q3k/linux
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
linux/fs/ncpfs/ncplib_kernel.h

255 lines
7.7 KiB
C
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
/*
* ncplib_kernel.h
*
* Copyright (C) 1995, 1996 by Volker Lendecke
* Modified for big endian by J.F. Chadima and David S. Miller
* Modified 1997 Peter Waltenberg, Bill Hawes, David Woodhouse for 2.1 dcache
* Modified 1998, 1999 Wolfram Pienkoss for NLS
* Modified 1999 Wolfram Pienkoss for directory caching
*
*/
#ifndef _NCPLIB_H
#define _NCPLIB_H
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/stat.h>
#include <linux/fcntl.h>
#include <linux/pagemap.h>
#include <asm/uaccess.h>
#include <asm/byteorder.h>
#include <asm/unaligned.h>
#include <asm/string.h>
#ifdef CONFIG_NCPFS_NLS
#include <linux/nls.h>
#else
#include <linux/ctype.h>
#endif /* CONFIG_NCPFS_NLS */
#define NCP_MIN_SYMLINK_SIZE 8
#define NCP_MAX_SYMLINK_SIZE 512
#define NCP_BLOCK_SHIFT 9
#define NCP_BLOCK_SIZE (1 << (NCP_BLOCK_SHIFT))
int ncp_negotiate_buffersize(struct ncp_server *, int, int *);
int ncp_negotiate_size_and_options(struct ncp_server *server, int size,
int options, int *ret_size, int *ret_options);
int ncp_get_volume_info_with_number(struct ncp_server* server, int n,
struct ncp_volume_info *target);
int ncp_get_directory_info(struct ncp_server* server, __u8 dirhandle,
struct ncp_volume_info* target);
int ncp_close_file(struct ncp_server *, const char *);
static inline int ncp_read_bounce_size(__u32 size) {
return sizeof(struct ncp_reply_header) + 2 + 2 + size + 8;
};
int ncp_read_bounce(struct ncp_server *, const char *, __u32, __u16,
char __user *, int *, void* bounce, __u32 bouncelen);
int ncp_read_kernel(struct ncp_server *, const char *, __u32, __u16,
char *, int *);
int ncp_write_kernel(struct ncp_server *, const char *, __u32, __u16,
const char *, int *);
static inline void ncp_inode_close(struct inode *inode) {
atomic_dec(&NCP_FINFO(inode)->opened);
}
BKL: Remove BKL from ncpfs Dozen of changes in ncpfs to provide some locking other than BKL. In readdir cache unlock and mark complete first page as last operation, so it can be used for synchronization, as code intended. When updating dentry name on case insensitive filesystems do at least some basic locking... Hold i_mutex when updating inode fields. Push some ncp_conn_is_valid down to ncp_request. Connection can become invalid at any moment, and fewer error code paths to test the better. Use i_size_{read,write} to modify file size. Set inode's backing_dev_info as ncpfs has its own special bdi. In ioctl unbreak ioctls invoked on filesystem mounted 'ro' - tests are for inode writeable or owner match, but were turned to filesystem writeable and inode writeable or owner match. Also collect all permission checks in single place. Add some locking, and remove comments saying that it would be cool to add some locks to the code. Constify some pointers. Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2010-09-26 23:47:33 +00:00
void ncp_extract_file_info(const void* src, struct nw_info_struct* target);
int ncp_obtain_info(struct ncp_server *server, struct inode *, const char *,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
struct nw_info_struct *target);
int ncp_obtain_nfs_info(struct ncp_server *server, struct nw_info_struct *target);
BKL: Remove BKL from ncpfs Dozen of changes in ncpfs to provide some locking other than BKL. In readdir cache unlock and mark complete first page as last operation, so it can be used for synchronization, as code intended. When updating dentry name on case insensitive filesystems do at least some basic locking... Hold i_mutex when updating inode fields. Push some ncp_conn_is_valid down to ncp_request. Connection can become invalid at any moment, and fewer error code paths to test the better. Use i_size_{read,write} to modify file size. Set inode's backing_dev_info as ncpfs has its own special bdi. In ioctl unbreak ioctls invoked on filesystem mounted 'ro' - tests are for inode writeable or owner match, but were turned to filesystem writeable and inode writeable or owner match. Also collect all permission checks in single place. Add some locking, and remove comments saying that it would be cool to add some locks to the code. Constify some pointers. Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2010-09-26 23:47:33 +00:00
