]>
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 }; |
0982368b | 346 | char mb[14]; |
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 | |
0982368b | 355 | snprintf(mb, 14, "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 | ||
26a96110 | 390 | 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 PR |
444 | |
445 | return fp; | |
446 | } | |
447 | ||
7bd000cb LE |
448 | static bool resource_is_supended(struct drbd_resource *resource) |
449 | { | |
450 | return resource->susp || resource->susp_fen || resource->susp_nod; | |
451 | } | |
452 | ||
bde89a9e | 453 | bool conn_try_outdate_peer(struct drbd_connection *connection) |
b411b363 | 454 | { |
7bd000cb | 455 | struct drbd_resource * const resource = connection->resource; |
28e448bb | 456 | unsigned int connect_cnt; |
cb703454 PR |
457 | union drbd_state mask = { }; |
458 | union drbd_state val = { }; | |
459 | enum drbd_fencing_p fp; | |
b411b363 PR |
460 | char *ex_to_string; |
461 | int r; | |
b411b363 | 462 | |
7bd000cb | 463 | spin_lock_irq(&resource->req_lock); |
bde89a9e | 464 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
1ec861eb | 465 | drbd_err(connection, "Expected cstate < C_WF_REPORT_PARAMS\n"); |
7bd000cb | 466 | spin_unlock_irq(&resource->req_lock); |
cb703454 PR |
467 | return false; |
468 | } | |
b411b363 | 469 | |
bde89a9e | 470 | connect_cnt = connection->connect_cnt; |
7bd000cb | 471 | spin_unlock_irq(&resource->req_lock); |
28e448bb | 472 | |
bde89a9e | 473 | fp = highest_fencing_policy(connection); |
cb703454 PR |
474 | switch (fp) { |
475 | case FP_NOT_AVAIL: | |
1ec861eb | 476 | drbd_warn(connection, "Not fencing peer, I'm not even Consistent myself.\n"); |
7bd000cb LE |
477 | spin_lock_irq(&resource->req_lock); |
478 | if (connection->cstate < C_WF_REPORT_PARAMS) { | |
479 | _conn_request_state(connection, | |
480 | (union drbd_state) { { .susp_fen = 1 } }, | |
481 | (union drbd_state) { { .susp_fen = 0 } }, | |
482 | CS_VERBOSE | CS_HARD | CS_DC_SUSP); | |
483 | /* We are no longer suspended due to the fencing policy. | |
484 | * We may still be suspended due to the on-no-data-accessible policy. | |
485 | * If that was OND_IO_ERROR, fail pending requests. */ | |
486 | if (!resource_is_supended(resource)) | |
487 | _tl_restart(connection, CONNECTION_LOST_WHILE_PENDING); | |
488 | } | |
489 | /* Else: in case we raced with a connection handshake, | |
490 | * let the handshake figure out if we maybe can RESEND, | |
491 | * and do not resume/fail pending requests here. | |
492 | * Worst case is we stay suspended for now, which may be | |
493 | * resolved by either re-establishing the replication link, or | |
494 | * the next link failure, or eventually the administrator. */ | |
495 | spin_unlock_irq(&resource->req_lock); | |
496 | return false; | |
497 | ||
cb703454 PR |
498 | case FP_DONT_CARE: |
499 | return true; | |
500 | default: ; | |
b411b363 PR |
501 | } |
502 | ||
bde89a9e | 503 | r = conn_khelper(connection, "fence-peer"); |
b411b363 PR |
504 | |
505 | switch ((r>>8) & 0xff) { | |
506 | case 3: /* peer is inconsistent */ | |
507 | ex_to_string = "peer is inconsistent or worse"; | |
cb703454 PR |
508 | mask.pdsk = D_MASK; |
509 | val.pdsk = D_INCONSISTENT; | |
b411b363 PR |
510 | break; |
511 | case 4: /* peer got outdated, or was already outdated */ | |
512 | ex_to_string = "peer was fenced"; | |
cb703454 PR |
513 | mask.pdsk = D_MASK; |
514 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
515 | break; |
516 | case 5: /* peer was down */ | |
bde89a9e | 517 | if (conn_highest_disk(connection) == D_UP_TO_DATE) { |
b411b363 PR |
518 | /* we will(have) create(d) a new UUID anyways... */ |
519 | ex_to_string = "peer is unreachable, assumed to be dead"; | |
cb703454 PR |
520 | mask.pdsk = D_MASK; |
521 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
522 | } else { |
523 | ex_to_string = "peer unreachable, doing nothing since disk != UpToDate"; | |
b411b363 PR |
524 | } |
525 | break; | |
526 | case 6: /* Peer is primary, voluntarily outdate myself. | |
527 | * This is useful when an unconnected R_SECONDARY is asked to | |
528 | * become R_PRIMARY, but finds the other peer being active. */ | |
529 | ex_to_string = "peer is active"; | |
1ec861eb | 530 | drbd_warn(connection, "Peer is primary, outdating myself.\n"); |
cb703454 PR |
531 | mask.disk = D_MASK; |
532 | val.disk = D_OUTDATED; | |
b411b363 PR |
533 | break; |
534 | case 7: | |
535 | if (fp != FP_STONITH) | |
1ec861eb | 536 | drbd_err(connection, "fence-peer() = 7 && fencing != Stonith !!!\n"); |
b411b363 | 537 | ex_to_string = "peer was stonithed"; |
cb703454 PR |
538 | mask.pdsk = D_MASK; |
539 | val.pdsk = D_OUTDATED; | |
b411b363 PR |
540 | break; |
541 | default: | |
542 | /* The script is broken ... */ | |
1ec861eb | 543 | drbd_err(connection, "fence-peer helper broken, returned %d\n", (r>>8)&0xff); |
cb703454 | 544 | return false; /* Eventually leave IO frozen */ |
b411b363 PR |
545 | } |
546 | ||
1ec861eb | 547 | drbd_info(connection, "fence-peer helper returned %d (%s)\n", |
cb703454 | 548 | (r>>8) & 0xff, ex_to_string); |
fb22c402 | 549 | |
cb703454 | 550 | /* Not using |
bde89a9e | 551 | conn_request_state(connection, mask, val, CS_VERBOSE); |
cb703454 PR |
552 | here, because we might were able to re-establish the connection in the |
553 | meantime. */ | |
7bd000cb | 554 | spin_lock_irq(&resource->req_lock); |
bde89a9e AG |
555 | if (connection->cstate < C_WF_REPORT_PARAMS && !test_bit(STATE_SENT, &connection->flags)) { |
556 | if (connection->connect_cnt != connect_cnt) | |
28e448bb PR |
557 | /* In case the connection was established and droped |
558 | while the fence-peer handler was running, ignore it */ | |
1ec861eb | 559 | drbd_info(connection, "Ignoring fence-peer exit code\n"); |
28e448bb | 560 | else |
bde89a9e | 561 | _conn_request_state(connection, mask, val, CS_VERBOSE); |
28e448bb | 562 | } |
7bd000cb | 563 | spin_unlock_irq(&resource->req_lock); |
cb703454 | 564 | |
bde89a9e | 565 | return conn_highest_pdsk(connection) <= D_OUTDATED; |
b411b363 PR |
566 | } |
567 | ||
87f7be4c PR |
568 | static int _try_outdate_peer_async(void *data) |
569 | { | |
bde89a9e | 570 | struct drbd_connection *connection = (struct drbd_connection *)data; |
87f7be4c | 571 | |
bde89a9e | 572 | conn_try_outdate_peer(connection); |
87f7be4c | 573 | |
05a10ec7 | 574 | kref_put(&connection->kref, drbd_destroy_connection); |
87f7be4c PR |
575 | return 0; |
576 | } | |
577 | ||
bde89a9e | 578 | void conn_try_outdate_peer_async(struct drbd_connection *connection) |
87f7be4c PR |
579 | { |
580 | struct task_struct *opa; | |
581 | ||
bde89a9e | 582 | kref_get(&connection->kref); |
bbc1c5e8 LE |
583 | /* We may just have force_sig()'ed this thread |
584 | * to get it out of some blocking network function. | |
585 | * Clear signals; otherwise kthread_run(), which internally uses | |
586 | * wait_on_completion_killable(), will mistake our pending signal | |
587 | * for a new fatal signal and fail. */ | |
588 | flush_signals(current); | |
bde89a9e | 589 | opa = kthread_run(_try_outdate_peer_async, connection, "drbd_async_h"); |
9dc9fbb3 | 590 | if (IS_ERR(opa)) { |
1ec861eb | 591 | drbd_err(connection, "out of mem, failed to invoke fence-peer helper\n"); |
05a10ec7 | 592 | kref_put(&connection->kref, drbd_destroy_connection); |
9dc9fbb3 | 593 | } |
87f7be4c | 594 | } |
b411b363 | 595 | |
bf885f8a | 596 | enum drbd_state_rv |
44a4d551 | 597 | drbd_set_role(struct drbd_device *const device, enum drbd_role new_role, int force) |
b411b363 | 598 | { |
44a4d551 LE |
599 | struct drbd_peer_device *const peer_device = first_peer_device(device); |
600 | struct drbd_connection *const connection = peer_device ? peer_device->connection : NULL; | |
b411b363 | 601 | const int max_tries = 4; |
bf885f8a | 602 | enum drbd_state_rv rv = SS_UNKNOWN_ERROR; |
44ed167d | 603 | struct net_conf *nc; |
b411b363 PR |
604 | int try = 0; |
605 | int forced = 0; | |
606 | union drbd_state mask, val; | |
b411b363 | 607 | |
b6f85ef9 AG |
608 | if (new_role == R_PRIMARY) { |
609 | struct drbd_connection *connection; | |
610 | ||
611 | /* Detect dead peers as soon as possible. */ | |
612 | ||
613 | rcu_read_lock(); | |
614 | for_each_connection(connection, device->resource) | |
615 | request_ping(connection); | |
616 | rcu_read_unlock(); | |
617 | } | |
b411b363 | 618 | |
b30ab791 | 619 | mutex_lock(device->state_mutex); |
b411b363 PR |
620 | |
621 | mask.i = 0; mask.role = R_MASK; | |
622 | val.i = 0; val.role = new_role; | |
623 | ||
624 | while (try++ < max_tries) { | |
a8821531 | 625 | rv = _drbd_request_state_holding_state_mutex(device, mask, val, CS_WAIT_COMPLETE); |
b411b363 PR |
626 | |
627 | /* in case we first succeeded to outdate, | |
628 | * but now suddenly could establish a connection */ | |
bf885f8a | 629 | if (rv == SS_CW_FAILED_BY_PEER && mask.pdsk != 0) { |
b411b363 PR |
630 | val.pdsk = 0; |
631 | mask.pdsk = 0; | |
632 | continue; | |
633 | } | |
634 | ||
bf885f8a | 635 | if (rv == SS_NO_UP_TO_DATE_DISK && force && |
b30ab791 AG |
636 | (device->state.disk < D_UP_TO_DATE && |
637 | device->state.disk >= D_INCONSISTENT)) { | |
b411b363 PR |
638 | mask.disk = D_MASK; |
639 | val.disk = D_UP_TO_DATE; | |
640 | forced = 1; | |
641 | continue; | |
642 | } | |
643 | ||
bf885f8a | 644 | if (rv == SS_NO_UP_TO_DATE_DISK && |
b30ab791 | 645 | device->state.disk == D_CONSISTENT && mask.pdsk == 0) { |
0b0ba1ef | 646 | D_ASSERT(device, device->state.pdsk == D_UNKNOWN); |
b411b363 | 647 | |
44a4d551 | 648 | if (conn_try_outdate_peer(connection)) { |
b411b363 PR |
649 | val.disk = D_UP_TO_DATE; |
650 | mask.disk = D_MASK; | |
651 | } | |
b411b363 PR |
652 | continue; |
653 | } | |
654 | ||
bf885f8a | 655 | if (rv == SS_NOTHING_TO_DO) |
3b98c0c2 | 656 | goto out; |
bf885f8a | 657 | if (rv == SS_PRIMARY_NOP && mask.pdsk == 0) { |
44a4d551 | 658 | if (!conn_try_outdate_peer(connection) && force) { |
d0180171 | 659 | drbd_warn(device, "Forced into split brain situation!\n"); |
cb703454 PR |
660 | mask.pdsk = D_MASK; |
661 | val.pdsk = D_OUTDATED; | |
b411b363 | 662 | |
cb703454 | 663 | } |
b411b363 PR |
664 | continue; |
665 | } | |
bf885f8a | 666 | if (rv == SS_TWO_PRIMARIES) { |
b411b363 PR |
667 | /* Maybe the peer is detected as dead very soon... |
668 | retry at most once more in this case. */ | |
44ed167d PR |
669 | int timeo; |
670 | rcu_read_lock(); | |
44a4d551 | 671 | nc = rcu_dereference(connection->net_conf); |
44ed167d PR |
672 | timeo = nc ? (nc->ping_timeo + 1) * HZ / 10 : 1; |
673 | rcu_read_unlock(); | |
674 | schedule_timeout_interruptible(timeo); | |
b411b363 PR |
675 | if (try < max_tries) |
676 | try = max_tries - 1; | |
677 | continue; | |
678 | } | |
bf885f8a | 679 | if (rv < SS_SUCCESS) { |
b30ab791 | 680 | rv = _drbd_request_state(device, mask, val, |
b411b363 | 681 | CS_VERBOSE + CS_WAIT_COMPLETE); |
bf885f8a | 682 | if (rv < SS_SUCCESS) |
3b98c0c2 | 683 | goto out; |
b411b363 PR |
684 | } |
685 | break; | |
686 | } | |
687 | ||
bf885f8a | 688 | if (rv < SS_SUCCESS) |
3b98c0c2 | 689 | goto out; |
b411b363 PR |
690 | |
691 | if (forced) | |
d0180171 | 692 | drbd_warn(device, "Forced to consider local data as UpToDate!\n"); |
b411b363 PR |
693 | |
694 | /* Wait until nothing is on the fly :) */ | |
b30ab791 | 695 | wait_event(device->misc_wait, atomic_read(&device->ap_pending_cnt) == 0); |
b411b363 | 696 | |
b6dd1a89 LE |
697 | /* FIXME also wait for all pending P_BARRIER_ACK? */ |
698 | ||
b411b363 | 699 | if (new_role == R_SECONDARY) { |
b30ab791 AG |
700 | if (get_ldev(device)) { |
701 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; | |
702 | put_ldev(device); | |
b411b363 PR |
703 | } |
704 | } else { | |
66ce6dbc | 705 | mutex_lock(&device->resource->conf_update); |
44a4d551 | 706 | nc = connection->net_conf; |
44ed167d | 707 | if (nc) |
6139f60d | 708 | nc->discard_my_data = 0; /* without copy; single bit op is atomic */ |
66ce6dbc | 709 | mutex_unlock(&device->resource->conf_update); |
91fd4dad | 710 | |
b30ab791 AG |
711 | if (get_ldev(device)) { |
712 | if (((device->state.conn < C_CONNECTED || | |
713 | device->state.pdsk <= D_FAILED) | |
714 | && device->ldev->md.uuid[UI_BITMAP] == 0) || forced) | |
715 | drbd_uuid_new_current(device); | |
b411b363 | 716 | |
b30ab791 AG |
717 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; |
718 | put_ldev(device); | |
b411b363 PR |
719 | } |
720 | } | |
721 | ||
19f843aa LE |
722 | /* writeout of activity log covered areas of the bitmap |
723 | * to stable storage done in after state change already */ | |
b411b363 | 724 | |
b30ab791 | 725 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
b411b363 PR |
726 | /* if this was forced, we should consider sync */ |
727 | if (forced) | |
44a4d551 LE |
728 | drbd_send_uuids(peer_device); |
729 | drbd_send_current_state(peer_device); | |
b411b363 PR |
730 | } |
731 | ||
b30ab791 | 732 | drbd_md_sync(device); |
720979fb | 733 | set_disk_ro(device->vdisk, new_role == R_SECONDARY); |
b30ab791 | 734 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
3b98c0c2 | 735 | out: |
b30ab791 | 736 | mutex_unlock(device->state_mutex); |
bf885f8a | 737 | return rv; |
b411b363 PR |
738 | } |
739 | ||
3b98c0c2 | 740 | static const char *from_attrs_err_to_txt(int err) |
ef50a3e3 | 741 | { |
3b98c0c2 LE |
742 | return err == -ENOMSG ? "required attribute missing" : |
743 | err == -EOPNOTSUPP ? "unknown mandatory attribute" : | |
f399002e | 744 | err == -EEXIST ? "can not change invariant setting" : |
3b98c0c2 | 745 | "invalid attribute value"; |
ef50a3e3 | 746 | } |
b411b363 | 747 | |
3b98c0c2 | 748 | int drbd_adm_set_role(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 749 | { |
a910b123 | 750 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
751 | struct set_role_parms parms; |
752 | int err; | |
753 | enum drbd_ret_code retcode; | |
b411b363 | 754 | |
a910b123 | 755 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
756 | if (!adm_ctx.reply_skb) |
757 | return retcode; | |
758 | if (retcode != NO_ERROR) | |
759 | goto out; | |
b411b363 | 760 | |
3b98c0c2 LE |
761 | memset(&parms, 0, sizeof(parms)); |
762 | if (info->attrs[DRBD_NLA_SET_ROLE_PARMS]) { | |
f399002e | 763 | err = set_role_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
764 | if (err) { |
765 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 766 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
767 | goto out; |
768 | } | |
769 | } | |
9e276872 LE |
770 | genl_unlock(); |
771 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 772 | |
3b98c0c2 | 773 | if (info->genlhdr->cmd == DRBD_ADM_PRIMARY) |
b30ab791 | 774 | retcode = drbd_set_role(adm_ctx.device, R_PRIMARY, parms.assume_uptodate); |
3b98c0c2 | 775 | else |
b30ab791 | 776 | retcode = drbd_set_role(adm_ctx.device, R_SECONDARY, 0); |
9e276872 LE |
777 | |
778 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
779 | genl_lock(); | |
3b98c0c2 | 780 | out: |
a910b123 | 781 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
782 | return 0; |
783 | } | |
784 | ||
ae8bf312 LE |
785 | /* Initializes the md.*_offset members, so we are able to find |
786 | * the on disk meta data. | |
787 | * | |
788 | * We currently have two possible layouts: | |
789 | * external: | |
790 | * |----------- md_size_sect ------------------| | |
791 | * [ 4k superblock ][ activity log ][ Bitmap ] | |
792 | * | al_offset == 8 | | |
793 | * | bm_offset = al_offset + X | | |
794 | * ==> bitmap sectors = md_size_sect - bm_offset | |
795 | * | |
796 | * internal: | |
797 | * |----------- md_size_sect ------------------| | |
798 | * [data.....][ Bitmap ][ activity log ][ 4k superblock ] | |
799 | * | al_offset < 0 | | |
800 | * | bm_offset = al_offset - Y | | |
801 | * ==> bitmap sectors = Y = al_offset - bm_offset | |
802 | * | |
803 | * Activity log size used to be fixed 32kB, | |
804 | * but is about to become configurable. | |
805 | */ | |
b30ab791 | 806 | static void drbd_md_set_sector_offsets(struct drbd_device *device, |
b411b363 PR |
807 | struct drbd_backing_dev *bdev) |
808 | { | |
809 | sector_t md_size_sect = 0; | |
c04ccaa6 | 810 | unsigned int al_size_sect = bdev->md.al_size_4k * 8; |
daeda1cc | 811 | |
3a4d4eb3 LE |
812 | bdev->md.md_offset = drbd_md_ss(bdev); |
813 | ||
68e41a43 | 814 | switch (bdev->md.meta_dev_idx) { |
b411b363 PR |
815 | default: |
816 | /* v07 style fixed size indexed meta data */ | |
ae8bf312 | 817 | bdev->md.md_size_sect = MD_128MB_SECT; |
ae8bf312 LE |
818 | bdev->md.al_offset = MD_4kB_SECT; |
819 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
820 | break; |
821 | case DRBD_MD_INDEX_FLEX_EXT: | |
822 | /* just occupy the full device; unit: sectors */ | |
823 | bdev->md.md_size_sect = drbd_get_capacity(bdev->md_bdev); | |
ae8bf312 LE |
824 | bdev->md.al_offset = MD_4kB_SECT; |
825 | bdev->md.bm_offset = MD_4kB_SECT + al_size_sect; | |
b411b363 PR |
826 | break; |
827 | case DRBD_MD_INDEX_INTERNAL: | |
828 | case DRBD_MD_INDEX_FLEX_INT: | |
b411b363 | 829 | /* al size is still fixed */ |
ae8bf312 | 830 | bdev->md.al_offset = -al_size_sect; |
b411b363 PR |
831 | /* we need (slightly less than) ~ this much bitmap sectors: */ |
832 | md_size_sect = drbd_get_capacity(bdev->backing_bdev); | |
833 | md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT); | |
834 | md_size_sect = BM_SECT_TO_EXT(md_size_sect); | |
835 | md_size_sect = ALIGN(md_size_sect, 8); | |
836 | ||
837 | /* plus the "drbd meta data super block", | |
838 | * and the activity log; */ | |
ae8bf312 | 839 | md_size_sect += MD_4kB_SECT + al_size_sect; |
b411b363 PR |
840 | |
841 | bdev->md.md_size_sect = md_size_sect; | |
842 | /* bitmap offset is adjusted by 'super' block size */ | |
ae8bf312 | 843 | bdev->md.bm_offset = -md_size_sect + MD_4kB_SECT; |
b411b363 PR |
844 | break; |
845 | } | |
846 | } | |
847 | ||
4b0715f0 | 848 | /* input size is expected to be in KB */ |
b411b363 PR |
849 | char *ppsize(char *buf, unsigned long long size) |
850 | { | |
4b0715f0 LE |
851 | /* Needs 9 bytes at max including trailing NUL: |
852 | * -1ULL ==> "16384 EB" */ | |
b411b363 PR |
853 | static char units[] = { 'K', 'M', 'G', 'T', 'P', 'E' }; |
854 | int base = 0; | |
4b0715f0 | 855 | while (size >= 10000 && base < sizeof(units)-1) { |
b411b363 PR |
856 | /* shift + round */ |
857 | size = (size >> 10) + !!(size & (1<<9)); | |
858 | base++; | |
859 | } | |
4b0715f0 | 860 | sprintf(buf, "%u %cB", (unsigned)size, units[base]); |
b411b363 PR |
861 | |
862 | return buf; | |
863 | } | |
864 | ||
865 | /* there is still a theoretical deadlock when called from receiver | |
866 | * on an D_INCONSISTENT R_PRIMARY: | |
867 | * remote READ does inc_ap_bio, receiver would need to receive answer | |
868 | * packet from remote to dec_ap_bio again. | |
869 | * receiver receive_sizes(), comes here, | |
870 | * waits for ap_bio_cnt == 0. -> deadlock. | |
871 | * but this cannot happen, actually, because: | |
872 | * R_PRIMARY D_INCONSISTENT, and peer's disk is unreachable | |
873 | * (not connected, or bad/no disk on peer): | |
874 | * see drbd_fail_request_early, ap_bio_cnt is zero. | |
875 | * R_PRIMARY D_INCONSISTENT, and C_SYNC_TARGET: | |
876 | * peer may not initiate a resize. | |
877 | */ | |
3b98c0c2 LE |
878 | /* Note these are not to be confused with |
879 | * drbd_adm_suspend_io/drbd_adm_resume_io, | |
880 | * which are (sub) state changes triggered by admin (drbdsetup), | |
881 | * and can be long lived. | |
b30ab791 | 882 | * This changes an device->flag, is triggered by drbd internals, |
3b98c0c2 | 883 | * and should be short-lived. */ |
7dbb4386 PR |
884 | /* It needs to be a counter, since multiple threads might |
885 | independently suspend and resume IO. */ | |
b30ab791 | 886 | void drbd_suspend_io(struct drbd_device *device) |
b411b363 | 887 | { |
7dbb4386 | 888 | atomic_inc(&device->suspend_cnt); |
b30ab791 | 889 | if (drbd_suspended(device)) |
265be2d0 | 890 | return; |
b30ab791 | 891 | wait_event(device->misc_wait, !atomic_read(&device->ap_bio_cnt)); |
b411b363 PR |
892 | } |
893 | ||
b30ab791 | 894 | void drbd_resume_io(struct drbd_device *device) |
b411b363 | 895 | { |
7dbb4386 PR |
896 | if (atomic_dec_and_test(&device->suspend_cnt)) |
897 | wake_up(&device->misc_wait); | |
b411b363 PR |
898 | } |
899 | ||
900 | /** | |
901 | * drbd_determine_dev_size() - Sets the right device size obeying all constraints | |
b30ab791 | 902 | * @device: DRBD device. |
b411b363 PR |
903 | * |
904 | * Returns 0 on success, negative return values indicate errors. | |
905 | * You should call drbd_md_sync() after calling this function. | |
906 | */ | |
d752b269 | 907 | enum determine_dev_size |
b30ab791 | 908 | drbd_determine_dev_size(struct drbd_device *device, enum dds_flags flags, struct resize_parms *rs) __must_hold(local) |
b411b363 | 909 | { |
8011e249 LE |
910 | struct md_offsets_and_sizes { |
911 | u64 last_agreed_sect; | |
912 | u64 md_offset; | |
913 | s32 al_offset; | |
914 | s32 bm_offset; | |
915 | u32 md_size_sect; | |
916 | ||
917 | u32 al_stripes; | |
918 | u32 al_stripe_size_4k; | |
919 | } prev; | |
920 | sector_t u_size, size; | |
b30ab791 | 921 | struct drbd_md *md = &device->ldev->md; |
b411b363 | 922 | char ppb[10]; |
d752b269 | 923 | void *buffer; |
b411b363 PR |
924 | |
925 | int md_moved, la_size_changed; | |
e96c9633 | 926 | enum determine_dev_size rv = DS_UNCHANGED; |
b411b363 | 927 | |
5f7c0124 LE |
928 | /* We may change the on-disk offsets of our meta data below. Lock out |
929 | * anything that may cause meta data IO, to avoid acting on incomplete | |
930 | * layout changes or scribbling over meta data that is in the process | |
931 | * of being moved. | |
b411b363 | 932 | * |
5f7c0124 LE |
933 | * Move is not exactly correct, btw, currently we have all our meta |
934 | * data in core memory, to "move" it we just write it all out, there | |
935 | * are no reads. */ | |
b30ab791 | 936 | drbd_suspend_io(device); |
e37d2438 | 937 | buffer = drbd_md_get_buffer(device, __func__); /* Lock meta-data IO */ |
d752b269 | 938 | if (!buffer) { |
b30ab791 | 939 | drbd_resume_io(device); |
d752b269 PR |
940 | return DS_ERROR; |
941 | } | |
b411b363 | 942 | |
8011e249 LE |
943 | /* remember current offset and sizes */ |
944 | prev.last_agreed_sect = md->la_size_sect; | |
945 | prev.md_offset = md->md_offset; | |
946 | prev.al_offset = md->al_offset; | |
947 | prev.bm_offset = md->bm_offset; | |
948 | prev.md_size_sect = md->md_size_sect; | |
949 | prev.al_stripes = md->al_stripes; | |
950 | prev.al_stripe_size_4k = md->al_stripe_size_4k; | |
b411b363 | 951 | |
d752b269 PR |
952 | if (rs) { |
953 | /* rs is non NULL if we should change the AL layout only */ | |
d752b269 PR |
954 | md->al_stripes = rs->al_stripes; |
955 | md->al_stripe_size_4k = rs->al_stripe_size / 4; | |
956 | md->al_size_4k = (u64)rs->al_stripes * rs->al_stripe_size / 4; | |
957 | } | |
958 | ||
b30ab791 | 959 | drbd_md_set_sector_offsets(device, device->ldev); |
b411b363 | 960 | |
daeda1cc | 961 | rcu_read_lock(); |
b30ab791 | 962 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc | 963 | rcu_read_unlock(); |
b30ab791 | 964 | size = drbd_new_dev_size(device, device->ldev, u_size, flags & DDSF_FORCED); |
b411b363 | 965 | |
8011e249 | 966 | if (size < prev.last_agreed_sect) { |
d752b269 PR |
967 | if (rs && u_size == 0) { |
968 | /* Remove "rs &&" later. This check should always be active, but | |
969 | right now the receiver expects the permissive behavior */ | |
d0180171 | 970 | drbd_warn(device, "Implicit shrink not allowed. " |
d752b269 PR |
971 | "Use --size=%llus for explicit shrink.\n", |
972 | (unsigned long long)size); | |
973 | rv = DS_ERROR_SHRINK; | |
974 | } | |
975 | if (u_size > size) | |
976 | rv = DS_ERROR_SPACE_MD; | |
977 | if (rv != DS_UNCHANGED) | |
978 | goto err_out; | |
979 | } | |
980 | ||
b30ab791 AG |
981 | if (drbd_get_capacity(device->this_bdev) != size || |
982 | drbd_bm_capacity(device) != size) { | |
b411b363 | 983 | int err; |
b30ab791 | 984 | err = drbd_bm_resize(device, size, !(flags & DDSF_NO_RESYNC)); |
b411b363 PR |
985 | if (unlikely(err)) { |
986 | /* currently there is only one error: ENOMEM! */ | |
8011e249 | 987 | size = drbd_bm_capacity(device); |
b411b363 | 988 | if (size == 0) { |
d0180171 | 989 | drbd_err(device, "OUT OF MEMORY! " |
b411b363 PR |
