[linux.git] / fs / nfs / fscache.c

// SPDX-License-Identifier: GPL-2.0-or-later
/* NFS filesystem cache interface
 *
 * Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells ([email protected])
 */

#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_fs_sb.h>
#include <linux/in6.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/iversion.h>

#include "internal.h"
#include "iostat.h"
#include "fscache.h"

#define NFSDBG_FACILITY		NFSDBG_FSCACHE

static struct rb_root nfs_fscache_keys = RB_ROOT;
static DEFINE_SPINLOCK(nfs_fscache_keys_lock);

/*
 * Layout of the key for an NFS server cache object.
 */
struct nfs_server_key {
	struct {
		uint16_t	nfsversion;		/* NFS protocol version */
		uint32_t	minorversion;		/* NFSv4 minor version */
		uint16_t	family;			/* address family */
		__be16		port;			/* IP port */
	} hdr;
	union {
		struct in_addr	ipv4_addr;	/* IPv4 address */
		struct in6_addr ipv6_addr;	/* IPv6 address */
	};
} __packed;

/*
 * Get the per-client index cookie for an NFS client if the appropriate mount
 * flag was set
 * - We always try and get an index cookie for the client, but get filehandle
 *   cookies on a per-superblock basis, depending on the mount flags
 */
void nfs_fscache_get_client_cookie(struct nfs_client *clp)
{
	const struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) &clp->cl_addr;
	const struct sockaddr_in *sin = (struct sockaddr_in *) &clp->cl_addr;
	struct nfs_server_key key;
	uint16_t len = sizeof(key.hdr);

	memset(&key, 0, sizeof(key));
	key.hdr.nfsversion = clp->rpc_ops->version;
	key.hdr.minorversion = clp->cl_minorversion;
	key.hdr.family = clp->cl_addr.ss_family;

	switch (clp->cl_addr.ss_family) {
	case AF_INET:
		key.hdr.port = sin->sin_port;
		key.ipv4_addr = sin->sin_addr;
		len += sizeof(key.ipv4_addr);
		break;

	case AF_INET6:
		key.hdr.port = sin6->sin6_port;
		key.ipv6_addr = sin6->sin6_addr;
		len += sizeof(key.ipv6_addr);
		break;

	default:
		printk(KERN_WARNING "NFS: Unknown network family '%d'\n",
		       clp->cl_addr.ss_family);
		clp->fscache = NULL;
		return;
	}

	/* create a cache index for looking up filehandles */
	clp->fscache = fscache_acquire_cookie(nfs_fscache_netfs.primary_index,
					      &nfs_fscache_server_index_def,
					      &key, len,
					      NULL, 0,
					      clp, 0, true);
	dfprintk(FSCACHE, "NFS: get client cookie (0x%p/0x%p)\n",
		 clp, clp->fscache);
}

/*
 * Dispose of a per-client cookie
 */
void nfs_fscache_release_client_cookie(struct nfs_client *clp)
{
	dfprintk(FSCACHE, "NFS: releasing client cookie (0x%p/0x%p)\n",
		 clp, clp->fscache);

	fscache_relinquish_cookie(clp->fscache, NULL, false);
	clp->fscache = NULL;
}

/*
 * Get the cache cookie for an NFS superblock.  We have to handle
 * uniquification here because the cache doesn't do it for us.
 *
 * The default uniquifier is just an empty string, but it may be overridden
 * either by the 'fsc=xxx' option to mount, or by inheriting it from the parent
 * superblock across an automount point of some nature.
 */
void nfs_fscache_get_super_cookie(struct super_block *sb, const char *uniq, int ulen)
{
	struct nfs_fscache_key *key, *xkey;
	struct nfs_server *nfss = NFS_SB(sb);
	struct rb_node **p, *parent;
	int diff;

	nfss->fscache_key = NULL;
	nfss->fscache = NULL;
	if (!uniq) {
		uniq = "";
		ulen = 1;
	}

	key = kzalloc(sizeof(*key) + ulen, GFP_KERNEL);
	if (!key)
		return;

	key->nfs_client = nfss->nfs_client;
	key->key.super.s_flags = sb->s_flags & NFS_SB_MASK;
	key->key.nfs_server.flags = nfss->flags;
	key->key.nfs_server.rsize = nfss->rsize;
	key->key.nfs_server.wsize = nfss->wsize;
	key->key.nfs_server.acregmin = nfss->acregmin;
	key->key.nfs_server.acregmax = nfss->acregmax;
	key->key.nfs_server.acdirmin = nfss->acdirmin;
	key->key.nfs_server.acdirmax = nfss->acdirmax;
	key->key.nfs_server.fsid = nfss->fsid;
	key->key.rpc_auth.au_flavor = nfss->client->cl_auth->au_flavor;

	key->key.uniq_len = ulen;
	memcpy(key->key.uniquifier, uniq, ulen);

	spin_lock(&nfs_fscache_keys_lock);
	p = &nfs_fscache_keys.rb_node;
	parent = NULL;
	while (*p) {
		parent = *p;
		xkey = rb_entry(parent, struct nfs_fscache_key, node);

