[mirror_ubuntu-artful-kernel.git] / fs / ceph / cache.c

/*
 * Ceph cache definitions.
 *
 *  Copyright (C) 2013 by Adfin Solutions, Inc. All Rights Reserved.
 *  Written by Milosz Tanski (milosz@adfin.com)
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License version 2
 *  as published by the Free Software Foundation.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to:
 *  Free Software Foundation
 *  51 Franklin Street, Fifth Floor
 *  Boston, MA  02111-1301  USA
 *
 */

#include "super.h"
#include "cache.h"

struct ceph_aux_inode {
	u64 		version;
	struct timespec	mtime;
	loff_t          size;
};

struct fscache_netfs ceph_cache_netfs = {
	.name		= "ceph",
	.version	= 0,
};

static DEFINE_MUTEX(ceph_fscache_lock);
static LIST_HEAD(ceph_fscache_list);

struct ceph_fscache_entry {
	struct list_head list;
	struct fscache_cookie *fscache;
	struct ceph_fsid fsid;
	size_t uniq_len;
	char uniquifier[0];
};

static uint16_t ceph_fscache_session_get_key(const void *cookie_netfs_data,
					     void *buffer, uint16_t maxbuf)
{
	const struct ceph_fs_client* fsc = cookie_netfs_data;
	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
	uint16_t fsid_len, uniq_len;

	fsid_len = sizeof(fsc->client->fsid);
	uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
	if (fsid_len + uniq_len > maxbuf)
		return 0;

	memcpy(buffer, &fsc->client->fsid, fsid_len);
	if (uniq_len)
		memcpy(buffer + fsid_len, fscache_uniq, uniq_len);

	return fsid_len + uniq_len;
}

static const struct fscache_cookie_def ceph_fscache_fsid_object_def = {
	.name		= "CEPH.fsid",
	.type		= FSCACHE_COOKIE_TYPE_INDEX,
	.get_key	= ceph_fscache_session_get_key,
};

int ceph_fscache_register(void)
{
	return fscache_register_netfs(&ceph_cache_netfs);
}

void ceph_fscache_unregister(void)
{
	fscache_unregister_netfs(&ceph_cache_netfs);
}

int ceph_fscache_register_fs(struct ceph_fs_client* fsc)
{
	const struct ceph_fsid *fsid = &fsc->client->fsid;
	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
	size_t uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
	struct ceph_fscache_entry *ent;
	int err = 0;

	mutex_lock(&ceph_fscache_lock);
	list_for_each_entry(ent, &ceph_fscache_list, list) {
		if (memcmp(&ent->fsid, fsid, sizeof(*fsid)))
			continue;
		if (ent->uniq_len != uniq_len)
			continue;
		if (uniq_len && memcmp(ent->uniquifier, fscache_uniq, uniq_len))
			continue;

		pr_err("fscache cookie already registered for fsid %pU\n", fsid);
		pr_err("  use fsc=%%s mount option to specify a uniquifier\n");
		err = -EBUSY;
		goto out_unlock;
	}

	ent = kzalloc(sizeof(*ent) + uniq_len, GFP_KERNEL);
	if (!ent) {
		err = -ENOMEM;
		goto out_unlock;
	}

	fsc->fscache = fscache_acquire_cookie(ceph_cache_netfs.primary_index,
					      &ceph_fscache_fsid_object_def,
					      fsc, true);

	if (fsc->fscache) {
		memcpy(&ent->fsid, fsid, sizeof(*fsid));
		if (uniq_len > 0) {
			memcpy(&ent->uniquifier, fscache_uniq, uniq_len);
			ent->uniq_len = uniq_len;
		}
		ent->fscache = fsc->fscache;
		list_add_tail(&ent->list, &ceph_fscache_list);
	} else {
		kfree(ent);
		pr_err("unable to register fscache cookie for fsid %pU\n",
		       fsid);
		/* all other fs ignore this error */
	}
out_unlock:
	mutex_unlock(&ceph_fscache_lock);
	return err;
}

static uint16_t ceph_fscache_inode_get_key(const void *cookie_netfs_data,
					   void *buffer, uint16_t maxbuf)
{
	const struct ceph_inode_info* ci = cookie_netfs_data;
	uint16_t klen;

