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

/* NFS filesystem cache interface
 *
 * Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells ([email protected])
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public Licence
 * as published by the Free Software Foundation; either version
 * 2 of the Licence, or (at your option) any later version.
 */

#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 */
		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.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;

	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_MS_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, sizeof(*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;
	}
}

/*
 * 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_inode *nfsi = NFS_I(inode);

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

	memset(&auxdata, 0, sizeof(auxdata));
	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);

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

	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);

	memset(&auxdata, 0, sizeof(auxdata));
	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
	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;

	memset(&auxdata, 0, sizeof(auxdata));
	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);

	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);
	} else {
		error = nfs_readpage_async(context, page->mapping->host, page);
		if (error)
			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
14727281 DH	1	/* NFS filesystem cache interface
	2	*
	3	* Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
	4	* Written by David Howells ([email protected])
	5	*
	6	* This program is free software; you can redistribute it and/or
	7	* modify it under the terms of the GNU General Public Licence
	8	* as published by the Free Software Foundation; either version
	9	* 2 of the Licence, or (at your option) any later version.
	10	*/
	11
	12	#include <linux/init.h>
	13	#include <linux/kernel.h>
	14	#include <linux/sched.h>
	15	#include <linux/mm.h>
	16	#include <linux/nfs_fs.h>
	17	#include <linux/nfs_fs_sb.h>
	18	#include <linux/in6.h>
	19	#include <linux/seq_file.h>
5a0e3ad6	20	#include <linux/slab.h>
402cb8dd	21	#include <linux/iversion.h>
14727281 DH	22
14727281 DH	23	#include "internal.h"
545db45f	24	#include "iostat.h"
14727281 DH	25	#include "fscache.h"
	26
	27	#define NFSDBG_FACILITY NFSDBG_FSCACHE
	28
08734048 DH	29	static struct rb_root nfs_fscache_keys = RB_ROOT;
	30	static DEFINE_SPINLOCK(nfs_fscache_keys_lock);
	31
402cb8dd DH	32	/*
	33	* Layout of the key for an NFS server cache object.
	34	*/
	35	struct nfs_server_key {
	36	struct {
	37	uint16_t nfsversion; /* NFS protocol version */
	38	uint16_t family; /* address family */
	39	__be16 port; /* IP port */
	40	} hdr;
	41	union {
	42	struct in_addr ipv4_addr; /* IPv4 address */
	43	struct in6_addr ipv6_addr; /* IPv6 address */
	44	};
	45	} __packed;
	46
14727281 DH	47	/*
	48	* Get the per-client index cookie for an NFS client if the appropriate mount
	49	* flag was set
	50	* - We always try and get an index cookie for the client, but get filehandle
	51	* cookies on a per-superblock basis, depending on the mount flags
	52	*/
	53	void nfs_fscache_get_client_cookie(struct nfs_client *clp)
	54	{
402cb8dd DH	55	const struct sockaddr_in6 sin6 = (struct sockaddr_in6 ) &clp->cl_addr;