990 | "Could not allocate bitmap!\n"); |
991 | } else { | |
d0180171 | 992 | drbd_err(device, "BM resizing failed. " |
8011e249 | 993 | "Leaving size unchanged\n"); |
b411b363 | 994 | } |
e96c9633 | 995 | rv = DS_ERROR; |
b411b363 PR |
996 | } |
997 | /* racy, see comments above. */ | |
b30ab791 | 998 | drbd_set_my_capacity(device, size); |
8011e249 | 999 | md->la_size_sect = size; |
d0180171 | 1000 | drbd_info(device, "size = %s (%llu KB)\n", ppsize(ppb, size>>1), |
b411b363 PR |
1001 | (unsigned long long)size>>1); |
1002 | } | |
d752b269 PR |
1003 | if (rv <= DS_ERROR) |
1004 | goto err_out; | |
b411b363 | 1005 | |
8011e249 | 1006 | la_size_changed = (prev.last_agreed_sect != md->la_size_sect); |
b411b363 | 1007 | |
8011e249 LE |
1008 | md_moved = prev.md_offset != md->md_offset |
1009 | || prev.md_size_sect != md->md_size_sect; | |
b411b363 | 1010 | |
d752b269 PR |
1011 | if (la_size_changed || md_moved || rs) { |
1012 | u32 prev_flags; | |
24dccabb | 1013 | |
fcb09674 LE |
1014 | /* We do some synchronous IO below, which may take some time. |
1015 | * Clear the timer, to avoid scary "timer expired!" messages, | |
1016 | * "Superblock" is written out at least twice below, anyways. */ | |
1017 | del_timer(&device->md_sync_timer); | |
d752b269 | 1018 | |
5f7c0124 LE |
1019 | /* We won't change the "al-extents" setting, we just may need |
1020 | * to move the on-disk location of the activity log ringbuffer. | |
1021 | * Lock for transaction is good enough, it may well be "dirty" | |
1022 | * or even "starving". */ | |
1023 | wait_event(device->al_wait, lc_try_lock_for_transaction(device->act_log)); | |
1024 | ||
1025 | /* mark current on-disk bitmap and activity log as unreliable */ | |
d752b269 | 1026 | prev_flags = md->flags; |
5f7c0124 | 1027 | md->flags |= MDF_FULL_SYNC | MDF_AL_DISABLED; |
b30ab791 | 1028 | drbd_md_write(device, buffer); |
d752b269 | 1029 | |
5f7c0124 LE |
1030 | drbd_al_initialize(device, buffer); |
1031 | ||
d0180171 | 1032 | drbd_info(device, "Writing the whole bitmap, %s\n", |
b411b363 PR |
1033 | la_size_changed && md_moved ? "size changed and md moved" : |
1034 | la_size_changed ? "size changed" : "md moved"); | |
20ceb2b2 | 1035 | /* next line implicitly does drbd_suspend_io()+drbd_resume_io() */ |
b30ab791 | 1036 | drbd_bitmap_io(device, md_moved ? &drbd_bm_write_all : &drbd_bm_write, |
d752b269 | 1037 | "size changed", BM_LOCKED_MASK); |
d752b269 | 1038 | |
5f7c0124 LE |
1039 | /* on-disk bitmap and activity log is authoritative again |
1040 | * (unless there was an IO error meanwhile...) */ | |
d752b269 | 1041 | md->flags = prev_flags; |
b30ab791 | 1042 | drbd_md_write(device, buffer); |
d752b269 PR |
1043 | |
1044 | if (rs) | |
d0180171 AG |
1045 | drbd_info(device, "Changed AL layout to al-stripes = %d, al-stripe-size-kB = %d\n", |
1046 | md->al_stripes, md->al_stripe_size_4k * 4); | |
b411b363 PR |
1047 | } |
1048 | ||
8011e249 LE |
1049 | if (size > prev.last_agreed_sect) |
1050 | rv = prev.last_agreed_sect ? DS_GREW : DS_GREW_FROM_ZERO; | |
1051 | if (size < prev.last_agreed_sect) | |
e96c9633 | 1052 | rv = DS_SHRUNK; |
d752b269 PR |
1053 | |
1054 | if (0) { | |
1055 | err_out: | |
8011e249 LE |
1056 | /* restore previous offset and sizes */ |
1057 | md->la_size_sect = prev.last_agreed_sect; | |
1058 | md->md_offset = prev.md_offset; | |
1059 | md->al_offset = prev.al_offset; | |
1060 | md->bm_offset = prev.bm_offset; | |
1061 | md->md_size_sect = prev.md_size_sect; | |
1062 | md->al_stripes = prev.al_stripes; | |
1063 | md->al_stripe_size_4k = prev.al_stripe_size_4k; | |
1064 | md->al_size_4k = (u64)prev.al_stripes * prev.al_stripe_size_4k; | |
d752b269 | 1065 | } |
b30ab791 AG |
1066 | lc_unlock(device->act_log); |
1067 | wake_up(&device->al_wait); | |
1068 | drbd_md_put_buffer(device); | |
1069 | drbd_resume_io(device); | |
b411b363 PR |
1070 | |
1071 | return rv; | |
1072 | } | |
1073 | ||
1074 | sector_t | |
b30ab791 | 1075 | drbd_new_dev_size(struct drbd_device *device, struct drbd_backing_dev *bdev, |
ef5e44a6 | 1076 | sector_t u_size, int assume_peer_has_space) |
b411b363 | 1077 | { |
b30ab791 | 1078 | sector_t p_size = device->p_size; /* partner's disk size. */ |
cccac985 | 1079 | sector_t la_size_sect = bdev->md.la_size_sect; /* last agreed size. */ |
b411b363 | 1080 | sector_t m_size; /* my size */ |
b411b363 PR |
1081 | sector_t size = 0; |
1082 | ||
1083 | m_size = drbd_get_max_capacity(bdev); | |
1084 | ||
b30ab791 | 1085 | if (device->state.conn < C_CONNECTED && assume_peer_has_space) { |
d0180171 | 1086 | drbd_warn(device, "Resize while not connected was forced by the user!\n"); |
a393db6f PR |
1087 | p_size = m_size; |
1088 | } | |
1089 | ||
b411b363 PR |
1090 | if (p_size && m_size) { |
1091 | size = min_t(sector_t, p_size, m_size); | |
1092 | } else { | |
cccac985 LE |
1093 | if (la_size_sect) { |
1094 | size = la_size_sect; | |
b411b363 PR |
1095 | if (m_size && m_size < size) |
1096 | size = m_size; | |
1097 | if (p_size && p_size < size) | |
1098 | size = p_size; | |
1099 | } else { | |
1100 | if (m_size) | |
1101 | size = m_size; | |
1102 | if (p_size) | |
1103 | size = p_size; | |
1104 | } | |
1105 | } | |
1106 | ||
1107 | if (size == 0) | |
d0180171 | 1108 | drbd_err(device, "Both nodes diskless!\n"); |
b411b363 PR |
1109 | |
1110 | if (u_size) { | |
1111 | if (u_size > size) | |
d0180171 | 1112 | drbd_err(device, "Requested disk size is too big (%lu > %lu)\n", |
b411b363 PR |
1113 | (unsigned long)u_size>>1, (unsigned long)size>>1); |
1114 | else | |
1115 | size = u_size; | |
1116 | } | |
1117 | ||
1118 | return size; | |
1119 | } | |
1120 | ||
1121 | /** | |
1122 | * drbd_check_al_size() - Ensures that the AL is of the right size | |
b30ab791 | 1123 | * @device: DRBD device. |
b411b363 PR |
1124 | * |
1125 | * Returns -EBUSY if current al lru is still used, -ENOMEM when allocation | |
1126 | * failed, and 0 on success. You should call drbd_md_sync() after you called | |
1127 | * this function. | |
1128 | */ | |
b30ab791 | 1129 | static int drbd_check_al_size(struct drbd_device *device, struct disk_conf *dc) |
b411b363 PR |
1130 | { |
1131 | struct lru_cache *n, *t; | |
1132 | struct lc_element *e; | |
1133 | unsigned int in_use; | |
1134 | int i; | |
1135 | ||
b30ab791 AG |
1136 | if (device->act_log && |
1137 | device->act_log->nr_elements == dc->al_extents) | |
b411b363 PR |
1138 | return 0; |
1139 | ||
1140 | in_use = 0; | |
b30ab791 | 1141 | t = device->act_log; |
7ad651b5 | 1142 | n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION, |
f399002e | 1143 | dc->al_extents, sizeof(struct lc_element), 0); |
b411b363 PR |
1144 | |
1145 | if (n == NULL) { | |
d0180171 | 1146 | drbd_err(device, "Cannot allocate act_log lru!\n"); |
b411b363 PR |
1147 | return -ENOMEM; |
1148 | } | |
b30ab791 | 1149 | spin_lock_irq(&device->al_lock); |
b411b363 PR |
1150 | if (t) { |
1151 | for (i = 0; i < t->nr_elements; i++) { | |
1152 | e = lc_element_by_index(t, i); | |
1153 | if (e->refcnt) | |
d0180171 | 1154 | drbd_err(device, "refcnt(%d)==%d\n", |
b411b363 PR |
1155 | e->lc_number, e->refcnt); |
1156 | in_use += e->refcnt; | |
1157 | } | |
1158 | } | |
1159 | if (!in_use) | |
b30ab791 AG |
1160 | device->act_log = n; |
1161 | spin_unlock_irq(&device->al_lock); | |
b411b363 | 1162 | if (in_use) { |
d0180171 | 1163 | drbd_err(device, "Activity log still in use!\n"); |
b411b363 PR |
1164 | lc_destroy(n); |
1165 | return -EBUSY; | |
1166 | } else { | |
d01efcee | 1167 | lc_destroy(t); |
b411b363 | 1168 | } |
b30ab791 | 1169 | drbd_md_mark_dirty(device); /* we changed device->act_log->nr_elemens */ |
b411b363 PR |
1170 | return 0; |
1171 | } | |
1172 | ||
69ba1ee9 LE |
1173 | static void blk_queue_discard_granularity(struct request_queue *q, unsigned int granularity) |
1174 | { | |
1175 | q->limits.discard_granularity = granularity; | |
1176 | } | |
9104d31a LE |
1177 | |
1178 | static unsigned int drbd_max_discard_sectors(struct drbd_connection *connection) | |
1179 | { | |
1180 | /* when we introduced REQ_WRITE_SAME support, we also bumped | |
1181 | * our maximum supported batch bio size used for discards. */ | |
1182 | if (connection->agreed_features & DRBD_FF_WSAME) | |
1183 | return DRBD_MAX_BBIO_SECTORS; | |
1184 | /* before, with DRBD <= 8.4.6, we only allowed up to one AL_EXTENT_SIZE. */ | |
1185 | return AL_EXTENT_SIZE >> 9; | |
1186 | } | |
1187 | ||
69ba1ee9 LE |
1188 | static void decide_on_discard_support(struct drbd_device *device, |
1189 | struct request_queue *q, | |
1190 | struct request_queue *b, | |
1191 | bool discard_zeroes_if_aligned) | |
1192 | { | |
1193 | /* q = drbd device queue (device->rq_queue) | |
1194 | * b = backing device queue (device->ldev->backing_bdev->bd_disk->queue), | |
1195 | * or NULL if diskless | |
1196 | */ | |
1197 | struct drbd_connection *connection = first_peer_device(device)->connection; | |
1198 | bool can_do = b ? blk_queue_discard(b) : true; | |
1199 | ||
1200 | if (can_do && b && !b->limits.discard_zeroes_data && !discard_zeroes_if_aligned) { | |
1201 | can_do = false; | |
1202 | drbd_info(device, "discard_zeroes_data=0 and discard_zeroes_if_aligned=no: disabling discards\n"); | |
1203 | } | |
9104d31a | 1204 | if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & DRBD_FF_TRIM)) { |
69ba1ee9 LE |
1205 | can_do = false; |
1206 | drbd_info(connection, "peer DRBD too old, does not support TRIM: disabling discards\n"); | |
1207 | } | |
1208 | if (can_do) { | |
1209 | /* We don't care for the granularity, really. | |
1210 | * Stacking limits below should fix it for the local | |
1211 | * device. Whether or not it is a suitable granularity | |
1212 | * on the remote device is not our problem, really. If | |
1213 | * you care, you need to use devices with similar | |
1214 | * topology on all peers. */ | |
1215 | blk_queue_discard_granularity(q, 512); | |
9104d31a | 1216 | q->limits.max_discard_sectors = drbd_max_discard_sectors(connection); |
69ba1ee9 LE |
1217 | queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q); |
1218 | } else { | |
1219 | queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q); | |
1220 | blk_queue_discard_granularity(q, 0); | |
1221 | q->limits.max_discard_sectors = 0; | |
1222 | } | |
1223 | } | |
1224 | ||
1225 | static void fixup_discard_if_not_supported(struct request_queue *q) | |
1226 | { | |
1227 | /* To avoid confusion, if this queue does not support discard, clear | |
1228 | * max_discard_sectors, which is what lsblk -D reports to the user. | |
1229 | * Older kernels got this wrong in "stack limits". | |
1230 | * */ | |
1231 | if (!blk_queue_discard(q)) { | |
1232 | blk_queue_max_discard_sectors(q, 0); | |
1233 | blk_queue_discard_granularity(q, 0); | |
1234 | } | |
1235 | } | |
1236 | ||
9104d31a LE |
1237 | static void decide_on_write_same_support(struct drbd_device *device, |
1238 | struct request_queue *q, | |
1239 | struct request_queue *b, struct o_qlim *o) | |
1240 | { | |
1241 | struct drbd_peer_device *peer_device = first_peer_device(device); | |
1242 | struct drbd_connection *connection = peer_device->connection; | |
1243 | bool can_do = b ? b->limits.max_write_same_sectors : true; | |
1244 | ||
1245 | if (can_do && connection->cstate >= C_CONNECTED && !(connection->agreed_features & DRBD_FF_WSAME)) { | |
1246 | can_do = false; | |
1247 | drbd_info(peer_device, "peer does not support WRITE_SAME\n"); | |
1248 | } | |
1249 | ||
1250 | if (o) { | |
1251 | /* logical block size; queue_logical_block_size(NULL) is 512 */ | |
1252 | unsigned int peer_lbs = be32_to_cpu(o->logical_block_size); | |
1253 | unsigned int me_lbs_b = queue_logical_block_size(b); | |
1254 | unsigned int me_lbs = queue_logical_block_size(q); | |
1255 | ||
1256 | if (me_lbs_b != me_lbs) { | |
1257 | drbd_warn(device, | |
1258 | "logical block size of local backend does not match (drbd:%u, backend:%u); was this a late attach?\n", | |
1259 | me_lbs, me_lbs_b); | |
1260 | /* rather disable write same than trigger some BUG_ON later in the scsi layer. */ | |
1261 | can_do = false; | |
1262 | } | |
1263 | if (me_lbs_b != peer_lbs) { | |
1264 | drbd_warn(peer_device, "logical block sizes do not match (me:%u, peer:%u); this may cause problems.\n", | |
1265 | me_lbs, peer_lbs); | |
1266 | if (can_do) { | |
1267 | drbd_dbg(peer_device, "logical block size mismatch: WRITE_SAME disabled.\n"); | |
1268 | can_do = false; | |
1269 | } | |
1270 | me_lbs = max(me_lbs, me_lbs_b); | |
1271 | /* We cannot change the logical block size of an in-use queue. | |
1272 | * We can only hope that access happens to be properly aligned. | |
1273 | * If not, the peer will likely produce an IO error, and detach. */ | |
1274 | if (peer_lbs > me_lbs) { | |
1275 | if (device->state.role != R_PRIMARY) { | |
1276 | blk_queue_logical_block_size(q, peer_lbs); | |
1277 | drbd_warn(peer_device, "logical block size set to %u\n", peer_lbs); | |
1278 | } else { | |
1279 | drbd_warn(peer_device, | |
1280 | "current Primary must NOT adjust logical block size (%u -> %u); hope for the best.\n", | |
1281 | me_lbs, peer_lbs); | |
1282 | } | |
1283 | } | |
1284 | } | |
1285 | if (can_do && !o->write_same_capable) { | |
1286 | /* If we introduce an open-coded write-same loop on the receiving side, | |
1287 | * the peer would present itself as "capable". */ | |
1288 | drbd_dbg(peer_device, "WRITE_SAME disabled (peer device not capable)\n"); | |
1289 | can_do = false; | |
1290 | } | |
1291 | } | |
1292 | ||
1293 | blk_queue_max_write_same_sectors(q, can_do ? DRBD_MAX_BBIO_SECTORS : 0); | |
1294 | } | |
1295 | ||
8fe39aac | 1296 | static void drbd_setup_queue_param(struct drbd_device *device, struct drbd_backing_dev *bdev, |
9104d31a | 1297 | unsigned int max_bio_size, struct o_qlim *o) |
b411b363 | 1298 | { |
b30ab791 | 1299 | struct request_queue * const q = device->rq_queue; |
db141b2f LE |
1300 | unsigned int max_hw_sectors = max_bio_size >> 9; |
1301 | unsigned int max_segments = 0; | |
c1b3156f | 1302 | struct request_queue *b = NULL; |
dd4f699d LE |
1303 | struct disk_conf *dc; |
1304 | bool discard_zeroes_if_aligned = true; | |
99432fcc | 1305 | |
8fe39aac PR |
1306 | if (bdev) { |
1307 | b = bdev->backing_bdev->bd_disk->queue; | |
99432fcc PR |
1308 | |
1309 | max_hw_sectors = min(queue_max_hw_sectors(b), max_bio_size >> 9); | |
daeda1cc | 1310 | rcu_read_lock(); |
dd4f699d LE |
1311 | dc = rcu_dereference(device->ldev->disk_conf); |
1312 | max_segments = dc->max_bio_bvecs; | |
1313 | discard_zeroes_if_aligned = dc->discard_zeroes_if_aligned; | |
daeda1cc | 1314 | rcu_read_unlock(); |
c1b3156f PR |
1315 | |
1316 | blk_set_stacking_limits(&q->limits); | |
99432fcc | 1317 | } |
b411b363 | 1318 | |
1816a2b4 LE |
1319 | blk_queue_max_hw_sectors(q, max_hw_sectors); |
1320 | /* This is the workaround for "bio would need to, but cannot, be split" */ | |
1321 | blk_queue_max_segments(q, max_segments ? max_segments : BLK_MAX_SEGMENTS); | |
09cbfeaf | 1322 | blk_queue_segment_boundary(q, PAGE_SIZE-1); |
69ba1ee9 | 1323 | decide_on_discard_support(device, q, b, discard_zeroes_if_aligned); |
9104d31a LE |
1324 | decide_on_write_same_support(device, q, b, o); |
1325 | ||
c1b3156f | 1326 | if (b) { |
99432fcc PR |
1327 | blk_queue_stack_limits(q, b); |
1328 | ||
1329 | if (q->backing_dev_info.ra_pages != b->backing_dev_info.ra_pages) { | |
d0180171 | 1330 | drbd_info(device, "Adjusting my ra_pages to backing device's (%lu -> %lu)\n", |
99432fcc PR |
1331 | q->backing_dev_info.ra_pages, |
1332 | b->backing_dev_info.ra_pages); | |
1333 | q->backing_dev_info.ra_pages = b->backing_dev_info.ra_pages; | |
1334 | } | |
99432fcc | 1335 | } |
69ba1ee9 | 1336 | fixup_discard_if_not_supported(q); |
99432fcc PR |
1337 | } |
1338 | ||
9104d31a | 1339 | void drbd_reconsider_queue_parameters(struct drbd_device *device, struct drbd_backing_dev *bdev, struct o_qlim *o) |
99432fcc | 1340 | { |
db141b2f | 1341 | unsigned int now, new, local, peer; |
99432fcc | 1342 | |
b30ab791 AG |
1343 | now = queue_max_hw_sectors(device->rq_queue) << 9; |
1344 | local = device->local_max_bio_size; /* Eventually last known value, from volatile memory */ | |
1345 | peer = device->peer_max_bio_size; /* Eventually last known value, from meta data */ | |
b411b363 | 1346 | |
8fe39aac PR |
1347 | if (bdev) { |
1348 | local = queue_max_hw_sectors(bdev->backing_bdev->bd_disk->queue) << 9; | |
b30ab791 | 1349 | device->local_max_bio_size = local; |
b411b363 | 1350 | } |
db141b2f | 1351 | local = min(local, DRBD_MAX_BIO_SIZE); |
99432fcc PR |
1352 | |
1353 | /* We may ignore peer limits if the peer is modern enough. | |
1354 | Because new from 8.3.8 onwards the peer can use multiple | |
1355 | BIOs for a single peer_request */ | |
b30ab791 | 1356 | if (device->state.conn >= C_WF_REPORT_PARAMS) { |
a6b32bc3 | 1357 | if (first_peer_device(device)->connection->agreed_pro_version < 94) |
b30ab791 | 1358 | peer = min(device->peer_max_bio_size, DRBD_MAX_SIZE_H80_PACKET); |
6809384c | 1359 | /* Correct old drbd (up to 8.3.7) if it believes it can do more than 32KiB */ |
a6b32bc3 | 1360 | else if (first_peer_device(device)->connection->agreed_pro_version == 94) |
99432fcc | 1361 | peer = DRBD_MAX_SIZE_H80_PACKET; |
a6b32bc3 | 1362 | else if (first_peer_device(device)->connection->agreed_pro_version < 100) |
2ffca4f3 PR |
1363 | peer = DRBD_MAX_BIO_SIZE_P95; /* drbd 8.3.8 onwards, before 8.4.0 */ |
1364 | else | |
99432fcc | 1365 | peer = DRBD_MAX_BIO_SIZE; |
99432fcc | 1366 | |
fa090e70 LE |
1367 | /* We may later detach and re-attach on a disconnected Primary. |
1368 | * Avoid this setting to jump back in that case. | |
1369 | * We want to store what we know the peer DRBD can handle, | |
1370 | * not what the peer IO backend can handle. */ | |
1371 | if (peer > device->peer_max_bio_size) | |
1372 | device->peer_max_bio_size = peer; | |
1373 | } | |
db141b2f | 1374 | new = min(local, peer); |
99432fcc | 1375 | |
b30ab791 | 1376 | if (device->state.role == R_PRIMARY && new < now) |
d0180171 | 1377 | drbd_err(device, "ASSERT FAILED new < now; (%u < %u)\n", new, now); |
99432fcc PR |
1378 | |
1379 | if (new != now) | |
d0180171 | 1380 | drbd_info(device, "max BIO size = %u\n", new); |
99432fcc | 1381 | |
9104d31a | 1382 | drbd_setup_queue_param(device, bdev, new, o); |
b411b363 PR |
1383 | } |
1384 | ||
a18e9d1e | 1385 | /* Starts the worker thread */ |
bde89a9e | 1386 | static void conn_reconfig_start(struct drbd_connection *connection) |
b411b363 | 1387 | { |
bde89a9e | 1388 | drbd_thread_start(&connection->worker); |
b5043c5e | 1389 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 PR |
1390 | } |
1391 | ||
a18e9d1e | 1392 | /* if still unconfigured, stops worker again. */ |
bde89a9e | 1393 | static void conn_reconfig_done(struct drbd_connection *connection) |
b411b363 | 1394 | { |
992d6e91 | 1395 | bool stop_threads; |
0500813f | 1396 | spin_lock_irq(&connection->resource->req_lock); |
bde89a9e AG |
1397 | stop_threads = conn_all_vols_unconf(connection) && |
1398 | connection->cstate == C_STANDALONE; | |
0500813f | 1399 | spin_unlock_irq(&connection->resource->req_lock); |
992d6e91 | 1400 | if (stop_threads) { |
668700b4 PR |
1401 | /* ack_receiver thread and ack_sender workqueue are implicitly |
1402 | * stopped by receiver in conn_disconnect() */ | |
bde89a9e AG |
1403 | drbd_thread_stop(&connection->receiver); |
1404 | drbd_thread_stop(&connection->worker); | |
992d6e91 | 1405 | } |
b411b363 PR |
1406 | } |
1407 | ||
0778286a | 1408 | /* Make sure IO is suspended before calling this function(). */ |
b30ab791 | 1409 | static void drbd_suspend_al(struct drbd_device *device) |
0778286a PR |
1410 | { |
1411 | int s = 0; | |
1412 | ||
b30ab791 | 1413 | if (!lc_try_lock(device->act_log)) { |
d0180171 | 1414 | drbd_warn(device, "Failed to lock al in drbd_suspend_al()\n"); |
0778286a PR |
1415 | return; |
1416 | } | |
1417 | ||
b30ab791 | 1418 | drbd_al_shrink(device); |
0500813f | 1419 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 AG |
1420 | if (device->state.conn < C_CONNECTED) |
1421 | s = !test_and_set_bit(AL_SUSPENDED, &device->flags); | |
0500813f | 1422 | spin_unlock_irq(&device->resource->req_lock); |
b30ab791 | 1423 | lc_unlock(device->act_log); |
0778286a PR |
1424 | |
1425 | if (s) | |
d0180171 | 1426 | drbd_info(device, "Suspended AL updates\n"); |
0778286a PR |
1427 | } |
1428 | ||
5979e361 LE |
1429 | |
1430 | static bool should_set_defaults(struct genl_info *info) | |
1431 | { | |
1432 | unsigned flags = ((struct drbd_genlmsghdr*)info->userhdr)->flags; | |
1433 | return 0 != (flags & DRBD_GENL_F_SET_DEFAULTS); | |
1434 | } | |
1435 | ||
5bbcf5e6 | 1436 | static unsigned int drbd_al_extents_max(struct drbd_backing_dev *bdev) |
d589a21e | 1437 | { |
5bbcf5e6 LE |
1438 | /* This is limited by 16 bit "slot" numbers, |
1439 | * and by available on-disk context storage. | |
1440 | * | |
1441 | * Also (u16)~0 is special (denotes a "free" extent). | |
1442 | * | |
1443 | * One transaction occupies one 4kB on-disk block, | |
1444 | * we have n such blocks in the on disk ring buffer, | |
1445 | * the "current" transaction may fail (n-1), | |
1446 | * and there is 919 slot numbers context information per transaction. | |
1447 | * | |
1448 | * 72 transaction blocks amounts to more than 2**16 context slots, | |
1449 | * so cap there first. | |
1450 | */ | |
1451 | const unsigned int max_al_nr = DRBD_AL_EXTENTS_MAX; | |
1452 | const unsigned int sufficient_on_disk = | |
1453 | (max_al_nr + AL_CONTEXT_PER_TRANSACTION -1) | |
1454 | /AL_CONTEXT_PER_TRANSACTION; | |
d589a21e | 1455 | |
5bbcf5e6 LE |
1456 | unsigned int al_size_4k = bdev->md.al_size_4k; |
1457 | ||
1458 | if (al_size_4k > sufficient_on_disk) | |
1459 | return max_al_nr; | |
1460 | ||
1461 | return (al_size_4k - 1) * AL_CONTEXT_PER_TRANSACTION; | |
d589a21e PR |
1462 | } |
1463 | ||
70df7092 LE |
1464 | static bool write_ordering_changed(struct disk_conf *a, struct disk_conf *b) |
1465 | { | |
1466 | return a->disk_barrier != b->disk_barrier || | |
1467 | a->disk_flushes != b->disk_flushes || | |
1468 | a->disk_drain != b->disk_drain; | |
1469 | } | |
1470 | ||
a5ca66c4 PR |
1471 | static void sanitize_disk_conf(struct drbd_device *device, struct disk_conf *disk_conf, |
1472 | struct drbd_backing_dev *nbc) | |
c5c23854 | 1473 | { |
a5ca66c4 PR |
1474 | struct request_queue * const q = nbc->backing_bdev->bd_disk->queue; |
1475 | ||
c5c23854 PR |
1476 | if (disk_conf->al_extents < DRBD_AL_EXTENTS_MIN) |
1477 | disk_conf->al_extents = DRBD_AL_EXTENTS_MIN; | |
1478 | if (disk_conf->al_extents > drbd_al_extents_max(nbc)) | |
1479 | disk_conf->al_extents = drbd_al_extents_max(nbc); | |
a5ca66c4 | 1480 | |
65f5be35 LE |
1481 | if (!blk_queue_discard(q) |
1482 | || (!q->limits.discard_zeroes_data && !disk_conf->discard_zeroes_if_aligned)) { | |
1483 | if (disk_conf->rs_discard_granularity) { | |
1484 | disk_conf->rs_discard_granularity = 0; /* disable feature */ | |
1485 | drbd_info(device, "rs_discard_granularity feature disabled\n"); | |
1486 | } | |
a5ca66c4 PR |
1487 | } |
1488 | ||
1489 | if (disk_conf->rs_discard_granularity) { | |
1490 | int orig_value = disk_conf->rs_discard_granularity; | |
1491 | int remainder; | |
1492 | ||
1493 | if (q->limits.discard_granularity > disk_conf->rs_discard_granularity) | |
1494 | disk_conf->rs_discard_granularity = q->limits.discard_granularity; | |
1495 | ||
1496 | remainder = disk_conf->rs_discard_granularity % q->limits.discard_granularity; | |
1497 | disk_conf->rs_discard_granularity += remainder; | |
1498 | ||
1499 | if (disk_conf->rs_discard_granularity > q->limits.max_discard_sectors << 9) | |
1500 | disk_conf->rs_discard_granularity = q->limits.max_discard_sectors << 9; | |
1501 | ||
1502 | if (disk_conf->rs_discard_granularity != orig_value) | |
1503 | drbd_info(device, "rs_discard_granularity changed to %d\n", | |
1504 | disk_conf->rs_discard_granularity); | |
1505 | } | |
c5c23854 PR |
1506 | } |
1507 | ||
f399002e LE |
1508 | int drbd_adm_disk_opts(struct sk_buff *skb, struct genl_info *info) |
1509 | { | |
a910b123 | 1510 | struct drbd_config_context adm_ctx; |
f399002e | 1511 | enum drbd_ret_code retcode; |
b30ab791 | 1512 | struct drbd_device *device; |
daeda1cc | 1513 | struct disk_conf *new_disk_conf, *old_disk_conf; |
813472ce | 1514 | struct fifo_buffer *old_plan = NULL, *new_plan = NULL; |