	/* use ceph virtual inode (id + snapshot) */
	klen = sizeof(ci->i_vino);
	if (klen > maxbuf)
		return 0;

	memcpy(buffer, &ci->i_vino, klen);
	return klen;
}

static uint16_t ceph_fscache_inode_get_aux(const void *cookie_netfs_data,
					   void *buffer, uint16_t bufmax)
{
	struct ceph_aux_inode aux;
	const struct ceph_inode_info* ci = cookie_netfs_data;
	const struct inode* inode = &ci->vfs_inode;

	memset(&aux, 0, sizeof(aux));
	aux.version = ci->i_version;
	aux.mtime = inode->i_mtime;
	aux.size = i_size_read(inode);

	memcpy(buffer, &aux, sizeof(aux));

	return sizeof(aux);
}

static void ceph_fscache_inode_get_attr(const void *cookie_netfs_data,
					uint64_t *size)
{
	const struct ceph_inode_info* ci = cookie_netfs_data;
	*size = i_size_read(&ci->vfs_inode);
}

static enum fscache_checkaux ceph_fscache_inode_check_aux(
	void *cookie_netfs_data, const void *data, uint16_t dlen)
{
	struct ceph_aux_inode aux;
	struct ceph_inode_info* ci = cookie_netfs_data;
	struct inode* inode = &ci->vfs_inode;

	if (dlen != sizeof(aux))
		return FSCACHE_CHECKAUX_OBSOLETE;

	memset(&aux, 0, sizeof(aux));
	aux.version = ci->i_version;
	aux.mtime = inode->i_mtime;
	aux.size = i_size_read(inode);

	if (memcmp(data, &aux, sizeof(aux)) != 0)
		return FSCACHE_CHECKAUX_OBSOLETE;

	dout("ceph inode 0x%p cached okay", ci);
	return FSCACHE_CHECKAUX_OKAY;
}

static void ceph_fscache_inode_now_uncached(void* cookie_netfs_data)
{
	struct ceph_inode_info* ci = cookie_netfs_data;
	struct pagevec pvec;
	pgoff_t first;
	int loop, nr_pages;

	pagevec_init(&pvec, 0);
	first = 0;

	dout("ceph inode 0x%p now uncached", ci);

	while (1) {
		nr_pages = pagevec_lookup(&pvec, ci->vfs_inode.i_mapping, first,
					  PAGEVEC_SIZE - pagevec_count(&pvec));

		if (!nr_pages)
			break;

		for (loop = 0; loop < nr_pages; loop++)
			ClearPageFsCache(pvec.pages[loop]);

		first = pvec.pages[nr_pages - 1]->index + 1;

		pvec.nr = nr_pages;
		pagevec_release(&pvec);
		cond_resched();
	}
}

static const struct fscache_cookie_def ceph_fscache_inode_object_def = {
	.name		= "CEPH.inode",
	.type		= FSCACHE_COOKIE_TYPE_DATAFILE,
	.get_key	= ceph_fscache_inode_get_key,
	.get_attr	= ceph_fscache_inode_get_attr,
	.get_aux	= ceph_fscache_inode_get_aux,
	.check_aux	= ceph_fscache_inode_check_aux,
	.now_uncached	= ceph_fscache_inode_now_uncached,
};

void ceph_fscache_register_inode_cookie(struct inode *inode)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);

	/* No caching for filesystem */
	if (fsc->fscache == NULL)
		return;