		if (key->nfs_client < xkey->nfs_client)
			goto go_left;
		if (key->nfs_client > xkey->nfs_client)
			goto go_right;

		diff = memcmp(&key->key, &xkey->key, sizeof(key->key));
		if (diff < 0)
			goto go_left;
		if (diff > 0)
			goto go_right;

		if (key->key.uniq_len == 0)
			goto non_unique;
		diff = memcmp(key->key.uniquifier,
			      xkey->key.uniquifier,
			      key->key.uniq_len);
		if (diff < 0)
			goto go_left;
		if (diff > 0)
			goto go_right;
		goto non_unique;

	go_left:
		p = &(*p)->rb_left;
		continue;
	go_right:
		p = &(*p)->rb_right;
	}

	rb_link_node(&key->node, parent, p);
	rb_insert_color(&key->node, &nfs_fscache_keys);
	spin_unlock(&nfs_fscache_keys_lock);
	nfss->fscache_key = key;

	/* create a cache index for looking up filehandles */
	nfss->fscache = fscache_acquire_cookie(nfss->nfs_client->fscache,
					       &nfs_fscache_super_index_def,
					       &key->key,
					       sizeof(key->key) + ulen,
					       NULL, 0,
					       nfss, 0, true);
	dfprintk(FSCACHE, "NFS: get superblock cookie (0x%p/0x%p)\n",
		 nfss, nfss->fscache);
	return;

non_unique:
	spin_unlock(&nfs_fscache_keys_lock);
	kfree(key);
	nfss->fscache_key = NULL;
	nfss->fscache = NULL;
	printk(KERN_WARNING "NFS:"
	       " Cache request denied due to non-unique superblock keys\n");
}

/*
 * release a per-superblock cookie
 */
void nfs_fscache_release_super_cookie(struct super_block *sb)
{
	struct nfs_server *nfss = NFS_SB(sb);

	dfprintk(FSCACHE, "NFS: releasing superblock cookie (0x%p/0x%p)\n",
		 nfss, nfss->fscache);

	fscache_relinquish_cookie(nfss->fscache, NULL, false);
	nfss->fscache = NULL;

	if (nfss->fscache_key) {
		spin_lock(&nfs_fscache_keys_lock);
		rb_erase(&nfss->fscache_key->node, &nfs_fscache_keys);
		spin_unlock(&nfs_fscache_keys_lock);
		kfree(nfss->fscache_key);
		nfss->fscache_key = NULL;
	}
}

static void nfs_fscache_update_auxdata(struct nfs_fscache_inode_auxdata *auxdata,
				  struct nfs_inode *nfsi)
{
	memset(auxdata, 0, sizeof(*auxdata));
	auxdata->mtime_sec  = nfsi->vfs_inode.i_mtime.tv_sec;
	auxdata->mtime_nsec = nfsi->vfs_inode.i_mtime.tv_nsec;
	auxdata->ctime_sec  = nfsi->vfs_inode.i_ctime.tv_sec;
	auxdata->ctime_nsec = nfsi->vfs_inode.i_ctime.tv_nsec;

	if (NFS_SERVER(&nfsi->vfs_inode)->nfs_client->rpc_ops->version == 4)
		auxdata->change_attr = inode_peek_iversion_raw(&nfsi->vfs_inode);
}

/*
 * Initialise the per-inode cache cookie pointer for an NFS inode.
 */
void nfs_fscache_init_inode(struct inode *inode)
{
	struct nfs_fscache_inode_auxdata auxdata;
	struct nfs_server *nfss = NFS_SERVER(inode);
	struct nfs_inode *nfsi = NFS_I(inode);

	nfsi->fscache = NULL;
	if (!(nfss->fscache && S_ISREG(inode->i_mode)))
		return;

	nfs_fscache_update_auxdata(&auxdata, nfsi);

	nfsi->fscache = fscache_acquire_cookie(NFS_SB(inode->i_sb)->fscache,
					       &nfs_fscache_inode_object_def,
					       nfsi->fh.data, nfsi->fh.size,
					       &auxdata, sizeof(auxdata),
					       nfsi, nfsi->vfs_inode.i_size, false);
}

/*
 * Release a per-inode cookie.
 */
void nfs_fscache_clear_inode(struct inode *inode)
{
	struct nfs_fscache_inode_auxdata auxdata;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct fscache_cookie *cookie = nfs_i_fscache(inode);

	dfprintk(FSCACHE, "NFS: clear cookie (0x%p/0x%p)\n", nfsi, cookie);

	nfs_fscache_update_auxdata(&auxdata, nfsi);
	fscache_relinquish_cookie(cookie, &auxdata, false);
	nfsi->fscache = NULL;
}

static bool nfs_fscache_can_enable(void *data)
{
	struct inode *inode = data;