f399002e | 1515 | int err, fifo_size; |
f399002e | 1516 | |
a910b123 | 1517 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
f399002e LE |
1518 | if (!adm_ctx.reply_skb) |
1519 | return retcode; | |
1520 | if (retcode != NO_ERROR) | |
9e276872 | 1521 | goto finish; |
f399002e | 1522 | |
b30ab791 | 1523 | device = adm_ctx.device; |
9e276872 | 1524 | mutex_lock(&adm_ctx.resource->adm_mutex); |
f399002e LE |
1525 | |
1526 | /* we also need a disk | |
1527 | * to change the options on */ | |
b30ab791 | 1528 | if (!get_ldev(device)) { |
f399002e LE |
1529 | retcode = ERR_NO_DISK; |
1530 | goto out; | |
1531 | } | |
1532 | ||
daeda1cc | 1533 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); |
5ecc72c3 | 1534 | if (!new_disk_conf) { |
f399002e LE |
1535 | retcode = ERR_NOMEM; |
1536 | goto fail; | |
1537 | } | |
1538 | ||
0500813f | 1539 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 1540 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc | 1541 | *new_disk_conf = *old_disk_conf; |
5979e361 | 1542 | if (should_set_defaults(info)) |
b966b5dd | 1543 | set_disk_conf_defaults(new_disk_conf); |
5979e361 | 1544 | |
5ecc72c3 | 1545 | err = disk_conf_from_attrs_for_change(new_disk_conf, info); |
c75b9b10 | 1546 | if (err && err != -ENOMSG) { |
f399002e | 1547 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1548 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
8e229434 | 1549 | goto fail_unlock; |
f399002e LE |
1550 | } |
1551 | ||
5ecc72c3 LE |
1552 | if (!expect(new_disk_conf->resync_rate >= 1)) |
1553 | new_disk_conf->resync_rate = 1; | |
f399002e | 1554 | |
a5ca66c4 | 1555 | sanitize_disk_conf(device, new_disk_conf, device->ldev); |
5bbcf5e6 LE |
1556 | |
1557 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) | |
1558 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
f399002e | 1559 | |
5ecc72c3 | 1560 | fifo_size = (new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ; |
b30ab791 | 1561 | if (fifo_size != device->rs_plan_s->size) { |
813472ce PR |
1562 | new_plan = fifo_alloc(fifo_size); |
1563 | if (!new_plan) { | |
d0180171 | 1564 | drbd_err(device, "kmalloc of fifo_buffer failed"); |
f399002e | 1565 | retcode = ERR_NOMEM; |
daeda1cc | 1566 | goto fail_unlock; |
f399002e LE |
1567 | } |
1568 | } | |
1569 | ||
b30ab791 AG |
1570 | drbd_suspend_io(device); |
1571 | wait_event(device->al_wait, lc_try_lock(device->act_log)); | |
1572 | drbd_al_shrink(device); | |
1573 | err = drbd_check_al_size(device, new_disk_conf); | |
1574 | lc_unlock(device->act_log); | |
1575 | wake_up(&device->al_wait); | |
1576 | drbd_resume_io(device); | |
f399002e LE |
1577 | |
1578 | if (err) { | |
1579 | retcode = ERR_NOMEM; | |
daeda1cc | 1580 | goto fail_unlock; |
f399002e LE |
1581 | } |
1582 | ||
28bc3b8c | 1583 | lock_all_resources(); |
b30ab791 | 1584 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); |
dc97b708 | 1585 | if (retcode == NO_ERROR) { |
b30ab791 AG |
1586 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
1587 | drbd_resync_after_changed(device); | |
dc97b708 | 1588 | } |
28bc3b8c | 1589 | unlock_all_resources(); |
f399002e | 1590 | |
daeda1cc PR |
1591 | if (retcode != NO_ERROR) |
1592 | goto fail_unlock; | |
f399002e | 1593 | |
813472ce | 1594 | if (new_plan) { |
b30ab791 AG |
1595 | old_plan = device->rs_plan_s; |
1596 | rcu_assign_pointer(device->rs_plan_s, new_plan); | |
9958c857 | 1597 | } |
9958c857 | 1598 | |
0500813f | 1599 | mutex_unlock(&device->resource->conf_update); |
27eb13e9 | 1600 | |
9a51ab1c | 1601 | if (new_disk_conf->al_updates) |
b30ab791 | 1602 | device->ldev->md.flags &= ~MDF_AL_DISABLED; |
9a51ab1c | 1603 | else |
b30ab791 | 1604 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c | 1605 | |
691631c0 | 1606 | if (new_disk_conf->md_flushes) |
b30ab791 | 1607 | clear_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1608 | else |
b30ab791 | 1609 | set_bit(MD_NO_FUA, &device->flags); |
691631c0 | 1610 | |
70df7092 | 1611 | if (write_ordering_changed(old_disk_conf, new_disk_conf)) |
f6ba8636 | 1612 | drbd_bump_write_ordering(device->resource, NULL, WO_BDEV_FLUSH); |
27eb13e9 | 1613 | |
dd4f699d | 1614 | if (old_disk_conf->discard_zeroes_if_aligned != new_disk_conf->discard_zeroes_if_aligned) |
9104d31a | 1615 | drbd_reconsider_queue_parameters(device, device->ldev, NULL); |
dd4f699d | 1616 | |
b30ab791 | 1617 | drbd_md_sync(device); |
f399002e | 1618 | |
69a22773 AG |
1619 | if (device->state.conn >= C_CONNECTED) { |
1620 | struct drbd_peer_device *peer_device; | |
1621 | ||
1622 | for_each_peer_device(peer_device, device) | |
1623 | drbd_send_sync_param(peer_device); | |
1624 | } | |
f399002e | 1625 | |
daeda1cc PR |
1626 | synchronize_rcu(); |
1627 | kfree(old_disk_conf); | |
813472ce | 1628 | kfree(old_plan); |
b30ab791 | 1629 | mod_timer(&device->request_timer, jiffies + HZ); |
daeda1cc PR |
1630 | goto success; |
1631 | ||
1632 | fail_unlock: | |
0500813f | 1633 | mutex_unlock(&device->resource->conf_update); |
f399002e | 1634 | fail: |
5ecc72c3 | 1635 | kfree(new_disk_conf); |
813472ce | 1636 | kfree(new_plan); |
daeda1cc | 1637 | success: |
b30ab791 | 1638 | put_ldev(device); |
f399002e | 1639 | out: |
9e276872 LE |
1640 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
1641 | finish: | |
a910b123 | 1642 | drbd_adm_finish(&adm_ctx, info, retcode); |
f399002e LE |
1643 | return 0; |
1644 | } | |
1645 | ||
63a7c8ad LE |
1646 | static struct block_device *open_backing_dev(struct drbd_device *device, |
1647 | const char *bdev_path, void *claim_ptr, bool do_bd_link) | |
1648 | { | |
1649 | struct block_device *bdev; | |
1650 | int err = 0; | |
1651 | ||
1652 | bdev = blkdev_get_by_path(bdev_path, | |
1653 | FMODE_READ | FMODE_WRITE | FMODE_EXCL, claim_ptr); | |
1654 | if (IS_ERR(bdev)) { | |
1655 | drbd_err(device, "open(\"%s\") failed with %ld\n", | |
1656 | bdev_path, PTR_ERR(bdev)); | |
1657 | return bdev; | |
1658 | } | |
1659 | ||
1660 | if (!do_bd_link) | |
1661 | return bdev; | |
1662 | ||
1663 | err = bd_link_disk_holder(bdev, device->vdisk); | |
1664 | if (err) { | |
1665 | blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1666 | drbd_err(device, "bd_link_disk_holder(\"%s\", ...) failed with %d\n", | |
1667 | bdev_path, err); | |
1668 | bdev = ERR_PTR(err); | |
1669 | } | |
1670 | return bdev; | |
1671 | } | |
1672 | ||
1673 | static int open_backing_devices(struct drbd_device *device, | |
1674 | struct disk_conf *new_disk_conf, | |
1675 | struct drbd_backing_dev *nbc) | |
1676 | { | |
1677 | struct block_device *bdev; | |
1678 | ||
1679 | bdev = open_backing_dev(device, new_disk_conf->backing_dev, device, true); | |
1680 | if (IS_ERR(bdev)) | |
1681 | return ERR_OPEN_DISK; | |
1682 | nbc->backing_bdev = bdev; | |
1683 | ||
1684 | /* | |
1685 | * meta_dev_idx >= 0: external fixed size, possibly multiple | |
1686 | * drbd sharing one meta device. TODO in that case, paranoia | |
1687 | * check that [md_bdev, meta_dev_idx] is not yet used by some | |
1688 | * other drbd minor! (if you use drbd.conf + drbdadm, that | |
1689 | * should check it for you already; but if you don't, or | |
1690 | * someone fooled it, we need to double check here) | |
1691 | */ | |
1692 | bdev = open_backing_dev(device, new_disk_conf->meta_dev, | |
1693 | /* claim ptr: device, if claimed exclusively; shared drbd_m_holder, | |
1694 | * if potentially shared with other drbd minors */ | |
1695 | (new_disk_conf->meta_dev_idx < 0) ? (void*)device : (void*)drbd_m_holder, | |
1696 | /* avoid double bd_claim_by_disk() for the same (source,target) tuple, | |
1697 | * as would happen with internal metadata. */ | |
1698 | (new_disk_conf->meta_dev_idx != DRBD_MD_INDEX_FLEX_INT && | |
1699 | new_disk_conf->meta_dev_idx != DRBD_MD_INDEX_INTERNAL)); | |
1700 | if (IS_ERR(bdev)) | |
1701 | return ERR_OPEN_MD_DISK; | |
1702 | nbc->md_bdev = bdev; | |
1703 | return NO_ERROR; | |
1704 | } | |
1705 | ||
1706 | static void close_backing_dev(struct drbd_device *device, struct block_device *bdev, | |
1707 | bool do_bd_unlink) | |
1708 | { | |
1709 | if (!bdev) | |
1710 | return; | |
1711 | if (do_bd_unlink) | |
1712 | bd_unlink_disk_holder(bdev, device->vdisk); | |
1713 | blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL); | |
1714 | } | |
1715 | ||
1716 | void drbd_backing_dev_free(struct drbd_device *device, struct drbd_backing_dev *ldev) | |
1717 | { | |
1718 | if (ldev == NULL) | |
1719 | return; | |
1720 | ||
1721 | close_backing_dev(device, ldev->md_bdev, ldev->md_bdev != ldev->backing_bdev); | |
1722 | close_backing_dev(device, ldev->backing_bdev, true); | |
1723 | ||
1724 | kfree(ldev->disk_conf); | |
1725 | kfree(ldev); | |
1726 | } | |
1727 | ||
3b98c0c2 | 1728 | int drbd_adm_attach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 1729 | { |
a910b123 | 1730 | struct drbd_config_context adm_ctx; |
b30ab791 | 1731 | struct drbd_device *device; |
44a4d551 LE |
1732 | struct drbd_peer_device *peer_device; |
1733 | struct drbd_connection *connection; | |
3b98c0c2 | 1734 | int err; |
116676ca | 1735 | enum drbd_ret_code retcode; |
b411b363 PR |
1736 | enum determine_dev_size dd; |
1737 | sector_t max_possible_sectors; | |
1738 | sector_t min_md_device_sectors; | |
1739 | struct drbd_backing_dev *nbc = NULL; /* new_backing_conf */ | |
daeda1cc | 1740 | struct disk_conf *new_disk_conf = NULL; |
b411b363 | 1741 | struct lru_cache *resync_lru = NULL; |
9958c857 | 1742 | struct fifo_buffer *new_plan = NULL; |
b411b363 | 1743 | union drbd_state ns, os; |
f2024e7c | 1744 | enum drbd_state_rv rv; |
44ed167d | 1745 | struct net_conf *nc; |
b411b363 | 1746 | |
a910b123 | 1747 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
1748 | if (!adm_ctx.reply_skb) |
1749 | return retcode; | |
1750 | if (retcode != NO_ERROR) | |
40cbf085 | 1751 | goto finish; |
b411b363 | 1752 | |
b30ab791 | 1753 | device = adm_ctx.device; |
9e276872 | 1754 | mutex_lock(&adm_ctx.resource->adm_mutex); |
44a4d551 | 1755 | peer_device = first_peer_device(device); |
3b8a44f8 | 1756 | connection = peer_device->connection; |
44a4d551 | 1757 | conn_reconfig_start(connection); |
b411b363 PR |
1758 | |
1759 | /* if you want to reconfigure, please tear down first */ | |
b30ab791 | 1760 | if (device->state.disk > D_DISKLESS) { |
b411b363 PR |
1761 | retcode = ERR_DISK_CONFIGURED; |
1762 | goto fail; | |
1763 | } | |
82f59cc6 LE |
1764 | /* It may just now have detached because of IO error. Make sure |
1765 | * drbd_ldev_destroy is done already, we may end up here very fast, | |
1766 | * e.g. if someone calls attach from the on-io-error handler, | |
1767 | * to realize a "hot spare" feature (not that I'd recommend that) */ | |
e334f550 | 1768 | wait_event(device->misc_wait, !test_bit(GOING_DISKLESS, &device->flags)); |
b411b363 | 1769 | |
383606e0 | 1770 | /* make sure there is no leftover from previous force-detach attempts */ |
b30ab791 AG |
1771 | clear_bit(FORCE_DETACH, &device->flags); |
1772 | clear_bit(WAS_IO_ERROR, &device->flags); | |
1773 | clear_bit(WAS_READ_ERROR, &device->flags); | |
383606e0 | 1774 | |
0029d624 | 1775 | /* and no leftover from previously aborted resync or verify, either */ |
b30ab791 AG |
1776 | device->rs_total = 0; |
1777 | device->rs_failed = 0; | |
1778 | atomic_set(&device->rs_pending_cnt, 0); | |
0029d624 | 1779 | |
3b98c0c2 | 1780 | /* allocation not in the IO path, drbdsetup context */ |
b411b363 PR |
1781 | nbc = kzalloc(sizeof(struct drbd_backing_dev), GFP_KERNEL); |
1782 | if (!nbc) { | |
1783 | retcode = ERR_NOMEM; | |
1784 | goto fail; | |
1785 | } | |
9f2247bb PR |
1786 | spin_lock_init(&nbc->md.uuid_lock); |
1787 | ||
daeda1cc PR |
1788 | new_disk_conf = kzalloc(sizeof(struct disk_conf), GFP_KERNEL); |
1789 | if (!new_disk_conf) { | |
1790 | retcode = ERR_NOMEM; | |
b411b363 PR |
1791 | goto fail; |
1792 | } | |
daeda1cc | 1793 | nbc->disk_conf = new_disk_conf; |
b411b363 | 1794 | |
daeda1cc PR |
1795 | set_disk_conf_defaults(new_disk_conf); |
1796 | err = disk_conf_from_attrs(new_disk_conf, info); | |
3b98c0c2 | 1797 | if (err) { |
b411b363 | 1798 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 1799 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
1800 | goto fail; |
1801 | } | |
1802 | ||
5bbcf5e6 LE |
1803 | if (new_disk_conf->c_plan_ahead > DRBD_C_PLAN_AHEAD_MAX) |
1804 | new_disk_conf->c_plan_ahead = DRBD_C_PLAN_AHEAD_MAX; | |
d589a21e | 1805 | |
9958c857 PR |
1806 | new_plan = fifo_alloc((new_disk_conf->c_plan_ahead * 10 * SLEEP_TIME) / HZ); |
1807 | if (!new_plan) { | |
1808 | retcode = ERR_NOMEM; | |
1809 | goto fail; | |
1810 | } | |
1811 | ||
daeda1cc | 1812 | if (new_disk_conf->meta_dev_idx < DRBD_MD_INDEX_FLEX_INT) { |
b411b363 PR |
1813 | retcode = ERR_MD_IDX_INVALID; |
1814 | goto fail; | |
1815 | } | |
1816 | ||
44ed167d | 1817 | rcu_read_lock(); |
44a4d551 | 1818 | nc = rcu_dereference(connection->net_conf); |
44ed167d | 1819 | if (nc) { |
daeda1cc | 1820 | if (new_disk_conf->fencing == FP_STONITH && nc->wire_protocol == DRBD_PROT_A) { |
44ed167d | 1821 | rcu_read_unlock(); |
47ff2d0a | 1822 | retcode = ERR_STONITH_AND_PROT_A; |
28bc3b8c | 1823 | goto fail; |
47ff2d0a PR |
1824 | } |
1825 | } | |
44ed167d | 1826 | rcu_read_unlock(); |
47ff2d0a | 1827 | |
63a7c8ad LE |
1828 | retcode = open_backing_devices(device, new_disk_conf, nbc); |
1829 | if (retcode != NO_ERROR) | |
28bc3b8c | 1830 | goto fail; |
b411b363 | 1831 | |
e525fd89 | 1832 | if ((nbc->backing_bdev == nbc->md_bdev) != |
daeda1cc PR |
1833 | (new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_INTERNAL || |
1834 | new_disk_conf->meta_dev_idx == DRBD_MD_INDEX_FLEX_INT)) { | |
e525fd89 | 1835 | retcode = ERR_MD_IDX_INVALID; |
28bc3b8c | 1836 | goto fail; |
b411b363 PR |
1837 | } |
1838 | ||
1839 | resync_lru = lc_create("resync", drbd_bm_ext_cache, | |
46a15bc3 | 1840 | 1, 61, sizeof(struct bm_extent), |
b411b363 PR |
1841 | offsetof(struct bm_extent, lce)); |
1842 | if (!resync_lru) { | |
1843 | retcode = ERR_NOMEM; | |
28bc3b8c | 1844 | goto fail; |
b411b363 PR |
1845 | } |
1846 | ||
c04ccaa6 LE |
1847 | /* Read our meta data super block early. |
1848 | * This also sets other on-disk offsets. */ | |
b30ab791 | 1849 | retcode = drbd_md_read(device, nbc); |
c04ccaa6 | 1850 | if (retcode != NO_ERROR) |
28bc3b8c | 1851 | goto fail; |
b411b363 | 1852 | |
a5ca66c4 | 1853 | sanitize_disk_conf(device, new_disk_conf, nbc); |
5bbcf5e6 | 1854 | |
daeda1cc | 1855 | if (drbd_get_max_capacity(nbc) < new_disk_conf->disk_size) { |
d0180171 | 1856 | drbd_err(device, "max capacity %llu smaller than disk size %llu\n", |
b411b363 | 1857 | (unsigned long long) drbd_get_max_capacity(nbc), |
daeda1cc | 1858 | (unsigned long long) new_disk_conf->disk_size); |
7948bcdc | 1859 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1860 | goto fail; |
b411b363 PR |
1861 | } |
1862 | ||
daeda1cc | 1863 | if (new_disk_conf->meta_dev_idx < 0) { |
b411b363 PR |
1864 | max_possible_sectors = DRBD_MAX_SECTORS_FLEX; |
1865 | /* at least one MB, otherwise it does not make sense */ | |
1866 | min_md_device_sectors = (2<<10); | |
1867 | } else { | |
1868 | max_possible_sectors = DRBD_MAX_SECTORS; | |
ae8bf312 | 1869 | min_md_device_sectors = MD_128MB_SECT * (new_disk_conf->meta_dev_idx + 1); |
b411b363 PR |
1870 | } |
1871 | ||
b411b363 | 1872 | if (drbd_get_capacity(nbc->md_bdev) < min_md_device_sectors) { |
7948bcdc | 1873 | retcode = ERR_MD_DISK_TOO_SMALL; |
d0180171 | 1874 | drbd_warn(device, "refusing attach: md-device too small, " |
b411b363 PR |
1875 | "at least %llu sectors needed for this meta-disk type\n", |
1876 | (unsigned long long) min_md_device_sectors); | |
28bc3b8c | 1877 | goto fail; |
b411b363 PR |
1878 | } |
1879 | ||
1880 | /* Make sure the new disk is big enough | |
1881 | * (we may currently be R_PRIMARY with no local disk...) */ | |
1882 | if (drbd_get_max_capacity(nbc) < | |
b30ab791 | 1883 | drbd_get_capacity(device->this_bdev)) { |
7948bcdc | 1884 | retcode = ERR_DISK_TOO_SMALL; |
28bc3b8c | 1885 | goto fail; |
b411b363 PR |
1886 | } |
1887 | ||
1888 | nbc->known_size = drbd_get_capacity(nbc->backing_bdev); | |
1889 | ||
1352994b | 1890 | if (nbc->known_size > max_possible_sectors) { |
d0180171 | 1891 | drbd_warn(device, "==> truncating very big lower level device " |
1352994b LE |
1892 | "to currently maximum possible %llu sectors <==\n", |
1893 | (unsigned long long) max_possible_sectors); | |
daeda1cc | 1894 | if (new_disk_conf->meta_dev_idx >= 0) |
d0180171 | 1895 | drbd_warn(device, "==>> using internal or flexible " |
1352994b LE |
1896 | "meta data may help <<==\n"); |
1897 | } | |
1898 | ||
b30ab791 | 1899 | drbd_suspend_io(device); |
b411b363 | 1900 | /* also wait for the last barrier ack. */ |
b6dd1a89 LE |
1901 | /* FIXME see also https://daiquiri.linbit/cgi-bin/bugzilla/show_bug.cgi?id=171 |
1902 | * We need a way to either ignore barrier acks for barriers sent before a device | |
1903 | * was attached, or a way to wait for all pending barrier acks to come in. | |
1904 | * As barriers are counted per resource, | |
1905 | * we'd need to suspend io on all devices of a resource. | |
1906 | */ | |
b30ab791 | 1907 | wait_event(device->misc_wait, !atomic_read(&device->ap_pending_cnt) || drbd_suspended(device)); |
b411b363 | 1908 | /* and for any other previously queued work */ |
44a4d551 | 1909 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 1910 | |
b30ab791 | 1911 | rv = _drbd_request_state(device, NS(disk, D_ATTACHING), CS_VERBOSE); |
f2024e7c | 1912 | retcode = rv; /* FIXME: Type mismatch. */ |
b30ab791 | 1913 | drbd_resume_io(device); |
f2024e7c | 1914 | if (rv < SS_SUCCESS) |
28bc3b8c | 1915 | goto fail; |
b411b363 | 1916 | |
b30ab791 | 1917 | if (!get_ldev_if_state(device, D_ATTACHING)) |
b411b363 PR |
1918 | goto force_diskless; |
1919 | ||
b30ab791 AG |
1920 | if (!device->bitmap) { |
1921 | if (drbd_bm_init(device)) { | |
b411b363 PR |
1922 | retcode = ERR_NOMEM; |
1923 | goto force_diskless_dec; | |
1924 | } | |
1925 | } | |
1926 | ||
b30ab791 | 1927 | if (device->state.conn < C_CONNECTED && |
babea49e | 1928 | device->state.role == R_PRIMARY && device->ed_uuid && |
b30ab791 | 1929 | (device->ed_uuid & ~((u64)1)) != (nbc->md.uuid[UI_CURRENT] & ~((u64)1))) { |
d0180171 | 1930 | drbd_err(device, "Can only attach to data with current UUID=%016llX\n", |
b30ab791 | 1931 | (unsigned long long)device->ed_uuid); |
b411b363 PR |
1932 | retcode = ERR_DATA_NOT_CURRENT; |
1933 | goto force_diskless_dec; | |
1934 | } | |
1935 | ||
1936 | /* Since we are diskless, fix the activity log first... */ | |
b30ab791 | 1937 | if (drbd_check_al_size(device, new_disk_conf)) { |
b411b363 PR |
1938 | retcode = ERR_NOMEM; |
1939 | goto force_diskless_dec; | |
1940 | } | |
1941 | ||
1942 | /* Prevent shrinking of consistent devices ! */ | |
1943 | if (drbd_md_test_flag(nbc, MDF_CONSISTENT) && | |
b30ab791 | 1944 | drbd_new_dev_size(device, nbc, nbc->disk_conf->disk_size, 0) < nbc->md.la_size_sect) { |
d0180171 | 1945 | drbd_warn(device, "refusing to truncate a consistent device\n"); |
7948bcdc | 1946 | retcode = ERR_DISK_TOO_SMALL; |
b411b363 PR |
1947 | goto force_diskless_dec; |
1948 | } | |
1949 | ||
28bc3b8c AG |
1950 | lock_all_resources(); |
1951 | retcode = drbd_resync_after_valid(device, new_disk_conf->resync_after); | |
1952 | if (retcode != NO_ERROR) { | |
1953 | unlock_all_resources(); | |
1954 | goto force_diskless_dec; | |
1955 | } | |
1956 | ||
b411b363 PR |
1957 | /* Reset the "barriers don't work" bits here, then force meta data to |
1958 | * be written, to ensure we determine if barriers are supported. */ | |
e544046a | 1959 | if (new_disk_conf->md_flushes) |
b30ab791 | 1960 | clear_bit(MD_NO_FUA, &device->flags); |
b411b363 | 1961 | else |
b30ab791 | 1962 | set_bit(MD_NO_FUA, &device->flags); |
b411b363 PR |
1963 | |
1964 | /* Point of no return reached. | |
1965 | * Devices and memory are no longer released by error cleanup below. | |
b30ab791 | 1966 | * now device takes over responsibility, and the state engine should |
b411b363 | 1967 | * clean it up somewhere. */ |
0b0ba1ef | 1968 | D_ASSERT(device, device->ldev == NULL); |
b30ab791 AG |
1969 | device->ldev = nbc; |
1970 | device->resync = resync_lru; | |
1971 | device->rs_plan_s = new_plan; | |
b411b363 PR |
1972 | nbc = NULL; |
1973 | resync_lru = NULL; | |
daeda1cc | 1974 | new_disk_conf = NULL; |
9958c857 | 1975 | new_plan = NULL; |
b411b363 | 1976 | |
1ec317d3 | 1977 | drbd_resync_after_changed(device); |
f6ba8636 | 1978 | drbd_bump_write_ordering(device->resource, device->ldev, WO_BDEV_FLUSH); |
28bc3b8c | 1979 | unlock_all_resources(); |
b411b363 | 1980 | |
b30ab791 AG |
1981 | if (drbd_md_test_flag(device->ldev, MDF_CRASHED_PRIMARY)) |
1982 | set_bit(CRASHED_PRIMARY, &device->flags); | |
b411b363 | 1983 | else |
b30ab791 | 1984 | clear_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1985 | |
b30ab791 | 1986 | if (drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && |
6bbf53ca | 1987 | !(device->state.role == R_PRIMARY && device->resource->susp_nod)) |
b30ab791 | 1988 | set_bit(CRASHED_PRIMARY, &device->flags); |
b411b363 | 1989 | |
b30ab791 AG |
1990 | device->send_cnt = 0; |
1991 | device->recv_cnt = 0; | |
1992 | device->read_cnt = 0; | |
1993 | device->writ_cnt = 0; | |
b411b363 | 1994 | |
9104d31a | 1995 | drbd_reconsider_queue_parameters(device, device->ldev, NULL); |
b411b363 PR |
1996 | |
1997 | /* If I am currently not R_PRIMARY, | |
1998 | * but meta data primary indicator is set, | |
1999 | * I just now recover from a hard crash, | |
2000 | * and have been R_PRIMARY before that crash. | |
2001 | * | |
2002 | * Now, if I had no connection before that crash | |
2003 | * (have been degraded R_PRIMARY), chances are that | |
2004 | * I won't find my peer now either. | |
2005 | * | |
2006 | * In that case, and _only_ in that case, | |
2007 | * we use the degr-wfc-timeout instead of the default, | |
2008 | * so we can automatically recover from a crash of a | |
2009 | * degraded but active "cluster" after a certain timeout. | |
2010 | */ | |
b30ab791 AG |
2011 | clear_bit(USE_DEGR_WFC_T, &device->flags); |
2012 | if (device->state.role != R_PRIMARY && | |
2013 | drbd_md_test_flag(device->ldev, MDF_PRIMARY_IND) && | |
2014 | !drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND)) | |
2015 | set_bit(USE_DEGR_WFC_T, &device->flags); | |
b411b363 | 2016 | |
b30ab791 | 2017 | dd = drbd_determine_dev_size(device, 0, NULL); |
d752b269 | 2018 | if (dd <= DS_ERROR) { |
b411b363 PR |
2019 | retcode = ERR_NOMEM_BITMAP; |
2020 | goto force_diskless_dec; | |
e96c9633 | 2021 | } else if (dd == DS_GREW) |
b30ab791 | 2022 | set_bit(RESYNC_AFTER_NEG, &device->flags); |
b411b363 | 2023 | |
b30ab791 AG |
2024 | if (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) || |
2025 | (test_bit(CRASHED_PRIMARY, &device->flags) && | |
2026 | drbd_md_test_flag(device->ldev, MDF_AL_DISABLED))) { | |
d0180171 | 2027 | drbd_info(device, "Assuming that all blocks are out of sync " |
b411b363 | 2028 | "(aka FullSync)\n"); |
b30ab791 | 2029 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
20ceb2b2 | 2030 | "set_n_write from attaching", BM_LOCKED_MASK)) { |
b411b363 PR |
2031 | retcode = ERR_IO_MD_DISK; |
2032 | goto force_diskless_dec; | |
2033 | } | |
2034 | } else { | |
b30ab791 | 2035 | if (drbd_bitmap_io(device, &drbd_bm_read, |
22ab6a30 | 2036 | "read from attaching", BM_LOCKED_MASK)) { |
19f843aa LE |
2037 | retcode = ERR_IO_MD_DISK; |
2038 | goto force_diskless_dec; | |
2039 | } | |
b411b363 PR |
2040 | } |
2041 | ||
b30ab791 AG |
2042 | if (_drbd_bm_total_weight(device) == drbd_bm_bits(device)) |
2043 | drbd_suspend_al(device); /* IO is still suspended here... */ | |
0778286a | 2044 | |
0500813f | 2045 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 2046 | os = drbd_read_state(device); |
78bae59b | 2047 | ns = os; |
b411b363 PR |
2048 | /* If MDF_CONSISTENT is not set go into inconsistent state, |
2049 | otherwise investigate MDF_WasUpToDate... | |
2050 | If MDF_WAS_UP_TO_DATE is not set go into D_OUTDATED disk state, | |
2051 | otherwise into D_CONSISTENT state. | |
2052 | */ | |
b30ab791 AG |
2053 | if (drbd_md_test_flag(device->ldev, MDF_CONSISTENT)) { |
2054 | if (drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE)) | |
b411b363 PR |
2055 | ns.disk = D_CONSISTENT; |
