[people/ms/linux.git] / fs / afs / volume.c

/* AFS volume management
 *
 * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */

#include <linux/kernel.h>
#include <linux/slab.h>
#include "internal.h"

unsigned __read_mostly afs_volume_gc_delay = 10;
unsigned __read_mostly afs_volume_record_life = 60 * 60;

static const char *const afs_voltypes[] = { "R/W", "R/O", "BAK" };

/*
 * Allocate a volume record and load it up from a vldb record.
 */
static struct afs_volume *afs_alloc_volume(struct afs_mount_params *params,
					   struct afs_vldb_entry *vldb,
					   unsigned long type_mask)
{
	struct afs_server_list *slist;
	struct afs_volume *volume;
	int ret = -ENOMEM, nr_servers = 0, i;

	for (i = 0; i < vldb->nr_servers; i++)
		if (vldb->fs_mask[i] & type_mask)
			nr_servers++;

	volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
	if (!volume)
		goto error_0;

	volume->vid		= vldb->vid[params->type];
	volume->update_at	= ktime_get_real_seconds() + afs_volume_record_life;
	volume->cell		= afs_get_cell(params->cell);
	volume->type		= params->type;
	volume->type_force	= params->force;
	volume->name_len	= vldb->name_len;

	atomic_set(&volume->usage, 1);
	INIT_LIST_HEAD(&volume->proc_link);
	rwlock_init(&volume->servers_lock);
	memcpy(volume->name, vldb->name, vldb->name_len + 1);

	slist = afs_alloc_server_list(params->cell, params->key, vldb, type_mask);
	if (IS_ERR(slist)) {
		ret = PTR_ERR(slist);
		goto error_1;
	}

	refcount_set(&slist->usage, 1);
	volume->servers = slist;
	return volume;

error_1:
	afs_put_cell(params->net, volume->cell);
	kfree(volume);
error_0:
	return ERR_PTR(ret);
}

/*
 * Look up a VLDB record for a volume.
 */
static struct afs_vldb_entry *afs_vl_lookup_vldb(struct afs_cell *cell,
						 struct key *key,
						 const char *volname,
						 size_t volnamesz)
{
	struct afs_vldb_entry *vldb = ERR_PTR(-EDESTADDRREQ);
	struct afs_vl_cursor vc;
	int ret;

	if (!afs_begin_vlserver_operation(&vc, cell, key))
		return ERR_PTR(-ERESTARTSYS);

	while (afs_select_vlserver(&vc)) {
		vldb = afs_vl_get_entry_by_name_u(&vc, volname, volnamesz);
	}

	ret = afs_end_vlserver_operation(&vc);
	return ret < 0 ? ERR_PTR(ret) : vldb;
}

/*
 * Look up a volume in the VL server and create a candidate volume record for
 * it.
 *
 * The volume name can be one of the following:
 *	"%[cell:]volume[.]"		R/W volume
 *	"#[cell:]volume[.]"		R/O or R/W volume (rwparent=0),
 *					 or R/W (rwparent=1) volume
 *	"%[cell:]volume.readonly"	R/O volume
 *	"#[cell:]volume.readonly"	R/O volume
 *	"%[cell:]volume.backup"		Backup volume
 *	"#[cell:]volume.backup"		Backup volume
 *
 * The cell name is optional, and defaults to the current cell.
 *
 * See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
 * Guide
 * - Rule 1: Explicit type suffix forces access of that type or nothing
 *           (no suffix, then use Rule 2 & 3)
 * - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
 *           if not available
 * - Rule 3: If parent volume is R/W, then only mount R/W volume unless
 *           explicitly told otherwise
 */
struct afs_volume *afs_create_volume(struct afs_mount_params *params)
{
	struct afs_vldb_entry *vldb;
	struct afs_volume *volume;
	unsigned long type_mask = 1UL << params->type;

	vldb = afs_vl_lookup_vldb(params->cell, params->key,
				  params->volname, params->volnamesz);
	if (IS_ERR(vldb))
		return ERR_CAST(vldb);

	if (test_bit(AFS_VLDB_QUERY_ERROR, &vldb->flags)) {
		volume = ERR_PTR(vldb->error);
		goto error;
	}

