linux/fs/nfs/fscache.h
David Howells ee4cdf7ba8
netfs: Speed up buffered reading
Improve the efficiency of buffered reads in a number of ways:

 (1) Overhaul the algorithm in general so that it's a lot more compact and
     split the read submission code between buffered and unbuffered
     versions.  The unbuffered version can be vastly simplified.

 (2) Read-result collection is handed off to a work queue rather than being
     done in the I/O thread.  Multiple subrequests can be processes
     simultaneously.

 (3) When a subrequest is collected, any folios it fully spans are
     collected and "spare" data on either side is donated to either the
     previous or the next subrequest in the sequence.

Notes:

 (*) Readahead expansion is massively slows down fio, presumably because it
     causes a load of extra allocations, both folio and xarray, up front
     before RPC requests can be transmitted.

 (*) RDMA with cifs does appear to work, both with SIW and RXE.

 (*) PG_private_2-based reading and copy-to-cache is split out into its own
     file and altered to use folio_queue.  Note that the copy to the cache
     now creates a new write transaction against the cache and adds the
     folios to be copied into it.  This allows it to use part of the
     writeback I/O code.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/20240814203850.2240469-20-dhowells@redhat.com/ # v2
Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-09-12 12:20:41 +02:00

202 lines
6.5 KiB
C

/* SPDX-License-Identifier: GPL-2.0-or-later */
/* NFS filesystem cache interface definitions
*
* Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*/
#ifndef _NFS_FSCACHE_H
#define _NFS_FSCACHE_H
#include <linux/swap.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
#include <linux/nfs4_mount.h>
#include <linux/fscache.h>
#include <linux/iversion.h>
#ifdef CONFIG_NFS_FSCACHE
/*
* Definition of the auxiliary data attached to NFS inode storage objects
* within the cache.
*
* The contents of this struct are recorded in the on-disk local cache in the
* auxiliary data attached to the data storage object backing an inode. This
* permits coherency to be managed when a new inode binds to an already extant
* cache object.
*/
struct nfs_fscache_inode_auxdata {
s64 mtime_sec;
s64 mtime_nsec;
s64 ctime_sec;
s64 ctime_nsec;
u64 change_attr;
};
struct nfs_netfs_io_data {
/*
* NFS may split a netfs_io_subrequest into multiple RPCs, each
* with their own read completion. In netfs, we can only call
* netfs_subreq_terminated() once for each subrequest. Use the
* refcount here to double as a marker of the last RPC completion,
* and only call netfs via netfs_subreq_terminated() once.
*/
refcount_t refcount;
struct netfs_io_subrequest *sreq;
/*
* Final disposition of the netfs_io_subrequest, sent in
* netfs_subreq_terminated()
*/
atomic64_t transferred;
int error;
};
static inline void nfs_netfs_get(struct nfs_netfs_io_data *netfs)
{
refcount_inc(&netfs->refcount);
}
static inline void nfs_netfs_put(struct nfs_netfs_io_data *netfs)
{
/* Only the last RPC completion should call netfs_subreq_terminated() */
if (!refcount_dec_and_test(&netfs->refcount))
return;
/*
* The NFS pageio interface may read a complete page, even when netfs
* only asked for a partial page. Specifically, this may be seen when
* one thread is truncating a file while another one is reading the last
* page of the file.
* Correct the final length here to be no larger than the netfs subrequest
* length, and thus avoid netfs's "Subreq overread" warning message.
*/
netfs->sreq->transferred = min_t(s64, netfs->sreq->len,
atomic64_read(&netfs->transferred));
netfs_read_subreq_terminated(netfs->sreq, netfs->error, false);
kfree(netfs);
}
static inline void nfs_netfs_inode_init(struct nfs_inode *nfsi)
{
netfs_inode_init(&nfsi->netfs, &nfs_netfs_ops, false);
}
extern void nfs_netfs_initiate_read(struct nfs_pgio_header *hdr);
extern void nfs_netfs_read_completion(struct nfs_pgio_header *hdr);
