]>
Commit | Line | Data |
---|---|---|
b411b363 PR |
1 | /* |
2 | drbd_nl.c | |
3 | ||
4 | This file is part of DRBD by Philipp Reisner and Lars Ellenberg. | |
5 | ||
6 | Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. | |
7 | Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. | |
8 | Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. | |
9 | ||
10 | drbd is free software; you can redistribute it and/or modify | |
11 | it under the terms of the GNU General Public License as published by | |
12 | the Free Software Foundation; either version 2, or (at your option) | |
13 | any later version. | |
14 | ||
15 | drbd is distributed in the hope that it will be useful, | |
16 | but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
18 | GNU General Public License for more details. | |
19 | ||
20 | You should have received a copy of the GNU General Public License | |
21 | along with drbd; see the file COPYING. If not, write to | |
22 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. | |
23 | ||
24 | */ | |
25 | ||
f88c5d90 LE |
26 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
27 | ||
b411b363 PR |
28 | #include <linux/module.h> |
29 | #include <linux/drbd.h> | |
30 | #include <linux/in.h> | |
31 | #include <linux/fs.h> | |
32 | #include <linux/file.h> | |
33 | #include <linux/slab.h> | |
b411b363 PR |
34 | #include <linux/blkpg.h> |
35 | #include <linux/cpumask.h> | |
36 | #include "drbd_int.h" | |
a3603a6e | 37 | #include "drbd_protocol.h" |
265be2d0 | 38 | #include "drbd_req.h" |
a2972846 | 39 | #include "drbd_state_change.h" |
b411b363 | 40 | #include <asm/unaligned.h> |
b411b363 | 41 | #include <linux/drbd_limits.h> |
87f7be4c | 42 | #include <linux/kthread.h> |
b411b363 | 43 | |
3b98c0c2 LE |
44 | #include <net/genetlink.h> |
45 | ||
46 | /* .doit */ | |
47 | // int drbd_adm_create_resource(struct sk_buff *skb, struct genl_info *info); | |
48 | // int drbd_adm_delete_resource(struct sk_buff *skb, struct genl_info *info); | |
49 | ||
05a10ec7 AG |
50 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info); |
51 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info); | |
3b98c0c2 | 52 | |
789c1b62 AG |
53 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info); |
54 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info); | |
85f75dd7 | 55 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
56 | |
57 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info); | |
58 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 59 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
60 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info); |
61 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 62 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
63 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info); |
64 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info); | |
65 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info); | |
66 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info); | |
67 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info); | |
68 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info); | |
69 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info); | |
70 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info); | |
71 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info); | |
72 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info); | |
73 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info); | |
f399002e | 74 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info); |
3b98c0c2 LE |
75 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info); |
76 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info); | |
77 | /* .dumpit */ | |
78 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb); | |
a55bbd37 AG |
79 | int drbd_adm_dump_resources(struct sk_buff *skb, struct netlink_callback *cb); |
80 | int drbd_adm_dump_devices(struct sk_buff *skb, struct netlink_callback *cb); | |
81 | int drbd_adm_dump_devices_done(struct netlink_callback *cb); | |
82 | int drbd_adm_dump_connections(struct sk_buff *skb, struct netlink_callback *cb); | |
83 | int drbd_adm_dump_connections_done(struct netlink_callback *cb); | |
84 | int drbd_adm_dump_peer_devices(struct sk_buff *skb, struct netlink_callback *cb); | |
85 | int drbd_adm_dump_peer_devices_done(struct netlink_callback *cb); | |
a2972846 | 86 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb); |
3b98c0c2 LE |
87 | |
88 | #include <linux/drbd_genl_api.h> | |
01b39b50 | 89 | #include "drbd_nla.h" |
3b98c0c2 LE |
90 | #include <linux/genl_magic_func.h> |
91 | ||
a2972846 AG |
92 | static atomic_t drbd_genl_seq = ATOMIC_INIT(2); /* two. */ |
93 | static atomic_t notify_genl_seq = ATOMIC_INIT(2); /* two. */ | |
94 | ||
95 | DEFINE_MUTEX(notification_mutex); | |
96 | ||
3b98c0c2 | 97 | /* used blkdev_get_by_path, to claim our meta data device(s) */ |
b411b363 PR |
98 | static char *drbd_m_holder = "Hands off! this is DRBD's meta data device."; |
99 | ||
3b98c0c2 LE |
100 | static void drbd_adm_send_reply(struct sk_buff *skb, struct genl_info *info) |
101 | { | |
102 | genlmsg_end(skb, genlmsg_data(nlmsg_data(nlmsg_hdr(skb)))); | |
103 | if (genlmsg_reply(skb, info)) | |
f88c5d90 | 104 | pr_err("error sending genl reply\n"); |
b411b363 | 105 | } |
3b98c0c2 LE |
106 | |
107 | /* Used on a fresh "drbd_adm_prepare"d reply_skb, this cannot fail: The only | |
108 | * reason it could fail was no space in skb, and there are 4k available. */ | |
f221f4bc | 109 | static int drbd_msg_put_info(struct sk_buff *skb, const char *info) |
3b98c0c2 | 110 | { |
3b98c0c2 LE |
111 | struct nlattr *nla; |
112 | int err = -EMSGSIZE; | |
113 | ||
114 | if (!info || !info[0]) | |
115 | return 0; | |
116 | ||
117 | nla = nla_nest_start(skb, DRBD_NLA_CFG_REPLY); | |
118 | if (!nla) | |
119 | return err; | |
120 | ||
121 | err = nla_put_string(skb, T_info_text, info); | |
122 | if (err) { | |
123 | nla_nest_cancel(skb, nla); | |
124 | return err; | |
125 | } else | |
126 | nla_nest_end(skb, nla); | |
127 | return 0; | |
b411b363 PR |
128 | } |
129 | ||
3b98c0c2 LE |
130 | /* This would be a good candidate for a "pre_doit" hook, |
131 | * and per-family private info->pointers. | |
132 | * But we need to stay compatible with older kernels. | |
133 | * If it returns successfully, adm_ctx members are valid. | |
9e276872 LE |
134 | * |
135 | * At this point, we still rely on the global genl_lock(). | |
136 | * If we want to avoid that, and allow "genl_family.parallel_ops", we may need | |
137 | * to add additional synchronization against object destruction/modification. | |
3b98c0c2 LE |
138 | */ |
139 | #define DRBD_ADM_NEED_MINOR 1 | |
44e52cfa | 140 | #define DRBD_ADM_NEED_RESOURCE 2 |
089c075d | 141 | #define DRBD_ADM_NEED_CONNECTION 4 |
a910b123 LE |
142 | static int drbd_adm_prepare(struct drbd_config_context *adm_ctx, |
143 | struct sk_buff *skb, struct genl_info *info, unsigned flags) | |
3b98c0c2 LE |
144 | { |
145 | struct drbd_genlmsghdr *d_in = info->userhdr; | |
146 | const u8 cmd = info->genlhdr->cmd; | |
147 | int err; | |
148 | ||
a910b123 | 149 | memset(adm_ctx, 0, sizeof(*adm_ctx)); |
3b98c0c2 LE |
150 | |
151 | /* genl_rcv_msg only checks for CAP_NET_ADMIN on "GENL_ADMIN_PERM" :( */ | |
98683650 | 152 | if (cmd != DRBD_ADM_GET_STATUS && !capable(CAP_NET_ADMIN)) |
3b98c0c2 LE |
153 | return -EPERM; |
154 | ||
a910b123 LE |
155 | adm_ctx->reply_skb = genlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL); |
156 | if (!adm_ctx->reply_skb) { | |
1e2a2551 | 157 | err = -ENOMEM; |
3b98c0c2 | 158 | goto fail; |
1e2a2551 | 159 | } |
3b98c0c2 | 160 | |
a910b123 | 161 | adm_ctx->reply_dh = genlmsg_put_reply(adm_ctx->reply_skb, |
3b98c0c2 LE |
162 | info, &drbd_genl_family, 0, cmd); |
163 | /* put of a few bytes into a fresh skb of >= 4k will always succeed. | |
164 | * but anyways */ | |
a910b123 | 165 | if (!adm_ctx->reply_dh) { |
1e2a2551 | 166 | err = -ENOMEM; |
3b98c0c2 | 167 | goto fail; |
1e2a2551 | 168 | } |
3b98c0c2 | 169 | |
a910b123 LE |
170 | adm_ctx->reply_dh->minor = d_in->minor; |
171 | adm_ctx->reply_dh->ret_code = NO_ERROR; | |
3b98c0c2 | 172 | |
a910b123 | 173 | adm_ctx->volume = VOLUME_UNSPECIFIED; |
3b98c0c2 LE |
174 | if (info->attrs[DRBD_NLA_CFG_CONTEXT]) { |
175 | struct nlattr *nla; | |
176 | /* parse and validate only */ | |
f399002e | 177 | err = drbd_cfg_context_from_attrs(NULL, info); |
3b98c0c2 LE |
178 | if (err) |
179 | goto fail; | |
180 | ||
181 | /* It was present, and valid, | |
182 | * copy it over to the reply skb. */ | |
a910b123 | 183 | err = nla_put_nohdr(adm_ctx->reply_skb, |
3b98c0c2 LE |
184 | info->attrs[DRBD_NLA_CFG_CONTEXT]->nla_len, |
185 | info->attrs[DRBD_NLA_CFG_CONTEXT]); | |
186 | if (err) | |
187 | goto fail; | |
188 | ||
9e276872 | 189 | /* and assign stuff to the adm_ctx */ |
3b98c0c2 | 190 | nla = nested_attr_tb[__nla_type(T_ctx_volume)]; |
089c075d | 191 | if (nla) |
a910b123 | 192 | adm_ctx->volume = nla_get_u32(nla); |
7c3063cc | 193 | nla = nested_attr_tb[__nla_type(T_ctx_resource_name)]; |
3b98c0c2 | 194 | if (nla) |
a910b123 LE |
195 | adm_ctx->resource_name = nla_data(nla); |
196 | adm_ctx->my_addr = nested_attr_tb[__nla_type(T_ctx_my_addr)]; | |
197 | adm_ctx->peer_addr = nested_attr_tb[__nla_type(T_ctx_peer_addr)]; | |
198 | if ((adm_ctx->my_addr && | |
199 | nla_len(adm_ctx->my_addr) > sizeof(adm_ctx->connection->my_addr)) || | |
200 | (adm_ctx->peer_addr && | |
201 | nla_len(adm_ctx->peer_addr) > sizeof(adm_ctx->connection->peer_addr))) { | |
089c075d AG |
202 | err = -EINVAL; |
203 | goto fail; | |
204 | } | |
205 | } | |
3b98c0c2 | 206 | |
a910b123 LE |
207 | adm_ctx->minor = d_in->minor; |
208 | adm_ctx->device = minor_to_device(d_in->minor); | |
9e276872 LE |
209 | |
210 | /* We are protected by the global genl_lock(). | |
211 | * But we may explicitly drop it/retake it in drbd_adm_set_role(), | |
212 | * so make sure this object stays around. */ | |
213 | if (adm_ctx->device) | |
214 | kref_get(&adm_ctx->device->kref); | |
215 | ||
a910b123 LE |
216 | if (adm_ctx->resource_name) { |
217 | adm_ctx->resource = drbd_find_resource(adm_ctx->resource_name); | |
4bc76048 | 218 | } |
3b98c0c2 | 219 | |
a910b123 LE |
220 | if (!adm_ctx->device && (flags & DRBD_ADM_NEED_MINOR)) { |
221 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown minor"); | |
3b98c0c2 LE |
222 | return ERR_MINOR_INVALID; |
223 | } | |
a910b123 LE |
224 | if (!adm_ctx->resource && (flags & DRBD_ADM_NEED_RESOURCE)) { |
225 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown resource"); | |
226 | if (adm_ctx->resource_name) | |
a10f6b8a | 227 | return ERR_RES_NOT_KNOWN; |
3b98c0c2 LE |
228 | return ERR_INVALID_REQUEST; |
229 | } | |
230 | ||
089c075d | 231 | if (flags & DRBD_ADM_NEED_CONNECTION) { |
a910b123 LE |
232 | if (adm_ctx->resource) { |
233 | drbd_msg_put_info(adm_ctx->reply_skb, "no resource name expected"); | |
089c075d AG |
234 | return ERR_INVALID_REQUEST; |
235 | } | |
a910b123 LE |
236 | if (adm_ctx->device) { |
237 | drbd_msg_put_info(adm_ctx->reply_skb, "no minor number expected"); | |
089c075d AG |
238 | return ERR_INVALID_REQUEST; |
239 | } | |
a910b123 LE |
240 | if (adm_ctx->my_addr && adm_ctx->peer_addr) |
241 | adm_ctx->connection = conn_get_by_addrs(nla_data(adm_ctx->my_addr), | |
242 | nla_len(adm_ctx->my_addr), | |
243 | nla_data(adm_ctx->peer_addr), | |
244 | nla_len(adm_ctx->peer_addr)); | |
245 | if (!adm_ctx->connection) { | |
246 | drbd_msg_put_info(adm_ctx->reply_skb, "unknown connection"); | |
089c075d AG |
247 | return ERR_INVALID_REQUEST; |
248 | } | |
249 | } | |
250 | ||
3b98c0c2 | 251 | /* some more paranoia, if the request was over-determined */ |
a910b123 LE |
252 | if (adm_ctx->device && adm_ctx->resource && |
253 | adm_ctx->device->resource != adm_ctx->resource) { | |
4bc76048 | 254 | pr_warning("request: minor=%u, resource=%s; but that minor belongs to resource %s\n", |
a910b123 LE |
255 | adm_ctx->minor, adm_ctx->resource->name, |
256 | adm_ctx->device->resource->name); | |
257 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists in different resource"); | |
527f4b24 LE |
258 | return ERR_INVALID_REQUEST; |
259 | } | |
a910b123 LE |
260 | if (adm_ctx->device && |
261 | adm_ctx->volume != VOLUME_UNSPECIFIED && | |
262 | adm_ctx->volume != adm_ctx->device->vnr) { | |
3b98c0c2 | 263 | pr_warning("request: minor=%u, volume=%u; but that minor is volume %u in %s\n", |
a910b123 LE |
264 | adm_ctx->minor, adm_ctx->volume, |
265 | adm_ctx->device->vnr, | |
266 | adm_ctx->device->resource->name); | |
267 | drbd_msg_put_info(adm_ctx->reply_skb, "minor exists as different volume"); | |
3b98c0c2 LE |
268 | return ERR_INVALID_REQUEST; |
269 | } | |
0ace9dfa | 270 | |
9e276872 LE |
271 | /* still, provide adm_ctx->resource always, if possible. */ |
272 | if (!adm_ctx->resource) { | |
273 | adm_ctx->resource = adm_ctx->device ? adm_ctx->device->resource | |
274 | : adm_ctx->connection ? adm_ctx->connection->resource : NULL; | |
275 | if (adm_ctx->resource) | |
276 | kref_get(&adm_ctx->resource->kref); | |
277 | } | |
278 | ||
3b98c0c2 LE |
279 | return NO_ERROR; |
280 | ||
281 | fail: | |
a910b123 LE |
282 | nlmsg_free(adm_ctx->reply_skb); |
283 | adm_ctx->reply_skb = NULL; | |
1e2a2551 | 284 | return err; |
3b98c0c2 LE |
285 | } |
286 | ||
a910b123 LE |
287 | static int drbd_adm_finish(struct drbd_config_context *adm_ctx, |
288 | struct genl_info *info, int retcode) | |
3b98c0c2 | 289 | { |
9e276872 LE |
290 | if (adm_ctx->device) { |
291 | kref_put(&adm_ctx->device->kref, drbd_destroy_device); | |
292 | adm_ctx->device = NULL; | |
293 | } | |
a910b123 LE |
294 | if (adm_ctx->connection) { |
295 | kref_put(&adm_ctx->connection->kref, &drbd_destroy_connection); | |
296 | adm_ctx->connection = NULL; | |
0ace9dfa | 297 | } |
a910b123 LE |
298 | if (adm_ctx->resource) { |
299 | kref_put(&adm_ctx->resource->kref, drbd_destroy_resource); | |
300 | adm_ctx->resource = NULL; | |
4bc76048 | 301 | } |
0ace9dfa | 302 | |
a910b123 | 303 | if (!adm_ctx->reply_skb) |
3b98c0c2 LE |
304 | return -ENOMEM; |
305 | ||
a910b123 LE |
306 | adm_ctx->reply_dh->ret_code = retcode; |
307 | drbd_adm_send_reply(adm_ctx->reply_skb, info); | |
3b98c0c2 LE |
308 | return 0; |
309 | } | |
b411b363 | 310 | |
bde89a9e | 311 | static void setup_khelper_env(struct drbd_connection *connection, char **envp) |
b411b363 | 312 | { |
6b75dced | 313 | char *afs; |
b411b363 | 314 | |
089c075d | 315 | /* FIXME: A future version will not allow this case. */ |
bde89a9e | 316 | if (connection->my_addr_len == 0 || connection->peer_addr_len == 0) |
089c075d AG |
317 | return; |
318 | ||
bde89a9e | 319 | switch (((struct sockaddr *)&connection->peer_addr)->sa_family) { |
089c075d AG |
320 | case AF_INET6: |
321 | afs = "ipv6"; | |
322 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI6", | |
bde89a9e | 323 | &((struct sockaddr_in6 *)&connection->peer_addr)->sin6_addr); |
b411b363 | 324 | break; |
089c075d AG |
325 | case AF_INET: |
326 | afs = "ipv4"; | |
327 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 328 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 329 | break; |
089c075d AG |
330 | default: |
331 | afs = "ssocks"; | |
332 | snprintf(envp[4], 60, "DRBD_PEER_ADDRESS=%pI4", | |
bde89a9e | 333 | &((struct sockaddr_in *)&connection->peer_addr)->sin_addr); |
b411b363 | 334 | } |
089c075d | 335 | snprintf(envp[3], 20, "DRBD_PEER_AF=%s", afs); |
6b75dced | 336 | } |
b411b363 | 337 | |
b30ab791 | 338 | int drbd_khelper(struct drbd_device *device, char *cmd) |
b411b363 PR |
339 | { |
340 | char *envp[] = { "HOME=/", | |
341 | "TERM=linux", | |
342 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
6b75dced PR |
343 | (char[20]) { }, /* address family */ |
344 | (char[60]) { }, /* address */ | |
b411b363 | 345 | NULL }; |
6b75dced | 346 | char mb[12]; |
b411b363 | 347 | char *argv[] = {usermode_helper, cmd, mb, NULL }; |
a6b32bc3 | 348 | struct drbd_connection *connection = first_peer_device(device)->connection; |
6b75dced | 349 | struct sib_info sib; |
b411b363 PR |
350 | int ret; |
351 | ||
bde89a9e AG |
352 | if (current == connection->worker.task) |
353 | set_bit(CALLBACK_PENDING, &connection->flags); | |
c2ba686f | 354 | |
b30ab791 | 355 | snprintf(mb, 12, "minor-%d", device_to_minor(device)); |
bde89a9e | 356 | setup_khelper_env(connection, envp); |
b411b363 | 357 | |
1090c056 LE |
358 | /* The helper may take some time. |
359 | * write out any unsynced meta data changes now */ | |
b30ab791 | 360 | drbd_md_sync(device); |
1090c056 | 361 | |
d0180171 | 362 | drbd_info(device, "helper command: %s %s %s\n", usermode_helper, cmd, mb); |
3b98c0c2 LE |
363 | sib.sib_reason = SIB_HELPER_PRE; |
364 | sib.helper_name = cmd; | |
b30ab791 | 365 | drbd_bcast_event(device, &sib); |
a2972846 | 366 | notify_helper(NOTIFY_CALL, device, connection, cmd, 0); |
70834d30 | 367 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
b411b363 | 368 | if (ret) |
d0180171 | 369 | drbd_warn(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
370 | usermode_helper, cmd, mb, |
371 | (ret >> 8) & 0xff, ret); | |
372 | else | |
d0180171 | 373 | drbd_info(device, "helper command: %s %s %s exit code %u (0x%x)\n", |
b411b363 PR |
374 | usermode_helper, cmd, mb, |
375 | (ret >> 8) & 0xff, ret); | |
3b98c0c2 LE |
376 | sib.sib_reason = SIB_HELPER_POST; |
377 | sib.helper_exit_code = ret; | |
b30ab791 | 378 | drbd_bcast_event(device, &sib); |
a2972846 | 379 | notify_helper(NOTIFY_RESPONSE, device, connection, cmd, ret); |
b411b363 | 380 | |
bde89a9e AG |
381 | if (current == connection->worker.task) |
382 | clear_bit(CALLBACK_PENDING, &connection->flags); | |
b411b363 PR |
383 | |
384 | if (ret < 0) /* Ignore any ERRNOs we got. */ | |
385 | ret = 0; | |
386 | ||
387 | return ret; | |
388 | } | |
389 | ||
bde89a9e | 390 | static int conn_khelper(struct drbd_connection *connection, char *cmd) |
6b75dced PR |
391 | { |
392 | char *envp[] = { "HOME=/", | |
393 | "TERM=linux", | |
394 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
395 | (char[20]) { }, /* address family */ | |
396 | (char[60]) { }, /* address */ | |
397 | NULL }; | |
77c556f6 AG |
398 | char *resource_name = connection->resource->name; |
399 | char *argv[] = {usermode_helper, cmd, resource_name, NULL }; | |
6b75dced PR |
400 | int ret; |
401 | ||
bde89a9e AG |
402 | setup_khelper_env(connection, envp); |
403 | conn_md_sync(connection); | |
6b75dced | 404 | |
1ec861eb | 405 | drbd_info(connection, "helper command: %s %s %s\n", usermode_helper, cmd, resource_name); |
6b75dced | 406 | /* TODO: conn_bcast_event() ?? */ |
a2972846 | 407 | notify_helper(NOTIFY_CALL, NULL, connection, cmd, 0); |
6b75dced | 408 | |
98683650 | 409 | ret = call_usermodehelper(usermode_helper, argv, envp, UMH_WAIT_PROC); |
6b75dced | 410 | if (ret) |
1ec861eb | 411 | drbd_warn(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 412 | usermode_helper, cmd, resource_name, |
6b75dced PR |
413 | (ret >> 8) & 0xff, ret); |
414 | else | |
1ec861eb | 415 | drbd_info(connection, "helper command: %s %s %s exit code %u (0x%x)\n", |
77c556f6 | 416 | usermode_helper, cmd, resource_name, |
6b75dced PR |
417 | (ret >> 8) & 0xff, ret); |
418 | /* TODO: conn_bcast_event() ?? */ | |
a2972846 | 419 | notify_helper(NOTIFY_RESPONSE, NULL, connection, cmd, ret); |
c2ba686f | 420 | |
b411b363 PR |
421 | if (ret < 0) /* Ignore any ERRNOs we got. */ |
422 | ret = 0; | |
423 | ||
424 | return ret; | |
425 | } | |
426 | ||
bde89a9e | 427 | static enum drbd_fencing_p highest_fencing_policy(struct drbd_connection *connection) |
b411b363 | 428 | { |
cb703454 | 429 | enum drbd_fencing_p fp = FP_NOT_AVAIL; |
c06ece6b | 430 | struct drbd_peer_device *peer_device; |
cb703454 PR |
431 | int vnr; |
432 | ||
695d08fa | 433 | rcu_read_lock(); |
c06ece6b AG |
434 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
435 | struct drbd_device *device = peer_device->device; | |
b30ab791 | 436 | if (get_ldev_if_state(device, D_CONSISTENT)) { |
c06ece6b AG |
437 | struct disk_conf *disk_conf = |
438 | rcu_dereference(peer_device->device->ldev->disk_conf); | |
439 | fp = max_t(enum drbd_fencing_p, fp, disk_conf->fencing); | |
b30ab791 | 440 | put_ldev(device); |
cb703454 PR |
441 | } |
442 | } | |
695d08fa | 443 | rcu_read_unlock(); |
cb703454 | 444 | |
d7fe69c6 PR |
445 | if (fp == FP_NOT_AVAIL) { |
446 | /* IO Suspending works on the whole resource. | |
447 | Do it only for one device. */ | |
448 | vnr = 0; | |
449 | peer_device = idr_get_next(&connection->peer_devices, &vnr); | |
450 | drbd_change_state(peer_device->device, CS_VERBOSE | CS_HARD, NS(susp_fen, 0)); | |
451 | } | |
452 | ||
cb703454 PR |
453 | return fp; |
454 | } | |
455 | ||
bde89a9e | 456 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 457 | { |
28e448bb | 458 | unsigned int connect_cnt; |
cb703454 PR |
459 | union drbd_state mask = { }; |
460 | union drbd_state val = { }; | |
461 | enum drbd_fencing_p fp; | |
b411b363 PR |
462 | char *ex_to_string; |
463 | int r; | |
b411b363 | 464 | |
02df6fe1 | 465 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e | 466 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 467 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
02df6fe1 | 468 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 PR |
469 | return false; |
470 | } | |
b411b363 | 471 | |
bde89a9e | 472 | connect_cnt = connection->connect_cnt; |
0500813f | 473 | spin_unlock_irq(&connection->resource->req_lock); |
28e448bb | 474 | |
bde89a9e | 475 | fp = highest_fencing_policy(connection); |
cb703454 PR |
476 | switch (fp) { |
477 | case FP_NOT_AVAIL: | |
1ec861eb | 478 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
fb22c402 | 479 | goto out; |
cb703454 PR |
480 | case FP_DONT_CARE: |
481 | return true; | |
482 | default: ; | |
b411b363 PR |
483 | } |
484 | ||
bde89a9e | 485 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
486 | |
487 | switch ((r>>8) & 0xff) { | |
488 | case 3: /* peer is inconsistent */ | |
489 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
490 | mask.pdsk = D_MASK; |
491 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
492 | break; |
493 | case 4: /* peer got outdated, or was already outdated */ | |
494 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
495 | mask.pdsk = D_MASK; |
496 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
497 | break; |
498 | case 5: /* peer was down */ | |
bde89a9e | 499 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
500 | /* we will(have) create(d) a new UUID anyways... */ |
501 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
502 | mask.pdsk = D_MASK; |
503 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
504 | } else { |
505 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
506 | } |
507 | break; | |
508 | case 6: /* Peer is primary, voluntarily outdate myself. | |
509 | * This is useful when an unconnected R_SECONDARY is asked to | |
510 | * become R_PRIMARY, but finds the other peer being active. */ | |
511 | ex_to_string = "peer is active"; | |
1ec861eb | 512 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
513 | mask.disk = D_MASK; |
514 | val.disk = D_OUTDATED; | |
b411b363 PR |
515 | break; |
516 | case 7: | |
517 | if (fp != FP_STONITH) | |
1ec861eb | 518 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 519 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
520 | mask.pdsk = D_MASK; |
521 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
522 | break; |
523 | default: | |
524 | /* The script is broken ... */ | |
1ec861eb | 525 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 526 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
527 | } |
528 | ||
1ec861eb | 529 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 530 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 531 | |
