remarkable-linux/fs/ext3/acl.c
Linus Torvalds 9c64daff9d ext3: avoid unnecessary spinlock in critical POSIX ACL path
If a filesystem supports POSIX ACL's, the VFS layer expects the filesystem
to do POSIX ACL checks on any files not owned by the caller, and it does
this for every single pathname component that it looks up.

That obviously can be pretty expensive if the filesystem isn't careful
about it, especially with locking. That's doubly sad, since the common
case tends to be that there are no ACL's associated with the files in
question.

ext3 already caches the ACL data so that it doesn't have to look it up
over and over again, but it does so by taking the inode->i_lock spinlock
on every lookup. Which is a noticeable overhead even if it's a private
lock, especially on CPU's where the serialization is expensive (eg Intel
Netburst aka 'P4').

For the special case of not actually having any ACL's, all that locking is
unnecessary. Even if somebody else were to be changing the ACL's on
another CPU, we simply don't care - if we've seen a NULL ACL, we might as
well use it.

So just load the ACL speculatively without any locking, and if it was
NULL, just use it. If it's non-NULL (either because we had a cached
entry, or because the cache hasn't been filled in at all), it means that
we'll need to get the lock and re-load it properly.

This is noticeable even on Nehalem, which does locking quite well (much
better than P4). From lmbench:

	Processor, Processes - times in microseconds - smaller is better
	--------------------------------------------------------------------
	Host                 OS  Mhz null null      open slct fork exec sh
	                             call  I/O stat clos TCP  proc proc proc
	--------- ------------- ---- ---- ---- ---- ---- ---- ---- ---- ----
 - before:
	nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.95 1.45 2.18 69.1 273. 1141
	nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.95 1.48 2.28 69.9 253. 1140
	nehalem.l Linux 2.6.30- 3193 0.04 0.10 0.95 1.42 2.19 68.6 284. 1141
 - after:
	nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.92 1.44 2.12 68.3 282. 1094
	nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.92 1.39 2.20 67.0 308. 1123
	nehalem.l Linux 2.6.30- 3193 0.04 0.09 0.92 1.39 2.36 67.4 293. 1148

where you can see what appears to be a roughly 3% improvement in stat
and open/close latencies from just the removal of the locking overhead.

Of course, this only matters for files you don't own (the owner never
needs to do the ACL checks), but that's the common case for libraries,
header files, and executables. As well as for the base components of any
absolute pathname, even if you are the owner of the final file.

[ At some point we probably want to move this ACL caching logic entirely
  into the VFS layer (and only call down to the filesystem when
  uncached), but in the meantime this improves ext3 a bit.

  A similar fix to btrfs makes a much bigger difference (15x improvement
  in lmbench) due to broken caching. ]

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Acked-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2009-06-17 00:36:35 -04:00