extern int nfs_netfs_folio_unlock(struct folio *folio);
/*
* fscache.c
*/
extern int nfs_fscache_get_super_cookie(struct super_block *, const char *, int);
extern void nfs_fscache_release_super_cookie(struct super_block *);
extern void nfs_fscache_init_inode(struct inode *);
extern void nfs_fscache_clear_inode(struct inode *);
extern void nfs_fscache_open_file(struct inode *, struct file *);
extern void nfs_fscache_release_file(struct inode *, struct file *);
extern int nfs_netfs_readahead(struct readahead_control *ractl);
extern int nfs_netfs_read_folio(struct file *file, struct folio *folio);
static inline bool nfs_fscache_release_folio(struct folio *folio, gfp_t gfp)
{
if (folio_test_private_2(folio)) { /* [DEPRECATED] */
if (current_is_kswapd() || !(gfp & __GFP_FS))
return false;
folio_wait_private_2(folio);
}
fscache_note_page_release(netfs_i_cookie(netfs_inode(folio->mapping->host)));
return true;
}
static inline void nfs_fscache_update_auxdata(struct nfs_fscache_inode_auxdata *auxdata,
struct inode *inode)
{
memset(auxdata, 0, sizeof(*auxdata));
auxdata->mtime_sec = inode_get_mtime(inode).tv_sec;
auxdata->mtime_nsec = inode_get_mtime(inode).tv_nsec;
auxdata->ctime_sec = inode_get_ctime(inode).tv_sec;
auxdata->ctime_nsec = inode_get_ctime(inode).tv_nsec;
if (NFS_SERVER(inode)->nfs_client->rpc_ops->version == 4)
auxdata->change_attr = inode_peek_iversion_raw(inode);
}
/*
* Invalidate the contents of fscache for this inode. This will not sleep.
*/
static inline void nfs_fscache_invalidate(struct inode *inode, int flags)
{
struct nfs_fscache_inode_auxdata auxdata;
struct fscache_cookie *cookie = netfs_i_cookie(&NFS_I(inode)->netfs);
nfs_fscache_update_auxdata(&auxdata, inode);
fscache_invalidate(cookie, &auxdata, i_size_read(inode), flags);
}
/*
* indicate the client caching state as readable text
*/
static inline const char *nfs_server_fscache_state(struct nfs_server *server)
{
if (server->fscache)
return "yes";
return "no ";
}
static inline void nfs_netfs_set_pgio_header(struct nfs_pgio_header *hdr,
struct nfs_pageio_descriptor *desc)
{
hdr->netfs = desc->pg_netfs;
}
static inline void nfs_netfs_set_pageio_descriptor(struct nfs_pageio_descriptor *desc,
struct nfs_pgio_header *hdr)
{
desc->pg_netfs = hdr->netfs;
}
static inline void nfs_netfs_reset_pageio_descriptor(struct nfs_pageio_descriptor *desc)
{
desc->pg_netfs = NULL;
}
#else /* CONFIG_NFS_FSCACHE */
static inline void nfs_netfs_inode_init(struct nfs_inode *nfsi) {}
static inline void nfs_netfs_initiate_read(struct nfs_pgio_header *hdr) {}
static inline void nfs_netfs_read_completion(struct nfs_pgio_header *hdr) {}
static inline int nfs_netfs_folio_unlock(struct folio *folio)
{
return 1;
}
static inline void nfs_fscache_release_super_cookie(struct super_block *sb) {}
static inline void nfs_fscache_init_inode(struct inode *inode) {}
static inline void nfs_fscache_clear_inode(struct inode *inode) {}
static inline void nfs_fscache_open_file(struct inode *inode,
struct file *filp) {}
static inline void nfs_fscache_release_file(struct inode *inode, struct file *file) {}
static inline int nfs_netfs_readahead(struct readahead_control *ractl)
{
return -ENOBUFS;
}
static inline int nfs_netfs_read_folio(struct file *file, struct folio *folio)
{
return -ENOBUFS;
}
static inline bool nfs_fscache_release_folio(struct folio *folio, gfp_t gfp)
{
return true; /* may release folio */
}
static inline void nfs_fscache_invalidate(struct inode *inode, int flags) {}
static inline const char *nfs_server_fscache_state(struct nfs_server *server)
{
return "no ";
}
static inline void nfs_netfs_set_pgio_header(struct nfs_pgio_header *hdr,
struct nfs_pageio_descriptor *desc) {}
static inline void nfs_netfs_set_pageio_descriptor(struct nfs_pageio_descriptor *desc,
struct nfs_pgio_header *hdr) {}
static inline void nfs_netfs_reset_pageio_descriptor(struct nfs_pageio_descriptor *desc) {}
#endif /* CONFIG_NFS_FSCACHE */
#endif /* _NFS_FSCACHE_H */