	56	const struct sockaddr_in sin = (struct sockaddr_in ) &clp->cl_addr;
	57	struct nfs_server_key key;
	58	uint16_t len = sizeof(key.hdr);
	59
	60	memset(&key, 0, sizeof(key));
	61	key.hdr.nfsversion = clp->rpc_ops->version;
	62	key.hdr.family = clp->cl_addr.ss_family;
	63
	64	switch (clp->cl_addr.ss_family) {
	65	case AF_INET:
	66	key.hdr.port = sin->sin_port;
	67	key.ipv4_addr = sin->sin_addr;
	68	len += sizeof(key.ipv4_addr);
	69	break;
	70
	71	case AF_INET6:
	72	key.hdr.port = sin6->sin6_port;
	73	key.ipv6_addr = sin6->sin6_addr;
	74	len += sizeof(key.ipv6_addr);
	75	break;
	76
	77	default:
	78	printk(KERN_WARNING "NFS: Unknown network family '%d'\n",
	79	clp->cl_addr.ss_family);
	80	clp->fscache = NULL;
	81	return;
	82	}
	83
14727281 DH	84	/* create a cache index for looking up filehandles */
	85	clp->fscache = fscache_acquire_cookie(nfs_fscache_netfs.primary_index,
	86	&nfs_fscache_server_index_def,
402cb8dd DH	87	&key, len,
402cb8dd DH	88	NULL, 0,
ee1235a9	89	clp, 0, true);
14727281 DH	90	dfprintk(FSCACHE, "NFS: get client cookie (0x%p/0x%p)\n",
	91	clp, clp->fscache);
	92	}
	93
	94	/*
	95	* Dispose of a per-client cookie
	96	*/
	97	void nfs_fscache_release_client_cookie(struct nfs_client *clp)
	98	{
	99	dfprintk(FSCACHE, "NFS: releasing client cookie (0x%p/0x%p)\n",
	100	clp, clp->fscache);
	101
402cb8dd	102	fscache_relinquish_cookie(clp->fscache, NULL, false);
14727281 DH	103	clp->fscache = NULL;
14727281 DH	104	}
08734048 DH	105
	106	/*
	107	* Get the cache cookie for an NFS superblock. We have to handle
	108	* uniquification here because the cache doesn't do it for us.
2df54806 DH	109	*
	110	* The default uniquifier is just an empty string, but it may be overridden
	111	* either by the 'fsc=xxx' option to mount, or by inheriting it from the parent
	112	* superblock across an automount point of some nature.
08734048	113	*/
2311b943	114	void nfs_fscache_get_super_cookie(struct super_block sb, const char uniq, int ulen)
08734048 DH	115	{
	116	struct nfs_fscache_key key, xkey;
	117	struct nfs_server *nfss = NFS_SB(sb);
	118	struct rb_node *p, parent;
2311b943	119	int diff;
2df54806 DH	120
	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;
	131	key->key.super.s_flags = sb->s_flags & NFS_MS_MASK;
	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,
402cb8dd DH	189	key, sizeof(*key) + ulen,
402cb8dd DH	190	NULL, 0,
ee1235a9	191	nfss, 0, true);
08734048 DH	192	dfprintk(FSCACHE, "NFS: get superblock cookie (0x%p/0x%p)\n",
	193	nfss, nfss->fscache);
	194	return;
	195
	196	non_unique:
	197	spin_unlock(&nfs_fscache_keys_lock);
	198	kfree(key);
	199	nfss->fscache_key = NULL;
	200	nfss->fscache = NULL;
	201	printk(KERN_WARNING "NFS:"
	202	" Cache request denied due to non-unique superblock keys\n");
	203	}
	204
	205	/*
	206	* release a per-superblock cookie
	207	*/
	208	void nfs_fscache_release_super_cookie(struct super_block *sb)
	209	{
	210	struct nfs_server *nfss = NFS_SB(sb);
	211
	212	dfprintk(FSCACHE, "NFS: releasing superblock cookie (0x%p/0x%p)\n",
	213	nfss, nfss->fscache);
	214
402cb8dd	215	fscache_relinquish_cookie(nfss->fscache, NULL, false);
08734048 DH	216	nfss->fscache = NULL;
	217
	218	if (nfss->fscache_key) {
	219	spin_lock(&nfs_fscache_keys_lock);
	220	rb_erase(&nfss->fscache_key->node, &nfs_fscache_keys);
	221	spin_unlock(&nfs_fscache_keys_lock);
	222	kfree(nfss->fscache_key);
	223	nfss->fscache_key = NULL;
