kernel_optimize_test/fs/ubifs/xattr.c
Richard Weinberger 272eda8298 ubifs: Correctly evict xattr inodes
UBIFS handles extended attributes just like files, as consequence of
that, they also have inodes.
Therefore UBIFS does all the inode machinery also for xattrs. Since new
inodes have i_nlink of 1, a file or xattr inode will be evicted
if i_nlink goes down to 0 after an unlink. UBIFS assumes this model also
for xattrs, which is not correct.
One can create a file "foo" with xattr "user.test". By reading
"user.test" an inode will be created, and by deleting "user.test" it
will get evicted later. The assumption breaks if the file "foo", which
hosts the xattrs, will be removed. VFS nor UBIFS does not remove each
xattr via ubifs_xattr_remove(), it just removes the host inode from
the TNC and all underlying xattr nodes too and the inode will remain
in the cache and wastes memory.

To solve this problem, remove xattr inodes from the VFS inode cache in
ubifs_xattr_remove() to make sure that they get evicted.

Fixes: 1e51764a3c ("UBIFS: add new flash file system")
Cc: <stable@vger.kernel.org>
Signed-off-by: Richard Weinberger <richard@nod.at>
2017-07-14 22:49:04 +02:00

682 lines
18 KiB
C

/*
* This file is part of UBIFS.
*
* Copyright (C) 2006-2008 Nokia Corporation.
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 as published by
* the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along with
* this program; if not, write to the Free Software Foundation, Inc., 51
* Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*
* Authors: Artem Bityutskiy (Битюцкий Артём)
* Adrian Hunter
*/
/*
* This file implements UBIFS extended attributes support.
*
* Extended attributes are implemented as regular inodes with attached data,
* which limits extended attribute size to UBIFS block size (4KiB). Names of
* extended attributes are described by extended attribute entries (xentries),
* which are almost identical to directory entries, but have different key type.
*
* In other words, the situation with extended attributes is very similar to
* directories. Indeed, any inode (but of course not xattr inodes) may have a
* number of associated xentries, just like directory inodes have associated
* directory entries. Extended attribute entries store the name of the extended
* attribute, the host inode number, and the extended attribute inode number.
* Similarly, direntries store the name, the parent and the target inode
* numbers. Thus, most of the common UBIFS mechanisms may be re-used for
* extended attributes.
*
* The number of extended attributes is not limited, but there is Linux
* limitation on the maximum possible size of the list of all extended
* attributes associated with an inode (%XATTR_LIST_MAX), so UBIFS makes sure
* the sum of all extended attribute names of the inode does not exceed that
* limit.
*
* Extended attributes are synchronous, which means they are written to the
* flash media synchronously and there is no write-back for extended attribute
* inodes. The extended attribute values are not stored in compressed form on
* the media.
*
* Since extended attributes are represented by regular inodes, they are cached
* in the VFS inode cache. The xentries are cached in the LNC cache (see
* tnc.c).
*
* ACL support is not implemented.
*/
#include "ubifs.h"
#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/xattr.h>
/*
* Limit the number of extended attributes per inode so that the total size
* (@xattr_size) is guaranteeded to fit in an 'unsigned int'.
*/
#define MAX_XATTRS_PER_INODE 65535
/*
* Extended attribute type constants.
*
* USER_XATTR: user extended attribute ("user.*")
* TRUSTED_XATTR: trusted extended attribute ("trusted.*)
* SECURITY_XATTR: security extended attribute ("security.*")
*/
enum {
USER_XATTR,
TRUSTED_XATTR,
SECURITY_XATTR,
};
static const struct inode_operations empty_iops;
static const struct file_operations empty_fops;
/**
* create_xattr - create an extended attribute.
* @c: UBIFS file-system description object
* @host: host inode
* @nm: extended attribute name
* @value: extended attribute value
* @size: size of extended attribute value
*
* This is a helper function which creates an extended attribute of name @nm
* and value @value for inode @host. The host inode is also updated on flash
* because the ctime and extended attribute accounting data changes. This
* function returns zero in case of success and a negative error code in case
* of failure.
*/
static int create_xattr(struct ubifs_info *c, struct inode *host,
const struct fscrypt_name *nm, const void *value, int size)
{
int err, names_len;
struct inode *inode;
struct ubifs_inode *ui, *host_ui = ubifs_inode(host);
struct ubifs_budget_req req = { .new_ino = 1, .new_dent = 1,
.new_ino_d = ALIGN(size, 8), .dirtied_ino = 1,
.dirtied_ino_d = ALIGN(host_ui->data_len, 8) };
if (host_ui->xattr_cnt >= MAX_XATTRS_PER_INODE) {
ubifs_err(c, "inode %lu already has too many xattrs (%d), cannot create more",
host->i_ino, host_ui->xattr_cnt);
return -ENOSPC;
}
/*
* Linux limits the maximum size of the extended attribute names list
