1/* AFS volume management
2 *
3 * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11
12#include <linux/kernel.h>
13#include <linux/module.h>
14#include <linux/init.h>
15#include <linux/slab.h>
16#include <linux/fs.h>
17#include <linux/pagemap.h>
18#include <linux/sched.h>
19#include "internal.h"
20
21static const char *afs_voltypes[] = { "R/W", "R/O", "BAK" };
22
23/*
24 * lookup a volume by name
25 * - this can be one of the following:
26 *	"%[cell:]volume[.]"		R/W volume
27 *	"#[cell:]volume[.]"		R/O or R/W volume (rwparent=0),
28 *					 or R/W (rwparent=1) volume
29 *	"%[cell:]volume.readonly"	R/O volume
30 *	"#[cell:]volume.readonly"	R/O volume
31 *	"%[cell:]volume.backup"		Backup volume
32 *	"#[cell:]volume.backup"		Backup volume
33 *
34 * The cell name is optional, and defaults to the current cell.
35 *
36 * See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
37 * Guide
38 * - Rule 1: Explicit type suffix forces access of that type or nothing
39 *           (no suffix, then use Rule 2 & 3)
40 * - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
41 *           if not available
42 * - Rule 3: If parent volume is R/W, then only mount R/W volume unless
43 *           explicitly told otherwise
44 */
45struct afs_volume *afs_volume_lookup(struct afs_mount_params *params)
46{
47	struct afs_vlocation *vlocation = NULL;
48	struct afs_volume *volume = NULL;
49	struct afs_server *server = NULL;
50	char srvtmask;
51	int ret, loop;
52
53	_enter("{%*.*s,%d}",
54	       params->volnamesz, params->volnamesz, params->volname, params->rwpath);
55
56	/* lookup the volume location record */
57	vlocation = afs_vlocation_lookup(params->cell, params->key,
58					 params->volname, params->volnamesz);
59	if (IS_ERR(vlocation)) {
60		ret = PTR_ERR(vlocation);
61		vlocation = NULL;
62		goto error;
63	}
64
65	/* make the final decision on the type we want */
66	ret = -ENOMEDIUM;
67	if (params->force && !(vlocation->vldb.vidmask & (1 << params->type)))
68		goto error;
69
70	srvtmask = 0;
71	for (loop = 0; loop < vlocation->vldb.nservers; loop++)
72		srvtmask |= vlocation->vldb.srvtmask[loop];
73
74	if (params->force) {
75		if (!(srvtmask & (1 << params->type)))
76			goto error;
77	} else if (srvtmask & AFS_VOL_VTM_RO) {
78		params->type = AFSVL_ROVOL;
79	} else if (srvtmask & AFS_VOL_VTM_RW) {
80		params->type = AFSVL_RWVOL;
81	} else {
82		goto error;
83	}
84
85	down_write(&params->cell->vl_sem);
86
87	/* is the volume already active? */
88	if (vlocation->vols[params->type]) {
89		/* yes - re-use it */
90		volume = vlocation->vols[params->type];
91		afs_get_volume(volume);
92		goto success;
93	}
94
95	/* create a new volume record */
96	_debug("creating new volume record");
97
98	ret = -ENOMEM;
99	volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
100	if (!volume)
101		goto error_up;
102
103	atomic_set(&volume->usage, 1);
104	volume->type		= params->type;
105	volume->type_force	= params->force;
106	volume->cell		= params->cell;
107	volume->vid		= vlocation->vldb.vid[params->type];
108
109	ret = bdi_setup_and_register(&volume->bdi, "afs");
110	if (ret)
111		goto error_bdi;
112
113	init_rwsem(&volume->server_sem);
114
115	/* look up all the applicable server records */
116	for (loop = 0; loop < 8; loop++) {
117		if (vlocation->vldb.srvtmask[loop] & (1 << volume->type)) {
118			server = afs_lookup_server(
119			       volume->cell, &vlocation->vldb.servers[loop]);
120			if (IS_ERR(server)) {
121				ret = PTR_ERR(server);
122				goto error_discard;
123			}
124
125			volume->servers[volume->nservers] = server;
126			volume->nservers++;
127		}
128	}
129