cb703454 | 532 | out: |
fb22c402 | 533 | |
cb703454 | 534 | /* Not using |
bde89a9e | 535 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
536 | here, because we might were able to re-establish the connection in the |
537 | meantime. */ | |
0500813f | 538 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
539 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
540 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
541 | /* In case the connection was established and droped |
542 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 543 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 544 | else |
bde89a9e | 545 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 546 | } |
0500813f | 547 | spin_unlock_irq(&connection->resource->req_lock); |
cb703454 | 548 | |
bde89a9e | 549 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
550 | } |
551 | ||
87f7be4c PR |
552 | static int _try_outdate_peer_async(void *data) |
553 | { | |
bde89a9e | 554 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 555 | |
bde89a9e | 556 | conn_try_outdate_peer(connection); |
87f7be4c | 557 | |
05a10ec7 | 558 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
559 | return 0; |
560 | } | |
561 | ||
bde89a9e | 562 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
563 | { |
564 | struct task_struct *opa; | |
565 | ||
bde89a9e | 566 | kref_get(&connection->kref); |
bbc1c5e8 LE |
567 | /* We may just have force_sig()'ed this thread |
568 | * to get it out of some blocking network function. | |
569 | * Clear signals; otherwise kthread_run(), which internally uses | |
570 | * wait_on_completion_killable(), will mistake our pending signal | |
571 | * for a new fatal signal and fail. */ | |
572 | flush_signals(current); | |
bde89a9e | 573 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); |
9dc9fbb3 | 574 | if (IS_ERR(opa)) { |
1ec861eb | 575 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 576 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 577 | } |
87f7be4c | 578 | } |
b411b363 | 579 | |
bf885f8a | 580 | enum drbd_state_rv |
44a4d551 | 581 | drbd_set_role(struct drbd_device *const device, enum drbd_role new_role, int force) |
b411b363 | 582 | { |
44a4d551 LE |
583 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
584 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; | |
b411b363 | 585 | const int max_tries = 4; |
bf885f8a | 586 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 587 | struct net_conf *nc; |
b411b363 PR |
588 | int try = 0; |
589 | int forced = 0; | |
590 | union drbd_state mask, val; | |
b411b363 | 591 | |
b6f85ef9 AG |
592 | if (new_role == R_PRIMARY) { |
593 | struct drbd_connection *connection; | |
594 | ||
595 | /* Detect dead peers as soon as possible. */ | |
596 | ||
597 | rcu_read_lock(); | |
598 | for_each_connection(connection, device->resource) | |
599 | request_ping(connection); | |
600 | rcu_read_unlock(); | |
601 | } | |
b411b363 | 602 | |
b30ab791 | 603 | mutex_lock(device->state_mutex); |
b411b363 PR |
604 | |
605 | mask.i = 0; mask.role = R_MASK; | |
606 | val.i = 0; val.role = new_role; | |
607 | ||
608 | while (try++ < max_tries) { | |
a8821531 | 609 | rv = _drbd_request_state_holding_state_mutex(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
610 | |
611 | /* in case we first succeeded to outdate, | |
612 | * but now suddenly could establish a connection */ | |
bf885f8a | 613 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
614 | val.pdsk = 0; |
615 | mask.pdsk = 0; | |
616 | continue; | |
617 | } | |
618 | ||
bf885f8a | 619 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
620 | (device->state.disk < D_UP_TO_DATE && |
621 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
622 | mask.disk = D_MASK; |
623 | val.disk = D_UP_TO_DATE; | |
624 | forced = 1; | |
625 | continue; | |
626 | } | |
627 | ||
bf885f8a | 628 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 629 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 630 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 631 | |
44a4d551 | 632 | if (conn_try_outdate_peer(connection)) { |
b411b363 PR |
633 | val.disk = D_UP_TO_DATE; |
634 | mask.disk = D_MASK; | |
635 | } | |
b411b363 PR |
636 | continue; |
637 | } | |
638 | ||
bf885f8a | 639 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 640 | goto out; |
bf885f8a | 641 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
44a4d551 | 642 | if (!conn_try_outdate_peer(connection) && force) { |
d0180171 | 643 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
644 | mask.pdsk = D_MASK; |
645 | val.pdsk = D_OUTDATED; | |
b411b363 | 646 | |
cb703454 | 647 | } |
b411b363 PR |
648 | continue; |
649 | } | |
bf885f8a | 650 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
651 | /* Maybe the peer is detected as dead very soon... |
652 | retry at most once more in this case. */ | |
44ed167d PR |
653 | int timeo; |
654 | rcu_read_lock(); | |
44a4d551 | 655 | nc = rcu_dereference(connection->net_conf); |
44ed167d PR |
656 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
657 | rcu_read_unlock(); | |
658 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
659 | if (try < max_tries) |
660 | try = max_tries - 1; | |
661 | continue; | |
662 | } | |
bf885f8a | 663 | if (rv < SS_SUCCESS) { |
b30ab791 | 664 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 665 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 666 | if (rv < SS_SUCCESS) |
3b98c0c2 | 667 | goto out; |
b411b363 PR |
668 | } |
669 | break; | |
670 | } | |
671 | ||
bf885f8a | 672 | if (rv < SS_SUCCESS) |
3b98c0c2 | 673 | goto out; |
b411b363 PR |
674 | |
675 | if (forced) | |
d0180171 | 676 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
677 | |
678 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 679 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 680 | |
b6dd1a89 LE |
681 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
682 | ||
b411b363 | 683 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
684 | if (get_ldev(device)) { |
685 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
686 | put_ldev(device); | |
b411b363 PR |
687 | } |
688 | } else { | |
66ce6dbc | 689 | mutex_lock(&device->resource->conf_update); |
44a4d551 | 690 | nc = connection->net_conf; |
44ed167d | 691 | if (nc) |
6139f60d | 692 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
66ce6dbc | 693 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 694 | |
b30ab791 AG |
695 | if (get_ldev(device)) { |
696 | if (((device->state.conn < C_CONNECTED || | |
697 | device->state.pdsk <= D_FAILED) | |
698 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
699 | drbd_uuid_new_current(device); | |
b411b363 | 700 | |
b30ab791 AG |
701 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
702 | put_ldev(device); | |
b411b363 PR |
703 | } |
704 | } | |
705 | ||
19f843aa LE |
706 | /* writeout of activity log covered areas of the bitmap |
707 | * to stable storage done in after state change already */ | |
b411b363 | 708 | |
b30ab791 | 709 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
710 | /* if this was forced, we should consider sync */ |
711 | if (forced) | |
44a4d551 LE |
712 | drbd_send_uuids(peer_device); |
713 | drbd_send_current_state(peer_device); | |
b411b363 PR |
714 | } |
715 | ||
b30ab791 | 716 | drbd_md_sync(device); |
720979fb | 717 | set_disk_ro(device->vdisk, new_role == R_SECONDARY); |
b30ab791 | 718 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 719 | out: |
b30ab791 | 720 | mutex_unlock(device->state_mutex); |
bf885f8a | 721 | return rv; |
b411b363 PR |
722 | } |
723 | ||
3b98c0c2 | 724 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 725 | { |
3b98c0c2 LE |
726 | return err == -ENOMSG ? "required attribute missing" : |
727 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 728 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 729 | "invalid attribute value"; |
ef50a3e3 | 730 | } |
b411b363 | 731 | |
3b98c0c2 | 732 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 733 | { |
a910b123 | 734 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
735 | struct set_role_parms parms; |
736 | int err; | |
737 | enum drbd_ret_code retcode; | |
b411b363 | 738 | |
a910b123 | 739 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
740 | if (!adm_ctx.reply_skb) |
741 | return retcode; | |
742 | if (retcode != NO_ERROR) | |
743 | goto out; | |
b411b363 | 744 | |
3b98c0c2 LE |
745 | memset(&parms, 0, sizeof(parms)); |
746 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 747 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
748 | if (err) { |
749 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 750 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
751 | goto out; |
752 | } | |
753 | } | |
9e276872 LE |
754 | genl_unlock(); |
755 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 756 | |
3b98c0c2 | 757 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 758 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 759 | else |
b30ab791 | 760 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
761 | |
762 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
763 | genl_lock(); | |
3b98c0c2 | 764 | out: |
a910b123 | 765 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
766 | return 0; |
767 | } | |
768 | ||
ae8bf312 LE |
769 | /* Initializes the md.*_offset members, so we are able to find |
770 | * the on disk meta data. | |
771 | * | |
772 | * We currently have two possible layouts: | |
773 | * external: | |
774 | * |----------- md_size_sect ------------------| | |
775 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
776 | * | al_offset == 8 | | |
777 | * | bm_offset = al_offset + X | | |
778 | * ==> bitmap sectors = md_size_sect - bm_offset | |
779 | * | |
780 | * internal: | |
781 | * |----------- md_size_sect ------------------| | |
782 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
783 | * | al_offset < 0 | | |
784 | * | bm_offset = al_offset - Y | | |
785 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
786 | * | |
787 | * Activity log size used to be fixed 32kB, | |
788 | * but is about to become configurable. | |
789 | */ | |
b30ab791 | 790 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
791 | struct drbd_backing_dev *bdev) |
792 | { | |
793 | sector_t md_size_sect = 0; | |
c04ccaa6 | 794 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 795 | |
3a4d4eb3 LE |
796 | bdev->md.md_offset = drbd_md_ss(bdev); |
797 | ||
68e41a43 | 798 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
799 | default: |
800 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 801 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
802 | bdev->md.al_offset = MD_4kB_SECT; |
803 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
804 | break; |
805 | case DRBD_MD_INDEX_FLEX_EXT: | |
806 | /* just occupy the full device; unit: sectors */ | |
807 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
808 | bdev->md.al_offset = MD_4kB_SECT; |
809 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
810 | break; |
811 | case DRBD_MD_INDEX_INTERNAL: | |
812 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 813 | /* al size is still fixed */ |
ae8bf312 | 814 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
815 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
816 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
817 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
818 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
819 | md_size_sect = ALIGN(md_size_sect, 8); | |
820 | ||
821 | /* plus the "drbd meta data super block", | |
822 | * and the activity log; */ | |
ae8bf312 | 823 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
824 | |
825 | bdev->md.md_size_sect = md_size_sect; | |
826 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 827 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
828 | break; |
829 | } | |
830 | } | |
831 | ||
4b0715f0 | 832 | /* input size is expected to be in KB */ |
b411b363 PR |
833 | char *ppsize(char *buf, unsigned long long size) |
834 | { | |
4b0715f0 LE |
835 | /* Needs 9 bytes at max including trailing NUL: |
836 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
837 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
838 | int base = 0; | |
4b0715f0 | 839 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
840 | /* shift + round */ |
841 | size = (size >> 10) + !!(size & (1<<9)); | |
842 | base++; | |
843 | } | |
4b0715f0 | 844 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
845 | |
846 | return buf; | |
847 | } | |
848 | ||
849 | /* there is still a theoretical deadlock when called from receiver | |
850 | * on an D_INCONSISTENT R_PRIMARY: | |
851 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
852 | * packet from remote to dec_ap_bio again. | |
853 | * receiver receive_sizes(), comes here, | |
854 | * waits for ap_bio_cnt == 0. -> deadlock. | |
855 | * but this cannot happen, actually, because: | |
856 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
857 | * (not connected, or bad/no disk on peer): | |
858 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
859 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
860 | * peer may not initiate a resize. | |
861 | */ | |
3b98c0c2 LE |
862 | /* Note these are not to be confused with |
863 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
864 | * which are (sub) state changes triggered by admin (drbdsetup), | |
865 | * and can be long lived. | |
b30ab791 | 866 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 867 | * and should be short-lived. */ |
b30ab791 | 868 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 869 | { |
b30ab791 AG |
870 | set_bit(SUSPEND_IO, &device->flags); |
871 | if (drbd_suspended(device)) | |
265be2d0 | 872 | return; |
b30ab791 | 873 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
874 | } |
875 | ||
b30ab791 | 876 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 877 | { |
b30ab791 AG |
878 | clear_bit(SUSPEND_IO, &device->flags); |
879 | wake_up(&device->misc_wait); | |
b411b363 PR |
880 | } |
881 | ||
882 | /** | |
883 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 884 | * @device: DRBD device. |
b411b363 PR |
885 | * |
886 | * Returns 0 on success, negative return values indicate errors. | |
887 | * You should call drbd_md_sync() after calling this function. | |
888 | */ | |
d752b269 | 889 | enum determine_dev_size |
b30ab791 | 890 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 PR |
891 | { |
892 | sector_t prev_first_sect, prev_size; /* previous meta location */ | |
cccac985 | 893 | sector_t la_size_sect, u_size; |
b30ab791 | 894 | struct drbd_md *md = &device->ldev->md; |
d752b269 PR |
895 | u32 prev_al_stripe_size_4k; |
896 | u32 prev_al_stripes; | |
b411b363 PR |
897 | sector_t size; |
898 | char ppb[10]; | |
d752b269 | 899 | void *buffer; |
b411b363 PR |
900 | |
901 | int md_moved, la_size_changed; | |
e96c9633 | 902 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 PR |
903 | |
904 | /* race: | |
905 | * application request passes inc_ap_bio, | |
906 | * but then cannot get an AL-reference. | |
907 | * this function later may wait on ap_bio_cnt == 0. -> deadlock. | |
908 | * | |
909 | * to avoid that: | |
910 | * Suspend IO right here. | |
911 | * still lock the act_log to not trigger ASSERTs there. | |
912 | */ | |
b30ab791 | 913 | drbd_suspend_io(device); |
e37d2438 | 914 | buffer = drbd_md_get_buffer(device, __func__); /* Lock meta-data IO */ |
d752b269 | 915 | if (!buffer) { |
b30ab791 | 916 | drbd_resume_io(device); |
d752b269 PR |
917 | return DS_ERROR; |
918 | } | |
b411b363 PR |
919 | |
920 | /* no wait necessary anymore, actually we could assert that */ | |
b30ab791 | 921 | wait_event(device->al_wait, lc_try_lock(device->act_log)); |
b411b363 | 922 | |
b30ab791 AG |
923 | prev_first_sect = drbd_md_first_sector(device->ldev); |
924 | prev_size = device->ldev->md.md_size_sect; | |
925 | la_size_sect = device->ldev->md.la_size_sect; | |
b411b363 | 926 | |
d752b269 PR |
927 | if (rs) { |
928 | /* rs is non NULL if we should change the AL layout only */ | |
929 | ||
930 | prev_al_stripes = md->al_stripes; | |
931 | prev_al_stripe_size_4k = md->al_stripe_size_4k; | |
932 | ||
933 | md->al_stripes = rs->al_stripes; | |
934 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
935 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
936 | } | |
937 | ||
b30ab791 | 938 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 939 | |
daeda1cc | 940 | rcu_read_lock(); |
b30ab791 | 941 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 942 | rcu_read_unlock(); |
b30ab791 | 943 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 944 | |
d752b269 PR |
945 | if (size < la_size_sect) { |
946 | if (rs && u_size == 0) { | |
947 | /* Remove "rs &&" later. This check should always be active, but | |
948 | right now the receiver expects the permissive behavior */ | |
d0180171 | 949 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
950 | "Use --size=%llus for explicit shrink.\n", |
951 | (unsigned long long)size); | |
952 | rv = DS_ERROR_SHRINK; | |
953 | } | |
954 | if (u_size > size) | |
955 | rv = DS_ERROR_SPACE_MD; | |
956 | if (rv != DS_UNCHANGED) | |
957 | goto err_out; | |
958 | } | |
959 | ||
b30ab791 AG |
960 | if (drbd_get_capacity(device->this_bdev) != size || |
961 | drbd_bm_capacity(device) != size) { | |
b411b363 | 962 | int err; |
b30ab791 | 963 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
964 | if (unlikely(err)) { |
965 | /* currently there is only one error: ENOMEM! */ | |
b30ab791 | 966 | size = drbd_bm_capacity(device)>>1; |
b411b363 | 967 | if (size == 0) { |
d0180171 | 968 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
969 | "Could not allocate bitmap!\n"); |
970 | } else { | |
d0180171 | 971 | drbd_err(device, "BM resizing failed. " |
b411b363 PR |
972 | "Leaving size unchanged at size = %lu KB\n", |
973 | (unsigned long)size); | |
974 | } | |
e96c9633 | 975 | rv = DS_ERROR; |
b411b363 PR |
976 | } |
977 | /* racy, see comments above. */ | |
b30ab791 AG |
978 | drbd_set_my_capacity(device, size); |
979 | device->ldev->md.la_size_sect = size; | |
d0180171 | 980 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
981 | (unsigned long long)size>>1); |
982 | } | |
d752b269 PR |
983 | if (rv <= DS_ERROR) |
984 | goto err_out; | |
b411b363 | 985 | |
b30ab791 | 986 | la_size_changed = (la_size_sect != device->ldev->md.la_size_sect); |
b411b363 | 987 | |
b30ab791 AG |
988 | md_moved = prev_first_sect != drbd_md_first_sector(device->ldev) |
989 | || prev_size != device->ldev->md.md_size_sect; | |
b411b363 | 990 | |
d752b269 PR |
991 | if (la_size_changed || md_moved || rs) { |
992 | u32 prev_flags; | |
24dccabb | 993 | |
fcb09674 LE |
994 | /* We do some synchronous IO below, which may take some time. |
995 | * Clear the timer, to avoid scary "timer expired!" messages, | |
996 | * "Superblock" is written out at least twice below, anyways. */ | |
997 | del_timer(&device->md_sync_timer); | |
b30ab791 | 998 | drbd_al_shrink(device); /* All extents inactive. */ |
d752b269 PR |
999 | |
1000 | prev_flags = md->flags; | |
1001 | md->flags &= ~MDF_PRIMARY_IND; | |
b30ab791 | 1002 | drbd_md_write(device, buffer); |
d752b269 | 1003 | |
d0180171 | 1004 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
1005 | la_size_changed && md_moved ? "size changed and md moved" : |
1006 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 1007 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 1008 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 1009 | "size changed", BM_LOCKED_MASK); |
b30ab791 | 1010 | drbd_initialize_al(device, buffer); |
d752b269 PR |
1011 | |
1012 | md->flags = prev_flags; | |
b30ab791 | 1013 | drbd_md_write(device, buffer); |
d752b269 PR |
1014 | |
1015 | if (rs) | |
d0180171 AG |
1016 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
1017 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
1018 | } |
1019 | ||
cccac985 | 1020 | if (size > la_size_sect) |
57737adc | 1021 | rv = la_size_sect ? DS_GREW : DS_GREW_FROM_ZERO; |
cccac985 | 1022 | if (size < la_size_sect) |
e96c9633 | 1023 | rv = DS_SHRUNK; |
d752b269 PR |
1024 | |
1025 | if (0) { | |
1026 | err_out: | |
1027 | if (rs) { | |
1028 | md->al_stripes = prev_al_stripes; | |
1029 | md->al_stripe_size_4k = prev_al_stripe_size_4k; | |
1030 | md->al_size_4k = (u64)prev_al_stripes * prev_al_stripe_size_4k; | |
1031 | ||
b30ab791 | 1032 | drbd_md_set_sector_offsets(device, device->ldev); |
d752b269 PR |
1033 | } |
1034 | } | |
b30ab791 AG |
1035 | lc_unlock(device->act_log); |
1036 | wake_up(&device->al_wait); | |
1037 | drbd_md_put_buffer(device); | |
1038 | drbd_resume_io(device); | |
b411b363 PR |
1039 | |
1040 | return rv; | |
1041 | } | |
1042 | ||
1043 | sector_t | |
b30ab791 | 1044 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1045 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1046 | { |
b30ab791 | 1047 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1048 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1049 | sector_t m_size; /* my size */ |
b411b363 PR |
1050 | sector_t size = 0; |
1051 | ||
1052 | m_size = drbd_get_max_capacity(bdev); | |
1053 | ||
b30ab791 | 1054 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1055 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1056 | p_size = m_size; |
1057 | } | |
1058 | ||
b411b363 PR |
1059 | if (p_size && m_size) { |
1060 | size = min_t(sector_t, p_size, m_size); | |
1061 | } else { | |
cccac985 LE |
1062 | if (la_size_sect) { |
1063 | size = la_size_sect; | |
b411b363 PR |
1064 | if (m_size && m_size < size) |
1065 | size = m_size; | |
1066 | if (p_size && p_size < size) | |
1067 | size = p_size; | |
1068 | } else { | |
1069 | if (m_size) | |
1070 | size = m_size; | |
1071 | if (p_size) | |
1072 | size = p_size; | |
1073 | } | |
1074 | } | |
1075 | ||
1076 | if (size == 0) | |
d0180171 | 1077 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1078 | |
1079 | if (u_size) { | |
1080 | if (u_size > size) | |
d0180171 | 1081 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1082 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1083 | else | |
1084 | size = u_size; | |
1085 | } | |
1086 | ||
1087 | return size; | |
1088 | } | |
1089 | ||
1090 | /** | |
1091 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1092 | * @device: DRBD device. |
b411b363 PR |
1093 | * |
1094 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1095 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1096 | * this function. | |
1097 | */ | |
b30ab791 | 1098 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1099 | { |
1100 | struct lru_cache *n, *t; | |
1101 | struct lc_element *e; | |
1102 | unsigned int in_use; | |
1103 | int i; | |
1104 | ||
b30ab791 AG |
1105 | if (device->act_log && |
1106 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1107 | return 0; |
1108 | ||
1109 | in_use = 0; | |
b30ab791 | 1110 | t = device->act_log; |
7ad651b5 | 1111 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1112 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1113 | |
1114 | if (n == NULL) { | |
d0180171 | 1115 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1116 | return -ENOMEM; |
1117 | } | |
b30ab791 | 1118 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1119 | if (t) { |
1120 | for (i = 0; i < t->nr_elements; i++) { | |
1121 | e = lc_element_by_index(t, i); | |
1122 | if (e->refcnt) | |
d0180171 | 1123 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1124 | e->lc_number, e->refcnt); |
1125 | in_use += e->refcnt; | |
1126 | } | |
1127 | } | |
1128 | if (!in_use) | |
b30ab791 AG |
1129 | device->act_log = n; |
1130 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1131 | if (in_use) { |