	return !inode_is_open_for_write(inode);
}

/*
 * Enable or disable caching for a file that is being opened as appropriate.
 * The cookie is allocated when the inode is initialised, but is not enabled at
 * that time.  Enablement is deferred to file-open time to avoid stat() and
 * access() thrashing the cache.
 *
 * For now, with NFS, only regular files that are open read-only will be able
 * to use the cache.
 *
 * We enable the cache for an inode if we open it read-only and it isn't
 * currently open for writing.  We disable the cache if the inode is open
 * write-only.
 *
 * The caller uses the file struct to pin i_writecount on the inode before
 * calling us when a file is opened for writing, so we can make use of that.
 *
 * Note that this may be invoked multiple times in parallel by parallel
 * nfs_open() functions.
 */
void nfs_fscache_open_file(struct inode *inode, struct file *filp)
{
	struct nfs_fscache_inode_auxdata auxdata;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct fscache_cookie *cookie = nfs_i_fscache(inode);

	if (!fscache_cookie_valid(cookie))
		return;

	nfs_fscache_update_auxdata(&auxdata, nfsi);

	if (inode_is_open_for_write(inode)) {
		dfprintk(FSCACHE, "NFS: nfsi 0x%p disabling cache\n", nfsi);
		clear_bit(NFS_INO_FSCACHE, &nfsi->flags);
		fscache_disable_cookie(cookie, &auxdata, true);
		fscache_uncache_all_inode_pages(cookie, inode);
	} else {
		dfprintk(FSCACHE, "NFS: nfsi 0x%p enabling cache\n", nfsi);
		fscache_enable_cookie(cookie, &auxdata, nfsi->vfs_inode.i_size,
				      nfs_fscache_can_enable, inode);
		if (fscache_cookie_enabled(cookie))
			set_bit(NFS_INO_FSCACHE, &NFS_I(inode)->flags);
	}
}
EXPORT_SYMBOL_GPL(nfs_fscache_open_file);

/*
 * Release the caching state associated with a page, if the page isn't busy
 * interacting with the cache.
 * - Returns true (can release page) or false (page busy).
 */
int nfs_fscache_release_page(struct page *page, gfp_t gfp)
{
	if (PageFsCache(page)) {
		struct fscache_cookie *cookie = nfs_i_fscache(page->mapping->host);

		BUG_ON(!cookie);
		dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
			 cookie, page, NFS_I(page->mapping->host));

		if (!fscache_maybe_release_page(cookie, page, gfp))
			return 0;

		nfs_inc_fscache_stats(page->mapping->host,
				      NFSIOS_FSCACHE_PAGES_UNCACHED);
	}

	return 1;
}

/*
 * Release the caching state associated with a page if undergoing complete page
 * invalidation.
 */
void __nfs_fscache_invalidate_page(struct page *page, struct inode *inode)
{
	struct fscache_cookie *cookie = nfs_i_fscache(inode);

	BUG_ON(!cookie);

	dfprintk(FSCACHE, "NFS: fscache invalidatepage (0x%p/0x%p/0x%p)\n",
		 cookie, page, NFS_I(inode));

	fscache_wait_on_page_write(cookie, page);

	BUG_ON(!PageLocked(page));
	fscache_uncache_page(cookie, page);
	nfs_inc_fscache_stats(page->mapping->host,
			      NFSIOS_FSCACHE_PAGES_UNCACHED);
}

/*
 * Handle completion of a page being read from the cache.
 * - Called in process (keventd) context.
 */
static void nfs_readpage_from_fscache_complete(struct page *page,
					       void *context,
					       int error)
{
	dfprintk(FSCACHE,
		 "NFS: readpage_from_fscache_complete (0x%p/0x%p/%d)\n",
		 page, context, error);

	/* if the read completes with an error, we just unlock the page and let
	 * the VM reissue the readpage */
	if (!error) {
		SetPageUptodate(page);
		unlock_page(page);
	}
}

/*
 * Retrieve a page from fscache
 */
int __nfs_readpage_from_fscache(struct nfs_open_context *ctx,
				struct inode *inode, struct page *page)
{
	int ret;

	dfprintk(FSCACHE,
		 "NFS: readpage_from_fscache(fsc:%p/p:%p(i:%lx f:%lx)/0x%p)\n",
		 nfs_i_fscache(inode), page, page->index, page->flags, inode);

	ret = fscache_read_or_alloc_page(nfs_i_fscache(inode),
					 page,
					 nfs_readpage_from_fscache_complete,
					 ctx,
					 GFP_KERNEL);

	switch (ret) {
	case 0: /* read BIO submitted (page in fscache) */
		dfprintk(FSCACHE,
			 "NFS:    readpage_from_fscache: BIO submitted\n");
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK);
		return ret;

	case -ENOBUFS: /* inode not in cache */
	case -ENODATA: /* page not in cache */
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
		dfprintk(FSCACHE,
			 "NFS:    readpage_from_fscache %d\n", ret);
		return 1;

	default:
		dfprintk(FSCACHE, "NFS:    readpage_from_fscache %d\n", ret);
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
	}
	return ret;
}