555 lines
12 KiB
C

/*
* linux/fs/ext3/acl.c
*
* Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
*/
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/capability.h>
#include <linux/fs.h>
#include <linux/ext3_jbd.h>
#include <linux/ext3_fs.h>
#include "xattr.h"
#include "acl.h"
/*
* Convert from filesystem to in-memory representation.
*/
static struct posix_acl *
ext3_acl_from_disk(const void *value, size_t size)
{
const char *end = (char *)value + size;
int n, count;
struct posix_acl *acl;
if (!value)
return NULL;
if (size < sizeof(ext3_acl_header))
return ERR_PTR(-EINVAL);
if (((ext3_acl_header *)value)->a_version !=
cpu_to_le32(EXT3_ACL_VERSION))
return ERR_PTR(-EINVAL);
value = (char *)value + sizeof(ext3_acl_header);
count = ext3_acl_count(size);
if (count < 0)
return ERR_PTR(-EINVAL);
if (count == 0)
return NULL;
acl = posix_acl_alloc(count, GFP_NOFS);
if (!acl)
return ERR_PTR(-ENOMEM);
for (n=0; n < count; n++) {
ext3_acl_entry *entry =
(ext3_acl_entry *)value;
if ((char *)value + sizeof(ext3_acl_entry_short) > end)
goto fail;
acl->a_entries[n].e_tag = le16_to_cpu(entry->e_tag);
acl->a_entries[n].e_perm = le16_to_cpu(entry->e_perm);
switch(acl->a_entries[n].e_tag) {
case ACL_USER_OBJ:
case ACL_GROUP_OBJ:
case ACL_MASK:
case ACL_OTHER:
value = (char *)value +
sizeof(ext3_acl_entry_short);
acl->a_entries[n].e_id = ACL_UNDEFINED_ID;
break;
case ACL_USER:
case ACL_GROUP:
value = (char *)value + sizeof(ext3_acl_entry);
if ((char *)value > end)
goto fail;
acl->a_entries[n].e_id =
le32_to_cpu(entry->e_id);
break;
default:
goto fail;
}
}
if (value != end)
goto fail;
return acl;
fail:
posix_acl_release(acl);
return ERR_PTR(-EINVAL);
}
/*
* Convert from in-memory to filesystem representation.
*/
static void *
ext3_acl_to_disk(const struct posix_acl *acl, size_t *size)
{
ext3_acl_header *ext_acl;
char *e;
size_t n;
*size = ext3_acl_size(acl->a_count);
ext_acl = kmalloc(sizeof(ext3_acl_header) + acl->a_count *
sizeof(ext3_acl_entry), GFP_NOFS);
if (!ext_acl)
return ERR_PTR(-ENOMEM);
ext_acl->a_version = cpu_to_le32(EXT3_ACL_VERSION);
e = (char *)ext_acl + sizeof(ext3_acl_header);
for (n=0; n < acl->a_count; n++) {
ext3_acl_entry *entry = (ext3_acl_entry *)e;
entry->e_tag = cpu_to_le16(acl->a_entries[n].e_tag);
entry->e_perm = cpu_to_le16(acl->a_entries[n].e_perm);
switch(acl->a_entries[n].e_tag) {
case ACL_USER:
case ACL_GROUP:
entry->e_id =
cpu_to_le32(acl->a_entries[n].e_id);
e += sizeof(ext3_acl_entry);
break;
case ACL_USER_OBJ:
case ACL_GROUP_OBJ:
case ACL_MASK:
case ACL_OTHER:
e += sizeof(ext3_acl_entry_short);
break;
default:
goto fail;
}
}
return (char *)ext_acl;
fail:
kfree(ext_acl);
return ERR_PTR(-EINVAL);
}
static inline struct posix_acl *
ext3_iget_acl(struct inode *inode, struct posix_acl **i_acl)
{
struct posix_acl *acl = ACCESS_ONCE(*i_acl);
if (acl) {
spin_lock(&inode->i_lock);
acl = *i_acl;
if (acl != EXT3_ACL_NOT_CACHED)
acl = posix_acl_dup(acl);
spin_unlock(&inode->i_lock);
}
return acl;
}
static inline void
ext3_iset_acl(struct inode *inode, struct posix_acl **i_acl,
struct posix_acl *acl)
{
spin_lock(&inode->i_lock);
if (*i_acl != EXT3_ACL_NOT_CACHED)
posix_acl_release(*i_acl);
*i_acl = posix_acl_dup(acl);
spin_unlock(&inode->i_lock);
}
/*
* Inode operation get_posix_acl().
*
* inode->i_mutex: don't care
*/
static struct posix_acl *
ext3_get_acl(struct inode *inode, int type)
{
struct ext3_inode_info *ei = EXT3_I(inode);
int name_index;
char *value = NULL;
struct posix_acl *acl;
int retval;
if (!test_opt(inode->i_sb, POSIX_ACL))
return NULL;
switch(type) {
case ACL_TYPE_ACCESS:
acl = ext3_iget_acl(inode, &ei->i_acl);
if (acl != EXT3_ACL_NOT_CACHED)
return acl;
name_index = EXT3_XATTR_INDEX_POSIX_ACL_ACCESS;
break;
case ACL_TYPE_DEFAULT:
acl = ext3_iget_acl(inode, &ei->i_default_acl);
if (acl != EXT3_ACL_NOT_CACHED)
return acl;
name_index = EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT;
break;
default:
return ERR_PTR(-EINVAL);
}
retval = ext3_xattr_get(inode, name_index, "", NULL, 0);
if (retval > 0) {
value = kmalloc(retval, GFP_NOFS);
if (!value)
return ERR_PTR(-ENOMEM);
retval = ext3_xattr_get(inode, name_index, "", value, retval);
}
if (retval > 0)
acl = ext3_acl_from_disk(value, retval);
else if (retval == -ENODATA || retval == -ENOSYS)
acl = NULL;
else
acl = ERR_PTR(retval);
kfree(value);
if (!IS_ERR(acl)) {
switch(type) {
case ACL_TYPE_ACCESS:
ext3_iset_acl(inode, &ei->i_acl, acl);
break;
case ACL_TYPE_DEFAULT:
ext3_iset_acl(inode, &ei->i_default_acl, acl);
break;
}
}
return acl;
}
/*
* Set the access or default ACL of an inode.
*
* inode->i_mutex: down unless called from ext3_new_inode
*/
static int
ext3_set_acl(handle_t *handle, struct inode *inode, int type,
struct posix_acl *acl)
{
struct ext3_inode_info *ei = EXT3_I(inode);
int name_index;
void *value = NULL;
size_t size = 0;
int error;
if (S_ISLNK(inode->i_mode))
return -EOPNOTSUPP;
switch(type) {
case ACL_TYPE_ACCESS:
name_index = EXT3_XATTR_INDEX_POSIX_ACL_ACCESS;
if (acl) {
mode_t mode = inode->i_mode;
error = posix_acl_equiv_mode(acl, &mode);
if (error < 0)
return error;
else {
inode->i_mode = mode;
ext3_mark_inode_dirty(handle, inode);
if (error == 0)
acl = NULL;
}
}
break;
case ACL_TYPE_DEFAULT:
name_index = EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT;
if (!S_ISDIR(inode->i_mode))
return acl ? -EACCES : 0;
break;
default:
return -EINVAL;
}
if (acl) {
value = ext3_acl_to_disk(acl, &size);
if (IS_ERR(value))
return (int)PTR_ERR(value);
}
error = ext3_xattr_set_handle(handle, inode, name_index, "",
value, size, 0);
kfree(value);
if (!error) {
switch(type) {
case ACL_TYPE_ACCESS:
ext3_iset_acl(inode, &ei->i_acl, acl);
break;
case ACL_TYPE_DEFAULT:
ext3_iset_acl(inode, &ei->i_default_acl, acl);
break;
}
}
return error;
}
static int
ext3_check_acl(struct inode *inode, int mask)
{
struct posix_acl *acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
if (IS_ERR(acl))
return PTR_ERR(acl);
if (acl) {
int error = posix_acl_permission(inode, acl, mask);
posix_acl_release(acl);
return error;
}
return -EAGAIN;
}
int
ext3_permission(struct inode *inode, int mask)
{
return generic_permission(inode, mask, ext3_check_acl);
}
/*
* Initialize the ACLs of a new inode. Called from ext3_new_inode.
*
* dir->i_mutex: down
* inode->i_mutex: up (access to inode is still exclusive)
*/
int
ext3_init_acl(handle_t *handle, struct inode *inode, struct inode *dir)
{
struct posix_acl *acl = NULL;
int error = 0;
if (!S_ISLNK(inode->i_mode)) {
if (test_opt(dir->i_sb, POSIX_ACL)) {
acl = ext3_get_acl(dir, ACL_TYPE_DEFAULT);
if (IS_ERR(acl))
return PTR_ERR(acl);
}
if (!acl)
inode->i_mode &= ~current_umask();
}
if (test_opt(inode->i_sb, POSIX_ACL) && acl) {
struct posix_acl *clone;
mode_t mode;
if (S_ISDIR(inode->i_mode)) {
error = ext3_set_acl(handle, inode,
ACL_TYPE_DEFAULT, acl);
if (error)
goto cleanup;
}
clone = posix_acl_clone(acl, GFP_NOFS);
error = -ENOMEM;
if (!clone)
goto cleanup;
mode = inode->i_mode;
error = posix_acl_create_masq(clone, &mode);
if (error >= 0) {
inode->i_mode = mode;
if (error > 0) {
/* This is an extended ACL */
error = ext3_set_acl(handle, inode,
ACL_TYPE_ACCESS, clone);
}
}
posix_acl_release(clone);
}
cleanup:
posix_acl_release(acl);
return error;
}
/*
* Does chmod for an inode that may have an Access Control List. The
* inode->i_mode field must be updated to the desired value by the caller
* before calling this function.
* Returns 0 on success, or a negative error number.
*
* We change the ACL rather than storing some ACL entries in the file
* mode permission bits (which would be more efficient), because that
* would break once additional permissions (like ACL_APPEND, ACL_DELETE
* for directories) are added. There are no more bits available in the
* file mode.
*
* inode->i_mutex: down
*/
int
ext3_acl_chmod(struct inode *inode)
{
struct posix_acl *acl, *clone;
int error;