int ncp_update_known_namespace(struct ncp_server *server, __u8 volume, int *ret_ns);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
int ncp_get_volume_root(struct ncp_server *server, const char *volname,
__u32 *volume, __le32 *dirent, __le32 *dosdirent);
int ncp_lookup_volume(struct ncp_server *, const char *, struct nw_info_struct *);
int ncp_modify_file_or_subdir_dos_info(struct ncp_server *, struct inode *,
__le32, const struct nw_modify_dos_info *info);
int ncp_modify_file_or_subdir_dos_info_path(struct ncp_server *, struct inode *,
const char* path, __le32, const struct nw_modify_dos_info *info);
int ncp_modify_nfs_info(struct ncp_server *, __u8 volnum, __le32 dirent,
__u32 mode, __u32 rdev);
int ncp_del_file_or_subdir2(struct ncp_server *, struct dentry*);
BKL: Remove BKL from ncpfs Dozen of changes in ncpfs to provide some locking other than BKL. In readdir cache unlock and mark complete first page as last operation, so it can be used for synchronization, as code intended. When updating dentry name on case insensitive filesystems do at least some basic locking... Hold i_mutex when updating inode fields. Push some ncp_conn_is_valid down to ncp_request. Connection can become invalid at any moment, and fewer error code paths to test the better. Use i_size_{read,write} to modify file size. Set inode's backing_dev_info as ncpfs has its own special bdi. In ioctl unbreak ioctls invoked on filesystem mounted 'ro' - tests are for inode writeable or owner match, but were turned to filesystem writeable and inode writeable or owner match. Also collect all permission checks in single place. Add some locking, and remove comments saying that it would be cool to add some locks to the code. Constify some pointers. Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2010-09-26 23:47:33 +00:00
int ncp_del_file_or_subdir(struct ncp_server *, struct inode *, const char *);
int ncp_open_create_file_or_subdir(struct ncp_server *, struct inode *, const char *,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
int, __le32, __le16, struct ncp_entry_info *);
int ncp_initialize_search(struct ncp_server *, struct inode *,
struct nw_search_sequence *target);
int ncp_search_for_fileset(struct ncp_server *server,
struct nw_search_sequence *seq,
int* more, int* cnt,
char* buffer, size_t bufsize,
char** rbuf, size_t* rsize);
int ncp_ren_or_mov_file_or_subdir(struct ncp_server *server,
BKL: Remove BKL from ncpfs Dozen of changes in ncpfs to provide some locking other than BKL. In readdir cache unlock and mark complete first page as last operation, so it can be used for synchronization, as code intended. When updating dentry name on case insensitive filesystems do at least some basic locking... Hold i_mutex when updating inode fields. Push some ncp_conn_is_valid down to ncp_request. Connection can become invalid at any moment, and fewer error code paths to test the better. Use i_size_{read,write} to modify file size. Set inode's backing_dev_info as ncpfs has its own special bdi. In ioctl unbreak ioctls invoked on filesystem mounted 'ro' - tests are for inode writeable or owner match, but were turned to filesystem writeable and inode writeable or owner match. Also collect all permission checks in single place. Add some locking, and remove comments saying that it would be cool to add some locks to the code. Constify some pointers. Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2010-09-26 23:47:33 +00:00
struct inode *, const char *, struct inode *, const char *);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
int
ncp_LogPhysicalRecord(struct ncp_server *server,
const char *file_id, __u8 locktype,
__u32 offset, __u32 length, __u16 timeout);
#ifdef CONFIG_NCPFS_IOCTL_LOCKING
int
ncp_ClearPhysicalRecord(struct ncp_server *server,
const char *file_id,
__u32 offset, __u32 length);
#endif /* CONFIG_NCPFS_IOCTL_LOCKING */
int
ncp_mount_subdir(struct ncp_server *, __u8, __u8, __le32,
__u32* volume, __le32* dirent, __le32* dosdirent);
int ncp_dirhandle_alloc(struct ncp_server *, __u8 vol, __le32 dirent, __u8 *dirhandle);
int ncp_dirhandle_free(struct ncp_server *, __u8 dirhandle);
int ncp_create_new(struct inode *dir, struct dentry *dentry,
ncpfs: propagate umode_t Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2011-07-26 07:12:16 +00:00
umode_t mode, dev_t rdev, __le32 attributes);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
