blob: e26fed84b1aaf4ec0c4869f80bf7fe6122235207 [file] [log] [blame]
Evgeniy Polyakovce0d9d72009-01-14 02:05:27 +03001/*
2 * 2007+ Copyright (c) Evgeniy Polyakov <johnpol@2ka.mipt.ru>
3 * All rights reserved.
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License as published by
7 * the Free Software Foundation; either version 2 of the License, or
8 * (at your option) any later version.
9 *
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
14 */
15
16#ifndef __DST_H
17#define __DST_H
18
19#include <linux/types.h>
20#include <linux/connector.h>
21
22#define DST_NAMELEN 32
23#define DST_NAME "dst"
24
25enum {
26 /* Remove node with given id from storage */
27 DST_DEL_NODE = 0,
28 /* Add remote node with given id to the storage */
29 DST_ADD_REMOTE,
30 /* Add local node with given id to the storage to be exported and used by remote peers */
31 DST_ADD_EXPORT,
32 /* Crypto initialization command (hash/cipher used to protect the connection) */
33 DST_CRYPTO,
34 /* Security attributes for given connection (permissions for example) */
35 DST_SECURITY,
36 /* Register given node in the block layer subsystem */
37 DST_START,
38 DST_CMD_MAX
39};
40
41struct dst_ctl
42{
43 /* Storage name */
44 char name[DST_NAMELEN];
45 /* Command flags */
46 __u32 flags;
47 /* Command itself (see above) */
48 __u32 cmd;
49 /* Maximum number of pages per single request in this device */
50 __u32 max_pages;
51 /* Stale/error transaction scanning timeout in milliseconds */
52 __u32 trans_scan_timeout;
53 /* Maximum number of retry sends before completing transaction as broken */
54 __u32 trans_max_retries;
55 /* Storage size */
56 __u64 size;
57};
58
59/* Reply command carries completion status */
60struct dst_ctl_ack
61{
62 struct cn_msg msg;
63 int error;
64 int unused[3];
65};
66
67/*
68 * Unfortunaltely socket address structure is not exported to userspace
69 * and is redefined there.
70 */
71#define SADDR_MAX_DATA 128
72
73struct saddr {
74 /* address family, AF_xxx */
75 unsigned short sa_family;
76 /* 14 bytes of protocol address */
77 char sa_data[SADDR_MAX_DATA];
78 /* Number of bytes used in sa_data */
79 unsigned short sa_data_len;
80};
81
82/* Address structure */
83struct dst_network_ctl
84{
85 /* Socket type: datagram, stream...*/
86 unsigned int type;
87 /* Let me guess, is it a Jupiter diameter? */
88 unsigned int proto;
89 /* Peer's address */
90 struct saddr addr;
91};
92
93struct dst_crypto_ctl
94{
95 /* Cipher and hash names */
96 char cipher_algo[DST_NAMELEN];
97 char hash_algo[DST_NAMELEN];
98
99 /* Key sizes. Can be zero for digest for example */
100 unsigned int cipher_keysize, hash_keysize;
101 /* Alignment. Calculated by the DST itself. */
102 unsigned int crypto_attached_size;
103 /* Number of threads to perform crypto operations */
104 int thread_num;
105};
106
107/* Export security attributes have this bits checked in when client connects */
108#define DST_PERM_READ (1<<0)
109#define DST_PERM_WRITE (1<<1)
110
111/*
112 * Right now it is simple model, where each remote address
113 * is assigned to set of permissions it is allowed to perform.
114 * In real world block device does not know anything but
115 * reading and writing, so it should be more than enough.
116 */
117struct dst_secure_user
118{
119 unsigned int permissions;
120 struct saddr addr;
121};
122
123/*
124 * Export control command: device to export and network address to accept
125 * clients to work with given device
126 */
127struct dst_export_ctl
128{
129 char device[DST_NAMELEN];
130 struct dst_network_ctl ctl;
131};
132
133enum {
134 DST_CFG = 1, /* Request remote configuration */
135 DST_IO, /* IO command */
136 DST_IO_RESPONSE, /* IO response */
137 DST_PING, /* Keepalive message */
138 DST_NCMD_MAX,
139};
140
141struct dst_cmd
142{
143 /* Network command itself, see above */
144 __u32 cmd;
145 /*
146 * Size of the attached data
147 * (in most cases, for READ command it means how many bytes were requested)
148 */
149 __u32 size;
150 /* Crypto size: number of attached bytes with digest/hmac */
151 __u32 csize;
152 /* Here we can carry secret data */
153 __u32 reserved;
154 /* Read/write bits, see how they are encoded in bio structure */
155 __u64 rw;
156 /* BIO flags */
157 __u64 flags;
158 /* Unique command id (like transaction ID) */
159 __u64 id;
160 /* Sector to start IO from */
161 __u64 sector;
162 /* Hash data is placed after this header */
163 __u8 hash[0];
164};
165
166/*
167 * Convert command to/from network byte order.
