_PARAMS) { | |
646 | /* if this was forced, we should consider sync */ | |
647 | if (forced) | |
648 | drbd_send_uuids(mdev); | |
f479ea06 | 649 | drbd_send_current_state(mdev); |
b411b363 PR |
650 | } |
651 | ||
652 | drbd_md_sync(mdev); | |
653 | ||
654 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
3b98c0c2 | 655 | out: |
8410da8f | 656 | mutex_unlock(mdev->state_mutex); |
bf885f8a | 657 | return rv; |
b411b363 PR |
658 | } |
659 | ||
3b98c0c2 | 660 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 661 | { |
3b98c0c2 LE |
662 | return err == -ENOMSG ? "required attribute missing" : |
663 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 664 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 665 | "invalid attribute value"; |
ef50a3e3 | 666 | } |
b411b363 | 667 | |
3b98c0c2 | 668 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 669 | { |
3b98c0c2 LE |
670 | struct set_role_parms parms; |
671 | int err; | |
672 | enum drbd_ret_code retcode; | |
b411b363 | 673 | |
3b98c0c2 LE |
674 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
675 | if (!adm_ctx.reply_skb) | |
676 | return retcode; | |
677 | if (retcode != NO_ERROR) | |
678 | goto out; | |
b411b363 | 679 | |
3b98c0c2 LE |
680 | memset(&parms, 0, sizeof(parms)); |
681 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 682 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
683 | if (err) { |
684 | retcode = ERR_MANDATORY_TAG; | |
685 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
686 | goto out; | |
687 | } | |
688 | } | |
b411b363 | 689 | |
3b98c0c2 LE |
690 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
691 | retcode = drbd_set_role(adm_ctx.mdev, R_PRIMARY, parms.assume_uptodate); | |
692 | else | |
693 | retcode = drbd_set_role(adm_ctx.mdev, R_SECONDARY, 0); | |
694 | out: | |
695 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
696 | return 0; |
697 | } | |
698 | ||
ae8bf312 LE |
699 | /* Initializes the md.*_offset members, so we are able to find |
700 | * the on disk meta data. | |
701 | * | |
702 | * We currently have two possible layouts: | |
703 | * external: | |
704 | * |----------- md_size_sect ------------------| | |
705 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
706 | * | al_offset == 8 | | |
707 | * | bm_offset = al_offset + X | | |
708 | * ==> bitmap sectors = md_size_sect - bm_offset | |
709 | * | |
710 | * internal: | |
711 | * |----------- md_size_sect ------------------| | |
712 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
713 | * | al_offset < 0 | | |
714 | * | bm_offset = al_offset - Y | | |
715 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
716 | * | |
717 | * Activity log size used to be fixed 32kB, | |
718 | * but is about to become configurable. | |
719 | */ | |
b411b363 PR |
720 | static void drbd_md_set_sector_offsets(struct drbd_conf *mdev, |
721 | struct drbd_backing_dev *bdev) | |
722 | { | |
723 | sector_t md_size_sect = 0; | |
c04ccaa6 | 724 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 725 | |
3a4d4eb3 LE |
726 | bdev->md.md_offset = drbd_md_ss(bdev); |
727 | ||
68e41a43 | 728 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
729 | default: |
730 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 731 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
732 | bdev->md.al_offset = MD_4kB_SECT; |
733 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
734 | break; |
735 | case DRBD_MD_INDEX_FLEX_EXT: | |
736 | /* just occupy the full device; unit: sectors */ | |
737 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
738 | bdev->md.al_offset = MD_4kB_SECT; |
739 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
740 | break; |
741 | case DRBD_MD_INDEX_INTERNAL: | |
742 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 743 | /* al size is still fixed */ |
ae8bf312 | 744 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
745 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
746 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
747 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
748 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
749 | md_size_sect = ALIGN(md_size_sect, 8); | |
750 | ||
751 | /* plus the "drbd meta data super block", | |
752 | * and the activity log; */ | |
ae8bf312 | 753 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
754 | |
755 | bdev->md.md_size_sect = md_size_sect; | |
756 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 757 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
758 | break; |
759 | } | |
760 | } | |
761 | ||
4b0715f0 | 762 | /* input size is expected to be in KB */ |
b411b363 PR |
763 | char *ppsize(char *buf, unsigned long long size) |
764 | { | |
4b0715f0 LE |
765 | /* Needs 9 bytes at max including trailing NUL: |
766 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
767 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
768 | int base = 0; | |
4b0715f0 | 769 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
770 | /* shift + round */ |
771 | size = (size >> 10) + !!(size & (1<<9)); | |
772 | base++; | |
773 | } | |
4b0715f0 | 774 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
775 | |
776 | return buf; | |
777 | } | |
778 | ||
779 | /* there is still a theoretical deadlock when called from receiver | |
780 | * on an D_INCONSISTENT R_PRIMARY: | |
781 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
782 | * packet from remote to dec_ap_bio again. | |
783 | * receiver receive_sizes(), comes here, | |
784 | * waits for ap_bio_cnt == 0. -> deadlock. | |
785 | * but this cannot happen, actually, because: | |
786 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
787 | * (not connected, or bad/no disk on peer): | |
788 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
789 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
790 | * peer may not initiate a resize. | |
791 | */ | |
3b98c0c2 LE |
792 | /* Note these are not to be confused with |
793 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
794 | * which are (sub) state changes triggered by admin (drbdsetup), | |
795 | * and can be long lived. | |
796 | * This changes an mdev->flag, is triggered by drbd internals, | |
797 | * and should be short-lived. */ | |
b411b363 PR |
798 | void drbd_suspend_io(struct drbd_conf *mdev) |
799 | { | |
800 | set_bit(SUSPEND_IO, &mdev->flags); | |
2aebfabb | 801 | if (drbd_suspended(mdev)) |
265be2d0 | 802 | return; |
b411b363 PR |
803 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_bio_cnt)); |
804 | } | |
805 | ||
806 | void drbd_resume_io(struct drbd_conf *mdev) | |
807 | { | |
808 | clear_bit(SUSPEND_IO, &mdev->flags); | |
809 | wake_up(&mdev->misc_wait); | |
810 | } | |
811 | ||
812 | /** | |
813 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
814 | * @mdev: DRBD device. | |
815 | * | |
816 | * Returns 0 on success, negative return values indicate errors. | |
817 | * You should call drbd_md_sync() after calling this function. | |
818 | */ | |
24c4830c | 819 | enum determine_dev_size drbd_determine_dev_size(struct drbd_conf *mdev, enum dds_flags flags) __must_hold(local) |
b411b363 PR |
820 | { |
821 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 822 | sector_t la_size_sect, u_size; |
b411b363 PR |
823 | sector_t size; |
824 | char ppb[10]; | |
825 | ||
826 | int md_moved, la_size_changed; | |
827 | enum determine_dev_size rv = unchanged; | |
828 | ||
829 | /* race: | |
830 | * application request passes inc_ap_bio, | |
831 | * but then cannot get an AL-reference. | |
832 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
833 | * | |
834 | * to avoid that: | |
835 | * Suspend IO right here. | |
836 | * still lock the act_log to not trigger ASSERTs there. | |
837 | */ | |
838 | drbd_suspend_io(mdev); | |
839 | ||
840 | /* no wait necessary anymore, actually we could assert that */ | |
841 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); | |
842 | ||
843 | prev_first_sect = drbd_md_first_sector(mdev->ldev); | |
844 | prev_size = mdev->ldev->md.md_size_sect; | |
cccac985 | 845 | la_size_sect = mdev->ldev->md.la_size_sect; |
b411b363 PR |
846 | |
847 | /* TODO: should only be some assert here, not (re)init... */ | |
848 | drbd_md_set_sector_offsets(mdev, mdev->ldev); | |
849 | ||
daeda1cc PR |
850 | rcu_read_lock(); |
851 | u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size; | |
852 | rcu_read_unlock(); | |
ef5e44a6 | 853 | size = drbd_new_dev_size(mdev, mdev->ldev, u_size, flags & DDSF_FORCED); |
b411b363 PR |
854 | |
855 | if (drbd_get_capacity(mdev->this_bdev) != size || | |
856 | drbd_bm_capacity(mdev) != size) { | |
857 | int err; | |
02d9a94b | 858 | err = drbd_bm_resize(mdev, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
859 | if (unlikely(err)) { |
860 | /* currently there is only one error: ENOMEM! */ | |
861 | size = drbd_bm_capacity(mdev)>>1; | |
862 | if (size == 0) { | |
863 | dev_err(DEV, "OUT OF MEMORY! " | |
864 | "Could not allocate bitmap!\n"); | |
865 | } else { | |
866 | dev_err(DEV, "BM resizing failed. " | |
867 | "Leaving size unchanged at size = %lu KB\n", | |
868 | (unsigned long)size); | |
869 | } | |
870 | rv = dev_size_error; | |
871 | } | |
872 | /* racy, see comments above. */ | |
873 | drbd_set_my_capacity(mdev, size); | |
874 | mdev->ldev->md.la_size_sect = size; | |
875 | dev_info(DEV, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), | |
876 | (unsigned long long)size>>1); | |
877 | } | |
878 | if (rv == dev_size_error) | |
879 | goto out; | |
880 | ||
cccac985 | 881 | la_size_changed = (la_size_sect != mdev->ldev->md.la_size_sect); |
b411b363 PR |
882 | |
883 | md_moved = prev_first_sect != drbd_md_first_sector(mdev->ldev) | |
884 | || prev_size != mdev->ldev->md.md_size_sect; | |
885 | ||
886 | if (la_size_changed || md_moved) { | |
24dccabb AG |
887 | int err; |
888 | ||
b411b363 PR |
889 | drbd_al_shrink(mdev); /* All extents inactive. */ |
890 | dev_info(DEV, "Writing the whole bitmap, %s\n", | |
891 | la_size_changed && md_moved ? "size changed and md moved" : | |
892 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 893 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
d1aa4d04 PR |
894 | err = drbd_bitmap_io(mdev, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
895 | "size changed", BM_LOCKED_MASK); | |
24dccabb AG |
896 | if (err) { |
897 | rv = dev_size_error; | |
898 | goto out; | |
899 | } | |
b411b363 PR |
900 | drbd_md_mark_dirty(mdev); |
901 | } | |
902 | ||
cccac985 | 903 | if (size > la_size_sect) |
b411b363 | 904 | rv = grew; |
cccac985 | 905 | if (size < la_size_sect) |
b411b363 PR |
906 | rv = shrunk; |
907 | out: | |
908 | lc_unlock(mdev->act_log); | |
909 | wake_up(&mdev->al_wait); | |
910 | drbd_resume_io(mdev); | |
911 | ||
912 | return rv; | |
913 | } | |
914 | ||
915 | sector_t | |
ef5e44a6 PR |
916 | drbd_new_dev_size(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, |
917 | sector_t u_size, int assume_peer_has_space) | |
b411b363 PR |
918 | { |
919 | sector_t p_size = mdev->p_size; /* partner's disk size. */ | |
cccac985 | 920 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 921 | sector_t m_size; /* my size */ |
b411b363 PR |
922 | sector_t size = 0; |
923 | ||
924 | m_size = drbd_get_max_capacity(bdev); | |
925 | ||
a393db6f PR |
926 | if (mdev->state.conn < C_CONNECTED && assume_peer_has_space) { |
927 | dev_warn(DEV, "Resize while not connected was forced by the user!\n"); | |
928 | p_size = m_size; | |
929 | } | |
930 | ||
b411b363 PR |
931 | if (p_size && m_size) { |
932 | size = min_t(sector_t, p_size, m_size); | |
933 | } else { | |
cccac985 LE |
934 | if (la_size_sect) { |
935 | size = la_size_sect; | |
b411b363 PR |
936 | if (m_size && m_size < size) |
937 | size = m_size; | |
938 | if (p_size && p_size < size) | |
939 | size = p_size; | |
940 | } else { | |
941 | if (m_size) | |
942 | size = m_size; | |
943 | if (p_size) | |
944 | size = p_size; | |
945 | } | |
946 | } | |
947 | ||
948 | if (size == 0) | |
949 | dev_err(DEV, "Both nodes diskless!\n"); | |
950 | ||
951 | if (u_size) { | |
952 | if (u_size > size) | |
953 | dev_err(DEV, "Requested disk size is too big (%lu > %lu)\n", | |
954 | (unsigned long)u_size>>1, (unsigned long)size>>1); | |
955 | else | |
956 | size = u_size; | |
957 | } | |
958 | ||
959 | return size; | |
960 | } | |
961 | ||
962 | /** | |
963 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
964 | * @mdev: DRBD device. | |
965 | * | |
966 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
967 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
968 | * this function. | |
969 | */ | |
f399002e | 970 | static int drbd_check_al_size(struct drbd_conf *mdev, struct disk_conf *dc) |
b411b363 PR |
971 | { |
972 | struct lru_cache *n, *t; | |
973 | struct lc_element *e; | |
974 | unsigned int in_use; | |
975 | int i; | |
976 | ||
b411b363 | 977 | if (mdev->act_log && |
f399002e | 978 | mdev->act_log->nr_elements == dc->al_extents) |
b411b363 PR |
979 | return 0; |
980 | ||
981 | in_use = 0; | |
982 | t = mdev->act_log; | |
7ad651b5 | 983 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 984 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
985 | |
986 | if (n == NULL) { | |
987 | dev_err(DEV, "Cannot allocate act_log lru!\n"); | |
988 | return -ENOMEM; | |
989 | } | |
990 | spin_lock_irq(&mdev->al_lock); | |
991 | if (t) { | |
992 | for (i = 0; i < t->nr_elements; i++) { | |
993 | e = lc_element_by_index(t, i); | |
994 | if (e->refcnt) | |
995 | dev_err(DEV, "refcnt(%d)==%d\n", | |
996 | e->lc_number, e->refcnt); | |
997 | in_use += e->refcnt; | |
998 | } | |
999 | } | |
1000 | if (!in_use) | |
1001 | mdev->act_log = n; | |
1002 | spin_unlock_irq(&mdev->al_lock); | |