static inline int ncp_is_nfs_extras(struct ncp_server* server, unsigned int volnum) {
#ifdef CONFIG_NCPFS_NFS_NS
return (server->m.flags & NCP_MOUNT_NFS_EXTRAS) &&
(server->name_space[volnum] == NW_NS_NFS);
#else
return 0;
#endif
}
#ifdef CONFIG_NCPFS_NLS
int ncp__io2vol(struct ncp_server *, unsigned char *, unsigned int *,
const unsigned char *, unsigned int, int);
int ncp__vol2io(struct ncp_server *, unsigned char *, unsigned int *,
const unsigned char *, unsigned int, int);
#define NCP_ESC ':'
fs: change d_compare for rcu-walk Change d_compare so it may be called from lock-free RCU lookups. This does put significant restrictions on what may be done from the callback, however there don't seem to have been any problems with in-tree fses. If some strange use case pops up that _really_ cannot cope with the rcu-walk rules, we can just add new rcu-unaware callbacks, which would cause name lookup to drop out of rcu-walk mode. For in-tree filesystems, this is just a mechanical change. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:27 +00:00
#define NCP_IO_TABLE(sb) (NCP_SBP(sb)->nls_io)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
#define ncp_tolower(t, c) nls_tolower(t, c)
#define ncp_toupper(t, c) nls_toupper(t, c)
#define ncp_strnicmp(t, s1, s2, len) \
nls_strnicmp(t, s1, s2, len)
#define ncp_io2vol(S,m,i,n,k,U) ncp__io2vol(S,m,i,n,k,U)
#define ncp_vol2io(S,m,i,n,k,U) ncp__vol2io(S,m,i,n,k,U)
#else
int ncp__io2vol(unsigned char *, unsigned int *,
const unsigned char *, unsigned int, int);
int ncp__vol2io(unsigned char *, unsigned int *,
const unsigned char *, unsigned int, int);
fs: change d_compare for rcu-walk Change d_compare so it may be called from lock-free RCU lookups. This does put significant restrictions on what may be done from the callback, however there don't seem to have been any problems with in-tree fses. If some strange use case pops up that _really_ cannot cope with the rcu-walk rules, we can just add new rcu-unaware callbacks, which would cause name lookup to drop out of rcu-walk mode. For in-tree filesystems, this is just a mechanical change. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:27 +00:00
#define NCP_IO_TABLE(sb) NULL
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
#define ncp_tolower(t, c) tolower(c)
#define ncp_toupper(t, c) toupper(c)
#define ncp_io2vol(S,m,i,n,k,U) ncp__io2vol(m,i,n,k,U)
#define ncp_vol2io(S,m,i,n,k,U) ncp__vol2io(m,i,n,k,U)
fs: change d_compare for rcu-walk Change d_compare so it may be called from lock-free RCU lookups. This does put significant restrictions on what may be done from the callback, however there don't seem to have been any problems with in-tree fses. If some strange use case pops up that _really_ cannot cope with the rcu-walk rules, we can just add new rcu-unaware callbacks, which would cause name lookup to drop out of rcu-walk mode. For in-tree filesystems, this is just a mechanical change. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:27 +00:00
static inline int ncp_strnicmp(const struct nls_table *t,
const unsigned char *s1, const unsigned char *s2, int len)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
{
while (len--) {
if (tolower(*s1++) != tolower(*s2++))
return 1;
}
return 0;
}
#endif /* CONFIG_NCPFS_NLS */
#define NCP_GET_AGE(dentry) (jiffies - (dentry)->d_time)
BKL: Remove BKL from ncpfs Dozen of changes in ncpfs to provide some locking other than BKL. In readdir cache unlock and mark complete first page as last operation, so it can be used for synchronization, as code intended. When updating dentry name on case insensitive filesystems do at least some basic locking... Hold i_mutex when updating inode fields. Push some ncp_conn_is_valid down to ncp_request. Connection can become invalid at any moment, and fewer error code paths to test the better. Use i_size_{read,write} to modify file size. Set inode's backing_dev_info as ncpfs has its own special bdi. In ioctl unbreak ioctls invoked on filesystem mounted 'ro' - tests are for inode writeable or owner match, but were turned to filesystem writeable and inode writeable or owner match. Also collect all permission checks in single place. Add some locking, and remove comments saying that it would be cool to add some locks to the code. Constify some pointers. Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2010-09-26 23:47:33 +00:00