2056 | else | |
2057 | ns.disk = D_OUTDATED; | |
2058 | } else { | |
2059 | ns.disk = D_INCONSISTENT; | |
2060 | } | |
2061 | ||
b30ab791 | 2062 | if (drbd_md_test_flag(device->ldev, MDF_PEER_OUT_DATED)) |
b411b363 PR |
2063 | ns.pdsk = D_OUTDATED; |
2064 | ||
daeda1cc PR |
2065 | rcu_read_lock(); |
2066 | if (ns.disk == D_CONSISTENT && | |
b30ab791 | 2067 | (ns.pdsk == D_OUTDATED || rcu_dereference(device->ldev->disk_conf)->fencing == FP_DONT_CARE)) |
b411b363 PR |
2068 | ns.disk = D_UP_TO_DATE; |
2069 | ||
2070 | /* All tests on MDF_PRIMARY_IND, MDF_CONNECTED_IND, | |
2071 | MDF_CONSISTENT and MDF_WAS_UP_TO_DATE must happen before | |
2072 | this point, because drbd_request_state() modifies these | |
2073 | flags. */ | |
2074 | ||
b30ab791 AG |
2075 | if (rcu_dereference(device->ldev->disk_conf)->al_updates) |
2076 | device->ldev->md.flags &= ~MDF_AL_DISABLED; | |
9a51ab1c | 2077 | else |
b30ab791 | 2078 | device->ldev->md.flags |= MDF_AL_DISABLED; |
9a51ab1c PR |
2079 | |
2080 | rcu_read_unlock(); | |
2081 | ||
b411b363 PR |
2082 | /* In case we are C_CONNECTED postpone any decision on the new disk |
2083 | state after the negotiation phase. */ | |
b30ab791 AG |
2084 | if (device->state.conn == C_CONNECTED) { |
2085 | device->new_state_tmp.i = ns.i; | |
b411b363 PR |
2086 | ns.i = os.i; |
2087 | ns.disk = D_NEGOTIATING; | |
dc66c74d PR |
2088 | |
2089 | /* We expect to receive up-to-date UUIDs soon. | |
2090 | To avoid a race in receive_state, free p_uuid while | |
2091 | holding req_lock. I.e. atomic with the state change */ | |
b30ab791 AG |
2092 | kfree(device->p_uuid); |
2093 | device->p_uuid = NULL; | |
b411b363 PR |
2094 | } |
2095 | ||
b30ab791 | 2096 | rv = _drbd_set_state(device, ns, CS_VERBOSE, NULL); |
0500813f | 2097 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
2098 | |
2099 | if (rv < SS_SUCCESS) | |
2100 | goto force_diskless_dec; | |
2101 | ||
b30ab791 | 2102 | mod_timer(&device->request_timer, jiffies + HZ); |
cdfda633 | 2103 | |
b30ab791 AG |
2104 | if (device->state.role == R_PRIMARY) |
2105 | device->ldev->md.uuid[UI_CURRENT] |= (u64)1; | |
b411b363 | 2106 | else |
b30ab791 | 2107 | device->ldev->md.uuid[UI_CURRENT] &= ~(u64)1; |
b411b363 | 2108 | |
b30ab791 AG |
2109 | drbd_md_mark_dirty(device); |
2110 | drbd_md_sync(device); | |
b411b363 | 2111 | |
b30ab791 AG |
2112 | kobject_uevent(&disk_to_dev(device->vdisk)->kobj, KOBJ_CHANGE); |
2113 | put_ldev(device); | |
44a4d551 | 2114 | conn_reconfig_done(connection); |
9e276872 | 2115 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2116 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2117 | return 0; |
2118 | ||
2119 | force_diskless_dec: | |
b30ab791 | 2120 | put_ldev(device); |
b411b363 | 2121 | force_diskless: |
b30ab791 AG |
2122 | drbd_force_state(device, NS(disk, D_DISKLESS)); |
2123 | drbd_md_sync(device); | |
b411b363 | 2124 | fail: |
44a4d551 | 2125 | conn_reconfig_done(connection); |
b411b363 | 2126 | if (nbc) { |
63a7c8ad LE |
2127 | close_backing_dev(device, nbc->md_bdev, nbc->md_bdev != nbc->backing_bdev); |
2128 | close_backing_dev(device, nbc->backing_bdev, true); | |
b411b363 PR |
2129 | kfree(nbc); |
2130 | } | |
daeda1cc | 2131 | kfree(new_disk_conf); |
b411b363 | 2132 | lc_destroy(resync_lru); |
9958c857 | 2133 | kfree(new_plan); |
9e276872 | 2134 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
40cbf085 | 2135 | finish: |
a910b123 | 2136 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2137 | return 0; |
2138 | } | |
2139 | ||
b30ab791 | 2140 | static int adm_detach(struct drbd_device *device, int force) |
b411b363 | 2141 | { |
19f83c76 | 2142 | enum drbd_state_rv retcode; |
6434f404 | 2143 | void *buffer; |
9a0d9d03 | 2144 | int ret; |
02ee8f95 | 2145 | |
cdfda633 | 2146 | if (force) { |
b30ab791 AG |
2147 | set_bit(FORCE_DETACH, &device->flags); |
2148 | drbd_force_state(device, NS(disk, D_FAILED)); | |
cdfda633 | 2149 | retcode = SS_SUCCESS; |
02ee8f95 PR |
2150 | goto out; |
2151 | } | |
2152 | ||
b30ab791 | 2153 | drbd_suspend_io(device); /* so no-one is stuck in drbd_al_begin_io */ |
6434f404 LE |
2154 | buffer = drbd_md_get_buffer(device, __func__); /* make sure there is no in-flight meta-data IO */ |
2155 | if (buffer) { | |
2156 | retcode = drbd_request_state(device, NS(disk, D_FAILED)); | |
2157 | drbd_md_put_buffer(device); | |
2158 | } else /* already <= D_FAILED */ | |
2159 | retcode = SS_NOTHING_TO_DO; | |
9a0d9d03 | 2160 | /* D_FAILED will transition to DISKLESS. */ |
05a72772 | 2161 | drbd_resume_io(device); |
b30ab791 AG |
2162 | ret = wait_event_interruptible(device->misc_wait, |
2163 | device->state.disk != D_FAILED); | |
9b2f61ae | 2164 | if ((int)retcode == (int)SS_IS_DISKLESS) |
9a0d9d03 LE |
2165 | retcode = SS_NOTHING_TO_DO; |
2166 | if (ret) | |
2167 | retcode = ERR_INTR; | |
02ee8f95 | 2168 | out: |
85f75dd7 | 2169 | return retcode; |
b411b363 PR |
2170 | } |
2171 | ||
82f59cc6 LE |
2172 | /* Detaching the disk is a process in multiple stages. First we need to lock |
2173 | * out application IO, in-flight IO, IO stuck in drbd_al_begin_io. | |
2174 | * Then we transition to D_DISKLESS, and wait for put_ldev() to return all | |
2175 | * internal references as well. | |
2176 | * Only then we have finally detached. */ | |
3b98c0c2 | 2177 | int drbd_adm_detach(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2178 | { |
a910b123 | 2179 | struct drbd_config_context adm_ctx; |
116676ca | 2180 | enum drbd_ret_code retcode; |
cdfda633 PR |
2181 | struct detach_parms parms = { }; |
2182 | int err; | |
b411b363 | 2183 | |
a910b123 | 2184 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2185 | if (!adm_ctx.reply_skb) |
2186 | return retcode; | |
2187 | if (retcode != NO_ERROR) | |
2188 | goto out; | |
b411b363 | 2189 | |
cdfda633 PR |
2190 | if (info->attrs[DRBD_NLA_DETACH_PARMS]) { |
2191 | err = detach_parms_from_attrs(&parms, info); | |
2192 | if (err) { | |
2193 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2194 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
cdfda633 PR |
2195 | goto out; |
2196 | } | |
b411b363 PR |
2197 | } |
2198 | ||
9e276872 | 2199 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 2200 | retcode = adm_detach(adm_ctx.device, parms.force_detach); |
9e276872 | 2201 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2202 | out: |
a910b123 | 2203 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2204 | return 0; |
2205 | } | |
b411b363 | 2206 | |
bde89a9e | 2207 | static bool conn_resync_running(struct drbd_connection *connection) |
f399002e | 2208 | { |
c06ece6b | 2209 | struct drbd_peer_device *peer_device; |
695d08fa | 2210 | bool rv = false; |
f399002e LE |
2211 | int vnr; |
2212 | ||
695d08fa | 2213 | rcu_read_lock(); |
c06ece6b AG |
2214 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2215 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2216 | if (device->state.conn == C_SYNC_SOURCE || |
2217 | device->state.conn == C_SYNC_TARGET || | |
2218 | device->state.conn == C_PAUSED_SYNC_S || | |
2219 | device->state.conn == C_PAUSED_SYNC_T) { | |
695d08fa PR |
2220 | rv = true; |
2221 | break; | |
2222 | } | |
b411b363 | 2223 | } |
695d08fa | 2224 | rcu_read_unlock(); |
b411b363 | 2225 | |
695d08fa | 2226 | return rv; |
f399002e | 2227 | } |
47ff2d0a | 2228 | |
bde89a9e | 2229 | static bool conn_ov_running(struct drbd_connection *connection) |
f399002e | 2230 | { |
c06ece6b | 2231 | struct drbd_peer_device *peer_device; |
695d08fa | 2232 | bool rv = false; |
f399002e LE |
2233 | int vnr; |
2234 | ||
695d08fa | 2235 | rcu_read_lock(); |
c06ece6b AG |
2236 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) { |
2237 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2238 | if (device->state.conn == C_VERIFY_S || |
2239 | device->state.conn == C_VERIFY_T) { | |
695d08fa PR |
2240 | rv = true; |
2241 | break; | |
47ff2d0a PR |
2242 | } |
2243 | } | |
695d08fa | 2244 | rcu_read_unlock(); |
b411b363 | 2245 | |
695d08fa | 2246 | return rv; |
f399002e | 2247 | } |
422028b1 | 2248 | |
cd64397c | 2249 | static enum drbd_ret_code |
270eb5c9 | 2250 | _check_net_options(struct drbd_connection *connection, struct net_conf *old_net_conf, struct net_conf *new_net_conf) |
cd64397c | 2251 | { |
c06ece6b | 2252 | struct drbd_peer_device *peer_device; |
cd64397c | 2253 | int i; |
b411b363 | 2254 | |
270eb5c9 AG |
2255 | if (old_net_conf && connection->cstate == C_WF_REPORT_PARAMS && connection->agreed_pro_version < 100) { |
2256 | if (new_net_conf->wire_protocol != old_net_conf->wire_protocol) | |
dcb20d1a | 2257 | return ERR_NEED_APV_100; |
b411b363 | 2258 | |
270eb5c9 | 2259 | if (new_net_conf->two_primaries != old_net_conf->two_primaries) |
dcb20d1a PR |
2260 | return ERR_NEED_APV_100; |
2261 | ||
270eb5c9 | 2262 | if (strcmp(new_net_conf->integrity_alg, old_net_conf->integrity_alg)) |
dcb20d1a | 2263 | return ERR_NEED_APV_100; |
b411b363 PR |
2264 | } |
2265 | ||
270eb5c9 | 2266 | if (!new_net_conf->two_primaries && |
bde89a9e AG |
2267 | conn_highest_role(connection) == R_PRIMARY && |
2268 | conn_highest_peer(connection) == R_PRIMARY) | |
dcb20d1a | 2269 | return ERR_NEED_ALLOW_TWO_PRI; |
b411b363 | 2270 | |
270eb5c9 AG |
2271 | if (new_net_conf->two_primaries && |
2272 | (new_net_conf->wire_protocol != DRBD_PROT_C)) | |
cd64397c PR |
2273 | return ERR_NOT_PROTO_C; |
2274 | ||
c06ece6b AG |
2275 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2276 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2277 | if (get_ldev(device)) { |
2278 | enum drbd_fencing_p fp = rcu_dereference(device->ldev->disk_conf)->fencing; | |
2279 | put_ldev(device); | |
270eb5c9 | 2280 | if (new_net_conf->wire_protocol == DRBD_PROT_A && fp == FP_STONITH) |
cd64397c | 2281 | return ERR_STONITH_AND_PROT_A; |
b411b363 | 2282 | } |
270eb5c9 | 2283 | if (device->state.role == R_PRIMARY && new_net_conf->discard_my_data) |
eb12010e | 2284 | return ERR_DISCARD_IMPOSSIBLE; |
b411b363 PR |
2285 | } |
2286 | ||
270eb5c9 | 2287 | if (new_net_conf->on_congestion != OC_BLOCK && new_net_conf->wire_protocol != DRBD_PROT_A) |
cd64397c | 2288 | return ERR_CONG_NOT_PROTO_A; |
b411b363 | 2289 | |
cd64397c PR |
2290 | return NO_ERROR; |
2291 | } | |
b411b363 | 2292 | |
44ed167d | 2293 | static enum drbd_ret_code |
270eb5c9 | 2294 | check_net_options(struct drbd_connection *connection, struct net_conf *new_net_conf) |
44ed167d PR |
2295 | { |
2296 | static enum drbd_ret_code rv; | |
c06ece6b | 2297 | struct drbd_peer_device *peer_device; |
44ed167d | 2298 | int i; |
b411b363 | 2299 | |
44ed167d | 2300 | rcu_read_lock(); |
270eb5c9 | 2301 | rv = _check_net_options(connection, rcu_dereference(connection->net_conf), new_net_conf); |
44ed167d | 2302 | rcu_read_unlock(); |
b411b363 | 2303 | |
179e20b8 | 2304 | /* connection->peer_devices protected by genl_lock() here */ |
c06ece6b AG |
2305 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2306 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2307 | if (!device->bitmap) { |
2308 | if (drbd_bm_init(device)) | |
44ed167d | 2309 | return ERR_NOMEM; |
b411b363 PR |
2310 | } |
2311 | } | |
2312 | ||
44ed167d PR |
2313 | return rv; |
2314 | } | |
b411b363 | 2315 | |
0fd0ea06 | 2316 | struct crypto { |
9534d671 HX |
2317 | struct crypto_ahash *verify_tfm; |
2318 | struct crypto_ahash *csums_tfm; | |
2319 | struct crypto_shash *cram_hmac_tfm; | |
2320 | struct crypto_ahash *integrity_tfm; | |
0fd0ea06 | 2321 | }; |
b411b363 | 2322 | |
0fd0ea06 | 2323 | static int |
9534d671 | 2324 | alloc_shash(struct crypto_shash **tfm, char *tfm_name, int err_alg) |
0fd0ea06 PR |
2325 | { |
2326 | if (!tfm_name[0]) | |
2327 | return NO_ERROR; | |
b411b363 | 2328 | |
9534d671 HX |
2329 | *tfm = crypto_alloc_shash(tfm_name, 0, 0); |
2330 | if (IS_ERR(*tfm)) { | |
2331 | *tfm = NULL; | |
2332 | return err_alg; | |
2333 | } | |
2334 | ||
2335 | return NO_ERROR; | |
2336 | } | |
2337 | ||
2338 | static int | |
2339 | alloc_ahash(struct crypto_ahash **tfm, char *tfm_name, int err_alg) | |
2340 | { | |
2341 | if (!tfm_name[0]) | |
2342 | return NO_ERROR; | |
2343 | ||
2344 | *tfm = crypto_alloc_ahash(tfm_name, 0, CRYPTO_ALG_ASYNC); | |
0fd0ea06 PR |
2345 | if (IS_ERR(*tfm)) { |
2346 | *tfm = NULL; | |
2347 | return err_alg; | |
b411b363 | 2348 | } |
b411b363 | 2349 | |
0fd0ea06 PR |
2350 | return NO_ERROR; |
2351 | } | |
b411b363 | 2352 | |
0fd0ea06 | 2353 | static enum drbd_ret_code |
270eb5c9 | 2354 | alloc_crypto(struct crypto *crypto, struct net_conf *new_net_conf) |
0fd0ea06 PR |
2355 | { |
2356 | char hmac_name[CRYPTO_MAX_ALG_NAME]; | |
2357 | enum drbd_ret_code rv; | |
0fd0ea06 | 2358 | |
9534d671 HX |
2359 | rv = alloc_ahash(&crypto->csums_tfm, new_net_conf->csums_alg, |
2360 | ERR_CSUMS_ALG); | |
0fd0ea06 PR |
2361 | if (rv != NO_ERROR) |
2362 | return rv; | |
9534d671 HX |
2363 | rv = alloc_ahash(&crypto->verify_tfm, new_net_conf->verify_alg, |
2364 | ERR_VERIFY_ALG); | |
0fd0ea06 PR |
2365 | if (rv != NO_ERROR) |
2366 | return rv; | |
9534d671 HX |
2367 | rv = alloc_ahash(&crypto->integrity_tfm, new_net_conf->integrity_alg, |
2368 | ERR_INTEGRITY_ALG); | |
0fd0ea06 PR |
2369 | if (rv != NO_ERROR) |
2370 | return rv; | |
270eb5c9 | 2371 | if (new_net_conf->cram_hmac_alg[0] != 0) { |
0fd0ea06 | 2372 | snprintf(hmac_name, CRYPTO_MAX_ALG_NAME, "hmac(%s)", |
270eb5c9 | 2373 | new_net_conf->cram_hmac_alg); |
b411b363 | 2374 | |
9534d671 HX |
2375 | rv = alloc_shash(&crypto->cram_hmac_tfm, hmac_name, |
2376 | ERR_AUTH_ALG); | |
b411b363 PR |
2377 | } |
2378 | ||
0fd0ea06 PR |
2379 | return rv; |
2380 | } | |
b411b363 | 2381 | |
0fd0ea06 PR |
2382 | static void free_crypto(struct crypto *crypto) |
2383 | { | |
9534d671 HX |
2384 | crypto_free_shash(crypto->cram_hmac_tfm); |
2385 | crypto_free_ahash(crypto->integrity_tfm); | |
2386 | crypto_free_ahash(crypto->csums_tfm); | |
2387 | crypto_free_ahash(crypto->verify_tfm); | |
0fd0ea06 | 2388 | } |
b411b363 | 2389 | |
f399002e LE |
2390 | int drbd_adm_net_opts(struct sk_buff *skb, struct genl_info *info) |
2391 | { | |
a910b123 | 2392 | struct drbd_config_context adm_ctx; |
f399002e | 2393 | enum drbd_ret_code retcode; |
bde89a9e | 2394 | struct drbd_connection *connection; |
270eb5c9 | 2395 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
f399002e LE |
2396 | int err; |
2397 | int ovr; /* online verify running */ | |
2398 | int rsr; /* re-sync running */ | |
0fd0ea06 | 2399 | struct crypto crypto = { }; |
b411b363 | 2400 | |
a910b123 | 2401 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
f399002e LE |
2402 | if (!adm_ctx.reply_skb) |
2403 | return retcode; | |
2404 | if (retcode != NO_ERROR) | |
9e276872 | 2405 | goto finish; |
b411b363 | 2406 | |
bde89a9e | 2407 | connection = adm_ctx.connection; |
9e276872 | 2408 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2409 | |
270eb5c9 AG |
2410 | new_net_conf = kzalloc(sizeof(struct net_conf), GFP_KERNEL); |
2411 | if (!new_net_conf) { | |
f399002e LE |
2412 | retcode = ERR_NOMEM; |
2413 | goto out; | |
2414 | } | |
b411b363 | 2415 | |
bde89a9e | 2416 | conn_reconfig_start(connection); |
b411b363 | 2417 | |
bde89a9e | 2418 | mutex_lock(&connection->data.mutex); |
0500813f | 2419 | mutex_lock(&connection->resource->conf_update); |
270eb5c9 | 2420 | old_net_conf = connection->net_conf; |
2561b9c1 | 2421 | |
270eb5c9 | 2422 | if (!old_net_conf) { |
a910b123 | 2423 | drbd_msg_put_info(adm_ctx.reply_skb, "net conf missing, try connect"); |
f399002e | 2424 | retcode = ERR_INVALID_REQUEST; |
2561b9c1 PR |
2425 | goto fail; |
2426 | } | |
2427 | ||
270eb5c9 | 2428 | *new_net_conf = *old_net_conf; |
5979e361 | 2429 | if (should_set_defaults(info)) |
270eb5c9 | 2430 | set_net_conf_defaults(new_net_conf); |
f399002e | 2431 | |
270eb5c9 | 2432 | err = net_conf_from_attrs_for_change(new_net_conf, info); |
c75b9b10 | 2433 | if (err && err != -ENOMSG) { |
f399002e | 2434 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2435 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
f399002e | 2436 | goto fail; |
2561b9c1 | 2437 | } |
b411b363 | 2438 | |
270eb5c9 | 2439 | retcode = check_net_options(connection, new_net_conf); |
cd64397c PR |
2440 | if (retcode != NO_ERROR) |
2441 | goto fail; | |
b411b363 | 2442 | |
f399002e | 2443 | /* re-sync running */ |
bde89a9e | 2444 | rsr = conn_resync_running(connection); |
270eb5c9 | 2445 | if (rsr && strcmp(new_net_conf->csums_alg, old_net_conf->csums_alg)) { |
f399002e | 2446 | retcode = ERR_CSUMS_RESYNC_RUNNING; |
91fd4dad | 2447 | goto fail; |
b411b363 PR |
2448 | } |
2449 | ||
f399002e | 2450 | /* online verify running */ |
bde89a9e | 2451 | ovr = conn_ov_running(connection); |
270eb5c9 | 2452 | if (ovr && strcmp(new_net_conf->verify_alg, old_net_conf->verify_alg)) { |
0fd0ea06 | 2453 | retcode = ERR_VERIFY_RUNNING; |
b411b363 | 2454 | goto fail; |
f399002e | 2455 | } |
b411b363 | 2456 | |
270eb5c9 | 2457 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 | 2458 | if (retcode != NO_ERROR) |
b411b363 | 2459 | goto fail; |
f399002e | 2460 | |
270eb5c9 | 2461 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
f399002e LE |
2462 | |
2463 | if (!rsr) { | |
9534d671 | 2464 | crypto_free_ahash(connection->csums_tfm); |
bde89a9e | 2465 | connection->csums_tfm = crypto.csums_tfm; |
0fd0ea06 | 2466 | crypto.csums_tfm = NULL; |
f399002e LE |
2467 | } |
2468 | if (!ovr) { | |
9534d671 | 2469 | crypto_free_ahash(connection->verify_tfm); |
bde89a9e | 2470 | connection->verify_tfm = crypto.verify_tfm; |
0fd0ea06 | 2471 | crypto.verify_tfm = NULL; |
b411b363 PR |
2472 | } |
2473 | ||
9534d671 | 2474 | crypto_free_ahash(connection->integrity_tfm); |
bde89a9e AG |
2475 | connection->integrity_tfm = crypto.integrity_tfm; |
2476 | if (connection->cstate >= C_WF_REPORT_PARAMS && connection->agreed_pro_version >= 100) | |
2477 | /* Do this without trying to take connection->data.mutex again. */ | |
2478 | __drbd_send_protocol(connection, P_PROTOCOL_UPDATE); | |
0fd0ea06 | 2479 | |
9534d671 | 2480 | crypto_free_shash(connection->cram_hmac_tfm); |
bde89a9e | 2481 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; |
0fd0ea06 | 2482 | |
0500813f | 2483 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2484 | mutex_unlock(&connection->data.mutex); |
91fd4dad | 2485 | synchronize_rcu(); |
270eb5c9 | 2486 | kfree(old_net_conf); |
91fd4dad | 2487 | |
69a22773 AG |
2488 | if (connection->cstate >= C_WF_REPORT_PARAMS) { |
2489 | struct drbd_peer_device *peer_device; | |
2490 | int vnr; | |
2491 | ||
2492 | idr_for_each_entry(&connection->peer_devices, peer_device, vnr) | |
2493 | drbd_send_sync_param(peer_device); | |
2494 | } | |
f399002e | 2495 | |
91fd4dad PR |
2496 | goto done; |
2497 | ||
b411b363 | 2498 | fail: |
0500813f | 2499 | mutex_unlock(&connection->resource->conf_update); |
bde89a9e | 2500 | mutex_unlock(&connection->data.mutex); |
0fd0ea06 | 2501 | free_crypto(&crypto); |
270eb5c9 | 2502 | kfree(new_net_conf); |
91fd4dad | 2503 | done: |
bde89a9e | 2504 | conn_reconfig_done(connection); |
f399002e | 2505 | out: |
9e276872 LE |
2506 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2507 | finish: | |
a910b123 | 2508 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2509 | return 0; |
2510 | } | |
2511 | ||
a2972846 AG |
2512 | static void connection_to_info(struct connection_info *info, |
2513 | struct drbd_connection *connection) | |
2514 | { | |
2515 | info->conn_connection_state = connection->cstate; | |
2516 | info->conn_role = conn_highest_peer(connection); | |
2517 | } | |
2518 | ||
2519 | static void peer_device_to_info(struct peer_device_info *info, | |
2520 | struct drbd_peer_device *peer_device) | |
2521 | { | |
2522 | struct drbd_device *device = peer_device->device; | |
2523 | ||
2524 | info->peer_repl_state = | |
2525 | max_t(enum drbd_conns, C_WF_REPORT_PARAMS, device->state.conn); | |
2526 | info->peer_disk_state = device->state.pdsk; | |
2527 | info->peer_resync_susp_user = device->state.user_isp; | |
2528 | info->peer_resync_susp_peer = device->state.peer_isp; | |
2529 | info->peer_resync_susp_dependency = device->state.aftr_isp; | |
2530 | } | |
2531 | ||
3b98c0c2 | 2532 | int drbd_adm_connect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2533 | { |
a2972846 AG |
2534 | struct connection_info connection_info; |
2535 | enum drbd_notification_type flags; | |
2536 | unsigned int peer_devices = 0; | |
a910b123 | 2537 | struct drbd_config_context adm_ctx; |
c06ece6b | 2538 | struct drbd_peer_device *peer_device; |
270eb5c9 | 2539 | struct net_conf *old_net_conf, *new_net_conf = NULL; |
0fd0ea06 | 2540 | struct crypto crypto = { }; |
77c556f6 | 2541 | struct drbd_resource *resource; |
bde89a9e | 2542 | struct drbd_connection *connection; |
3b98c0c2 LE |
2543 | enum drbd_ret_code retcode; |
2544 | int i; | |
2545 | int err; | |
b411b363 | 2546 | |
a910b123 | 2547 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
b411b363 | 2548 | |
3b98c0c2 LE |
2549 | if (!adm_ctx.reply_skb) |
2550 | return retcode; | |
2551 | if (retcode != NO_ERROR) | |
2552 | goto out; | |
089c075d | 2553 | if (!(adm_ctx.my_addr && adm_ctx.peer_addr)) { |
a910b123 | 2554 | drbd_msg_put_info(adm_ctx.reply_skb, "connection endpoint(s) missing"); |
089c075d AG |
2555 | retcode = ERR_INVALID_REQUEST; |
2556 | goto out; | |
2557 | } | |
b411b363 | 2558 | |
089c075d AG |
2559 | /* No need for _rcu here. All reconfiguration is |
2560 | * strictly serialized on genl_lock(). We are protected against | |
2561 | * concurrent reconfiguration/addition/deletion */ | |
77c556f6 AG |
2562 | for_each_resource(resource, &drbd_resources) { |
2563 | for_each_connection(connection, resource) { | |
2564 | if (nla_len(adm_ctx.my_addr) == connection->my_addr_len && | |
2565 | !memcmp(nla_data(adm_ctx.my_addr), &connection->my_addr, | |
2566 | connection->my_addr_len)) { | |
2567 | retcode = ERR_LOCAL_ADDR; | |
2568 | goto out; | |
2569 | } | |
b411b363 | 2570 | |
77c556f6 AG |
2571 | if (nla_len(adm_ctx.peer_addr) == connection->peer_addr_len && |
2572 | !memcmp(nla_data(adm_ctx.peer_addr), &connection->peer_addr, | |
2573 | connection->peer_addr_len)) { | |
2574 | retcode = ERR_PEER_ADDR; | |
2575 | goto out; | |
2576 | } | |
089c075d | 2577 | } |
b411b363 PR |
2578 | } |
2579 | ||
9e276872 | 2580 | mutex_lock(&adm_ctx.resource->adm_mutex); |
3ab706fe | 2581 | connection = first_connection(adm_ctx.resource); |
bde89a9e | 2582 | conn_reconfig_start(connection); |
b411b363 | 2583 | |
bde89a9e | 2584 | if (connection->cstate > C_STANDALONE) { |
b411b363 | 2585 | retcode = ERR_NET_CONFIGURED; |
b411b363 PR |
2586 | goto fail; |
2587 | } | |
2588 | ||
a209b4ae | 2589 | /* allocation not in the IO path, drbdsetup / netlink process context */ |
270eb5c9 AG |
2590 | new_net_conf = kzalloc(sizeof(*new_net_conf), GFP_KERNEL); |
2591 | if (!new_net_conf) { | |
b411b363 | 2592 | retcode = ERR_NOMEM; |
b411b363 PR |
2593 | goto fail; |
2594 | } | |
2595 | ||
270eb5c9 | 2596 | set_net_conf_defaults(new_net_conf); |
b411b363 | 2597 | |
270eb5c9 | 2598 | err = net_conf_from_attrs(new_net_conf, info); |
25e40932 | 2599 | if (err && err != -ENOMSG) { |
b411b363 | 2600 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2601 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2602 | goto fail; |
2603 | } | |
2604 | ||
270eb5c9 | 2605 | retcode = check_net_options(connection, new_net_conf); |
cd64397c | 2606 | if (retcode != NO_ERROR) |
422028b1 | 2607 | goto fail; |
b411b363 | 2608 | |
270eb5c9 | 2609 | retcode = alloc_crypto(&crypto, new_net_conf); |
0fd0ea06 PR |
2610 | if (retcode != NO_ERROR) |
2611 | goto fail; | |
b411b363 | 2612 | |
270eb5c9 | 2613 | ((char *)new_net_conf->shared_secret)[SHARED_SECRET_MAX-1] = 0; |
7b4e4d31 | 2614 | |
b5043c5e | 2615 | drbd_flush_workqueue(&connection->sender_work); |
b411b363 | 2616 | |
0500813f | 2617 | mutex_lock(&adm_ctx.resource->conf_update); |
270eb5c9 AG |
2618 | old_net_conf = connection->net_conf; |
2619 | if (old_net_conf) { | |
b411b363 | 2620 | retcode = ERR_NET_CONFIGURED; |
0500813f | 2621 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 PR |
2622 | goto fail; |
2623 | } | |
270eb5c9 | 2624 | rcu_assign_pointer(connection->net_conf, new_net_conf); |
b411b363 | 2625 | |
bde89a9e AG |
2626 | conn_free_crypto(connection); |
2627 | connection->cram_hmac_tfm = crypto.cram_hmac_tfm; | |