1003 | if (in_use) { | |
1004 | dev_err(DEV, "Activity log still in use!\n"); | |
1005 | lc_destroy(n); | |
1006 | return -EBUSY; | |
1007 | } else { | |
1008 | if (t) | |
1009 | lc_destroy(t); | |
1010 | } | |
1011 | drbd_md_mark_dirty(mdev); /* we changed mdev->act_log->nr_elemens */ | |
1012 | return 0; | |
1013 | } | |
1014 | ||
99432fcc | 1015 | static void drbd_setup_queue_param(struct drbd_conf *mdev, unsigned int max_bio_size) |
b411b363 PR |
1016 | { |
1017 | struct request_queue * const q = mdev->rq_queue; | |
db141b2f LE |
1018 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1019 | unsigned int max_segments = 0; | |
99432fcc PR |
1020 | |
1021 | if (get_ldev_if_state(mdev, D_ATTACHING)) { | |
1022 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
1023 | ||
1024 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc PR |
1025 | rcu_read_lock(); |
1026 | max_segments = rcu_dereference(mdev->ldev->disk_conf)->max_bio_bvecs; | |
1027 | rcu_read_unlock(); | |
99432fcc PR |
1028 | put_ldev(mdev); |
1029 | } | |
b411b363 | 1030 | |
b411b363 | 1031 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1032 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1033 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1034 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1035 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1036 | |
99432fcc PR |
1037 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1038 | struct request_queue * const b = mdev->ldev->backing_bdev->bd_disk->queue; | |
1039 | ||
1040 | blk_queue_stack_limits(q, b); | |
1041 | ||
1042 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
1043 | dev_info(DEV, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", | |
1044 | q->backing_dev_info.ra_pages, | |
1045 | b->backing_dev_info.ra_pages); | |
1046 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1047 | } | |
1048 | put_ldev(mdev); | |
1049 | } | |
1050 | } | |
1051 | ||
1052 | void drbd_reconsider_max_bio_size(struct drbd_conf *mdev) | |
1053 | { | |
db141b2f | 1054 | unsigned int now, new, local, peer; |
99432fcc PR |
1055 | |
1056 | now = queue_max_hw_sectors(mdev->rq_queue) << 9; | |
1057 | local = mdev->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1058 | peer = mdev->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1059 | |
99432fcc PR |
1060 | if (get_ldev_if_state(mdev, D_ATTACHING)) { |
1061 | local = queue_max_hw_sectors(mdev->ldev->backing_bdev->bd_disk->queue) << 9; | |
1062 | mdev->local_max_bio_size = local; | |
1063 | put_ldev(mdev); | |
b411b363 | 1064 | } |
db141b2f | 1065 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1066 | |
1067 | /* We may ignore peer limits if the peer is modern enough. | |
1068 | Because new from 8.3.8 onwards the peer can use multiple | |
1069 | BIOs for a single peer_request */ | |
1070 | if (mdev->state.conn >= C_CONNECTED) { | |
31890f4a | 1071 | if (mdev->tconn->agreed_pro_version < 94) |
98683650 | 1072 | peer = min( mdev->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1073 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
31890f4a | 1074 | else if (mdev->tconn->agreed_pro_version == 94) |
99432fcc | 1075 | peer = DRBD_MAX_SIZE_H80_PACKET; |
2ffca4f3 PR |
1076 | else if (mdev->tconn->agreed_pro_version < 100) |
1077 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ | |
1078 | else | |
99432fcc PR |
1079 | peer = DRBD_MAX_BIO_SIZE; |
1080 | } | |
1081 | ||
db141b2f | 1082 | new = min(local, peer); |
99432fcc PR |
1083 | |
1084 | if (mdev->state.role == R_PRIMARY && new < now) | |
db141b2f | 1085 | dev_err(DEV, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1086 | |
1087 | if (new != now) | |
1088 | dev_info(DEV, "max BIO size = %u\n", new); | |
1089 | ||
1090 | drbd_setup_queue_param(mdev, new); | |
b411b363 PR |
1091 | } |
1092 | ||
a18e9d1e | 1093 | /* Starts the worker thread */ |
0e29d163 | 1094 | static void conn_reconfig_start(struct drbd_tconn *tconn) |
b411b363 | 1095 | { |
0e29d163 PR |
1096 | drbd_thread_start(&tconn->worker); |
1097 | conn_flush_workqueue(tconn); | |
b411b363 PR |
1098 | } |
1099 | ||
a18e9d1e | 1100 | /* if still unconfigured, stops worker again. */ |
0e29d163 | 1101 | static void conn_reconfig_done(struct drbd_tconn *tconn) |
b411b363 | 1102 | { |
992d6e91 | 1103 | bool stop_threads; |
0e29d163 | 1104 | spin_lock_irq(&tconn->req_lock); |
e0e16653 PR |
1105 | stop_threads = conn_all_vols_unconf(tconn) && |
1106 | tconn->cstate == C_STANDALONE; | |
0e29d163 | 1107 | spin_unlock_irq(&tconn->req_lock); |
992d6e91 LE |
1108 | if (stop_threads) { |
1109 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1110 | * in conn_disconnect() */ |
992d6e91 LE |
1111 | drbd_thread_stop(&tconn->receiver); |
1112 | drbd_thread_stop(&tconn->worker); | |
1113 | } | |
b411b363 PR |
1114 | } |
1115 | ||
0778286a PR |
1116 | /* Make sure IO is suspended before calling this function(). */ |
1117 | static void drbd_suspend_al(struct drbd_conf *mdev) | |
1118 | { | |
1119 | int s = 0; | |
1120 | ||
61610420 | 1121 | if (!lc_try_lock(mdev->act_log)) { |
0778286a PR |
1122 | dev_warn(DEV, "Failed to lock al in drbd_suspend_al()\n"); |
1123 | return; | |
1124 | } | |
1125 | ||
61610420 | 1126 | drbd_al_shrink(mdev); |
87eeee41 | 1127 | spin_lock_irq(&mdev->tconn->req_lock); |
0778286a PR |
1128 | if (mdev->state.conn < C_CONNECTED) |
1129 | s = !test_and_set_bit(AL_SUSPENDED, &mdev->flags); | |
87eeee41 | 1130 | spin_unlock_irq(&mdev->tconn->req_lock); |
61610420 | 1131 | lc_unlock(mdev->act_log); |
0778286a PR |
1132 | |
1133 | if (s) | |
1134 | dev_info(DEV, "Suspended AL updates\n"); | |
1135 | } | |
1136 | ||
5979e361 LE |
1137 | |
1138 | static bool should_set_defaults(struct genl_info *info) | |
1139 | { | |
1140 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1141 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1142 | } | |
1143 | ||
5bbcf5e6 | 1144 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1145 | { |
5bbcf5e6 LE |
1146 | /* This is limited by 16 bit "slot" numbers, |
1147 | * and by available on-disk context storage. | |
1148 | * | |
1149 | * Also (u16)~0 is special (denotes a "free" extent). | |
1150 | * | |
1151 | * One transaction occupies one 4kB on-disk block, | |
1152 | * we have n such blocks in the on disk ring buffer, | |
1153 | * the "current" transaction may fail (n-1), | |
1154 | * and there is 919 slot numbers context information per transaction. | |
1155 | * | |
1156 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1157 | * so cap there first. | |
1158 | */ | |
1159 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1160 | const unsigned int sufficient_on_disk = | |
1161 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1162 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1163 | |
5bbcf5e6 LE |
1164 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1165 | ||
1166 | if (al_size_4k > sufficient_on_disk) | |
1167 | return max_al_nr; | |
1168 | ||
1169 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1170 | } |
1171 | ||
f399002e LE |
1172 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1173 | { | |
1174 | enum drbd_ret_code retcode; | |
1175 | struct drbd_conf *mdev; | |
daeda1cc | 1176 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1177 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1178 | int err, fifo_size; |
f399002e LE |
1179 | |
1180 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
1181 | if (!adm_ctx.reply_skb) | |
1182 | return retcode; | |
1183 | if (retcode != NO_ERROR) | |
1184 | goto out; | |
1185 | ||
1186 | mdev = adm_ctx.mdev; | |
1187 | ||
1188 | /* we also need a disk | |
1189 | * to change the options on */ | |
1190 | if (!get_ldev(mdev)) { | |
1191 | retcode = ERR_NO_DISK; | |
1192 | goto out; | |
1193 | } | |
1194 | ||
daeda1cc | 1195 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1196 | if (!new_disk_conf) { |
f399002e LE |
1197 | retcode = ERR_NOMEM; |
1198 | goto fail; | |
1199 | } | |
1200 | ||
daeda1cc PR |
1201 | mutex_lock(&mdev->tconn->conf_update); |
1202 | old_disk_conf = mdev->ldev->disk_conf; | |
1203 | *new_disk_conf = *old_disk_conf; | |
5979e361 | 1204 | if (should_set_defaults(info)) |
b966b5dd | 1205 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1206 | |
5ecc72c3 | 1207 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1208 | if (err && err != -ENOMSG) { |
f399002e LE |
1209 | retcode = ERR_MANDATORY_TAG; |
1210 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1211 | } | |
1212 | ||
5ecc72c3 LE |
1213 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1214 | new_disk_conf->resync_rate = 1; | |
f399002e | 1215 | |
5bbcf5e6 LE |
1216 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1217 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1218 | if (new_disk_conf->al_extents > drbd_al_extents_max(mdev->ldev)) | |
1219 | new_disk_conf->al_extents = drbd_al_extents_max(mdev->ldev); | |
1220 | ||
1221 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1222 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1223 | |
5ecc72c3 | 1224 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
9958c857 | 1225 | if (fifo_size != mdev->rs_plan_s->size) { |
813472ce PR |
1226 | new_plan = fifo_alloc(fifo_size); |
1227 | if (!new_plan) { | |
f399002e LE |
1228 | dev_err(DEV, "kmalloc of fifo_buffer failed"); |
1229 | retcode = ERR_NOMEM; | |
daeda1cc | 1230 | goto fail_unlock; |
f399002e LE |
1231 | } |
1232 | } | |
1233 | ||
0ee98e2e | 1234 | drbd_suspend_io(mdev); |
f399002e LE |
1235 | wait_event(mdev->al_wait, lc_try_lock(mdev->act_log)); |
1236 | drbd_al_shrink(mdev); | |
5ecc72c3 | 1237 | err = drbd_check_al_size(mdev, new_disk_conf); |
f399002e LE |
1238 | lc_unlock(mdev->act_log); |
1239 | wake_up(&mdev->al_wait); | |
0ee98e2e | 1240 | drbd_resume_io(mdev); |
f399002e LE |
1241 | |
1242 | if (err) { | |
1243 | retcode = ERR_NOMEM; | |
daeda1cc | 1244 | goto fail_unlock; |
f399002e LE |
1245 | } |
1246 | ||
dc97b708 | 1247 | write_lock_irq(&global_state_lock); |
95f8efd0 | 1248 | retcode = drbd_resync_after_valid(mdev, new_disk_conf->resync_after); |
dc97b708 | 1249 | if (retcode == NO_ERROR) { |
daeda1cc | 1250 | rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf); |
95f8efd0 | 1251 | drbd_resync_after_changed(mdev); |
dc97b708 PR |
1252 | } |
1253 | write_unlock_irq(&global_state_lock); | |
f399002e | 1254 | |
daeda1cc PR |
1255 | if (retcode != NO_ERROR) |
1256 | goto fail_unlock; | |
f399002e | 1257 | |
813472ce PR |
1258 | if (new_plan) { |
1259 | old_plan = mdev->rs_plan_s; | |
1260 | rcu_assign_pointer(mdev->rs_plan_s, new_plan); | |
9958c857 | 1261 | } |
9958c857 | 1262 | |
c141ebda | 1263 | mutex_unlock(&mdev->tconn->conf_update); |
27eb13e9 | 1264 | |
9a51ab1c | 1265 | if (new_disk_conf->al_updates) |
4035e4c2 | 1266 | mdev->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c PR |
1267 | else |
1268 | mdev->ldev->md.flags |= MDF_AL_DISABLED; | |
1269 | ||
691631c0 LE |
1270 | if (new_disk_conf->md_flushes) |
1271 | clear_bit(MD_NO_FUA, &mdev->flags); | |
1272 | else | |
1273 | set_bit(MD_NO_FUA, &mdev->flags); | |
1274 | ||
27eb13e9 PR |
1275 | drbd_bump_write_ordering(mdev->tconn, WO_bdev_flush); |
1276 | ||
daeda1cc | 1277 | drbd_md_sync(mdev); |
f399002e LE |
1278 | |
1279 | if (mdev->state.conn >= C_CONNECTED) | |
1280 | drbd_send_sync_param(mdev); | |
1281 | ||
daeda1cc PR |
1282 | synchronize_rcu(); |
1283 | kfree(old_disk_conf); | |
813472ce | 1284 | kfree(old_plan); |
cdfda633 | 1285 | mod_timer(&mdev->request_timer, jiffies + HZ); |
daeda1cc PR |
1286 | goto success; |
1287 | ||
1288 | fail_unlock: | |
1289 | mutex_unlock(&mdev->tconn->conf_update); | |
f399002e | 1290 | fail: |
5ecc72c3 | 1291 | kfree(new_disk_conf); |
813472ce | 1292 | kfree(new_plan); |
daeda1cc PR |
1293 | success: |
1294 | put_ldev(mdev); | |
f399002e LE |
1295 | out: |
1296 | drbd_adm_finish(info, retcode); | |
1297 | return 0; | |
1298 | } | |
1299 | ||
3b98c0c2 | 1300 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1301 | { |
3b98c0c2 LE |
1302 | struct drbd_conf *mdev; |
1303 | int err; | |
116676ca | 1304 | enum drbd_ret_code retcode; |
b411b363 PR |
1305 | enum determine_dev_size dd; |
1306 | sector_t max_possible_sectors; | |
1307 | sector_t min_md_device_sectors; | |
1308 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1309 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1310 | struct block_device *bdev; |
b411b363 | 1311 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1312 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1313 | union drbd_state ns, os; |
f2024e7c | 1314 | enum drbd_state_rv rv; |
44ed167d | 1315 | struct net_conf *nc; |
b411b363 | 1316 | |
3b98c0c2 LE |
1317 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1318 | if (!adm_ctx.reply_skb) | |
1319 | return retcode; | |
1320 | if (retcode != NO_ERROR) | |
40cbf085 | 1321 | goto finish; |
b411b363 | 1322 | |
3b98c0c2 | 1323 | mdev = adm_ctx.mdev; |
0e29d163 | 1324 | conn_reconfig_start(mdev->tconn); |
b411b363 PR |
1325 | |
1326 | /* if you want to reconfigure, please tear down first */ | |
1327 | if (mdev->state.disk > D_DISKLESS) { | |
1328 | retcode = ERR_DISK_CONFIGURED; | |
1329 | goto fail; | |
1330 | } | |
82f59cc6 LE |
1331 | /* It may just now have detached because of IO error. Make sure |
1332 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1333 | * e.g. if someone calls attach from the on-io-error handler, | |
1334 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
1335 | wait_event(mdev->misc_wait, !atomic_read(&mdev->local_cnt)); | |
b411b363 | 1336 | |
383606e0 | 1337 | /* make sure there is no leftover from previous force-detach attempts */ |