d0180171 | 1132 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1133 | lc_destroy(n); |
1134 | return -EBUSY; | |
1135 | } else { | |
d01efcee | 1136 | lc_destroy(t); |
b411b363 | 1137 | } |
b30ab791 | 1138 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1139 | return 0; |
1140 | } | |
1141 | ||
8fe39aac PR |
1142 | static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev, |
1143 | unsigned int max_bio_size) | |
b411b363 | 1144 | { |
b30ab791 | 1145 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1146 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1147 | unsigned int max_segments = 0; | |
c1b3156f | 1148 | struct request_queue *b = NULL; |
99432fcc | 1149 | |
8fe39aac PR |
1150 | if (bdev) { |
1151 | b = bdev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1152 | |
1153 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1154 | rcu_read_lock(); |
b30ab791 | 1155 | max_segments = rcu_dereference(device->ldev->disk_conf)->max_bio_bvecs; |
daeda1cc | 1156 | rcu_read_unlock(); |
c1b3156f PR |
1157 | |
1158 | blk_set_stacking_limits(&q->limits); | |
1159 | blk_queue_max_write_same_sectors(q, 0); | |
99432fcc | 1160 | } |
b411b363 | 1161 | |
b411b363 | 1162 | blk_queue_logical_block_size(q, 512); |
1816a2b4 LE |
1163 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1164 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1165 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
1166 | blk_queue_segment_boundary(q, PAGE_CACHE_SIZE-1); | |
b411b363 | 1167 | |
c1b3156f | 1168 | if (b) { |
20c68fde LE |
1169 | struct drbd_connection *connection = first_peer_device(device)->connection; |
1170 | ||
1171 | if (blk_queue_discard(b) && | |
1172 | (connection->cstate < C_CONNECTED || connection->agreed_features & FF_TRIM)) { | |
20c68fde LE |
1173 | /* For now, don't allow more than one activity log extent worth of data |
1174 | * to be discarded in one go. We may need to rework drbd_al_begin_io() | |
1175 | * to allow for even larger discard ranges */ | |
2bb4cd5c | 1176 | blk_queue_max_discard_sectors(q, DRBD_MAX_DISCARD_SECTORS); |
20c68fde LE |
1177 | |
1178 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); | |
1179 | /* REALLY? Is stacking secdiscard "legal"? */ | |
1180 | if (blk_queue_secdiscard(b)) | |
1181 | queue_flag_set_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1182 | } else { | |
2bb4cd5c | 1183 | blk_queue_max_discard_sectors(q, 0); |
20c68fde LE |
1184 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); |
1185 | queue_flag_clear_unlocked(QUEUE_FLAG_SECDISCARD, q); | |
1186 | } | |
99432fcc PR |
1187 | |
1188 | blk_queue_stack_limits(q, b); | |
1189 | ||
1190 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1191 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1192 | q->backing_dev_info.ra_pages, |
1193 | b->backing_dev_info.ra_pages); | |
1194 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1195 | } | |
99432fcc PR |
1196 | } |
1197 | } | |
1198 | ||
8fe39aac | 1199 | void drbd_reconsider_max_bio_size(struct drbd_device *device, struct drbd_backing_dev *bdev) |
99432fcc | 1200 | { |
db141b2f | 1201 | unsigned int now, new, local, peer; |
99432fcc | 1202 | |
b30ab791 AG |
1203 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1204 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1205 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1206 | |
8fe39aac PR |
1207 | if (bdev) { |
1208 | local = queue_max_hw_sectors(bdev->backing_bdev->bd_disk->queue) << 9; | |
b30ab791 | 1209 | device->local_max_bio_size = local; |
b411b363 | 1210 | } |
db141b2f | 1211 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1212 | |
1213 | /* We may ignore peer limits if the peer is modern enough. | |
1214 | Because new from 8.3.8 onwards the peer can use multiple | |
1215 | BIOs for a single peer_request */ | |
b30ab791 | 1216 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1217 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1218 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1219 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1220 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1221 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1222 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1223 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1224 | else | |
99432fcc | 1225 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1226 | |
fa090e70 LE |
1227 | /* We may later detach and re-attach on a disconnected Primary. |
1228 | * Avoid this setting to jump back in that case. | |
1229 | * We want to store what we know the peer DRBD can handle, | |
1230 | * not what the peer IO backend can handle. */ | |
1231 | if (peer > device->peer_max_bio_size) | |
1232 | device->peer_max_bio_size = peer; | |
1233 | } | |
db141b2f | 1234 | new = min(local, peer); |
99432fcc | 1235 | |
b30ab791 | 1236 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1237 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1238 | |
1239 | if (new != now) | |
d0180171 | 1240 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1241 | |
8fe39aac | 1242 | drbd_setup_queue_param(device, bdev, new); |
b411b363 PR |
1243 | } |
1244 | ||
a18e9d1e | 1245 | /* Starts the worker thread */ |
bde89a9e | 1246 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1247 | { |
bde89a9e | 1248 | drbd_thread_start(&connection->worker); |
b5043c5e | 1249 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1250 | } |
1251 | ||
a18e9d1e | 1252 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1253 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1254 | { |
992d6e91 | 1255 | bool stop_threads; |
0500813f | 1256 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1257 | stop_threads = conn_all_vols_unconf(connection) && |
1258 | connection->cstate == C_STANDALONE; | |
0500813f | 1259 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 LE |
1260 | if (stop_threads) { |
1261 | /* asender is implicitly stopped by receiver | |
81fa2e67 | 1262 | * in conn_disconnect() */ |
bde89a9e AG |
1263 | drbd_thread_stop(&connection->receiver); |
1264 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1265 | } |
b411b363 PR |
1266 | } |
1267 | ||
0778286a | 1268 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1269 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1270 | { |
1271 | int s = 0; | |
1272 | ||
b30ab791 | 1273 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1274 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1275 | return; |
1276 | } | |
1277 | ||
b30ab791 | 1278 | drbd_al_shrink(device); |
0500813f | 1279 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1280 | if (device->state.conn < C_CONNECTED) |
1281 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1282 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1283 | lc_unlock(device->act_log); |
0778286a PR |
1284 | |
1285 | if (s) | |
d0180171 | 1286 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1287 | } |
1288 | ||
5979e361 LE |
1289 | |
1290 | static bool should_set_defaults(struct genl_info *info) | |
1291 | { | |
1292 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1293 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1294 | } | |
1295 | ||
5bbcf5e6 | 1296 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1297 | { |
5bbcf5e6 LE |
1298 | /* This is limited by 16 bit "slot" numbers, |
1299 | * and by available on-disk context storage. | |
1300 | * | |
1301 | * Also (u16)~0 is special (denotes a "free" extent). | |
1302 | * | |
1303 | * One transaction occupies one 4kB on-disk block, | |
1304 | * we have n such blocks in the on disk ring buffer, | |
1305 | * the "current" transaction may fail (n-1), | |
1306 | * and there is 919 slot numbers context information per transaction. | |
1307 | * | |
1308 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1309 | * so cap there first. | |
1310 | */ | |
1311 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1312 | const unsigned int sufficient_on_disk = | |
1313 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1314 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1315 | |
5bbcf5e6 LE |
1316 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1317 | ||
1318 | if (al_size_4k > sufficient_on_disk) | |
1319 | return max_al_nr; | |
1320 | ||
1321 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1322 | } |
1323 | ||
70df7092 LE |
1324 | static bool write_ordering_changed(struct disk_conf *a, struct disk_conf *b) |
1325 | { | |
1326 | return a->disk_barrier != b->disk_barrier || | |
1327 | a->disk_flushes != b->disk_flushes || | |
1328 | a->disk_drain != b->disk_drain; | |
1329 | } | |
1330 | ||
f399002e LE |
1331 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1332 | { | |
a910b123 | 1333 | struct drbd_config_context adm_ctx; |
f399002e | 1334 | enum drbd_ret_code retcode; |
b30ab791 | 1335 | struct drbd_device *device; |
daeda1cc | 1336 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1337 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1338 | int err, fifo_size; |
f399002e | 1339 | |
a910b123 | 1340 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1341 | if (!adm_ctx.reply_skb) |
1342 | return retcode; | |
1343 | if (retcode != NO_ERROR) | |
9e276872 | 1344 | goto finish; |
f399002e | 1345 | |
b30ab791 | 1346 | device = adm_ctx.device; |
9e276872 | 1347 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1348 | |
1349 | /* we also need a disk | |
1350 | * to change the options on */ | |
b30ab791 | 1351 | if (!get_ldev(device)) { |
f399002e LE |
1352 | retcode = ERR_NO_DISK; |
1353 | goto out; | |
1354 | } | |
1355 | ||
daeda1cc | 1356 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1357 | if (!new_disk_conf) { |
f399002e LE |
1358 | retcode = ERR_NOMEM; |
1359 | goto fail; | |
1360 | } | |
1361 | ||
0500813f | 1362 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1363 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1364 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1365 | if (should_set_defaults(info)) |
b966b5dd | 1366 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1367 | |
5ecc72c3 | 1368 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1369 | if (err && err != -ENOMSG) { |
f399002e | 1370 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1371 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1372 | goto fail_unlock; |
f399002e LE |
1373 | } |
1374 | ||
5ecc72c3 LE |
1375 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1376 | new_disk_conf->resync_rate = 1; | |
f399002e | 1377 | |
5bbcf5e6 LE |
1378 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1379 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
b30ab791 AG |
1380 | if (new_disk_conf->al_extents > drbd_al_extents_max(device->ldev)) |
1381 | new_disk_conf->al_extents = drbd_al_extents_max(device->ldev); | |
5bbcf5e6 LE |
1382 | |
1383 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1384 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1385 | |
5ecc72c3 | 1386 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1387 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1388 | new_plan = fifo_alloc(fifo_size); |
1389 | if (!new_plan) { | |
d0180171 | 1390 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1391 | retcode = ERR_NOMEM; |
daeda1cc | 1392 | goto fail_unlock; |
f399002e LE |
1393 | } |
1394 | } | |
1395 | ||
b30ab791 AG |
1396 | drbd_suspend_io(device); |
1397 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1398 | drbd_al_shrink(device); | |
1399 | err = drbd_check_al_size(device, new_disk_conf); | |
1400 | lc_unlock(device->act_log); | |
1401 | wake_up(&device->al_wait); | |
1402 | drbd_resume_io(device); | |
f399002e LE |
1403 | |
1404 | if (err) { | |
1405 | retcode = ERR_NOMEM; | |
daeda1cc | 1406 | goto fail_unlock; |
f399002e LE |
1407 | } |
1408 | ||
28bc3b8c | 1409 | lock_all_resources(); |
b30ab791 | 1410 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1411 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1412 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1413 | drbd_resync_after_changed(device); | |
dc97b708 | 1414 | } |
28bc3b8c | 1415 | unlock_all_resources(); |
f399002e | 1416 | |
daeda1cc PR |
1417 | if (retcode != NO_ERROR) |
1418 | goto fail_unlock; | |
f399002e | 1419 | |
813472ce | 1420 | if (new_plan) { |
b30ab791 AG |
1421 | old_plan = device->rs_plan_s; |
1422 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1423 | } |
9958c857 | 1424 | |
0500813f | 1425 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1426 | |
9a51ab1c | 1427 | if (new_disk_conf->al_updates) |
b30ab791 | 1428 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1429 | else |
b30ab791 | 1430 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1431 | |
691631c0 | 1432 | if (new_disk_conf->md_flushes) |
b30ab791 | 1433 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1434 | else |
b30ab791 | 1435 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1436 | |
70df7092 | 1437 | if (write_ordering_changed(old_disk_conf, new_disk_conf)) |
f6ba8636 | 1438 | drbd_bump_write_ordering(device->resource, NULL, WO_BDEV_FLUSH); |
27eb13e9 | 1439 | |
b30ab791 | 1440 | drbd_md_sync(device); |
f399002e | 1441 | |
69a22773 AG |
1442 | if (device->state.conn >= C_CONNECTED) { |
1443 | struct drbd_peer_device *peer_device; | |
1444 | ||
1445 | for_each_peer_device(peer_device, device) | |
1446 | drbd_send_sync_param(peer_device); | |
1447 | } | |
f399002e | 1448 | |
daeda1cc PR |
1449 | synchronize_rcu(); |
1450 | kfree(old_disk_conf); | |
813472ce | 1451 | kfree(old_plan); |
b30ab791 | 1452 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1453 | goto success; |
1454 | ||
1455 | fail_unlock: | |
0500813f | 1456 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1457 | fail: |
5ecc72c3 | 1458 | kfree(new_disk_conf); |
813472ce | 1459 | kfree(new_plan); |
daeda1cc | 1460 | success: |
b30ab791 | 1461 | put_ldev(device); |
f399002e | 1462 | out: |
9e276872 LE |
1463 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1464 | finish: | |
a910b123 | 1465 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1466 | return 0; |
1467 | } | |
1468 | ||
3b98c0c2 | 1469 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1470 | { |
a910b123 | 1471 | struct drbd_config_context adm_ctx; |
b30ab791 | 1472 | struct drbd_device *device; |
44a4d551 LE |
1473 | struct drbd_peer_device *peer_device; |
1474 | struct drbd_connection *connection; | |
3b98c0c2 | 1475 | int err; |
116676ca | 1476 | enum drbd_ret_code retcode; |
b411b363 PR |
1477 | enum determine_dev_size dd; |
1478 | sector_t max_possible_sectors; | |
1479 | sector_t min_md_device_sectors; | |
1480 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1481 | struct disk_conf *new_disk_conf = NULL; |
e525fd89 | 1482 | struct block_device *bdev; |
b411b363 | 1483 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1484 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1485 | union drbd_state ns, os; |
f2024e7c | 1486 | enum drbd_state_rv rv; |
44ed167d | 1487 | struct net_conf *nc; |
b411b363 | 1488 | |
a910b123 | 1489 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1490 | if (!adm_ctx.reply_skb) |
1491 | return retcode; | |
1492 | if (retcode != NO_ERROR) | |
40cbf085 | 1493 | goto finish; |
b411b363 | 1494 | |
b30ab791 | 1495 | device = adm_ctx.device; |
9e276872 | 1496 | mutex_lock(&adm_ctx.resource->adm_mutex); |
44a4d551 | 1497 | peer_device = first_peer_device(device); |
3b8a44f8 | 1498 | connection = peer_device->connection; |
44a4d551 | 1499 | conn_reconfig_start(connection); |
b411b363 PR |
1500 | |
1501 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1502 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1503 | retcode = ERR_DISK_CONFIGURED; |
1504 | goto fail; | |
1505 | } | |
82f59cc6 LE |
1506 | /* It may just now have detached because of IO error. Make sure |
1507 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1508 | * e.g. if someone calls attach from the on-io-error handler, | |
1509 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
e334f550 | 1510 | wait_event(device->misc_wait, !test_bit(GOING_DISKLESS, &device->flags)); |
b411b363 | 1511 | |
383606e0 | 1512 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1513 | clear_bit(FORCE_DETACH, &device->flags); |
1514 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1515 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1516 | |
0029d624 | 1517 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1518 | device->rs_total = 0; |
1519 | device->rs_failed = 0; | |
1520 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1521 | |
3b98c0c2 | 1522 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1523 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1524 | if (!nbc) { | |
1525 | retcode = ERR_NOMEM; | |
1526 | goto fail; | |
1527 | } | |
9f2247bb PR |
1528 | spin_lock_init(&nbc->md.uuid_lock); |
1529 | ||
daeda1cc PR |
1530 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1531 | if (!new_disk_conf) { | |
1532 | retcode = ERR_NOMEM; | |
b411b363 PR |
1533 | goto fail; |
1534 | } | |
daeda1cc | 1535 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1536 | |
daeda1cc PR |
1537 | set_disk_conf_defaults(new_disk_conf); |
1538 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1539 | if (err) { |
b411b363 | 1540 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1541 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1542 | goto fail; |
1543 | } | |
1544 | ||
5bbcf5e6 LE |
1545 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1546 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1547 | |
9958c857 PR |
1548 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1549 | if (!new_plan) { | |
1550 | retcode = ERR_NOMEM; | |
1551 | goto fail; | |
1552 | } | |
1553 | ||
daeda1cc | 1554 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1555 | retcode = ERR_MD_IDX_INVALID; |
1556 | goto fail; | |
1557 | } | |
1558 | ||
44ed167d | 1559 | rcu_read_lock(); |
44a4d551 | 1560 | nc = rcu_dereference(connection->net_conf); |
44ed167d | 1561 | if (nc) { |
daeda1cc | 1562 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1563 | rcu_read_unlock(); |
47ff2d0a | 1564 | retcode = ERR_STONITH_AND_PROT_A; |
28bc3b8c | 1565 | goto fail; |
47ff2d0a PR |
1566 | } |
1567 | } | |
44ed167d | 1568 | rcu_read_unlock(); |
47ff2d0a | 1569 | |
daeda1cc | 1570 | bdev = blkdev_get_by_path(new_disk_conf->backing_dev, |
b30ab791 | 1571 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, device); |
e525fd89 | 1572 | if (IS_ERR(bdev)) { |
d0180171 | 1573 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->backing_dev, |
e525fd89 | 1574 | PTR_ERR(bdev)); |
b411b363 | 1575 | retcode = ERR_OPEN_DISK; |
28bc3b8c | 1576 | goto fail; |
b411b363 | 1577 | } |
e525fd89 TH |
1578 | nbc->backing_bdev = bdev; |
1579 | ||
1580 | /* | |
1581 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1582 | * drbd sharing one meta device. TODO in that case, paranoia | |
1583 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1584 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1585 | * should check it for you already; but if you don't, or | |
1586 | * someone fooled it, we need to double check here) | |
1587 | */ | |
daeda1cc | 1588 | bdev = blkdev_get_by_path(new_disk_conf->meta_dev, |
d4d77629 | 1589 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, |
daeda1cc | 1590 | (new_disk_conf->meta_dev_idx < 0) ? |
b30ab791 | 1591 | (void *)device : (void *)drbd_m_holder); |
e525fd89 | 1592 | if (IS_ERR(bdev)) { |
d0180171 | 1593 | drbd_err(device, "open(\"%s\") failed with %ld\n", new_disk_conf->meta_dev, |
e525fd89 | 1594 | PTR_ERR(bdev)); |
b411b363 | 1595 | retcode = ERR_OPEN_MD_DISK; |
28bc3b8c | 1596 | goto fail; |
b411b363 | 1597 | } |
e525fd89 | 1598 | nbc->md_bdev = bdev; |
b411b363 | 1599 | |
e525fd89 | 1600 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1601 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1602 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1603 | retcode = ERR_MD_IDX_INVALID; |
28bc3b8c | 1604 | goto fail; |
b411b363 PR |
1605 | } |
1606 | ||
1607 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1608 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1609 | offsetof(struct bm_extent, lce)); |
1610 | if (!resync_lru) { | |
1611 | retcode = ERR_NOMEM; | |
28bc3b8c | 1612 | goto fail; |
b411b363 PR |
1613 | } |
1614 | ||
c04ccaa6 LE |
1615 | /* Read our meta data super block early. |
1616 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1617 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 | 1618 | if (retcode != NO_ERROR) |
28bc3b8c | 1619 | goto fail; |
b411b363 | 1620 | |
5bbcf5e6 LE |
1621 | if (new_disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1622 | new_disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1623 | if (new_disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1624 | new_disk_conf->al_extents = drbd_al_extents_max(nbc); | |
1625 | ||
daeda1cc | 1626 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1627 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1628 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1629 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1630 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1631 | goto fail; |
b411b363 PR |
1632 | } |
1633 | ||
daeda1cc | 1634 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1635 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1636 | /* at least one MB, otherwise it does not make sense */ | |
1637 | min_md_device_sectors = (2<<10); | |
1638 | } else { | |
1639 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1640 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1641 | } |
1642 | ||
b411b363 | 1643 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1644 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1645 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1646 | "at least %llu sectors needed for this meta-disk type\n", |
1647 | (unsigned long long) min_md_device_sectors); | |
28bc3b8c | 1648 | goto fail; |
b411b363 PR |
1649 | } |
1650 | ||
1651 | /* Make sure the new disk is big enough | |
1652 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1653 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1654 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1655 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1656 | goto fail; |
b411b363 PR |
1657 | } |
1658 | ||
1659 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1660 | ||
1352994b | 1661 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1662 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1663 | "to currently maximum possible %llu sectors <==\n", |
1664 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1665 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1666 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1667 | "meta data may help <<==\n"); |
1668 | } | |
1669 | ||
b30ab791 | 1670 | drbd_suspend_io(device); |
b411b363 | 1671 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1672 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1673 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1674 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1675 | * As barriers are counted per resource, | |
1676 | * we'd need to suspend io on all devices of a resource. | |
1677 | */ | |
b30ab791 | 1678 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1679 | /* and for any other previously queued work */ |
44a4d551 | 1680 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 1681 | |
b30ab791 | 1682 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1683 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1684 | drbd_resume_io(device); |