2628 | connection->integrity_tfm = crypto.integrity_tfm; | |
2629 | connection->csums_tfm = crypto.csums_tfm; | |
2630 | connection->verify_tfm = crypto.verify_tfm; | |
b411b363 | 2631 | |
bde89a9e AG |
2632 | connection->my_addr_len = nla_len(adm_ctx.my_addr); |
2633 | memcpy(&connection->my_addr, nla_data(adm_ctx.my_addr), connection->my_addr_len); | |
2634 | connection->peer_addr_len = nla_len(adm_ctx.peer_addr); | |
2635 | memcpy(&connection->peer_addr, nla_data(adm_ctx.peer_addr), connection->peer_addr_len); | |
b411b363 | 2636 | |
a2972846 AG |
2637 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2638 | peer_devices++; | |
2639 | } | |
2640 | ||
2641 | connection_to_info(&connection_info, connection); | |
2642 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2643 | mutex_lock(¬ification_mutex); | |
2644 | notify_connection_state(NULL, 0, connection, &connection_info, NOTIFY_CREATE | flags); | |
2645 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
2646 | struct peer_device_info peer_device_info; | |
2647 | ||
2648 | peer_device_to_info(&peer_device_info, peer_device); | |
2649 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
2650 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, NOTIFY_CREATE | flags); | |
2651 | } | |
2652 | mutex_unlock(¬ification_mutex); | |
0500813f | 2653 | mutex_unlock(&adm_ctx.resource->conf_update); |
b411b363 | 2654 | |
695d08fa | 2655 | rcu_read_lock(); |
c06ece6b AG |
2656 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { |
2657 | struct drbd_device *device = peer_device->device; | |
b30ab791 AG |
2658 | device->send_cnt = 0; |
2659 | device->recv_cnt = 0; | |
b411b363 | 2660 | } |
695d08fa | 2661 | rcu_read_unlock(); |
b411b363 | 2662 | |
bde89a9e | 2663 | retcode = conn_request_state(connection, NS(conn, C_UNCONNECTED), CS_VERBOSE); |
b411b363 | 2664 | |
bde89a9e | 2665 | conn_reconfig_done(connection); |
9e276872 | 2666 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
a910b123 | 2667 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2668 | return 0; |
b411b363 | 2669 | |
b411b363 | 2670 | fail: |
0fd0ea06 | 2671 | free_crypto(&crypto); |
270eb5c9 | 2672 | kfree(new_net_conf); |
b411b363 | 2673 | |
bde89a9e | 2674 | conn_reconfig_done(connection); |
9e276872 | 2675 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 2676 | out: |
a910b123 | 2677 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2678 | return 0; |
2679 | } | |
2680 | ||
bde89a9e | 2681 | static enum drbd_state_rv conn_try_disconnect(struct drbd_connection *connection, bool force) |
85f75dd7 LE |
2682 | { |
2683 | enum drbd_state_rv rv; | |
85f75dd7 | 2684 | |
bde89a9e | 2685 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2686 | force ? CS_HARD : 0); |
85f75dd7 LE |
2687 | |
2688 | switch (rv) { | |
2689 | case SS_NOTHING_TO_DO: | |
f3dfa40a | 2690 | break; |
85f75dd7 LE |
2691 | case SS_ALREADY_STANDALONE: |
2692 | return SS_SUCCESS; | |
2693 | case SS_PRIMARY_NOP: | |
2694 | /* Our state checking code wants to see the peer outdated. */ | |
bde89a9e | 2695 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, pdsk, D_OUTDATED), 0); |
2bd5ed5d PR |
2696 | |
2697 | if (rv == SS_OUTDATE_WO_CONN) /* lost connection before graceful disconnect succeeded */ | |
bde89a9e | 2698 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), CS_VERBOSE); |
2bd5ed5d | 2699 | |
85f75dd7 LE |
2700 | break; |
2701 | case SS_CW_FAILED_BY_PEER: | |
2702 | /* The peer probably wants to see us outdated. */ | |
bde89a9e | 2703 | rv = conn_request_state(connection, NS2(conn, C_DISCONNECTING, |
85f75dd7 LE |
2704 | disk, D_OUTDATED), 0); |
2705 | if (rv == SS_IS_DISKLESS || rv == SS_LOWER_THAN_OUTDATED) { | |
bde89a9e | 2706 | rv = conn_request_state(connection, NS(conn, C_DISCONNECTING), |
f3dfa40a | 2707 | CS_HARD); |
b411b363 | 2708 | } |
85f75dd7 LE |
2709 | break; |
2710 | default:; | |
2711 | /* no special handling necessary */ | |
2712 | } | |
2713 | ||
f3dfa40a LE |
2714 | if (rv >= SS_SUCCESS) { |
2715 | enum drbd_state_rv rv2; | |
2716 | /* No one else can reconfigure the network while I am here. | |
2717 | * The state handling only uses drbd_thread_stop_nowait(), | |
2718 | * we want to really wait here until the receiver is no more. | |
2719 | */ | |
9693da23 | 2720 | drbd_thread_stop(&connection->receiver); |
f3dfa40a LE |
2721 | |
2722 | /* Race breaker. This additional state change request may be | |
2723 | * necessary, if this was a forced disconnect during a receiver | |
2724 | * restart. We may have "killed" the receiver thread just | |
8fe60551 | 2725 | * after drbd_receiver() returned. Typically, we should be |
f3dfa40a LE |
2726 | * C_STANDALONE already, now, and this becomes a no-op. |
2727 | */ | |
bde89a9e | 2728 | rv2 = conn_request_state(connection, NS(conn, C_STANDALONE), |
f3dfa40a LE |
2729 | CS_VERBOSE | CS_HARD); |
2730 | if (rv2 < SS_SUCCESS) | |
1ec861eb | 2731 | drbd_err(connection, |
f3dfa40a LE |
2732 | "unexpected rv2=%d in conn_try_disconnect()\n", |
2733 | rv2); | |
a2972846 AG |
2734 | /* Unlike in DRBD 9, the state engine has generated |
2735 | * NOTIFY_DESTROY events before clearing connection->net_conf. */ | |
b411b363 | 2736 | } |
85f75dd7 LE |
2737 | return rv; |
2738 | } | |
b411b363 | 2739 | |
3b98c0c2 | 2740 | int drbd_adm_disconnect(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2741 | { |
a910b123 | 2742 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2743 | struct disconnect_parms parms; |
bde89a9e | 2744 | struct drbd_connection *connection; |
85f75dd7 | 2745 | enum drbd_state_rv rv; |
3b98c0c2 LE |
2746 | enum drbd_ret_code retcode; |
2747 | int err; | |
2561b9c1 | 2748 | |
a910b123 | 2749 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_CONNECTION); |
3b98c0c2 LE |
2750 | if (!adm_ctx.reply_skb) |
2751 | return retcode; | |
2752 | if (retcode != NO_ERROR) | |
2561b9c1 | 2753 | goto fail; |
b411b363 | 2754 | |
bde89a9e | 2755 | connection = adm_ctx.connection; |
3b98c0c2 LE |
2756 | memset(&parms, 0, sizeof(parms)); |
2757 | if (info->attrs[DRBD_NLA_DISCONNECT_PARMS]) { | |
f399002e | 2758 | err = disconnect_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
2759 | if (err) { |
2760 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 2761 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2762 | goto fail; |
2763 | } | |
2764 | } | |
2765 | ||
9e276872 | 2766 | mutex_lock(&adm_ctx.resource->adm_mutex); |
bde89a9e | 2767 | rv = conn_try_disconnect(connection, parms.force_disconnect); |
85f75dd7 | 2768 | if (rv < SS_SUCCESS) |
f3dfa40a LE |
2769 | retcode = rv; /* FIXME: Type mismatch. */ |
2770 | else | |
2771 | retcode = NO_ERROR; | |
9e276872 | 2772 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2773 | fail: |
a910b123 | 2774 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2775 | return 0; |
2776 | } | |
2777 | ||
b30ab791 | 2778 | void resync_after_online_grow(struct drbd_device *device) |
b411b363 PR |
2779 | { |
2780 | int iass; /* I am sync source */ | |
2781 | ||
d0180171 | 2782 | drbd_info(device, "Resync of new storage after online grow\n"); |
b30ab791 AG |
2783 | if (device->state.role != device->state.peer) |
2784 | iass = (device->state.role == R_PRIMARY); | |
b411b363 | 2785 | else |
a6b32bc3 | 2786 | iass = test_bit(RESOLVE_CONFLICTS, &first_peer_device(device)->connection->flags); |
b411b363 PR |
2787 | |
2788 | if (iass) | |
b30ab791 | 2789 | drbd_start_resync(device, C_SYNC_SOURCE); |
b411b363 | 2790 | else |
b30ab791 | 2791 | _drbd_request_state(device, NS(conn, C_WF_SYNC_UUID), CS_VERBOSE + CS_SERIALIZE); |
b411b363 PR |
2792 | } |
2793 | ||
3b98c0c2 | 2794 | int drbd_adm_resize(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2795 | { |
a910b123 | 2796 | struct drbd_config_context adm_ctx; |
daeda1cc | 2797 | struct disk_conf *old_disk_conf, *new_disk_conf = NULL; |
3b98c0c2 | 2798 | struct resize_parms rs; |
b30ab791 | 2799 | struct drbd_device *device; |
3b98c0c2 | 2800 | enum drbd_ret_code retcode; |
b411b363 | 2801 | enum determine_dev_size dd; |
d752b269 | 2802 | bool change_al_layout = false; |
6495d2c6 | 2803 | enum dds_flags ddsf; |
daeda1cc | 2804 | sector_t u_size; |
3b98c0c2 | 2805 | int err; |
b411b363 | 2806 | |
a910b123 | 2807 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2808 | if (!adm_ctx.reply_skb) |
2809 | return retcode; | |
2810 | if (retcode != NO_ERROR) | |
9e276872 | 2811 | goto finish; |
3b98c0c2 | 2812 | |
9e276872 | 2813 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
2814 | device = adm_ctx.device; |
2815 | if (!get_ldev(device)) { | |
d752b269 PR |
2816 | retcode = ERR_NO_DISK; |
2817 | goto fail; | |
2818 | } | |
2819 | ||
3b98c0c2 | 2820 | memset(&rs, 0, sizeof(struct resize_parms)); |
b30ab791 AG |
2821 | rs.al_stripes = device->ldev->md.al_stripes; |
2822 | rs.al_stripe_size = device->ldev->md.al_stripe_size_4k * 4; | |
3b98c0c2 | 2823 | if (info->attrs[DRBD_NLA_RESIZE_PARMS]) { |
f399002e | 2824 | err = resize_parms_from_attrs(&rs, info); |
b411b363 | 2825 | if (err) { |
3b98c0c2 | 2826 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2827 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
d752b269 | 2828 | goto fail_ldev; |
b411b363 PR |
2829 | } |
2830 | } | |
2831 | ||
b30ab791 | 2832 | if (device->state.conn > C_CONNECTED) { |
b411b363 | 2833 | retcode = ERR_RESIZE_RESYNC; |
d752b269 | 2834 | goto fail_ldev; |
b411b363 | 2835 | } |
b411b363 | 2836 | |
b30ab791 AG |
2837 | if (device->state.role == R_SECONDARY && |
2838 | device->state.peer == R_SECONDARY) { | |
b411b363 | 2839 | retcode = ERR_NO_PRIMARY; |
d752b269 | 2840 | goto fail_ldev; |
b411b363 | 2841 | } |
b411b363 | 2842 | |
a6b32bc3 | 2843 | if (rs.no_resync && first_peer_device(device)->connection->agreed_pro_version < 93) { |
6495d2c6 | 2844 | retcode = ERR_NEED_APV_93; |
9bcd2521 | 2845 | goto fail_ldev; |
6495d2c6 PR |
2846 | } |
2847 | ||
daeda1cc | 2848 | rcu_read_lock(); |
b30ab791 | 2849 | u_size = rcu_dereference(device->ldev->disk_conf)->disk_size; |
daeda1cc PR |
2850 | rcu_read_unlock(); |
2851 | if (u_size != (sector_t)rs.resize_size) { | |
2852 | new_disk_conf = kmalloc(sizeof(struct disk_conf), GFP_KERNEL); | |
2853 | if (!new_disk_conf) { | |
778f271d | 2854 | retcode = ERR_NOMEM; |
9bcd2521 | 2855 | goto fail_ldev; |
778f271d PR |
2856 | } |
2857 | } | |
2858 | ||
b30ab791 AG |
2859 | if (device->ldev->md.al_stripes != rs.al_stripes || |
2860 | device->ldev->md.al_stripe_size_4k != rs.al_stripe_size / 4) { | |
d752b269 PR |
2861 | u32 al_size_k = rs.al_stripes * rs.al_stripe_size; |
2862 | ||
2863 | if (al_size_k > (16 * 1024 * 1024)) { | |
2864 | retcode = ERR_MD_LAYOUT_TOO_BIG; | |
2865 | goto fail_ldev; | |
2866 | } | |
2867 | ||
2868 | if (al_size_k < MD_32kB_SECT/2) { | |
2869 | retcode = ERR_MD_LAYOUT_TOO_SMALL; | |
2870 | goto fail_ldev; | |
2871 | } | |
2872 | ||
cdc6af8d | 2873 | if (device->state.conn != C_CONNECTED && !rs.resize_force) { |
d752b269 PR |
2874 | retcode = ERR_MD_LAYOUT_CONNECTED; |
2875 | goto fail_ldev; | |
2876 | } | |
2877 | ||
2878 | change_al_layout = true; | |
2879 | } | |
2880 | ||
b30ab791 AG |
2881 | if (device->ldev->known_size != drbd_get_capacity(device->ldev->backing_bdev)) |
2882 | device->ldev->known_size = drbd_get_capacity(device->ldev->backing_bdev); | |
b411b363 | 2883 | |
daeda1cc | 2884 | if (new_disk_conf) { |
0500813f | 2885 | mutex_lock(&device->resource->conf_update); |
b30ab791 | 2886 | old_disk_conf = device->ldev->disk_conf; |
daeda1cc PR |
2887 | *new_disk_conf = *old_disk_conf; |
2888 | new_disk_conf->disk_size = (sector_t)rs.resize_size; | |
b30ab791 | 2889 | rcu_assign_pointer(device->ldev->disk_conf, new_disk_conf); |
0500813f | 2890 | mutex_unlock(&device->resource->conf_update); |
daeda1cc PR |
2891 | synchronize_rcu(); |
2892 | kfree(old_disk_conf); | |
70644786 | 2893 | new_disk_conf = NULL; |
b411b363 PR |
2894 | } |
2895 | ||
6495d2c6 | 2896 | ddsf = (rs.resize_force ? DDSF_FORCED : 0) | (rs.no_resync ? DDSF_NO_RESYNC : 0); |
b30ab791 AG |
2897 | dd = drbd_determine_dev_size(device, ddsf, change_al_layout ? &rs : NULL); |
2898 | drbd_md_sync(device); | |
2899 | put_ldev(device); | |
e96c9633 | 2900 | if (dd == DS_ERROR) { |
b411b363 PR |
2901 | retcode = ERR_NOMEM_BITMAP; |
2902 | goto fail; | |
d752b269 PR |
2903 | } else if (dd == DS_ERROR_SPACE_MD) { |
2904 | retcode = ERR_MD_LAYOUT_NO_FIT; | |
2905 | goto fail; | |
2906 | } else if (dd == DS_ERROR_SHRINK) { | |
2907 | retcode = ERR_IMPLICIT_SHRINK; | |
2908 | goto fail; | |
b411b363 | 2909 | } |
778f271d | 2910 | |
b30ab791 | 2911 | if (device->state.conn == C_CONNECTED) { |
e96c9633 | 2912 | if (dd == DS_GREW) |
b30ab791 | 2913 | set_bit(RESIZE_PENDING, &device->flags); |
b411b363 | 2914 | |
69a22773 AG |
2915 | drbd_send_uuids(first_peer_device(device)); |
2916 | drbd_send_sizes(first_peer_device(device), 1, ddsf); | |
778f271d PR |
2917 | } |
2918 | ||
b411b363 | 2919 | fail: |
9e276872 LE |
2920 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
2921 | finish: | |
a910b123 | 2922 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 | 2923 | return 0; |
b411b363 | 2924 | |
9bcd2521 | 2925 | fail_ldev: |
b30ab791 | 2926 | put_ldev(device); |
70644786 | 2927 | kfree(new_disk_conf); |
9bcd2521 | 2928 | goto fail; |
b411b363 | 2929 | } |
b411b363 | 2930 | |
f399002e | 2931 | int drbd_adm_resource_opts(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2932 | { |
a910b123 | 2933 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 2934 | enum drbd_ret_code retcode; |
b57a1e27 | 2935 | struct res_opts res_opts; |
f399002e | 2936 | int err; |
b411b363 | 2937 | |
a910b123 | 2938 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
2939 | if (!adm_ctx.reply_skb) |
2940 | return retcode; | |
2941 | if (retcode != NO_ERROR) | |
2942 | goto fail; | |
b411b363 | 2943 | |
eb6bea67 | 2944 | res_opts = adm_ctx.resource->res_opts; |
5979e361 | 2945 | if (should_set_defaults(info)) |
b966b5dd | 2946 | set_res_opts_defaults(&res_opts); |
b411b363 | 2947 | |
b57a1e27 | 2948 | err = res_opts_from_attrs(&res_opts, info); |
c75b9b10 | 2949 | if (err && err != -ENOMSG) { |
b411b363 | 2950 | retcode = ERR_MANDATORY_TAG; |
a910b123 | 2951 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
b411b363 PR |
2952 | goto fail; |
2953 | } | |
2954 | ||
9e276872 | 2955 | mutex_lock(&adm_ctx.resource->adm_mutex); |
eb6bea67 | 2956 | err = set_resource_options(adm_ctx.resource, &res_opts); |
afbbfa88 AG |
2957 | if (err) { |
2958 | retcode = ERR_INVALID_REQUEST; | |
2959 | if (err == -ENOMEM) | |
2960 | retcode = ERR_NOMEM; | |
b411b363 | 2961 | } |
9e276872 | 2962 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
b411b363 | 2963 | |
b411b363 | 2964 | fail: |
a910b123 | 2965 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
2966 | return 0; |
2967 | } | |
2968 | ||
3b98c0c2 | 2969 | int drbd_adm_invalidate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 2970 | { |
a910b123 | 2971 | struct drbd_config_context adm_ctx; |
b30ab791 | 2972 | struct drbd_device *device; |
3b98c0c2 LE |
2973 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
2974 | ||
a910b123 | 2975 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
2976 | if (!adm_ctx.reply_skb) |
2977 | return retcode; | |
2978 | if (retcode != NO_ERROR) | |
2979 | goto out; | |
2980 | ||
b30ab791 | 2981 | device = adm_ctx.device; |
8fe39aac PR |
2982 | if (!get_ldev(device)) { |
2983 | retcode = ERR_NO_DISK; | |
2984 | goto out; | |
2985 | } | |
2986 | ||
2987 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 2988 | |
194bfb32 | 2989 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
2990 | * resync just being finished, wait for it before requesting a new resync. |
2991 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
2992 | drbd_suspend_io(device); |
2993 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 2994 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 2995 | |
0b2dafcd PR |
2996 | /* If we happen to be C_STANDALONE R_SECONDARY, just change to |
2997 | * D_INCONSISTENT, and set all bits in the bitmap. Otherwise, | |
2998 | * try to start a resync handshake as sync target for full sync. | |
9376d9f8 | 2999 | */ |
b30ab791 AG |
3000 | if (device->state.conn == C_STANDALONE && device->state.role == R_SECONDARY) { |
3001 | retcode = drbd_request_state(device, NS(disk, D_INCONSISTENT)); | |
0b2dafcd | 3002 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 3003 | if (drbd_bitmap_io(device, &drbd_bmio_set_n_write, |
0b2dafcd PR |
3004 | "set_n_write from invalidate", BM_LOCKED_MASK)) |
3005 | retcode = ERR_IO_MD_DISK; | |
3006 | } | |
3007 | } else | |
b30ab791 AG |
3008 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_T)); |
3009 | drbd_resume_io(device); | |
9e276872 | 3010 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 3011 | put_ldev(device); |
3b98c0c2 | 3012 | out: |
a910b123 | 3013 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3014 | return 0; |
3015 | } | |
3016 | ||
3b98c0c2 LE |
3017 | static int drbd_adm_simple_request_state(struct sk_buff *skb, struct genl_info *info, |
3018 | union drbd_state mask, union drbd_state val) | |
b411b363 | 3019 | { |
a910b123 | 3020 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3021 | enum drbd_ret_code retcode; |
194bfb32 | 3022 | |
a910b123 | 3023 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3024 | if (!adm_ctx.reply_skb) |
3025 | return retcode; | |
3026 | if (retcode != NO_ERROR) | |
3027 | goto out; | |
b411b363 | 3028 | |
9e276872 | 3029 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3030 | retcode = drbd_request_state(adm_ctx.device, mask, val); |
9e276872 | 3031 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3032 | out: |
a910b123 | 3033 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3034 | return 0; |
3035 | } | |
3036 | ||
8fe39aac | 3037 | static int drbd_bmio_set_susp_al(struct drbd_device *device) __must_hold(local) |
0778286a PR |
3038 | { |
3039 | int rv; | |
3040 | ||
b30ab791 AG |
3041 | rv = drbd_bmio_set_n_write(device); |
3042 | drbd_suspend_al(device); | |
0778286a PR |
3043 | return rv; |
3044 | } | |
3045 | ||
3b98c0c2 | 3046 | int drbd_adm_invalidate_peer(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3047 | { |
a910b123 | 3048 | struct drbd_config_context adm_ctx; |
25b0d6c8 | 3049 | int retcode; /* drbd_ret_code, drbd_state_rv */ |
b30ab791 | 3050 | struct drbd_device *device; |
25b0d6c8 | 3051 | |
a910b123 | 3052 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
25b0d6c8 PR |
3053 | if (!adm_ctx.reply_skb) |
3054 | return retcode; | |
3055 | if (retcode != NO_ERROR) | |
3056 | goto out; | |
3057 | ||
b30ab791 | 3058 | device = adm_ctx.device; |
8fe39aac PR |
3059 | if (!get_ldev(device)) { |
3060 | retcode = ERR_NO_DISK; | |
3061 | goto out; | |
3062 | } | |
3063 | ||
3064 | mutex_lock(&adm_ctx.resource->adm_mutex); | |
b411b363 | 3065 | |
194bfb32 | 3066 | /* If there is still bitmap IO pending, probably because of a previous |
7ee1fb93 LE |
3067 | * resync just being finished, wait for it before requesting a new resync. |
3068 | * Also wait for it's after_state_ch(). */ | |
b30ab791 AG |
3069 | drbd_suspend_io(device); |
3070 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
b5043c5e | 3071 | drbd_flush_workqueue(&first_peer_device(device)->connection->sender_work); |
194bfb32 | 3072 | |
0b2dafcd PR |
3073 | /* If we happen to be C_STANDALONE R_PRIMARY, just set all bits |
3074 | * in the bitmap. Otherwise, try to start a resync handshake | |
3075 | * as sync source for full sync. | |
3076 | */ | |
b30ab791 | 3077 | if (device->state.conn == C_STANDALONE && device->state.role == R_PRIMARY) { |
0b2dafcd PR |
3078 | /* The peer will get a resync upon connect anyways. Just make that |
3079 | into a full resync. */ | |
b30ab791 | 3080 | retcode = drbd_request_state(device, NS(pdsk, D_INCONSISTENT)); |
0b2dafcd | 3081 | if (retcode >= SS_SUCCESS) { |
b30ab791 | 3082 | if (drbd_bitmap_io(device, &drbd_bmio_set_susp_al, |
0b2dafcd PR |
3083 | "set_n_write from invalidate_peer", |
3084 | BM_LOCKED_SET_ALLOWED)) | |
3085 | retcode = ERR_IO_MD_DISK; | |
3086 | } | |
3087 | } else | |
b30ab791 AG |
3088 | retcode = drbd_request_state(device, NS(conn, C_STARTING_SYNC_S)); |
3089 | drbd_resume_io(device); | |
9e276872 | 3090 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
8fe39aac | 3091 | put_ldev(device); |
25b0d6c8 | 3092 | out: |
a910b123 | 3093 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3094 | return 0; |
3095 | } | |
3096 | ||
3b98c0c2 | 3097 | int drbd_adm_pause_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3098 | { |
a910b123 | 3099 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 3100 | enum drbd_ret_code retcode; |
b411b363 | 3101 | |
a910b123 | 3102 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3103 | if (!adm_ctx.reply_skb) |
3104 | return retcode; | |
3105 | if (retcode != NO_ERROR) | |
3106 | goto out; | |
b411b363 | 3107 | |
9e276872 | 3108 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 3109 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 1)) == SS_NOTHING_TO_DO) |
3b98c0c2 | 3110 | retcode = ERR_PAUSE_IS_SET; |
9e276872 | 3111 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3112 | out: |
a910b123 | 3113 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3114 | return 0; |
3115 | } | |
3116 | ||
3b98c0c2 | 3117 | int drbd_adm_resume_sync(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3118 | { |
a910b123 | 3119 | struct drbd_config_context adm_ctx; |
da9fbc27 | 3120 | union drbd_dev_state s; |
3b98c0c2 LE |
3121 | enum drbd_ret_code retcode; |
3122 | ||
a910b123 | 3123 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3124 | if (!adm_ctx.reply_skb) |
3125 | return retcode; | |
3126 | if (retcode != NO_ERROR) | |
3127 | goto out; | |
b411b363 | 3128 | |
9e276872 | 3129 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
3130 | if (drbd_request_state(adm_ctx.device, NS(user_isp, 0)) == SS_NOTHING_TO_DO) { |
3131 | s = adm_ctx.device->state; | |
cd88d030 PR |
3132 | if (s.conn == C_PAUSED_SYNC_S || s.conn == C_PAUSED_SYNC_T) { |
3133 | retcode = s.aftr_isp ? ERR_PIC_AFTER_DEP : | |
3134 | s.peer_isp ? ERR_PIC_PEER_DEP : ERR_PAUSE_IS_CLEAR; | |
3135 | } else { | |
3136 | retcode = ERR_PAUSE_IS_CLEAR; | |
3137 | } | |
3138 | } | |
9e276872 | 3139 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3140 | out: |
a910b123 | 3141 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3142 | return 0; |
3143 | } | |
3144 | ||
3b98c0c2 | 3145 | int drbd_adm_suspend_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3146 | { |
3b98c0c2 | 3147 | return drbd_adm_simple_request_state(skb, info, NS(susp, 1)); |
b411b363 PR |
3148 | } |
3149 | ||
3b98c0c2 | 3150 | int drbd_adm_resume_io(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3151 | { |
a910b123 | 3152 | struct drbd_config_context adm_ctx; |
b30ab791 | 3153 | struct drbd_device *device; |
3b98c0c2 LE |
3154 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
3155 | ||
a910b123 | 3156 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3157 | if (!adm_ctx.reply_skb) |
3158 | return retcode; | |
3159 | if (retcode != NO_ERROR) | |
3160 | goto out; | |
3161 | ||
9e276872 | 3162 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 AG |
3163 | device = adm_ctx.device; |
3164 | if (test_bit(NEW_CUR_UUID, &device->flags)) { | |
9fa48269 LE |
3165 | if (get_ldev_if_state(device, D_ATTACHING)) { |
3166 | drbd_uuid_new_current(device); | |
3167 | put_ldev(device); | |
3168 | } else { | |
3169 | /* This is effectively a multi-stage "forced down". | |
3170 | * The NEW_CUR_UUID bit is supposedly only set, if we | |
3171 | * lost the replication connection, and are configured | |
3172 | * to freeze IO and wait for some fence-peer handler. | |
3173 | * So we still don't have a replication connection. | |
3174 | * And now we don't have a local disk either. After | |
3175 | * resume, we will fail all pending and new IO, because | |
3176 | * we don't have any data anymore. Which means we will | |
3177 | * eventually be able to terminate all users of this | |
3178 | * device, and then take it down. By bumping the | |
3179 | * "effective" data uuid, we make sure that you really | |
3180 | * need to tear down before you reconfigure, we will | |
