[linux-2.6-block.git] / fs / afs / volume.c

// SPDX-License-Identifier: GPL-2.0-or-later
/* AFS volume management
 *
 * Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 */

#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;

/*
 * Allocate a volume record and load it up from a vldb record.
 */
static struct afs_volume *afs_alloc_volume(struct afs_fs_context *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);
	rwlock_init(&volume->cb_v_break_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_fs_context *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
2874c5fd	1	// SPDX-License-Identifier: GPL-2.0-or-later
ec26815a	2	/* AFS volume management
1da177e4	3	*
08e0e7c8	4	* Copyright (C) 2002, 2007 Red Hat, Inc. All Rights Reserved.
1da177e4	5	* Written by David Howells (dhowells@redhat.com)
1da177e4 LT	6	*/
	7
	8	#include <linux/kernel.h>
1da177e4	9	#include <linux/slab.h>
1da177e4 LT	10	#include "internal.h"
1da177e4 LT	11
d2ddc776 DH	12	unsigned __read_mostly afs_volume_gc_delay = 10;
	13	unsigned __read_mostly afs_volume_record_life = 60 * 60;
	14
1da177e4	15	/*
d2ddc776 DH	16	* Allocate a volume record and load it up from a vldb record.
d2ddc776 DH	17	*/
13fcc683	18	static struct afs_volume afs_alloc_volume(struct afs_fs_context params,
d2ddc776 DH	19	struct afs_vldb_entry *vldb,
	20	unsigned long type_mask)
	21	{
	22	struct afs_server_list *slist;
d2ddc776	23	struct afs_volume *volume;
45df8462	24	int ret = -ENOMEM, nr_servers = 0, i;
d2ddc776 DH	25
	26	for (i = 0; i < vldb->nr_servers; i++)
	27	if (vldb->fs_mask[i] & type_mask)
	28	nr_servers++;
	29
	30	volume = kzalloc(sizeof(struct afs_volume), GFP_KERNEL);
	31	if (!volume)
	32	goto error_0;
	33
	34	volume->vid = vldb->vid[params->type];
	35	volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
	36	volume->cell = afs_get_cell(params->cell);
	37	volume->type = params->type;
	38	volume->type_force = params->force;
	39	volume->name_len = vldb->name_len;
	40
	41	atomic_set(&volume->usage, 1);
	42	INIT_LIST_HEAD(&volume->proc_link);
	43	rwlock_init(&volume->servers_lock);
90fa9b64	44	rwlock_init(&volume->cb_v_break_lock);
d2ddc776 DH	45	memcpy(volume->name, vldb->name, vldb->name_len + 1);
	46
	47	slist = afs_alloc_server_list(params->cell, params->key, vldb, type_mask);
	48	if (IS_ERR(slist)) {
	49	ret = PTR_ERR(slist);
	50	goto error_1;
	51	}
	52
	53	refcount_set(&slist->usage, 1);
	54	volume->servers = slist;
d2ddc776 DH	55	return volume;
d2ddc776 DH	56
d2ddc776	57	error_1:
e4415015	58	afs_put_cell(params->net, volume->cell);
d2ddc776 DH	59	kfree(volume);
	60	error_0:
	61	return ERR_PTR(ret);
	62	}
	63
	64	/*
	65	* Look up a VLDB record for a volume.
	66	*/
	67	static struct afs_vldb_entry afs_vl_lookup_vldb(struct afs_cell cell,
	68	struct key *key,
	69	const char *volname,
	70	size_t volnamesz)
	71	{
0a5143f2 DH	72	struct afs_vldb_entry *vldb = ERR_PTR(-EDESTADDRREQ);
0a5143f2 DH	73	struct afs_vl_cursor vc;
d2ddc776 DH	74	int ret;
d2ddc776 DH	75
0a5143f2 DH	76	if (!afs_begin_vlserver_operation(&vc, cell, key))
0a5143f2 DH	77	return ERR_PTR(-ERESTARTSYS);
d2ddc776	78
0a5143f2	79	while (afs_select_vlserver(&vc)) {
0a5143f2	80	vldb = afs_vl_get_entry_by_name_u(&vc, volname, volnamesz);
d2ddc776 DH	81	}
d2ddc776 DH	82
0a5143f2 DH	83	ret = afs_end_vlserver_operation(&vc);
0a5143f2 DH	84	return ret < 0 ? ERR_PTR(ret) : vldb;
d2ddc776 DH	85	}
	86
	87	/*
	88	* Look up a volume in the VL server and create a candidate volume record for
	89	* it.
	90	*
	91	* The volume name can be one of the following:
1da177e4 LT	92	* "%[cell:]volume[.]" R/W volume
	93	* "#[cell:]volume[.]" R/O or R/W volume (rwparent=0),
	94	* or R/W (rwparent=1) volume
	95	* "%[cell:]volume.readonly" R/O volume
	96	* "#[cell:]volume.readonly" R/O volume
	97	* "%[cell:]volume.backup" Backup volume
	98	* "#[cell:]volume.backup" Backup volume
	99	*
	100	* The cell name is optional, and defaults to the current cell.