f2024e7c | 1685 | if (rv < SS_SUCCESS) |
28bc3b8c | 1686 | goto fail; |
b411b363 | 1687 | |
b30ab791 | 1688 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1689 | goto force_diskless; |
1690 | ||
b30ab791 AG |
1691 | if (!device->bitmap) { |
1692 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1693 | retcode = ERR_NOMEM; |
1694 | goto force_diskless_dec; | |
1695 | } | |
1696 | } | |
1697 | ||
b30ab791 | 1698 | if (device->state.conn < C_CONNECTED && |
babea49e | 1699 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1700 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1701 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1702 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1703 | retcode = ERR_DATA_NOT_CURRENT; |
1704 | goto force_diskless_dec; | |
1705 | } | |
1706 | ||
1707 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1708 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1709 | retcode = ERR_NOMEM; |
1710 | goto force_diskless_dec; | |
1711 | } | |
1712 | ||
1713 | /* Prevent shrinking of consistent devices ! */ | |
1714 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1715 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1716 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1717 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1718 | goto force_diskless_dec; |
1719 | } | |
1720 | ||
28bc3b8c AG |
1721 | lock_all_resources(); |
1722 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); | |
1723 | if (retcode != NO_ERROR) { | |
1724 | unlock_all_resources(); | |
1725 | goto force_diskless_dec; | |
1726 | } | |
1727 | ||
b411b363 PR |
1728 | /* Reset the "barriers don't work" bits here, then force meta data to |
1729 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1730 | if (new_disk_conf->md_flushes) |
b30ab791 | 1731 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1732 | else |
b30ab791 | 1733 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1734 | |
1735 | /* Point of no return reached. | |
1736 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1737 | * now device takes over responsibility, and the state engine should |
b411b363 | 1738 | * clean it up somewhere. */ |
0b0ba1ef | 1739 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1740 | device->ldev = nbc; |
1741 | device->resync = resync_lru; | |
1742 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1743 | nbc = NULL; |
1744 | resync_lru = NULL; | |
daeda1cc | 1745 | new_disk_conf = NULL; |
9958c857 | 1746 | new_plan = NULL; |
b411b363 | 1747 | |
1ec317d3 | 1748 | drbd_resync_after_changed(device); |
f6ba8636 | 1749 | drbd_bump_write_ordering(device->resource, device->ldev, WO_BDEV_FLUSH); |
28bc3b8c | 1750 | unlock_all_resources(); |
b411b363 | 1751 | |
b30ab791 AG |
1752 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1753 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1754 | else |
b30ab791 | 1755 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1756 | |
b30ab791 | 1757 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1758 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1759 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1760 | |
b30ab791 AG |
1761 | device->send_cnt = 0; |
1762 | device->recv_cnt = 0; | |
1763 | device->read_cnt = 0; | |
1764 | device->writ_cnt = 0; | |
b411b363 | 1765 | |
8fe39aac | 1766 | drbd_reconsider_max_bio_size(device, device->ldev); |
b411b363 PR |
1767 | |
1768 | /* If I am currently not R_PRIMARY, | |
1769 | * but meta data primary indicator is set, | |
1770 | * I just now recover from a hard crash, | |
1771 | * and have been R_PRIMARY before that crash. | |
1772 | * | |
1773 | * Now, if I had no connection before that crash | |
1774 | * (have been degraded R_PRIMARY), chances are that | |
1775 | * I won't find my peer now either. | |
1776 | * | |
1777 | * In that case, and _only_ in that case, | |
1778 | * we use the degr-wfc-timeout instead of the default, | |
1779 | * so we can automatically recover from a crash of a | |
1780 | * degraded but active "cluster" after a certain timeout. | |
1781 | */ | |
b30ab791 AG |
1782 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
1783 | if (device->state.role != R_PRIMARY && | |
1784 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
1785 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
1786 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 1787 | |
b30ab791 | 1788 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 1789 | if (dd <= DS_ERROR) { |
b411b363 PR |
1790 | retcode = ERR_NOMEM_BITMAP; |
1791 | goto force_diskless_dec; | |
e96c9633 | 1792 | } else if (dd == DS_GREW) |
b30ab791 | 1793 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 1794 | |
b30ab791 AG |
1795 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
1796 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
1797 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 1798 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 1799 | "(aka FullSync)\n"); |
b30ab791 | 1800 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 1801 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
1802 | retcode = ERR_IO_MD_DISK; |
1803 | goto force_diskless_dec; | |
1804 | } | |
1805 | } else { | |
b30ab791 | 1806 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 1807 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
1808 | retcode = ERR_IO_MD_DISK; |
1809 | goto force_diskless_dec; | |
1810 | } | |
b411b363 PR |
1811 | } |
1812 | ||
b30ab791 AG |
1813 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
1814 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 1815 | |
0500813f | 1816 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 1817 | os = drbd_read_state(device); |
78bae59b | 1818 | ns = os; |
b411b363 PR |
1819 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
1820 | otherwise investigate MDF_WasUpToDate... | |
1821 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
1822 | otherwise into D_CONSISTENT state. | |
1823 | */ | |
b30ab791 AG |
1824 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
1825 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
1826 | ns.disk = D_CONSISTENT; |
1827 | else | |
1828 | ns.disk = D_OUTDATED; | |
1829 | } else { | |
1830 | ns.disk = D_INCONSISTENT; | |
1831 | } | |
1832 | ||
b30ab791 | 1833 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
1834 | ns.pdsk = D_OUTDATED; |
1835 | ||
daeda1cc PR |
1836 | rcu_read_lock(); |
1837 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 1838 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
1839 | ns.disk = D_UP_TO_DATE; |
1840 | ||
1841 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
1842 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
1843 | this point, because drbd_request_state() modifies these | |
1844 | flags. */ | |
1845 | ||
b30ab791 AG |
1846 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
1847 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 1848 | else |
b30ab791 | 1849 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
1850 | |
1851 | rcu_read_unlock(); | |
1852 | ||
b411b363 PR |
1853 | /* In case we are C_CONNECTED postpone any decision on the new disk |
1854 | state after the negotiation phase. */ | |
b30ab791 AG |
1855 | if (device->state.conn == C_CONNECTED) { |
1856 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
1857 | ns.i = os.i; |
1858 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
1859 | |
1860 | /* We expect to receive up-to-date UUIDs soon. | |
1861 | To avoid a race in receive_state, free p_uuid while | |
1862 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
1863 | kfree(device->p_uuid); |
1864 | device->p_uuid = NULL; | |
b411b363 PR |
1865 | } |
1866 | ||
b30ab791 | 1867 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 1868 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
1869 | |
1870 | if (rv < SS_SUCCESS) | |
1871 | goto force_diskless_dec; | |
1872 | ||
b30ab791 | 1873 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 1874 | |
b30ab791 AG |
1875 | if (device->state.role == R_PRIMARY) |
1876 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 1877 | else |
b30ab791 | 1878 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 1879 | |
b30ab791 AG |
1880 | drbd_md_mark_dirty(device); |
1881 | drbd_md_sync(device); | |
b411b363 | 1882 | |
b30ab791 AG |
1883 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
1884 | put_ldev(device); | |
44a4d551 | 1885 | conn_reconfig_done(connection); |
9e276872 | 1886 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 1887 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1888 | return 0; |
1889 | ||
1890 | force_diskless_dec: | |
b30ab791 | 1891 | put_ldev(device); |
b411b363 | 1892 | force_diskless: |
b30ab791 AG |
1893 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
1894 | drbd_md_sync(device); | |
b411b363 | 1895 | fail: |
44a4d551 | 1896 | conn_reconfig_done(connection); |
b411b363 | 1897 | if (nbc) { |
e525fd89 TH |
1898 | if (nbc->backing_bdev) |
1899 | blkdev_put(nbc->backing_bdev, | |
1900 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1901 | if (nbc->md_bdev) | |
1902 | blkdev_put(nbc->md_bdev, | |
1903 | FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
b411b363 PR |
1904 | kfree(nbc); |
1905 | } | |
daeda1cc | 1906 | kfree(new_disk_conf); |
b411b363 | 1907 | lc_destroy(resync_lru); |
9958c857 | 1908 | kfree(new_plan); |
9e276872 | 1909 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 1910 | finish: |
a910b123 | 1911 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1912 | return 0; |
1913 | } | |
1914 | ||
b30ab791 | 1915 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 1916 | { |
19f83c76 | 1917 | enum drbd_state_rv retcode; |
9a0d9d03 | 1918 | int ret; |
02ee8f95 | 1919 | |
cdfda633 | 1920 | if (force) { |
b30ab791 AG |
1921 | set_bit(FORCE_DETACH, &device->flags); |
1922 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 1923 | retcode = SS_SUCCESS; |
02ee8f95 PR |
1924 | goto out; |
1925 | } | |
1926 | ||
b30ab791 | 1927 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
e37d2438 | 1928 | drbd_md_get_buffer(device, __func__); /* make sure there is no in-flight meta-data IO */ |
b30ab791 AG |
1929 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); |
1930 | drbd_md_put_buffer(device); | |
9a0d9d03 | 1931 | /* D_FAILED will transition to DISKLESS. */ |
b30ab791 AG |
1932 | ret = wait_event_interruptible(device->misc_wait, |
1933 | device->state.disk != D_FAILED); | |
1934 | drbd_resume_io(device); | |
9b2f61ae | 1935 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
1936 | retcode = SS_NOTHING_TO_DO; |
1937 | if (ret) | |
1938 | retcode = ERR_INTR; | |
02ee8f95 | 1939 | out: |
85f75dd7 | 1940 | return retcode; |
b411b363 PR |
1941 | } |
1942 | ||
82f59cc6 LE |
1943 | /* Detaching the disk is a process in multiple stages. First we need to lock |
1944 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
1945 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
1946 | * internal references as well. | |
1947 | * Only then we have finally detached. */ | |
3b98c0c2 | 1948 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1949 | { |
a910b123 | 1950 | struct drbd_config_context adm_ctx; |
116676ca | 1951 | enum drbd_ret_code retcode; |
cdfda633 PR |
1952 | struct detach_parms parms = { }; |
1953 | int err; | |
b411b363 | 1954 | |
a910b123 | 1955 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1956 | if (!adm_ctx.reply_skb) |
1957 | return retcode; | |
1958 | if (retcode != NO_ERROR) | |
1959 | goto out; | |
b411b363 | 1960 | |
cdfda633 PR |
1961 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
1962 | err = detach_parms_from_attrs(&parms, info); | |
1963 | if (err) { | |
1964 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 1965 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
1966 | goto out; |
1967 | } | |
b411b363 PR |
1968 | } |
1969 | ||
9e276872 | 1970 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 1971 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 1972 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 1973 | out: |
a910b123 | 1974 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
1975 | return 0; |
1976 | } | |
b411b363 | 1977 | |
bde89a9e | 1978 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 1979 | { |
c06ece6b | 1980 | struct drbd_peer_device *peer_device; |
695d08fa | 1981 | bool rv = false; |
f399002e LE |
1982 | int vnr; |
1983 | ||
695d08fa | 1984 | rcu_read_lock(); |
c06ece6b AG |
1985 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
1986 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
1987 | if (device->state.conn == C_SYNC_SOURCE || |
1988 | device->state.conn == C_SYNC_TARGET || | |
1989 | device->state.conn == C_PAUSED_SYNC_S || | |
1990 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
1991 | rv = true; |
1992 | break; | |
1993 | } | |
b411b363 | 1994 | } |
695d08fa | 1995 | rcu_read_unlock(); |
b411b363 | 1996 | |
695d08fa | 1997 | return rv; |
f399002e | 1998 | } |
47ff2d0a | 1999 | |
bde89a9e | 2000 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 2001 | { |
c06ece6b | 2002 | struct drbd_peer_device *peer_device; |
695d08fa | 2003 | bool rv = false; |
f399002e LE |
2004 | int vnr; |
2005 | ||
695d08fa | 2006 | rcu_read_lock(); |
c06ece6b AG |
2007 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2008 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2009 | if (device->state.conn == C_VERIFY_S || |
2010 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
2011 | rv = true; |
2012 | break; | |
47ff2d0a PR |
2013 | } |
2014 | } | |
695d08fa | 2015 | rcu_read_unlock(); |
b411b363 | 2016 | |
695d08fa | 2017 | return rv; |
f399002e | 2018 | } |
422028b1 | 2019 | |
cd64397c | 2020 | static enum drbd_ret_code |
270eb5c9 | 2021 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 2022 | { |
c06ece6b | 2023 | struct drbd_peer_device *peer_device; |
cd64397c | 2024 | int i; |
b411b363 | 2025 | |
270eb5c9 AG |
2026 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
2027 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 2028 | return ERR_NEED_APV_100; |
b411b363 | 2029 | |
270eb5c9 | 2030 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
2031 | return ERR_NEED_APV_100; |
2032 | ||
270eb5c9 | 2033 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 2034 | return ERR_NEED_APV_100; |
b411b363 PR |
2035 | } |
2036 | ||
270eb5c9 | 2037 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
2038 | conn_highest_role(connection) == R_PRIMARY && |
2039 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 2040 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 2041 | |
270eb5c9 AG |
2042 | if (new_net_conf->two_primaries && |
2043 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2044 | return ERR_NOT_PROTO_C; |
2045 | ||
c06ece6b AG |
2046 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2047 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2048 | if (get_ldev(device)) { |
2049 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2050 | put_ldev(device); | |
270eb5c9 | 2051 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2052 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2053 | } |
270eb5c9 | 2054 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2055 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2056 | } |
2057 | ||
270eb5c9 | 2058 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2059 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2060 | |
cd64397c PR |
2061 | return NO_ERROR; |
2062 | } | |
b411b363 | 2063 | |
44ed167d | 2064 | static enum drbd_ret_code |
270eb5c9 | 2065 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
2066 | { |
2067 | static enum drbd_ret_code rv; | |
c06ece6b | 2068 | struct drbd_peer_device *peer_device; |
44ed167d | 2069 | int i; |
b411b363 | 2070 | |
44ed167d | 2071 | rcu_read_lock(); |
270eb5c9 | 2072 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2073 | rcu_read_unlock(); |
b411b363 | 2074 | |
179e20b8 | 2075 | /* connection->peer_devices protected by genl_lock() here */ |
c06ece6b AG |
2076 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2077 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2078 | if (!device->bitmap) { |
2079 | if (drbd_bm_init(device)) | |
44ed167d | 2080 | return ERR_NOMEM; |
b411b363 PR |
2081 | } |
2082 | } | |
2083 | ||
44ed167d PR |
2084 | return rv; |
2085 | } | |
b411b363 | 2086 | |
0fd0ea06 PR |
2087 | struct crypto { |
2088 | struct crypto_hash *verify_tfm; | |
2089 | struct crypto_hash *csums_tfm; | |
2090 | struct crypto_hash *cram_hmac_tfm; | |
8d412fc6 | 2091 | struct crypto_hash *integrity_tfm; |
0fd0ea06 | 2092 | }; |
b411b363 | 2093 | |
0fd0ea06 | 2094 | static int |
4b6ad6d4 | 2095 | alloc_hash(struct crypto_hash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2096 | { |
2097 | if (!tfm_name[0]) | |
2098 | return NO_ERROR; | |
b411b363 | 2099 | |
0fd0ea06 PR |
2100 | *tfm = crypto_alloc_hash(tfm_name, 0, CRYPTO_ALG_ASYNC); |
2101 | if (IS_ERR(*tfm)) { | |
2102 | *tfm = NULL; | |
2103 | return err_alg; | |
b411b363 | 2104 | } |
b411b363 | 2105 | |
0fd0ea06 PR |
2106 | return NO_ERROR; |
2107 | } | |
b411b363 | 2108 | |
0fd0ea06 | 2109 | static enum drbd_ret_code |
270eb5c9 | 2110 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2111 | { |
2112 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2113 | enum drbd_ret_code rv; | |
0fd0ea06 | 2114 | |
270eb5c9 | 2115 | rv = alloc_hash(&crypto->csums_tfm, new_net_conf->csums_alg, |
4b6ad6d4 | 2116 | ERR_CSUMS_ALG); |
0fd0ea06 PR |
2117 | if (rv != NO_ERROR) |
2118 | return rv; | |
270eb5c9 | 2119 | rv = alloc_hash(&crypto->verify_tfm, new_net_conf->verify_alg, |
4b6ad6d4 | 2120 | ERR_VERIFY_ALG); |
0fd0ea06 PR |
2121 | if (rv != NO_ERROR) |
2122 | return rv; | |
270eb5c9 | 2123 | rv = alloc_hash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
4b6ad6d4 | 2124 | ERR_INTEGRITY_ALG); |
0fd0ea06 PR |
2125 | if (rv != NO_ERROR) |
2126 | return rv; | |
270eb5c9 | 2127 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2128 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2129 | new_net_conf->cram_hmac_alg); |
b411b363 | 2130 | |
4b6ad6d4 AG |
2131 | rv = alloc_hash(&crypto->cram_hmac_tfm, hmac_name, |
2132 | ERR_AUTH_ALG); | |
b411b363 PR |
2133 | } |
2134 | ||
0fd0ea06 PR |
2135 | return rv; |
2136 | } | |
b411b363 | 2137 | |
0fd0ea06 PR |
2138 | static void free_crypto(struct crypto *crypto) |
2139 | { | |
0fd0ea06 | 2140 | crypto_free_hash(crypto->cram_hmac_tfm); |
8d412fc6 | 2141 | crypto_free_hash(crypto->integrity_tfm); |
0fd0ea06 PR |
2142 | crypto_free_hash(crypto->csums_tfm); |
2143 | crypto_free_hash(crypto->verify_tfm); | |
2144 | } | |
b411b363 | 2145 | |
f399002e LE |
2146 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2147 | { | |
a910b123 | 2148 | struct drbd_config_context adm_ctx; |
f399002e | 2149 | enum drbd_ret_code retcode; |
bde89a9e | 2150 | struct drbd_connection *connection; |
270eb5c9 | 2151 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2152 | int err; |
2153 | int ovr; /* online verify running */ | |
2154 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2155 | struct crypto crypto = { }; |
b411b363 | 2156 | |
a910b123 | 2157 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2158 | if (!adm_ctx.reply_skb) |
2159 | return retcode; | |
2160 | if (retcode != NO_ERROR) | |
9e276872 | 2161 | goto finish; |
b411b363 | 2162 | |
bde89a9e | 2163 | connection = adm_ctx.connection; |
9e276872 | 2164 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2165 | |
270eb5c9 AG |
2166 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2167 | if (!new_net_conf) { | |
f399002e LE |
2168 | retcode = ERR_NOMEM; |
2169 | goto out; | |
2170 | } | |
b411b363 | 2171 | |
bde89a9e | 2172 | conn_reconfig_start(connection); |
b411b363 | 2173 | |
bde89a9e | 2174 | mutex_lock(&connection->data.mutex); |
0500813f | 2175 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2176 | old_net_conf = connection->net_conf; |
2561b9c1 | 2177 | |
270eb5c9 | 2178 | if (!old_net_conf) { |
a910b123 | 2179 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2180 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2181 | goto fail; |
2182 | } | |
2183 | ||
270eb5c9 | 2184 | *new_net_conf = *old_net_conf; |
5979e361 | 2185 | if (should_set_defaults(info)) |
270eb5c9 | 2186 | set_net_conf_defaults(new_net_conf); |
f399002e | 2187 | |
270eb5c9 | 2188 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2189 | if (err && err != -ENOMSG) { |
f399002e | 2190 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2191 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2192 | goto fail; |
2561b9c1 | 2193 | } |
b411b363 | 2194 | |
270eb5c9 | 2195 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2196 | if (retcode != NO_ERROR) |
2197 | goto fail; | |
b411b363 | 2198 | |
f399002e | 2199 | /* re-sync running */ |
bde89a9e | 2200 | rsr = conn_resync_running(connection); |
270eb5c9 | 2201 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2202 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2203 | goto fail; |
b411b363 PR |
2204 | } |
2205 | ||
f399002e | 2206 | /* online verify running */ |
bde89a9e | 2207 | ovr = conn_ov_running(connection); |
270eb5c9 | 2208 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2209 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2210 | goto fail; |
f399002e | 2211 | } |
b411b363 | 2212 | |
270eb5c9 | 2213 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2214 | if (retcode != NO_ERROR) |
b411b363 | 2215 | goto fail; |
f399002e | 2216 | |
270eb5c9 | 2217 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2218 | |
2219 | if (!rsr) { | |
bde89a9e AG |
2220 | crypto_free_hash(connection->csums_tfm); |
2221 | connection->csums_tfm = crypto.csums_tfm; | |
0fd0ea06 | 2222 | crypto.csums_tfm = NULL; |
f399002e LE |
2223 | } |
2224 | if (!ovr) { | |
bde89a9e AG |
2225 | crypto_free_hash(connection->verify_tfm); |
2226 | connection->verify_tfm = crypto.verify_tfm; | |
0fd0ea06 | 2227 | crypto.verify_tfm = NULL; |
b411b363 PR |
2228 | } |
2229 | ||
bde89a9e AG |
2230 | crypto_free_hash(connection->integrity_tfm); |
2231 | connection->integrity_tfm = crypto.integrity_tfm; | |
2232 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2233 | /* Do this without trying to take connection->data.mutex again. */ | |
2234 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2235 | |
bde89a9e AG |
2236 | crypto_free_hash(connection->cram_hmac_tfm); |
2237 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
0fd0ea06 | 2238 | |
0500813f | 2239 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2240 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2241 | synchronize_rcu(); |
270eb5c9 | 2242 | kfree(old_net_conf); |
91fd4dad | 2243 | |
69a22773 AG |
2244 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2245 | struct drbd_peer_device *peer_device; | |
2246 | int vnr; | |
2247 | ||
2248 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2249 | drbd_send_sync_param(peer_device); | |
2250 | } | |
f399002e | 2251 | |
91fd4dad PR |
2252 | goto done; |
2253 | ||
b411b363 | 2254 | fail: |
0500813f | 2255 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2256 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2257 | free_crypto(&crypto); |
270eb5c9 | 2258 | kfree(new_net_conf); |
91fd4dad | 2259 | done: |
bde89a9e | 2260 | conn_reconfig_done(connection); |
f399002e | 2261 | out: |
9e276872 LE |
2262 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2263 | finish: | |
a910b123 | 2264 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2265 | return 0; |
2266 | } | |
2267 | ||
a2972846 AG |
2268 | static void connection_to_info(struct connection_info *info, |
2269 | struct drbd_connection *connection) | |