130	/* attach the cache and volume location */
131#ifdef CONFIG_AFS_FSCACHE
132	volume->cache = fscache_acquire_cookie(vlocation->cache,
133					       &afs_volume_cache_index_def,
134					       volume, true);
135#endif
136	afs_get_vlocation(vlocation);
137	volume->vlocation = vlocation;
138
139	vlocation->vols[volume->type] = volume;
140
141success:
142	_debug("kAFS selected %s volume %08x",
143	       afs_voltypes[volume->type], volume->vid);
144	up_write(&params->cell->vl_sem);
145	afs_put_vlocation(vlocation);
146	_leave(" = %p", volume);
147	return volume;
148
149	/* clean up */
150error_up:
151	up_write(&params->cell->vl_sem);
152error:
153	afs_put_vlocation(vlocation);
154	_leave(" = %d", ret);
155	return ERR_PTR(ret);
156
157error_discard:
158	bdi_destroy(&volume->bdi);
159error_bdi:
160	up_write(&params->cell->vl_sem);
161
162	for (loop = volume->nservers - 1; loop >= 0; loop--)
163		afs_put_server(volume->servers[loop]);
164
165	kfree(volume);
166	goto error;
167}
168
169/*
170 * destroy a volume record
171 */
172void afs_put_volume(struct afs_volume *volume)
173{
174	struct afs_vlocation *vlocation;
175	int loop;
176
177	if (!volume)
178		return;
179
180	_enter("%p", volume);
181
182	ASSERTCMP(atomic_read(&volume->usage), >, 0);
183
184	vlocation = volume->vlocation;
185
186	/* to prevent a race, the decrement and the dequeue must be effectively
187	 * atomic */
188	down_write(&vlocation->cell->vl_sem);
189
190	if (likely(!atomic_dec_and_test(&volume->usage))) {
191		up_write(&vlocation->cell->vl_sem);
192		_leave("");
193		return;
194	}
195
196	vlocation->vols[volume->type] = NULL;
197
198	up_write(&vlocation->cell->vl_sem);
199
200	/* finish cleaning up the volume */
201#ifdef CONFIG_AFS_FSCACHE
202	fscache_relinquish_cookie(volume->cache, 0);
203#endif
204	afs_put_vlocation(vlocation);
205
206	for (loop = volume->nservers - 1; loop >= 0; loop--)
207		afs_put_server(volume->servers[loop]);
208
209	bdi_destroy(&volume->bdi);
210	kfree(volume);
211
212	_leave(" [destroyed]");
213}
214
215/*
216 * pick a server to use to try accessing this volume
217 * - returns with an elevated usage count on the server chosen
218 */
219struct afs_server *afs_volume_pick_fileserver(struct afs_vnode *vnode)
220{
221	struct afs_volume *volume = vnode->volume;
222	struct afs_server *server;
223	int ret, state, loop;
224
225	_enter("%s", volume->vlocation->vldb.name);
226
227	/* stick with the server we're already using if we can */
228	if (vnode->server && vnode->server->fs_state == 0) {
229		afs_get_server(vnode->server);
230		_leave(" = %p [current]", vnode->server);
231		return vnode->server;
232	}
233
234	down_read(&volume->server_sem);
235
236	/* handle the no-server case */
237	if (volume->nservers == 0) {
238		ret = volume->rjservers ? -ENOMEDIUM : -ESTALE;
239		up_read(&volume->server_sem);
240		_leave(" = %d [no servers]", ret);
241		return ERR_PTR(ret);
242	}
243
244	/* basically, just search the list for the first live server and use
245	 * that */
246	ret = 0;
247	for (loop = 0; loop < volume->nservers; loop++) {
248		server = volume->servers[loop];
249		state = server->fs_state;
250
251		_debug("consider %d [%d]", loop, state);
252
253		switch (state) {
254			/* found an apparently healthy server */
255		case 0:
256			afs_get_server(server);
257			up_read(&volume->server_sem);
258			_leave(" = %p (picked %08x)",
259			       server, ntohl(server->addr.s_addr));
260			return server;
261
262		case -ENETUNREACH:
263			if (ret == 0)
264				ret = state;
265			break;
266
267		case -EHOSTUNREACH:
268			if (ret == 0 ||
269			    ret == -ENETUNREACH)
270				ret = state;