#define NCP_MAX_AGE(server) atomic_read(&(server)->dentry_ttl)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
#define NCP_TEST_AGE(server,dentry) (NCP_GET_AGE(dentry) < NCP_MAX_AGE(server))
static inline void
ncp_age_dentry(struct ncp_server* server, struct dentry* dentry)
{
BKL: Remove BKL from ncpfs Dozen of changes in ncpfs to provide some locking other than BKL. In readdir cache unlock and mark complete first page as last operation, so it can be used for synchronization, as code intended. When updating dentry name on case insensitive filesystems do at least some basic locking... Hold i_mutex when updating inode fields. Push some ncp_conn_is_valid down to ncp_request. Connection can become invalid at any moment, and fewer error code paths to test the better. Use i_size_{read,write} to modify file size. Set inode's backing_dev_info as ncpfs has its own special bdi. In ioctl unbreak ioctls invoked on filesystem mounted 'ro' - tests are for inode writeable or owner match, but were turned to filesystem writeable and inode writeable or owner match. Also collect all permission checks in single place. Add some locking, and remove comments saying that it would be cool to add some locks to the code. Constify some pointers. Signed-off-by: Petr Vandrovec <petr@vandrovec.name> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2010-09-26 23:47:33 +00:00
dentry->d_time = jiffies - NCP_MAX_AGE(server);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
}
static inline void
ncp_new_dentry(struct dentry* dentry)
{
dentry->d_time = jiffies;
}
static inline void
ncp_renew_dentries(struct dentry *parent)
{
struct ncp_server *server = NCP_SERVER(parent->d_inode);
struct list_head *next;
struct dentry *dentry;
fs: dcache scale subdirs Protect d_subdirs and d_child with d_lock, except in filesystems that aren't using dcache_lock for these anyway (eg. using i_mutex). Note: if we change the locking rule in future so that ->d_child protection is provided only with ->d_parent->d_lock, it may allow us to reduce some locking. But it would be an exception to an otherwise regular locking scheme, so we'd have to see some good results. Probably not worthwhile. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:34 +00:00
spin_lock(&parent->d_lock);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
next = parent->d_subdirs.next;
while (next != &parent->d_subdirs) {
[PATCH] shrink dentry struct Some long time ago, dentry struct was carefully tuned so that on 32 bits UP, sizeof(struct dentry) was exactly 128, ie a power of 2, and a multiple of memory cache lines. Then RCU was added and dentry struct enlarged by two pointers, with nice results for SMP, but not so good on UP, because breaking the above tuning (128 + 8 = 136 bytes) This patch reverts this unwanted side effect, by using an union (d_u), where d_rcu and d_child are placed so that these two fields can share their memory needs. At the time d_free() is called (and d_rcu is really used), d_child is known to be empty and not touched by the dentry freeing. Lockless lookups only access d_name, d_parent, d_lock, d_op, d_flags (so the previous content of d_child is not needed if said dentry was unhashed but still accessed by a CPU because of RCU constraints) As dentry cache easily contains millions of entries, a size reduction is worth the extra complexity of the ugly C union. Signed-off-by: Eric Dumazet <dada1@cosmosbay.com> Cc: Dipankar Sarma <dipankar@in.ibm.com> Cc: Maneesh Soni <maneesh@in.ibm.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Ian Kent <raven@themaw.net> Cc: Paul Jackson <pj@sgi.com> Cc: Al Viro <viro@ftp.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Neil Brown <neilb@cse.unsw.edu.au> Cc: James Morris <jmorris@namei.org> Cc: Stephen Smalley <sds@epoch.ncsc.mil> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-08 09:03:32 +00:00
dentry = list_entry(next, struct dentry, d_u.d_child);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
if (dentry->d_fsdata == NULL)
ncp_age_dentry(server, dentry);
else
ncp_new_dentry(dentry);
next = next->next;
}