2270 | { | |
2271 | info->conn_connection_state = connection->cstate; | |
2272 | info->conn_role = conn_highest_peer(connection); | |
2273 | } | |
2274 | ||
2275 | static void peer_device_to_info(struct peer_device_info *info, | |
2276 | struct drbd_peer_device *peer_device) | |
2277 | { | |
2278 | struct drbd_device *device = peer_device->device; | |
2279 | ||
2280 | info->peer_repl_state = | |
2281 | max_t(enum drbd_conns, C_WF_REPORT_PARAMS, device->state.conn); | |
2282 | info->peer_disk_state = device->state.pdsk; | |
2283 | info->peer_resync_susp_user = device->state.user_isp; | |
2284 | info->peer_resync_susp_peer = device->state.peer_isp; | |
2285 | info->peer_resync_susp_dependency = device->state.aftr_isp; | |
2286 | } | |
2287 | ||
3b98c0c2 | 2288 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2289 | { |
a2972846 AG |
2290 | struct connection_info connection_info; |
2291 | enum drbd_notification_type flags; | |
2292 | unsigned int peer_devices = 0; | |
a910b123 | 2293 | struct drbd_config_context adm_ctx; |
c06ece6b | 2294 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2295 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2296 | struct crypto crypto = { }; |
77c556f6 | 2297 | struct drbd_resource *resource; |
bde89a9e | 2298 | struct drbd_connection *connection; |
3b98c0c2 LE |
2299 | enum drbd_ret_code retcode; |
2300 | int i; | |
2301 | int err; | |
b411b363 | 2302 | |
a910b123 | 2303 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2304 | |
3b98c0c2 LE |
2305 | if (!adm_ctx.reply_skb) |
2306 | return retcode; | |
2307 | if (retcode != NO_ERROR) | |
2308 | goto out; | |
089c075d | 2309 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2310 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2311 | retcode = ERR_INVALID_REQUEST; |
2312 | goto out; | |
2313 | } | |
b411b363 | 2314 | |
089c075d AG |
2315 | /* No need for _rcu here. All reconfiguration is |
2316 | * strictly serialized on genl_lock(). We are protected against | |
2317 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2318 | for_each_resource(resource, &drbd_resources) { |
2319 | for_each_connection(connection, resource) { | |
2320 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2321 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2322 | connection->my_addr_len)) { | |
2323 | retcode = ERR_LOCAL_ADDR; | |
2324 | goto out; | |
2325 | } | |
b411b363 | 2326 | |
77c556f6 AG |
2327 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2328 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2329 | connection->peer_addr_len)) { | |
2330 | retcode = ERR_PEER_ADDR; | |
2331 | goto out; | |
2332 | } | |
089c075d | 2333 | } |
b411b363 PR |
2334 | } |
2335 | ||
9e276872 | 2336 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2337 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2338 | conn_reconfig_start(connection); |
b411b363 | 2339 | |
bde89a9e | 2340 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2341 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2342 | goto fail; |
2343 | } | |
2344 | ||
a209b4ae | 2345 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2346 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2347 | if (!new_net_conf) { | |
b411b363 | 2348 | retcode = ERR_NOMEM; |
b411b363 PR |
2349 | goto fail; |
2350 | } | |
2351 | ||
270eb5c9 | 2352 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2353 | |
270eb5c9 | 2354 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2355 | if (err && err != -ENOMSG) { |
b411b363 | 2356 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2357 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2358 | goto fail; |
2359 | } | |
2360 | ||
270eb5c9 | 2361 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2362 | if (retcode != NO_ERROR) |
422028b1 | 2363 | goto fail; |
b411b363 | 2364 | |
270eb5c9 | 2365 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2366 | if (retcode != NO_ERROR) |
2367 | goto fail; | |
b411b363 | 2368 | |
270eb5c9 | 2369 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2370 | |
b5043c5e | 2371 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2372 | |
0500813f | 2373 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2374 | old_net_conf = connection->net_conf; |
2375 | if (old_net_conf) { | |
b411b363 | 2376 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2377 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2378 | goto fail; |
2379 | } | |
270eb5c9 | 2380 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2381 | |
bde89a9e AG |
2382 | conn_free_crypto(connection); |
2383 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2384 | connection->integrity_tfm = crypto.integrity_tfm; | |
2385 | connection->csums_tfm = crypto.csums_tfm; | |
2386 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2387 | |
bde89a9e AG |
2388 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2389 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2390 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2391 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2392 | |
a2972846 AG |
2393 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2394 | peer_devices++; | |
2395 | } | |
2396 | ||
2397 | connection_to_info(&connection_info, connection); | |
2398 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2399 | mutex_lock(¬ification_mutex); | |
2400 | notify_connection_state(NULL, 0, connection, &connection_info, NOTIFY_CREATE | flags); | |
2401 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
2402 | struct peer_device_info peer_device_info; | |
2403 | ||
2404 | peer_device_to_info(&peer_device_info, peer_device); | |
2405 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2406 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, NOTIFY_CREATE | flags); | |
2407 | } | |
2408 | mutex_unlock(¬ification_mutex); | |
0500813f | 2409 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2410 | |
695d08fa | 2411 | rcu_read_lock(); |
c06ece6b AG |
2412 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2413 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2414 | device->send_cnt = 0; |
2415 | device->recv_cnt = 0; | |
b411b363 | 2416 | } |
695d08fa | 2417 | rcu_read_unlock(); |
b411b363 | 2418 | |
bde89a9e | 2419 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2420 | |
bde89a9e | 2421 | conn_reconfig_done(connection); |
9e276872 | 2422 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2423 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2424 | return 0; |
b411b363 | 2425 | |
b411b363 | 2426 | fail: |
0fd0ea06 | 2427 | free_crypto(&crypto); |
270eb5c9 | 2428 | kfree(new_net_conf); |
b411b363 | 2429 | |
bde89a9e | 2430 | conn_reconfig_done(connection); |
9e276872 | 2431 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2432 | out: |
a910b123 | 2433 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2434 | return 0; |
2435 | } | |
2436 | ||
bde89a9e | 2437 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2438 | { |
2439 | enum drbd_state_rv rv; | |
85f75dd7 | 2440 | |
bde89a9e | 2441 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2442 | force ? CS_HARD : 0); |
85f75dd7 LE |
2443 | |
2444 | switch (rv) { | |
2445 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2446 | break; |
85f75dd7 LE |
2447 | case SS_ALREADY_STANDALONE: |
2448 | return SS_SUCCESS; | |
2449 | case SS_PRIMARY_NOP: | |
2450 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2451 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2452 | |
2453 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2454 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2455 | |
85f75dd7 LE |
2456 | break; |
2457 | case SS_CW_FAILED_BY_PEER: | |
2458 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2459 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2460 | disk, D_OUTDATED), 0); |
2461 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2462 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2463 | CS_HARD); |
b411b363 | 2464 | } |
85f75dd7 LE |
2465 | break; |
2466 | default:; | |
2467 | /* no special handling necessary */ | |
2468 | } | |
2469 | ||
f3dfa40a LE |
2470 | if (rv >= SS_SUCCESS) { |
2471 | enum drbd_state_rv rv2; | |
2472 | /* No one else can reconfigure the network while I am here. | |
2473 | * The state handling only uses drbd_thread_stop_nowait(), | |
2474 | * we want to really wait here until the receiver is no more. | |
2475 | */ | |
9693da23 | 2476 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2477 | |
2478 | /* Race breaker. This additional state change request may be | |
2479 | * necessary, if this was a forced disconnect during a receiver | |
2480 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2481 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2482 | * C_STANDALONE already, now, and this becomes a no-op. |
2483 | */ | |
bde89a9e | 2484 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2485 | CS_VERBOSE | CS_HARD); |
2486 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2487 | drbd_err(connection, |
f3dfa40a LE |
2488 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2489 | rv2); | |
a2972846 AG |
2490 | /* Unlike in DRBD 9, the state engine has generated |
2491 | * NOTIFY_DESTROY events before clearing connection->net_conf. */ | |
b411b363 | 2492 | } |
85f75dd7 LE |
2493 | return rv; |
2494 | } | |
b411b363 | 2495 | |
3b98c0c2 | 2496 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2497 | { |
a910b123 | 2498 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2499 | struct disconnect_parms parms; |
bde89a9e | 2500 | struct drbd_connection *connection; |
85f75dd7 | 2501 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2502 | enum drbd_ret_code retcode; |
2503 | int err; | |
2561b9c1 | 2504 | |
a910b123 | 2505 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2506 | if (!adm_ctx.reply_skb) |
2507 | return retcode; | |
2508 | if (retcode != NO_ERROR) | |
2561b9c1 | 2509 | goto fail; |
b411b363 | 2510 | |
bde89a9e | 2511 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2512 | memset(&parms, 0, sizeof(parms)); |
2513 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2514 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2515 | if (err) { |
2516 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2517 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2518 | goto fail; |
2519 | } | |
2520 | } | |
2521 | ||
9e276872 | 2522 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2523 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2524 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2525 | retcode = rv; /* FIXME: Type mismatch. */ |
2526 | else | |
2527 | retcode = NO_ERROR; | |
9e276872 | 2528 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2529 | fail: |
a910b123 | 2530 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2531 | return 0; |
2532 | } | |
2533 | ||
b30ab791 | 2534 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2535 | { |
2536 | int iass; /* I am sync source */ | |
2537 | ||
d0180171 | 2538 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2539 | if (device->state.role != device->state.peer) |
2540 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2541 | else |
a6b32bc3 | 2542 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2543 | |
2544 | if (iass) | |
b30ab791 | 2545 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2546 | else |
b30ab791 | 2547 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2548 | } |
2549 | ||
3b98c0c2 | 2550 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2551 | { |
a910b123 | 2552 | struct drbd_config_context adm_ctx; |
daeda1cc | 2553 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2554 | struct resize_parms rs; |
b30ab791 | 2555 | struct drbd_device *device; |
3b98c0c2 | 2556 | enum drbd_ret_code retcode; |
b411b363 | 2557 | enum determine_dev_size dd; |
d752b269 | 2558 | bool change_al_layout = false; |
6495d2c6 | 2559 | enum dds_flags ddsf; |
daeda1cc | 2560 | sector_t u_size; |
3b98c0c2 | 2561 | int err; |
b411b363 | 2562 | |
a910b123 | 2563 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2564 | if (!adm_ctx.reply_skb) |
2565 | return retcode; | |
2566 | if (retcode != NO_ERROR) | |
9e276872 | 2567 | goto finish; |
3b98c0c2 | 2568 | |
9e276872 | 2569 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2570 | device = adm_ctx.device; |
2571 | if (!get_ldev(device)) { | |
d752b269 PR |
2572 | retcode = ERR_NO_DISK; |
2573 | goto fail; | |
2574 | } | |
2575 | ||
3b98c0c2 | 2576 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2577 | rs.al_stripes = device->ldev->md.al_stripes; |
2578 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2579 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2580 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2581 | if (err) { |
3b98c0c2 | 2582 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2583 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2584 | goto fail_ldev; |
b411b363 PR |
2585 | } |
2586 | } | |
2587 | ||
b30ab791 | 2588 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2589 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2590 | goto fail_ldev; |
b411b363 | 2591 | } |
b411b363 | 2592 | |
b30ab791 AG |
2593 | if (device->state.role == R_SECONDARY && |
2594 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2595 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2596 | goto fail_ldev; |
b411b363 | 2597 | } |
b411b363 | 2598 | |
a6b32bc3 | 2599 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2600 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2601 | goto fail_ldev; |
6495d2c6 PR |
2602 | } |
2603 | ||
daeda1cc | 2604 | rcu_read_lock(); |
b30ab791 | 2605 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2606 | rcu_read_unlock(); |
2607 | if (u_size != (sector_t)rs.resize_size) { | |
2608 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2609 | if (!new_disk_conf) { | |
778f271d | 2610 | retcode = ERR_NOMEM; |
9bcd2521 | 2611 | goto fail_ldev; |
778f271d PR |
2612 | } |
2613 | } | |
2614 | ||
b30ab791 AG |
2615 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2616 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2617 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2618 | ||
2619 | if (al_size_k > (16 * 1024 * 1024)) { | |
2620 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2621 | goto fail_ldev; | |
2622 | } | |
2623 | ||
2624 | if (al_size_k < MD_32kB_SECT/2) { | |
2625 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2626 | goto fail_ldev; | |
2627 | } | |
2628 | ||
cdc6af8d | 2629 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2630 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2631 | goto fail_ldev; | |
2632 | } | |
2633 | ||
2634 | change_al_layout = true; | |
2635 | } | |
2636 | ||
b30ab791 AG |
2637 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2638 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2639 | |
daeda1cc | 2640 | if (new_disk_conf) { |
0500813f | 2641 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2642 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2643 | *new_disk_conf = *old_disk_conf; |
2644 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2645 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2646 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2647 | synchronize_rcu(); |
2648 | kfree(old_disk_conf); | |
b411b363 PR |
2649 | } |
2650 | ||
6495d2c6 | 2651 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2652 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2653 | drbd_md_sync(device); | |
2654 | put_ldev(device); | |
e96c9633 | 2655 | if (dd == DS_ERROR) { |
b411b363 PR |
2656 | retcode = ERR_NOMEM_BITMAP; |
2657 | goto fail; | |
d752b269 PR |
2658 | } else if (dd == DS_ERROR_SPACE_MD) { |
2659 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2660 | goto fail; | |
2661 | } else if (dd == DS_ERROR_SHRINK) { | |
2662 | retcode = ERR_IMPLICIT_SHRINK; | |
2663 | goto fail; | |
b411b363 | 2664 | } |
778f271d | 2665 | |
b30ab791 | 2666 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2667 | if (dd == DS_GREW) |
b30ab791 | 2668 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2669 | |
69a22773 AG |
2670 | drbd_send_uuids(first_peer_device(device)); |
2671 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2672 | } |
2673 | ||
b411b363 | 2674 | fail: |
9e276872 LE |
2675 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2676 | finish: | |
a910b123 | 2677 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2678 | return 0; |
b411b363 | 2679 | |
9bcd2521 | 2680 | fail_ldev: |
b30ab791 | 2681 | put_ldev(device); |
9bcd2521 | 2682 | goto fail; |
b411b363 | 2683 | } |
b411b363 | 2684 | |
f399002e | 2685 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2686 | { |
a910b123 | 2687 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2688 | enum drbd_ret_code retcode; |
b57a1e27 | 2689 | struct res_opts res_opts; |
f399002e | 2690 | int err; |
b411b363 | 2691 | |
a910b123 | 2692 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2693 | if (!adm_ctx.reply_skb) |
2694 | return retcode; | |
2695 | if (retcode != NO_ERROR) | |
2696 | goto fail; | |
b411b363 | 2697 | |
eb6bea67 | 2698 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2699 | if (should_set_defaults(info)) |
b966b5dd | 2700 | set_res_opts_defaults(&res_opts); |
b411b363 | 2701 | |
b57a1e27 | 2702 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2703 | if (err && err != -ENOMSG) { |
b411b363 | 2704 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2705 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2706 | goto fail; |
2707 | } | |
2708 | ||
9e276872 | 2709 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2710 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2711 | if (err) { |
2712 | retcode = ERR_INVALID_REQUEST; | |
2713 | if (err == -ENOMEM) | |
2714 | retcode = ERR_NOMEM; | |
b411b363 | 2715 | } |
9e276872 | 2716 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2717 | |
b411b363 | 2718 | fail: |
a910b123 | 2719 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2720 | return 0; |
2721 | } | |
2722 | ||
3b98c0c2 | 2723 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2724 | { |
a910b123 | 2725 | struct drbd_config_context adm_ctx; |
b30ab791 | 2726 | struct drbd_device *device; |
3b98c0c2 LE |
2727 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2728 | ||
a910b123 | 2729 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2730 | if (!adm_ctx.reply_skb) |
2731 | return retcode; | |
2732 | if (retcode != NO_ERROR) | |
2733 | goto out; | |
2734 | ||
b30ab791 | 2735 | device = adm_ctx.device; |
8fe39aac PR |
2736 | if (!get_ldev(device)) { |
2737 | retcode = ERR_NO_DISK; | |
2738 | goto out; | |
2739 | } | |
2740 | ||
2741 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2742 | |
194bfb32 | 2743 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2744 | * resync just being finished, wait for it before requesting a new resync. |
2745 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2746 | drbd_suspend_io(device); |
2747 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2748 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2749 | |
0b2dafcd PR |
2750 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2751 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2752 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2753 | */ |
b30ab791 AG |
2754 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
2755 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 2756 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2757 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
2758 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
2759 | retcode = ERR_IO_MD_DISK; | |
2760 | } | |
2761 | } else | |
b30ab791 AG |
2762 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
2763 | drbd_resume_io(device); | |
9e276872 | 2764 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2765 | put_ldev(device); |
3b98c0c2 | 2766 | out: |
a910b123 | 2767 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2768 | return 0; |
2769 | } | |
2770 | ||
3b98c0c2 LE |
2771 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
2772 | union drbd_state mask, union drbd_state val) | |
b411b363 | 2773 | { |
a910b123 | 2774 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2775 | enum drbd_ret_code retcode; |
194bfb32 | 2776 | |
a910b123 | 2777 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2778 | if (!adm_ctx.reply_skb) |
2779 | return retcode; | |
2780 | if (retcode != NO_ERROR) | |
2781 | goto out; | |
b411b363 | 2782 | |
9e276872 | 2783 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2784 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 2785 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2786 | out: |
a910b123 | 2787 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2788 | return 0; |
2789 | } | |
2790 | ||
8fe39aac | 2791 | static int drbd_bmio_set_susp_al(struct drbd_device *device) __must_hold(local) |
0778286a PR |
2792 | { |
2793 | int rv; | |
2794 | ||
b30ab791 AG |
2795 | rv = drbd_bmio_set_n_write(device); |
2796 | drbd_suspend_al(device); | |
0778286a PR |
2797 | return rv; |
2798 | } | |
2799 | ||
3b98c0c2 | 2800 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2801 | { |
a910b123 | 2802 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 2803 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 2804 | struct drbd_device *device; |
25b0d6c8 | 2805 | |
a910b123 | 2806 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
2807 | if (!adm_ctx.reply_skb) |
2808 | return retcode; | |
2809 | if (retcode != NO_ERROR) | |
2810 | goto out; | |
2811 | ||
b30ab791 | 2812 | device = adm_ctx.device; |
8fe39aac PR |
2813 | if (!get_ldev(device)) { |
2814 | retcode = ERR_NO_DISK; | |
2815 | goto out; | |
2816 | } | |
2817 | ||
2818 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2819 | |
194bfb32 | 2820 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2821 | * resync just being finished, wait for it before requesting a new resync. |
2822 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2823 | drbd_suspend_io(device); |
2824 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2825 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2826 | |
0b2dafcd PR |
2827 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
2828 | * in the bitmap. Otherwise, try to start a resync handshake | |
2829 | * as sync source for full sync. | |
2830 | */ | |
b30ab791 | 2831 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
2832 | /* The peer will get a resync upon connect anyways. Just make that |
2833 | into a full resync. */ | |
b30ab791 | 2834 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 2835 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 2836 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
2837 | "set_n_write from invalidate_peer", |
2838 | BM_LOCKED_SET_ALLOWED)) | |
2839 | retcode = ERR_IO_MD_DISK; | |
2840 | } | |
2841 | } else | |
b30ab791 AG |
2842 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
2843 | drbd_resume_io(device); | |
9e276872 | 2844 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 2845 | put_ldev(device); |
25b0d6c8 | 2846 | out: |
a910b123 | 2847 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2848 | return 0; |
2849 | } | |
2850 | ||
3b98c0c2 | 2851 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2852 | { |
a910b123 | 2853 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2854 | enum drbd_ret_code retcode; |
b411b363 | 2855 | |
a910b123 | 2856 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2857 | if (!adm_ctx.reply_skb) |
2858 | return retcode; | |
2859 | if (retcode != NO_ERROR) | |
2860 | goto out; | |
b411b363 | 2861 | |