271			break;
272
273		case -ECONNREFUSED:
274			if (ret == 0 ||
275			    ret == -ENETUNREACH ||
276			    ret == -EHOSTUNREACH)
277				ret = state;
278			break;
279
280		default:
281		case -EREMOTEIO:
282			if (ret == 0 ||
283			    ret == -ENETUNREACH ||
284			    ret == -EHOSTUNREACH ||
285			    ret == -ECONNREFUSED)
286				ret = state;
287			break;
288		}
289	}
290
291	/* no available servers
292	 * - TODO: handle the no active servers case better
293	 */
294	up_read(&volume->server_sem);
295	_leave(" = %d", ret);
296	return ERR_PTR(ret);
297}
298
299/*
300 * release a server after use
301 * - releases the ref on the server struct that was acquired by picking
302 * - records result of using a particular server to access a volume
303 * - return 0 to try again, 1 if okay or to issue error
304 * - the caller must release the server struct if result was 0
305 */
306int afs_volume_release_fileserver(struct afs_vnode *vnode,
307				  struct afs_server *server,
308				  int result)
309{
310	struct afs_volume *volume = vnode->volume;
311	unsigned loop;
312
313	_enter("%s,%08x,%d",
314	       volume->vlocation->vldb.name, ntohl(server->addr.s_addr),
315	       result);
316
317	switch (result) {
318		/* success */
319	case 0:
320		server->fs_act_jif = jiffies;
321		server->fs_state = 0;
322		_leave("");
323		return 1;
324
325		/* the fileserver denied all knowledge of the volume */
326	case -ENOMEDIUM:
327		server->fs_act_jif = jiffies;
328		down_write(&volume->server_sem);
329
330		/* firstly, find where the server is in the active list (if it
331		 * is) */
332		for (loop = 0; loop < volume->nservers; loop++)
333			if (volume->servers[loop] == server)
334				goto present;
335
336		/* no longer there - may have been discarded by another op */
337		goto try_next_server_upw;
338
339	present:
340		volume->nservers--;
341		memmove(&volume->servers[loop],
342			&volume->servers[loop + 1],
343			sizeof(volume->servers[loop]) *
344			(volume->nservers - loop));
345		volume->servers[volume->nservers] = NULL;
346		afs_put_server(server);
347		volume->rjservers++;
348
349		if (volume->nservers > 0)
350			/* another server might acknowledge its existence */
351			goto try_next_server_upw;
352
353		/* handle the case where all the fileservers have rejected the
354		 * volume
355		 * - TODO: try asking the fileservers for volume information
356		 * - TODO: contact the VL server again to see if the volume is
357		 *         no longer registered
358		 */
359		up_write(&volume->server_sem);
360		afs_put_server(server);
361		_leave(" [completely rejected]");
362		return 1;
363
364		/* problem reaching the server */
365	case -ENETUNREACH:
366	case -EHOSTUNREACH:
367	case -ECONNREFUSED:
368	case -ETIME:
369	case -ETIMEDOUT:
370	case -EREMOTEIO:
371		/* mark the server as dead
372		 * TODO: vary dead timeout depending on error
373		 */
374		spin_lock(&server->fs_lock);
375		if (!server->fs_state) {
376			server->fs_dead_jif = jiffies + HZ * 10;
377			server->fs_state = result;
378			printk("kAFS: SERVER DEAD state=%d\n", result);
379		}
380		spin_unlock(&server->fs_lock);
381		goto try_next_server;
382
383		/* miscellaneous error */
384	default:
385		server->fs_act_jif = jiffies;
386	case -ENOMEM:
387	case -ENONET:
388		/* tell the caller to accept the result */
389		afs_put_server(server);
390		_leave(" [local failure]");
391		return 1;
392	}
393
394	/* tell the caller to loop around and try the next server */
395try_next_server_upw:
396	up_write(&volume->server_sem);
397try_next_server:
398	afs_put_server(server);
399	_leave(" [try next server]");
400	return 0;
401}
402