	/* Only cache for regular files that are read only */
	if (!S_ISREG(inode->i_mode))
		return;

	inode_lock_nested(inode, I_MUTEX_CHILD);
	if (!ci->fscache) {
		ci->fscache = fscache_acquire_cookie(fsc->fscache,
					&ceph_fscache_inode_object_def,
					ci, false);
	}
	inode_unlock(inode);
}

void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci)
{
	struct fscache_cookie* cookie;

	if ((cookie = ci->fscache) == NULL)
		return;

	ci->fscache = NULL;

	fscache_uncache_all_inode_pages(cookie, &ci->vfs_inode);
	fscache_relinquish_cookie(cookie, 0);
}

static bool ceph_fscache_can_enable(void *data)
{
	struct inode *inode = data;
	return !inode_is_open_for_write(inode);
}

void ceph_fscache_file_set_cookie(struct inode *inode, struct file *filp)
{
	struct ceph_inode_info *ci = ceph_inode(inode);

	if (!fscache_cookie_valid(ci->fscache))
		return;

	if (inode_is_open_for_write(inode)) {
		dout("fscache_file_set_cookie %p %p disabling cache\n",
		     inode, filp);
		fscache_disable_cookie(ci->fscache, false);
		fscache_uncache_all_inode_pages(ci->fscache, inode);
	} else {
		fscache_enable_cookie(ci->fscache, ceph_fscache_can_enable,
				inode);
		if (fscache_cookie_enabled(ci->fscache)) {
			dout("fscache_file_set_cookie %p %p enabling cache\n",
			     inode, filp);
		}
	}
}

static void ceph_vfs_readpage_complete(struct page *page, void *data, int error)
{
	if (!error)
		SetPageUptodate(page);
}

static void ceph_vfs_readpage_complete_unlock(struct page *page, void *data, int error)
{
	if (!error)
		SetPageUptodate(page);

	unlock_page(page);
}

static inline bool cache_valid(struct ceph_inode_info *ci)
{
	return ci->i_fscache_gen == ci->i_rdcache_gen;
}


/* Atempt to read from the fscache,
 *
 * This function is called from the readpage_nounlock context. DO NOT attempt to
 * unlock the page here (or in the callback).
 */
int ceph_readpage_from_fscache(struct inode *inode, struct page *page)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int ret;

	if (!cache_valid(ci))
		return -ENOBUFS;

	ret = fscache_read_or_alloc_page(ci->fscache, page,
					 ceph_vfs_readpage_complete, NULL,
					 GFP_KERNEL);

	switch (ret) {
		case 0: /* Page found */
			dout("page read submitted\n");
			return 0;
		case -ENOBUFS: /* Pages were not found, and can't be */
		case -ENODATA: /* Pages were not found */
			dout("page/inode not in cache\n");
			return ret;
		default:
			dout("%s: unknown error ret = %i\n", __func__, ret);
			return ret;
	}
}

int ceph_readpages_from_fscache(struct inode *inode,
				  struct address_space *mapping,
				  struct list_head *pages,
				  unsigned *nr_pages)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int ret;

	if (!cache_valid(ci))
		return -ENOBUFS;

	ret = fscache_read_or_alloc_pages(ci->fscache, mapping, pages, nr_pages,
					  ceph_vfs_readpage_complete_unlock,
					  NULL, mapping_gfp_mask(mapping));

	switch (ret) {
		case 0: /* All pages found */
			dout("all-page read submitted\n");
			return 0;
		case -ENOBUFS: /* Some pages were not found, and can't be */
		case -ENODATA: /* some pages were not found */
			dout("page/inode not in cache\n");
			return ret;
		default:
			dout("%s: unknown error ret = %i\n", __func__, ret);
			return ret;
	}
}

void ceph_readpage_to_fscache(struct inode *inode, struct page *page)
{
	struct ceph_inode_info *ci = ceph_inode(inode);
	int ret;

	if (!PageFsCache(page))
		return;

	if (!cache_valid(ci))
		return;

	ret = fscache_write_page(ci->fscache, page, GFP_KERNEL);
	if (ret)
		 fscache_uncache_page(ci->fscache, page);
}