	/* Make the final decision on the type we want */
	volume = ERR_PTR(-ENOMEDIUM);
	if (params->force) {
		if (!(vldb->flags & type_mask))
			goto error;
	} else if (test_bit(AFS_VLDB_HAS_RO, &vldb->flags)) {
		params->type = AFSVL_ROVOL;
	} else if (test_bit(AFS_VLDB_HAS_RW, &vldb->flags)) {
		params->type = AFSVL_RWVOL;
	} else {
		goto error;
	}

	type_mask = 1UL << params->type;
	volume = afs_alloc_volume(params, vldb, type_mask);

error:
	kfree(vldb);
	return volume;
}

/*
 * Destroy a volume record
 */
static void afs_destroy_volume(struct afs_net *net, struct afs_volume *volume)
{
	_enter("%p", volume);

#ifdef CONFIG_AFS_FSCACHE
	ASSERTCMP(volume->cache, ==, NULL);
#endif

	afs_put_serverlist(net, volume->servers);
	afs_put_cell(net, volume->cell);
	kfree(volume);

	_leave(" [destroyed]");
}

/*
 * Drop a reference on a volume record.
 */
void afs_put_volume(struct afs_cell *cell, struct afs_volume *volume)
{
	if (volume) {
		_enter("%s", volume->name);

		if (atomic_dec_and_test(&volume->usage))
			afs_destroy_volume(cell->net, volume);
	}
}

/*
 * Activate a volume.
 */
void afs_activate_volume(struct afs_volume *volume)
{
#ifdef CONFIG_AFS_FSCACHE
	volume->cache = fscache_acquire_cookie(volume->cell->cache,
					       &afs_volume_cache_index_def,
					       &volume->vid, sizeof(volume->vid),
					       NULL, 0,
					       volume, 0, true);
#endif

	write_lock(&volume->cell->proc_lock);
	list_add_tail(&volume->proc_link, &volume->cell->proc_volumes);
	write_unlock(&volume->cell->proc_lock);
}

/*
 * Deactivate a volume.
 */
void afs_deactivate_volume(struct afs_volume *volume)
{
	_enter("%s", volume->name);

	write_lock(&volume->cell->proc_lock);
	list_del_init(&volume->proc_link);
	write_unlock(&volume->cell->proc_lock);

#ifdef CONFIG_AFS_FSCACHE
	fscache_relinquish_cookie(volume->cache, NULL,
				  test_bit(AFS_VOLUME_DELETED, &volume->flags));
	volume->cache = NULL;
#endif

	_leave("");
}

/*
 * Query the VL service to update the volume status.
 */
static int afs_update_volume_status(struct afs_volume *volume, struct key *key)
{
	struct afs_server_list *new, *old, *discard;
	struct afs_vldb_entry *vldb;
	char idbuf[16];
	int ret, idsz;

	_enter("");

	/* We look up an ID by passing it as a decimal string in the
	 * operation's name parameter.
	 */
	idsz = sprintf(idbuf, "%llu", volume->vid);

	vldb = afs_vl_lookup_vldb(volume->cell, key, idbuf, idsz);
	if (IS_ERR(vldb)) {
		ret = PTR_ERR(vldb);
		goto error;
	}

	/* See if the volume got renamed. */
	if (vldb->name_len != volume->name_len ||
	    memcmp(vldb->name, volume->name, vldb->name_len) != 0) {
		/* TODO: Use RCU'd string. */
		memcpy(volume->name, vldb->name, AFS_MAXVOLNAME);
		volume->name_len = vldb->name_len;
	}