	101	*
	102	* See "The Rules of Mount Point Traversal" in Chapter 5 of the AFS SysAdmin
	103	* Guide
	104	* - Rule 1: Explicit type suffix forces access of that type or nothing
	105	* (no suffix, then use Rule 2 & 3)
	106	* - Rule 2: If parent volume is R/O, then mount R/O volume by preference, R/W
	107	* if not available
	108	* - Rule 3: If parent volume is R/W, then only mount R/W volume unless
	109	* explicitly told otherwise
	110	*/
13fcc683	111	struct afs_volume afs_create_volume(struct afs_fs_context params)
1da177e4	112	{
d2ddc776 DH	113	struct afs_vldb_entry *vldb;
	114	struct afs_volume *volume;
	115	unsigned long type_mask = 1UL << params->type;
1da177e4	116
d2ddc776 DH	117	vldb = afs_vl_lookup_vldb(params->cell, params->key,
	118	params->volname, params->volnamesz);
	119	if (IS_ERR(vldb))
	120	return ERR_CAST(vldb);
1da177e4	121
d2ddc776 DH	122	if (test_bit(AFS_VLDB_QUERY_ERROR, &vldb->flags)) {
	123	volume = ERR_PTR(vldb->error);
	124	goto error;
	125	}
1da177e4	126
d2ddc776 DH	127	/* Make the final decision on the type we want */
d2ddc776 DH	128	volume = ERR_PTR(-ENOMEDIUM);
00d3b7a4	129	if (params->force) {
d2ddc776	130	if (!(vldb->flags & type_mask))
1da177e4	131	goto error;
d2ddc776	132	} else if (test_bit(AFS_VLDB_HAS_RO, &vldb->flags)) {
00d3b7a4	133	params->type = AFSVL_ROVOL;
d2ddc776	134	} else if (test_bit(AFS_VLDB_HAS_RW, &vldb->flags)) {
00d3b7a4	135	params->type = AFSVL_RWVOL;
ec26815a	136	} else {
1da177e4 LT	137	goto error;
	138	}
	139
d2ddc776 DH	140	type_mask = 1UL << params->type;
d2ddc776 DH	141	volume = afs_alloc_volume(params, vldb, type_mask);
1da177e4	142
d2ddc776 DH	143	error:
	144	kfree(vldb);
	145	return volume;
	146	}
1da177e4	147
d2ddc776 DH	148	/*
	149	* Destroy a volume record
	150	*/
	151	static void afs_destroy_volume(struct afs_net net, struct afs_volume volume)
	152	{
	153	_enter("%p", volume);
1da177e4	154
d2ddc776 DH	155	#ifdef CONFIG_AFS_FSCACHE
	156	ASSERTCMP(volume->cache, ==, NULL);
	157	#endif
1da177e4	158
d2ddc776 DH	159	afs_put_serverlist(net, volume->servers);
	160	afs_put_cell(net, volume->cell);
	161	kfree(volume);
1da177e4	162
d2ddc776 DH	163	_leave(" [destroyed]");
	164	}
	165
	166	/*
	167	* Drop a reference on a volume record.
	168	*/
	169	void afs_put_volume(struct afs_cell cell, struct afs_volume volume)
	170	{
	171	if (volume) {
	172	_enter("%s", volume->name);
	173
	174	if (atomic_dec_and_test(&volume->usage))
	175	afs_destroy_volume(cell->net, volume);
1da177e4	176	}
d2ddc776	177	}
1da177e4	178
d2ddc776 DH	179	/*
	180	* Activate a volume.
	181	*/
	182	void afs_activate_volume(struct afs_volume *volume)
	183	{
9b3f26c9	184	#ifdef CONFIG_AFS_FSCACHE
ad6a942a	185	volume->cache = fscache_acquire_cookie(volume->cell->cache,
9b3f26c9	186	&afs_volume_cache_index_def,
402cb8dd DH	187	&volume->vid, sizeof(volume->vid),
402cb8dd DH	188	NULL, 0,
ee1235a9	189	volume, 0, true);
1da177e4	190	#endif
1da177e4	191
d2ddc776 DH	192	write_lock(&volume->cell->proc_lock);
	193	list_add_tail(&volume->proc_link, &volume->cell->proc_volumes);
	194	write_unlock(&volume->cell->proc_lock);
	195	}
1da177e4	196
d2ddc776 DH	197	/*
	198	* Deactivate a volume.