/*
 * Retrieve a set of pages from fscache
 */
int __nfs_readpages_from_fscache(struct nfs_open_context *ctx,
				 struct inode *inode,
				 struct address_space *mapping,
				 struct list_head *pages,
				 unsigned *nr_pages)
{
	unsigned npages = *nr_pages;
	int ret;

	dfprintk(FSCACHE, "NFS: nfs_getpages_from_fscache (0x%p/%u/0x%p)\n",
		 nfs_i_fscache(inode), npages, inode);

	ret = fscache_read_or_alloc_pages(nfs_i_fscache(inode),
					  mapping, pages, nr_pages,
					  nfs_readpage_from_fscache_complete,
					  ctx,
					  mapping_gfp_mask(mapping));
	if (*nr_pages < npages)
		nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK,
				      npages);
	if (*nr_pages > 0)
		nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL,
				      *nr_pages);

	switch (ret) {
	case 0: /* read submitted to the cache for all pages */
		BUG_ON(!list_empty(pages));
		BUG_ON(*nr_pages != 0);
		dfprintk(FSCACHE,
			 "NFS: nfs_getpages_from_fscache: submitted\n");

		return ret;

	case -ENOBUFS: /* some pages aren't cached and can't be */
	case -ENODATA: /* some pages aren't cached */
		dfprintk(FSCACHE,
			 "NFS: nfs_getpages_from_fscache: no page: %d\n", ret);
		return 1;

	default:
		dfprintk(FSCACHE,
			 "NFS: nfs_getpages_from_fscache: ret  %d\n", ret);
	}

	return ret;
}

/*
 * Store a newly fetched page in fscache
 * - PG_fscache must be set on the page
 */
void __nfs_readpage_to_fscache(struct inode *inode, struct page *page, int sync)
{
	int ret;

	dfprintk(FSCACHE,
		 "NFS: readpage_to_fscache(fsc:%p/p:%p(i:%lx f:%lx)/%d)\n",
		 nfs_i_fscache(inode), page, page->index, page->flags, sync);

	ret = fscache_write_page(nfs_i_fscache(inode), page,
				 inode->i_size, GFP_KERNEL);
	dfprintk(FSCACHE,
		 "NFS:     readpage_to_fscache: p:%p(i:%lu f:%lx) ret %d\n",
		 page, page->index, page->flags, ret);