	/* See if the volume's server list got updated. */
	new = afs_alloc_server_list(volume->cell, key,
				    vldb, (1 << volume->type));
	if (IS_ERR(new)) {
		ret = PTR_ERR(new);
		goto error_vldb;
	}

	write_lock(&volume->servers_lock);

	discard = new;
	old = volume->servers;
	if (afs_annotate_server_list(new, old)) {
		new->seq = volume->servers_seq + 1;
		volume->servers = new;
		smp_wmb();
		volume->servers_seq++;
		discard = old;
	}

	volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
	clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
	write_unlock(&volume->servers_lock);
	ret = 0;

	afs_put_serverlist(volume->cell->net, discard);
error_vldb:
	kfree(vldb);
error:
	_leave(" = %d", ret);
	return ret;
}

/*
 * Make sure the volume record is up to date.
 */
int afs_check_volume_status(struct afs_volume *volume, struct key *key)
{
	time64_t now = ktime_get_real_seconds();
	int ret, retries = 0;

	_enter("");

	if (volume->update_at <= now)
		set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);

retry:
	if (!test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags) &&
	    !test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
		_leave(" = 0");
		return 0;
	}

	if (!test_and_set_bit_lock(AFS_VOLUME_UPDATING, &volume->flags)) {
		ret = afs_update_volume_status(volume, key);
		clear_bit_unlock(AFS_VOLUME_WAIT, &volume->flags);
		clear_bit_unlock(AFS_VOLUME_UPDATING, &volume->flags);
		wake_up_bit(&volume->flags, AFS_VOLUME_WAIT);
		_leave(" = %d", ret);
		return ret;
	}

	if (!test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
		_leave(" = 0 [no wait]");
		return 0;
	}

	ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT, TASK_INTERRUPTIBLE);
	if (ret == -ERESTARTSYS) {
		_leave(" = %d", ret);
		return ret;
	}