9e276872 | 2862 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2863 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 2864 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 2865 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2866 | out: |
a910b123 | 2867 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2868 | return 0; |
2869 | } | |
2870 | ||
3b98c0c2 | 2871 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2872 | { |
a910b123 | 2873 | struct drbd_config_context adm_ctx; |
da9fbc27 | 2874 | union drbd_dev_state s; |
3b98c0c2 LE |
2875 | enum drbd_ret_code retcode; |
2876 | ||
a910b123 | 2877 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2878 | if (!adm_ctx.reply_skb) |
2879 | return retcode; | |
2880 | if (retcode != NO_ERROR) | |
2881 | goto out; | |
b411b363 | 2882 | |
9e276872 | 2883 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2884 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
2885 | s = adm_ctx.device->state; | |
cd88d030 PR |
2886 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
2887 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
2888 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
2889 | } else { | |
2890 | retcode = ERR_PAUSE_IS_CLEAR; | |
2891 | } | |
2892 | } | |
9e276872 | 2893 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2894 | out: |
a910b123 | 2895 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2896 | return 0; |
2897 | } | |
2898 | ||
3b98c0c2 | 2899 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2900 | { |
3b98c0c2 | 2901 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
2902 | } |
2903 | ||
3b98c0c2 | 2904 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2905 | { |
a910b123 | 2906 | struct drbd_config_context adm_ctx; |
b30ab791 | 2907 | struct drbd_device *device; |
3b98c0c2 LE |
2908 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2909 | ||
a910b123 | 2910 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2911 | if (!adm_ctx.reply_skb) |
2912 | return retcode; | |
2913 | if (retcode != NO_ERROR) | |
2914 | goto out; | |
2915 | ||
9e276872 | 2916 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2917 | device = adm_ctx.device; |
2918 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
2919 | drbd_uuid_new_current(device); | |
2920 | clear_bit(NEW_CUR_UUID, &device->flags); | |
43a5182c | 2921 | } |
b30ab791 AG |
2922 | drbd_suspend_io(device); |
2923 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 2924 | if (retcode == SS_SUCCESS) { |
b30ab791 | 2925 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 2926 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 2927 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 2928 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 2929 | } |
b30ab791 | 2930 | drbd_resume_io(device); |
9e276872 | 2931 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2932 | out: |
a910b123 | 2933 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2934 | return 0; |
2935 | } | |
2936 | ||
3b98c0c2 | 2937 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2938 | { |
3b98c0c2 | 2939 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
2940 | } |
2941 | ||
251b8f8e AG |
2942 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
2943 | struct drbd_resource *resource, | |
2944 | struct drbd_connection *connection, | |
2945 | struct drbd_device *device) | |
b411b363 | 2946 | { |
543cc10b LE |
2947 | struct nlattr *nla; |
2948 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
2949 | if (!nla) | |
2950 | goto nla_put_failure; | |
251b8f8e AG |
2951 | if (device && |
2952 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 2953 | goto nla_put_failure; |
f597f6b8 | 2954 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 2955 | goto nla_put_failure; |
251b8f8e AG |
2956 | if (connection) { |
2957 | if (connection->my_addr_len && | |
2958 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
2959 | goto nla_put_failure; | |
2960 | if (connection->peer_addr_len && | |
2961 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
2962 | goto nla_put_failure; | |
2963 | } | |
543cc10b LE |
2964 | nla_nest_end(skb, nla); |
2965 | return 0; | |
b411b363 | 2966 | |
543cc10b LE |
2967 | nla_put_failure: |
2968 | if (nla) | |
2969 | nla_nest_cancel(skb, nla); | |
2970 | return -EMSGSIZE; | |
2971 | } | |
b411b363 | 2972 | |
a55bbd37 AG |
2973 | /* |
2974 | * The generic netlink dump callbacks are called outside the genl_lock(), so | |
2975 | * they cannot use the simple attribute parsing code which uses global | |
2976 | * attribute tables. | |
2977 | */ | |
2978 | static struct nlattr *find_cfg_context_attr(const struct nlmsghdr *nlh, int attr) | |
2979 | { | |
2980 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; | |
2981 | const int maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; | |
2982 | struct nlattr *nla; | |
2983 | ||
2984 | nla = nla_find(nlmsg_attrdata(nlh, hdrlen), nlmsg_attrlen(nlh, hdrlen), | |
2985 | DRBD_NLA_CFG_CONTEXT); | |
2986 | if (!nla) | |
2987 | return NULL; | |
2988 | return drbd_nla_find_nested(maxtype, nla, __nla_type(attr)); | |
2989 | } | |
2990 | ||
2991 | static void resource_to_info(struct resource_info *, struct drbd_resource *); | |
2992 | ||
2993 | int drbd_adm_dump_resources(struct sk_buff *skb, struct netlink_callback *cb) | |
2994 | { | |
2995 | struct drbd_genlmsghdr *dh; | |
2996 | struct drbd_resource *resource; | |
2997 | struct resource_info resource_info; | |
2998 | struct resource_statistics resource_statistics; | |
2999 | int err; | |
3000 | ||
3001 | rcu_read_lock(); | |
3002 | if (cb->args[0]) { | |
3003 | for_each_resource_rcu(resource, &drbd_resources) | |
3004 | if (resource == (struct drbd_resource *)cb->args[0]) | |
3005 | goto found_resource; | |
3006 | err = 0; /* resource was probably deleted */ | |
3007 | goto out; | |
3008 | } | |
3009 | resource = list_entry(&drbd_resources, | |
3010 | struct drbd_resource, resources); | |
3011 | ||
3012 | found_resource: | |
3013 | list_for_each_entry_continue_rcu(resource, &drbd_resources, resources) { | |
3014 | goto put_result; | |
3015 | } | |
3016 | err = 0; | |
3017 | goto out; | |
3018 | ||
3019 | put_result: | |
3020 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3021 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3022 | NLM_F_MULTI, DRBD_ADM_GET_RESOURCES); | |
3023 | err = -ENOMEM; | |
3024 | if (!dh) | |
3025 | goto out; | |
3026 | dh->minor = -1U; | |
3027 | dh->ret_code = NO_ERROR; | |
3028 | err = nla_put_drbd_cfg_context(skb, resource, NULL, NULL); | |
3029 | if (err) | |
3030 | goto out; | |
3031 | err = res_opts_to_skb(skb, &resource->res_opts, !capable(CAP_SYS_ADMIN)); | |
3032 | if (err) | |
3033 | goto out; | |
3034 | resource_to_info(&resource_info, resource); | |
3035 | err = resource_info_to_skb(skb, &resource_info, !capable(CAP_SYS_ADMIN)); | |
3036 | if (err) | |
3037 | goto out; | |
3038 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
3039 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
3040 | if (err) | |
3041 | goto out; | |
3042 | cb->args[0] = (long)resource; | |
3043 | genlmsg_end(skb, dh); | |
3044 | err = 0; | |
3045 | ||
3046 | out: | |
3047 | rcu_read_unlock(); | |
3048 | if (err) | |
3049 | return err; | |
3050 | return skb->len; | |
3051 | } | |
3052 | ||
3053 | static void device_to_statistics(struct device_statistics *s, | |
3054 | struct drbd_device *device) | |
3055 | { | |
3056 | memset(s, 0, sizeof(*s)); | |
3057 | s->dev_upper_blocked = !may_inc_ap_bio(device); | |
3058 | if (get_ldev(device)) { | |
3059 | struct drbd_md *md = &device->ldev->md; | |
3060 | u64 *history_uuids = (u64 *)s->history_uuids; | |
3061 | struct request_queue *q; | |
3062 | int n; | |
3063 | ||
3064 | spin_lock_irq(&md->uuid_lock); | |
3065 | s->dev_current_uuid = md->uuid[UI_CURRENT]; | |
3066 | BUILD_BUG_ON(sizeof(s->history_uuids) < UI_HISTORY_END - UI_HISTORY_START + 1); | |
3067 | for (n = 0; n < UI_HISTORY_END - UI_HISTORY_START + 1; n++) | |
3068 | history_uuids[n] = md->uuid[UI_HISTORY_START + n]; | |
3069 | for (; n < HISTORY_UUIDS; n++) | |
3070 | history_uuids[n] = 0; | |
3071 | s->history_uuids_len = HISTORY_UUIDS; | |
3072 | spin_unlock_irq(&md->uuid_lock); | |
3073 | ||
3074 | s->dev_disk_flags = md->flags; | |
3075 | q = bdev_get_queue(device->ldev->backing_bdev); | |
3076 | s->dev_lower_blocked = | |
3077 | bdi_congested(&q->backing_dev_info, | |
3078 | (1 << WB_async_congested) | | |
3079 | (1 << WB_sync_congested)); | |
3080 | put_ldev(device); | |
3081 | } | |
3082 | s->dev_size = drbd_get_capacity(device->this_bdev); | |
3083 | s->dev_read = device->read_cnt; | |
3084 | s->dev_write = device->writ_cnt; | |
3085 | s->dev_al_writes = device->al_writ_cnt; | |
3086 | s->dev_bm_writes = device->bm_writ_cnt; | |
3087 | s->dev_upper_pending = atomic_read(&device->ap_bio_cnt); | |
3088 | s->dev_lower_pending = atomic_read(&device->local_cnt); | |
3089 | s->dev_al_suspended = test_bit(AL_SUSPENDED, &device->flags); | |
3090 | s->dev_exposed_data_uuid = device->ed_uuid; | |
3091 | } | |
3092 | ||
3093 | static int put_resource_in_arg0(struct netlink_callback *cb, int holder_nr) | |
3094 | { | |
3095 | if (cb->args[0]) { | |
3096 | struct drbd_resource *resource = | |
3097 | (struct drbd_resource *)cb->args[0]; | |
3098 | kref_put(&resource->kref, drbd_destroy_resource); | |
3099 | } | |
3100 | ||
3101 | return 0; | |
3102 | } | |
3103 | ||
3104 | int drbd_adm_dump_devices_done(struct netlink_callback *cb) { | |
3105 | return put_resource_in_arg0(cb, 7); | |
3106 | } | |
3107 | ||
3108 | static void device_to_info(struct device_info *, struct drbd_device *); | |
3109 | ||
3110 | int drbd_adm_dump_devices(struct sk_buff *skb, struct netlink_callback *cb) | |
3111 | { | |
3112 | struct nlattr *resource_filter; | |
3113 | struct drbd_resource *resource; | |
3114 | struct drbd_device *uninitialized_var(device); | |
3115 | int minor, err, retcode; | |
3116 | struct drbd_genlmsghdr *dh; | |
3117 | struct device_info device_info; | |
3118 | struct device_statistics device_statistics; | |
3119 | struct idr *idr_to_search; | |
3120 | ||
3121 | resource = (struct drbd_resource *)cb->args[0]; | |
3122 | if (!cb->args[0] && !cb->args[1]) { | |
3123 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3124 | if (resource_filter) { | |
3125 | retcode = ERR_RES_NOT_KNOWN; | |
3126 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3127 | if (!resource) | |
3128 | goto put_result; | |
3129 | cb->args[0] = (long)resource; | |
3130 | } | |
3131 | } | |
3132 | ||
3133 | rcu_read_lock(); | |
3134 | minor = cb->args[1]; | |
3135 | idr_to_search = resource ? &resource->devices : &drbd_devices; | |
3136 | device = idr_get_next(idr_to_search, &minor); | |
3137 | if (!device) { | |
3138 | err = 0; | |
3139 | goto out; | |
3140 | } | |
3141 | idr_for_each_entry_continue(idr_to_search, device, minor) { | |
3142 | retcode = NO_ERROR; | |
3143 | goto put_result; /* only one iteration */ | |
3144 | } | |
3145 | err = 0; | |
3146 | goto out; /* no more devices */ | |
3147 | ||
3148 | put_result: | |
3149 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3150 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3151 | NLM_F_MULTI, DRBD_ADM_GET_DEVICES); | |
3152 | err = -ENOMEM; | |
3153 | if (!dh) | |
3154 | goto out; | |
3155 | dh->ret_code = retcode; | |
3156 | dh->minor = -1U; | |
3157 | if (retcode == NO_ERROR) { | |
3158 | dh->minor = device->minor; | |
3159 | err = nla_put_drbd_cfg_context(skb, device->resource, NULL, device); | |
3160 | if (err) | |
3161 | goto out; | |
3162 | if (get_ldev(device)) { | |
3163 | struct disk_conf *disk_conf = | |
3164 | rcu_dereference(device->ldev->disk_conf); | |
3165 | ||
3166 | err = disk_conf_to_skb(skb, disk_conf, !capable(CAP_SYS_ADMIN)); | |
3167 | put_ldev(device); | |
3168 | if (err) | |
3169 | goto out; | |
3170 | } | |
3171 | device_to_info(&device_info, device); | |
3172 | err = device_info_to_skb(skb, &device_info, !capable(CAP_SYS_ADMIN)); | |
3173 | if (err) | |
3174 | goto out; | |
3175 | ||
3176 | device_to_statistics(&device_statistics, device); | |
3177 | err = device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
3178 | if (err) | |
3179 | goto out; | |
3180 | cb->args[1] = minor + 1; | |
3181 | } | |
3182 | genlmsg_end(skb, dh); | |
3183 | err = 0; | |
3184 | ||
3185 | out: | |
3186 | rcu_read_unlock(); | |
3187 | if (err) | |
3188 | return err; | |
3189 | return skb->len; | |
3190 | } | |
3191 | ||
3192 | int drbd_adm_dump_connections_done(struct netlink_callback *cb) | |
3193 | { | |
3194 | return put_resource_in_arg0(cb, 6); | |
3195 | } | |
3196 | ||
3197 | enum { SINGLE_RESOURCE, ITERATE_RESOURCES }; | |
3198 | ||
3199 | int drbd_adm_dump_connections(struct sk_buff *skb, struct netlink_callback *cb) | |
3200 | { | |
3201 | struct nlattr *resource_filter; | |
3202 | struct drbd_resource *resource = NULL, *next_resource; | |
3203 | struct drbd_connection *uninitialized_var(connection); | |
3204 | int err = 0, retcode; | |
3205 | struct drbd_genlmsghdr *dh; | |
3206 | struct connection_info connection_info; | |
3207 | struct connection_statistics connection_statistics; | |
3208 | ||
3209 | rcu_read_lock(); | |
3210 | resource = (struct drbd_resource *)cb->args[0]; | |
3211 | if (!cb->args[0]) { | |
3212 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3213 | if (resource_filter) { | |
3214 | retcode = ERR_RES_NOT_KNOWN; | |
3215 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3216 | if (!resource) | |
3217 | goto put_result; | |
3218 | cb->args[0] = (long)resource; | |
3219 | cb->args[1] = SINGLE_RESOURCE; | |
3220 | } | |
3221 | } | |
3222 | if (!resource) { | |
3223 | if (list_empty(&drbd_resources)) | |
3224 | goto out; | |
3225 | resource = list_first_entry(&drbd_resources, struct drbd_resource, resources); | |
3226 | kref_get(&resource->kref); | |
3227 | cb->args[0] = (long)resource; | |
3228 | cb->args[1] = ITERATE_RESOURCES; | |
3229 | } | |
3230 | ||
3231 | next_resource: | |
3232 | rcu_read_unlock(); | |
3233 | mutex_lock(&resource->conf_update); | |
3234 | rcu_read_lock(); | |
3235 | if (cb->args[2]) { | |
3236 | for_each_connection_rcu(connection, resource) | |
3237 | if (connection == (struct drbd_connection *)cb->args[2]) | |
3238 | goto found_connection; | |
3239 | /* connection was probably deleted */ | |
3240 | goto no_more_connections; | |
3241 | } | |
3242 | connection = list_entry(&resource->connections, struct drbd_connection, connections); | |
3243 | ||
3244 | found_connection: | |
3245 | list_for_each_entry_continue_rcu(connection, &resource->connections, connections) { | |
3246 | if (!has_net_conf(connection)) | |
3247 | continue; | |
3248 | retcode = NO_ERROR; | |
3249 | goto put_result; /* only one iteration */ | |
3250 | } | |
3251 | ||
3252 | no_more_connections: | |
3253 | if (cb->args[1] == ITERATE_RESOURCES) { | |
3254 | for_each_resource_rcu(next_resource, &drbd_resources) { | |
3255 | if (next_resource == resource) | |
3256 | goto found_resource; | |
3257 | } | |
3258 | /* resource was probably deleted */ | |
3259 | } | |
3260 | goto out; | |
3261 | ||
3262 | found_resource: | |
3263 | list_for_each_entry_continue_rcu(next_resource, &drbd_resources, resources) { | |
3264 | mutex_unlock(&resource->conf_update); | |
3265 | kref_put(&resource->kref, drbd_destroy_resource); | |
3266 | resource = next_resource; | |
3267 | kref_get(&resource->kref); | |
3268 | cb->args[0] = (long)resource; | |
3269 | cb->args[2] = 0; | |
3270 | goto next_resource; | |
3271 | } | |
3272 | goto out; /* no more resources */ | |
3273 | ||
3274 | put_result: | |
3275 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3276 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3277 | NLM_F_MULTI, DRBD_ADM_GET_CONNECTIONS); | |
3278 | err = -ENOMEM; | |
3279 | if (!dh) | |
3280 | goto out; | |
3281 | dh->ret_code = retcode; | |
3282 | dh->minor = -1U; | |
3283 | if (retcode == NO_ERROR) { | |
3284 | struct net_conf *net_conf; | |
3285 | ||
3286 | err = nla_put_drbd_cfg_context(skb, resource, connection, NULL); | |
3287 | if (err) | |
3288 | goto out; | |
3289 | net_conf = rcu_dereference(connection->net_conf); | |
3290 | if (net_conf) { | |
3291 | err = net_conf_to_skb(skb, net_conf, !capable(CAP_SYS_ADMIN)); | |
3292 | if (err) | |
3293 | goto out; | |
3294 | } | |
3295 | connection_to_info(&connection_info, connection); | |
3296 | err = connection_info_to_skb(skb, &connection_info, !capable(CAP_SYS_ADMIN)); | |
3297 | if (err) | |
3298 | goto out; | |
3299 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
3300 | err = connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
3301 | if (err) | |
3302 | goto out; | |
3303 | cb->args[2] = (long)connection; | |
3304 | } | |
3305 | genlmsg_end(skb, dh); | |
3306 | err = 0; | |
3307 | ||
3308 | out: | |
3309 | rcu_read_unlock(); | |
3310 | if (resource) | |
3311 | mutex_unlock(&resource->conf_update); | |
3312 | if (err) | |
3313 | return err; | |
3314 | return skb->len; | |
3315 | } | |
3316 | ||
3317 | enum mdf_peer_flag { | |
3318 | MDF_PEER_CONNECTED = 1 << 0, | |
3319 | MDF_PEER_OUTDATED = 1 << 1, | |
3320 | MDF_PEER_FENCING = 1 << 2, | |
3321 | MDF_PEER_FULL_SYNC = 1 << 3, | |
3322 | }; | |
3323 | ||
3324 | static void peer_device_to_statistics(struct peer_device_statistics *s, | |
3325 | struct drbd_peer_device *peer_device) | |
3326 | { | |
3327 | struct drbd_device *device = peer_device->device; | |
3328 | ||
3329 | memset(s, 0, sizeof(*s)); | |
3330 | s->peer_dev_received = device->recv_cnt; | |
3331 | s->peer_dev_sent = device->send_cnt; | |
3332 | s->peer_dev_pending = atomic_read(&device->ap_pending_cnt) + | |
3333 | atomic_read(&device->rs_pending_cnt); | |
3334 | s->peer_dev_unacked = atomic_read(&device->unacked_cnt); | |
3335 | s->peer_dev_out_of_sync = drbd_bm_total_weight(device) << (BM_BLOCK_SHIFT - 9); | |
3336 | s->peer_dev_resync_failed = device->rs_failed << (BM_BLOCK_SHIFT - 9); | |
3337 | if (get_ldev(device)) { | |
3338 | struct drbd_md *md = &device->ldev->md; | |
3339 | ||
3340 | spin_lock_irq(&md->uuid_lock); | |
3341 | s->peer_dev_bitmap_uuid = md->uuid[UI_BITMAP]; | |
3342 | spin_unlock_irq(&md->uuid_lock); | |
3343 | s->peer_dev_flags = | |
3344 | (drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND) ? | |
3345 | MDF_PEER_CONNECTED : 0) + | |
3346 | (drbd_md_test_flag(device->ldev, MDF_CONSISTENT) && | |
3347 | !drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE) ? | |
3348 | MDF_PEER_OUTDATED : 0) + | |
3349 | /* FIXME: MDF_PEER_FENCING? */ | |
3350 | (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) ? | |
3351 | MDF_PEER_FULL_SYNC : 0); | |
3352 | put_ldev(device); | |
3353 | } | |
3354 | } | |
3355 | ||
3356 | int drbd_adm_dump_peer_devices_done(struct netlink_callback *cb) | |
3357 | { | |
3358 | return put_resource_in_arg0(cb, 9); | |
3359 | } | |
3360 | ||
3361 | int drbd_adm_dump_peer_devices(struct sk_buff *skb, struct netlink_callback *cb) | |
3362 | { | |
3363 | struct nlattr *resource_filter; | |
3364 | struct drbd_resource *resource; | |
3365 | struct drbd_device *uninitialized_var(device); | |
3366 | struct drbd_peer_device *peer_device = NULL; | |
3367 | int minor, err, retcode; | |
3368 | struct drbd_genlmsghdr *dh; | |
3369 | struct idr *idr_to_search; | |
3370 | ||
3371 | resource = (struct drbd_resource *)cb->args[0]; | |
3372 | if (!cb->args[0] && !cb->args[1]) { | |
3373 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3374 | if (resource_filter) { | |
3375 | retcode = ERR_RES_NOT_KNOWN; | |
3376 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3377 | if (!resource) | |
3378 | goto put_result; | |
3379 | } | |
3380 | cb->args[0] = (long)resource; | |
3381 | } | |
3382 | ||
3383 | rcu_read_lock(); | |
3384 | minor = cb->args[1]; | |
3385 | idr_to_search = resource ? &resource->devices : &drbd_devices; | |
3386 | device = idr_find(idr_to_search, minor); | |
3387 | if (!device) { | |
3388 | next_device: | |
3389 | minor++; | |
3390 | cb->args[2] = 0; | |
3391 | device = idr_get_next(idr_to_search, &minor); | |
3392 | if (!device) { | |
3393 | err = 0; | |
3394 | goto out; | |
3395 | } | |
3396 | } | |
3397 | if (cb->args[2]) { | |
3398 | for_each_peer_device(peer_device, device) | |
3399 | if (peer_device == (struct drbd_peer_device *)cb->args[2]) | |
3400 | goto found_peer_device; | |
3401 | /* peer device was probably deleted */ | |
3402 | goto next_device; | |
3403 | } | |
3404 | /* Make peer_device point to the list head (not the first entry). */ | |
3405 | peer_device = list_entry(&device->peer_devices, struct drbd_peer_device, peer_devices); | |
3406 | ||
3407 | found_peer_device: | |
3408 | list_for_each_entry_continue_rcu(peer_device, &device->peer_devices, peer_devices) { | |
3409 | if (!has_net_conf(peer_device->connection)) | |
3410 | continue; | |
3411 | retcode = NO_ERROR; | |
3412 | goto put_result; /* only one iteration */ | |
3413 | } | |
3414 | goto next_device; | |
3415 | ||
3416 | put_result: | |
3417 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3418 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3419 | NLM_F_MULTI, DRBD_ADM_GET_PEER_DEVICES); | |
3420 | err = -ENOMEM; | |
3421 | if (!dh) | |
3422 | goto out; | |
3423 | dh->ret_code = retcode; | |
3424 | dh->minor = -1U; | |
3425 | if (retcode == NO_ERROR) { | |
3426 | struct peer_device_info peer_device_info; | |
3427 | struct peer_device_statistics peer_device_statistics; | |
3428 | ||
3429 | dh->minor = minor; | |
3430 | err = nla_put_drbd_cfg_context(skb, device->resource, peer_device->connection, device); | |
3431 | if (err) | |
3432 | goto out; | |
3433 | peer_device_to_info(&peer_device_info, peer_device); | |
3434 | err = peer_device_info_to_skb(skb, &peer_device_info, !capable(CAP_SYS_ADMIN)); | |
3435 | if (err) | |
3436 | goto out; | |
3437 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
3438 | err = peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
3439 | if (err) | |
3440 | goto out; | |
3441 | cb->args[1] = minor; | |
3442 | cb->args[2] = (long)peer_device; | |
3443 | } | |
3444 | genlmsg_end(skb, dh); | |
3445 | err = 0; | |
3446 | ||
3447 | out: | |
3448 | rcu_read_unlock(); | |
3449 | if (err) | |
3450 | return err; | |
3451 | return skb->len; | |
3452 | } | |
251b8f8e AG |
3453 | /* |
3454 | * Return the connection of @resource if @resource has exactly one connection. | |
3455 | */ | |
3456 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
3457 | { | |
3458 | struct list_head *connections = &resource->connections; | |
3459 | ||
3460 | if (list_empty(connections) || connections->next->next != connections) | |
3461 | return NULL; | |
3462 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
3463 | } | |
3464 | ||
8ce953aa | 3465 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 3466 | const struct sib_info *sib) |
b411b363 | 3467 | { |
251b8f8e | 3468 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
3469 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
3470 | struct nlattr *nla; | |
3471 | int got_ldev; | |
3b98c0c2 LE |
3472 | int err = 0; |
3473 | int exclude_sensitive; | |
3474 | ||
3475 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
3476 | * to. So we better exclude_sensitive information. | |
3477 | * | |
3478 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
3479 | * in the context of the requesting user process. Exclude sensitive | |
3480 | * information, unless current has superuser. | |
3481 | * | |
3482 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
3483 | * relies on the current implementation of netlink_dump(), which | |
3484 | * executes the dump callback successively from netlink_recvmsg(), | |
3485 | * always in the context of the receiving process */ | |
3486 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