	if (ret != 0) {
		fscache_uncache_page(nfs_i_fscache(inode), page);
		nfs_inc_fscache_stats(inode,
				      NFSIOS_FSCACHE_PAGES_WRITTEN_FAIL);
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_UNCACHED);
	} else {
		nfs_inc_fscache_stats(inode,
				      NFSIOS_FSCACHE_PAGES_WRITTEN_OK);
	}
}
Commit	Line	Data
b4d0d230	1	// SPDX-License-Identifier: GPL-2.0-or-later
14727281 DH	2	/* NFS filesystem cache interface
	3	*
	4	* Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
	5	* Written by David Howells ([email protected])
14727281 DH	6	*/
	7
	8	#include <linux/init.h>
	9	#include <linux/kernel.h>
	10	#include <linux/sched.h>
	11	#include <linux/mm.h>
	12	#include <linux/nfs_fs.h>
	13	#include <linux/nfs_fs_sb.h>
	14	#include <linux/in6.h>
	15	#include <linux/seq_file.h>
5a0e3ad6	16	#include <linux/slab.h>
402cb8dd	17	#include <linux/iversion.h>
14727281 DH	18
14727281 DH	19	#include "internal.h"
545db45f	20	#include "iostat.h"
14727281 DH	21	#include "fscache.h"
	22
	23	#define NFSDBG_FACILITY NFSDBG_FSCACHE
	24
08734048 DH	25	static struct rb_root nfs_fscache_keys = RB_ROOT;
	26	static DEFINE_SPINLOCK(nfs_fscache_keys_lock);
	27
402cb8dd DH	28	/*
	29	* Layout of the key for an NFS server cache object.
	30	*/
	31	struct nfs_server_key {
	32	struct {
	33	uint16_t nfsversion; /* NFS protocol version */
55dee1bc	34	uint32_t minorversion; /* NFSv4 minor version */
402cb8dd DH	35	uint16_t family; /* address family */
	36	__be16 port; /* IP port */
	37	} hdr;
	38	union {
	39	struct in_addr ipv4_addr; /* IPv4 address */
	40	struct in6_addr ipv6_addr; /* IPv6 address */
	41	};
	42	} __packed;
	43
14727281 DH	44	/*
	45	* Get the per-client index cookie for an NFS client if the appropriate mount
	46	* flag was set
	47	* - We always try and get an index cookie for the client, but get filehandle
	48	* cookies on a per-superblock basis, depending on the mount flags
	49	*/
	50	void nfs_fscache_get_client_cookie(struct nfs_client *clp)
	51	{
402cb8dd DH	52	const struct sockaddr_in6 sin6 = (struct sockaddr_in6 ) &clp->cl_addr;
	53	const struct sockaddr_in sin = (struct sockaddr_in ) &clp->cl_addr;
	54	struct nfs_server_key key;
	55	uint16_t len = sizeof(key.hdr);
	56
	57	memset(&key, 0, sizeof(key));
	58	key.hdr.nfsversion = clp->rpc_ops->version;
55dee1bc	59	key.hdr.minorversion = clp->cl_minorversion;
402cb8dd DH	60	key.hdr.family = clp->cl_addr.ss_family;
	61
	62	switch (clp->cl_addr.ss_family) {
	63	case AF_INET:
	64	key.hdr.port = sin->sin_port;
	65	key.ipv4_addr = sin->sin_addr;
	66	len += sizeof(key.ipv4_addr);
	67	break;
	68
	69	case AF_INET6:
	70	key.hdr.port = sin6->sin6_port;
	71	key.ipv6_addr = sin6->sin6_addr;
	72	len += sizeof(key.ipv6_addr);
	73	break;
	74
	75	default:
	76	printk(KERN_WARNING "NFS: Unknown network family '%d'\n",
	77	clp->cl_addr.ss_family);
	78	clp->fscache = NULL;
	79	return;
	80	}
	81
14727281 DH	82	/* create a cache index for looking up filehandles */
	83	clp->fscache = fscache_acquire_cookie(nfs_fscache_netfs.primary_index,
	84	&nfs_fscache_server_index_def,
402cb8dd DH	85	&key, len,
402cb8dd DH	86	NULL, 0,
ee1235a9	87	clp, 0, true);
14727281 DH	88	dfprintk(FSCACHE, "NFS: get client cookie (0x%p/0x%p)\n",
	89	clp, clp->fscache);
	90	}
	91
	92	/*
	93	* Dispose of a per-client cookie
	94	*/
	95	void nfs_fscache_release_client_cookie(struct nfs_client *clp)
	96	{
	97	dfprintk(FSCACHE, "NFS: releasing client cookie (0x%p/0x%p)\n",
	98	clp, clp->fscache);
	99
402cb8dd	100	fscache_relinquish_cookie(clp->fscache, NULL, false);
14727281 DH	101	clp->fscache = NULL;
14727281 DH	102	}
08734048 DH	103
	104	/*
	105	* Get the cache cookie for an NFS superblock. We have to handle
	106	* uniquification here because the cache doesn't do it for us.
2df54806 DH	107	*
	108	* The default uniquifier is just an empty string, but it may be overridden
	109	* either by the 'fsc=xxx' option to mount, or by inheriting it from the parent
	110	* superblock across an automount point of some nature.
08734048	111	*/
2311b943	112	void nfs_fscache_get_super_cookie(struct super_block sb, const char uniq, int ulen)
08734048 DH	113	{
	114	struct nfs_fscache_key key, xkey;
	115	struct nfs_server *nfss = NFS_SB(sb);
	116	struct rb_node *p, parent;
2311b943	117	int diff;
2df54806	118
dea1bb35 TM	119	nfss->fscache_key = NULL;
dea1bb35 TM	120	nfss->fscache = NULL;
2df54806 DH	121	if (!uniq) {
	122	uniq = "";
	123	ulen = 1;
	124	}