	199	*/
	200	void afs_deactivate_volume(struct afs_volume *volume)
	201	{
	202	_enter("%s", volume->name);
1da177e4	203
d2ddc776 DH	204	write_lock(&volume->cell->proc_lock);
	205	list_del_init(&volume->proc_link);
	206	write_unlock(&volume->cell->proc_lock);
1da177e4	207
d2ddc776	208	#ifdef CONFIG_AFS_FSCACHE
402cb8dd	209	fscache_relinquish_cookie(volume->cache, NULL,
d2ddc776 DH	210	test_bit(AFS_VOLUME_DELETED, &volume->flags));
	211	volume->cache = NULL;
	212	#endif
1da177e4	213
d2ddc776	214	_leave("");
ec26815a	215	}
1da177e4	216
1da177e4	217	/*
d2ddc776	218	* Query the VL service to update the volume status.
1da177e4	219	*/
d2ddc776	220	static int afs_update_volume_status(struct afs_volume volume, struct key key)
1da177e4	221	{
d2ddc776 DH	222	struct afs_server_list new, old, *discard;
	223	struct afs_vldb_entry *vldb;
	224	char idbuf[16];
	225	int ret, idsz;
1da177e4	226
d2ddc776	227	_enter("");
1da177e4	228
d2ddc776 DH	229	/* We look up an ID by passing it as a decimal string in the
	230	* operation's name parameter.
	231	*/
3b6492df	232	idsz = sprintf(idbuf, "%llu", volume->vid);
1da177e4	233
d2ddc776 DH	234	vldb = afs_vl_lookup_vldb(volume->cell, key, idbuf, idsz);
	235	if (IS_ERR(vldb)) {
	236	ret = PTR_ERR(vldb);
	237	goto error;
	238	}
1da177e4	239
d2ddc776 DH	240	/* See if the volume got renamed. */
	241	if (vldb->name_len != volume->name_len \|\|
	242	memcmp(vldb->name, volume->name, vldb->name_len) != 0) {
	243	/* TODO: Use RCU'd string. */
	244	memcpy(volume->name, vldb->name, AFS_MAXVOLNAME);
	245	volume->name_len = vldb->name_len;
	246	}
	247
	248	/* See if the volume's server list got updated. */
	249	new = afs_alloc_server_list(volume->cell, key,
45df8462	250	vldb, (1 << volume->type));
d2ddc776 DH	251	if (IS_ERR(new)) {
	252	ret = PTR_ERR(new);
	253	goto error_vldb;
	254	}
1da177e4	255
d2ddc776	256	write_lock(&volume->servers_lock);
1da177e4	257
d2ddc776 DH	258	discard = new;
	259	old = volume->servers;
	260	if (afs_annotate_server_list(new, old)) {
	261	new->seq = volume->servers_seq + 1;
	262	volume->servers = new;
	263	smp_wmb();
	264	volume->servers_seq++;
	265	discard = old;
1da177e4 LT	266	}
1da177e4 LT	267
d2ddc776 DH	268	volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
	269	clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
	270	write_unlock(&volume->servers_lock);
	271	ret = 0;
1da177e4	272
d2ddc776 DH	273	afs_put_serverlist(volume->cell->net, discard);
	274	error_vldb:
	275	kfree(vldb);
	276	error:
	277	_leave(" = %d", ret);
	278	return ret;
	279	}
1da177e4	280
d2ddc776 DH	281	/*
	282	* Make sure the volume record is up to date.
	283	*/
	284	int afs_check_volume_status(struct afs_volume volume, struct key key)
	285	{
	286	time64_t now = ktime_get_real_seconds();
	287	int ret, retries = 0;
1da177e4	288
d2ddc776 DH	289	_enter("");
	290
	291	if (volume->update_at <= now)
	292	set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
	293
	294	retry:
	295	if (!test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags) &&
	296	!test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
	297	_leave(" = 0");
	298	return 0;
c435ee34	299	}
1da177e4	300
d2ddc776 DH	301	if (!test_and_set_bit_lock(AFS_VOLUME_UPDATING, &volume->flags)) {
	302	ret = afs_update_volume_status(volume, key);
	303	clear_bit_unlock(AFS_VOLUME_WAIT, &volume->flags);
	304	clear_bit_unlock(AFS_VOLUME_UPDATING, &volume->flags);
	305	wake_up_bit(&volume->flags, AFS_VOLUME_WAIT);
	306	_leave(" = %d", ret);
	307	return ret;
	308	}
1da177e4	309
d2ddc776 DH	310	if (!test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
	311	_leave(" = 0 [no wait]");
	312	return 0;
	313	}
	314
	315	ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT, TASK_INTERRUPTIBLE);
	316	if (ret == -ERESTARTSYS) {
	317	_leave(" = %d", ret);
	318	return ret;
	319	}
	320
	321	retries++;
	322	if (retries == 4) {
	323	_leave(" = -ESTALE");
	324	return -ESTALE;
	325	}
	326	goto retry;
ec26815a	327	}