0c849666 | 1338 | clear_bit(FORCE_DETACH, &mdev->flags); |
edc9f5eb LE |
1339 | clear_bit(WAS_IO_ERROR, &mdev->flags); |
1340 | clear_bit(WAS_READ_ERROR, &mdev->flags); | |
383606e0 | 1341 | |
0029d624 LE |
1342 | /* and no leftover from previously aborted resync or verify, either */ |
1343 | mdev->rs_total = 0; | |
1344 | mdev->rs_failed = 0; | |
1345 | atomic_set(&mdev->rs_pending_cnt, 0); | |
1346 | ||
3b98c0c2 | 1347 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1348 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1349 | if (!nbc) { | |
1350 | retcode = ERR_NOMEM; | |
1351 | goto fail; | |
1352 | } | |
9f2247bb PR |
1353 | spin_lock_init(&nbc->md.uuid_lock); |
1354 | ||
daeda1cc PR |
1355 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1356 | if (!new_disk_conf) { | |
1357 | retcode = ERR_NOMEM; | |
b411b363 PR |
1358 | goto fail; |
1359 | } | |
daeda1cc | 1360 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1361 | |
daeda1cc PR |
1362 | set_disk_conf_defaults(new_disk_conf); |
1363 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1364 | if (err) { |
b411b363 | 1365 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 1366 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
1367 | goto fail; |
1368 | } | |
1369 | ||
5bbcf5e6 LE |
1370 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1371 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1372 | |
9958c857 PR |
1373 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1374 | if (!new_plan) { | |
1375 | retcode = ERR_NOMEM; | |
1376 | goto fail; | |
1377 | } | |
1378 | ||
daeda1cc | 1379 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1380 | retcode = ERR_MD_IDX_INVALID; |
1381 | goto fail; | |
1382 | } | |
1383 | ||
44ed167d PR |
1384 | rcu_read_lock(); |
1385 | nc = rcu_dereference(mdev->tconn->net_conf); | |
1386 | if (nc) { | |
daeda1cc | 1387 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1388 | rcu_read_unlock(); |
47ff2d0a PR |
1389 | retcode = ERR_STONITH_AND_PROT_A; |
1390 | goto fail; | |
1391 | } | |
1392 | } | |
44ed167d | 1393 | rcu_read_unlock(); |
47ff2d0a | 1394 | |
daeda1cc | 1395 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
d4d77629 | 1396 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, mdev); |
e525fd89 | 1397 | if (IS_ERR(bdev)) { |
daeda1cc | 1398 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1399 | PTR_ERR(bdev)); |
b411b363 PR |
1400 | retcode = ERR_OPEN_DISK; |
1401 | goto fail; | |
1402 | } | |
e525fd89 TH |
1403 | nbc->backing_bdev = bdev; |
1404 | ||
1405 | /* | |
1406 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1407 | * drbd sharing one meta device. TODO in that case, paranoia | |
1408 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1409 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1410 | * should check it for you already; but if you don't, or | |
1411 | * someone fooled it, we need to double check here) | |
1412 | */ | |
daeda1cc | 1413 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1414 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1415 | (new_disk_conf->meta_dev_idx < 0) ? |
d4d77629 | 1416 | (void *)mdev : (void *)drbd_m_holder); |
e525fd89 | 1417 | if (IS_ERR(bdev)) { |
daeda1cc | 1418 | dev_err(DEV, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1419 | PTR_ERR(bdev)); |
b411b363 PR |
1420 | retcode = ERR_OPEN_MD_DISK; |
1421 | goto fail; | |
1422 | } | |
e525fd89 | 1423 | nbc->md_bdev = bdev; |
b411b363 | 1424 | |
e525fd89 | 1425 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1426 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1427 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1428 | retcode = ERR_MD_IDX_INVALID; |
b411b363 PR |
1429 | goto fail; |
1430 | } | |
1431 | ||
1432 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1433 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1434 | offsetof(struct bm_extent, lce)); |
1435 | if (!resync_lru) { | |
1436 | retcode = ERR_NOMEM; | |
e525fd89 | 1437 | goto fail; |
b411b363 PR |
1438 | } |
1439 | ||
c04ccaa6 LE |
1440 | /* Read our meta data super block early. |
1441 | * This also sets other on-disk offsets. */ | |
1442 | retcode = drbd_md_read(mdev, nbc); | |
1443 | if (retcode != NO_ERROR) | |
1444 | goto fail; | |
b411b363 | 1445 | |
5bbcf5e6 LE |
1446 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1447 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1448 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1449 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1450 | ||
daeda1cc | 1451 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
b411b363 PR |
1452 | dev_err(DEV, "max capacity %llu smaller than disk size %llu\n", |
1453 | (unsigned long long) drbd_get_max_capacity(nbc), | |
daeda1cc | 1454 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1455 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1456 | goto fail; |
b411b363 PR |
1457 | } |
1458 | ||
daeda1cc | 1459 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1460 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1461 | /* at least one MB, otherwise it does not make sense */ | |
1462 | min_md_device_sectors = (2<<10); | |
1463 | } else { | |
1464 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1465 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1466 | } |
1467 | ||
b411b363 | 1468 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1469 | retcode = ERR_MD_DISK_TOO_SMALL; |
b411b363 PR |
1470 | dev_warn(DEV, "refusing attach: md-device too small, " |
1471 | "at least %llu sectors needed for this meta-disk type\n", | |
1472 | (unsigned long long) min_md_device_sectors); | |
e525fd89 | 1473 | goto fail; |
b411b363 PR |
1474 | } |
1475 | ||
1476 | /* Make sure the new disk is big enough | |
1477 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1478 | if (drbd_get_max_capacity(nbc) < | |
1479 | drbd_get_capacity(mdev->this_bdev)) { | |
7948bcdc | 1480 | retcode = ERR_DISK_TOO_SMALL; |
e525fd89 | 1481 | goto fail; |
b411b363 PR |
1482 | } |
1483 | ||
1484 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1485 | ||
1352994b LE |
1486 | if (nbc->known_size > max_possible_sectors) { |
1487 | dev_warn(DEV, "==> truncating very big lower level device " | |
1488 | "to currently maximum possible %llu sectors <==\n", | |
1489 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1490 | if (new_disk_conf->meta_dev_idx >= 0) |
1352994b LE |
1491 | dev_warn(DEV, "==>> using internal or flexible " |
1492 | "meta data may help <<==\n"); | |
1493 | } | |
1494 | ||
b411b363 PR |
1495 | drbd_suspend_io(mdev); |
1496 | /* also wait for the last barrier ack. */ | |
b6dd1a89 LE |
1497 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1498 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1499 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1500 | * As barriers are counted per resource, | |
1501 | * we'd need to suspend io on all devices of a resource. | |
1502 | */ | |
2aebfabb | 1503 | wait_event(mdev->misc_wait, !atomic_read(&mdev->ap_pending_cnt) || drbd_suspended(mdev)); |
b411b363 PR |
1504 | /* and for any other previously queued work */ |
1505 | drbd_flush_workqueue(mdev); | |
1506 | ||
f2024e7c AG |
1507 | rv = _drbd_request_state(mdev, NS(disk, D_ATTACHING), CS_VERBOSE); |
1508 | retcode = rv; /* FIXME: Type mismatch. */ | |
b411b363 | 1509 | drbd_resume_io(mdev); |
f2024e7c | 1510 | if (rv < SS_SUCCESS) |
e525fd89 | 1511 | goto fail; |
b411b363 PR |
1512 | |
1513 | if (!get_ldev_if_state(mdev, D_ATTACHING)) | |
1514 | goto force_diskless; | |
1515 | ||
b411b363 PR |
1516 | if (!mdev->bitmap) { |
1517 | if (drbd_bm_init(mdev)) { | |
1518 | retcode = ERR_NOMEM; | |
1519 | goto force_diskless_dec; | |
1520 | } | |
1521 | } | |
1522 | ||
b411b363 PR |
1523 | if (mdev->state.conn < C_CONNECTED && |
1524 | mdev->state.role == R_PRIMARY && | |
1525 | (mdev->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { | |
1526 | dev_err(DEV, "Can only attach to data with current UUID=%016llX\n", | |
1527 | (unsigned long long)mdev->ed_uuid); | |
1528 | retcode = ERR_DATA_NOT_CURRENT; | |
1529 | goto force_diskless_dec; | |
1530 | } | |
1531 | ||
1532 | /* Since we are diskless, fix the activity log first... */ | |
daeda1cc | 1533 | if (drbd_check_al_size(mdev, new_disk_conf)) { |
b411b363 PR |
1534 | retcode = ERR_NOMEM; |
1535 | goto force_diskless_dec; | |
1536 | } | |
1537 | ||
1538 | /* Prevent shrinking of consistent devices ! */ | |
1539 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
daeda1cc | 1540 | drbd_new_dev_size(mdev, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
b411b363 | 1541 | dev_warn(DEV, "refusing to truncate a consistent device\n"); |
7948bcdc | 1542 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1543 | goto force_diskless_dec; |
1544 | } | |
1545 | ||
b411b363 PR |
1546 | /* Reset the "barriers don't work" bits here, then force meta data to |
1547 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1548 | if (new_disk_conf->md_flushes) |
a8a4e51e | 1549 | clear_bit(MD_NO_FUA, &mdev->flags); |
b411b363 | 1550 | else |
e544046a | 1551 | set_bit(MD_NO_FUA, &mdev->flags); |
b411b363 PR |
1552 | |
1553 | /* Point of no return reached. | |
1554 | * Devices and memory are no longer released by error cleanup below. | |
1555 | * now mdev takes over responsibility, and the state engine should | |
1556 | * clean it up somewhere. */ | |
1557 | D_ASSERT(mdev->ldev == NULL); | |
1558 | mdev->ldev = nbc; | |
1559 | mdev->resync = resync_lru; | |
9958c857 | 1560 | mdev->rs_plan_s = new_plan; |
b411b363 PR |
1561 | nbc = NULL; |
1562 | resync_lru = NULL; | |
daeda1cc | 1563 | new_disk_conf = NULL; |
9958c857 | 1564 | new_plan = NULL; |
b411b363 | 1565 | |
4b0007c0 | 1566 | drbd_bump_write_ordering(mdev->tconn, WO_bdev_flush); |
b411b363 PR |
1567 | |
1568 | if (drbd_md_test_flag(mdev->ldev, MDF_CRASHED_PRIMARY)) | |
1569 | set_bit(CRASHED_PRIMARY, &mdev->flags); | |
1570 | else | |
1571 | clear_bit(CRASHED_PRIMARY, &mdev->flags); | |
1572 | ||
894c6a94 | 1573 | if (drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && |
d5d7ebd4 | 1574 | !(mdev->state.role == R_PRIMARY && mdev->tconn->susp_nod)) |
b411b363 | 1575 | set_bit(CRASHED_PRIMARY, &mdev->flags); |
b411b363 PR |
1576 | |
1577 | mdev->send_cnt = 0; | |
1578 | mdev->recv_cnt = 0; | |
1579 | mdev->read_cnt = 0; | |
1580 | mdev->writ_cnt = 0; | |
1581 | ||
99432fcc | 1582 | drbd_reconsider_max_bio_size(mdev); |
b411b363 PR |
1583 | |
1584 | /* If I am currently not R_PRIMARY, | |
1585 | * but meta data primary indicator is set, | |
1586 | * I just now recover from a hard crash, | |
1587 | * and have been R_PRIMARY before that crash. | |
1588 | * | |
1589 | * Now, if I had no connection before that crash | |
1590 | * (have been degraded R_PRIMARY), chances are that | |
1591 | * I won't find my peer now either. | |
1592 | * | |
1593 | * In that case, and _only_ in that case, | |
1594 | * we use the degr-wfc-timeout instead of the default, | |
1595 | * so we can automatically recover from a crash of a | |
1596 | * degraded but active "cluster" after a certain timeout. | |
1597 | */ | |
1598 | clear_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1599 | if (mdev->state.role != R_PRIMARY && | |
1600 | drbd_md_test_flag(mdev->ldev, MDF_PRIMARY_IND) && | |
1601 | !drbd_md_test_flag(mdev->ldev, MDF_CONNECTED_IND)) | |
1602 | set_bit(USE_DEGR_WFC_T, &mdev->flags); | |
1603 | ||
24c4830c | 1604 | dd = drbd_determine_dev_size(mdev, 0); |
b411b363 PR |
1605 | if (dd == dev_size_error) { |
1606 | retcode = ERR_NOMEM_BITMAP; | |
1607 | goto force_diskless_dec; | |
1608 | } else if (dd == grew) | |
1609 | set_bit(RESYNC_AFTER_NEG, &mdev->flags); | |
1610 | ||
9a51ab1c PR |
1611 | if (drbd_md_test_flag(mdev->ldev, MDF_FULL_SYNC) || |
1612 | (test_bit(CRASHED_PRIMARY, &mdev->flags) && | |
1613 | drbd_md_test_flag(mdev->ldev, MDF_AL_DISABLED))) { | |
b411b363 PR |
1614 | dev_info(DEV, "Assuming that all blocks are out of sync " |
1615 | "(aka FullSync)\n"); | |
20ceb2b2 LE |
1616 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, |
1617 | "set_n_write from attaching", BM_LOCKED_MASK)) { | |
b411b363 PR |
1618 | retcode = ERR_IO_MD_DISK; |
1619 | goto force_diskless_dec; | |
1620 | } | |
1621 | } else { | |
20ceb2b2 | 1622 | if (drbd_bitmap_io(mdev, &drbd_bm_read, |
22ab6a30 | 1623 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1624 | retcode = ERR_IO_MD_DISK; |
1625 | goto force_diskless_dec; | |
1626 | } | |
b411b363 PR |
1627 | } |
1628 | ||
0778286a PR |
1629 | if (_drbd_bm_total_weight(mdev) == drbd_bm_bits(mdev)) |
1630 | drbd_suspend_al(mdev); /* IO is still suspended here... */ | |
1631 | ||
87eeee41 | 1632 | spin_lock_irq(&mdev->tconn->req_lock); |
78bae59b PR |
1633 | os = drbd_read_state(mdev); |
1634 | ns = os; | |
b411b363 PR |
1635 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1636 | otherwise investigate MDF_WasUpToDate... | |
1637 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1638 | otherwise into D_CONSISTENT state. | |
1639 | */ | |
1640 | if (drbd_md_test_flag(mdev->ldev, MDF_CONSISTENT)) { | |
1641 | if (drbd_md_test_flag(mdev->ldev, MDF_WAS_UP_TO_DATE)) | |
1642 | ns.disk = D_CONSISTENT; | |
1643 | else | |
1644 | ns.disk = D_OUTDATED; | |
1645 | } else { | |
1646 | ns.disk = D_INCONSISTENT; | |
1647 | } | |
1648 | ||
1649 | if (drbd_md_test_flag(mdev->ldev, MDF_PEER_OUT_DATED)) | |
1650 | ns.pdsk = D_OUTDATED; | |
1651 | ||
daeda1cc PR |
1652 | rcu_read_lock(); |
1653 | if (ns.disk == D_CONSISTENT && | |
1654 | (ns.pdsk == D_OUTDATED || rcu_dereference(mdev->ldev->disk_conf)->fencing == FP_DONT_CARE)) | |
b411b363 PR |
1655 | ns.disk = D_UP_TO_DATE; |
1656 | ||