	retries++;
	if (retries == 4) {
		_leave(" = -ESTALE");
		return -ESTALE;
	}
	goto retry;
}
Commit	Line	Data
ec26815a	1	/* AFS volume management
1da177e4	2	*
08e0e7c8	3	* Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
1da177e4 LT	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>
1da177e4	13	#include <linux/slab.h>
1da177e4 LT	14	#include "internal.h"
1da177e4 LT	15
d2ddc776 DH	16	unsigned __read_mostly afs_volume_gc_delay = 10;
	17	unsigned __read_mostly afs_volume_record_life = 60 * 60;
	18
	19	static const char *const afs_voltypes[] = { "R/W", "R/O", "BAK" };
1da177e4	20
1da177e4	21	/*
d2ddc776 DH	22	* Allocate a volume record and load it up from a vldb record.
	23	*/
	24	static struct afs_volume afs_alloc_volume(struct afs_mount_params params,
	25	struct afs_vldb_entry *vldb,
	26	unsigned long type_mask)
	27	{
	28	struct afs_server_list *slist;
d2ddc776	29	struct afs_volume *volume;
45df8462	30	int ret = -ENOMEM, nr_servers = 0, i;
d2ddc776 DH	31
	32	for (i = 0; i < vldb->nr_servers; i++)
	33	if (vldb->fs_mask[i] & type_mask)
	34	nr_servers++;
	35
	36	volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
	37	if (!volume)
	38	goto error_0;
	39
	40	volume->vid = vldb->vid[params->type];
	41	volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
	42	volume->cell = afs_get_cell(params->cell);
	43	volume->type = params->type;
	44	volume->type_force = params->force;
	45	volume->name_len = vldb->name_len;
	46
	47	atomic_set(&volume->usage, 1);
	48	INIT_LIST_HEAD(&volume->proc_link);
	49	rwlock_init(&volume->servers_lock);
	50	memcpy(volume->name, vldb->name, vldb->name_len + 1);
	51
	52	slist = afs_alloc_server_list(params->cell, params->key, vldb, type_mask);
	53	if (IS_ERR(slist)) {
	54	ret = PTR_ERR(slist);
	55	goto error_1;
	56	}
	57
	58	refcount_set(&slist->usage, 1);
	59	volume->servers = slist;
d2ddc776 DH	60	return volume;
d2ddc776 DH	61
d2ddc776	62	error_1:
e4415015	63	afs_put_cell(params->net, volume->cell);
d2ddc776 DH	64	kfree(volume);
	65	error_0:
	66	return ERR_PTR(ret);
	67	}
	68
	69	/*
	70	* Look up a VLDB record for a volume.
	71	*/
	72	static struct afs_vldb_entry afs_vl_lookup_vldb(struct afs_cell cell,
	73	struct key *key,
	74	const char *volname,
	75	size_t volnamesz)
	76	{
0a5143f2 DH	77	struct afs_vldb_entry *vldb = ERR_PTR(-EDESTADDRREQ);
0a5143f2 DH	78	struct afs_vl_cursor vc;
d2ddc776 DH	79	int ret;
d2ddc776 DH	80
0a5143f2 DH	81	if (!afs_begin_vlserver_operation(&vc, cell, key))
0a5143f2 DH	82	return ERR_PTR(-ERESTARTSYS);
d2ddc776	83
0a5143f2	84	while (afs_select_vlserver(&vc)) {
0a5143f2	85	vldb = afs_vl_get_entry_by_name_u(&vc, volname, volnamesz);
d2ddc776 DH	86	}
d2ddc776 DH	87
0a5143f2 DH	88	ret = afs_end_vlserver_operation(&vc);
0a5143f2 DH	89	return ret < 0 ? ERR_PTR(ret) : vldb;
d2ddc776 DH	90	}
	91
	92	/*
	93	* Look up a volume in the VL server and create a candidate volume record for
	94	* it.
	95	*
	96	* The volume name can be one of the following:
1da177e4 LT	97	* "%[cell:]volume[.]" R/W volume
	98	* "#[cell:]volume[.]" R/O or R/W volume (rwparent=0),
	99	* or R/W (rwparent=1) volume
	100	* "%[cell:]volume.readonly" R/O volume
	101	* "#[cell:]volume.readonly" R/O volume
	102	* "%[cell:]volume.backup" Backup volume
	103	* "#[cell:]volume.backup" Backup volume
	104	*
	105	* The cell name is optional, and defaults to the current cell.
	106	*
	107	* See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
	108	* Guide
	109	* - Rule 1: Explicit type suffix forces access of that type or nothing