3181 | * the refuse to re-connect or re-attach (because no | |
3182 | * matching real data uuid exists). | |
3183 | */ | |
3184 | u64 val; | |
3185 | get_random_bytes(&val, sizeof(u64)); | |
3186 | drbd_set_ed_uuid(device, val); | |
3187 | drbd_warn(device, "Resumed without access to data; please tear down before attempting to re-configure.\n"); | |
3188 | } | |
b30ab791 | 3189 | clear_bit(NEW_CUR_UUID, &device->flags); |
43a5182c | 3190 | } |
b30ab791 AG |
3191 | drbd_suspend_io(device); |
3192 | retcode = drbd_request_state(device, NS3(susp, 0, susp_nod, 0, susp_fen, 0)); | |
3b98c0c2 | 3193 | if (retcode == SS_SUCCESS) { |
b30ab791 | 3194 | if (device->state.conn < C_CONNECTED) |
a6b32bc3 | 3195 | tl_clear(first_peer_device(device)->connection); |
b30ab791 | 3196 | if (device->state.disk == D_DISKLESS || device->state.disk == D_FAILED) |
a6b32bc3 | 3197 | tl_restart(first_peer_device(device)->connection, FAIL_FROZEN_DISK_IO); |
265be2d0 | 3198 | } |
b30ab791 | 3199 | drbd_resume_io(device); |
9e276872 | 3200 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 3201 | out: |
a910b123 | 3202 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
3203 | return 0; |
3204 | } | |
3205 | ||
3b98c0c2 | 3206 | int drbd_adm_outdate(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3207 | { |
3b98c0c2 | 3208 | return drbd_adm_simple_request_state(skb, info, NS(disk, D_OUTDATED)); |
b411b363 PR |
3209 | } |
3210 | ||
251b8f8e AG |
3211 | static int nla_put_drbd_cfg_context(struct sk_buff *skb, |
3212 | struct drbd_resource *resource, | |
3213 | struct drbd_connection *connection, | |
3214 | struct drbd_device *device) | |
b411b363 | 3215 | { |
543cc10b LE |
3216 | struct nlattr *nla; |
3217 | nla = nla_nest_start(skb, DRBD_NLA_CFG_CONTEXT); | |
3218 | if (!nla) | |
3219 | goto nla_put_failure; | |
251b8f8e AG |
3220 | if (device && |
3221 | nla_put_u32(skb, T_ctx_volume, device->vnr)) | |
26ec9287 | 3222 | goto nla_put_failure; |
f597f6b8 | 3223 | if (nla_put_string(skb, T_ctx_resource_name, resource->name)) |
26ec9287 | 3224 | goto nla_put_failure; |
251b8f8e AG |
3225 | if (connection) { |
3226 | if (connection->my_addr_len && | |
3227 | nla_put(skb, T_ctx_my_addr, connection->my_addr_len, &connection->my_addr)) | |
3228 | goto nla_put_failure; | |
3229 | if (connection->peer_addr_len && | |
3230 | nla_put(skb, T_ctx_peer_addr, connection->peer_addr_len, &connection->peer_addr)) | |
3231 | goto nla_put_failure; | |
3232 | } | |
543cc10b LE |
3233 | nla_nest_end(skb, nla); |
3234 | return 0; | |
b411b363 | 3235 | |
543cc10b LE |
3236 | nla_put_failure: |
3237 | if (nla) | |
3238 | nla_nest_cancel(skb, nla); | |
3239 | return -EMSGSIZE; | |
3240 | } | |
b411b363 | 3241 | |
a55bbd37 AG |
3242 | /* |
3243 | * The generic netlink dump callbacks are called outside the genl_lock(), so | |
3244 | * they cannot use the simple attribute parsing code which uses global | |
3245 | * attribute tables. | |
3246 | */ | |
3247 | static struct nlattr *find_cfg_context_attr(const struct nlmsghdr *nlh, int attr) | |
3248 | { | |
3249 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; | |
3250 | const int maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; | |
3251 | struct nlattr *nla; | |
3252 | ||
3253 | nla = nla_find(nlmsg_attrdata(nlh, hdrlen), nlmsg_attrlen(nlh, hdrlen), | |
3254 | DRBD_NLA_CFG_CONTEXT); | |
3255 | if (!nla) | |
3256 | return NULL; | |
3257 | return drbd_nla_find_nested(maxtype, nla, __nla_type(attr)); | |
3258 | } | |
3259 | ||
3260 | static void resource_to_info(struct resource_info *, struct drbd_resource *); | |
3261 | ||
3262 | int drbd_adm_dump_resources(struct sk_buff *skb, struct netlink_callback *cb) | |
3263 | { | |
3264 | struct drbd_genlmsghdr *dh; | |
3265 | struct drbd_resource *resource; | |
3266 | struct resource_info resource_info; | |
3267 | struct resource_statistics resource_statistics; | |
3268 | int err; | |
3269 | ||
3270 | rcu_read_lock(); | |
3271 | if (cb->args[0]) { | |
3272 | for_each_resource_rcu(resource, &drbd_resources) | |
3273 | if (resource == (struct drbd_resource *)cb->args[0]) | |
3274 | goto found_resource; | |
3275 | err = 0; /* resource was probably deleted */ | |
3276 | goto out; | |
3277 | } | |
3278 | resource = list_entry(&drbd_resources, | |
3279 | struct drbd_resource, resources); | |
3280 | ||
3281 | found_resource: | |
3282 | list_for_each_entry_continue_rcu(resource, &drbd_resources, resources) { | |
3283 | goto put_result; | |
3284 | } | |
3285 | err = 0; | |
3286 | goto out; | |
3287 | ||
3288 | put_result: | |
3289 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3290 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3291 | NLM_F_MULTI, DRBD_ADM_GET_RESOURCES); | |
3292 | err = -ENOMEM; | |
3293 | if (!dh) | |
3294 | goto out; | |
3295 | dh->minor = -1U; | |
3296 | dh->ret_code = NO_ERROR; | |
3297 | err = nla_put_drbd_cfg_context(skb, resource, NULL, NULL); | |
3298 | if (err) | |
3299 | goto out; | |
3300 | err = res_opts_to_skb(skb, &resource->res_opts, !capable(CAP_SYS_ADMIN)); | |
3301 | if (err) | |
3302 | goto out; | |
3303 | resource_to_info(&resource_info, resource); | |
3304 | err = resource_info_to_skb(skb, &resource_info, !capable(CAP_SYS_ADMIN)); | |
3305 | if (err) | |
3306 | goto out; | |
3307 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
3308 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
3309 | if (err) | |
3310 | goto out; | |
3311 | cb->args[0] = (long)resource; | |
3312 | genlmsg_end(skb, dh); | |
3313 | err = 0; | |
3314 | ||
3315 | out: | |
3316 | rcu_read_unlock(); | |
3317 | if (err) | |
3318 | return err; | |
3319 | return skb->len; | |
3320 | } | |
3321 | ||
3322 | static void device_to_statistics(struct device_statistics *s, | |
3323 | struct drbd_device *device) | |
3324 | { | |
3325 | memset(s, 0, sizeof(*s)); | |
3326 | s->dev_upper_blocked = !may_inc_ap_bio(device); | |
3327 | if (get_ldev(device)) { | |
3328 | struct drbd_md *md = &device->ldev->md; | |
3329 | u64 *history_uuids = (u64 *)s->history_uuids; | |
3330 | struct request_queue *q; | |
3331 | int n; | |
3332 | ||
3333 | spin_lock_irq(&md->uuid_lock); | |
3334 | s->dev_current_uuid = md->uuid[UI_CURRENT]; | |
3335 | BUILD_BUG_ON(sizeof(s->history_uuids) < UI_HISTORY_END - UI_HISTORY_START + 1); | |
3336 | for (n = 0; n < UI_HISTORY_END - UI_HISTORY_START + 1; n++) | |
3337 | history_uuids[n] = md->uuid[UI_HISTORY_START + n]; | |
3338 | for (; n < HISTORY_UUIDS; n++) | |
3339 | history_uuids[n] = 0; | |
3340 | s->history_uuids_len = HISTORY_UUIDS; | |
3341 | spin_unlock_irq(&md->uuid_lock); | |
3342 | ||
3343 | s->dev_disk_flags = md->flags; | |
3344 | q = bdev_get_queue(device->ldev->backing_bdev); | |
3345 | s->dev_lower_blocked = | |
3346 | bdi_congested(&q->backing_dev_info, | |
3347 | (1 << WB_async_congested) | | |
3348 | (1 << WB_sync_congested)); | |
3349 | put_ldev(device); | |
3350 | } | |
3351 | s->dev_size = drbd_get_capacity(device->this_bdev); | |
3352 | s->dev_read = device->read_cnt; | |
3353 | s->dev_write = device->writ_cnt; | |
3354 | s->dev_al_writes = device->al_writ_cnt; | |
3355 | s->dev_bm_writes = device->bm_writ_cnt; | |
3356 | s->dev_upper_pending = atomic_read(&device->ap_bio_cnt); | |
3357 | s->dev_lower_pending = atomic_read(&device->local_cnt); | |
3358 | s->dev_al_suspended = test_bit(AL_SUSPENDED, &device->flags); | |
3359 | s->dev_exposed_data_uuid = device->ed_uuid; | |
3360 | } | |
3361 | ||
3362 | static int put_resource_in_arg0(struct netlink_callback *cb, int holder_nr) | |
3363 | { | |
3364 | if (cb->args[0]) { | |
3365 | struct drbd_resource *resource = | |
3366 | (struct drbd_resource *)cb->args[0]; | |
3367 | kref_put(&resource->kref, drbd_destroy_resource); | |
3368 | } | |
3369 | ||
3370 | return 0; | |
3371 | } | |
3372 | ||
3373 | int drbd_adm_dump_devices_done(struct netlink_callback *cb) { | |
3374 | return put_resource_in_arg0(cb, 7); | |
3375 | } | |
3376 | ||
3377 | static void device_to_info(struct device_info *, struct drbd_device *); | |
3378 | ||
3379 | int drbd_adm_dump_devices(struct sk_buff *skb, struct netlink_callback *cb) | |
3380 | { | |
3381 | struct nlattr *resource_filter; | |
3382 | struct drbd_resource *resource; | |
3383 | struct drbd_device *uninitialized_var(device); | |
3384 | int minor, err, retcode; | |
3385 | struct drbd_genlmsghdr *dh; | |
3386 | struct device_info device_info; | |
3387 | struct device_statistics device_statistics; | |
3388 | struct idr *idr_to_search; | |
3389 | ||
3390 | resource = (struct drbd_resource *)cb->args[0]; | |
3391 | if (!cb->args[0] && !cb->args[1]) { | |
3392 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3393 | if (resource_filter) { | |
3394 | retcode = ERR_RES_NOT_KNOWN; | |
3395 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3396 | if (!resource) | |
3397 | goto put_result; | |
3398 | cb->args[0] = (long)resource; | |
3399 | } | |
3400 | } | |
3401 | ||
3402 | rcu_read_lock(); | |
3403 | minor = cb->args[1]; | |
3404 | idr_to_search = resource ? &resource->devices : &drbd_devices; | |
3405 | device = idr_get_next(idr_to_search, &minor); | |
3406 | if (!device) { | |
3407 | err = 0; | |
3408 | goto out; | |
3409 | } | |
3410 | idr_for_each_entry_continue(idr_to_search, device, minor) { | |
3411 | retcode = NO_ERROR; | |
3412 | goto put_result; /* only one iteration */ | |
3413 | } | |
3414 | err = 0; | |
3415 | goto out; /* no more devices */ | |
3416 | ||
3417 | put_result: | |
3418 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3419 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3420 | NLM_F_MULTI, DRBD_ADM_GET_DEVICES); | |
3421 | err = -ENOMEM; | |
3422 | if (!dh) | |
3423 | goto out; | |
3424 | dh->ret_code = retcode; | |
3425 | dh->minor = -1U; | |
3426 | if (retcode == NO_ERROR) { | |
3427 | dh->minor = device->minor; | |
3428 | err = nla_put_drbd_cfg_context(skb, device->resource, NULL, device); | |
3429 | if (err) | |
3430 | goto out; | |
3431 | if (get_ldev(device)) { | |
3432 | struct disk_conf *disk_conf = | |
3433 | rcu_dereference(device->ldev->disk_conf); | |
3434 | ||
3435 | err = disk_conf_to_skb(skb, disk_conf, !capable(CAP_SYS_ADMIN)); | |
3436 | put_ldev(device); | |
3437 | if (err) | |
3438 | goto out; | |
3439 | } | |
3440 | device_to_info(&device_info, device); | |
3441 | err = device_info_to_skb(skb, &device_info, !capable(CAP_SYS_ADMIN)); | |
3442 | if (err) | |
3443 | goto out; | |
3444 | ||
3445 | device_to_statistics(&device_statistics, device); | |
3446 | err = device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
3447 | if (err) | |
3448 | goto out; | |
3449 | cb->args[1] = minor + 1; | |
3450 | } | |
3451 | genlmsg_end(skb, dh); | |
3452 | err = 0; | |
3453 | ||
3454 | out: | |
3455 | rcu_read_unlock(); | |
3456 | if (err) | |
3457 | return err; | |
3458 | return skb->len; | |
3459 | } | |
3460 | ||
3461 | int drbd_adm_dump_connections_done(struct netlink_callback *cb) | |
3462 | { | |
3463 | return put_resource_in_arg0(cb, 6); | |
3464 | } | |
3465 | ||
3466 | enum { SINGLE_RESOURCE, ITERATE_RESOURCES }; | |
3467 | ||
3468 | int drbd_adm_dump_connections(struct sk_buff *skb, struct netlink_callback *cb) | |
3469 | { | |
3470 | struct nlattr *resource_filter; | |
3471 | struct drbd_resource *resource = NULL, *next_resource; | |
3472 | struct drbd_connection *uninitialized_var(connection); | |
3473 | int err = 0, retcode; | |
3474 | struct drbd_genlmsghdr *dh; | |
3475 | struct connection_info connection_info; | |
3476 | struct connection_statistics connection_statistics; | |
3477 | ||
3478 | rcu_read_lock(); | |
3479 | resource = (struct drbd_resource *)cb->args[0]; | |
3480 | if (!cb->args[0]) { | |
3481 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3482 | if (resource_filter) { | |
3483 | retcode = ERR_RES_NOT_KNOWN; | |
3484 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3485 | if (!resource) | |
3486 | goto put_result; | |
3487 | cb->args[0] = (long)resource; | |
3488 | cb->args[1] = SINGLE_RESOURCE; | |
3489 | } | |
3490 | } | |
3491 | if (!resource) { | |
3492 | if (list_empty(&drbd_resources)) | |
3493 | goto out; | |
3494 | resource = list_first_entry(&drbd_resources, struct drbd_resource, resources); | |
3495 | kref_get(&resource->kref); | |
3496 | cb->args[0] = (long)resource; | |
3497 | cb->args[1] = ITERATE_RESOURCES; | |
3498 | } | |
3499 | ||
3500 | next_resource: | |
3501 | rcu_read_unlock(); | |
3502 | mutex_lock(&resource->conf_update); | |
3503 | rcu_read_lock(); | |
3504 | if (cb->args[2]) { | |
3505 | for_each_connection_rcu(connection, resource) | |
3506 | if (connection == (struct drbd_connection *)cb->args[2]) | |
3507 | goto found_connection; | |
3508 | /* connection was probably deleted */ | |
3509 | goto no_more_connections; | |
3510 | } | |
3511 | connection = list_entry(&resource->connections, struct drbd_connection, connections); | |
3512 | ||
3513 | found_connection: | |
3514 | list_for_each_entry_continue_rcu(connection, &resource->connections, connections) { | |
3515 | if (!has_net_conf(connection)) | |
3516 | continue; | |
3517 | retcode = NO_ERROR; | |
3518 | goto put_result; /* only one iteration */ | |
3519 | } | |
3520 | ||
3521 | no_more_connections: | |
3522 | if (cb->args[1] == ITERATE_RESOURCES) { | |
3523 | for_each_resource_rcu(next_resource, &drbd_resources) { | |
3524 | if (next_resource == resource) | |
3525 | goto found_resource; | |
3526 | } | |
3527 | /* resource was probably deleted */ | |
3528 | } | |
3529 | goto out; | |
3530 | ||
3531 | found_resource: | |
3532 | list_for_each_entry_continue_rcu(next_resource, &drbd_resources, resources) { | |
3533 | mutex_unlock(&resource->conf_update); | |
3534 | kref_put(&resource->kref, drbd_destroy_resource); | |
3535 | resource = next_resource; | |
3536 | kref_get(&resource->kref); | |
3537 | cb->args[0] = (long)resource; | |
3538 | cb->args[2] = 0; | |
3539 | goto next_resource; | |
3540 | } | |
3541 | goto out; /* no more resources */ | |
3542 | ||
3543 | put_result: | |
3544 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3545 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3546 | NLM_F_MULTI, DRBD_ADM_GET_CONNECTIONS); | |
3547 | err = -ENOMEM; | |
3548 | if (!dh) | |
3549 | goto out; | |
3550 | dh->ret_code = retcode; | |
3551 | dh->minor = -1U; | |
3552 | if (retcode == NO_ERROR) { | |
3553 | struct net_conf *net_conf; | |
3554 | ||
3555 | err = nla_put_drbd_cfg_context(skb, resource, connection, NULL); | |
3556 | if (err) | |
3557 | goto out; | |
3558 | net_conf = rcu_dereference(connection->net_conf); | |
3559 | if (net_conf) { | |
3560 | err = net_conf_to_skb(skb, net_conf, !capable(CAP_SYS_ADMIN)); | |
3561 | if (err) | |
3562 | goto out; | |
3563 | } | |
3564 | connection_to_info(&connection_info, connection); | |
3565 | err = connection_info_to_skb(skb, &connection_info, !capable(CAP_SYS_ADMIN)); | |
3566 | if (err) | |
3567 | goto out; | |
3568 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
3569 | err = connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
3570 | if (err) | |
3571 | goto out; | |
3572 | cb->args[2] = (long)connection; | |
3573 | } | |
3574 | genlmsg_end(skb, dh); | |
3575 | err = 0; | |
3576 | ||
3577 | out: | |
3578 | rcu_read_unlock(); | |
3579 | if (resource) | |
3580 | mutex_unlock(&resource->conf_update); | |
3581 | if (err) | |
3582 | return err; | |
3583 | return skb->len; | |
3584 | } | |
3585 | ||
3586 | enum mdf_peer_flag { | |
3587 | MDF_PEER_CONNECTED = 1 << 0, | |
3588 | MDF_PEER_OUTDATED = 1 << 1, | |
3589 | MDF_PEER_FENCING = 1 << 2, | |
3590 | MDF_PEER_FULL_SYNC = 1 << 3, | |
3591 | }; | |
3592 | ||
3593 | static void peer_device_to_statistics(struct peer_device_statistics *s, | |
3594 | struct drbd_peer_device *peer_device) | |
3595 | { | |
3596 | struct drbd_device *device = peer_device->device; | |
3597 | ||
3598 | memset(s, 0, sizeof(*s)); | |
3599 | s->peer_dev_received = device->recv_cnt; | |
3600 | s->peer_dev_sent = device->send_cnt; | |
3601 | s->peer_dev_pending = atomic_read(&device->ap_pending_cnt) + | |
3602 | atomic_read(&device->rs_pending_cnt); | |
3603 | s->peer_dev_unacked = atomic_read(&device->unacked_cnt); | |
3604 | s->peer_dev_out_of_sync = drbd_bm_total_weight(device) << (BM_BLOCK_SHIFT - 9); | |
3605 | s->peer_dev_resync_failed = device->rs_failed << (BM_BLOCK_SHIFT - 9); | |
3606 | if (get_ldev(device)) { | |
3607 | struct drbd_md *md = &device->ldev->md; | |
3608 | ||
3609 | spin_lock_irq(&md->uuid_lock); | |
3610 | s->peer_dev_bitmap_uuid = md->uuid[UI_BITMAP]; | |
3611 | spin_unlock_irq(&md->uuid_lock); | |
3612 | s->peer_dev_flags = | |
3613 | (drbd_md_test_flag(device->ldev, MDF_CONNECTED_IND) ? | |
3614 | MDF_PEER_CONNECTED : 0) + | |
3615 | (drbd_md_test_flag(device->ldev, MDF_CONSISTENT) && | |
3616 | !drbd_md_test_flag(device->ldev, MDF_WAS_UP_TO_DATE) ? | |
3617 | MDF_PEER_OUTDATED : 0) + | |
3618 | /* FIXME: MDF_PEER_FENCING? */ | |
3619 | (drbd_md_test_flag(device->ldev, MDF_FULL_SYNC) ? | |
3620 | MDF_PEER_FULL_SYNC : 0); | |
3621 | put_ldev(device); | |
3622 | } | |
3623 | } | |
3624 | ||
3625 | int drbd_adm_dump_peer_devices_done(struct netlink_callback *cb) | |
3626 | { | |
3627 | return put_resource_in_arg0(cb, 9); | |
3628 | } | |
3629 | ||
3630 | int drbd_adm_dump_peer_devices(struct sk_buff *skb, struct netlink_callback *cb) | |
3631 | { | |
3632 | struct nlattr *resource_filter; | |
3633 | struct drbd_resource *resource; | |
3634 | struct drbd_device *uninitialized_var(device); | |
3635 | struct drbd_peer_device *peer_device = NULL; | |
3636 | int minor, err, retcode; | |
3637 | struct drbd_genlmsghdr *dh; | |
3638 | struct idr *idr_to_search; | |
3639 | ||
3640 | resource = (struct drbd_resource *)cb->args[0]; | |
3641 | if (!cb->args[0] && !cb->args[1]) { | |
3642 | resource_filter = find_cfg_context_attr(cb->nlh, T_ctx_resource_name); | |
3643 | if (resource_filter) { | |
3644 | retcode = ERR_RES_NOT_KNOWN; | |
3645 | resource = drbd_find_resource(nla_data(resource_filter)); | |
3646 | if (!resource) | |
3647 | goto put_result; | |
3648 | } | |
3649 | cb->args[0] = (long)resource; | |
3650 | } | |
3651 | ||
3652 | rcu_read_lock(); | |
3653 | minor = cb->args[1]; | |
3654 | idr_to_search = resource ? &resource->devices : &drbd_devices; | |
3655 | device = idr_find(idr_to_search, minor); | |
3656 | if (!device) { | |
3657 | next_device: | |
3658 | minor++; | |
3659 | cb->args[2] = 0; | |
3660 | device = idr_get_next(idr_to_search, &minor); | |
3661 | if (!device) { | |
3662 | err = 0; | |
3663 | goto out; | |
3664 | } | |
3665 | } | |
3666 | if (cb->args[2]) { | |
3667 | for_each_peer_device(peer_device, device) | |
3668 | if (peer_device == (struct drbd_peer_device *)cb->args[2]) | |
3669 | goto found_peer_device; | |
3670 | /* peer device was probably deleted */ | |
3671 | goto next_device; | |
3672 | } | |
3673 | /* Make peer_device point to the list head (not the first entry). */ | |
3674 | peer_device = list_entry(&device->peer_devices, struct drbd_peer_device, peer_devices); | |
3675 | ||
3676 | found_peer_device: | |
3677 | list_for_each_entry_continue_rcu(peer_device, &device->peer_devices, peer_devices) { | |
3678 | if (!has_net_conf(peer_device->connection)) | |
3679 | continue; | |
3680 | retcode = NO_ERROR; | |
3681 | goto put_result; /* only one iteration */ | |
3682 | } | |
3683 | goto next_device; | |
3684 | ||
3685 | put_result: | |
3686 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, | |
3687 | cb->nlh->nlmsg_seq, &drbd_genl_family, | |
3688 | NLM_F_MULTI, DRBD_ADM_GET_PEER_DEVICES); | |
3689 | err = -ENOMEM; | |
3690 | if (!dh) | |
3691 | goto out; | |
3692 | dh->ret_code = retcode; | |
3693 | dh->minor = -1U; | |
3694 | if (retcode == NO_ERROR) { | |
3695 | struct peer_device_info peer_device_info; | |
3696 | struct peer_device_statistics peer_device_statistics; | |
3697 | ||
3698 | dh->minor = minor; | |
3699 | err = nla_put_drbd_cfg_context(skb, device->resource, peer_device->connection, device); | |
3700 | if (err) | |
3701 | goto out; | |
3702 | peer_device_to_info(&peer_device_info, peer_device); | |
3703 | err = peer_device_info_to_skb(skb, &peer_device_info, !capable(CAP_SYS_ADMIN)); | |
3704 | if (err) | |
3705 | goto out; | |
3706 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
3707 | err = peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
3708 | if (err) | |
3709 | goto out; | |
3710 | cb->args[1] = minor; | |
3711 | cb->args[2] = (long)peer_device; | |
3712 | } | |
3713 | genlmsg_end(skb, dh); | |
3714 | err = 0; | |
3715 | ||
3716 | out: | |
3717 | rcu_read_unlock(); | |
3718 | if (err) | |
3719 | return err; | |
3720 | return skb->len; | |
3721 | } | |
251b8f8e AG |
3722 | /* |
3723 | * Return the connection of @resource if @resource has exactly one connection. | |
3724 | */ | |
3725 | static struct drbd_connection *the_only_connection(struct drbd_resource *resource) | |
3726 | { | |
3727 | struct list_head *connections = &resource->connections; | |
3728 | ||
3729 | if (list_empty(connections) || connections->next->next != connections) | |
3730 | return NULL; | |
3731 | return list_first_entry(&resource->connections, struct drbd_connection, connections); | |
3732 | } | |
3733 | ||
8ce953aa | 3734 | static int nla_put_status_info(struct sk_buff *skb, struct drbd_device *device, |
3b98c0c2 | 3735 | const struct sib_info *sib) |
b411b363 | 3736 | { |
251b8f8e | 3737 | struct drbd_resource *resource = device->resource; |
3b98c0c2 LE |
3738 | struct state_info *si = NULL; /* for sizeof(si->member); */ |
3739 | struct nlattr *nla; | |
3740 | int got_ldev; | |
3b98c0c2 LE |
3741 | int err = 0; |
3742 | int exclude_sensitive; | |
3743 | ||
3744 | /* If sib != NULL, this is drbd_bcast_event, which anyone can listen | |
3745 | * to. So we better exclude_sensitive information. | |
3746 | * | |
3747 | * If sib == NULL, this is drbd_adm_get_status, executed synchronously | |
3748 | * in the context of the requesting user process. Exclude sensitive | |
3749 | * information, unless current has superuser. | |
3750 | * | |
3751 | * NOTE: for drbd_adm_get_status_all(), this is a netlink dump, and | |
3752 | * relies on the current implementation of netlink_dump(), which | |
3753 | * executes the dump callback successively from netlink_recvmsg(), | |
3754 | * always in the context of the receiving process */ | |
3755 | exclude_sensitive = sib || !capable(CAP_SYS_ADMIN); | |
3756 | ||
b30ab791 | 3757 | got_ldev = get_ldev(device); |
3b98c0c2 LE |
3758 | |
3759 | /* We need to add connection name and volume number information still. | |
3760 | * Minor number is in drbd_genlmsghdr. */ | |
251b8f8e | 3761 | if (nla_put_drbd_cfg_context(skb, resource, the_only_connection(resource), device)) |
3b98c0c2 | 3762 | goto nla_put_failure; |
3b98c0c2 | 3763 | |
eb6bea67 | 3764 | if (res_opts_to_skb(skb, &device->resource->res_opts, exclude_sensitive)) |
f399002e LE |
3765 | goto nla_put_failure; |
3766 | ||
daeda1cc | 3767 | rcu_read_lock(); |
f9eb7bf4 AG |
3768 | if (got_ldev) { |
3769 | struct disk_conf *disk_conf; | |
44ed167d | 3770 | |
b30ab791 | 3771 | disk_conf = rcu_dereference(device->ldev->disk_conf); |
f9eb7bf4 AG |
3772 | err = disk_conf_to_skb(skb, disk_conf, exclude_sensitive); |
3773 | } | |
3774 | if (!err) { | |
3775 | struct net_conf *nc; | |
3776 | ||
a6b32bc3 | 3777 | nc = rcu_dereference(first_peer_device(device)->connection->net_conf); |
f9eb7bf4 AG |
3778 | if (nc) |
3779 | err = net_conf_to_skb(skb, nc, exclude_sensitive); | |
3780 | } | |
44ed167d PR |
3781 | rcu_read_unlock(); |
3782 | if (err) | |
3783 | goto nla_put_failure; | |
3b98c0c2 | 3784 | |
3b98c0c2 LE |
3785 | nla = nla_nest_start(skb, DRBD_NLA_STATE_INFO); |
3786 | if (!nla) | |
3787 | goto nla_put_failure; | |
26ec9287 | 3788 | if (nla_put_u32(skb, T_sib_reason, sib ? sib->sib_reason : SIB_GET_STATUS_REPLY) || |
b30ab791 | 3789 | nla_put_u32(skb, T_current_state, device->state.i) || |
1dee3f59 ND |
3790 | nla_put_u64_0pad(skb, T_ed_uuid, device->ed_uuid) || |
3791 | nla_put_u64_0pad(skb, T_capacity, | |
3792 | drbd_get_capacity(device->this_bdev)) || | |
3793 | nla_put_u64_0pad(skb, T_send_cnt, device->send_cnt) || | |
3794 | nla_put_u64_0pad(skb, T_recv_cnt, device->recv_cnt) || | |
3795 | nla_put_u64_0pad(skb, T_read_cnt, device->read_cnt) || | |