1657 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1658 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1659 | this point, because drbd_request_state() modifies these | |
1660 | flags. */ | |
1661 | ||
9a51ab1c | 1662 | if (rcu_dereference(mdev->ldev->disk_conf)->al_updates) |
4035e4c2 | 1663 | mdev->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c PR |
1664 | else |
1665 | mdev->ldev->md.flags |= MDF_AL_DISABLED; | |
1666 | ||
1667 | rcu_read_unlock(); | |
1668 | ||
b411b363 PR |
1669 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1670 | state after the negotiation phase. */ | |
1671 | if (mdev->state.conn == C_CONNECTED) { | |
1672 | mdev->new_state_tmp.i = ns.i; | |
1673 | ns.i = os.i; | |
1674 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1675 | |
1676 | /* We expect to receive up-to-date UUIDs soon. | |
1677 | To avoid a race in receive_state, free p_uuid while | |
1678 | holding req_lock. I.e. atomic with the state change */ | |
1679 | kfree(mdev->p_uuid); | |
1680 | mdev->p_uuid = NULL; | |
b411b363 PR |
1681 | } |
1682 | ||
1683 | rv = _drbd_set_state(mdev, ns, CS_VERBOSE, NULL); | |
87eeee41 | 1684 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
1685 | |
1686 | if (rv < SS_SUCCESS) | |
1687 | goto force_diskless_dec; | |
1688 | ||
cdfda633 PR |
1689 | mod_timer(&mdev->request_timer, jiffies + HZ); |
1690 | ||
b411b363 PR |
1691 | if (mdev->state.role == R_PRIMARY) |
1692 | mdev->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
1693 | else | |
1694 | mdev->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
1695 | ||
1696 | drbd_md_mark_dirty(mdev); | |
1697 | drbd_md_sync(mdev); | |
1698 | ||
1699 | kobject_uevent(&disk_to_dev(mdev->vdisk)->kobj, KOBJ_CHANGE); | |
1700 | put_ldev(mdev); | |
0e29d163 | 1701 | conn_reconfig_done(mdev->tconn); |
3b98c0c2 | 1702 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1703 | return 0; |
1704 | ||
1705 | force_diskless_dec: | |
1706 | put_ldev(mdev); | |
1707 | force_diskless: | |
9510b241 | 1708 | drbd_force_state(mdev, NS(disk, D_DISKLESS)); |
b411b363 | 1709 | drbd_md_sync(mdev); |
b411b363 | 1710 | fail: |
40cbf085 | 1711 | conn_reconfig_done(mdev->tconn); |
b411b363 | 1712 | if (nbc) { |
e525fd89 TH |
1713 | if (nbc->backing_bdev) |
1714 | blkdev_put(nbc->backing_bdev, | |
1715 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1716 | if (nbc->md_bdev) | |
1717 | blkdev_put(nbc->md_bdev, | |
1718 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1719 | kfree(nbc); |
1720 | } | |
daeda1cc | 1721 | kfree(new_disk_conf); |
b411b363 | 1722 | lc_destroy(resync_lru); |
9958c857 | 1723 | kfree(new_plan); |
b411b363 | 1724 | |
40cbf085 | 1725 | finish: |
3b98c0c2 | 1726 | drbd_adm_finish(info, retcode); |
b411b363 PR |
1727 | return 0; |
1728 | } | |
1729 | ||
cdfda633 | 1730 | static int adm_detach(struct drbd_conf *mdev, int force) |
b411b363 | 1731 | { |
19f83c76 | 1732 | enum drbd_state_rv retcode; |
9a0d9d03 | 1733 | int ret; |
02ee8f95 | 1734 | |
cdfda633 | 1735 | if (force) { |
0c849666 | 1736 | set_bit(FORCE_DETACH, &mdev->flags); |
02ee8f95 | 1737 | drbd_force_state(mdev, NS(disk, D_FAILED)); |
cdfda633 | 1738 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1739 | goto out; |
1740 | } | |
1741 | ||
82f59cc6 | 1742 | drbd_suspend_io(mdev); /* so no-one is stuck in drbd_al_begin_io */ |
a2e91381 | 1743 | drbd_md_get_buffer(mdev); /* make sure there is no in-flight meta-data IO */ |
9a0d9d03 | 1744 | retcode = drbd_request_state(mdev, NS(disk, D_FAILED)); |
a2e91381 | 1745 | drbd_md_put_buffer(mdev); |
9a0d9d03 LE |
1746 | /* D_FAILED will transition to DISKLESS. */ |
1747 | ret = wait_event_interruptible(mdev->misc_wait, | |
1748 | mdev->state.disk != D_FAILED); | |
82f59cc6 | 1749 | drbd_resume_io(mdev); |
9b2f61ae | 1750 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1751 | retcode = SS_NOTHING_TO_DO; |
1752 | if (ret) | |
1753 | retcode = ERR_INTR; | |
02ee8f95 | 1754 | out: |
85f75dd7 | 1755 | return retcode; |
b411b363 PR |
1756 | } |
1757 | ||
82f59cc6 LE |
1758 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1759 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1760 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1761 | * internal references as well. | |
1762 | * Only then we have finally detached. */ | |
3b98c0c2 | 1763 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1764 | { |
116676ca | 1765 | enum drbd_ret_code retcode; |
cdfda633 PR |
1766 | struct detach_parms parms = { }; |
1767 | int err; | |
b411b363 | 1768 | |
3b98c0c2 LE |
1769 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
1770 | if (!adm_ctx.reply_skb) | |
1771 | return retcode; | |
1772 | if (retcode != NO_ERROR) | |
1773 | goto out; | |
b411b363 | 1774 | |
cdfda633 PR |
1775 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1776 | err = detach_parms_from_attrs(&parms, info); | |
1777 | if (err) { | |
1778 | retcode = ERR_MANDATORY_TAG; | |
1779 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1780 | goto out; | |
1781 | } | |
b411b363 PR |
1782 | } |
1783 | ||
cdfda633 | 1784 | retcode = adm_detach(adm_ctx.mdev, parms.force_detach); |
3b98c0c2 LE |
1785 | out: |
1786 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
1787 | return 0; |
1788 | } | |
b411b363 | 1789 | |
f399002e LE |
1790 | static bool conn_resync_running(struct drbd_tconn *tconn) |
1791 | { | |
1792 | struct drbd_conf *mdev; | |
695d08fa | 1793 | bool rv = false; |
f399002e LE |
1794 | int vnr; |
1795 | ||
695d08fa | 1796 | rcu_read_lock(); |
f399002e LE |
1797 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1798 | if (mdev->state.conn == C_SYNC_SOURCE || | |
1799 | mdev->state.conn == C_SYNC_TARGET || | |
1800 | mdev->state.conn == C_PAUSED_SYNC_S || | |
695d08fa PR |
1801 | mdev->state.conn == C_PAUSED_SYNC_T) { |
1802 | rv = true; | |
1803 | break; | |
1804 | } | |
b411b363 | 1805 | } |
695d08fa | 1806 | rcu_read_unlock(); |
b411b363 | 1807 | |
695d08fa | 1808 | return rv; |
f399002e | 1809 | } |
47ff2d0a | 1810 | |
f399002e LE |
1811 | static bool conn_ov_running(struct drbd_tconn *tconn) |
1812 | { | |
1813 | struct drbd_conf *mdev; | |
695d08fa | 1814 | bool rv = false; |
f399002e LE |
1815 | int vnr; |
1816 | ||
695d08fa | 1817 | rcu_read_lock(); |
f399002e LE |
1818 | idr_for_each_entry(&tconn->volumes, mdev, vnr) { |
1819 | if (mdev->state.conn == C_VERIFY_S || | |
695d08fa PR |
1820 | mdev->state.conn == C_VERIFY_T) { |
1821 | rv = true; | |
1822 | break; | |
47ff2d0a PR |
1823 | } |
1824 | } | |
695d08fa | 1825 | rcu_read_unlock(); |
b411b363 | 1826 | |
695d08fa | 1827 | return rv; |
f399002e | 1828 | } |
422028b1 | 1829 | |
cd64397c | 1830 | static enum drbd_ret_code |
44ed167d | 1831 | _check_net_options(struct drbd_tconn *tconn, struct net_conf *old_conf, struct net_conf *new_conf) |
cd64397c PR |
1832 | { |
1833 | struct drbd_conf *mdev; | |
1834 | int i; | |
b411b363 | 1835 | |
dcb20d1a PR |
1836 | if (old_conf && tconn->cstate == C_WF_REPORT_PARAMS && tconn->agreed_pro_version < 100) { |
1837 | if (new_conf->wire_protocol != old_conf->wire_protocol) | |
1838 | return ERR_NEED_APV_100; | |
b411b363 | 1839 | |
dcb20d1a PR |
1840 | if (new_conf->two_primaries != old_conf->two_primaries) |
1841 | return ERR_NEED_APV_100; | |
1842 | ||
dcb20d1a PR |
1843 | if (strcmp(new_conf->integrity_alg, old_conf->integrity_alg)) |
1844 | return ERR_NEED_APV_100; | |
b411b363 PR |
1845 | } |
1846 | ||
dcb20d1a PR |
1847 | if (!new_conf->two_primaries && |
1848 | conn_highest_role(tconn) == R_PRIMARY && | |
1849 | conn_highest_peer(tconn) == R_PRIMARY) | |
1850 | return ERR_NEED_ALLOW_TWO_PRI; | |
b411b363 | 1851 | |
cd64397c PR |
1852 | if (new_conf->two_primaries && |
1853 | (new_conf->wire_protocol != DRBD_PROT_C)) | |
1854 | return ERR_NOT_PROTO_C; | |
1855 | ||
cd64397c PR |
1856 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
1857 | if (get_ldev(mdev)) { | |
daeda1cc | 1858 | enum drbd_fencing_p fp = rcu_dereference(mdev->ldev->disk_conf)->fencing; |
cd64397c | 1859 | put_ldev(mdev); |
44ed167d | 1860 | if (new_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 1861 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 1862 | } |
6139f60d | 1863 | if (mdev->state.role == R_PRIMARY && new_conf->discard_my_data) |
eb12010e | 1864 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
1865 | } |
1866 | ||
cd64397c PR |
1867 | if (new_conf->on_congestion != OC_BLOCK && new_conf->wire_protocol != DRBD_PROT_A) |
1868 | return ERR_CONG_NOT_PROTO_A; | |
b411b363 | 1869 | |
cd64397c PR |
1870 | return NO_ERROR; |
1871 | } | |
b411b363 | 1872 | |
44ed167d PR |
1873 | static enum drbd_ret_code |
1874 | check_net_options(struct drbd_tconn *tconn, struct net_conf *new_conf) | |
1875 | { | |
1876 | static enum drbd_ret_code rv; | |
1877 | struct drbd_conf *mdev; | |
1878 | int i; | |
b411b363 | 1879 | |
44ed167d PR |
1880 | rcu_read_lock(); |
1881 | rv = _check_net_options(tconn, rcu_dereference(tconn->net_conf), new_conf); | |
1882 | rcu_read_unlock(); | |
b411b363 | 1883 | |
44ed167d PR |
1884 | /* tconn->volumes protected by genl_lock() here */ |
1885 | idr_for_each_entry(&tconn->volumes, mdev, i) { | |
1886 | if (!mdev->bitmap) { | |
1887 | if(drbd_bm_init(mdev)) | |
1888 | return ERR_NOMEM; | |
b411b363 PR |
1889 | } |
1890 | } | |
1891 | ||
44ed167d PR |
1892 | return rv; |
1893 | } | |
b411b363 | 1894 | |
0fd0ea06 PR |
1895 | struct crypto { |
1896 | struct crypto_hash *verify_tfm; | |
1897 | struct crypto_hash *csums_tfm; | |
1898 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 1899 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 1900 | }; |
b411b363 | 1901 | |
0fd0ea06 | 1902 | static int |
4b6ad6d4 | 1903 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
1904 | { |
1905 | if (!tfm_name[0]) | |
1906 | return NO_ERROR; | |
b411b363 | 1907 | |
0fd0ea06 PR |
1908 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
1909 | if (IS_ERR(*tfm)) { | |
1910 | *tfm = NULL; | |
1911 | return err_alg; | |
b411b363 | 1912 | } |
b411b363 | 1913 | |
0fd0ea06 PR |
1914 | return NO_ERROR; |
1915 | } | |
b411b363 | 1916 | |
0fd0ea06 PR |
1917 | static enum drbd_ret_code |
1918 | alloc_crypto(struct crypto *crypto, struct net_conf *new_conf) | |
1919 | { | |
1920 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
1921 | enum drbd_ret_code rv; | |
0fd0ea06 | 1922 | |
4b6ad6d4 AG |
1923 | rv = alloc_hash(&crypto->csums_tfm, new_conf->csums_alg, |
1924 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
1925 | if (rv != NO_ERROR) |
1926 | return rv; | |
4b6ad6d4 AG |
1927 | rv = alloc_hash(&crypto->verify_tfm, new_conf->verify_alg, |
1928 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
1929 | if (rv != NO_ERROR) |
1930 | return rv; | |
4b6ad6d4 AG |
1931 | rv = alloc_hash(&crypto->integrity_tfm, new_conf->integrity_alg, |
1932 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
1933 | if (rv != NO_ERROR) |
1934 | return rv; | |
0fd0ea06 PR |
1935 | if (new_conf->cram_hmac_alg[0] != 0) { |
1936 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", | |
1937 | new_conf->cram_hmac_alg); | |
b411b363 | 1938 | |
4b6ad6d4 AG |
1939 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
1940 | ERR_AUTH_ALG); | |
b411b363 PR |
1941 | } |
1942 | ||
0fd0ea06 PR |
1943 | return rv; |
1944 | } | |
b411b363 | 1945 | |
0fd0ea06 PR |
1946 | static void free_crypto(struct crypto *crypto) |
1947 | { | |
0fd0ea06 | 1948 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 1949 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
1950 | crypto_free_hash(crypto->csums_tfm); |
1951 | crypto_free_hash(crypto->verify_tfm); | |
1952 | } | |
b411b363 | 1953 | |
f399002e LE |
1954 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
1955 | { | |
1956 | enum drbd_ret_code retcode; | |
1957 | struct drbd_tconn *tconn; | |
44ed167d | 1958 | struct net_conf *old_conf, *new_conf = NULL; |
f399002e LE |
1959 | int err; |
1960 | int ovr; /* online verify running */ | |
1961 | int rsr; /* re-sync running */ | |
0fd0ea06 | 1962 | struct crypto crypto = { }; |
b411b363 | 1963 | |
089c075d | 1964 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
1965 | if (!adm_ctx.reply_skb) |
1966 | return retcode; | |
1967 | if (retcode != NO_ERROR) | |
1968 | goto out; | |
b411b363 | 1969 | |
f399002e | 1970 | tconn = adm_ctx.tconn; |
b411b363 | 1971 | |
f399002e LE |
1972 | new_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
1973 | if (!new_conf) { | |
1974 | retcode = ERR_NOMEM; | |
1975 | goto out; | |
1976 | } | |
b411b363 | 1977 | |
44ed167d | 1978 | conn_reconfig_start(tconn); |
b411b363 | 1979 | |
88104ca4 | 1980 | mutex_lock(&tconn->data.mutex); |
a0095508 | 1981 | mutex_lock(&tconn->conf_update); |
91fd4dad | 1982 | old_conf = tconn->net_conf; |
2561b9c1 | 1983 | |
44ed167d | 1984 | if (!old_conf) { |
f399002e LE |
1985 | drbd_msg_put_info("net conf missing, try connect"); |
1986 | retcode = ERR_INVALID_REQUEST; | |
2561b9c1 PR |
1987 | goto fail; |
1988 | } | |
1989 | ||
44ed167d | 1990 | *new_conf = *old_conf; |
5979e361 | 1991 | if (should_set_defaults(info)) |
b966b5dd | 1992 | set_net_conf_defaults(new_conf); |
f399002e | 1993 | |
f399002e | 1994 | err = net_conf_from_attrs_for_change(new_conf, info); |
c75b9b10 | 1995 | if (err && err != -ENOMSG) { |
f399002e LE |
1996 | retcode = ERR_MANDATORY_TAG; |
1997 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
1998 | goto fail; | |
2561b9c1 | 1999 | } |
b411b363 | 2000 | |
cd64397c PR |
2001 | retcode = check_net_options(tconn, new_conf); |
2002 | if (retcode != NO_ERROR) | |
2003 | goto fail; | |
b411b363 | 2004 | |
f399002e LE |
2005 | /* re-sync running */ |
2006 | rsr = conn_resync_running(tconn); | |