	110	* (no suffix, then use Rule 2 & 3)
	111	* - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
	112	* if not available
	113	* - Rule 3: If parent volume is R/W, then only mount R/W volume unless
	114	* explicitly told otherwise
	115	*/
d2ddc776	116	struct afs_volume afs_create_volume(struct afs_mount_params params)
1da177e4	117	{
d2ddc776 DH	118	struct afs_vldb_entry *vldb;
	119	struct afs_volume *volume;
	120	unsigned long type_mask = 1UL << params->type;
1da177e4	121
d2ddc776 DH	122	vldb = afs_vl_lookup_vldb(params->cell, params->key,
	123	params->volname, params->volnamesz);
	124	if (IS_ERR(vldb))
	125	return ERR_CAST(vldb);
1da177e4	126
d2ddc776 DH	127	if (test_bit(AFS_VLDB_QUERY_ERROR, &vldb->flags)) {
	128	volume = ERR_PTR(vldb->error);
	129	goto error;
	130	}
1da177e4	131
d2ddc776 DH	132	/* Make the final decision on the type we want */
d2ddc776 DH	133	volume = ERR_PTR(-ENOMEDIUM);
00d3b7a4	134	if (params->force) {
d2ddc776	135	if (!(vldb->flags & type_mask))
1da177e4	136	goto error;
d2ddc776	137	} else if (test_bit(AFS_VLDB_HAS_RO, &vldb->flags)) {
00d3b7a4	138	params->type = AFSVL_ROVOL;
d2ddc776	139	} else if (test_bit(AFS_VLDB_HAS_RW, &vldb->flags)) {
00d3b7a4	140	params->type = AFSVL_RWVOL;
ec26815a	141	} else {
1da177e4 LT	142	goto error;
	143	}
	144
d2ddc776 DH	145	type_mask = 1UL << params->type;
d2ddc776 DH	146	volume = afs_alloc_volume(params, vldb, type_mask);
1da177e4	147
d2ddc776 DH	148	error:
	149	kfree(vldb);
	150	return volume;
	151	}
1da177e4	152
d2ddc776 DH	153	/*
	154	* Destroy a volume record
	155	*/
	156	static void afs_destroy_volume(struct afs_net net, struct afs_volume volume)
	157	{
	158	_enter("%p", volume);
1da177e4	159
d2ddc776 DH	160	#ifdef CONFIG_AFS_FSCACHE
	161	ASSERTCMP(volume->cache, ==, NULL);
	162	#endif
1da177e4	163
d2ddc776 DH	164	afs_put_serverlist(net, volume->servers);
	165	afs_put_cell(net, volume->cell);
	166	kfree(volume);
1da177e4	167
d2ddc776 DH	168	_leave(" [destroyed]");
	169	}
	170
	171	/*
	172	* Drop a reference on a volume record.
	173	*/
	174	void afs_put_volume(struct afs_cell cell, struct afs_volume volume)
	175	{
	176	if (volume) {
	177	_enter("%s", volume->name);
	178
	179	if (atomic_dec_and_test(&volume->usage))
	180	afs_destroy_volume(cell->net, volume);
1da177e4	181	}
d2ddc776	182	}
1da177e4	183
d2ddc776 DH	184	/*
	185	* Activate a volume.
	186	*/
	187	void afs_activate_volume(struct afs_volume *volume)
	188	{
9b3f26c9	189	#ifdef CONFIG_AFS_FSCACHE
ad6a942a	190	volume->cache = fscache_acquire_cookie(volume->cell->cache,
9b3f26c9	191	&afs_volume_cache_index_def,
402cb8dd DH	192	&volume->vid, sizeof(volume->vid),
402cb8dd DH	193	NULL, 0,
ee1235a9	194	volume, 0, true);
1da177e4	195	#endif
1da177e4	196
d2ddc776 DH	197	write_lock(&volume->cell->proc_lock);
	198	list_add_tail(&volume->proc_link, &volume->cell->proc_volumes);
	199	write_unlock(&volume->cell->proc_lock);
	200	}
1da177e4	201
d2ddc776 DH	202	/*
	203	* Deactivate a volume.
	204	*/
	205	void afs_deactivate_volume(struct afs_volume *volume)
	206	{
	207	_enter("%s", volume->name);
1da177e4	208
d2ddc776 DH	209	write_lock(&volume->cell->proc_lock);
	210	list_del_init(&volume->proc_link);
	211	write_unlock(&volume->cell->proc_lock);
1da177e4	212
d2ddc776	213	#ifdef CONFIG_AFS_FSCACHE
402cb8dd	214	fscache_relinquish_cookie(volume->cache, NULL,
d2ddc776 DH	215	test_bit(AFS_VOLUME_DELETED, &volume->flags));
	216	volume->cache = NULL;
	217	#endif