	224	}
	225	}
ef79c097 DH	226
	227	/*
	228	* Initialise the per-inode cache cookie pointer for an NFS inode.
	229	*/
f1fe29b4	230	void nfs_fscache_init_inode(struct inode *inode)
ef79c097	231	{
402cb8dd	232	struct nfs_fscache_inode_auxdata auxdata;
ef79c097 DH	233	struct nfs_inode *nfsi = NFS_I(inode);
ef79c097 DH	234
f1fe29b4 DH	235	nfsi->fscache = NULL;
f1fe29b4 DH	236	if (!S_ISREG(inode->i_mode))
ef79c097	237	return;
402cb8dd DH	238
402cb8dd DH	239	memset(&auxdata, 0, sizeof(auxdata));
95582b00 DD	240	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
95582b00 DD	241	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
402cb8dd DH	242
	243	if (NFS_SERVER(&nfsi->vfs_inode)->nfs_client->rpc_ops->version == 4)
	244	auxdata.change_attr = inode_peek_iversion_raw(&nfsi->vfs_inode);
	245
f1fe29b4 DH	246	nfsi->fscache = fscache_acquire_cookie(NFS_SB(inode->i_sb)->fscache,
f1fe29b4 DH	247	&nfs_fscache_inode_object_def,
402cb8dd DH	248	nfsi->fh.data, nfsi->fh.size,
402cb8dd DH	249	&auxdata, sizeof(auxdata),
ee1235a9	250	nfsi, nfsi->vfs_inode.i_size, false);
ef79c097 DH	251	}
	252
	253	/*
	254	* Release a per-inode cookie.
	255	*/
f1fe29b4	256	void nfs_fscache_clear_inode(struct inode *inode)
ef79c097	257	{
402cb8dd	258	struct nfs_fscache_inode_auxdata auxdata;
ef79c097	259	struct nfs_inode *nfsi = NFS_I(inode);
f1fe29b4	260	struct fscache_cookie *cookie = nfs_i_fscache(inode);
ef79c097	261
f1fe29b4	262	dfprintk(FSCACHE, "NFS: clear cookie (0x%p/0x%p)\n", nfsi, cookie);
ef79c097	263
402cb8dd	264	memset(&auxdata, 0, sizeof(auxdata));
95582b00 DD	265	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
95582b00 DD	266	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
402cb8dd	267	fscache_relinquish_cookie(cookie, &auxdata, false);
ef79c097 DH	268	nfsi->fscache = NULL;
	269	}
	270
f1fe29b4	271	static bool nfs_fscache_can_enable(void *data)
ef79c097	272	{
f1fe29b4	273	struct inode *inode = data;
ef79c097	274
f1fe29b4	275	return !inode_is_open_for_write(inode);
ef79c097 DH	276	}
	277
	278	/*
f1fe29b4 DH	279	* Enable or disable caching for a file that is being opened as appropriate.
	280	* The cookie is allocated when the inode is initialised, but is not enabled at
	281	* that time. Enablement is deferred to file-open time to avoid stat() and
	282	* access() thrashing the cache.
	283	*
	284	* For now, with NFS, only regular files that are open read-only will be able
	285	* to use the cache.
	286	*
	287	* We enable the cache for an inode if we open it read-only and it isn't
	288	* currently open for writing. We disable the cache if the inode is open
	289	* write-only.
	290	*
	291	* The caller uses the file struct to pin i_writecount on the inode before
	292	* calling us when a file is opened for writing, so we can make use of that.
	293	*
	294	* Note that this may be invoked multiple times in parallel by parallel
	295	* nfs_open() functions.
ef79c097	296	*/
f1fe29b4	297	void nfs_fscache_open_file(struct inode inode, struct file filp)
ef79c097	298	{
402cb8dd	299	struct nfs_fscache_inode_auxdata auxdata;
ef79c097	300	struct nfs_inode *nfsi = NFS_I(inode);
f1fe29b4	301	struct fscache_cookie *cookie = nfs_i_fscache(inode);
ef79c097	302
f1fe29b4 DH	303	if (!fscache_cookie_valid(cookie))
f1fe29b4 DH	304	return;
ef79c097	305
402cb8dd	306	memset(&auxdata, 0, sizeof(auxdata));
95582b00 DD	307	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