void ceph_invalidate_fscache_page(struct inode* inode, struct page *page)
{
	struct ceph_inode_info *ci = ceph_inode(inode);

	if (!PageFsCache(page))
		return;

	fscache_wait_on_page_write(ci->fscache, page);
	fscache_uncache_page(ci->fscache, page);
}

void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc)
{
	if (fscache_cookie_valid(fsc->fscache)) {
		struct ceph_fscache_entry *ent;
		bool found = false;

		mutex_lock(&ceph_fscache_lock);
		list_for_each_entry(ent, &ceph_fscache_list, list) {
			if (ent->fscache == fsc->fscache) {
				list_del(&ent->list);
				kfree(ent);
				found = true;
				break;
			}
		}
		WARN_ON_ONCE(!found);
		mutex_unlock(&ceph_fscache_lock);

		__fscache_relinquish_cookie(fsc->fscache, 0);
	}
	fsc->fscache = NULL;
}

/*
 * caller should hold CEPH_CAP_FILE_{RD,CACHE}
 */
void ceph_fscache_revalidate_cookie(struct ceph_inode_info *ci)
{
	if (cache_valid(ci))
		return;

	/* resue i_truncate_mutex. There should be no pending
	 * truncate while the caller holds CEPH_CAP_FILE_RD */
	mutex_lock(&ci->i_truncate_mutex);
	if (!cache_valid(ci)) {
		if (fscache_check_consistency(ci->fscache))
			fscache_invalidate(ci->fscache);
		spin_lock(&ci->i_ceph_lock);
		ci->i_fscache_gen = ci->i_rdcache_gen;
		spin_unlock(&ci->i_ceph_lock);
	}
	mutex_unlock(&ci->i_truncate_mutex);
}
Commit	Line	Data
99ccbd22 MT	1	/*
	2	* Ceph cache definitions.
	3	*
	4	* Copyright (C) 2013 by Adfin Solutions, Inc. All Rights Reserved.
	5	* Written by Milosz Tanski (milosz@adfin.com)
	6	*
	7	* This program is free software; you can redistribute it and/or modify
	8	* it under the terms of the GNU General Public License version 2
	9	* as published by the Free Software Foundation.
	10	*
	11	* This program is distributed in the hope that it will be useful,
	12	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	13	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	14	* GNU General Public License for more details.
	15	*
	16	* You should have received a copy of the GNU General Public License
	17	* along with this program; if not, write to:
	18	* Free Software Foundation
	19	* 51 Franklin Street, Fifth Floor
	20	* Boston, MA 02111-1301 USA
	21	*
	22	*/
	23
99ccbd22 MT	24	#include "super.h"
	25	#include "cache.h"
	26
	27	struct ceph_aux_inode {
f6973c09	28	u64 version;
99ccbd22 MT	29	struct timespec mtime;
	30	loff_t size;
	31	};
	32
	33	struct fscache_netfs ceph_cache_netfs = {
	34	.name = "ceph",
	35	.version = 0,
	36	};
	37
1d8f8360 YZ	38	static DEFINE_MUTEX(ceph_fscache_lock);
	39	static LIST_HEAD(ceph_fscache_list);
	40
	41	struct ceph_fscache_entry {
	42	struct list_head list;
	43	struct fscache_cookie *fscache;
	44	struct ceph_fsid fsid;
	45	size_t uniq_len;
	46	char uniquifier[0];
	47	};
	48
99ccbd22 MT	49	static uint16_t ceph_fscache_session_get_key(const void *cookie_netfs_data,
	50	void *buffer, uint16_t maxbuf)
	51	{
	52	const struct ceph_fs_client* fsc = cookie_netfs_data;