0fd0ea06 | 2007 | if (rsr && strcmp(new_conf->csums_alg, old_conf->csums_alg)) { |
f399002e | 2008 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2009 | goto fail; |
b411b363 PR |
2010 | } |
2011 | ||
f399002e LE |
2012 | /* online verify running */ |
2013 | ovr = conn_ov_running(tconn); | |
0fd0ea06 PR |
2014 | if (ovr && strcmp(new_conf->verify_alg, old_conf->verify_alg)) { |
2015 | retcode = ERR_VERIFY_RUNNING; | |
b411b363 | 2016 | goto fail; |
f399002e | 2017 | } |
b411b363 | 2018 | |
0fd0ea06 PR |
2019 | retcode = alloc_crypto(&crypto, new_conf); |
2020 | if (retcode != NO_ERROR) | |
b411b363 | 2021 | goto fail; |
f399002e | 2022 | |
44ed167d | 2023 | rcu_assign_pointer(tconn->net_conf, new_conf); |
f399002e LE |
2024 | |
2025 | if (!rsr) { | |
2026 | crypto_free_hash(tconn->csums_tfm); | |
0fd0ea06 PR |
2027 | tconn->csums_tfm = crypto.csums_tfm; |
2028 | crypto.csums_tfm = NULL; | |
f399002e LE |
2029 | } |
2030 | if (!ovr) { | |
2031 | crypto_free_hash(tconn->verify_tfm); | |
0fd0ea06 PR |
2032 | tconn->verify_tfm = crypto.verify_tfm; |
2033 | crypto.verify_tfm = NULL; | |
b411b363 PR |
2034 | } |
2035 | ||
8d412fc6 AG |
2036 | crypto_free_hash(tconn->integrity_tfm); |
2037 | tconn->integrity_tfm = crypto.integrity_tfm; | |
d659f2aa | 2038 | if (tconn->cstate >= C_WF_REPORT_PARAMS && tconn->agreed_pro_version >= 100) |
88104ca4 | 2039 | /* Do this without trying to take tconn->data.mutex again. */ |
d659f2aa | 2040 | __drbd_send_protocol(tconn, P_PROTOCOL_UPDATE); |
0fd0ea06 | 2041 | |
0fd0ea06 PR |
2042 | crypto_free_hash(tconn->cram_hmac_tfm); |
2043 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2044 | ||
a0095508 | 2045 | mutex_unlock(&tconn->conf_update); |
88104ca4 | 2046 | mutex_unlock(&tconn->data.mutex); |
91fd4dad PR |
2047 | synchronize_rcu(); |
2048 | kfree(old_conf); | |
2049 | ||
f399002e LE |
2050 | if (tconn->cstate >= C_WF_REPORT_PARAMS) |
2051 | drbd_send_sync_param(minor_to_mdev(conn_lowest_minor(tconn))); | |
2052 | ||
91fd4dad PR |
2053 | goto done; |
2054 | ||
b411b363 | 2055 | fail: |
a0095508 | 2056 | mutex_unlock(&tconn->conf_update); |
88104ca4 | 2057 | mutex_unlock(&tconn->data.mutex); |
0fd0ea06 | 2058 | free_crypto(&crypto); |
f399002e | 2059 | kfree(new_conf); |
91fd4dad | 2060 | done: |
f399002e LE |
2061 | conn_reconfig_done(tconn); |
2062 | out: | |
2063 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2064 | return 0; |
2065 | } | |
2066 | ||
3b98c0c2 | 2067 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2068 | { |
3b98c0c2 | 2069 | struct drbd_conf *mdev; |
44ed167d | 2070 | struct net_conf *old_conf, *new_conf = NULL; |
0fd0ea06 | 2071 | struct crypto crypto = { }; |
3b98c0c2 | 2072 | struct drbd_tconn *tconn; |
3b98c0c2 LE |
2073 | enum drbd_ret_code retcode; |
2074 | int i; | |
2075 | int err; | |
b411b363 | 2076 | |
44e52cfa | 2077 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2078 | |
3b98c0c2 LE |
2079 | if (!adm_ctx.reply_skb) |
2080 | return retcode; | |
2081 | if (retcode != NO_ERROR) | |
2082 | goto out; | |
089c075d AG |
2083 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
2084 | drbd_msg_put_info("connection endpoint(s) missing"); | |
2085 | retcode = ERR_INVALID_REQUEST; | |
2086 | goto out; | |
2087 | } | |
b411b363 | 2088 | |
089c075d AG |
2089 | /* No need for _rcu here. All reconfiguration is |
2090 | * strictly serialized on genl_lock(). We are protected against | |
2091 | * concurrent reconfiguration/addition/deletion */ | |
2092 | list_for_each_entry(tconn, &drbd_tconns, all_tconn) { | |
2093 | if (nla_len(adm_ctx.my_addr) == tconn->my_addr_len && | |
2094 | !memcmp(nla_data(adm_ctx.my_addr), &tconn->my_addr, tconn->my_addr_len)) { | |
2095 | retcode = ERR_LOCAL_ADDR; | |
2096 | goto out; | |
2097 | } | |
b411b363 | 2098 | |
089c075d AG |
2099 | if (nla_len(adm_ctx.peer_addr) == tconn->peer_addr_len && |
2100 | !memcmp(nla_data(adm_ctx.peer_addr), &tconn->peer_addr, tconn->peer_addr_len)) { | |
2101 | retcode = ERR_PEER_ADDR; | |
2102 | goto out; | |
2103 | } | |
b411b363 PR |
2104 | } |
2105 | ||
3b98c0c2 | 2106 | tconn = adm_ctx.tconn; |
80883197 | 2107 | conn_reconfig_start(tconn); |
b411b363 | 2108 | |
80883197 | 2109 | if (tconn->cstate > C_STANDALONE) { |
b411b363 | 2110 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2111 | goto fail; |
2112 | } | |
2113 | ||
a209b4ae | 2114 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
5979e361 | 2115 | new_conf = kzalloc(sizeof(*new_conf), GFP_KERNEL); |
b411b363 PR |
2116 | if (!new_conf) { |
2117 | retcode = ERR_NOMEM; | |
b411b363 PR |
2118 | goto fail; |
2119 | } | |
2120 | ||
b966b5dd | 2121 | set_net_conf_defaults(new_conf); |
b411b363 | 2122 | |
f399002e | 2123 | err = net_conf_from_attrs(new_conf, info); |
25e40932 | 2124 | if (err && err != -ENOMSG) { |
b411b363 | 2125 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2126 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2127 | goto fail; |
2128 | } | |
2129 | ||
cd64397c PR |
2130 | retcode = check_net_options(tconn, new_conf); |
2131 | if (retcode != NO_ERROR) | |
422028b1 | 2132 | goto fail; |
b411b363 | 2133 | |
0fd0ea06 PR |
2134 | retcode = alloc_crypto(&crypto, new_conf); |
2135 | if (retcode != NO_ERROR) | |
2136 | goto fail; | |
b411b363 | 2137 | |
b411b363 | 2138 | ((char *)new_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2139 | |
80883197 | 2140 | conn_flush_workqueue(tconn); |
b411b363 | 2141 | |
a0095508 | 2142 | mutex_lock(&tconn->conf_update); |
91fd4dad PR |
2143 | old_conf = tconn->net_conf; |
2144 | if (old_conf) { | |
b411b363 | 2145 | retcode = ERR_NET_CONFIGURED; |
a0095508 | 2146 | mutex_unlock(&tconn->conf_update); |
b411b363 PR |
2147 | goto fail; |
2148 | } | |
44ed167d | 2149 | rcu_assign_pointer(tconn->net_conf, new_conf); |
b411b363 | 2150 | |
91fd4dad | 2151 | conn_free_crypto(tconn); |
0fd0ea06 | 2152 | tconn->cram_hmac_tfm = crypto.cram_hmac_tfm; |
8d412fc6 | 2153 | tconn->integrity_tfm = crypto.integrity_tfm; |
0fd0ea06 PR |
2154 | tconn->csums_tfm = crypto.csums_tfm; |
2155 | tconn->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2156 | |
089c075d AG |
2157 | tconn->my_addr_len = nla_len(adm_ctx.my_addr); |
2158 | memcpy(&tconn->my_addr, nla_data(adm_ctx.my_addr), tconn->my_addr_len); | |
2159 | tconn->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2160 | memcpy(&tconn->peer_addr, nla_data(adm_ctx.peer_addr), tconn->peer_addr_len); | |
b411b363 | 2161 | |
a0095508 | 2162 | mutex_unlock(&tconn->conf_update); |
b411b363 | 2163 | |
695d08fa | 2164 | rcu_read_lock(); |
80883197 PR |
2165 | idr_for_each_entry(&tconn->volumes, mdev, i) { |
2166 | mdev->send_cnt = 0; | |
2167 | mdev->recv_cnt = 0; | |
b411b363 | 2168 | } |
695d08fa | 2169 | rcu_read_unlock(); |
b411b363 | 2170 | |
5ee743e9 | 2171 | retcode = conn_request_state(tconn, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2172 | |
80883197 | 2173 | conn_reconfig_done(tconn); |
3b98c0c2 | 2174 | drbd_adm_finish(info, retcode); |
b411b363 | 2175 | return 0; |
b411b363 | 2176 | |
b411b363 | 2177 | fail: |
0fd0ea06 | 2178 | free_crypto(&crypto); |
b411b363 | 2179 | kfree(new_conf); |
b411b363 | 2180 | |
80883197 | 2181 | conn_reconfig_done(tconn); |
3b98c0c2 LE |
2182 | out: |
2183 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2184 | return 0; |
2185 | } | |
2186 | ||
85f75dd7 LE |
2187 | static enum drbd_state_rv conn_try_disconnect(struct drbd_tconn *tconn, bool force) |
2188 | { | |
2189 | enum drbd_state_rv rv; | |
85f75dd7 | 2190 | |
f3dfa40a LE |
2191 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2192 | force ? CS_HARD : 0); | |
85f75dd7 LE |
2193 | |
2194 | switch (rv) { | |
2195 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2196 | break; |
85f75dd7 LE |
2197 | case SS_ALREADY_STANDALONE: |
2198 | return SS_SUCCESS; | |
2199 | case SS_PRIMARY_NOP: | |
2200 | /* Our state checking code wants to see the peer outdated. */ | |
2201 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
f3dfa40a | 2202 | pdsk, D_OUTDATED), CS_VERBOSE); |
85f75dd7 LE |
2203 | break; |
2204 | case SS_CW_FAILED_BY_PEER: | |
2205 | /* The peer probably wants to see us outdated. */ | |
2206 | rv = conn_request_state(tconn, NS2(conn, C_DISCONNECTING, | |
2207 | disk, D_OUTDATED), 0); | |
2208 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
f3dfa40a LE |
2209 | rv = conn_request_state(tconn, NS(conn, C_DISCONNECTING), |
2210 | CS_HARD); | |
b411b363 | 2211 | } |
85f75dd7 LE |
2212 | break; |
2213 | default:; | |
2214 | /* no special handling necessary */ | |
2215 | } | |
2216 | ||
f3dfa40a LE |
2217 | if (rv >= SS_SUCCESS) { |
2218 | enum drbd_state_rv rv2; | |
2219 | /* No one else can reconfigure the network while I am here. | |
2220 | * The state handling only uses drbd_thread_stop_nowait(), | |
2221 | * we want to really wait here until the receiver is no more. | |
2222 | */ | |
2223 | drbd_thread_stop(&adm_ctx.tconn->receiver); | |
2224 | ||
2225 | /* Race breaker. This additional state change request may be | |
2226 | * necessary, if this was a forced disconnect during a receiver | |
2227 | * restart. We may have "killed" the receiver thread just | |
2228 | * after drbdd_init() returned. Typically, we should be | |
2229 | * C_STANDALONE already, now, and this becomes a no-op. | |
2230 | */ | |
2231 | rv2 = conn_request_state(tconn, NS(conn, C_STANDALONE), | |
2232 | CS_VERBOSE | CS_HARD); | |
2233 | if (rv2 < SS_SUCCESS) | |
2234 | conn_err(tconn, | |
2235 | "unexpected rv2=%d in conn_try_disconnect()\n", | |
2236 | rv2); | |
b411b363 | 2237 | } |
85f75dd7 LE |
2238 | return rv; |
2239 | } | |
b411b363 | 2240 | |
3b98c0c2 | 2241 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2242 | { |
3b98c0c2 LE |
2243 | struct disconnect_parms parms; |
2244 | struct drbd_tconn *tconn; | |
85f75dd7 | 2245 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2246 | enum drbd_ret_code retcode; |
2247 | int err; | |
2561b9c1 | 2248 | |
089c075d | 2249 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2250 | if (!adm_ctx.reply_skb) |
2251 | return retcode; | |
2252 | if (retcode != NO_ERROR) | |
2561b9c1 | 2253 | goto fail; |
b411b363 | 2254 | |
3b98c0c2 LE |
2255 | tconn = adm_ctx.tconn; |
2256 | memset(&parms, 0, sizeof(parms)); | |
2257 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2258 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2259 | if (err) { |
2260 | retcode = ERR_MANDATORY_TAG; | |
2261 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
b411b363 PR |
2262 | goto fail; |
2263 | } | |
2264 | } | |
2265 | ||
85f75dd7 LE |
2266 | rv = conn_try_disconnect(tconn, parms.force_disconnect); |
2267 | if (rv < SS_SUCCESS) | |
f3dfa40a LE |
2268 | retcode = rv; /* FIXME: Type mismatch. */ |
2269 | else | |
2270 | retcode = NO_ERROR; | |
b411b363 | 2271 | fail: |
3b98c0c2 | 2272 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2273 | return 0; |
2274 | } | |
2275 | ||
2276 | void resync_after_online_grow(struct drbd_conf *mdev) | |
2277 | { | |
2278 | int iass; /* I am sync source */ | |
2279 | ||
2280 | dev_info(DEV, "Resync of new storage after online grow\n"); | |
2281 | if (mdev->state.role != mdev->state.peer) | |
2282 | iass = (mdev->state.role == R_PRIMARY); | |
2283 | else | |
427c0434 | 2284 | iass = test_bit(RESOLVE_CONFLICTS, &mdev->tconn->flags); |
b411b363 PR |
2285 | |
2286 | if (iass) | |
2287 | drbd_start_resync(mdev, C_SYNC_SOURCE); | |
2288 | else | |
2289 | _drbd_request_state(mdev, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); | |
2290 | } | |
2291 | ||
3b98c0c2 | 2292 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2293 | { |
daeda1cc | 2294 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 LE |
2295 | struct resize_parms rs; |
2296 | struct drbd_conf *mdev; | |
2297 | enum drbd_ret_code retcode; | |
b411b363 | 2298 | enum determine_dev_size dd; |
6495d2c6 | 2299 | enum dds_flags ddsf; |
daeda1cc | 2300 | sector_t u_size; |
3b98c0c2 | 2301 | int err; |
b411b363 | 2302 | |
3b98c0c2 LE |
2303 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2304 | if (!adm_ctx.reply_skb) | |
2305 | return retcode; | |
2306 | if (retcode != NO_ERROR) | |
b411b363 | 2307 | goto fail; |
3b98c0c2 LE |
2308 | |
2309 | memset(&rs, 0, sizeof(struct resize_parms)); | |
2310 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { | |
f399002e | 2311 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2312 | if (err) { |
3b98c0c2 LE |
2313 | retcode = ERR_MANDATORY_TAG; |
2314 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
b411b363 PR |
2315 | goto fail; |
2316 | } | |
2317 | } | |
2318 | ||
3b98c0c2 | 2319 | mdev = adm_ctx.mdev; |
b411b363 PR |
2320 | if (mdev->state.conn > C_CONNECTED) { |
2321 | retcode = ERR_RESIZE_RESYNC; | |
2322 | goto fail; | |
b411b363 | 2323 | } |
b411b363 | 2324 | |
b411b363 PR |
2325 | if (mdev->state.role == R_SECONDARY && |
2326 | mdev->state.peer == R_SECONDARY) { | |
2327 | retcode = ERR_NO_PRIMARY; | |
2328 | goto fail; | |
2329 | } | |
ef50a3e3 | 2330 | |
b411b363 PR |
2331 | if (!get_ldev(mdev)) { |
2332 | retcode = ERR_NO_DISK; | |
b411b363 | 2333 | goto fail; |
b411b363 | 2334 | } |
b411b363 | 2335 | |
31890f4a | 2336 | if (rs.no_resync && mdev->tconn->agreed_pro_version < 93) { |
6495d2c6 | 2337 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2338 | goto fail_ldev; |
6495d2c6 PR |
2339 | } |
2340 | ||
daeda1cc PR |
2341 | rcu_read_lock(); |
2342 | u_size = rcu_dereference(mdev->ldev->disk_conf)->disk_size; | |
2343 | rcu_read_unlock(); | |
2344 | if (u_size != (sector_t)rs.resize_size) { | |
2345 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2346 | if (!new_disk_conf) { | |
778f271d | 2347 | retcode = ERR_NOMEM; |