95582b00 DD	308	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
402cb8dd	309
f1fe29b4 DH	310	if (inode_is_open_for_write(inode)) {
	311	dfprintk(FSCACHE, "NFS: nfsi 0x%p disabling cache\n", nfsi);
	312	clear_bit(NFS_INO_FSCACHE, &nfsi->flags);
402cb8dd	313	fscache_disable_cookie(cookie, &auxdata, true);
f1fe29b4 DH	314	fscache_uncache_all_inode_pages(cookie, inode);
	315	} else {
	316	dfprintk(FSCACHE, "NFS: nfsi 0x%p enabling cache\n", nfsi);
ee1235a9	317	fscache_enable_cookie(cookie, &auxdata, nfsi->vfs_inode.i_size,
402cb8dd	318	nfs_fscache_can_enable, inode);
f1fe29b4 DH	319	if (fscache_cookie_enabled(cookie))
f1fe29b4 DH	320	set_bit(NFS_INO_FSCACHE, &NFS_I(inode)->flags);
ef79c097	321	}
ef79c097	322	}
f1fe29b4	323	EXPORT_SYMBOL_GPL(nfs_fscache_open_file);
545db45f DH	324
	325	/*
	326	* Release the caching state associated with a page, if the page isn't busy
	327	* interacting with the cache.
	328	* - Returns true (can release page) or false (page busy).
	329	*/
	330	int nfs_fscache_release_page(struct page *page, gfp_t gfp)
	331	{
545db45f	332	if (PageFsCache(page)) {
f1fe29b4	333	struct fscache_cookie *cookie = nfs_i_fscache(page->mapping->host);
2c174009 TM	334
2c174009 TM	335	BUG_ON(!cookie);
545db45f	336	dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
f1fe29b4	337	cookie, page, NFS_I(page->mapping->host));
545db45f	338
201a1542 DH	339	if (!fscache_maybe_release_page(cookie, page, gfp))
	340	return 0;
	341
e9f456ca LR	342	nfs_inc_fscache_stats(page->mapping->host,
e9f456ca LR	343	NFSIOS_FSCACHE_PAGES_UNCACHED);
545db45f DH	344	}
	345
	346	return 1;
	347	}
	348
	349	/*
	350	* Release the caching state associated with a page if undergoing complete page
	351	* invalidation.
	352	*/
	353	void __nfs_fscache_invalidate_page(struct page page, struct inode inode)
	354	{
f1fe29b4	355	struct fscache_cookie *cookie = nfs_i_fscache(inode);
545db45f DH	356
	357	BUG_ON(!cookie);
	358
	359	dfprintk(FSCACHE, "NFS: fscache invalidatepage (0x%p/0x%p/0x%p)\n",
f1fe29b4	360	cookie, page, NFS_I(inode));
545db45f DH	361
	362	fscache_wait_on_page_write(cookie, page);
	363
	364	BUG_ON(!PageLocked(page));
	365	fscache_uncache_page(cookie, page);
e9f456ca LR	366	nfs_inc_fscache_stats(page->mapping->host,
e9f456ca LR	367	NFSIOS_FSCACHE_PAGES_UNCACHED);
545db45f	368	}
9a9fc1c0 DH	369
	370	/*
	371	* Handle completion of a page being read from the cache.
	372	* - Called in process (keventd) context.
	373	*/
	374	static void nfs_readpage_from_fscache_complete(struct page *page,
	375	void *context,
	376	int error)
	377	{
	378	dfprintk(FSCACHE,
	379	"NFS: readpage_from_fscache_complete (0x%p/0x%p/%d)\n",
	380	page, context, error);
	381
	382	/* if the read completes with an error, we just unlock the page and let
	383	* the VM reissue the readpage */
	384	if (!error) {
	385	SetPageUptodate(page);
	386	unlock_page(page);
	387	} else {
	388	error = nfs_readpage_async(context, page->mapping->host, page);
	389	if (error)
	390	unlock_page(page);
	391	}
	392	}
	393
	394	/*
	395	* Retrieve a page from fscache
	396	*/
	397	int __nfs_readpage_from_fscache(struct nfs_open_context *ctx,
	398	struct inode inode, struct page page)
	399	{
	400	int ret;
	401
	402	dfprintk(FSCACHE,
	403	"NFS: readpage_from_fscache(fsc:%p/p:%p(i:%lx f:%lx)/0x%p)\n",
f1fe29b4	404	nfs_i_fscache(inode), page, page->index, page->flags, inode);
9a9fc1c0	405