1d8f8360 YZ	53	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
1d8f8360 YZ	54	uint16_t fsid_len, uniq_len;
99ccbd22	55
1d8f8360 YZ	56	fsid_len = sizeof(fsc->client->fsid);
	57	uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
	58	if (fsid_len + uniq_len > maxbuf)
99ccbd22 MT	59	return 0;
99ccbd22 MT	60
1d8f8360 YZ	61	memcpy(buffer, &fsc->client->fsid, fsid_len);
	62	if (uniq_len)
	63	memcpy(buffer + fsid_len, fscache_uniq, uniq_len);
	64
	65	return fsid_len + uniq_len;
99ccbd22 MT	66	}
	67
	68	static const struct fscache_cookie_def ceph_fscache_fsid_object_def = {
	69	.name = "CEPH.fsid",
	70	.type = FSCACHE_COOKIE_TYPE_INDEX,
	71	.get_key = ceph_fscache_session_get_key,
	72	};
	73
971f0bde	74	int ceph_fscache_register(void)
99ccbd22 MT	75	{
	76	return fscache_register_netfs(&ceph_cache_netfs);
	77	}
	78
971f0bde	79	void ceph_fscache_unregister(void)
99ccbd22 MT	80	{
	81	fscache_unregister_netfs(&ceph_cache_netfs);
	82	}
	83
	84	int ceph_fscache_register_fs(struct ceph_fs_client* fsc)
	85	{
1d8f8360 YZ	86	const struct ceph_fsid *fsid = &fsc->client->fsid;
	87	const char *fscache_uniq = fsc->mount_options->fscache_uniq;
	88	size_t uniq_len = fscache_uniq ? strlen(fscache_uniq) : 0;
	89	struct ceph_fscache_entry *ent;
	90	int err = 0;
	91
	92	mutex_lock(&ceph_fscache_lock);
	93	list_for_each_entry(ent, &ceph_fscache_list, list) {
	94	if (memcmp(&ent->fsid, fsid, sizeof(*fsid)))
	95	continue;
	96	if (ent->uniq_len != uniq_len)
	97	continue;
	98	if (uniq_len && memcmp(ent->uniquifier, fscache_uniq, uniq_len))
	99	continue;
	100
	101	pr_err("fscache cookie already registered for fsid %pU\n", fsid);
	102	pr_err(" use fsc=%%s mount option to specify a uniquifier\n");
	103	err = -EBUSY;
	104	goto out_unlock;
	105	}
	106
	107	ent = kzalloc(sizeof(*ent) + uniq_len, GFP_KERNEL);
	108	if (!ent) {
	109	err = -ENOMEM;
	110	goto out_unlock;
	111	}
	112
99ccbd22 MT	113	fsc->fscache = fscache_acquire_cookie(ceph_cache_netfs.primary_index,
99ccbd22 MT	114	&ceph_fscache_fsid_object_def,
94d30ae9	115	fsc, true);
99ccbd22	116
1d8f8360 YZ	117	if (fsc->fscache) {
	118	memcpy(&ent->fsid, fsid, sizeof(*fsid));
	119	if (uniq_len > 0) {
	120	memcpy(&ent->uniquifier, fscache_uniq, uniq_len);
	121	ent->uniq_len = uniq_len;
	122	}
	123	ent->fscache = fsc->fscache;
	124	list_add_tail(&ent->list, &ceph_fscache_list);
	125	} else {
	126	kfree(ent);
	127	pr_err("unable to register fscache cookie for fsid %pU\n",
	128	fsid);
	129	/* all other fs ignore this error */
	130	}
	131	out_unlock:
	132	mutex_unlock(&ceph_fscache_lock);
	133	return err;
99ccbd22 MT	134	}
	135
	136	static uint16_t ceph_fscache_inode_get_key(const void *cookie_netfs_data,
	137	void *buffer, uint16_t maxbuf)
	138	{
	139	const struct ceph_inode_info* ci = cookie_netfs_data;
	140	uint16_t klen;
	141
1291fb95	142	/* use ceph virtual inode (id + snapshot) */
99ccbd22 MT	143	klen = sizeof(ci->i_vino);
	144	if (klen > maxbuf)
	145	return 0;
	146