9bcd2521 | 2348 | goto fail_ldev; |
778f271d PR |
2349 | } |
2350 | } | |
2351 | ||
087c2492 | 2352 | if (mdev->ldev->known_size != drbd_get_capacity(mdev->ldev->backing_bdev)) |
b411b363 | 2353 | mdev->ldev->known_size = drbd_get_capacity(mdev->ldev->backing_bdev); |
b411b363 | 2354 | |
daeda1cc PR |
2355 | if (new_disk_conf) { |
2356 | mutex_lock(&mdev->tconn->conf_update); | |
2357 | old_disk_conf = mdev->ldev->disk_conf; | |
2358 | *new_disk_conf = *old_disk_conf; | |
2359 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
2360 | rcu_assign_pointer(mdev->ldev->disk_conf, new_disk_conf); | |
2361 | mutex_unlock(&mdev->tconn->conf_update); | |
2362 | synchronize_rcu(); | |
2363 | kfree(old_disk_conf); | |
b411b363 PR |
2364 | } |
2365 | ||
6495d2c6 | 2366 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
24c4830c | 2367 | dd = drbd_determine_dev_size(mdev, ddsf); |
b411b363 PR |
2368 | drbd_md_sync(mdev); |
2369 | put_ldev(mdev); | |
2370 | if (dd == dev_size_error) { | |
2371 | retcode = ERR_NOMEM_BITMAP; | |
2372 | goto fail; | |
b411b363 | 2373 | } |
778f271d | 2374 | |
087c2492 | 2375 | if (mdev->state.conn == C_CONNECTED) { |
b411b363 PR |
2376 | if (dd == grew) |
2377 | set_bit(RESIZE_PENDING, &mdev->flags); | |
2378 | ||
2379 | drbd_send_uuids(mdev); | |
6495d2c6 | 2380 | drbd_send_sizes(mdev, 1, ddsf); |
778f271d PR |
2381 | } |
2382 | ||
b411b363 | 2383 | fail: |
3b98c0c2 | 2384 | drbd_adm_finish(info, retcode); |
b411b363 | 2385 | return 0; |
b411b363 | 2386 | |
9bcd2521 PR |
2387 | fail_ldev: |
2388 | put_ldev(mdev); | |
2389 | goto fail; | |
b411b363 | 2390 | } |
b411b363 | 2391 | |
f399002e | 2392 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2393 | { |
3b98c0c2 | 2394 | enum drbd_ret_code retcode; |
f399002e | 2395 | struct drbd_tconn *tconn; |
b57a1e27 | 2396 | struct res_opts res_opts; |
f399002e | 2397 | int err; |
b411b363 | 2398 | |
44e52cfa | 2399 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2400 | if (!adm_ctx.reply_skb) |
2401 | return retcode; | |
2402 | if (retcode != NO_ERROR) | |
2403 | goto fail; | |
f399002e | 2404 | tconn = adm_ctx.tconn; |
b411b363 | 2405 | |
b57a1e27 | 2406 | res_opts = tconn->res_opts; |
5979e361 | 2407 | if (should_set_defaults(info)) |
b966b5dd | 2408 | set_res_opts_defaults(&res_opts); |
b411b363 | 2409 | |
b57a1e27 | 2410 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2411 | if (err && err != -ENOMSG) { |
b411b363 | 2412 | retcode = ERR_MANDATORY_TAG; |
3b98c0c2 | 2413 | drbd_msg_put_info(from_attrs_err_to_txt(err)); |
b411b363 PR |
2414 | goto fail; |
2415 | } | |
2416 | ||
afbbfa88 AG |
2417 | err = set_resource_options(tconn, &res_opts); |
2418 | if (err) { | |
2419 | retcode = ERR_INVALID_REQUEST; | |
2420 | if (err == -ENOMEM) | |
2421 | retcode = ERR_NOMEM; | |
b411b363 PR |
2422 | } |
2423 | ||
b411b363 | 2424 | fail: |
3b98c0c2 | 2425 | drbd_adm_finish(info, retcode); |
b411b363 PR |
2426 | return 0; |
2427 | } | |
2428 | ||
3b98c0c2 | 2429 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2430 | { |
3b98c0c2 LE |
2431 | struct drbd_conf *mdev; |
2432 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2433 | ||
2434 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2435 | if (!adm_ctx.reply_skb) | |
2436 | return retcode; | |
2437 | if (retcode != NO_ERROR) | |
2438 | goto out; | |
2439 | ||
2440 | mdev = adm_ctx.mdev; | |
b411b363 | 2441 | |
194bfb32 | 2442 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2443 | * resync just being finished, wait for it before requesting a new resync. |
2444 | * Also wait for it's after_state_ch(). */ | |
a574daf5 | 2445 | drbd_suspend_io(mdev); |
194bfb32 | 2446 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); |
7ee1fb93 | 2447 | drbd_flush_workqueue(mdev); |
194bfb32 | 2448 | |
b411b363 PR |
2449 | retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T), CS_ORDERED); |
2450 | ||
9376d9f8 PR |
2451 | /* If that did not work, try again, |
2452 | * but log failures this time (implicit CS_VERBOSE). | |
2453 | * | |
2454 | * If we happen to be C_STANDALONE R_SECONDARY, | |
2455 | * just change to D_INCONSISTENT, and set all bits in the bitmap. | |
2456 | * Otherwise, we just fail, to avoid races with the resync handshake. | |
2457 | */ | |
2458 | if (retcode < SS_SUCCESS) { | |
2459 | if (mdev->state.conn == C_STANDALONE && mdev->state.role == R_SECONDARY) { | |
2460 | retcode = drbd_request_state(mdev, NS(disk, D_INCONSISTENT)); | |
2461 | if (retcode >= SS_SUCCESS) { | |
2462 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_n_write, | |
2463 | "set_n_write from invalidate", BM_LOCKED_MASK)) | |
2464 | retcode = ERR_IO_MD_DISK; | |
2465 | } | |
2466 | } else | |
2467 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_T)); | |
b411b363 | 2468 | } |
a574daf5 | 2469 | drbd_resume_io(mdev); |
b411b363 | 2470 | |
3b98c0c2 LE |
2471 | out: |
2472 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2473 | return 0; |
2474 | } | |
2475 | ||
3b98c0c2 LE |
2476 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2477 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2478 | { |
3b98c0c2 | 2479 | enum drbd_ret_code retcode; |
194bfb32 | 2480 | |
3b98c0c2 LE |
2481 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2482 | if (!adm_ctx.reply_skb) | |
2483 | return retcode; | |
2484 | if (retcode != NO_ERROR) | |
2485 | goto out; | |
b411b363 | 2486 | |
3b98c0c2 LE |
2487 | retcode = drbd_request_state(adm_ctx.mdev, mask, val); |
2488 | out: | |
2489 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2490 | return 0; |
2491 | } | |
2492 | ||
0778286a PR |
2493 | static int drbd_bmio_set_susp_al(struct drbd_conf *mdev) |
2494 | { | |
2495 | int rv; | |
2496 | ||
2497 | rv = drbd_bmio_set_n_write(mdev); | |
2498 | drbd_suspend_al(mdev); | |
2499 | return rv; | |
2500 | } | |
2501 | ||
3b98c0c2 | 2502 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2503 | { |
25b0d6c8 PR |
2504 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
2505 | struct drbd_conf *mdev; | |
2506 | ||
2507 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2508 | if (!adm_ctx.reply_skb) | |
2509 | return retcode; | |
2510 | if (retcode != NO_ERROR) | |
2511 | goto out; | |
2512 | ||
2513 | mdev = adm_ctx.mdev; | |
b411b363 | 2514 | |
194bfb32 | 2515 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2516 | * resync just being finished, wait for it before requesting a new resync. |
2517 | * Also wait for it's after_state_ch(). */ | |
a574daf5 | 2518 | drbd_suspend_io(mdev); |
5016b82a | 2519 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); |
7ee1fb93 | 2520 | drbd_flush_workqueue(mdev); |
194bfb32 | 2521 | |
0778286a | 2522 | retcode = _drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S), CS_ORDERED); |
0778286a | 2523 | if (retcode < SS_SUCCESS) { |
9376d9f8 PR |
2524 | if (mdev->state.conn == C_STANDALONE && mdev->state.role == R_PRIMARY) { |
2525 | /* The peer will get a resync upon connect anyways. Just make that | |
2526 | into a full resync. */ | |
0778286a PR |
2527 | retcode = drbd_request_state(mdev, NS(pdsk, D_INCONSISTENT)); |
2528 | if (retcode >= SS_SUCCESS) { | |
0778286a | 2529 | if (drbd_bitmap_io(mdev, &drbd_bmio_set_susp_al, |
25b0d6c8 PR |
2530 | "set_n_write from invalidate_peer", |
2531 | BM_LOCKED_SET_ALLOWED)) | |
0778286a PR |
2532 | retcode = ERR_IO_MD_DISK; |
2533 | } | |
2534 | } else | |
2535 | retcode = drbd_request_state(mdev, NS(conn, C_STARTING_SYNC_S)); | |
2536 | } | |
a574daf5 | 2537 | drbd_resume_io(mdev); |
b411b363 | 2538 | |
25b0d6c8 PR |
2539 | out: |
2540 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2541 | return 0; |
2542 | } | |
2543 | ||
3b98c0c2 | 2544 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2545 | { |
3b98c0c2 | 2546 | enum drbd_ret_code retcode; |
b411b363 | 2547 | |
3b98c0c2 LE |
2548 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2549 | if (!adm_ctx.reply_skb) | |
2550 | return retcode; | |
2551 | if (retcode != NO_ERROR) | |
2552 | goto out; | |
b411b363 | 2553 | |
3b98c0c2 LE |
2554 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
2555 | retcode = ERR_PAUSE_IS_SET; | |
2556 | out: | |
2557 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2558 | return 0; |
2559 | } | |
2560 | ||
3b98c0c2 | 2561 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2562 | { |
da9fbc27 | 2563 | union drbd_dev_state s; |
3b98c0c2 LE |
2564 | enum drbd_ret_code retcode; |
2565 | ||
2566 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2567 | if (!adm_ctx.reply_skb) | |
2568 | return retcode; | |
2569 | if (retcode != NO_ERROR) | |
2570 | goto out; | |
b411b363 | 2571 | |
3b98c0c2 LE |
2572 | if (drbd_request_state(adm_ctx.mdev, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2573 | s = adm_ctx.mdev->state; | |
cd88d030 PR |
2574 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2575 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2576 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2577 | } else { | |
2578 | retcode = ERR_PAUSE_IS_CLEAR; | |
2579 | } | |
2580 | } | |
b411b363 | 2581 | |
3b98c0c2 LE |
2582 | out: |
2583 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2584 | return 0; |
2585 | } | |
2586 | ||
3b98c0c2 | 2587 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2588 | { |
3b98c0c2 | 2589 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2590 | } |
2591 | ||
3b98c0c2 | 2592 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2593 | { |
3b98c0c2 LE |
2594 | struct drbd_conf *mdev; |
2595 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ | |
2596 | ||
2597 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); | |
2598 | if (!adm_ctx.reply_skb) | |
2599 | return retcode; | |
2600 | if (retcode != NO_ERROR) | |
2601 | goto out; | |
2602 | ||
2603 | mdev = adm_ctx.mdev; | |
43a5182c PR |
2604 | if (test_bit(NEW_CUR_UUID, &mdev->flags)) { |
2605 | drbd_uuid_new_current(mdev); | |
2606 | clear_bit(NEW_CUR_UUID, &mdev->flags); | |
43a5182c | 2607 | } |
265be2d0 | 2608 | drbd_suspend_io(mdev); |
3b98c0c2 LE |
2609 | retcode = drbd_request_state(mdev, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); |
2610 | if (retcode == SS_SUCCESS) { | |
265be2d0 | 2611 | if (mdev->state.conn < C_CONNECTED) |
2f5cdd0b | 2612 | tl_clear(mdev->tconn); |
265be2d0 | 2613 | if (mdev->state.disk == D_DISKLESS || mdev->state.disk == D_FAILED) |
2f5cdd0b | 2614 | tl_restart(mdev->tconn, FAIL_FROZEN_DISK_IO); |
265be2d0 PR |
2615 | } |
2616 | drbd_resume_io(mdev); | |
2617 | ||
3b98c0c2 LE |
2618 | out: |
2619 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
2620 | return 0; |
2621 | } | |
2622 | ||
3b98c0c2 | 2623 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2624 | { |
3b98c0c2 | 2625 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2626 | } |
2627 | ||
089c075d | 2628 | int nla_put_drbd_cfg_context(struct sk_buff *skb, struct drbd_tconn *tconn, unsigned vnr) |
b411b363 | 2629 | { |
543cc10b LE |
2630 | struct nlattr *nla; |
2631 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2632 | if (!nla) | |
2633 | goto nla_put_failure; | |
26ec9287 AG |
2634 | if (vnr != VOLUME_UNSPECIFIED && |
2635 | nla_put_u32(skb, T_ctx_volume, vnr)) | |
2636 | goto nla_put_failure; | |
2637 | if (nla_put_string(skb, T_ctx_resource_name, tconn->name)) | |
2638 | goto nla_put_failure; | |
2639 | if (tconn->my_addr_len && | |
2640 | nla_put(skb, T_ctx_my_addr, tconn->my_addr_len, &tconn->my_addr)) | |
2641 | goto nla_put_failure; | |
2642 | if (tconn->peer_addr_len && | |
2643 | nla_put(skb, T_ctx_peer_addr, tconn->peer_addr_len, &tconn->peer_addr)) | |
2644 | goto nla_put_failure; | |
543cc10b LE |
2645 | nla_nest_end(skb, nla); |
2646 | return 0; | |
b411b363 | 2647 | |
543cc10b LE |
2648 | nla_put_failure: |
2649 | if (nla) | |
2650 | nla_nest_cancel(skb, nla); | |
2651 | return -EMSGSIZE; | |
2652 | } | |
b411b363 | 2653 | |
3b98c0c2 LE |
2654 | int nla_put_status_info(struct sk_buff *skb, struct drbd_conf *mdev, |
2655 | const struct sib_info *sib) | |
b411b363 | 2656 | { |
3b98c0c2 | 2657 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
44ed167d | 2658 | struct net_conf *nc; |
3b98c0c2 LE |
2659 | struct nlattr *nla; |
2660 | int got_ldev; | |
3b98c0c2 LE |
2661 | int err = 0; |
2662 | int exclude_sensitive; | |
2663 | ||
2664 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
2665 | * to. So we better exclude_sensitive information. | |
2666 | * | |
2667 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
2668 | * in the context of the requesting user process. Exclude sensitive | |
2669 | * information, unless current has superuser. | |
2670 | * | |
2671 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
2672 | * relies on the current implementation of netlink_dump(), which | |
2673 | * executes the dump callback successively from netlink_recvmsg(), | |
2674 | * always in the context of the receiving process */ | |
2675 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
2676 | ||
2677 | got_ldev = get_ldev(mdev); | |
3b98c0c2 LE |
2678 | |
2679 | /* We need to add connection name and volume number information still. | |
2680 | * Minor number is in drbd_genlmsghdr. */ | |
089c075d | 2681 | if (nla_put_drbd_cfg_context(skb, mdev->tconn, mdev->vnr)) |
3b98c0c2 | 2682 | goto nla_put_failure; |
3b98c0c2 | 2683 | |
f399002e LE |
2684 | if (res_opts_to_skb(skb, &mdev->tconn->res_opts, exclude_sensitive)) |
2685 | goto nla_put_failure; | |
2686 | ||
daeda1cc | 2687 | rcu_read_lock(); |
3b98c0c2 | 2688 | if (got_ldev) |
daeda1cc | 2689 | if (disk_conf_to_skb(skb, rcu_dereference(mdev->ldev->disk_conf), exclude_sensitive)) |
3b98c0c2 | 2690 | goto nla_put_failure; |
44ed167d | 2691 | |