168 * We do not use hton*() functions, since there is
169 * no 64-bit implementation.
170 */
171static inline void dst_convert_cmd(struct dst_cmd *c)
172{
173 c->cmd = __cpu_to_be32(c->cmd);
174 c->csize = __cpu_to_be32(c->csize);
175 c->size = __cpu_to_be32(c->size);
176 c->sector = __cpu_to_be64(c->sector);
177 c->id = __cpu_to_be64(c->id);
178 c->flags = __cpu_to_be64(c->flags);
179 c->rw = __cpu_to_be64(c->rw);
180}
181
182/* Transaction id */
183typedef __u64 dst_gen_t;
184
185#ifdef __KERNEL__
186
187#include <linux/blkdev.h>
188#include <linux/bio.h>
189#include <linux/device.h>
190#include <linux/mempool.h>
191#include <linux/net.h>
192#include <linux/poll.h>
193#include <linux/rbtree.h>
194
195#ifdef CONFIG_DST_DEBUG
196#define dprintk(f, a...) printk(KERN_NOTICE f, ##a)
197#else
198static inline void __attribute__ ((format (printf, 1, 2)))
199 dprintk(const char *fmt, ...) {}
200#endif
201
202struct dst_node;
203
204struct dst_trans
205{
206 /* DST node we are working with */
207 struct dst_node *n;
208
209 /* Entry inside transaction tree */
210 struct rb_node trans_entry;
211
212 /* Merlin kills this transaction when this memory cell equals zero */
213 atomic_t refcnt;
214
215 /* How this transaction should be processed by crypto engine */
216 short enc;
217 /* How many times this transaction was resent */
218 short retries;
219 /* Completion status */
220 int error;
221
222 /* When did we send it to the remote peer */
223 long send_time;
224
225 /* My name is...
226 * Well, computers does not speak, they have unique id instead */
227 dst_gen_t gen;
228
229 /* Block IO we are working with */
230 struct bio *bio;
231
232 /* Network command for above block IO request */
233 struct dst_cmd cmd;
234};
235
236struct dst_crypto_engine
237{
238 /* What should we do with all block requests */
239 struct crypto_hash *hash;
240 struct crypto_ablkcipher *cipher;
241
242 /* Pool of pages used to encrypt data into before sending */
243 int page_num;
244 struct page **pages;
245
246 /* What to do with current request */
247 int enc;
248 /* Who we are and where do we go */
249 struct scatterlist *src, *dst;
250
251 /* Maximum timeout waiting for encryption to be completed */
252 long timeout;
253 /* IV is a 64-bit sequential counter */
254 u64 iv;
255
256 /* Secret data */
257 void *private;
258
259 /* Cached temporary data lives here */
260 int size;
261 void *data;
262};
263
264struct dst_state
265{
266 /* The main state protection */
267 struct mutex state_lock;
268
269 /* Polling machinery for sockets */
270 wait_queue_t wait;
271 wait_queue_head_t *whead;
272 /* Most of events are being waited here */
273 wait_queue_head_t thread_wait;
274
275 /* Who owns this? */
276 struct dst_node *node;
277
278 /* Network address for this state */
279 struct dst_network_ctl ctl;
280
281 /* Permissions to work with: read-only or rw connection */
282 u32 permissions;
283
284 /* Called when we need to clean private data */
285 void (* cleanup)(struct dst_state *st);
286
287 /* Used by the server: BIO completion queues BIOs here */
288 struct list_head request_list;
289 spinlock_t request_lock;
290
291 /* Guess what? No, it is not number of planets */
292 atomic_t refcnt;
293
294 /* This flags is set when connection should be dropped */
295 int need_exit;
296
297 /*
298 * Socket to work with. Second pointer is used for
299 * lockless check if socket was changed before performing
300 * next action (like working with cached polling result)
301 */
302 struct socket *socket, *read_socket;
303
304 /* Cached preallocated data */
305 void *data;
306 unsigned int size;
307
308 /* Currently processed command */
309 struct dst_cmd cmd;
310};
311
312struct dst_info
313{
314 /* Device size */
315 u64 size;
316
317 /* Local device name for export devices */
318 char local[DST_NAMELEN];
319
320 /* Network setup */
321 struct dst_network_ctl net;
322
323 /* Sysfs bits use this */
324 struct device device;
325};
326
327struct dst_node
328{
329 struct list_head node_entry;
330
331 /* Hi, my name is stored here */
332 char name[DST_NAMELEN];
333 /* My cache name is stored here */
334 char cache_name[DST_NAMELEN];
335
336 /* Block device attached to given node.