	125
08734048 DH	126	key = kzalloc(sizeof(*key) + ulen, GFP_KERNEL);
	127	if (!key)
	128	return;
	129
	130	key->nfs_client = nfss->nfs_client;
62a55d08	131	key->key.super.s_flags = sb->s_flags & NFS_SB_MASK;
08734048 DH	132	key->key.nfs_server.flags = nfss->flags;
	133	key->key.nfs_server.rsize = nfss->rsize;
	134	key->key.nfs_server.wsize = nfss->wsize;
	135	key->key.nfs_server.acregmin = nfss->acregmin;
	136	key->key.nfs_server.acregmax = nfss->acregmax;
	137	key->key.nfs_server.acdirmin = nfss->acdirmin;
	138	key->key.nfs_server.acdirmax = nfss->acdirmax;
	139	key->key.nfs_server.fsid = nfss->fsid;
	140	key->key.rpc_auth.au_flavor = nfss->client->cl_auth->au_flavor;
	141
	142	key->key.uniq_len = ulen;
	143	memcpy(key->key.uniquifier, uniq, ulen);
	144
	145	spin_lock(&nfs_fscache_keys_lock);
	146	p = &nfs_fscache_keys.rb_node;
	147	parent = NULL;
	148	while (*p) {
	149	parent = *p;
	150	xkey = rb_entry(parent, struct nfs_fscache_key, node);
	151
	152	if (key->nfs_client < xkey->nfs_client)
	153	goto go_left;
	154	if (key->nfs_client > xkey->nfs_client)
	155	goto go_right;
	156
	157	diff = memcmp(&key->key, &xkey->key, sizeof(key->key));
	158	if (diff < 0)
	159	goto go_left;
	160	if (diff > 0)
	161	goto go_right;
	162
	163	if (key->key.uniq_len == 0)
	164	goto non_unique;
	165	diff = memcmp(key->key.uniquifier,
	166	xkey->key.uniquifier,
	167	key->key.uniq_len);
	168	if (diff < 0)
	169	goto go_left;
	170	if (diff > 0)
	171	goto go_right;
	172	goto non_unique;
	173
	174	go_left:
	175	p = &(*p)->rb_left;
	176	continue;
	177	go_right:
	178	p = &(*p)->rb_right;
	179	}
	180
	181	rb_link_node(&key->node, parent, p);
	182	rb_insert_color(&key->node, &nfs_fscache_keys);
	183	spin_unlock(&nfs_fscache_keys_lock);
	184	nfss->fscache_key = key;
	185
	186	/* create a cache index for looking up filehandles */
	187	nfss->fscache = fscache_acquire_cookie(nfss->nfs_client->fscache,
	188	&nfs_fscache_super_index_def,
d9bfced1 DW	189	&key->key,
d9bfced1 DW	190	sizeof(key->key) + ulen,
402cb8dd	191	NULL, 0,
ee1235a9	192	nfss, 0, true);
08734048 DH	193	dfprintk(FSCACHE, "NFS: get superblock cookie (0x%p/0x%p)\n",
	194	nfss, nfss->fscache);
	195	return;
	196
	197	non_unique:
	198	spin_unlock(&nfs_fscache_keys_lock);
	199	kfree(key);
	200	nfss->fscache_key = NULL;
	201	nfss->fscache = NULL;
	202	printk(KERN_WARNING "NFS:"
	203	" Cache request denied due to non-unique superblock keys\n");
	204	}
	205
	206	/*
	207	* release a per-superblock cookie
	208	*/
	209	void nfs_fscache_release_super_cookie(struct super_block *sb)
	210	{
	211	struct nfs_server *nfss = NFS_SB(sb);
	212
	213	dfprintk(FSCACHE, "NFS: releasing superblock cookie (0x%p/0x%p)\n",
	214	nfss, nfss->fscache);
	215
402cb8dd	216	fscache_relinquish_cookie(nfss->fscache, NULL, false);
08734048 DH	217	nfss->fscache = NULL;
	218
	219	if (nfss->fscache_key) {
	220	spin_lock(&nfs_fscache_keys_lock);
	221	rb_erase(&nfss->fscache_key->node, &nfs_fscache_keys);
	222	spin_unlock(&nfs_fscache_keys_lock);
	223	kfree(nfss->fscache_key);
	224	nfss->fscache_key = NULL;
	225	}
	226	}
ef79c097	227
50eaa652 DW	228	static void nfs_fscache_update_auxdata(struct nfs_fscache_inode_auxdata *auxdata,
	229	struct nfs_inode *nfsi)
	230	{
	231	memset(auxdata, 0, sizeof(*auxdata));
	232	auxdata->mtime_sec = nfsi->vfs_inode.i_mtime.tv_sec;
	233	auxdata->mtime_nsec = nfsi->vfs_inode.i_mtime.tv_nsec;
	234	auxdata->ctime_sec = nfsi->vfs_inode.i_ctime.tv_sec;
	235	auxdata->ctime_nsec = nfsi->vfs_inode.i_ctime.tv_nsec;
	236
	237	if (NFS_SERVER(&nfsi->vfs_inode)->nfs_client->rpc_ops->version == 4)
	238	auxdata->change_attr = inode_peek_iversion_raw(&nfsi->vfs_inode);
	239	}
	240
ef79c097 DH	241	/*
	242	* Initialise the per-inode cache cookie pointer for an NFS inode.
	243	*/
f1fe29b4	244	void nfs_fscache_init_inode(struct inode *inode)
ef79c097	245	{
402cb8dd	246	struct nfs_fscache_inode_auxdata auxdata;
dea1bb35	247	struct nfs_server *nfss = NFS_SERVER(inode);
ef79c097 DH	248	struct nfs_inode *nfsi = NFS_I(inode);
ef79c097 DH	249
f1fe29b4	250	nfsi->fscache = NULL;
dea1bb35	251	if (!(nfss->fscache && S_ISREG(inode->i_mode)))
ef79c097	252	return;
402cb8dd	253
50eaa652	254	nfs_fscache_update_auxdata(&auxdata, nfsi);
402cb8dd	255
f1fe29b4 DH	256	nfsi->fscache = fscache_acquire_cookie(NFS_SB(inode->i_sb)->fscache,
f1fe29b4 DH	257	&nfs_fscache_inode_object_def,
402cb8dd DH	258	nfsi->fh.data, nfsi->fh.size,