44ed167d PR |
2692 | nc = rcu_dereference(mdev->tconn->net_conf); |
2693 | if (nc) | |
2694 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
2695 | rcu_read_unlock(); | |
2696 | if (err) | |
2697 | goto nla_put_failure; | |
3b98c0c2 | 2698 | |
3b98c0c2 LE |
2699 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
2700 | if (!nla) | |
2701 | goto nla_put_failure; | |
26ec9287 AG |
2702 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
2703 | nla_put_u32(skb, T_current_state, mdev->state.i) || | |
2704 | nla_put_u64(skb, T_ed_uuid, mdev->ed_uuid) || | |
3174f8c5 PM |
2705 | nla_put_u64(skb, T_capacity, drbd_get_capacity(mdev->this_bdev)) || |
2706 | nla_put_u64(skb, T_send_cnt, mdev->send_cnt) || | |
2707 | nla_put_u64(skb, T_recv_cnt, mdev->recv_cnt) || | |
2708 | nla_put_u64(skb, T_read_cnt, mdev->read_cnt) || | |
2709 | nla_put_u64(skb, T_writ_cnt, mdev->writ_cnt) || | |
2710 | nla_put_u64(skb, T_al_writ_cnt, mdev->al_writ_cnt) || | |
2711 | nla_put_u64(skb, T_bm_writ_cnt, mdev->bm_writ_cnt) || | |
2712 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&mdev->ap_bio_cnt)) || | |
2713 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&mdev->ap_pending_cnt)) || | |
2714 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&mdev->rs_pending_cnt))) | |
26ec9287 | 2715 | goto nla_put_failure; |
3b98c0c2 LE |
2716 | |
2717 | if (got_ldev) { | |
39a1aa7f | 2718 | int err; |
b411b363 | 2719 | |
39a1aa7f PR |
2720 | spin_lock_irq(&mdev->ldev->md.uuid_lock); |
2721 | err = nla_put(skb, T_uuids, sizeof(si->uuids), mdev->ldev->md.uuid); | |
2722 | spin_unlock_irq(&mdev->ldev->md.uuid_lock); | |
2723 | ||
2724 | if (err) | |
2725 | goto nla_put_failure; | |
2726 | ||
26ec9287 | 2727 | if (nla_put_u32(skb, T_disk_flags, mdev->ldev->md.flags) || |
26ec9287 AG |
2728 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(mdev)) || |
2729 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(mdev))) | |
2730 | goto nla_put_failure; | |
3b98c0c2 LE |
2731 | if (C_SYNC_SOURCE <= mdev->state.conn && |
2732 | C_PAUSED_SYNC_T >= mdev->state.conn) { | |
26ec9287 AG |
2733 | if (nla_put_u64(skb, T_bits_rs_total, mdev->rs_total) || |
2734 | nla_put_u64(skb, T_bits_rs_failed, mdev->rs_failed)) | |
2735 | goto nla_put_failure; | |
3b98c0c2 | 2736 | } |
b411b363 | 2737 | } |
b411b363 | 2738 | |
3b98c0c2 LE |
2739 | if (sib) { |
2740 | switch(sib->sib_reason) { | |
2741 | case SIB_SYNC_PROGRESS: | |
2742 | case SIB_GET_STATUS_REPLY: | |
2743 | break; | |
2744 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
2745 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
2746 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
2747 | goto nla_put_failure; | |
3b98c0c2 LE |
2748 | break; |
2749 | case SIB_HELPER_POST: | |
26ec9287 AG |
2750 | if (nla_put_u32(skb, T_helper_exit_code, |
2751 | sib->helper_exit_code)) | |
2752 | goto nla_put_failure; | |
3b98c0c2 LE |
2753 | /* fall through */ |
2754 | case SIB_HELPER_PRE: | |
26ec9287 AG |
2755 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
2756 | goto nla_put_failure; | |
3b98c0c2 LE |
2757 | break; |
2758 | } | |
b411b363 | 2759 | } |
3b98c0c2 | 2760 | nla_nest_end(skb, nla); |
b411b363 | 2761 | |
3b98c0c2 LE |
2762 | if (0) |
2763 | nla_put_failure: | |
2764 | err = -EMSGSIZE; | |
2765 | if (got_ldev) | |
2766 | put_ldev(mdev); | |
3b98c0c2 | 2767 | return err; |
b411b363 PR |
2768 | } |
2769 | ||
3b98c0c2 | 2770 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2771 | { |
3b98c0c2 LE |
2772 | enum drbd_ret_code retcode; |
2773 | int err; | |
b411b363 | 2774 | |
3b98c0c2 LE |
2775 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2776 | if (!adm_ctx.reply_skb) | |
2777 | return retcode; | |
2778 | if (retcode != NO_ERROR) | |
2779 | goto out; | |
b411b363 | 2780 | |
3b98c0c2 LE |
2781 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.mdev, NULL); |
2782 | if (err) { | |
2783 | nlmsg_free(adm_ctx.reply_skb); | |
2784 | return err; | |
b411b363 | 2785 | } |
3b98c0c2 LE |
2786 | out: |
2787 | drbd_adm_finish(info, retcode); | |
2788 | return 0; | |
b411b363 PR |
2789 | } |
2790 | ||
71932efc | 2791 | int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2792 | { |
3b98c0c2 LE |
2793 | struct drbd_conf *mdev; |
2794 | struct drbd_genlmsghdr *dh; | |
543cc10b LE |
2795 | struct drbd_tconn *pos = (struct drbd_tconn*)cb->args[0]; |
2796 | struct drbd_tconn *tconn = NULL; | |
2797 | struct drbd_tconn *tmp; | |
2798 | unsigned volume = cb->args[1]; | |
2799 | ||
2800 | /* Open coded, deferred, iteration: | |
2801 | * list_for_each_entry_safe(tconn, tmp, &drbd_tconns, all_tconn) { | |
2802 | * idr_for_each_entry(&tconn->volumes, mdev, i) { | |
2803 | * ... | |
2804 | * } | |
2805 | * } | |
2806 | * where tconn is cb->args[0]; | |
2807 | * and i is cb->args[1]; | |
2808 | * | |
71932efc LE |
2809 | * cb->args[2] indicates if we shall loop over all resources, |
2810 | * or just dump all volumes of a single resource. | |
2811 | * | |
3b98c0c2 LE |
2812 | * This may miss entries inserted after this dump started, |
2813 | * or entries deleted before they are reached. | |
543cc10b LE |
2814 | * |
2815 | * We need to make sure the mdev won't disappear while | |
2816 | * we are looking at it, and revalidate our iterators | |
2817 | * on each iteration. | |
2818 | */ | |
b411b363 | 2819 | |
9dc9fbb3 | 2820 | /* synchronize with conn_create()/conn_destroy() */ |
c141ebda | 2821 | rcu_read_lock(); |
543cc10b | 2822 | /* revalidate iterator position */ |
ec0bddbc | 2823 | list_for_each_entry_rcu(tmp, &drbd_tconns, all_tconn) { |
543cc10b LE |
2824 | if (pos == NULL) { |
2825 | /* first iteration */ | |
2826 | pos = tmp; | |
2827 | tconn = pos; | |
2828 | break; | |
2829 | } | |
2830 | if (tmp == pos) { | |
2831 | tconn = pos; | |
2832 | break; | |
2833 | } | |
b411b363 | 2834 | } |
543cc10b | 2835 | if (tconn) { |
71932efc | 2836 | next_tconn: |
543cc10b LE |
2837 | mdev = idr_get_next(&tconn->volumes, &volume); |
2838 | if (!mdev) { | |
2839 | /* No more volumes to dump on this tconn. | |
2840 | * Advance tconn iterator. */ | |
ec0bddbc PR |
2841 | pos = list_entry_rcu(tconn->all_tconn.next, |
2842 | struct drbd_tconn, all_tconn); | |
71932efc | 2843 | /* Did we dump any volume on this tconn yet? */ |
543cc10b | 2844 | if (volume != 0) { |
71932efc LE |
2845 | /* If we reached the end of the list, |
2846 | * or only a single resource dump was requested, | |
2847 | * we are done. */ | |
2848 | if (&pos->all_tconn == &drbd_tconns || cb->args[2]) | |
2849 | goto out; | |
543cc10b | 2850 | volume = 0; |
71932efc | 2851 | tconn = pos; |
543cc10b LE |
2852 | goto next_tconn; |
2853 | } | |
2854 | } | |
2855 | ||
98683650 | 2856 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
2857 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
2858 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
2859 | if (!dh) | |
543cc10b LE |
2860 | goto out; |
2861 | ||
2862 | if (!mdev) { | |
367d675d LE |
2863 | /* This is a tconn without a single volume. |
2864 | * Suprisingly enough, it may have a network | |
2865 | * configuration. */ | |
2866 | struct net_conf *nc; | |
543cc10b LE |
2867 | dh->minor = -1U; |
2868 | dh->ret_code = NO_ERROR; | |
089c075d | 2869 | if (nla_put_drbd_cfg_context(skb, tconn, VOLUME_UNSPECIFIED)) |
367d675d LE |
2870 | goto cancel; |
2871 | nc = rcu_dereference(tconn->net_conf); | |
2872 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
2873 | goto cancel; | |
2874 | goto done; | |
543cc10b | 2875 | } |
b411b363 | 2876 | |
543cc10b LE |
2877 | D_ASSERT(mdev->vnr == volume); |
2878 | D_ASSERT(mdev->tconn == tconn); | |
3b98c0c2 | 2879 | |
543cc10b | 2880 | dh->minor = mdev_to_minor(mdev); |
3b98c0c2 LE |
2881 | dh->ret_code = NO_ERROR; |
2882 | ||
2883 | if (nla_put_status_info(skb, mdev, NULL)) { | |
367d675d | 2884 | cancel: |
3b98c0c2 | 2885 | genlmsg_cancel(skb, dh); |
543cc10b | 2886 | goto out; |
3b98c0c2 | 2887 | } |
367d675d | 2888 | done: |
3b98c0c2 LE |
2889 | genlmsg_end(skb, dh); |
2890 | } | |
b411b363 | 2891 | |
543cc10b | 2892 | out: |
c141ebda | 2893 | rcu_read_unlock(); |
543cc10b LE |
2894 | /* where to start the next iteration */ |
2895 | cb->args[0] = (long)pos; | |
2896 | cb->args[1] = (pos == tconn) ? volume + 1 : 0; | |
b411b363 | 2897 | |
543cc10b LE |
2898 | /* No more tconns/volumes/minors found results in an empty skb. |
2899 | * Which will terminate the dump. */ | |
3b98c0c2 | 2900 | return skb->len; |
b411b363 PR |
2901 | } |
2902 | ||
71932efc LE |
2903 | /* |
2904 | * Request status of all resources, or of all volumes within a single resource. | |
2905 | * | |
2906 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
2907 | * Which means we cannot use the family->attrbuf or other such members, because | |
2908 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
2909 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
2910 | * | |
2911 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 2912 | */ |
71932efc | 2913 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 2914 | { |
71932efc LE |
2915 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
2916 | struct nlattr *nla; | |
7c3063cc | 2917 | const char *resource_name; |
71932efc | 2918 | struct drbd_tconn *tconn; |
7c3063cc | 2919 | int maxtype; |
71932efc LE |
2920 | |
2921 | /* Is this a followup call? */ | |
2922 | if (cb->args[0]) { | |
2923 | /* ... of a single resource dump, | |
2924 | * and the resource iterator has been advanced already? */ | |
2925 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
2926 | return 0; /* DONE. */ | |
2927 | goto dump; | |
2928 | } | |
2929 | ||
2930 | /* First call (from netlink_dump_start). We need to figure out | |
2931 | * which resource(s) the user wants us to dump. */ | |
2932 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
2933 | nlmsg_attrlen(cb->nlh, hdrlen), | |
2934 | DRBD_NLA_CFG_CONTEXT); | |
2935 | ||
2936 | /* No explicit context given. Dump all. */ | |
2937 | if (!nla) | |
2938 | goto dump; | |
7c3063cc AG |
2939 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
2940 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
2941 | if (IS_ERR(nla)) | |
2942 | return PTR_ERR(nla); | |
71932efc LE |
2943 | /* context given, but no name present? */ |
2944 | if (!nla) | |
2945 | return -EINVAL; | |
7c3063cc AG |
2946 | resource_name = nla_data(nla); |
2947 | tconn = conn_get_by_name(resource_name); | |
0ace9dfa | 2948 | |
71932efc LE |
2949 | if (!tconn) |
2950 | return -ENODEV; | |
2951 | ||
0ace9dfa PR |
2952 | kref_put(&tconn->kref, &conn_destroy); /* get_one_status() (re)validates tconn by itself */ |
2953 | ||
71932efc LE |
2954 | /* prime iterators, and set "filter" mode mark: |
2955 | * only dump this tconn. */ | |
2956 | cb->args[0] = (long)tconn; | |
2957 | /* cb->args[1] = 0; passed in this way. */ | |
2958 | cb->args[2] = (long)tconn; | |
2959 | ||
2960 | dump: | |
2961 | return get_one_status(skb, cb); | |
2962 | } | |
b411b363 | 2963 | |
3b98c0c2 | 2964 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2965 | { |
3b98c0c2 LE |
2966 | enum drbd_ret_code retcode; |
2967 | struct timeout_parms tp; | |
2968 | int err; | |
b411b363 | 2969 | |
3b98c0c2 LE |
2970 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2971 | if (!adm_ctx.reply_skb) | |
2972 | return retcode; | |
2973 | if (retcode != NO_ERROR) | |
2974 | goto out; | |
b411b363 | 2975 | |
3b98c0c2 LE |
2976 | tp.timeout_type = |
2977 | adm_ctx.mdev->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : | |
2978 | test_bit(USE_DEGR_WFC_T, &adm_ctx.mdev->flags) ? UT_DEGRADED : | |
2979 | UT_DEFAULT; | |
b411b363 | 2980 | |
3b98c0c2 LE |
2981 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
2982 | if (err) { | |
2983 | nlmsg_free(adm_ctx.reply_skb); | |
2984 | return err; | |
2985 | } | |
2986 | out: | |
2987 | drbd_adm_finish(info, retcode); | |
2988 | return 0; | |
b411b363 PR |
2989 | } |
2990 | ||
3b98c0c2 | 2991 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2992 | { |
3b98c0c2 LE |
2993 | struct drbd_conf *mdev; |
2994 | enum drbd_ret_code retcode; | |
58ffa580 | 2995 | struct start_ov_parms parms; |
b411b363 | 2996 | |
3b98c0c2 LE |
2997 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
2998 | if (!adm_ctx.reply_skb) | |
2999 | return retcode; | |
3000 | if (retcode != NO_ERROR) | |
3001 | goto out; | |
873b0d5f | 3002 | |
3b98c0c2 | 3003 | mdev = adm_ctx.mdev; |
58ffa580 LE |
3004 | |
3005 | /* resume from last known position, if possible */ | |
3006 | parms.ov_start_sector = mdev->ov_start_sector; | |
3007 | parms.ov_stop_sector = ULLONG_MAX; | |
3b98c0c2 | 3008 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3009 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3010 | if (err) { |
3011 | retcode = ERR_MANDATORY_TAG; | |
3012 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3013 | goto out; | |
3014 | } | |
b411b363 | 3015 | } |
58ffa580 LE |
3016 | /* w_make_ov_request expects position to be aligned */ |
3017 | mdev->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); | |
3018 | mdev->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3019 | |
3020 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3021 | * just being finished, wait for it before requesting a new resync. */ | |
a574daf5 | 3022 | drbd_suspend_io(mdev); |
873b0d5f | 3023 | wait_event(mdev->misc_wait, !test_bit(BITMAP_IO, &mdev->flags)); |
3b98c0c2 | 3024 | retcode = drbd_request_state(mdev,NS(conn,C_VERIFY_S)); |
a574daf5 | 3025 | drbd_resume_io(mdev); |
3b98c0c2 LE |
3026 | out: |
3027 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3028 | return 0; |
3029 | } | |
3030 | ||
3031 | ||
3b98c0c2 | 3032 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3033 | { |