337 * Only valid for exporting nodes */
338 struct block_device *bdev;
339 /* Network state machine for given peer */
340 struct dst_state *state;
341
342 /* Block IO machinery */
343 struct request_queue *queue;
344 struct gendisk *disk;
345
346 /* Number of threads in processing pool */
347 int thread_num;
348 /* Maximum number of pages in single IO */
349 int max_pages;
350
351 /* I'm that big in bytes */
352 loff_t size;
353
354 /* Exported to userspace node information */
355 struct dst_info *info;
356
357 /*
358 * Security attribute list.
359 * Used only by exporting node currently.
360 */
361 struct list_head security_list;
362 struct mutex security_lock;
363
364 /*
365 * When this unerflows below zero, university collapses.
366 * But this will not happen, since node will be freed,
367 * when reference counter reaches zero.
368 */
369 atomic_t refcnt;
370
371 /* How precisely should I be started? */
372 int (*start)(struct dst_node *);
373
374 /* Crypto capabilities */
375 struct dst_crypto_ctl crypto;
376 u8 *hash_key;
377 u8 *cipher_key;
378
379 /* Pool of processing thread */
380 struct thread_pool *pool;
381
382 /* Transaction IDs live here */
383 atomic_long_t gen;
384
385 /*
386 * How frequently and how many times transaction
387 * tree should be scanned to drop stale objects.
388 */
389 long trans_scan_timeout;
390 int trans_max_retries;
391
392 /* Small gnomes live here */
393 struct rb_root trans_root;
394 struct mutex trans_lock;
395
396 /*
397 * Transaction cache/memory pool.
398 * It is big enough to contain not only transaction
399 * itself, but additional crypto data (digest/hmac).
400 */
401 struct kmem_cache *trans_cache;
402 mempool_t *trans_pool;
403
404 /* This entity scans transaction tree */
405 struct delayed_work trans_work;
406
407 wait_queue_head_t wait;
408};
409
410/* Kernel representation of the security attribute */
411struct dst_secure
412{
413 struct list_head sec_entry;
414 struct dst_secure_user sec;
415};
416
417int dst_process_bio(struct dst_node *n, struct bio *bio);
418
419int dst_node_init_connected(struct dst_node *n, struct dst_network_ctl *r);
420int dst_node_init_listened(struct dst_node *n, struct dst_export_ctl *le);
421
422static inline struct dst_state *dst_state_get(struct dst_state *st)
423{
424 BUG_ON(atomic_read(&st->refcnt) == 0);
425 atomic_inc(&st->refcnt);
426 return st;
427}
428
429void dst_state_put(struct dst_state *st);
430
431struct dst_state *dst_state_alloc(struct dst_node *n);
432int dst_state_socket_create(struct dst_state *st);
433void dst_state_socket_release(struct dst_state *st);
434
435void dst_state_exit_connected(struct dst_state *st);
436
437int dst_state_schedule_receiver(struct dst_state *st);
438
439void dst_dump_addr(struct socket *sk, struct sockaddr *sa, char *str);
440
441static inline void dst_state_lock(struct dst_state *st)
442{
443 mutex_lock(&st->state_lock);
444}
445
446static inline void dst_state_unlock(struct dst_state *st)
447{
448 mutex_unlock(&st->state_lock);
449}
450
451void dst_poll_exit(struct dst_state *st);
452int dst_poll_init(struct dst_state *st);
453
454static inline unsigned int dst_state_poll(struct dst_state *st)
455{
456 unsigned int revents = POLLHUP | POLLERR;
457
458 dst_state_lock(st);
459 if (st->socket)
460 revents = st->socket->ops->poll(NULL, st->socket, NULL);
461 dst_state_unlock(st);
462
463 return revents;
464}
465
466static inline int dst_thread_setup(void *private, void *data)
467{
468 return 0;
469}
470
471void dst_node_put(struct dst_node *n);
472
473static inline struct dst_node *dst_node_get(struct dst_node *n)