fs: dcache scale subdirs Protect d_subdirs and d_child with d_lock, except in filesystems that aren't using dcache_lock for these anyway (eg. using i_mutex). Note: if we change the locking rule in future so that ->d_child protection is provided only with ->d_parent->d_lock, it may allow us to reduce some locking. But it would be an exception to an otherwise regular locking scheme, so we'd have to see some good results. Probably not worthwhile. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:34 +00:00
spin_unlock(&parent->d_lock);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
}
static inline void
ncp_invalidate_dircache_entries(struct dentry *parent)
{
struct ncp_server *server = NCP_SERVER(parent->d_inode);
struct list_head *next;
struct dentry *dentry;
fs: dcache scale subdirs Protect d_subdirs and d_child with d_lock, except in filesystems that aren't using dcache_lock for these anyway (eg. using i_mutex). Note: if we change the locking rule in future so that ->d_child protection is provided only with ->d_parent->d_lock, it may allow us to reduce some locking. But it would be an exception to an otherwise regular locking scheme, so we'd have to see some good results. Probably not worthwhile. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:34 +00:00
spin_lock(&parent->d_lock);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
next = parent->d_subdirs.next;
while (next != &parent->d_subdirs) {
[PATCH] shrink dentry struct Some long time ago, dentry struct was carefully tuned so that on 32 bits UP, sizeof(struct dentry) was exactly 128, ie a power of 2, and a multiple of memory cache lines. Then RCU was added and dentry struct enlarged by two pointers, with nice results for SMP, but not so good on UP, because breaking the above tuning (128 + 8 = 136 bytes) This patch reverts this unwanted side effect, by using an union (d_u), where d_rcu and d_child are placed so that these two fields can share their memory needs. At the time d_free() is called (and d_rcu is really used), d_child is known to be empty and not touched by the dentry freeing. Lockless lookups only access d_name, d_parent, d_lock, d_op, d_flags (so the previous content of d_child is not needed if said dentry was unhashed but still accessed by a CPU because of RCU constraints) As dentry cache easily contains millions of entries, a size reduction is worth the extra complexity of the ugly C union. Signed-off-by: Eric Dumazet <dada1@cosmosbay.com> Cc: Dipankar Sarma <dipankar@in.ibm.com> Cc: Maneesh Soni <maneesh@in.ibm.com> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: "Paul E. McKenney" <paulmck@us.ibm.com> Cc: Ian Kent <raven@themaw.net> Cc: Paul Jackson <pj@sgi.com> Cc: Al Viro <viro@ftp.linux.org.uk> Cc: Christoph Hellwig <hch@lst.de> Cc: Trond Myklebust <trond.myklebust@fys.uio.no> Cc: Neil Brown <neilb@cse.unsw.edu.au> Cc: James Morris <jmorris@namei.org> Cc: Stephen Smalley <sds@epoch.ncsc.mil> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-08 09:03:32 +00:00
dentry = list_entry(next, struct dentry, d_u.d_child);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
dentry->d_fsdata = NULL;
ncp_age_dentry(server, dentry);
next = next->next;
}
fs: dcache scale subdirs Protect d_subdirs and d_child with d_lock, except in filesystems that aren't using dcache_lock for these anyway (eg. using i_mutex). Note: if we change the locking rule in future so that ->d_child protection is provided only with ->d_parent->d_lock, it may allow us to reduce some locking. But it would be an exception to an otherwise regular locking scheme, so we'd have to see some good results. Probably not worthwhile. Signed-off-by: Nick Piggin <npiggin@kernel.dk>
2011-01-07 06:49:34 +00:00
spin_unlock(&parent->d_lock);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-16 22:20:36 +00:00
}
struct ncp_cache_head {
time_t mtime;
unsigned long time; /* cache age */
unsigned long end; /* last valid fpos in cache */
int eof;
};
#define NCP_DIRCACHE_SIZE ((int)(PAGE_CACHE_SIZE/sizeof(struct dentry *)))
union ncp_dir_cache {
struct ncp_cache_head head;
struct dentry *dentry[NCP_DIRCACHE_SIZE];
};
#define NCP_FIRSTCACHE_SIZE ((int)((NCP_DIRCACHE_SIZE * \
sizeof(struct dentry *) - sizeof(struct ncp_cache_head)) / \
sizeof(struct dentry *)))
#define NCP_DIRCACHE_START (NCP_DIRCACHE_SIZE - NCP_FIRSTCACHE_SIZE)
struct ncp_cache_control {
struct ncp_cache_head head;
struct page *page;
union ncp_dir_cache *cache;
unsigned long fpos, ofs;
int filled, valid, idx;
};
#endif /* _NCPLIB_H */
Reference in a new issue Copy permalink