3487 | ||
b30ab791 | 3488 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
3489 | |
3490 | /* We need to add connection name and volume number information still. | |
3491 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 3492 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 3493 | goto nla_put_failure; |
3b98c0c2 | 3494 | |
eb6bea67 | 3495 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
3496 | goto nla_put_failure; |
3497 | ||
daeda1cc | 3498 | rcu_read_lock(); |
f9eb7bf4 AG |
3499 | if (got_ldev) { |
3500 | struct disk_conf *disk_conf; | |
44ed167d | 3501 | |
b30ab791 | 3502 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
3503 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
3504 | } | |
3505 | if (!err) { | |
3506 | struct net_conf *nc; | |
3507 | ||
a6b32bc3 | 3508 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
3509 | if (nc) |
3510 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
3511 | } | |
44ed167d PR |
3512 | rcu_read_unlock(); |
3513 | if (err) | |
3514 | goto nla_put_failure; | |
3b98c0c2 | 3515 | |
3b98c0c2 LE |
3516 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
3517 | if (!nla) | |
3518 | goto nla_put_failure; | |
26ec9287 | 3519 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 AG |
3520 | nla_put_u32(skb, T_current_state, device->state.i) || |
3521 | nla_put_u64(skb, T_ed_uuid, device->ed_uuid) || | |
3522 | nla_put_u64(skb, T_capacity, drbd_get_capacity(device->this_bdev)) || | |
3523 | nla_put_u64(skb, T_send_cnt, device->send_cnt) || | |
3524 | nla_put_u64(skb, T_recv_cnt, device->recv_cnt) || | |
3525 | nla_put_u64(skb, T_read_cnt, device->read_cnt) || | |
3526 | nla_put_u64(skb, T_writ_cnt, device->writ_cnt) || | |
3527 | nla_put_u64(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
3528 | nla_put_u64(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
3529 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || | |
3530 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
3531 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 3532 | goto nla_put_failure; |
3b98c0c2 LE |
3533 | |
3534 | if (got_ldev) { | |
39a1aa7f | 3535 | int err; |
b411b363 | 3536 | |
b30ab791 AG |
3537 | spin_lock_irq(&device->ldev->md.uuid_lock); |
3538 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
3539 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
3540 | |
3541 | if (err) | |
3542 | goto nla_put_failure; | |
3543 | ||
b30ab791 AG |
3544 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
3545 | nla_put_u64(skb, T_bits_total, drbd_bm_bits(device)) || | |
3546 | nla_put_u64(skb, T_bits_oos, drbd_bm_total_weight(device))) | |
26ec9287 | 3547 | goto nla_put_failure; |
b30ab791 AG |
3548 | if (C_SYNC_SOURCE <= device->state.conn && |
3549 | C_PAUSED_SYNC_T >= device->state.conn) { | |
3550 | if (nla_put_u64(skb, T_bits_rs_total, device->rs_total) || | |
3551 | nla_put_u64(skb, T_bits_rs_failed, device->rs_failed)) | |
26ec9287 | 3552 | goto nla_put_failure; |
3b98c0c2 | 3553 | } |
b411b363 | 3554 | } |
b411b363 | 3555 | |
3b98c0c2 LE |
3556 | if (sib) { |
3557 | switch(sib->sib_reason) { | |
3558 | case SIB_SYNC_PROGRESS: | |
3559 | case SIB_GET_STATUS_REPLY: | |
3560 | break; | |
3561 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
3562 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
3563 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
3564 | goto nla_put_failure; | |
3b98c0c2 LE |
3565 | break; |
3566 | case SIB_HELPER_POST: | |
26ec9287 AG |
3567 | if (nla_put_u32(skb, T_helper_exit_code, |
3568 | sib->helper_exit_code)) | |
3569 | goto nla_put_failure; | |
3b98c0c2 LE |
3570 | /* fall through */ |
3571 | case SIB_HELPER_PRE: | |
26ec9287 AG |
3572 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
3573 | goto nla_put_failure; | |
3b98c0c2 LE |
3574 | break; |
3575 | } | |
b411b363 | 3576 | } |
3b98c0c2 | 3577 | nla_nest_end(skb, nla); |
b411b363 | 3578 | |
3b98c0c2 LE |
3579 | if (0) |
3580 | nla_put_failure: | |
3581 | err = -EMSGSIZE; | |
3582 | if (got_ldev) | |
b30ab791 | 3583 | put_ldev(device); |
3b98c0c2 | 3584 | return err; |
b411b363 PR |
3585 | } |
3586 | ||
3b98c0c2 | 3587 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3588 | { |
a910b123 | 3589 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3590 | enum drbd_ret_code retcode; |
3591 | int err; | |
b411b363 | 3592 | |
a910b123 | 3593 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3594 | if (!adm_ctx.reply_skb) |
3595 | return retcode; | |
3596 | if (retcode != NO_ERROR) | |
3597 | goto out; | |
b411b363 | 3598 | |
b30ab791 | 3599 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3600 | if (err) { |
3601 | nlmsg_free(adm_ctx.reply_skb); | |
3602 | return err; | |
b411b363 | 3603 | } |
3b98c0c2 | 3604 | out: |
a910b123 | 3605 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3606 | return 0; |
b411b363 PR |
3607 | } |
3608 | ||
4b7a530f | 3609 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3610 | { |
b30ab791 | 3611 | struct drbd_device *device; |
3b98c0c2 | 3612 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3613 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3614 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3615 | struct drbd_resource *tmp; |
543cc10b LE |
3616 | unsigned volume = cb->args[1]; |
3617 | ||
3618 | /* Open coded, deferred, iteration: | |
77c556f6 | 3619 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3620 | * connection = "first connection of resource or undefined"; |
3621 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3622 | * ... |
3623 | * } | |
3624 | * } | |
77c556f6 | 3625 | * where resource is cb->args[0]; |
543cc10b LE |
3626 | * and i is cb->args[1]; |
3627 | * | |
71932efc LE |
3628 | * cb->args[2] indicates if we shall loop over all resources, |
3629 | * or just dump all volumes of a single resource. | |
3630 | * | |
3b98c0c2 LE |
3631 | * This may miss entries inserted after this dump started, |
3632 | * or entries deleted before they are reached. | |
543cc10b | 3633 | * |
b30ab791 | 3634 | * We need to make sure the device won't disappear while |
543cc10b LE |
3635 | * we are looking at it, and revalidate our iterators |
3636 | * on each iteration. | |
3637 | */ | |
b411b363 | 3638 | |
05a10ec7 | 3639 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3640 | rcu_read_lock(); |
543cc10b | 3641 | /* revalidate iterator position */ |
77c556f6 | 3642 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3643 | if (pos == NULL) { |
3644 | /* first iteration */ | |
3645 | pos = tmp; | |
77c556f6 | 3646 | resource = pos; |
543cc10b LE |
3647 | break; |
3648 | } | |
3649 | if (tmp == pos) { | |
77c556f6 | 3650 | resource = pos; |
543cc10b LE |
3651 | break; |
3652 | } | |
b411b363 | 3653 | } |
77c556f6 AG |
3654 | if (resource) { |
3655 | next_resource: | |
251b8f8e AG |
3656 | device = idr_get_next(&resource->devices, &volume); |
3657 | if (!device) { | |
77c556f6 AG |
3658 | /* No more volumes to dump on this resource. |
3659 | * Advance resource iterator. */ | |
3660 | pos = list_entry_rcu(resource->resources.next, | |
3661 | struct drbd_resource, resources); | |
3662 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3663 | if (volume != 0) { |
71932efc LE |
3664 | /* If we reached the end of the list, |
3665 | * or only a single resource dump was requested, | |
3666 | * we are done. */ | |
77c556f6 | 3667 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3668 | goto out; |
543cc10b | 3669 | volume = 0; |
77c556f6 AG |
3670 | resource = pos; |
3671 | goto next_resource; | |
543cc10b LE |
3672 | } |
3673 | } | |
3674 | ||
98683650 | 3675 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3676 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3677 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3678 | if (!dh) | |
543cc10b LE |
3679 | goto out; |
3680 | ||
251b8f8e | 3681 | if (!device) { |
bde89a9e | 3682 | /* This is a connection without a single volume. |
367d675d LE |
3683 | * Suprisingly enough, it may have a network |
3684 | * configuration. */ | |
251b8f8e AG |
3685 | struct drbd_connection *connection; |
3686 | ||
543cc10b LE |
3687 | dh->minor = -1U; |
3688 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3689 | connection = the_only_connection(resource); |
3690 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3691 | goto cancel; |
251b8f8e AG |
3692 | if (connection) { |
3693 | struct net_conf *nc; | |
3694 | ||
3695 | nc = rcu_dereference(connection->net_conf); | |
3696 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3697 | goto cancel; | |
3698 | } | |
367d675d | 3699 | goto done; |
543cc10b | 3700 | } |
b411b363 | 3701 | |
0b0ba1ef | 3702 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3703 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3704 | |
b30ab791 | 3705 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3706 | dh->ret_code = NO_ERROR; |
3707 | ||
b30ab791 | 3708 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3709 | cancel: |
3b98c0c2 | 3710 | genlmsg_cancel(skb, dh); |
543cc10b | 3711 | goto out; |
3b98c0c2 | 3712 | } |
367d675d | 3713 | done: |
3b98c0c2 | 3714 | genlmsg_end(skb, dh); |
bde89a9e | 3715 | } |
b411b363 | 3716 | |
543cc10b | 3717 | out: |
c141ebda | 3718 | rcu_read_unlock(); |
543cc10b | 3719 | /* where to start the next iteration */ |
bde89a9e | 3720 | cb->args[0] = (long)pos; |
77c556f6 | 3721 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3722 | |
77c556f6 | 3723 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3724 | * Which will terminate the dump. */ |
3b98c0c2 | 3725 | return skb->len; |
b411b363 PR |
3726 | } |
3727 | ||
71932efc LE |
3728 | /* |
3729 | * Request status of all resources, or of all volumes within a single resource. | |
3730 | * | |
3731 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
3732 | * Which means we cannot use the family->attrbuf or other such members, because | |
3733 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
3734 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
3735 | * | |
3736 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 3737 | */ |
71932efc | 3738 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3739 | { |
71932efc LE |
3740 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
3741 | struct nlattr *nla; | |
7c3063cc | 3742 | const char *resource_name; |
4bc76048 | 3743 | struct drbd_resource *resource; |
7c3063cc | 3744 | int maxtype; |
71932efc LE |
3745 | |
3746 | /* Is this a followup call? */ | |
3747 | if (cb->args[0]) { | |
3748 | /* ... of a single resource dump, | |
3749 | * and the resource iterator has been advanced already? */ | |
3750 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
3751 | return 0; /* DONE. */ | |
3752 | goto dump; | |
3753 | } | |
3754 | ||
3755 | /* First call (from netlink_dump_start). We need to figure out | |
3756 | * which resource(s) the user wants us to dump. */ | |
3757 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
3758 | nlmsg_attrlen(cb->nlh, hdrlen), | |
3759 | DRBD_NLA_CFG_CONTEXT); | |
3760 | ||
3761 | /* No explicit context given. Dump all. */ | |
3762 | if (!nla) | |
3763 | goto dump; | |
7c3063cc AG |
3764 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
3765 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
3766 | if (IS_ERR(nla)) | |
3767 | return PTR_ERR(nla); | |
71932efc LE |
3768 | /* context given, but no name present? */ |
3769 | if (!nla) | |
3770 | return -EINVAL; | |
7c3063cc | 3771 | resource_name = nla_data(nla); |
4bc76048 AG |
3772 | if (!*resource_name) |
3773 | return -ENODEV; | |
3774 | resource = drbd_find_resource(resource_name); | |
3775 | if (!resource) | |
71932efc LE |
3776 | return -ENODEV; |
3777 | ||
4bc76048 | 3778 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 3779 | |
71932efc | 3780 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 3781 | * only dump this connection. */ |
4bc76048 | 3782 | cb->args[0] = (long)resource; |
71932efc | 3783 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 3784 | cb->args[2] = (long)resource; |
71932efc LE |
3785 | |
3786 | dump: | |
3787 | return get_one_status(skb, cb); | |
3788 | } | |
b411b363 | 3789 | |
3b98c0c2 | 3790 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3791 | { |
a910b123 | 3792 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3793 | enum drbd_ret_code retcode; |
3794 | struct timeout_parms tp; | |
3795 | int err; | |
b411b363 | 3796 | |
a910b123 | 3797 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3798 | if (!adm_ctx.reply_skb) |
3799 | return retcode; | |
3800 | if (retcode != NO_ERROR) | |
3801 | goto out; | |
b411b363 | 3802 | |
3b98c0c2 | 3803 | tp.timeout_type = |
b30ab791 AG |
3804 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
3805 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 3806 | UT_DEFAULT; |
b411b363 | 3807 | |
3b98c0c2 LE |
3808 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
3809 | if (err) { | |
3810 | nlmsg_free(adm_ctx.reply_skb); | |
3811 | return err; | |
3812 | } | |
3813 | out: | |
a910b123 | 3814 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3815 | return 0; |
b411b363 PR |
3816 | } |
3817 | ||
3b98c0c2 | 3818 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3819 | { |
a910b123 | 3820 | struct drbd_config_context adm_ctx; |
b30ab791 | 3821 | struct drbd_device *device; |
3b98c0c2 | 3822 | enum drbd_ret_code retcode; |
58ffa580 | 3823 | struct start_ov_parms parms; |
b411b363 | 3824 | |
a910b123 | 3825 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3826 | if (!adm_ctx.reply_skb) |
3827 | return retcode; | |
3828 | if (retcode != NO_ERROR) | |
3829 | goto out; | |
873b0d5f | 3830 | |
b30ab791 | 3831 | device = adm_ctx.device; |
58ffa580 LE |
3832 | |
3833 | /* resume from last known position, if possible */ | |
b30ab791 | 3834 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 3835 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 3836 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 3837 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
3838 | if (err) { |
3839 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3840 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3841 | goto out; |
3842 | } | |
b411b363 | 3843 | } |
9e276872 LE |
3844 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3845 | ||
58ffa580 | 3846 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
3847 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
3848 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
3849 | |
3850 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
3851 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
3852 | drbd_suspend_io(device); |
3853 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
3854 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
3855 | drbd_resume_io(device); | |
9e276872 LE |
3856 | |
3857 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 3858 | out: |
a910b123 | 3859 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3860 | return 0; |
3861 | } | |
3862 | ||
3863 | ||
3b98c0c2 | 3864 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3865 | { |
a910b123 | 3866 | struct drbd_config_context adm_ctx; |
b30ab791 | 3867 | struct drbd_device *device; |
3b98c0c2 | 3868 | enum drbd_ret_code retcode; |
b411b363 PR |
3869 | int skip_initial_sync = 0; |
3870 | int err; | |
3b98c0c2 | 3871 | struct new_c_uuid_parms args; |
b411b363 | 3872 | |
a910b123 | 3873 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3874 | if (!adm_ctx.reply_skb) |
3875 | return retcode; | |
3876 | if (retcode != NO_ERROR) | |
3877 | goto out_nolock; | |
b411b363 | 3878 | |
b30ab791 | 3879 | device = adm_ctx.device; |
3b98c0c2 LE |
3880 | memset(&args, 0, sizeof(args)); |
3881 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 3882 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
3883 | if (err) { |
3884 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3885 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
3886 | goto out_nolock; |
3887 | } | |
b411b363 PR |
3888 | } |
3889 | ||
9e276872 | 3890 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3891 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 3892 | |
b30ab791 | 3893 | if (!get_ldev(device)) { |
b411b363 PR |
3894 | retcode = ERR_NO_DISK; |
3895 | goto out; | |
3896 | } | |
3897 | ||
3898 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
3899 | if (device->state.conn == C_CONNECTED && |
3900 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 3901 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 3902 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 3903 | skip_initial_sync = 1; |
b30ab791 | 3904 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
3905 | retcode = ERR_CONNECTED; |
3906 | goto out_dec; | |
3907 | } | |
3908 | ||
b30ab791 AG |
3909 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
3910 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
3911 | |
3912 | if (args.clear_bm) { | |
b30ab791 | 3913 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 3914 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 3915 | if (err) { |
d0180171 | 3916 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
3917 | retcode = ERR_IO_MD_DISK; |
3918 | } | |
3919 | if (skip_initial_sync) { | |
69a22773 | 3920 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
3921 | _drbd_uuid_set(device, UI_BITMAP, 0); |
3922 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 3923 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 3924 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 3925 | CS_VERBOSE, NULL); |
0500813f | 3926 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
3927 | } |
3928 | } | |
3929 | ||
b30ab791 | 3930 | drbd_md_sync(device); |
b411b363 | 3931 | out_dec: |
b30ab791 | 3932 | put_ldev(device); |
b411b363 | 3933 | out: |
b30ab791 | 3934 | mutex_unlock(device->state_mutex); |
9e276872 | 3935 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3936 | out_nolock: |
a910b123 | 3937 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3938 | return 0; |
3939 | } | |
3940 | ||
3b98c0c2 | 3941 | static enum drbd_ret_code |
a910b123 | 3942 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 3943 | { |
a910b123 | 3944 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 3945 | if (!name || !name[0]) { |
a910b123 | 3946 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 3947 | return ERR_MANDATORY_TAG; |
b411b363 | 3948 | } |
3b98c0c2 LE |
3949 | /* if we want to use these in sysfs/configfs/debugfs some day, |
3950 | * we must not allow slashes */ | |
3951 | if (strchr(name, '/')) { | |
a910b123 | 3952 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 3953 | return ERR_INVALID_REQUEST; |
b411b363 | 3954 | } |
3b98c0c2 | 3955 | return NO_ERROR; |
774b3055 | 3956 | } |
b411b363 | 3957 | |
a2972846 AG |
3958 | static void resource_to_info(struct resource_info *info, |
3959 | struct drbd_resource *resource) | |
3960 | { | |
3961 | info->res_role = conn_highest_role(first_connection(resource)); | |
3962 | info->res_susp = resource->susp; | |
3963 | info->res_susp_nod = resource->susp_nod; | |
3964 | info->res_susp_fen = resource->susp_fen; | |
3965 | } | |
3966 | ||
789c1b62 | 3967 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3968 | { |
a2972846 | 3969 | struct drbd_connection *connection; |
a910b123 | 3970 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3971 | enum drbd_ret_code retcode; |
afbbfa88 AG |
3972 | struct res_opts res_opts; |
3973 | int err; | |
b411b363 | 3974 | |
a910b123 | 3975 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
3976 | if (!adm_ctx.reply_skb) |
3977 | return retcode; | |
3978 | if (retcode != NO_ERROR) | |
3979 | goto out; | |
b411b363 | 3980 | |
afbbfa88 AG |
3981 | set_res_opts_defaults(&res_opts); |
3982 | err = res_opts_from_attrs(&res_opts, info); | |
3983 | if (err && err != -ENOMSG) { | |
3984 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 3985 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 3986 | goto out; |
b411b363 PR |
3987 | } |
3988 | ||
a910b123 | 3989 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
3990 | if (retcode != NO_ERROR) |
3991 | goto out; | |
b411b363 | 3992 | |
5c661042 | 3993 | if (adm_ctx.resource) { |
38f19616 LE |
3994 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
3995 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 3996 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
3997 | } |
3998 | /* else: still NO_ERROR */ | |
3b98c0c2 | 3999 | goto out; |
b411b363 | 4000 | } |
b411b363 | 4001 | |
9e276872 | 4002 | /* not yet safe for genl_family.parallel_ops */ |
28bc3b8c | 4003 | mutex_lock(&resources_mutex); |
a2972846 | 4004 | connection = conn_create(adm_ctx.resource_name, &res_opts); |
28bc3b8c | 4005 | mutex_unlock(&resources_mutex); |
a2972846 AG |
4006 | |
4007 | if (connection) { | |
4008 | struct resource_info resource_info; | |
4009 | ||
4010 | mutex_lock(¬ification_mutex); | |
4011 | resource_to_info(&resource_info, connection->resource); | |
4012 | notify_resource_state(NULL, 0, connection->resource, | |
4013 | &resource_info, NOTIFY_CREATE); | |
4014 | mutex_unlock(¬ification_mutex); | |
4015 | } else | |
4016 | retcode = ERR_NOMEM; | |
4017 | ||
3b98c0c2 | 4018 | out: |
a910b123 | 4019 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4020 | return 0; |
b411b363 PR |
4021 | } |
4022 | ||
a2972846 AG |
4023 | static void device_to_info(struct device_info *info, |
4024 | struct drbd_device *device) | |
4025 | { | |
4026 | info->dev_disk_state = device->state.disk; | |
4027 | } | |
4028 | ||
4029 | ||
05a10ec7 | 4030 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4031 | { |
a910b123 | 4032 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
4033 | struct drbd_genlmsghdr *dh = info->userhdr; |
4034 | enum drbd_ret_code retcode; | |
b411b363 | 4035 | |
a910b123 | 4036 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
4037 | if (!adm_ctx.reply_skb) |
4038 | return retcode; | |
4039 | if (retcode != NO_ERROR) | |
4040 | goto out; | |
b411b363 | 4041 | |
f2257a56 | 4042 | if (dh->minor > MINORMASK) { |
a910b123 | 4043 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
4044 | retcode = ERR_INVALID_REQUEST; |
4045 | goto out; | |
b411b363 | 4046 | } |
0c8e36d9 | 4047 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 4048 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
4049 | retcode = ERR_INVALID_REQUEST; |
4050 | goto out; | |
b411b363 | 4051 | } |
b411b363 | 4052 | |
38f19616 | 4053 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 4054 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 4055 | if (adm_ctx.device) { |
38f19616 | 4056 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
179e20b8 | 4057 | retcode = ERR_MINOR_OR_VOLUME_EXISTS; |
38f19616 LE |
4058 | /* else: still NO_ERROR */ |
4059 | goto out; | |
b411b363 | 4060 | } |
38f19616 | 4061 | |
9e276872 | 4062 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 4063 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
a2972846 AG |
4064 | if (retcode == NO_ERROR) { |
4065 | struct drbd_device *device; | |
4066 | struct drbd_peer_device *peer_device; | |
4067 | struct device_info info; | |
4068 | unsigned int peer_devices = 0; | |
4069 | enum drbd_notification_type flags; | |
4070 | ||
4071 | device = minor_to_device(dh->minor); | |
4072 | for_each_peer_device(peer_device, device) { | |
4073 | if (!has_net_conf(peer_device->connection)) | |
4074 | continue; | |
4075 | peer_devices++; | |
4076 | } | |
4077 | ||
4078 | device_to_info(&info, device); | |
4079 | mutex_lock(¬ification_mutex); | |
4080 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
4081 | notify_device_state(NULL, 0, device, &info, NOTIFY_CREATE | flags); | |