if (S_ISLNK(inode->i_mode))
return -EOPNOTSUPP;
if (!test_opt(inode->i_sb, POSIX_ACL))
return 0;
acl = ext3_get_acl(inode, ACL_TYPE_ACCESS);
if (IS_ERR(acl) || !acl)
return PTR_ERR(acl);
clone = posix_acl_clone(acl, GFP_KERNEL);
posix_acl_release(acl);
if (!clone)
return -ENOMEM;
error = posix_acl_chmod_masq(clone, inode->i_mode);
if (!error) {
handle_t *handle;
int retries = 0;
retry:
handle = ext3_journal_start(inode,
EXT3_DATA_TRANS_BLOCKS(inode->i_sb));
if (IS_ERR(handle)) {
error = PTR_ERR(handle);
ext3_std_error(inode->i_sb, error);
goto out;
}
error = ext3_set_acl(handle, inode, ACL_TYPE_ACCESS, clone);
ext3_journal_stop(handle);
if (error == -ENOSPC &&
ext3_should_retry_alloc(inode->i_sb, &retries))
goto retry;
}
out:
posix_acl_release(clone);
return error;
}
/*
* Extended attribute handlers
*/
static size_t
ext3_xattr_list_acl_access(struct inode *inode, char *list, size_t list_len,
const char *name, size_t name_len)
{
const size_t size = sizeof(POSIX_ACL_XATTR_ACCESS);
if (!test_opt(inode->i_sb, POSIX_ACL))
return 0;
if (list && size <= list_len)
memcpy(list, POSIX_ACL_XATTR_ACCESS, size);
return size;
}
static size_t
ext3_xattr_list_acl_default(struct inode *inode, char *list, size_t list_len,
const char *name, size_t name_len)
{
const size_t size = sizeof(POSIX_ACL_XATTR_DEFAULT);
if (!test_opt(inode->i_sb, POSIX_ACL))
return 0;
if (list && size <= list_len)
memcpy(list, POSIX_ACL_XATTR_DEFAULT, size);
return size;
}
static int
ext3_xattr_get_acl(struct inode *inode, int type, void *buffer, size_t size)
{
struct posix_acl *acl;
int error;
if (!test_opt(inode->i_sb, POSIX_ACL))
return -EOPNOTSUPP;
acl = ext3_get_acl(inode, type);
if (IS_ERR(acl))
return PTR_ERR(acl);
if (acl == NULL)
return -ENODATA;
error = posix_acl_to_xattr(acl, buffer, size);
posix_acl_release(acl);
return error;
}
static int
ext3_xattr_get_acl_access(struct inode *inode, const char *name,
void *buffer, size_t size)
{
if (strcmp(name, "") != 0)
return -EINVAL;
return ext3_xattr_get_acl(inode, ACL_TYPE_ACCESS, buffer, size);
}
static int
ext3_xattr_get_acl_default(struct inode *inode, const char *name,
void *buffer, size_t size)
{
if (strcmp(name, "") != 0)
return -EINVAL;
return ext3_xattr_get_acl(inode, ACL_TYPE_DEFAULT, buffer, size);
}
static int
ext3_xattr_set_acl(struct inode *inode, int type, const void *value,
size_t size)
{
handle_t *handle;
struct posix_acl *acl;
int error, retries = 0;
if (!test_opt(inode->i_sb, POSIX_ACL))
return -EOPNOTSUPP;
if (!is_owner_or_cap(inode))
return -EPERM;
if (value) {
acl = posix_acl_from_xattr(value, size);
if (IS_ERR(acl))
return PTR_ERR(acl);
else if (acl) {
error = posix_acl_valid(acl);
if (error)
goto release_and_out;
}
} else
acl = NULL;
retry:
handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS(inode->i_sb));
if (IS_ERR(handle))
return PTR_ERR(handle);
error = ext3_set_acl(handle, inode, type, acl);
ext3_journal_stop(handle);
if (error == -ENOSPC && ext3_should_retry_alloc(inode->i_sb, &retries))
goto retry;
release_and_out:
posix_acl_release(acl);
return error;
}
static int
ext3_xattr_set_acl_access(struct inode *inode, const char *name,
const void *value, size_t size, int flags)
{
if (strcmp(name, "") != 0)
return -EINVAL;
return ext3_xattr_set_acl(inode, ACL_TYPE_ACCESS, value, size);
}
static int
ext3_xattr_set_acl_default(struct inode *inode, const char *name,
const void *value, size_t size, int flags)
{
if (strcmp(name, "") != 0)
return -EINVAL;
return ext3_xattr_set_acl(inode, ACL_TYPE_DEFAULT, value, size);
}
struct xattr_handler ext3_xattr_acl_access_handler = {
.prefix = POSIX_ACL_XATTR_ACCESS,
.list = ext3_xattr_list_acl_access,
.get = ext3_xattr_get_acl_access,
.set = ext3_xattr_set_acl_access,
};
struct xattr_handler ext3_xattr_acl_default_handler = {
.prefix = POSIX_ACL_XATTR_DEFAULT,
.list = ext3_xattr_list_acl_default,
.get = ext3_xattr_get_acl_default,
.set = ext3_xattr_set_acl_default,
};