	147	memcpy(buffer, &ci->i_vino, klen);
	148	return klen;
	149	}
	150
	151	static uint16_t ceph_fscache_inode_get_aux(const void *cookie_netfs_data,
	152	void *buffer, uint16_t bufmax)
	153	{
	154	struct ceph_aux_inode aux;
	155	const struct ceph_inode_info* ci = cookie_netfs_data;
	156	const struct inode* inode = &ci->vfs_inode;
	157
	158	memset(&aux, 0, sizeof(aux));
f6973c09	159	aux.version = ci->i_version;
99ccbd22	160	aux.mtime = inode->i_mtime;
99c88e69	161	aux.size = i_size_read(inode);
99ccbd22 MT	162
	163	memcpy(buffer, &aux, sizeof(aux));
	164
	165	return sizeof(aux);
	166	}
	167
	168	static void ceph_fscache_inode_get_attr(const void *cookie_netfs_data,
	169	uint64_t *size)
	170	{
	171	const struct ceph_inode_info* ci = cookie_netfs_data;
99c88e69	172	*size = i_size_read(&ci->vfs_inode);
99ccbd22 MT	173	}
	174
	175	static enum fscache_checkaux ceph_fscache_inode_check_aux(
	176	void cookie_netfs_data, const void data, uint16_t dlen)
	177	{
	178	struct ceph_aux_inode aux;
	179	struct ceph_inode_info* ci = cookie_netfs_data;
	180	struct inode* inode = &ci->vfs_inode;
	181
	182	if (dlen != sizeof(aux))
	183	return FSCACHE_CHECKAUX_OBSOLETE;
	184
	185	memset(&aux, 0, sizeof(aux));
f6973c09	186	aux.version = ci->i_version;
99ccbd22	187	aux.mtime = inode->i_mtime;
99c88e69	188	aux.size = i_size_read(inode);
99ccbd22 MT	189
	190	if (memcmp(data, &aux, sizeof(aux)) != 0)
	191	return FSCACHE_CHECKAUX_OBSOLETE;
	192
	193	dout("ceph inode 0x%p cached okay", ci);
	194	return FSCACHE_CHECKAUX_OKAY;
	195	}
	196
	197	static void ceph_fscache_inode_now_uncached(void* cookie_netfs_data)
	198	{
	199	struct ceph_inode_info* ci = cookie_netfs_data;
	200	struct pagevec pvec;
	201	pgoff_t first;
	202	int loop, nr_pages;
	203
	204	pagevec_init(&pvec, 0);
	205	first = 0;
	206
	207	dout("ceph inode 0x%p now uncached", ci);
	208
	209	while (1) {
	210	nr_pages = pagevec_lookup(&pvec, ci->vfs_inode.i_mapping, first,
	211	PAGEVEC_SIZE - pagevec_count(&pvec));
	212
	213	if (!nr_pages)
	214	break;
	215
	216	for (loop = 0; loop < nr_pages; loop++)
	217	ClearPageFsCache(pvec.pages[loop]);
	218
	219	first = pvec.pages[nr_pages - 1]->index + 1;
	220
	221	pvec.nr = nr_pages;
	222	pagevec_release(&pvec);
	223	cond_resched();
	224	}
	225	}
	226
	227	static const struct fscache_cookie_def ceph_fscache_inode_object_def = {
	228	.name = "CEPH.inode",
	229	.type = FSCACHE_COOKIE_TYPE_DATAFILE,
	230	.get_key = ceph_fscache_inode_get_key,
	231	.get_attr = ceph_fscache_inode_get_attr,
	232	.get_aux = ceph_fscache_inode_get_aux,
	233	.check_aux = ceph_fscache_inode_check_aux,
	234	.now_uncached = ceph_fscache_inode_now_uncached,
	235	};
	236
46b59b2b	237	void ceph_fscache_register_inode_cookie(struct inode *inode)
99ccbd22	238	{
46b59b2b YZ	239	struct ceph_inode_info *ci = ceph_inode(inode);
46b59b2b YZ	240	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
99ccbd22 MT	241
	242	/* No caching for filesystem */
	243	if (fsc->fscache == NULL)