f1fe29b4	406	ret = fscache_read_or_alloc_page(nfs_i_fscache(inode),
9a9fc1c0 DH	407	page,
	408	nfs_readpage_from_fscache_complete,
	409	ctx,
	410	GFP_KERNEL);
	411
	412	switch (ret) {
	413	case 0: /* read BIO submitted (page in fscache) */
	414	dfprintk(FSCACHE,
	415	"NFS: readpage_from_fscache: BIO submitted\n");
e9f456ca	416	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK);
9a9fc1c0 DH	417	return ret;
	418
	419	case -ENOBUFS: /* inode not in cache */
	420	case -ENODATA: /* page not in cache */
e9f456ca	421	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
9a9fc1c0 DH	422	dfprintk(FSCACHE,
	423	"NFS: readpage_from_fscache %d\n", ret);
	424	return 1;
	425
	426	default:
	427	dfprintk(FSCACHE, "NFS: readpage_from_fscache %d\n", ret);
e9f456ca	428	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
9a9fc1c0 DH	429	}
	430	return ret;
	431	}
	432
	433	/*
	434	* Retrieve a set of pages from fscache
	435	*/
	436	int __nfs_readpages_from_fscache(struct nfs_open_context *ctx,
	437	struct inode *inode,
	438	struct address_space *mapping,
	439	struct list_head *pages,
	440	unsigned *nr_pages)
	441	{
0f15c53d CL	442	unsigned npages = *nr_pages;
0f15c53d CL	443	int ret;
9a9fc1c0 DH	444
9a9fc1c0 DH	445	dfprintk(FSCACHE, "NFS: nfs_getpages_from_fscache (0x%p/%u/0x%p)\n",
f1fe29b4	446	nfs_i_fscache(inode), npages, inode);
9a9fc1c0	447
f1fe29b4	448	ret = fscache_read_or_alloc_pages(nfs_i_fscache(inode),
9a9fc1c0 DH	449	mapping, pages, nr_pages,
	450	nfs_readpage_from_fscache_complete,
	451	ctx,
	452	mapping_gfp_mask(mapping));
	453	if (*nr_pages < npages)
	454	nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK,
	455	npages);
	456	if (*nr_pages > 0)
	457	nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL,
	458	*nr_pages);
	459
	460	switch (ret) {
	461	case 0: /* read submitted to the cache for all pages */
	462	BUG_ON(!list_empty(pages));
	463	BUG_ON(*nr_pages != 0);
	464	dfprintk(FSCACHE,
	465	"NFS: nfs_getpages_from_fscache: submitted\n");
	466
	467	return ret;
	468
	469	case -ENOBUFS: /* some pages aren't cached and can't be */
	470	case -ENODATA: /* some pages aren't cached */
	471	dfprintk(FSCACHE,
	472	"NFS: nfs_getpages_from_fscache: no page: %d\n", ret);
	473	return 1;
	474
	475	default:
	476	dfprintk(FSCACHE,
	477	"NFS: nfs_getpages_from_fscache: ret %d\n", ret);
	478	}
	479
	480	return ret;
	481	}
7f8e05f6 DH	482
	483	/*
	484	* Store a newly fetched page in fscache
	485	* - PG_fscache must be set on the page
	486	*/
	487	void __nfs_readpage_to_fscache(struct inode inode, struct page page, int sync)
	488	{
	489	int ret;
	490
	491	dfprintk(FSCACHE,
	492	"NFS: readpage_to_fscache(fsc:%p/p:%p(i:%lx f:%lx)/%d)\n",
f1fe29b4	493	nfs_i_fscache(inode), page, page->index, page->flags, sync);
7f8e05f6	494
ee1235a9 DH	495	ret = fscache_write_page(nfs_i_fscache(inode), page,
ee1235a9 DH	496	inode->i_size, GFP_KERNEL);
7f8e05f6 DH	497	dfprintk(FSCACHE,
	498	"NFS: readpage_to_fscache: p:%p(i:%lu f:%lx) ret %d\n",
	499	page, page->index, page->flags, ret);
	500
	501	if (ret != 0) {
f1fe29b4	502	fscache_uncache_page(nfs_i_fscache(inode), page);
e9f456ca LR	503	nfs_inc_fscache_stats(inode,
	504	NFSIOS_FSCACHE_PAGES_WRITTEN_FAIL);
	505	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_UNCACHED);
7f8e05f6	506	} else {
e9f456ca LR	507	nfs_inc_fscache_stats(inode,
e9f456ca LR	508	NFSIOS_FSCACHE_PAGES_WRITTEN_OK);
7f8e05f6 DH	509	}
7f8e05f6 DH	510	}