402cb8dd DH	259	&auxdata, sizeof(auxdata),
ee1235a9	260	nfsi, nfsi->vfs_inode.i_size, false);
ef79c097 DH	261	}
	262
	263	/*
	264	* Release a per-inode cookie.
	265	*/
f1fe29b4	266	void nfs_fscache_clear_inode(struct inode *inode)
ef79c097	267	{
402cb8dd	268	struct nfs_fscache_inode_auxdata auxdata;
ef79c097	269	struct nfs_inode *nfsi = NFS_I(inode);
f1fe29b4	270	struct fscache_cookie *cookie = nfs_i_fscache(inode);
ef79c097	271
f1fe29b4	272	dfprintk(FSCACHE, "NFS: clear cookie (0x%p/0x%p)\n", nfsi, cookie);
ef79c097	273
50eaa652	274	nfs_fscache_update_auxdata(&auxdata, nfsi);
402cb8dd	275	fscache_relinquish_cookie(cookie, &auxdata, false);
ef79c097 DH	276	nfsi->fscache = NULL;
	277	}
	278
f1fe29b4	279	static bool nfs_fscache_can_enable(void *data)
ef79c097	280	{
f1fe29b4	281	struct inode *inode = data;
ef79c097	282
f1fe29b4	283	return !inode_is_open_for_write(inode);
ef79c097 DH	284	}
	285
	286	/*
f1fe29b4 DH	287	* Enable or disable caching for a file that is being opened as appropriate.
	288	* The cookie is allocated when the inode is initialised, but is not enabled at
	289	* that time. Enablement is deferred to file-open time to avoid stat() and
	290	* access() thrashing the cache.
	291	*
	292	* For now, with NFS, only regular files that are open read-only will be able
	293	* to use the cache.
	294	*
	295	* We enable the cache for an inode if we open it read-only and it isn't
	296	* currently open for writing. We disable the cache if the inode is open
	297	* write-only.
	298	*
	299	* The caller uses the file struct to pin i_writecount on the inode before
	300	* calling us when a file is opened for writing, so we can make use of that.
	301	*
	302	* Note that this may be invoked multiple times in parallel by parallel
	303	* nfs_open() functions.
ef79c097	304	*/
f1fe29b4	305	void nfs_fscache_open_file(struct inode inode, struct file filp)
ef79c097	306	{
402cb8dd	307	struct nfs_fscache_inode_auxdata auxdata;
ef79c097	308	struct nfs_inode *nfsi = NFS_I(inode);
f1fe29b4	309	struct fscache_cookie *cookie = nfs_i_fscache(inode);
ef79c097	310
f1fe29b4 DH	311	if (!fscache_cookie_valid(cookie))
f1fe29b4 DH	312	return;
ef79c097	313
50eaa652	314	nfs_fscache_update_auxdata(&auxdata, nfsi);
402cb8dd	315
f1fe29b4 DH	316	if (inode_is_open_for_write(inode)) {
	317	dfprintk(FSCACHE, "NFS: nfsi 0x%p disabling cache\n", nfsi);
	318	clear_bit(NFS_INO_FSCACHE, &nfsi->flags);
402cb8dd	319	fscache_disable_cookie(cookie, &auxdata, true);
f1fe29b4 DH	320	fscache_uncache_all_inode_pages(cookie, inode);
	321	} else {
	322	dfprintk(FSCACHE, "NFS: nfsi 0x%p enabling cache\n", nfsi);
ee1235a9	323	fscache_enable_cookie(cookie, &auxdata, nfsi->vfs_inode.i_size,
402cb8dd	324	nfs_fscache_can_enable, inode);
f1fe29b4 DH	325	if (fscache_cookie_enabled(cookie))
f1fe29b4 DH	326	set_bit(NFS_INO_FSCACHE, &NFS_I(inode)->flags);
ef79c097	327	}
ef79c097	328	}
f1fe29b4	329	EXPORT_SYMBOL_GPL(nfs_fscache_open_file);
545db45f DH	330
	331	/*
	332	* Release the caching state associated with a page, if the page isn't busy
	333	* interacting with the cache.
	334	* - Returns true (can release page) or false (page busy).
	335	*/
	336	int nfs_fscache_release_page(struct page *page, gfp_t gfp)
	337	{
545db45f	338	if (PageFsCache(page)) {
f1fe29b4	339	struct fscache_cookie *cookie = nfs_i_fscache(page->mapping->host);
2c174009 TM	340
2c174009 TM	341	BUG_ON(!cookie);
545db45f	342	dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
f1fe29b4	343	cookie, page, NFS_I(page->mapping->host));
545db45f	344
201a1542 DH	345	if (!fscache_maybe_release_page(cookie, page, gfp))
	346	return 0;
	347
e9f456ca LR	348	nfs_inc_fscache_stats(page->mapping->host,
e9f456ca LR	349	NFSIOS_FSCACHE_PAGES_UNCACHED);
545db45f DH	350	}
	351
	352	return 1;
	353	}
	354
	355	/*
	356	* Release the caching state associated with a page if undergoing complete page
	357	* invalidation.
	358	*/
	359	void __nfs_fscache_invalidate_page(struct page page, struct inode inode)
	360	{
f1fe29b4	361	struct fscache_cookie *cookie = nfs_i_fscache(inode);
545db45f DH	362
	363	BUG_ON(!cookie);
	364
	365	dfprintk(FSCACHE, "NFS: fscache invalidatepage (0x%p/0x%p/0x%p)\n",
f1fe29b4	366	cookie, page, NFS_I(inode));
545db45f DH	367
	368	fscache_wait_on_page_write(cookie, page);
	369
	370	BUG_ON(!PageLocked(page));
	371	fscache_uncache_page(cookie, page);
e9f456ca LR	372	nfs_inc_fscache_stats(page->mapping->host,
e9f456ca LR	373	NFSIOS_FSCACHE_PAGES_UNCACHED);