	244	return;
	245
	246	/* Only cache for regular files that are read only */
46b59b2b	247	if (!S_ISREG(inode->i_mode))
99ccbd22 MT	248	return;
99ccbd22 MT	249
46b59b2b YZ	250	inode_lock_nested(inode, I_MUTEX_CHILD);
	251	if (!ci->fscache) {
	252	ci->fscache = fscache_acquire_cookie(fsc->fscache,
	253	&ceph_fscache_inode_object_def,
	254	ci, false);
	255	}
5955102c	256	inode_unlock(inode);
99ccbd22 MT	257	}
	258
	259	void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci)
	260	{
	261	struct fscache_cookie* cookie;
	262
	263	if ((cookie = ci->fscache) == NULL)
	264	return;
	265
	266	ci->fscache = NULL;
	267
	268	fscache_uncache_all_inode_pages(cookie, &ci->vfs_inode);
	269	fscache_relinquish_cookie(cookie, 0);
	270	}
	271
46b59b2b YZ	272	static bool ceph_fscache_can_enable(void *data)
	273	{
	274	struct inode *inode = data;
	275	return !inode_is_open_for_write(inode);
	276	}
	277
	278	void ceph_fscache_file_set_cookie(struct inode inode, struct file filp)
	279	{
	280	struct ceph_inode_info *ci = ceph_inode(inode);
	281
	282	if (!fscache_cookie_valid(ci->fscache))
	283	return;
	284
	285	if (inode_is_open_for_write(inode)) {
	286	dout("fscache_file_set_cookie %p %p disabling cache\n",
	287	inode, filp);
	288	fscache_disable_cookie(ci->fscache, false);
	289	fscache_uncache_all_inode_pages(ci->fscache, inode);
	290	} else {
	291	fscache_enable_cookie(ci->fscache, ceph_fscache_can_enable,
	292	inode);
	293	if (fscache_cookie_enabled(ci->fscache)) {
0fbc5360	294	dout("fscache_file_set_cookie %p %p enabling cache\n",
46b59b2b YZ	295	inode, filp);
	296	}
	297	}
	298	}
	299
99ccbd22 MT	300	static void ceph_vfs_readpage_complete(struct page page, void data, int error)
	301	{
	302	if (!error)
	303	SetPageUptodate(page);
	304	}
	305
	306	static void ceph_vfs_readpage_complete_unlock(struct page page, void data, int error)
	307	{
	308	if (!error)
	309	SetPageUptodate(page);
	310
	311	unlock_page(page);
	312	}
	313
3b33f692	314	static inline bool cache_valid(struct ceph_inode_info *ci)
99ccbd22	315	{
f7f7e7a0	316	return ci->i_fscache_gen == ci->i_rdcache_gen;
99ccbd22 MT	317	}
	318
	319
	320	/* Atempt to read from the fscache,
	321	*
	322	* This function is called from the readpage_nounlock context. DO NOT attempt to
	323	* unlock the page here (or in the callback).
	324	*/
	325	int ceph_readpage_from_fscache(struct inode inode, struct page page)
	326	{
	327	struct ceph_inode_info *ci = ceph_inode(inode);
	328	int ret;
	329
	330	if (!cache_valid(ci))
	331	return -ENOBUFS;
	332
	333	ret = fscache_read_or_alloc_page(ci->fscache, page,
	334	ceph_vfs_readpage_complete, NULL,
	335	GFP_KERNEL);
	336
	337	switch (ret) {
	338	case 0: /* Page found */
	339	dout("page read submitted\n");
	340	return 0;
	341	case -ENOBUFS: /* Pages were not found, and can't be */
	342	case -ENODATA: /* Pages were not found */
	343	dout("page/inode not in cache\n");
	344	return ret;
	345	default:
	346	dout("%s: unknown error ret = %i\n", __func__, ret);
	347	return ret;
	348	}
	349	}
	350
	351	int ceph_readpages_from_fscache(struct inode *inode,
	352	struct address_space *mapping,
	353	struct list_head *pages,
	354	unsigned *nr_pages)
	355	{
	356	struct ceph_inode_info *ci = ceph_inode(inode);
	357	int ret;
	358
	359	if (!cache_valid(ci))
	360	return -ENOBUFS;
	361
	362	ret = fscache_read_or_alloc_pages(ci->fscache, mapping, pages, nr_pages,
	363	ceph_vfs_readpage_complete_unlock,
	364	NULL, mapping_gfp_mask(mapping));
	365
	366	switch (ret) {
	367	case 0: /* All pages found */
	368	dout("all-page read submitted\n");
	369	return 0;
	370	case -ENOBUFS: /* Some pages were not found, and can't be */
	371	case -ENODATA: /* some pages were not found */
	372	dout("page/inode not in cache\n");
	373	return ret;
	374	default:
	375	dout("%s: unknown error ret = %i\n", __func__, ret);
	376	return ret;
	377	}
	378	}
	379
	380	void ceph_readpage_to_fscache(struct inode inode, struct page page)
381	{
382	struct ceph_inode_info *ci = ceph_inode(inode);
383	int ret;
384
9b8dd1e8 MT	385	if (!PageFsCache(page))
	386	return;
	387
99ccbd22 MT	388	if (!cache_valid(ci))
	389	return;
	390
	391	ret = fscache_write_page(ci->fscache, page, GFP_KERNEL);
	392	if (ret)
	393	fscache_uncache_page(ci->fscache, page);
	394	}
	395
	396	void ceph_invalidate_fscache_page(struct inode* inode, struct page *page)
	397	{
	398	struct ceph_inode_info *ci = ceph_inode(inode);
	399
ffc79664 MT	400	if (!PageFsCache(page))
	401	return;
	402
99ccbd22 MT	403	fscache_wait_on_page_write(ci->fscache, page);
	404	fscache_uncache_page(ci->fscache, page);
	405	}
	406
	407	void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc)
	408	{
1d8f8360 YZ	409	if (fscache_cookie_valid(fsc->fscache)) {
	410	struct ceph_fscache_entry *ent;
	411	bool found = false;
	412
	413	mutex_lock(&ceph_fscache_lock);
	414	list_for_each_entry(ent, &ceph_fscache_list, list) {
	415	if (ent->fscache == fsc->fscache) {
	416	list_del(&ent->list);
	417	kfree(ent);
	418	found = true;
	419	break;
	420	}
	421	}
	422	WARN_ON_ONCE(!found);
	423	mutex_unlock(&ceph_fscache_lock);
	424
	425	__fscache_relinquish_cookie(fsc->fscache, 0);
	426	}
99ccbd22 MT	427	fsc->fscache = NULL;
	428	}
	429
f7f7e7a0 YZ	430	/*
	431	* caller should hold CEPH_CAP_FILE_{RD,CACHE}
	432	*/
	433	void ceph_fscache_revalidate_cookie(struct ceph_inode_info *ci)
99ccbd22	434	{
f7f7e7a0	435	if (cache_valid(ci))
e81568eb MT	436	return;
e81568eb MT	437
f7f7e7a0 YZ	438	/* resue i_truncate_mutex. There should be no pending
	439	* truncate while the caller holds CEPH_CAP_FILE_RD */
	440	mutex_lock(&ci->i_truncate_mutex);
	441	if (!cache_valid(ci)) {
	442	if (fscache_check_consistency(ci->fscache))
	443	fscache_invalidate(ci->fscache);
	444	spin_lock(&ci->i_ceph_lock);
	445	ci->i_fscache_gen = ci->i_rdcache_gen;
	446	spin_unlock(&ci->i_ceph_lock);
99ccbd22	447	}
f7f7e7a0	448	mutex_unlock(&ci->i_truncate_mutex);
99ccbd22	449	}