3796 | nla_put_u64_0pad(skb, T_writ_cnt, device->writ_cnt) || | |
3797 | nla_put_u64_0pad(skb, T_al_writ_cnt, device->al_writ_cnt) || | |
3798 | nla_put_u64_0pad(skb, T_bm_writ_cnt, device->bm_writ_cnt) || | |
b30ab791 AG |
3799 | nla_put_u32(skb, T_ap_bio_cnt, atomic_read(&device->ap_bio_cnt)) || |
3800 | nla_put_u32(skb, T_ap_pending_cnt, atomic_read(&device->ap_pending_cnt)) || | |
3801 | nla_put_u32(skb, T_rs_pending_cnt, atomic_read(&device->rs_pending_cnt))) | |
26ec9287 | 3802 | goto nla_put_failure; |
3b98c0c2 LE |
3803 | |
3804 | if (got_ldev) { | |
39a1aa7f | 3805 | int err; |
b411b363 | 3806 | |
b30ab791 AG |
3807 | spin_lock_irq(&device->ldev->md.uuid_lock); |
3808 | err = nla_put(skb, T_uuids, sizeof(si->uuids), device->ldev->md.uuid); | |
3809 | spin_unlock_irq(&device->ldev->md.uuid_lock); | |
39a1aa7f PR |
3810 | |
3811 | if (err) | |
3812 | goto nla_put_failure; | |
3813 | ||
b30ab791 | 3814 | if (nla_put_u32(skb, T_disk_flags, device->ldev->md.flags) || |
1dee3f59 ND |
3815 | nla_put_u64_0pad(skb, T_bits_total, drbd_bm_bits(device)) || |
3816 | nla_put_u64_0pad(skb, T_bits_oos, | |
3817 | drbd_bm_total_weight(device))) | |
26ec9287 | 3818 | goto nla_put_failure; |
b30ab791 AG |
3819 | if (C_SYNC_SOURCE <= device->state.conn && |
3820 | C_PAUSED_SYNC_T >= device->state.conn) { | |
1dee3f59 ND |
3821 | if (nla_put_u64_0pad(skb, T_bits_rs_total, |
3822 | device->rs_total) || | |
3823 | nla_put_u64_0pad(skb, T_bits_rs_failed, | |
3824 | device->rs_failed)) | |
26ec9287 | 3825 | goto nla_put_failure; |
3b98c0c2 | 3826 | } |
b411b363 | 3827 | } |
b411b363 | 3828 | |
3b98c0c2 LE |
3829 | if (sib) { |
3830 | switch(sib->sib_reason) { | |
3831 | case SIB_SYNC_PROGRESS: | |
3832 | case SIB_GET_STATUS_REPLY: | |
3833 | break; | |
3834 | case SIB_STATE_CHANGE: | |
26ec9287 AG |
3835 | if (nla_put_u32(skb, T_prev_state, sib->os.i) || |
3836 | nla_put_u32(skb, T_new_state, sib->ns.i)) | |
3837 | goto nla_put_failure; | |
3b98c0c2 LE |
3838 | break; |
3839 | case SIB_HELPER_POST: | |
26ec9287 AG |
3840 | if (nla_put_u32(skb, T_helper_exit_code, |
3841 | sib->helper_exit_code)) | |
3842 | goto nla_put_failure; | |
3b98c0c2 LE |
3843 | /* fall through */ |
3844 | case SIB_HELPER_PRE: | |
26ec9287 AG |
3845 | if (nla_put_string(skb, T_helper, sib->helper_name)) |
3846 | goto nla_put_failure; | |
3b98c0c2 LE |
3847 | break; |
3848 | } | |
b411b363 | 3849 | } |
3b98c0c2 | 3850 | nla_nest_end(skb, nla); |
b411b363 | 3851 | |
3b98c0c2 LE |
3852 | if (0) |
3853 | nla_put_failure: | |
3854 | err = -EMSGSIZE; | |
3855 | if (got_ldev) | |
b30ab791 | 3856 | put_ldev(device); |
3b98c0c2 | 3857 | return err; |
b411b363 PR |
3858 | } |
3859 | ||
3b98c0c2 | 3860 | int drbd_adm_get_status(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 3861 | { |
a910b123 | 3862 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
3863 | enum drbd_ret_code retcode; |
3864 | int err; | |
b411b363 | 3865 | |
a910b123 | 3866 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
3867 | if (!adm_ctx.reply_skb) |
3868 | return retcode; | |
3869 | if (retcode != NO_ERROR) | |
3870 | goto out; | |
b411b363 | 3871 | |
b30ab791 | 3872 | err = nla_put_status_info(adm_ctx.reply_skb, adm_ctx.device, NULL); |
3b98c0c2 LE |
3873 | if (err) { |
3874 | nlmsg_free(adm_ctx.reply_skb); | |
3875 | return err; | |
b411b363 | 3876 | } |
3b98c0c2 | 3877 | out: |
a910b123 | 3878 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 3879 | return 0; |
b411b363 PR |
3880 | } |
3881 | ||
4b7a530f | 3882 | static int get_one_status(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 3883 | { |
b30ab791 | 3884 | struct drbd_device *device; |
3b98c0c2 | 3885 | struct drbd_genlmsghdr *dh; |
77c556f6 AG |
3886 | struct drbd_resource *pos = (struct drbd_resource *)cb->args[0]; |
3887 | struct drbd_resource *resource = NULL; | |
77c556f6 | 3888 | struct drbd_resource *tmp; |
543cc10b LE |
3889 | unsigned volume = cb->args[1]; |
3890 | ||
3891 | /* Open coded, deferred, iteration: | |
77c556f6 | 3892 | * for_each_resource_safe(resource, tmp, &drbd_resources) { |
251b8f8e AG |
3893 | * connection = "first connection of resource or undefined"; |
3894 | * idr_for_each_entry(&resource->devices, device, i) { | |
543cc10b LE |
3895 | * ... |
3896 | * } | |
3897 | * } | |
77c556f6 | 3898 | * where resource is cb->args[0]; |
543cc10b LE |
3899 | * and i is cb->args[1]; |
3900 | * | |
71932efc LE |
3901 | * cb->args[2] indicates if we shall loop over all resources, |
3902 | * or just dump all volumes of a single resource. | |
3903 | * | |
3b98c0c2 LE |
3904 | * This may miss entries inserted after this dump started, |
3905 | * or entries deleted before they are reached. | |
543cc10b | 3906 | * |
b30ab791 | 3907 | * We need to make sure the device won't disappear while |
543cc10b LE |
3908 | * we are looking at it, and revalidate our iterators |
3909 | * on each iteration. | |
3910 | */ | |
b411b363 | 3911 | |
05a10ec7 | 3912 | /* synchronize with conn_create()/drbd_destroy_connection() */ |
c141ebda | 3913 | rcu_read_lock(); |
543cc10b | 3914 | /* revalidate iterator position */ |
77c556f6 | 3915 | for_each_resource_rcu(tmp, &drbd_resources) { |
543cc10b LE |
3916 | if (pos == NULL) { |
3917 | /* first iteration */ | |
3918 | pos = tmp; | |
77c556f6 | 3919 | resource = pos; |
543cc10b LE |
3920 | break; |
3921 | } | |
3922 | if (tmp == pos) { | |
77c556f6 | 3923 | resource = pos; |
543cc10b LE |
3924 | break; |
3925 | } | |
b411b363 | 3926 | } |
77c556f6 AG |
3927 | if (resource) { |
3928 | next_resource: | |
251b8f8e AG |
3929 | device = idr_get_next(&resource->devices, &volume); |
3930 | if (!device) { | |
77c556f6 AG |
3931 | /* No more volumes to dump on this resource. |
3932 | * Advance resource iterator. */ | |
3933 | pos = list_entry_rcu(resource->resources.next, | |
3934 | struct drbd_resource, resources); | |
3935 | /* Did we dump any volume of this resource yet? */ | |
543cc10b | 3936 | if (volume != 0) { |
71932efc LE |
3937 | /* If we reached the end of the list, |
3938 | * or only a single resource dump was requested, | |
3939 | * we are done. */ | |
77c556f6 | 3940 | if (&pos->resources == &drbd_resources || cb->args[2]) |
71932efc | 3941 | goto out; |
543cc10b | 3942 | volume = 0; |
77c556f6 AG |
3943 | resource = pos; |
3944 | goto next_resource; | |
543cc10b LE |
3945 | } |
3946 | } | |
3947 | ||
98683650 | 3948 | dh = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, |
3b98c0c2 LE |
3949 | cb->nlh->nlmsg_seq, &drbd_genl_family, |
3950 | NLM_F_MULTI, DRBD_ADM_GET_STATUS); | |
3951 | if (!dh) | |
543cc10b LE |
3952 | goto out; |
3953 | ||
251b8f8e | 3954 | if (!device) { |
bde89a9e | 3955 | /* This is a connection without a single volume. |
367d675d LE |
3956 | * Suprisingly enough, it may have a network |
3957 | * configuration. */ | |
251b8f8e AG |
3958 | struct drbd_connection *connection; |
3959 | ||
543cc10b LE |
3960 | dh->minor = -1U; |
3961 | dh->ret_code = NO_ERROR; | |
251b8f8e AG |
3962 | connection = the_only_connection(resource); |
3963 | if (nla_put_drbd_cfg_context(skb, resource, connection, NULL)) | |
367d675d | 3964 | goto cancel; |
251b8f8e AG |
3965 | if (connection) { |
3966 | struct net_conf *nc; | |
3967 | ||
3968 | nc = rcu_dereference(connection->net_conf); | |
3969 | if (nc && net_conf_to_skb(skb, nc, 1) != 0) | |
3970 | goto cancel; | |
3971 | } | |
367d675d | 3972 | goto done; |
543cc10b | 3973 | } |
b411b363 | 3974 | |
0b0ba1ef | 3975 | D_ASSERT(device, device->vnr == volume); |
251b8f8e | 3976 | D_ASSERT(device, device->resource == resource); |
3b98c0c2 | 3977 | |
b30ab791 | 3978 | dh->minor = device_to_minor(device); |
3b98c0c2 LE |
3979 | dh->ret_code = NO_ERROR; |
3980 | ||
b30ab791 | 3981 | if (nla_put_status_info(skb, device, NULL)) { |
367d675d | 3982 | cancel: |
3b98c0c2 | 3983 | genlmsg_cancel(skb, dh); |
543cc10b | 3984 | goto out; |
3b98c0c2 | 3985 | } |
367d675d | 3986 | done: |
3b98c0c2 | 3987 | genlmsg_end(skb, dh); |
bde89a9e | 3988 | } |
b411b363 | 3989 | |
543cc10b | 3990 | out: |
c141ebda | 3991 | rcu_read_unlock(); |
543cc10b | 3992 | /* where to start the next iteration */ |
bde89a9e | 3993 | cb->args[0] = (long)pos; |
77c556f6 | 3994 | cb->args[1] = (pos == resource) ? volume + 1 : 0; |
b411b363 | 3995 | |
77c556f6 | 3996 | /* No more resources/volumes/minors found results in an empty skb. |
543cc10b | 3997 | * Which will terminate the dump. */ |
3b98c0c2 | 3998 | return skb->len; |
b411b363 PR |
3999 | } |
4000 | ||
71932efc LE |
4001 | /* |
4002 | * Request status of all resources, or of all volumes within a single resource. | |
4003 | * | |
4004 | * This is a dump, as the answer may not fit in a single reply skb otherwise. | |
4005 | * Which means we cannot use the family->attrbuf or other such members, because | |
4006 | * dump is NOT protected by the genl_lock(). During dump, we only have access | |
4007 | * to the incoming skb, and need to opencode "parsing" of the nlattr payload. | |
4008 | * | |
4009 | * Once things are setup properly, we call into get_one_status(). | |
b411b363 | 4010 | */ |
71932efc | 4011 | int drbd_adm_get_status_all(struct sk_buff *skb, struct netlink_callback *cb) |
b411b363 | 4012 | { |
71932efc LE |
4013 | const unsigned hdrlen = GENL_HDRLEN + GENL_MAGIC_FAMILY_HDRSZ; |
4014 | struct nlattr *nla; | |
7c3063cc | 4015 | const char *resource_name; |
4bc76048 | 4016 | struct drbd_resource *resource; |
7c3063cc | 4017 | int maxtype; |
71932efc LE |
4018 | |
4019 | /* Is this a followup call? */ | |
4020 | if (cb->args[0]) { | |
4021 | /* ... of a single resource dump, | |
4022 | * and the resource iterator has been advanced already? */ | |
4023 | if (cb->args[2] && cb->args[2] != cb->args[0]) | |
4024 | return 0; /* DONE. */ | |
4025 | goto dump; | |
4026 | } | |
4027 | ||
4028 | /* First call (from netlink_dump_start). We need to figure out | |
4029 | * which resource(s) the user wants us to dump. */ | |
4030 | nla = nla_find(nlmsg_attrdata(cb->nlh, hdrlen), | |
4031 | nlmsg_attrlen(cb->nlh, hdrlen), | |
4032 | DRBD_NLA_CFG_CONTEXT); | |
4033 | ||
4034 | /* No explicit context given. Dump all. */ | |
4035 | if (!nla) | |
4036 | goto dump; | |
7c3063cc AG |
4037 | maxtype = ARRAY_SIZE(drbd_cfg_context_nl_policy) - 1; |
4038 | nla = drbd_nla_find_nested(maxtype, nla, __nla_type(T_ctx_resource_name)); | |
4039 | if (IS_ERR(nla)) | |
4040 | return PTR_ERR(nla); | |
71932efc LE |
4041 | /* context given, but no name present? */ |
4042 | if (!nla) | |
4043 | return -EINVAL; | |
7c3063cc | 4044 | resource_name = nla_data(nla); |
4bc76048 AG |
4045 | if (!*resource_name) |
4046 | return -ENODEV; | |
4047 | resource = drbd_find_resource(resource_name); | |
4048 | if (!resource) | |
71932efc LE |
4049 | return -ENODEV; |
4050 | ||
4bc76048 | 4051 | kref_put(&resource->kref, drbd_destroy_resource); /* get_one_status() revalidates the resource */ |
0ace9dfa | 4052 | |
71932efc | 4053 | /* prime iterators, and set "filter" mode mark: |
bde89a9e | 4054 | * only dump this connection. */ |
4bc76048 | 4055 | cb->args[0] = (long)resource; |
71932efc | 4056 | /* cb->args[1] = 0; passed in this way. */ |
4bc76048 | 4057 | cb->args[2] = (long)resource; |
71932efc LE |
4058 | |
4059 | dump: | |
4060 | return get_one_status(skb, cb); | |
4061 | } | |
b411b363 | 4062 | |
3b98c0c2 | 4063 | int drbd_adm_get_timeout_type(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4064 | { |
a910b123 | 4065 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
4066 | enum drbd_ret_code retcode; |
4067 | struct timeout_parms tp; | |
4068 | int err; | |
b411b363 | 4069 | |
a910b123 | 4070 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4071 | if (!adm_ctx.reply_skb) |
4072 | return retcode; | |
4073 | if (retcode != NO_ERROR) | |
4074 | goto out; | |
b411b363 | 4075 | |
3b98c0c2 | 4076 | tp.timeout_type = |
b30ab791 AG |
4077 | adm_ctx.device->state.pdsk == D_OUTDATED ? UT_PEER_OUTDATED : |
4078 | test_bit(USE_DEGR_WFC_T, &adm_ctx.device->flags) ? UT_DEGRADED : | |
3b98c0c2 | 4079 | UT_DEFAULT; |
b411b363 | 4080 | |
3b98c0c2 LE |
4081 | err = timeout_parms_to_priv_skb(adm_ctx.reply_skb, &tp); |
4082 | if (err) { | |
4083 | nlmsg_free(adm_ctx.reply_skb); | |
4084 | return err; | |
4085 | } | |
4086 | out: | |
a910b123 | 4087 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4088 | return 0; |
b411b363 PR |
4089 | } |
4090 | ||
3b98c0c2 | 4091 | int drbd_adm_start_ov(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4092 | { |
a910b123 | 4093 | struct drbd_config_context adm_ctx; |
b30ab791 | 4094 | struct drbd_device *device; |
3b98c0c2 | 4095 | enum drbd_ret_code retcode; |
58ffa580 | 4096 | struct start_ov_parms parms; |
b411b363 | 4097 | |
a910b123 | 4098 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4099 | if (!adm_ctx.reply_skb) |
4100 | return retcode; | |
4101 | if (retcode != NO_ERROR) | |
4102 | goto out; | |
873b0d5f | 4103 | |
b30ab791 | 4104 | device = adm_ctx.device; |
58ffa580 LE |
4105 | |
4106 | /* resume from last known position, if possible */ | |
b30ab791 | 4107 | parms.ov_start_sector = device->ov_start_sector; |
58ffa580 | 4108 | parms.ov_stop_sector = ULLONG_MAX; |
3b98c0c2 | 4109 | if (info->attrs[DRBD_NLA_START_OV_PARMS]) { |
f399002e | 4110 | int err = start_ov_parms_from_attrs(&parms, info); |
3b98c0c2 LE |
4111 | if (err) { |
4112 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 4113 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
4114 | goto out; |
4115 | } | |
b411b363 | 4116 | } |
9e276872 LE |
4117 | mutex_lock(&adm_ctx.resource->adm_mutex); |
4118 | ||
58ffa580 | 4119 | /* w_make_ov_request expects position to be aligned */ |
b30ab791 AG |
4120 | device->ov_start_sector = parms.ov_start_sector & ~(BM_SECT_PER_BIT-1); |
4121 | device->ov_stop_sector = parms.ov_stop_sector; | |
873b0d5f LE |
4122 | |
4123 | /* If there is still bitmap IO pending, e.g. previous resync or verify | |
4124 | * just being finished, wait for it before requesting a new resync. */ | |
b30ab791 AG |
4125 | drbd_suspend_io(device); |
4126 | wait_event(device->misc_wait, !test_bit(BITMAP_IO, &device->flags)); | |
4127 | retcode = drbd_request_state(device, NS(conn, C_VERIFY_S)); | |
4128 | drbd_resume_io(device); | |
9e276872 LE |
4129 | |
4130 | mutex_unlock(&adm_ctx.resource->adm_mutex); | |
3b98c0c2 | 4131 | out: |
a910b123 | 4132 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4133 | return 0; |
4134 | } | |
4135 | ||
4136 | ||
3b98c0c2 | 4137 | int drbd_adm_new_c_uuid(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4138 | { |
a910b123 | 4139 | struct drbd_config_context adm_ctx; |
b30ab791 | 4140 | struct drbd_device *device; |
3b98c0c2 | 4141 | enum drbd_ret_code retcode; |
b411b363 PR |
4142 | int skip_initial_sync = 0; |
4143 | int err; | |
3b98c0c2 | 4144 | struct new_c_uuid_parms args; |
b411b363 | 4145 | |
a910b123 | 4146 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4147 | if (!adm_ctx.reply_skb) |
4148 | return retcode; | |
4149 | if (retcode != NO_ERROR) | |
4150 | goto out_nolock; | |
b411b363 | 4151 | |
b30ab791 | 4152 | device = adm_ctx.device; |
3b98c0c2 LE |
4153 | memset(&args, 0, sizeof(args)); |
4154 | if (info->attrs[DRBD_NLA_NEW_C_UUID_PARMS]) { | |
f399002e | 4155 | err = new_c_uuid_parms_from_attrs(&args, info); |
3b98c0c2 LE |
4156 | if (err) { |
4157 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 4158 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
3b98c0c2 LE |
4159 | goto out_nolock; |
4160 | } | |
b411b363 PR |
4161 | } |
4162 | ||
9e276872 | 4163 | mutex_lock(&adm_ctx.resource->adm_mutex); |
b30ab791 | 4164 | mutex_lock(device->state_mutex); /* Protects us against serialized state changes. */ |
b411b363 | 4165 | |
b30ab791 | 4166 | if (!get_ldev(device)) { |
b411b363 PR |
4167 | retcode = ERR_NO_DISK; |
4168 | goto out; | |
4169 | } | |
4170 | ||
4171 | /* this is "skip initial sync", assume to be clean */ | |
a6b32bc3 AG |
4172 | if (device->state.conn == C_CONNECTED && |
4173 | first_peer_device(device)->connection->agreed_pro_version >= 90 && | |
b30ab791 | 4174 | device->ldev->md.uuid[UI_CURRENT] == UUID_JUST_CREATED && args.clear_bm) { |
d0180171 | 4175 | drbd_info(device, "Preparing to skip initial sync\n"); |
b411b363 | 4176 | skip_initial_sync = 1; |
b30ab791 | 4177 | } else if (device->state.conn != C_STANDALONE) { |
b411b363 PR |
4178 | retcode = ERR_CONNECTED; |
4179 | goto out_dec; | |
4180 | } | |
4181 | ||
b30ab791 AG |
4182 | drbd_uuid_set(device, UI_BITMAP, 0); /* Rotate UI_BITMAP to History 1, etc... */ |
4183 | drbd_uuid_new_current(device); /* New current, previous to UI_BITMAP */ | |
b411b363 PR |
4184 | |
4185 | if (args.clear_bm) { | |
b30ab791 | 4186 | err = drbd_bitmap_io(device, &drbd_bmio_clear_n_write, |
20ceb2b2 | 4187 | "clear_n_write from new_c_uuid", BM_LOCKED_MASK); |
b411b363 | 4188 | if (err) { |
d0180171 | 4189 | drbd_err(device, "Writing bitmap failed with %d\n", err); |
b411b363 PR |
4190 | retcode = ERR_IO_MD_DISK; |
4191 | } | |
4192 | if (skip_initial_sync) { | |
69a22773 | 4193 | drbd_send_uuids_skip_initial_sync(first_peer_device(device)); |
b30ab791 AG |
4194 | _drbd_uuid_set(device, UI_BITMAP, 0); |
4195 | drbd_print_uuids(device, "cleared bitmap UUID"); | |
0500813f | 4196 | spin_lock_irq(&device->resource->req_lock); |
b30ab791 | 4197 | _drbd_set_state(_NS2(device, disk, D_UP_TO_DATE, pdsk, D_UP_TO_DATE), |
b411b363 | 4198 | CS_VERBOSE, NULL); |
0500813f | 4199 | spin_unlock_irq(&device->resource->req_lock); |
b411b363 PR |
4200 | } |
4201 | } | |
4202 | ||
b30ab791 | 4203 | drbd_md_sync(device); |
b411b363 | 4204 | out_dec: |
b30ab791 | 4205 | put_ldev(device); |
b411b363 | 4206 | out: |
b30ab791 | 4207 | mutex_unlock(device->state_mutex); |
9e276872 | 4208 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 4209 | out_nolock: |
a910b123 | 4210 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4211 | return 0; |
4212 | } | |
4213 | ||
3b98c0c2 | 4214 | static enum drbd_ret_code |
a910b123 | 4215 | drbd_check_resource_name(struct drbd_config_context *adm_ctx) |
b411b363 | 4216 | { |
a910b123 | 4217 | const char *name = adm_ctx->resource_name; |
3b98c0c2 | 4218 | if (!name || !name[0]) { |
a910b123 | 4219 | drbd_msg_put_info(adm_ctx->reply_skb, "resource name missing"); |
3b98c0c2 | 4220 | return ERR_MANDATORY_TAG; |
b411b363 | 4221 | } |
3b98c0c2 LE |
4222 | /* if we want to use these in sysfs/configfs/debugfs some day, |
4223 | * we must not allow slashes */ | |
4224 | if (strchr(name, '/')) { | |
a910b123 | 4225 | drbd_msg_put_info(adm_ctx->reply_skb, "invalid resource name"); |
3b98c0c2 | 4226 | return ERR_INVALID_REQUEST; |
b411b363 | 4227 | } |
3b98c0c2 | 4228 | return NO_ERROR; |
774b3055 | 4229 | } |
b411b363 | 4230 | |
a2972846 AG |
4231 | static void resource_to_info(struct resource_info *info, |
4232 | struct drbd_resource *resource) | |
4233 | { | |
4234 | info->res_role = conn_highest_role(first_connection(resource)); | |
4235 | info->res_susp = resource->susp; | |
4236 | info->res_susp_nod = resource->susp_nod; | |
4237 | info->res_susp_fen = resource->susp_fen; | |
4238 | } | |
4239 | ||
789c1b62 | 4240 | int drbd_adm_new_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4241 | { |
a2972846 | 4242 | struct drbd_connection *connection; |
a910b123 | 4243 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 4244 | enum drbd_ret_code retcode; |
afbbfa88 AG |
4245 | struct res_opts res_opts; |
4246 | int err; | |
b411b363 | 4247 | |
a910b123 | 4248 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, 0); |
3b98c0c2 LE |
4249 | if (!adm_ctx.reply_skb) |
4250 | return retcode; | |
4251 | if (retcode != NO_ERROR) | |
4252 | goto out; | |
b411b363 | 4253 | |
afbbfa88 AG |
4254 | set_res_opts_defaults(&res_opts); |
4255 | err = res_opts_from_attrs(&res_opts, info); | |
4256 | if (err && err != -ENOMSG) { | |
4257 | retcode = ERR_MANDATORY_TAG; | |
a910b123 | 4258 | drbd_msg_put_info(adm_ctx.reply_skb, from_attrs_err_to_txt(err)); |
afbbfa88 | 4259 | goto out; |
b411b363 PR |
4260 | } |
4261 | ||
a910b123 | 4262 | retcode = drbd_check_resource_name(&adm_ctx); |
3b98c0c2 LE |
4263 | if (retcode != NO_ERROR) |
4264 | goto out; | |
b411b363 | 4265 | |
5c661042 | 4266 | if (adm_ctx.resource) { |
38f19616 LE |
4267 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) { |
4268 | retcode = ERR_INVALID_REQUEST; | |
a910b123 | 4269 | drbd_msg_put_info(adm_ctx.reply_skb, "resource exists"); |
38f19616 LE |
4270 | } |
4271 | /* else: still NO_ERROR */ | |
3b98c0c2 | 4272 | goto out; |
b411b363 | 4273 | } |
b411b363 | 4274 | |
9e276872 | 4275 | /* not yet safe for genl_family.parallel_ops */ |
28bc3b8c | 4276 | mutex_lock(&resources_mutex); |
a2972846 | 4277 | connection = conn_create(adm_ctx.resource_name, &res_opts); |
28bc3b8c | 4278 | mutex_unlock(&resources_mutex); |
a2972846 AG |
4279 | |
4280 | if (connection) { | |
4281 | struct resource_info resource_info; | |
4282 | ||
4283 | mutex_lock(¬ification_mutex); | |
4284 | resource_to_info(&resource_info, connection->resource); | |
4285 | notify_resource_state(NULL, 0, connection->resource, | |
4286 | &resource_info, NOTIFY_CREATE); | |
4287 | mutex_unlock(¬ification_mutex); | |
4288 | } else | |
4289 | retcode = ERR_NOMEM; | |
4290 | ||
3b98c0c2 | 4291 | out: |
a910b123 | 4292 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4293 | return 0; |
b411b363 PR |
4294 | } |
4295 | ||
a2972846 AG |
4296 | static void device_to_info(struct device_info *info, |
4297 | struct drbd_device *device) | |
4298 | { | |
4299 | info->dev_disk_state = device->state.disk; | |
4300 | } | |
4301 | ||
4302 | ||
05a10ec7 | 4303 | int drbd_adm_new_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4304 | { |
a910b123 | 4305 | struct drbd_config_context adm_ctx; |
3b98c0c2 LE |
4306 | struct drbd_genlmsghdr *dh = info->userhdr; |
4307 | enum drbd_ret_code retcode; | |
b411b363 | 4308 | |
a910b123 | 4309 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
4310 | if (!adm_ctx.reply_skb) |
4311 | return retcode; | |
4312 | if (retcode != NO_ERROR) | |
4313 | goto out; | |
b411b363 | 4314 | |
f2257a56 | 4315 | if (dh->minor > MINORMASK) { |
a910b123 | 4316 | drbd_msg_put_info(adm_ctx.reply_skb, "requested minor out of range"); |
3b98c0c2 LE |
4317 | retcode = ERR_INVALID_REQUEST; |
4318 | goto out; | |
b411b363 | 4319 | } |
0c8e36d9 | 4320 | if (adm_ctx.volume > DRBD_VOLUME_MAX) { |
a910b123 | 4321 | drbd_msg_put_info(adm_ctx.reply_skb, "requested volume id out of range"); |
3b98c0c2 LE |
4322 | retcode = ERR_INVALID_REQUEST; |
4323 | goto out; | |
b411b363 | 4324 | } |
b411b363 | 4325 | |
38f19616 | 4326 | /* drbd_adm_prepare made sure already |
a6b32bc3 | 4327 | * that first_peer_device(device)->connection and device->vnr match the request. */ |
b30ab791 | 4328 | if (adm_ctx.device) { |
38f19616 | 4329 | if (info->nlhdr->nlmsg_flags & NLM_F_EXCL) |
179e20b8 | 4330 | retcode = ERR_MINOR_OR_VOLUME_EXISTS; |
38f19616 LE |
4331 | /* else: still NO_ERROR */ |
4332 | goto out; | |
b411b363 | 4333 | } |
38f19616 | 4334 | |
9e276872 | 4335 | mutex_lock(&adm_ctx.resource->adm_mutex); |
a910b123 | 4336 | retcode = drbd_create_device(&adm_ctx, dh->minor); |
a2972846 AG |
4337 | if (retcode == NO_ERROR) { |
4338 | struct drbd_device *device; | |
4339 | struct drbd_peer_device *peer_device; | |
4340 | struct device_info info; | |
4341 | unsigned int peer_devices = 0; | |
4342 | enum drbd_notification_type flags; | |
4343 | ||
4344 | device = minor_to_device(dh->minor); | |
4345 | for_each_peer_device(peer_device, device) { | |
4346 | if (!has_net_conf(peer_device->connection)) | |
4347 | continue; | |
4348 | peer_devices++; | |
4349 | } | |
4350 | ||
4351 | device_to_info(&info, device); | |
4352 | mutex_lock(¬ification_mutex); | |
4353 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
4354 | notify_device_state(NULL, 0, device, &info, NOTIFY_CREATE | flags); | |
4355 | for_each_peer_device(peer_device, device) { | |
4356 | struct peer_device_info peer_device_info; | |
4357 | ||
4358 | if (!has_net_conf(peer_device->connection)) | |
4359 | continue; | |
4360 | peer_device_to_info(&peer_device_info, peer_device); | |
4361 | flags = (peer_devices--) ? NOTIFY_CONTINUES : 0; | |
4362 | notify_peer_device_state(NULL, 0, peer_device, &peer_device_info, | |