545db45f	374	}
9a9fc1c0 DH	375
	376	/*
	377	* Handle completion of a page being read from the cache.
	378	* - Called in process (keventd) context.
	379	*/
	380	static void nfs_readpage_from_fscache_complete(struct page *page,
	381	void *context,
	382	int error)
	383	{
	384	dfprintk(FSCACHE,
	385	"NFS: readpage_from_fscache_complete (0x%p/0x%p/%d)\n",
	386	page, context, error);
	387
	388	/* if the read completes with an error, we just unlock the page and let
	389	* the VM reissue the readpage */
	390	if (!error) {
	391	SetPageUptodate(page);
	392	unlock_page(page);
9a9fc1c0 DH	393	}
	394	}
	395
	396	/*
	397	* Retrieve a page from fscache
	398	*/
	399	int __nfs_readpage_from_fscache(struct nfs_open_context *ctx,
	400	struct inode inode, struct page page)
	401	{
	402	int ret;
	403
	404	dfprintk(FSCACHE,
	405	"NFS: readpage_from_fscache(fsc:%p/p:%p(i:%lx f:%lx)/0x%p)\n",
f1fe29b4	406	nfs_i_fscache(inode), page, page->index, page->flags, inode);
9a9fc1c0	407
f1fe29b4	408	ret = fscache_read_or_alloc_page(nfs_i_fscache(inode),
9a9fc1c0 DH	409	page,
	410	nfs_readpage_from_fscache_complete,
	411	ctx,
	412	GFP_KERNEL);
	413
	414	switch (ret) {
	415	case 0: /* read BIO submitted (page in fscache) */
	416	dfprintk(FSCACHE,
	417	"NFS: readpage_from_fscache: BIO submitted\n");
e9f456ca	418	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK);
9a9fc1c0 DH	419	return ret;
	420
	421	case -ENOBUFS: /* inode not in cache */
	422	case -ENODATA: /* page not in cache */
e9f456ca	423	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
9a9fc1c0 DH	424	dfprintk(FSCACHE,
	425	"NFS: readpage_from_fscache %d\n", ret);
	426	return 1;
	427
	428	default:
	429	dfprintk(FSCACHE, "NFS: readpage_from_fscache %d\n", ret);
e9f456ca	430	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
9a9fc1c0 DH	431	}
	432	return ret;
	433	}
	434
	435	/*
	436	* Retrieve a set of pages from fscache
	437	*/
	438	int __nfs_readpages_from_fscache(struct nfs_open_context *ctx,
	439	struct inode *inode,
	440	struct address_space *mapping,
	441	struct list_head *pages,
	442	unsigned *nr_pages)
	443	{
0f15c53d CL	444	unsigned npages = *nr_pages;
0f15c53d CL	445	int ret;
9a9fc1c0 DH	446
9a9fc1c0 DH	447	dfprintk(FSCACHE, "NFS: nfs_getpages_from_fscache (0x%p/%u/0x%p)\n",
f1fe29b4	448	nfs_i_fscache(inode), npages, inode);
9a9fc1c0	449
f1fe29b4	450	ret = fscache_read_or_alloc_pages(nfs_i_fscache(inode),
9a9fc1c0 DH	451	mapping, pages, nr_pages,
	452	nfs_readpage_from_fscache_complete,
	453	ctx,
	454	mapping_gfp_mask(mapping));
	455	if (*nr_pages < npages)
	456	nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK,
	457	npages);
	458	if (*nr_pages > 0)
	459	nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL,
	460	*nr_pages);
	461
	462	switch (ret) {
	463	case 0: /* read submitted to the cache for all pages */
	464	BUG_ON(!list_empty(pages));
	465	BUG_ON(*nr_pages != 0);
	466	dfprintk(FSCACHE,
	467	"NFS: nfs_getpages_from_fscache: submitted\n");
	468
	469	return ret;
	470
	471	case -ENOBUFS: /* some pages aren't cached and can't be */
	472	case -ENODATA: /* some pages aren't cached */
	473	dfprintk(FSCACHE,
	474	"NFS: nfs_getpages_from_fscache: no page: %d\n", ret);
	475	return 1;
	476
	477	default:
	478	dfprintk(FSCACHE,
	479	"NFS: nfs_getpages_from_fscache: ret %d\n", ret);
	480	}
	481
	482	return ret;
	483	}
7f8e05f6 DH	484
	485	/*
	486	* Store a newly fetched page in fscache
	487	* - PG_fscache must be set on the page
	488	*/
	489	void __nfs_readpage_to_fscache(struct inode inode, struct page page, int sync)
	490	{
	491	int ret;
	492
	493	dfprintk(FSCACHE,
	494	"NFS: readpage_to_fscache(fsc:%p/p:%p(i:%lx f:%lx)/%d)\n",
f1fe29b4	495	nfs_i_fscache(inode), page, page->index, page->flags, sync);
7f8e05f6	496
ee1235a9 DH	497	ret = fscache_write_page(nfs_i_fscache(inode), page,
ee1235a9 DH	498	inode->i_size, GFP_KERNEL);
7f8e05f6 DH	499	dfprintk(FSCACHE,
	500	"NFS: readpage_to_fscache: p:%p(i:%lu f:%lx) ret %d\n",
	501	page, page->index, page->flags, ret);
	502
	503	if (ret != 0) {
f1fe29b4	504	fscache_uncache_page(nfs_i_fscache(inode), page);
e9f456ca LR	505	nfs_inc_fscache_stats(inode,
	506	NFSIOS_FSCACHE_PAGES_WRITTEN_FAIL);
	507	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_UNCACHED);
7f8e05f6	508	} else {
e9f456ca LR	509	nfs_inc_fscache_stats(inode,
e9f456ca LR	510	NFSIOS_FSCACHE_PAGES_WRITTEN_OK);
7f8e05f6 DH	511	}
7f8e05f6 DH	512	}