1da177e4	218
d2ddc776	219	_leave("");
ec26815a	220	}
1da177e4	221
1da177e4	222	/*
d2ddc776	223	* Query the VL service to update the volume status.
1da177e4	224	*/
d2ddc776	225	static int afs_update_volume_status(struct afs_volume volume, struct key key)
1da177e4	226	{
d2ddc776 DH	227	struct afs_server_list new, old, *discard;
	228	struct afs_vldb_entry *vldb;
	229	char idbuf[16];
	230	int ret, idsz;
1da177e4	231
d2ddc776	232	_enter("");
1da177e4	233
d2ddc776 DH	234	/* We look up an ID by passing it as a decimal string in the
	235	* operation's name parameter.
	236	*/
3b6492df	237	idsz = sprintf(idbuf, "%llu", volume->vid);
1da177e4	238
d2ddc776 DH	239	vldb = afs_vl_lookup_vldb(volume->cell, key, idbuf, idsz);
	240	if (IS_ERR(vldb)) {
	241	ret = PTR_ERR(vldb);
	242	goto error;
	243	}
1da177e4	244
d2ddc776 DH	245	/* See if the volume got renamed. */
	246	if (vldb->name_len != volume->name_len \|\|
	247	memcmp(vldb->name, volume->name, vldb->name_len) != 0) {
	248	/* TODO: Use RCU'd string. */
	249	memcpy(volume->name, vldb->name, AFS_MAXVOLNAME);
	250	volume->name_len = vldb->name_len;
	251	}
	252
	253	/* See if the volume's server list got updated. */
	254	new = afs_alloc_server_list(volume->cell, key,
45df8462	255	vldb, (1 << volume->type));
d2ddc776 DH	256	if (IS_ERR(new)) {
	257	ret = PTR_ERR(new);
	258	goto error_vldb;
	259	}
1da177e4	260
d2ddc776	261	write_lock(&volume->servers_lock);
1da177e4	262
d2ddc776 DH	263	discard = new;
	264	old = volume->servers;
	265	if (afs_annotate_server_list(new, old)) {
	266	new->seq = volume->servers_seq + 1;
	267	volume->servers = new;
	268	smp_wmb();
	269	volume->servers_seq++;
	270	discard = old;
1da177e4 LT	271	}
1da177e4 LT	272
d2ddc776 DH	273	volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
	274	clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
	275	write_unlock(&volume->servers_lock);
	276	ret = 0;
1da177e4	277
d2ddc776 DH	278	afs_put_serverlist(volume->cell->net, discard);
	279	error_vldb:
	280	kfree(vldb);
	281	error:
	282	_leave(" = %d", ret);
	283	return ret;
	284	}
1da177e4	285
d2ddc776 DH	286	/*
	287	* Make sure the volume record is up to date.
	288	*/
	289	int afs_check_volume_status(struct afs_volume volume, struct key key)
	290	{
	291	time64_t now = ktime_get_real_seconds();
	292	int ret, retries = 0;
1da177e4	293
d2ddc776 DH	294	_enter("");
	295
	296	if (volume->update_at <= now)
	297	set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
	298
	299	retry:
	300	if (!test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags) &&
	301	!test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
	302	_leave(" = 0");
	303	return 0;
c435ee34	304	}
1da177e4	305
d2ddc776 DH	306	if (!test_and_set_bit_lock(AFS_VOLUME_UPDATING, &volume->flags)) {
	307	ret = afs_update_volume_status(volume, key);
	308	clear_bit_unlock(AFS_VOLUME_WAIT, &volume->flags);
	309	clear_bit_unlock(AFS_VOLUME_UPDATING, &volume->flags);
	310	wake_up_bit(&volume->flags, AFS_VOLUME_WAIT);
	311	_leave(" = %d", ret);
	312	return ret;
	313	}
1da177e4	314
d2ddc776 DH	315	if (!test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
	316	_leave(" = 0 [no wait]");
	317	return 0;
	318	}
	319
	320	ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT, TASK_INTERRUPTIBLE);
	321	if (ret == -ERESTARTSYS) {
	322	_leave(" = %d", ret);
	323	return ret;
	324	}
	325
	326	retries++;
	327	if (retries == 4) {
	328	_leave(" = -ESTALE");
	329	return -ESTALE;
	330	}
	331	goto retry;
ec26815a	332	}