4363 | NOTIFY_CREATE | flags); | |
4364 | } | |
4365 | mutex_unlock(¬ification_mutex); | |
4366 | } | |
9e276872 | 4367 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
3b98c0c2 | 4368 | out: |
a910b123 | 4369 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4370 | return 0; |
b411b363 PR |
4371 | } |
4372 | ||
05a10ec7 | 4373 | static enum drbd_ret_code adm_del_minor(struct drbd_device *device) |
b411b363 | 4374 | { |
a2972846 AG |
4375 | struct drbd_peer_device *peer_device; |
4376 | ||
b30ab791 AG |
4377 | if (device->state.disk == D_DISKLESS && |
4378 | /* no need to be device->state.conn == C_STANDALONE && | |
85f75dd7 LE |
4379 | * we may want to delete a minor from a live replication group. |
4380 | */ | |
b30ab791 | 4381 | device->state.role == R_SECONDARY) { |
a2972846 AG |
4382 | struct drbd_connection *connection = |
4383 | first_connection(device->resource); | |
4384 | ||
b30ab791 | 4385 | _drbd_request_state(device, NS(conn, C_WF_REPORT_PARAMS), |
369bea63 | 4386 | CS_VERBOSE + CS_WAIT_COMPLETE); |
a2972846 AG |
4387 | |
4388 | /* If the state engine hasn't stopped the sender thread yet, we | |
4389 | * need to flush the sender work queue before generating the | |
4390 | * DESTROY events here. */ | |
4391 | if (get_t_state(&connection->worker) == RUNNING) | |
4392 | drbd_flush_workqueue(&connection->sender_work); | |
4393 | ||
4394 | mutex_lock(¬ification_mutex); | |
4395 | for_each_peer_device(peer_device, device) { | |
4396 | if (!has_net_conf(peer_device->connection)) | |
4397 | continue; | |
4398 | notify_peer_device_state(NULL, 0, peer_device, NULL, | |
4399 | NOTIFY_DESTROY | NOTIFY_CONTINUES); | |
4400 | } | |
4401 | notify_device_state(NULL, 0, device, NULL, NOTIFY_DESTROY); | |
4402 | mutex_unlock(¬ification_mutex); | |
4403 | ||
f82795d6 | 4404 | drbd_delete_device(device); |
85f75dd7 LE |
4405 | return NO_ERROR; |
4406 | } else | |
4407 | return ERR_MINOR_CONFIGURED; | |
b411b363 PR |
4408 | } |
4409 | ||
05a10ec7 | 4410 | int drbd_adm_del_minor(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4411 | { |
a910b123 | 4412 | struct drbd_config_context adm_ctx; |
3b98c0c2 | 4413 | enum drbd_ret_code retcode; |
b411b363 | 4414 | |
a910b123 | 4415 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_MINOR); |
3b98c0c2 LE |
4416 | if (!adm_ctx.reply_skb) |
4417 | return retcode; | |
4418 | if (retcode != NO_ERROR) | |
4419 | goto out; | |
b411b363 | 4420 | |
9e276872 | 4421 | mutex_lock(&adm_ctx.resource->adm_mutex); |
05a10ec7 | 4422 | retcode = adm_del_minor(adm_ctx.device); |
9e276872 | 4423 | mutex_unlock(&adm_ctx.resource->adm_mutex); |
85f75dd7 | 4424 | out: |
a910b123 | 4425 | drbd_adm_finish(&adm_ctx, info, retcode); |
85f75dd7 | 4426 | return 0; |
b411b363 PR |
4427 | } |
4428 | ||
179e20b8 AG |
4429 | static int adm_del_resource(struct drbd_resource *resource) |
4430 | { | |
4431 | struct drbd_connection *connection; | |
4432 | ||
4433 | for_each_connection(connection, resource) { | |
4434 | if (connection->cstate > C_STANDALONE) | |
4435 | return ERR_NET_CONFIGURED; | |
4436 | } | |
4437 | if (!idr_is_empty(&resource->devices)) | |
4438 | return ERR_RES_IN_USE; | |
4439 | ||
a2972846 AG |
4440 | /* The state engine has stopped the sender thread, so we don't |
4441 | * need to flush the sender work queue before generating the | |
4442 | * DESTROY event here. */ | |
4443 | mutex_lock(¬ification_mutex); | |
4444 | notify_resource_state(NULL, 0, resource, NULL, NOTIFY_DESTROY); | |
4445 | mutex_unlock(¬ification_mutex); | |
4446 | ||
28bc3b8c | 4447 | mutex_lock(&resources_mutex); |
179e20b8 | 4448 | list_del_rcu(&resource->resources); |
28bc3b8c | 4449 | mutex_unlock(&resources_mutex); |
179e20b8 AG |
4450 | /* Make sure all threads have actually stopped: state handling only |
4451 | * does drbd_thread_stop_nowait(). */ | |
4452 | list_for_each_entry(connection, &resource->connections, connections) | |
4453 | drbd_thread_stop(&connection->worker); | |
4454 | synchronize_rcu(); | |
4455 | drbd_free_resource(resource); | |
4456 | return NO_ERROR; | |
4457 | } | |
4458 | ||
85f75dd7 | 4459 | int drbd_adm_down(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4460 | { |
a910b123 | 4461 | struct drbd_config_context adm_ctx; |
b6f85ef9 AG |
4462 | struct drbd_resource *resource; |
4463 | struct drbd_connection *connection; | |
4464 | struct drbd_device *device; | |
f3dfa40a | 4465 | int retcode; /* enum drbd_ret_code rsp. enum drbd_state_rv */ |
85f75dd7 | 4466 | unsigned i; |
b411b363 | 4467 | |
a910b123 | 4468 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
85f75dd7 LE |
4469 | if (!adm_ctx.reply_skb) |
4470 | return retcode; | |
4471 | if (retcode != NO_ERROR) | |
9e276872 | 4472 | goto finish; |
b411b363 | 4473 | |
b6f85ef9 | 4474 | resource = adm_ctx.resource; |
9e276872 | 4475 | mutex_lock(&resource->adm_mutex); |
85f75dd7 | 4476 | /* demote */ |
b6f85ef9 AG |
4477 | for_each_connection(connection, resource) { |
4478 | struct drbd_peer_device *peer_device; | |
4479 | ||
4480 | idr_for_each_entry(&connection->peer_devices, peer_device, i) { | |
4481 | retcode = drbd_set_role(peer_device->device, R_SECONDARY, 0); | |
4482 | if (retcode < SS_SUCCESS) { | |
a910b123 | 4483 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to demote"); |
b6f85ef9 AG |
4484 | goto out; |
4485 | } | |
4486 | } | |
4487 | ||
4488 | retcode = conn_try_disconnect(connection, 0); | |
85f75dd7 | 4489 | if (retcode < SS_SUCCESS) { |
a910b123 | 4490 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to disconnect"); |
c141ebda | 4491 | goto out; |
85f75dd7 | 4492 | } |
b411b363 | 4493 | } |
b411b363 | 4494 | |
85f75dd7 | 4495 | /* detach */ |
b6f85ef9 AG |
4496 | idr_for_each_entry(&resource->devices, device, i) { |
4497 | retcode = adm_detach(device, 0); | |
27012382 | 4498 | if (retcode < SS_SUCCESS || retcode > NO_ERROR) { |
a910b123 | 4499 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to detach"); |
c141ebda | 4500 | goto out; |
85f75dd7 LE |
4501 | } |
4502 | } | |
b411b363 | 4503 | |
85f75dd7 | 4504 | /* delete volumes */ |
b6f85ef9 AG |
4505 | idr_for_each_entry(&resource->devices, device, i) { |
4506 | retcode = adm_del_minor(device); | |
85f75dd7 LE |
4507 | if (retcode != NO_ERROR) { |
4508 | /* "can not happen" */ | |
a910b123 | 4509 | drbd_msg_put_info(adm_ctx.reply_skb, "failed to delete volume"); |
ef356262 | 4510 | goto out; |
85f75dd7 LE |
4511 | } |
4512 | } | |
b411b363 | 4513 | |
179e20b8 | 4514 | retcode = adm_del_resource(resource); |
3b98c0c2 | 4515 | out: |
9e276872 LE |
4516 | mutex_unlock(&resource->adm_mutex); |
4517 | finish: | |
a910b123 | 4518 | drbd_adm_finish(&adm_ctx, info, retcode); |
3b98c0c2 | 4519 | return 0; |
b411b363 PR |
4520 | } |
4521 | ||
789c1b62 | 4522 | int drbd_adm_del_resource(struct sk_buff *skb, struct genl_info *info) |
b411b363 | 4523 | { |
a910b123 | 4524 | struct drbd_config_context adm_ctx; |
77c556f6 | 4525 | struct drbd_resource *resource; |
3b98c0c2 | 4526 | enum drbd_ret_code retcode; |
b411b363 | 4527 | |
a910b123 | 4528 | retcode = drbd_adm_prepare(&adm_ctx, skb, info, DRBD_ADM_NEED_RESOURCE); |
3b98c0c2 LE |
4529 | if (!adm_ctx.reply_skb) |
4530 | return retcode; | |
4531 | if (retcode != NO_ERROR) | |
9e276872 | 4532 | goto finish; |
77c556f6 | 4533 | resource = adm_ctx.resource; |
b411b363 | 4534 | |
179e20b8 AG |
4535 | mutex_lock(&resource->adm_mutex); |
4536 | retcode = adm_del_resource(resource); | |
9e276872 LE |
4537 | mutex_unlock(&resource->adm_mutex); |
4538 | finish: | |
a910b123 | 4539 | drbd_adm_finish(&adm_ctx, info, retcode); |
b411b363 PR |
4540 | return 0; |
4541 | } | |
4542 | ||
b30ab791 | 4543 | void drbd_bcast_event(struct drbd_device *device, const struct sib_info *sib) |
b411b363 | 4544 | { |
3b98c0c2 LE |
4545 | struct sk_buff *msg; |
4546 | struct drbd_genlmsghdr *d_out; | |
4547 | unsigned seq; | |
4548 | int err = -ENOMEM; | |
4549 | ||
4550 | seq = atomic_inc_return(&drbd_genl_seq); | |
4551 | msg = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4552 | if (!msg) | |
4553 | goto failed; | |
4554 | ||
4555 | err = -EMSGSIZE; | |
4556 | d_out = genlmsg_put(msg, 0, seq, &drbd_genl_family, 0, DRBD_EVENT); | |
4557 | if (!d_out) /* cannot happen, but anyways. */ | |
4558 | goto nla_put_failure; | |
b30ab791 | 4559 | d_out->minor = device_to_minor(device); |
6f9b5f84 | 4560 | d_out->ret_code = NO_ERROR; |
3b98c0c2 | 4561 | |
b30ab791 | 4562 | if (nla_put_status_info(msg, device, sib)) |
3b98c0c2 LE |
4563 | goto nla_put_failure; |
4564 | genlmsg_end(msg, d_out); | |
d38f8612 | 4565 | err = drbd_genl_multicast_events(msg, GFP_NOWAIT); |
3b98c0c2 LE |
4566 | /* msg has been consumed or freed in netlink_broadcast() */ |
4567 | if (err && err != -ESRCH) | |
4568 | goto failed; | |
b411b363 | 4569 | |
3b98c0c2 | 4570 | return; |
b411b363 | 4571 | |
3b98c0c2 LE |
4572 | nla_put_failure: |
4573 | nlmsg_free(msg); | |
4574 | failed: | |
d0180171 | 4575 | drbd_err(device, "Error %d while broadcasting event. " |
3b98c0c2 LE |
4576 | "Event seq:%u sib_reason:%u\n", |
4577 | err, seq, sib->sib_reason); | |
b411b363 | 4578 | } |
a2972846 | 4579 | |
a2972846 AG |
4580 | static int nla_put_notification_header(struct sk_buff *msg, |
4581 | enum drbd_notification_type type) | |
4582 | { | |
4583 | struct drbd_notification_header nh = { | |
4584 | .nh_type = type, | |
4585 | }; | |
4586 | ||
4587 | return drbd_notification_header_to_skb(msg, &nh, true); | |
4588 | } | |
4589 | ||
4590 | void notify_resource_state(struct sk_buff *skb, | |
4591 | unsigned int seq, | |
4592 | struct drbd_resource *resource, | |
4593 | struct resource_info *resource_info, | |
4594 | enum drbd_notification_type type) | |
4595 | { | |
4596 | struct resource_statistics resource_statistics; | |
4597 | struct drbd_genlmsghdr *dh; | |
4598 | bool multicast = false; | |
4599 | int err; | |
4600 | ||
4601 | if (!skb) { | |
4602 | seq = atomic_inc_return(¬ify_genl_seq); | |
4603 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4604 | err = -ENOMEM; | |
4605 | if (!skb) | |
4606 | goto failed; | |
4607 | multicast = true; | |
4608 | } | |
4609 | ||
4610 | err = -EMSGSIZE; | |
4611 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_RESOURCE_STATE); | |
4612 | if (!dh) | |
4613 | goto nla_put_failure; | |
4614 | dh->minor = -1U; | |
4615 | dh->ret_code = NO_ERROR; | |
4616 | if (nla_put_drbd_cfg_context(skb, resource, NULL, NULL) || | |
4617 | nla_put_notification_header(skb, type) || | |
4618 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4619 | resource_info_to_skb(skb, resource_info, true))) | |
4620 | goto nla_put_failure; | |
4621 | resource_statistics.res_stat_write_ordering = resource->write_ordering; | |
4622 | err = resource_statistics_to_skb(skb, &resource_statistics, !capable(CAP_SYS_ADMIN)); | |
4623 | if (err) | |
4624 | goto nla_put_failure; | |
4625 | genlmsg_end(skb, dh); | |
4626 | if (multicast) { | |
d38f8612 | 4627 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4628 | /* skb has been consumed or freed in netlink_broadcast() */ |
4629 | if (err && err != -ESRCH) | |
4630 | goto failed; | |
4631 | } | |
4632 | return; | |
4633 | ||
4634 | nla_put_failure: | |
4635 | nlmsg_free(skb); | |
4636 | failed: | |
4637 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4638 | err, seq); | |
4639 | } | |
4640 | ||
4641 | void notify_device_state(struct sk_buff *skb, | |
4642 | unsigned int seq, | |
4643 | struct drbd_device *device, | |
4644 | struct device_info *device_info, | |
4645 | enum drbd_notification_type type) | |
4646 | { | |
4647 | struct device_statistics device_statistics; | |
4648 | struct drbd_genlmsghdr *dh; | |
4649 | bool multicast = false; | |
4650 | int err; | |
4651 | ||
4652 | if (!skb) { | |
4653 | seq = atomic_inc_return(¬ify_genl_seq); | |
4654 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4655 | err = -ENOMEM; | |
4656 | if (!skb) | |
4657 | goto failed; | |
4658 | multicast = true; | |
4659 | } | |
4660 | ||
4661 | err = -EMSGSIZE; | |
4662 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_DEVICE_STATE); | |
4663 | if (!dh) | |
4664 | goto nla_put_failure; | |
4665 | dh->minor = device->minor; | |
4666 | dh->ret_code = NO_ERROR; | |
4667 | if (nla_put_drbd_cfg_context(skb, device->resource, NULL, device) || | |
4668 | nla_put_notification_header(skb, type) || | |
4669 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4670 | device_info_to_skb(skb, device_info, true))) | |
4671 | goto nla_put_failure; | |
4672 | device_to_statistics(&device_statistics, device); | |
4673 | device_statistics_to_skb(skb, &device_statistics, !capable(CAP_SYS_ADMIN)); | |
4674 | genlmsg_end(skb, dh); | |
4675 | if (multicast) { | |
d38f8612 | 4676 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4677 | /* skb has been consumed or freed in netlink_broadcast() */ |
4678 | if (err && err != -ESRCH) | |
4679 | goto failed; | |
4680 | } | |
4681 | return; | |
4682 | ||
4683 | nla_put_failure: | |
4684 | nlmsg_free(skb); | |
4685 | failed: | |
4686 | drbd_err(device, "Error %d while broadcasting event. Event seq:%u\n", | |
4687 | err, seq); | |
4688 | } | |
4689 | ||
4690 | void notify_connection_state(struct sk_buff *skb, | |
4691 | unsigned int seq, | |
4692 | struct drbd_connection *connection, | |
4693 | struct connection_info *connection_info, | |
4694 | enum drbd_notification_type type) | |
4695 | { | |
4696 | struct connection_statistics connection_statistics; | |
4697 | struct drbd_genlmsghdr *dh; | |
4698 | bool multicast = false; | |
4699 | int err; | |
4700 | ||
4701 | if (!skb) { | |
4702 | seq = atomic_inc_return(¬ify_genl_seq); | |
4703 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4704 | err = -ENOMEM; | |
4705 | if (!skb) | |
4706 | goto failed; | |
4707 | multicast = true; | |
4708 | } | |
4709 | ||
4710 | err = -EMSGSIZE; | |
4711 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_CONNECTION_STATE); | |
4712 | if (!dh) | |
4713 | goto nla_put_failure; | |
4714 | dh->minor = -1U; | |
4715 | dh->ret_code = NO_ERROR; | |
4716 | if (nla_put_drbd_cfg_context(skb, connection->resource, connection, NULL) || | |
4717 | nla_put_notification_header(skb, type) || | |
4718 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4719 | connection_info_to_skb(skb, connection_info, true))) | |
4720 | goto nla_put_failure; | |
4721 | connection_statistics.conn_congested = test_bit(NET_CONGESTED, &connection->flags); | |
4722 | connection_statistics_to_skb(skb, &connection_statistics, !capable(CAP_SYS_ADMIN)); | |
4723 | genlmsg_end(skb, dh); | |
4724 | if (multicast) { | |
d38f8612 | 4725 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4726 | /* skb has been consumed or freed in netlink_broadcast() */ |
4727 | if (err && err != -ESRCH) | |
4728 | goto failed; | |
4729 | } | |
4730 | return; | |
4731 | ||
4732 | nla_put_failure: | |
4733 | nlmsg_free(skb); | |
4734 | failed: | |
4735 | drbd_err(connection, "Error %d while broadcasting event. Event seq:%u\n", | |
4736 | err, seq); | |
4737 | } | |
4738 | ||
4739 | void notify_peer_device_state(struct sk_buff *skb, | |
4740 | unsigned int seq, | |
4741 | struct drbd_peer_device *peer_device, | |
4742 | struct peer_device_info *peer_device_info, | |
4743 | enum drbd_notification_type type) | |
4744 | { | |
4745 | struct peer_device_statistics peer_device_statistics; | |
4746 | struct drbd_resource *resource = peer_device->device->resource; | |
4747 | struct drbd_genlmsghdr *dh; | |
4748 | bool multicast = false; | |
4749 | int err; | |
4750 | ||
4751 | if (!skb) { | |
4752 | seq = atomic_inc_return(¬ify_genl_seq); | |
4753 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4754 | err = -ENOMEM; | |
4755 | if (!skb) | |
4756 | goto failed; | |
4757 | multicast = true; | |
4758 | } | |
4759 | ||
4760 | err = -EMSGSIZE; | |
4761 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_PEER_DEVICE_STATE); | |
4762 | if (!dh) | |
4763 | goto nla_put_failure; | |
4764 | dh->minor = -1U; | |
4765 | dh->ret_code = NO_ERROR; | |
4766 | if (nla_put_drbd_cfg_context(skb, resource, peer_device->connection, peer_device->device) || | |
4767 | nla_put_notification_header(skb, type) || | |
4768 | ((type & ~NOTIFY_FLAGS) != NOTIFY_DESTROY && | |
4769 | peer_device_info_to_skb(skb, peer_device_info, true))) | |
4770 | goto nla_put_failure; | |
4771 | peer_device_to_statistics(&peer_device_statistics, peer_device); | |
4772 | peer_device_statistics_to_skb(skb, &peer_device_statistics, !capable(CAP_SYS_ADMIN)); | |
4773 | genlmsg_end(skb, dh); | |
4774 | if (multicast) { | |
d38f8612 | 4775 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4776 | /* skb has been consumed or freed in netlink_broadcast() */ |
4777 | if (err && err != -ESRCH) | |
4778 | goto failed; | |
4779 | } | |
4780 | return; | |
4781 | ||
4782 | nla_put_failure: | |
4783 | nlmsg_free(skb); | |
4784 | failed: | |
4785 | drbd_err(peer_device, "Error %d while broadcasting event. Event seq:%u\n", | |
4786 | err, seq); | |
4787 | } | |
4788 | ||
4789 | void notify_helper(enum drbd_notification_type type, | |
4790 | struct drbd_device *device, struct drbd_connection *connection, | |
4791 | const char *name, int status) | |
4792 | { | |
4793 | struct drbd_resource *resource = device ? device->resource : connection->resource; | |
4794 | struct drbd_helper_info helper_info; | |
4795 | unsigned int seq = atomic_inc_return(¬ify_genl_seq); | |
4796 | struct sk_buff *skb = NULL; | |
4797 | struct drbd_genlmsghdr *dh; | |
4798 | int err; | |
4799 | ||
4800 | strlcpy(helper_info.helper_name, name, sizeof(helper_info.helper_name)); | |
4801 | helper_info.helper_name_len = min(strlen(name), sizeof(helper_info.helper_name)); | |
4802 | helper_info.helper_status = status; | |
4803 | ||
4804 | skb = genlmsg_new(NLMSG_GOODSIZE, GFP_NOIO); | |
4805 | err = -ENOMEM; | |
4806 | if (!skb) | |
4807 | goto fail; | |
4808 | ||
4809 | err = -EMSGSIZE; | |
4810 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_HELPER); | |
4811 | if (!dh) | |
4812 | goto fail; | |
4813 | dh->minor = device ? device->minor : -1; | |
4814 | dh->ret_code = NO_ERROR; | |
4815 | mutex_lock(¬ification_mutex); | |
4816 | if (nla_put_drbd_cfg_context(skb, resource, connection, device) || | |
4817 | nla_put_notification_header(skb, type) || | |
4818 | drbd_helper_info_to_skb(skb, &helper_info, true)) | |
4819 | goto unlock_fail; | |
4820 | genlmsg_end(skb, dh); | |
d38f8612 | 4821 | err = drbd_genl_multicast_events(skb, GFP_NOWAIT); |
a2972846 AG |
4822 | skb = NULL; |
4823 | /* skb has been consumed or freed in netlink_broadcast() */ | |
4824 | if (err && err != -ESRCH) | |
4825 | goto unlock_fail; | |
4826 | mutex_unlock(¬ification_mutex); | |
4827 | return; | |
4828 | ||
4829 | unlock_fail: | |
4830 | mutex_unlock(¬ification_mutex); | |
4831 | fail: | |
4832 | nlmsg_free(skb); | |
4833 | drbd_err(resource, "Error %d while broadcasting event. Event seq:%u\n", | |
4834 | err, seq); | |
4835 | } | |
4836 | ||
4837 | static void notify_initial_state_done(struct sk_buff *skb, unsigned int seq) | |
4838 | { | |
4839 | struct drbd_genlmsghdr *dh; | |
4840 | int err; | |
4841 | ||
4842 | err = -EMSGSIZE; | |
4843 | dh = genlmsg_put(skb, 0, seq, &drbd_genl_family, 0, DRBD_INITIAL_STATE_DONE); | |
4844 | if (!dh) | |
4845 | goto nla_put_failure; | |
4846 | dh->minor = -1U; | |
4847 | dh->ret_code = NO_ERROR; | |
4848 | if (nla_put_notification_header(skb, NOTIFY_EXISTS)) | |
4849 | goto nla_put_failure; | |
4850 | genlmsg_end(skb, dh); | |
4851 | return; | |
4852 | ||
4853 | nla_put_failure: | |
4854 | nlmsg_free(skb); | |
4855 | pr_err("Error %d sending event. Event seq:%u\n", err, seq); | |
4856 | } | |
4857 | ||
4858 | static void free_state_changes(struct list_head *list) | |
4859 | { | |
4860 | while (!list_empty(list)) { | |
4861 | struct drbd_state_change *state_change = | |
4862 | list_first_entry(list, struct drbd_state_change, list); | |
4863 | list_del(&state_change->list); | |
4864 | forget_state_change(state_change); | |
4865 | } | |
4866 | } | |
4867 | ||
4868 | static unsigned int notifications_for_state_change(struct drbd_state_change *state_change) | |
4869 | { | |
4870 | return 1 + | |
4871 | state_change->n_connections + | |
4872 | state_change->n_devices + | |
4873 | state_change->n_devices * state_change->n_connections; | |
4874 | } | |
4875 | ||
4876 | static int get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4877 | { | |
4878 | struct drbd_state_change *state_change = (struct drbd_state_change *)cb->args[0]; | |
4879 | unsigned int seq = cb->args[2]; | |
4880 | unsigned int n; | |
4881 | enum drbd_notification_type flags = 0; | |
4882 | ||
4883 | /* There is no need for taking notification_mutex here: it doesn't | |
4884 | matter if the initial state events mix with later state chage | |
4885 | events; we can always tell the events apart by the NOTIFY_EXISTS | |
4886 | flag. */ | |
4887 | ||
4888 | cb->args[5]--; | |
4889 | if (cb->args[5] == 1) { | |
4890 | notify_initial_state_done(skb, seq); | |
4891 | goto out; | |
4892 | } | |
4893 | n = cb->args[4]++; | |
4894 | if (cb->args[4] < cb->args[3]) | |
4895 | flags |= NOTIFY_CONTINUES; | |
4896 | if (n < 1) { | |
4897 | notify_resource_state_change(skb, seq, state_change->resource, | |
4898 | NOTIFY_EXISTS | flags); | |
4899 | goto next; | |
4900 | } | |
4901 | n--; | |
4902 | if (n < state_change->n_connections) { | |
4903 | notify_connection_state_change(skb, seq, &state_change->connections[n], | |
4904 | NOTIFY_EXISTS | flags); | |
4905 | goto next; | |
4906 | } | |
4907 | n -= state_change->n_connections; | |
4908 | if (n < state_change->n_devices) { | |
4909 | notify_device_state_change(skb, seq, &state_change->devices[n], | |
4910 | NOTIFY_EXISTS | flags); | |
4911 | goto next; | |
4912 | } | |
4913 | n -= state_change->n_devices; | |
4914 | if (n < state_change->n_devices * state_change->n_connections) { | |
4915 | notify_peer_device_state_change(skb, seq, &state_change->peer_devices[n], | |
4916 | NOTIFY_EXISTS | flags); | |
4917 | goto next; | |
4918 | } | |
4919 | ||
4920 | next: | |
4921 | if (cb->args[4] == cb->args[3]) { | |
4922 | struct drbd_state_change *next_state_change = | |
4923 | list_entry(state_change->list.next, | |
4924 | struct drbd_state_change, list); | |
4925 | cb->args[0] = (long)next_state_change; | |
4926 | cb->args[3] = notifications_for_state_change(next_state_change); | |
4927 | cb->args[4] = 0; | |
4928 | } | |
4929 | out: | |
4930 | return skb->len; | |
4931 | } | |
4932 | ||
4933 | int drbd_adm_get_initial_state(struct sk_buff *skb, struct netlink_callback *cb) | |
4934 | { | |
4935 | struct drbd_resource *resource; | |
4936 | LIST_HEAD(head); | |
4937 | ||
4938 | if (cb->args[5] >= 1) { | |
4939 | if (cb->args[5] > 1) | |
4940 | return get_initial_state(skb, cb); | |
4941 | if (cb->args[0]) { | |
4942 | struct drbd_state_change *state_change = | |
4943 | (struct drbd_state_change *)cb->args[0]; | |
4944 | ||
4945 | /* connect list to head */ | |
4946 | list_add(&head, &state_change->list); | |
4947 | free_state_changes(&head); | |
4948 | } | |
4949 | return 0; | |
4950 | } | |
4951 | ||
4952 | cb->args[5] = 2; /* number of iterations */ | |
4953 | mutex_lock(&resources_mutex); | |
4954 | for_each_resource(resource, &drbd_resources) { | |
4955 | struct drbd_state_change *state_change; | |
4956 | ||
4957 | state_change = remember_old_state(resource, GFP_KERNEL); | |
4958 | if (!state_change) { | |
4959 | if (!list_empty(&head)) | |
4960 | free_state_changes(&head); | |
4961 | mutex_unlock(&resources_mutex); | |
4962 | return -ENOMEM; | |
4963 | } | |
4964 | copy_old_to_new_state_change(state_change); | |
4965 | list_add_tail(&state_change->list, &head); | |
4966 | cb->args[5] += notifications_for_state_change(state_change); | |
4967 | } | |
4968 | mutex_unlock(&resources_mutex); | |
4969 | ||
4970 | if (!list_empty(&head)) { | |
4971 | struct drbd_state_change *state_change = | |
4972 | list_entry(head.next, struct drbd_state_change, list); | |
4973 | cb->args[0] = (long)state_change; | |
4974 | cb->args[3] = notifications_for_state_change(state_change); | |
4975 | list_del(&head); /* detach list from head */ | |
4976 | } | |
4977 | ||
4978 | cb->args[2] = cb->nlh->nlmsg_seq; | |
4979 | return get_initial_state(skb, cb); | |
4980 | } |