3b98c0c2 LE |
3034 | struct drbd_conf *mdev; |
3035 | enum drbd_ret_code retcode; | |
b411b363 PR |
3036 | int skip_initial_sync = 0; |
3037 | int err; | |
3b98c0c2 | 3038 | struct new_c_uuid_parms args; |
b411b363 | 3039 | |
3b98c0c2 LE |
3040 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3041 | if (!adm_ctx.reply_skb) | |
3042 | return retcode; | |
3043 | if (retcode != NO_ERROR) | |
3044 | goto out_nolock; | |
b411b363 | 3045 | |
3b98c0c2 LE |
3046 | mdev = adm_ctx.mdev; |
3047 | memset(&args, 0, sizeof(args)); | |
3048 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3049 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3050 | if (err) { |
3051 | retcode = ERR_MANDATORY_TAG; | |
3052 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3053 | goto out_nolock; | |
3054 | } | |
b411b363 PR |
3055 | } |
3056 | ||
8410da8f | 3057 | mutex_lock(mdev->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 PR |
3058 | |
3059 | if (!get_ldev(mdev)) { | |
3060 | retcode = ERR_NO_DISK; | |
3061 | goto out; | |
3062 | } | |
3063 | ||
3064 | /* this is "skip initial sync", assume to be clean */ | |
31890f4a | 3065 | if (mdev->state.conn == C_CONNECTED && mdev->tconn->agreed_pro_version >= 90 && |
b411b363 PR |
3066 | mdev->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
3067 | dev_info(DEV, "Preparing to skip initial sync\n"); | |
3068 | skip_initial_sync = 1; | |
3069 | } else if (mdev->state.conn != C_STANDALONE) { | |
3070 | retcode = ERR_CONNECTED; | |
3071 | goto out_dec; | |
3072 | } | |
3073 | ||
3074 | drbd_uuid_set(mdev, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ | |
3075 | drbd_uuid_new_current(mdev); /* New current, previous to UI_BITMAP */ | |
3076 | ||
3077 | if (args.clear_bm) { | |
20ceb2b2 LE |
3078 | err = drbd_bitmap_io(mdev, &drbd_bmio_clear_n_write, |
3079 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); | |
b411b363 PR |
3080 | if (err) { |
3081 | dev_err(DEV, "Writing bitmap failed with %d\n",err); | |
3082 | retcode = ERR_IO_MD_DISK; | |
3083 | } | |
3084 | if (skip_initial_sync) { | |
3085 | drbd_send_uuids_skip_initial_sync(mdev); | |
3086 | _drbd_uuid_set(mdev, UI_BITMAP, 0); | |
62b0da3a | 3087 | drbd_print_uuids(mdev, "cleared bitmap UUID"); |
87eeee41 | 3088 | spin_lock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
3089 | _drbd_set_state(_NS2(mdev, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
3090 | CS_VERBOSE, NULL); | |
87eeee41 | 3091 | spin_unlock_irq(&mdev->tconn->req_lock); |
b411b363 PR |
3092 | } |
3093 | } | |
3094 | ||
3095 | drbd_md_sync(mdev); | |
3096 | out_dec: | |
3097 | put_ldev(mdev); | |
3098 | out: | |
8410da8f | 3099 | mutex_unlock(mdev->state_mutex); |
3b98c0c2 LE |
3100 | out_nolock: |
3101 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3102 | return 0; |
3103 | } | |
3104 | ||
3b98c0c2 | 3105 | static enum drbd_ret_code |
7c3063cc | 3106 | drbd_check_resource_name(const char *name) |
b411b363 | 3107 | { |
3b98c0c2 | 3108 | if (!name || !name[0]) { |
7c3063cc | 3109 | drbd_msg_put_info("resource name missing"); |
3b98c0c2 | 3110 | return ERR_MANDATORY_TAG; |
b411b363 | 3111 | } |
3b98c0c2 LE |
3112 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3113 | * we must not allow slashes */ | |
3114 | if (strchr(name, '/')) { | |
7c3063cc | 3115 | drbd_msg_put_info("invalid resource name"); |
3b98c0c2 | 3116 | return ERR_INVALID_REQUEST; |
b411b363 | 3117 | } |
3b98c0c2 | 3118 | return NO_ERROR; |
774b3055 | 3119 | } |
b411b363 | 3120 | |
789c1b62 | 3121 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3122 | { |
3b98c0c2 | 3123 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3124 | struct res_opts res_opts; |
3125 | int err; | |
b411b363 | 3126 | |
3b98c0c2 LE |
3127 | retcode = drbd_adm_prepare(skb, info, 0); |
3128 | if (!adm_ctx.reply_skb) | |
3129 | return retcode; | |
3130 | if (retcode != NO_ERROR) | |
3131 | goto out; | |
b411b363 | 3132 | |
afbbfa88 AG |
3133 | set_res_opts_defaults(&res_opts); |
3134 | err = res_opts_from_attrs(&res_opts, info); | |
3135 | if (err && err != -ENOMSG) { | |
3136 | retcode = ERR_MANDATORY_TAG; | |
3137 | drbd_msg_put_info(from_attrs_err_to_txt(err)); | |
3138 | goto out; | |
b411b363 PR |
3139 | } |
3140 | ||
7c3063cc | 3141 | retcode = drbd_check_resource_name(adm_ctx.resource_name); |
3b98c0c2 LE |
3142 | if (retcode != NO_ERROR) |
3143 | goto out; | |
b411b363 | 3144 | |
3b98c0c2 | 3145 | if (adm_ctx.tconn) { |
38f19616 LE |
3146 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3147 | retcode = ERR_INVALID_REQUEST; | |
789c1b62 | 3148 | drbd_msg_put_info("resource exists"); |
38f19616 LE |
3149 | } |
3150 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3151 | goto out; |
b411b363 | 3152 | } |
b411b363 | 3153 | |
afbbfa88 | 3154 | if (!conn_create(adm_ctx.resource_name, &res_opts)) |
b411b363 | 3155 | retcode = ERR_NOMEM; |
3b98c0c2 LE |
3156 | out: |
3157 | drbd_adm_finish(info, retcode); | |
3158 | return 0; | |
b411b363 PR |
3159 | } |
3160 | ||
3b98c0c2 | 3161 | int drbd_adm_add_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3162 | { |
3b98c0c2 LE |
3163 | struct drbd_genlmsghdr *dh = info->userhdr; |
3164 | enum drbd_ret_code retcode; | |
b411b363 | 3165 | |
44e52cfa | 3166 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3167 | if (!adm_ctx.reply_skb) |
3168 | return retcode; | |
3169 | if (retcode != NO_ERROR) | |
3170 | goto out; | |
b411b363 | 3171 | |
f2257a56 | 3172 | if (dh->minor > MINORMASK) { |
3b98c0c2 LE |
3173 | drbd_msg_put_info("requested minor out of range"); |
3174 | retcode = ERR_INVALID_REQUEST; | |
3175 | goto out; | |
b411b363 | 3176 | } |
0c8e36d9 | 3177 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
3b98c0c2 LE |
3178 | drbd_msg_put_info("requested volume id out of range"); |
3179 | retcode = ERR_INVALID_REQUEST; | |
3180 | goto out; | |
b411b363 | 3181 | } |
b411b363 | 3182 | |
38f19616 LE |
3183 | /* drbd_adm_prepare made sure already |
3184 | * that mdev->tconn and mdev->vnr match the request. */ | |
3185 | if (adm_ctx.mdev) { | |
3186 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) | |
3187 | retcode = ERR_MINOR_EXISTS; | |
3188 | /* else: still NO_ERROR */ | |
3189 | goto out; | |
b411b363 | 3190 | } |
38f19616 | 3191 | |
3b98c0c2 LE |
3192 | retcode = conn_new_minor(adm_ctx.tconn, dh->minor, adm_ctx.volume); |
3193 | out: | |
3194 | drbd_adm_finish(info, retcode); | |
3195 | return 0; | |
b411b363 PR |
3196 | } |
3197 | ||
85f75dd7 | 3198 | static enum drbd_ret_code adm_delete_minor(struct drbd_conf *mdev) |
b411b363 | 3199 | { |
85f75dd7 LE |
3200 | if (mdev->state.disk == D_DISKLESS && |
3201 | /* no need to be mdev->state.conn == C_STANDALONE && | |
3202 | * we may want to delete a minor from a live replication group. | |
3203 | */ | |
3204 | mdev->state.role == R_SECONDARY) { | |
369bea63 PR |
3205 | _drbd_request_state(mdev, NS(conn, C_WF_REPORT_PARAMS), |
3206 | CS_VERBOSE + CS_WAIT_COMPLETE); | |
81fa2e67 PR |
3207 | idr_remove(&mdev->tconn->volumes, mdev->vnr); |
3208 | idr_remove(&minors, mdev_to_minor(mdev)); | |
113fef9e | 3209 | destroy_workqueue(mdev->submit.wq); |
81fa2e67 PR |
3210 | del_gendisk(mdev->vdisk); |
3211 | synchronize_rcu(); | |
3212 | kref_put(&mdev->kref, &drbd_minor_destroy); | |
85f75dd7 LE |
3213 | return NO_ERROR; |
3214 | } else | |
3215 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
3216 | } |
3217 | ||
3b98c0c2 | 3218 | int drbd_adm_delete_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3219 | { |
3b98c0c2 | 3220 | enum drbd_ret_code retcode; |
b411b363 | 3221 | |
3b98c0c2 LE |
3222 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_MINOR); |
3223 | if (!adm_ctx.reply_skb) | |
3224 | return retcode; | |
3225 | if (retcode != NO_ERROR) | |
3226 | goto out; | |
b411b363 | 3227 | |
85f75dd7 | 3228 | retcode = adm_delete_minor(adm_ctx.mdev); |
85f75dd7 LE |
3229 | out: |
3230 | drbd_adm_finish(info, retcode); | |
3231 | return 0; | |
b411b363 PR |
3232 | } |
3233 | ||
85f75dd7 | 3234 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3235 | { |
f3dfa40a | 3236 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 LE |
3237 | struct drbd_conf *mdev; |
3238 | unsigned i; | |
b411b363 | 3239 | |
85f75dd7 LE |
3240 | retcode = drbd_adm_prepare(skb, info, 0); |
3241 | if (!adm_ctx.reply_skb) | |
3242 | return retcode; | |
3243 | if (retcode != NO_ERROR) | |
3244 | goto out; | |
b411b363 | 3245 | |
85f75dd7 | 3246 | if (!adm_ctx.tconn) { |
789c1b62 | 3247 | retcode = ERR_RES_NOT_KNOWN; |
85f75dd7 | 3248 | goto out; |
b411b363 PR |
3249 | } |
3250 | ||
85f75dd7 LE |
3251 | /* demote */ |
3252 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
3253 | retcode = drbd_set_role(mdev, R_SECONDARY, 0); | |
3254 | if (retcode < SS_SUCCESS) { | |
3255 | drbd_msg_put_info("failed to demote"); | |
c141ebda | 3256 | goto out; |
85f75dd7 | 3257 | } |
b411b363 | 3258 | } |
b411b363 | 3259 | |
f3dfa40a LE |
3260 | retcode = conn_try_disconnect(adm_ctx.tconn, 0); |
3261 | if (retcode < SS_SUCCESS) { | |
85f75dd7 | 3262 | drbd_msg_put_info("failed to disconnect"); |
f3dfa40a | 3263 | goto out; |
85f75dd7 | 3264 | } |
b411b363 | 3265 | |
85f75dd7 LE |
3266 | /* detach */ |
3267 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
cdfda633 | 3268 | retcode = adm_detach(mdev, 0); |
27012382 | 3269 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
85f75dd7 | 3270 | drbd_msg_put_info("failed to detach"); |
c141ebda | 3271 | goto out; |
85f75dd7 LE |
3272 | } |
3273 | } | |
b411b363 | 3274 | |
f3dfa40a LE |
3275 | /* If we reach this, all volumes (of this tconn) are Secondary, |
3276 | * Disconnected, Diskless, aka Unconfigured. Make sure all threads have | |
c141ebda | 3277 | * actually stopped, state handling only does drbd_thread_stop_nowait(). */ |
f3dfa40a | 3278 | drbd_thread_stop(&adm_ctx.tconn->worker); |
b411b363 | 3279 | |
f3dfa40a | 3280 | /* Now, nothing can fail anymore */ |
b411b363 | 3281 | |
85f75dd7 LE |
3282 | /* delete volumes */ |
3283 | idr_for_each_entry(&adm_ctx.tconn->volumes, mdev, i) { | |
3284 | retcode = adm_delete_minor(mdev); | |
3285 | if (retcode != NO_ERROR) { | |
3286 | /* "can not happen" */ | |
3287 | drbd_msg_put_info("failed to delete volume"); | |
ef356262 | 3288 | goto out; |
85f75dd7 LE |
3289 | } |
3290 | } | |
b411b363 | 3291 | |
85f75dd7 LE |
3292 | /* delete connection */ |
3293 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { | |
ec0bddbc PR |
3294 | list_del_rcu(&adm_ctx.tconn->all_tconn); |
3295 | synchronize_rcu(); | |
9dc9fbb3 | 3296 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); |
b411b363 | 3297 | |
85f75dd7 LE |
3298 | retcode = NO_ERROR; |
3299 | } else { | |
3300 | /* "can not happen" */ | |
789c1b62 | 3301 | retcode = ERR_RES_IN_USE; |
85f75dd7 | 3302 | drbd_msg_put_info("failed to delete connection"); |
85f75dd7 | 3303 | } |
ef356262 | 3304 | goto out; |
3b98c0c2 LE |
3305 | out: |
3306 | drbd_adm_finish(info, retcode); | |
3307 | return 0; | |
b411b363 PR |
3308 | } |
3309 | ||
789c1b62 | 3310 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3311 | { |
3b98c0c2 | 3312 | enum drbd_ret_code retcode; |
b411b363 | 3313 | |
44e52cfa | 3314 | retcode = drbd_adm_prepare(skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
3315 | if (!adm_ctx.reply_skb) |
3316 | return retcode; | |
3317 | if (retcode != NO_ERROR) | |
3318 | goto out; | |
b411b363 | 3319 | |
3b98c0c2 | 3320 | if (conn_lowest_minor(adm_ctx.tconn) < 0) { |
ec0bddbc PR |
3321 | list_del_rcu(&adm_ctx.tconn->all_tconn); |
3322 | synchronize_rcu(); | |
9dc9fbb3 PR |
3323 | kref_put(&adm_ctx.tconn->kref, &conn_destroy); |
3324 | ||
3b98c0c2 LE |
3325 | retcode = NO_ERROR; |
3326 | } else { | |
789c1b62 | 3327 | retcode = ERR_RES_IN_USE; |
b411b363 PR |
3328 | } |
3329 | ||
992d6e91 LE |
3330 | if (retcode == NO_ERROR) |
3331 | drbd_thread_stop(&adm_ctx.tconn->worker); | |
3b98c0c2 LE |
3332 | out: |
3333 | drbd_adm_finish(info, retcode); | |
b411b363 PR |
3334 | return 0; |
3335 | } | |
3336 | ||
3b98c0c2 | 3337 | void drbd_bcast_event(struct drbd_conf *mdev, const struct sib_info *sib) |
b411b363 | 3338 | { |
3b98c0c2 LE |
3339 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
3340 | struct sk_buff *msg; | |
3341 | struct drbd_genlmsghdr *d_out; | |
3342 | unsigned seq; | |
3343 | int err = -ENOMEM; | |
3344 | ||
ef86b779 PR |
3345 | if (sib->sib_reason == SIB_SYNC_PROGRESS) { |
3346 | if (time_after(jiffies, mdev->rs_last_bcast + HZ)) | |
3347 | mdev->rs_last_bcast = jiffies; | |
3348 | else | |
3349 | return; | |
3350 | } | |
b411b363 | 3351 | |
3b98c0c2 LE |
3352 | seq = atomic_inc_return(&drbd_genl_seq); |
3353 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
3354 | if (!msg) | |
3355 | goto failed; | |
3356 | ||
3357 | err = -EMSGSIZE; | |
3358 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
3359 | if (!d_out) /* cannot happen, but anyways. */ | |
3360 | goto nla_put_failure; | |
3361 | d_out->minor = mdev_to_minor(mdev); | |
6f9b5f84 | 3362 | d_out->ret_code = NO_ERROR; |
3b98c0c2 LE |
3363 | |
3364 | if (nla_put_status_info(msg, mdev, sib)) | |
3365 | goto nla_put_failure; | |
3366 | genlmsg_end(msg, d_out); | |
3367 | err = drbd_genl_multicast_events(msg, 0); | |
3368 | /* msg has been consumed or freed in netlink_broadcast() */ | |
3369 | if (err && err != -ESRCH) | |
3370 | goto failed; | |
b411b363 | 3371 | |
3b98c0c2 | 3372 | return; |
b411b363 | 3373 | |
3b98c0c2 LE |
3374 | nla_put_failure: |
3375 | nlmsg_free(msg); | |
3376 | failed: | |
3377 | dev_err(DEV, "Error %d while broadcasting event. " | |
3378 | "Event seq:%u sib_reason:%u\n", | |
3379 | err, seq, sib->sib_reason); | |
b411b363 | 3380 | } |