4082 | for_each_peer_device(peer_device, device) { | |
4083 | struct peer_device_info peer_device_info; | |
4084 | ||
4085 | if (!has_net_conf(peer_device->connection)) | |
4086 | continue; | |
4087 | peer_device_to_info(&peer_device_info, peer_device); | |
4088 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
4089 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, | |
4090 | NOTIFY_CREATE | flags); | |
4091 | } | |
4092 | mutex_unlock(¬ification_mutex); | |
4093 | } | |
9e276872 | 4094 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 4095 | out: |
a910b123 | 4096 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4097 | return 0; |
b411b363 PR |
4098 | } |
4099 | ||
05a10ec7 | 4100 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 4101 | { |
a2972846 AG |
4102 | struct drbd_peer_device *peer_device; |
4103 | ||
b30ab791 AG |
4104 | if (device->state.disk == D_DISKLESS && |
4105 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
4106 | * we may want to delete a minor from a live replication group. |
4107 | */ | |
b30ab791 | 4108 | device->state.role == R_SECONDARY) { |
a2972846 AG |
4109 | struct drbd_connection *connection = |
4110 | first_connection(device->resource); | |
4111 | ||
b30ab791 | 4112 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), |
369bea63 | 4113 | CS_VERBOSE + CS_WAIT_COMPLETE); |
a2972846 AG |
4114 | |
4115 | /* If the state engine hasn't stopped the sender thread yet, we | |
4116 | * need to flush the sender work queue before generating the | |
4117 | * DESTROY events here. */ | |
4118 | if (get_t_state(&connection->worker) == RUNNING) | |
4119 | drbd_flush_workqueue(&connection->sender_work); | |
4120 | ||
4121 | mutex_lock(¬ification_mutex); | |
4122 | for_each_peer_device(peer_device, device) { | |
4123 | if (!has_net_conf(peer_device->connection)) | |
4124 | continue; | |
4125 | notify_peer_device_state(NULL, 0, peer_device, NULL, | |
4126 | NOTIFY_DESTROY | NOTIFY_CONTINUES); | |
4127 | } | |
4128 | notify_device_state(NULL, 0, device, NULL, NOTIFY_DESTROY); | |
4129 | mutex_unlock(¬ification_mutex); | |
4130 | ||
f82795d6 | 4131 | drbd_delete_device(device); |
85f75dd7 LE |
4132 | return NO_ERROR; |
4133 | } else | |
4134 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
4135 | } |
4136 | ||
05a10ec7 | 4137 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4138 | { |
a910b123 | 4139 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 4140 | enum drbd_ret_code retcode; |
b411b363 | 4141 | |
a910b123 | 4142 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4143 | if (!adm_ctx.reply_skb) |
4144 | return retcode; | |
4145 | if (retcode != NO_ERROR) | |
4146 | goto out; | |
b411b363 | 4147 | |
9e276872 | 4148 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 4149 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 4150 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 4151 | out: |
a910b123 | 4152 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 4153 | return 0; |
b411b363 PR |
4154 | } |
4155 | ||
179e20b8 AG |
4156 | static int adm_del_resource(struct drbd_resource *resource) |
4157 | { | |
4158 | struct drbd_connection *connection; | |
4159 | ||
4160 | for_each_connection(connection, resource) { | |
4161 | if (connection->cstate > C_STANDALONE) | |
4162 | return ERR_NET_CONFIGURED; | |
4163 | } | |
4164 | if (!idr_is_empty(&resource->devices)) | |
4165 | return ERR_RES_IN_USE; | |
4166 | ||
a2972846 AG |
4167 | /* The state engine has stopped the sender thread, so we don't |
4168 | * need to flush the sender work queue before generating the | |
4169 | * DESTROY event here. */ | |
4170 | mutex_lock(¬ification_mutex); | |
4171 | notify_resource_state(NULL, 0, resource, NULL, NOTIFY_DESTROY); | |
4172 | mutex_unlock(¬ification_mutex); | |
4173 | ||
28bc3b8c | 4174 | mutex_lock(&resources_mutex); |
179e20b8 | 4175 | list_del_rcu(&resource->resources); |
28bc3b8c | 4176 | mutex_unlock(&resources_mutex); |
179e20b8 AG |
4177 | /* Make sure all threads have actually stopped: state handling only |
4178 | * does drbd_thread_stop_nowait(). */ | |
4179 | list_for_each_entry(connection, &resource->connections, connections) | |
4180 | drbd_thread_stop(&connection->worker); | |
4181 | synchronize_rcu(); | |
4182 | drbd_free_resource(resource); | |
4183 | return NO_ERROR; | |
4184 | } | |
4185 | ||
85f75dd7 | 4186 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4187 | { |
a910b123 | 4188 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
4189 | struct drbd_resource *resource; |
4190 | struct drbd_connection *connection; | |
4191 | struct drbd_device *device; | |
f3dfa40a | 4192 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 4193 | unsigned i; |
b411b363 | 4194 | |
a910b123 | 4195 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
4196 | if (!adm_ctx.reply_skb) |
4197 | return retcode; | |
4198 | if (retcode != NO_ERROR) | |
9e276872 | 4199 | goto finish; |
b411b363 | 4200 | |
b6f85ef9 | 4201 | resource = adm_ctx.resource; |
9e276872 | 4202 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 4203 | /* demote */ |
b6f85ef9 AG |
4204 | for_each_connection(connection, resource) { |
4205 | struct drbd_peer_device *peer_device; | |
4206 | ||
4207 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
4208 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
4209 | if (retcode < SS_SUCCESS) { | |
a910b123 | 4210 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
4211 | goto out; |
4212 | } | |
4213 | } | |
4214 | ||
4215 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 4216 | if (retcode < SS_SUCCESS) { |
a910b123 | 4217 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 4218 | goto out; |
85f75dd7 | 4219 | } |
b411b363 | 4220 | } |
b411b363 | 4221 | |
85f75dd7 | 4222 | /* detach */ |
b6f85ef9 AG |
4223 | idr_for_each_entry(&resource->devices, device, i) { |
4224 | retcode = adm_detach(device, 0); | |
27012382 | 4225 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 4226 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 4227 | goto out; |
85f75dd7 LE |
4228 | } |
4229 | } | |
b411b363 | 4230 | |
85f75dd7 | 4231 | /* delete volumes */ |
b6f85ef9 AG |
4232 | idr_for_each_entry(&resource->devices, device, i) { |
4233 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
4234 | if (retcode != NO_ERROR) { |
4235 | /* "can not happen" */ | |
a910b123 | 4236 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 4237 | goto out; |
85f75dd7 LE |
4238 | } |
4239 | } | |
b411b363 | 4240 | |
179e20b8 | 4241 | retcode = adm_del_resource(resource); |
3b98c0c2 | 4242 | out: |
9e276872 LE |
4243 | mutex_unlock(&resource->adm_mutex); |
4244 | finish: | |
a910b123 | 4245 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4246 | return 0; |
b411b363 PR |
4247 | } |
4248 | ||
789c1b62 | 4249 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4250 | { |
a910b123 | 4251 | struct drbd_config_context adm_ctx; |
77c556f6 | 4252 | struct drbd_resource *resource; |
3b98c0c2 | 4253 | enum drbd_ret_code retcode; |
b411b363 | 4254 | |
a910b123 | 4255 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
4256 | if (!adm_ctx.reply_skb) |
4257 | return retcode; | |
4258 | if (retcode != NO_ERROR) | |
9e276872 | 4259 | goto finish; |
77c556f6 | 4260 | resource = adm_ctx.resource; |
b411b363 | 4261 | |
179e20b8 AG |
4262 | mutex_lock(&resource->adm_mutex); |
4263 | retcode = adm_del_resource(resource); | |
9e276872 LE |
4264 | mutex_unlock(&resource->adm_mutex); |
4265 | finish: | |
a910b123 | 4266 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4267 | return 0; |
4268 | } | |
4269 | ||
b30ab791 | 4270 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 4271 | { |
3b98c0c2 LE |
4272 | struct sk_buff *msg; |
4273 | struct drbd_genlmsghdr *d_out; | |
4274 | unsigned seq; | |
4275 | int err = -ENOMEM; | |
4276 | ||
4277 | seq = atomic_inc_return(&drbd_genl_seq); | |
4278 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4279 | if (!msg) | |
4280 | goto failed; | |
4281 | ||
4282 | err = -EMSGSIZE; | |
4283 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
4284 | if (!d_out) /* cannot happen, but anyways. */ | |
4285 | goto nla_put_failure; | |
b30ab791 | 4286 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 4287 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 4288 | |
b30ab791 | 4289 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
4290 | goto nla_put_failure; |
4291 | genlmsg_end(msg, d_out); | |
4292 | err = drbd_genl_multicast_events(msg, 0); | |
4293 | /* msg has been consumed or freed in netlink_broadcast() */ | |
4294 | if (err && err != -ESRCH) | |
4295 | goto failed; | |
b411b363 | 4296 | |
3b98c0c2 | 4297 | return; |
b411b363 | 4298 | |
3b98c0c2 LE |
4299 | nla_put_failure: |
4300 | nlmsg_free(msg); | |
4301 | failed: | |
d0180171 | 4302 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
4303 | "Event seq:%u sib_reason:%u\n", |
4304 | err, seq, sib->sib_reason); | |
b411b363 | 4305 | } |
a2972846 | 4306 | |
a2972846 AG |
4307 | static int nla_put_notification_header(struct sk_buff *msg, |
4308 | enum drbd_notification_type type) | |
4309 | { | |
4310 | struct drbd_notification_header nh = { | |
4311 | .nh_type = type, | |
4312 | }; | |
4313 | ||
4314 | return drbd_notification_header_to_skb(msg, &nh, true); | |
4315 | } | |
4316 | ||
4317 | void notify_resource_state(struct sk_buff *skb, | |
4318 | unsigned int seq, | |
4319 | struct drbd_resource *resource, | |
4320 | struct resource_info *resource_info, | |
4321 | enum drbd_notification_type type) | |
4322 | { | |
4323 | struct resource_statistics resource_statistics; | |
4324 | struct drbd_genlmsghdr *dh; | |
4325 | bool multicast = false; | |
4326 | int err; | |
4327 | ||
4328 | if (!skb) { | |
4329 | seq = atomic_inc_return(¬ify_genl_seq); | |
4330 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4331 | err = -ENOMEM; | |
4332 | if (!skb) | |
4333 | goto failed; | |
4334 | multicast = true; | |
4335 | } | |
4336 | ||
4337 | err = -EMSGSIZE; | |
4338 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_RESOURCE_STATE); | |
4339 | if (!dh) | |
4340 | goto nla_put_failure; | |
4341 | dh->minor = -1U; | |
4342 | dh->ret_code = NO_ERROR; | |
4343 | if (nla_put_drbd_cfg_context(skb, resource, NULL, NULL) || | |
4344 | nla_put_notification_header(skb, type) || | |
4345 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4346 | resource_info_to_skb(skb, resource_info, true))) | |
4347 | goto nla_put_failure; | |
4348 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
4349 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
4350 | if (err) | |
4351 | goto nla_put_failure; | |
4352 | genlmsg_end(skb, dh); | |
4353 | if (multicast) { | |
4354 | err = drbd_genl_multicast_events(skb, 0); | |
4355 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4356 | if (err && err != -ESRCH) | |
4357 | goto failed; | |
4358 | } | |
4359 | return; | |
4360 | ||
4361 | nla_put_failure: | |
4362 | nlmsg_free(skb); | |
4363 | failed: | |
4364 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4365 | err, seq); | |
4366 | } | |
4367 | ||
4368 | void notify_device_state(struct sk_buff *skb, | |
4369 | unsigned int seq, | |
4370 | struct drbd_device *device, | |
4371 | struct device_info *device_info, | |
4372 | enum drbd_notification_type type) | |
4373 | { | |
4374 | struct device_statistics device_statistics; | |
4375 | struct drbd_genlmsghdr *dh; | |
4376 | bool multicast = false; | |
4377 | int err; | |
4378 | ||
4379 | if (!skb) { | |
4380 | seq = atomic_inc_return(¬ify_genl_seq); | |
4381 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4382 | err = -ENOMEM; | |
4383 | if (!skb) | |
4384 | goto failed; | |
4385 | multicast = true; | |
4386 | } | |
4387 | ||
4388 | err = -EMSGSIZE; | |
4389 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_DEVICE_STATE); | |
4390 | if (!dh) | |
4391 | goto nla_put_failure; | |
4392 | dh->minor = device->minor; | |
4393 | dh->ret_code = NO_ERROR; | |
4394 | if (nla_put_drbd_cfg_context(skb, device->resource, NULL, device) || | |
4395 | nla_put_notification_header(skb, type) || | |
4396 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4397 | device_info_to_skb(skb, device_info, true))) | |
4398 | goto nla_put_failure; | |
4399 | device_to_statistics(&device_statistics, device); | |
4400 | device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
4401 | genlmsg_end(skb, dh); | |
4402 | if (multicast) { | |
4403 | err = drbd_genl_multicast_events(skb, 0); | |
4404 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4405 | if (err && err != -ESRCH) | |
4406 | goto failed; | |
4407 | } | |
4408 | return; | |
4409 | ||
4410 | nla_put_failure: | |
4411 | nlmsg_free(skb); | |
4412 | failed: | |
4413 | drbd_err(device, "Error %d while broadcasting event. Event seq:%u\n", | |
4414 | err, seq); | |
4415 | } | |
4416 | ||
4417 | void notify_connection_state(struct sk_buff *skb, | |
4418 | unsigned int seq, | |
4419 | struct drbd_connection *connection, | |
4420 | struct connection_info *connection_info, | |
4421 | enum drbd_notification_type type) | |
4422 | { | |
4423 | struct connection_statistics connection_statistics; | |
4424 | struct drbd_genlmsghdr *dh; | |
4425 | bool multicast = false; | |
4426 | int err; | |
4427 | ||
4428 | if (!skb) { | |
4429 | seq = atomic_inc_return(¬ify_genl_seq); | |
4430 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4431 | err = -ENOMEM; | |
4432 | if (!skb) | |
4433 | goto failed; | |
4434 | multicast = true; | |
4435 | } | |
4436 | ||
4437 | err = -EMSGSIZE; | |
4438 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_CONNECTION_STATE); | |
4439 | if (!dh) | |
4440 | goto nla_put_failure; | |
4441 | dh->minor = -1U; | |
4442 | dh->ret_code = NO_ERROR; | |
4443 | if (nla_put_drbd_cfg_context(skb, connection->resource, connection, NULL) || | |
4444 | nla_put_notification_header(skb, type) || | |
4445 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4446 | connection_info_to_skb(skb, connection_info, true))) | |
4447 | goto nla_put_failure; | |
4448 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
4449 | connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
4450 | genlmsg_end(skb, dh); | |
4451 | if (multicast) { | |
4452 | err = drbd_genl_multicast_events(skb, 0); | |
4453 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4454 | if (err && err != -ESRCH) | |
4455 | goto failed; | |
4456 | } | |
4457 | return; | |
4458 | ||
4459 | nla_put_failure: | |
4460 | nlmsg_free(skb); | |
4461 | failed: | |
4462 | drbd_err(connection, "Error %d while broadcasting event. Event seq:%u\n", | |
4463 | err, seq); | |
4464 | } | |
4465 | ||
4466 | void notify_peer_device_state(struct sk_buff *skb, | |
4467 | unsigned int seq, | |
4468 | struct drbd_peer_device *peer_device, | |
4469 | struct peer_device_info *peer_device_info, | |
4470 | enum drbd_notification_type type) | |
4471 | { | |
4472 | struct peer_device_statistics peer_device_statistics; | |
4473 | struct drbd_resource *resource = peer_device->device->resource; | |
4474 | struct drbd_genlmsghdr *dh; | |
4475 | bool multicast = false; | |
4476 | int err; | |
4477 | ||
4478 | if (!skb) { | |
4479 | seq = atomic_inc_return(¬ify_genl_seq); | |
4480 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4481 | err = -ENOMEM; | |
4482 | if (!skb) | |
4483 | goto failed; | |
4484 | multicast = true; | |
4485 | } | |
4486 | ||
4487 | err = -EMSGSIZE; | |
4488 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_PEER_DEVICE_STATE); | |
4489 | if (!dh) | |
4490 | goto nla_put_failure; | |
4491 | dh->minor = -1U; | |
4492 | dh->ret_code = NO_ERROR; | |
4493 | if (nla_put_drbd_cfg_context(skb, resource, peer_device->connection, peer_device->device) || | |
4494 | nla_put_notification_header(skb, type) || | |
4495 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4496 | peer_device_info_to_skb(skb, peer_device_info, true))) | |
4497 | goto nla_put_failure; | |
4498 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
4499 | peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
4500 | genlmsg_end(skb, dh); | |
4501 | if (multicast) { | |
4502 | err = drbd_genl_multicast_events(skb, 0); | |
4503 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4504 | if (err && err != -ESRCH) | |
4505 | goto failed; | |
4506 | } | |
4507 | return; | |
4508 | ||
4509 | nla_put_failure: | |
4510 | nlmsg_free(skb); | |
4511 | failed: | |
4512 | drbd_err(peer_device, "Error %d while broadcasting event. Event seq:%u\n", | |
4513 | err, seq); | |
4514 | } | |
4515 | ||
4516 | void notify_helper(enum drbd_notification_type type, | |
4517 | struct drbd_device *device, struct drbd_connection *connection, | |
4518 | const char *name, int status) | |
4519 | { | |
4520 | struct drbd_resource *resource = device ? device->resource : connection->resource; | |
4521 | struct drbd_helper_info helper_info; | |
4522 | unsigned int seq = atomic_inc_return(¬ify_genl_seq); | |
4523 | struct sk_buff *skb = NULL; | |
4524 | struct drbd_genlmsghdr *dh; | |
4525 | int err; | |
4526 | ||
4527 | strlcpy(helper_info.helper_name, name, sizeof(helper_info.helper_name)); | |
4528 | helper_info.helper_name_len = min(strlen(name), sizeof(helper_info.helper_name)); | |
4529 | helper_info.helper_status = status; | |
4530 | ||
4531 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4532 | err = -ENOMEM; | |
4533 | if (!skb) | |
4534 | goto fail; | |
4535 | ||
4536 | err = -EMSGSIZE; | |
4537 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_HELPER); | |
4538 | if (!dh) | |
4539 | goto fail; | |
4540 | dh->minor = device ? device->minor : -1; | |
4541 | dh->ret_code = NO_ERROR; | |
4542 | mutex_lock(¬ification_mutex); | |
4543 | if (nla_put_drbd_cfg_context(skb, resource, connection, device) || | |
4544 | nla_put_notification_header(skb, type) || | |
4545 | drbd_helper_info_to_skb(skb, &helper_info, true)) | |
4546 | goto unlock_fail; | |
4547 | genlmsg_end(skb, dh); | |
4548 | err = drbd_genl_multicast_events(skb, 0); | |
4549 | skb = NULL; | |
4550 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4551 | if (err && err != -ESRCH) | |
4552 | goto unlock_fail; | |
4553 | mutex_unlock(¬ification_mutex); | |
4554 | return; | |
4555 | ||
4556 | unlock_fail: | |
4557 | mutex_unlock(¬ification_mutex); | |
4558 | fail: | |
4559 | nlmsg_free(skb); | |
4560 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4561 | err, seq); | |
4562 | } | |
4563 | ||
4564 | static void notify_initial_state_done(struct sk_buff *skb, unsigned int seq) | |
4565 | { | |
4566 | struct drbd_genlmsghdr *dh; | |
4567 | int err; | |
4568 | ||
4569 | err = -EMSGSIZE; | |
4570 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_INITIAL_STATE_DONE); | |
4571 | if (!dh) | |
4572 | goto nla_put_failure; | |
4573 | dh->minor = -1U; | |
4574 | dh->ret_code = NO_ERROR; | |
4575 | if (nla_put_notification_header(skb, NOTIFY_EXISTS)) | |
4576 | goto nla_put_failure; | |
4577 | genlmsg_end(skb, dh); | |
4578 | return; | |
4579 | ||
4580 | nla_put_failure: | |
4581 | nlmsg_free(skb); | |
4582 | pr_err("Error %d sending event. Event seq:%u\n", err, seq); | |
4583 | } | |
4584 | ||
4585 | static void free_state_changes(struct list_head *list) | |
4586 | { | |
4587 | while (!list_empty(list)) { | |
4588 | struct drbd_state_change *state_change = | |
4589 | list_first_entry(list, struct drbd_state_change, list); | |
4590 | list_del(&state_change->list); | |
4591 | forget_state_change(state_change); | |
4592 | } | |
4593 | } | |
4594 | ||
4595 | static unsigned int notifications_for_state_change(struct drbd_state_change *state_change) | |
4596 | { | |
4597 | return 1 + | |
4598 | state_change->n_connections + | |
4599 | state_change->n_devices + | |
4600 | state_change->n_devices * state_change->n_connections; | |
4601 | } | |
4602 | ||
4603 | static int get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4604 | { | |
4605 | struct drbd_state_change *state_change = (struct drbd_state_change *)cb->args[0]; | |
4606 | unsigned int seq = cb->args[2]; | |
4607 | unsigned int n; | |
4608 | enum drbd_notification_type flags = 0; | |
4609 | ||
4610 | /* There is no need for taking notification_mutex here: it doesn't | |
4611 | matter if the initial state events mix with later state chage | |
4612 | events; we can always tell the events apart by the NOTIFY_EXISTS | |
4613 | flag. */ | |
4614 | ||
4615 | cb->args[5]--; | |
4616 | if (cb->args[5] == 1) { | |
4617 | notify_initial_state_done(skb, seq); | |
4618 | goto out; | |
4619 | } | |
4620 | n = cb->args[4]++; | |
4621 | if (cb->args[4] < cb->args[3]) | |
4622 | flags |= NOTIFY_CONTINUES; | |
4623 | if (n < 1) { | |
4624 | notify_resource_state_change(skb, seq, state_change->resource, | |
4625 | NOTIFY_EXISTS | flags); | |
4626 | goto next; | |
4627 | } | |
4628 | n--; | |
4629 | if (n < state_change->n_connections) { | |
4630 | notify_connection_state_change(skb, seq, &state_change->connections[n], | |
4631 | NOTIFY_EXISTS | flags); | |
4632 | goto next; | |
4633 | } | |
4634 | n -= state_change->n_connections; | |
4635 | if (n < state_change->n_devices) { | |
4636 | notify_device_state_change(skb, seq, &state_change->devices[n], | |
4637 | NOTIFY_EXISTS | flags); | |
4638 | goto next; | |
4639 | } | |
4640 | n -= state_change->n_devices; | |
4641 | if (n < state_change->n_devices * state_change->n_connections) { | |
4642 | notify_peer_device_state_change(skb, seq, &state_change->peer_devices[n], | |
4643 | NOTIFY_EXISTS | flags); | |
4644 | goto next; | |
4645 | } | |
4646 | ||
4647 | next: | |
4648 | if (cb->args[4] == cb->args[3]) { | |
4649 | struct drbd_state_change *next_state_change = | |
4650 | list_entry(state_change->list.next, | |
4651 | struct drbd_state_change, list); | |
4652 | cb->args[0] = (long)next_state_change; | |
4653 | cb->args[3] = notifications_for_state_change(next_state_change); | |
4654 | cb->args[4] = 0; | |
4655 | } | |
4656 | out: | |
4657 | return skb->len; | |
4658 | } | |
4659 | ||
4660 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4661 | { | |
4662 | struct drbd_resource *resource; | |
4663 | LIST_HEAD(head); | |
4664 | ||
4665 | if (cb->args[5] >= 1) { | |
4666 | if (cb->args[5] > 1) | |
4667 | return get_initial_state(skb, cb); | |
4668 | if (cb->args[0]) { | |
4669 | struct drbd_state_change *state_change = | |
4670 | (struct drbd_state_change *)cb->args[0]; | |
4671 | ||
4672 | /* connect list to head */ | |
4673 | list_add(&head, &state_change->list); | |
4674 | free_state_changes(&head); | |
4675 | } | |
4676 | return 0; | |
4677 | } | |
4678 | ||
4679 | cb->args[5] = 2; /* number of iterations */ | |
4680 | mutex_lock(&resources_mutex); | |
4681 | for_each_resource(resource, &drbd_resources) { | |
4682 | struct drbd_state_change *state_change; | |
4683 | ||
4684 | state_change = remember_old_state(resource, GFP_KERNEL); | |
4685 | if (!state_change) { | |
4686 | if (!list_empty(&head)) | |
4687 | free_state_changes(&head); | |
4688 | mutex_unlock(&resources_mutex); | |
4689 | return -ENOMEM; | |
4690 | } | |
4691 | copy_old_to_new_state_change(state_change); | |
4692 | list_add_tail(&state_change->list, &head); | |
4693 | cb->args[5] += notifications_for_state_change(state_change); | |
4694 | } | |
4695 | mutex_unlock(&resources_mutex); | |
4696 | ||
4697 | if (!list_empty(&head)) { | |
4698 | struct drbd_state_change *state_change = | |
4699 | list_entry(head.next, struct drbd_state_change, list); | |
4700 | cb->args[0] = (long)state_change; | |
4701 | cb->args[3] = notifications_for_state_change(state_change); | |
4702 | list_del(&head); /* detach list from head */ | |
4703 | } | |
4704 | ||
4705 | cb->args[2] = cb->nlh->nlmsg_seq; | |
4706 | return get_initial_state(skb, cb); | |
4707 | } |