474{
475 atomic_inc(&n->refcnt);
476 return n;
477}
478
479int dst_data_recv(struct dst_state *st, void *data, unsigned int size);
480int dst_recv_cdata(struct dst_state *st, void *cdata);
481int dst_data_send_header(struct socket *sock,
482 void *data, unsigned int size, int more);
483
484int dst_send_bio(struct dst_state *st, struct dst_cmd *cmd, struct bio *bio);
485
486int dst_process_io(struct dst_state *st);
487int dst_export_crypto(struct dst_node *n, struct bio *bio);
488int dst_export_send_bio(struct bio *bio);
489int dst_start_export(struct dst_node *n);
490
491int __init dst_export_init(void);
492void dst_export_exit(void);
493
494/* Private structure for export block IO requests */
495struct dst_export_priv
496{
497 struct list_head request_entry;
498 struct dst_state *state;
499 struct bio *bio;
500 struct dst_cmd cmd;
501};
502
503static inline void dst_trans_get(struct dst_trans *t)
504{
505 atomic_inc(&t->refcnt);
506}
507
508struct dst_trans *dst_trans_search(struct dst_node *node, dst_gen_t gen);
509int dst_trans_remove(struct dst_trans *t);
510int dst_trans_remove_nolock(struct dst_trans *t);
511void dst_trans_put(struct dst_trans *t);
512
513/*
514 * Convert bio into network command.
515 */
516static inline void dst_bio_to_cmd(struct bio *bio, struct dst_cmd *cmd,
517 u32 command, u64 id)
518{
519 cmd->cmd = command;
520 cmd->flags = (bio->bi_flags << BIO_POOL_BITS) >> BIO_POOL_BITS;
521 cmd->rw = bio->bi_rw;
522 cmd->size = bio->bi_size;
523 cmd->csize = 0;
524 cmd->id = id;
525 cmd->sector = bio->bi_sector;
526};
527
528int dst_trans_send(struct dst_trans *t);
529int dst_trans_crypto(struct dst_trans *t);
530
531int dst_node_crypto_init(struct dst_node *n, struct dst_crypto_ctl *ctl);
532void dst_node_crypto_exit(struct dst_node *n);
533
534static inline int dst_need_crypto(struct dst_node *n)
535{
536 struct dst_crypto_ctl *c = &n->crypto;
537 /*
538 * Logical OR is appropriate here, but boolean one produces
539 * more optimal code, so it is used instead.
540 */
541 return (c->hash_algo[0] | c->cipher_algo[0]);
542}
543
544int dst_node_trans_init(struct dst_node *n, unsigned int size);
545void dst_node_trans_exit(struct dst_node *n);
546
547/*
548 * Pool of threads.
549 * Ready list contains threads currently free to be used,
550 * active one contains threads with some work scheduled for them.
551 * Caller can wait in given queue when thread is ready.
552 */
553struct thread_pool
554{
555 int thread_num;
556 struct mutex thread_lock;
557 struct list_head ready_list, active_list;
558
559 wait_queue_head_t wait;
560};
561
562void thread_pool_del_worker(struct thread_pool *p);
563void thread_pool_del_worker_id(struct thread_pool *p, unsigned int id);
564int thread_pool_add_worker(struct thread_pool *p,
565 char *name,
566 unsigned int id,
567 void *(* init)(void *data),
568 void (* cleanup)(void *data),
569 void *data);
570
571void thread_pool_destroy(struct thread_pool *p);
572struct thread_pool *thread_pool_create(int num, char *name,
573 void *(* init)(void *data),
574 void (* cleanup)(void *data),
575 void *data);
576
577int thread_pool_schedule(struct thread_pool *p,
578 int (* setup)(void *stored_private, void *setup_data),
579 int (* action)(void *stored_private, void *setup_data),
580 void *setup_data, long timeout);
581int thread_pool_schedule_private(struct thread_pool *p,
582 int (* setup)(void *private, void *data),
583 int (* action)(void *private, void *data),
584 void *data, long timeout, void *id);
585
586#endif /* __KERNEL__ */
587#endif /* __DST_H */