* to %XATTR_LIST_MAX. This means we should not allow creating more
* extended attributes if the name list becomes larger. This limitation
* is artificial for UBIFS, though.
*/
names_len = host_ui->xattr_names + host_ui->xattr_cnt + fname_len(nm) + 1;
if (names_len > XATTR_LIST_MAX) {
ubifs_err(c, "cannot add one more xattr name to inode %lu, total names length would become %d, max. is %d",
host->i_ino, names_len, XATTR_LIST_MAX);
return -ENOSPC;
}
err = ubifs_budget_space(c, &req);
if (err)
return err;
inode = ubifs_new_inode(c, host, S_IFREG | S_IRWXUGO);
if (IS_ERR(inode)) {
err = PTR_ERR(inode);
goto out_budg;
}
/* Re-define all operations to be "nothing" */
inode->i_mapping->a_ops = &empty_aops;
inode->i_op = &empty_iops;
inode->i_fop = &empty_fops;
inode->i_flags |= S_SYNC | S_NOATIME | S_NOCMTIME | S_NOQUOTA;
ui = ubifs_inode(inode);
ui->xattr = 1;
ui->flags |= UBIFS_XATTR_FL;
ui->data = kmemdup(value, size, GFP_NOFS);
if (!ui->data) {
err = -ENOMEM;
goto out_free;
}
inode->i_size = ui->ui_size = size;
ui->data_len = size;
mutex_lock(&host_ui->ui_mutex);
host->i_ctime = current_time(host);
host_ui->xattr_cnt += 1;
host_ui->xattr_size += CALC_DENT_SIZE(fname_len(nm));
host_ui->xattr_size += CALC_XATTR_BYTES(size);
host_ui->xattr_names += fname_len(nm);
/*
* We handle UBIFS_XATTR_NAME_ENCRYPTION_CONTEXT here because we
* have to set the UBIFS_CRYPT_FL flag on the host inode.
* To avoid multiple updates of the same inode in the same operation,
* let's do it here.
*/
if (strcmp(fname_name(nm), UBIFS_XATTR_NAME_ENCRYPTION_CONTEXT) == 0)
host_ui->flags |= UBIFS_CRYPT_FL;
err = ubifs_jnl_update(c, host, nm, inode, 0, 1);
if (err)
goto out_cancel;
mutex_unlock(&host_ui->ui_mutex);
ubifs_release_budget(c, &req);
insert_inode_hash(inode);
iput(inode);
return 0;
out_cancel:
host_ui->xattr_cnt -= 1;
host_ui->xattr_size -= CALC_DENT_SIZE(fname_len(nm));
host_ui->xattr_size -= CALC_XATTR_BYTES(size);
host_ui->xattr_names -= fname_len(nm);
host_ui->flags &= ~UBIFS_CRYPT_FL;
mutex_unlock(&host_ui->ui_mutex);
out_free:
make_bad_inode(inode);
iput(inode);
out_budg:
ubifs_release_budget(c, &req);
return err;
}
/**
* change_xattr - change an extended attribute.
* @c: UBIFS file-system description object
* @host: host inode
* @inode: extended attribute inode
* @value: extended attribute value
* @size: size of extended attribute value
*
* This helper function changes the value of extended attribute @inode with new
* data from @value. Returns zero in case of success and a negative error code
* in case of failure.
*/
static int change_xattr(struct ubifs_info *c, struct inode *host,
struct inode *inode, const void *value, int size)
{
int err;
struct ubifs_inode *host_ui = ubifs_inode(host);
struct ubifs_inode *ui = ubifs_inode(inode);
void *buf = NULL;
int old_size;
struct ubifs_budget_req req = { .dirtied_ino = 2,
.dirtied_ino_d = ALIGN(size, 8) + ALIGN(host_ui->data_len, 8) };
ubifs_assert(ui->data_len == inode->i_size);
err = ubifs_budget_space(c, &req);
if (err)
return err;
buf = kmemdup(value, size, GFP_NOFS);
if (!buf) {
err = -ENOMEM;
goto out_free;
}
mutex_lock(&ui->ui_mutex);
kfree(ui->data);
ui->data = buf;
inode->i_size = ui->ui_size = size;
old_size = ui->data_len;
ui->data_len = size;
mutex_unlock(&ui->ui_mutex);
mutex_lock(&host_ui->ui_mutex);
host->i_ctime = current_time(host);
host_ui->xattr_size -= CALC_XATTR_BYTES(old_size);
host_ui->xattr_size += CALC_XATTR_BYTES(size);
/*
* It is important to write the host inode after the xattr inode
* because if the host inode gets synchronized (via 'fsync()'), then
* the extended attribute inode gets synchronized, because it goes
* before the host inode in the write-buffer.
*/
err = ubifs_jnl_change_xattr(c, inode, host);
if (err)
goto out_cancel;
mutex_unlock(&host_ui->ui_mutex);
ubifs_release_budget(c, &req);
return 0;
out_cancel:
host_ui->xattr_size -= CALC_XATTR_BYTES(size);
host_ui->xattr_size += CALC_XATTR_BYTES(old_size);
mutex_unlock(&host_ui->ui_mutex);
make_bad_inode(inode);
out_free:
ubifs_release_budget(c, &req);
return err;
}
static struct inode *iget_xattr(struct ubifs_info *c, ino_t inum)
{
struct inode *inode;
inode = ubifs_iget(c->vfs_sb, inum);
if (IS_ERR(inode)) {
ubifs_err(c, "dead extended attribute entry, error %d",
(int)PTR_ERR(inode));
return inode;
}
if (ubifs_inode(inode)->xattr)
return inode;
ubifs_err(c, "corrupt extended attribute entry");
iput(inode);
return ERR_PTR(-EINVAL);
}
int ubifs_xattr_set(struct inode *host, const char *name, const void *value,
size_t size, int flags)
{
struct inode *inode;
struct ubifs_info *c = host->i_sb->s_fs_info;
struct fscrypt_name nm = { .disk_name = FSTR_INIT((char *)name, strlen(name))};
struct ubifs_dent_node *xent;
union ubifs_key key;
int err;
/*
* Creating an encryption context is done unlocked since we
* operate on a new inode which is not visible to other users
* at this point.
*/
if (strcmp(name, UBIFS_XATTR_NAME_ENCRYPTION_CONTEXT) != 0)
ubifs_assert(inode_is_locked(host));
if (size > UBIFS_MAX_INO_DATA)
return -ERANGE;
if (fname_len(&nm) > UBIFS_MAX_NLEN)
return -ENAMETOOLONG;
xent = kmalloc(UBIFS_MAX_XENT_NODE_SZ, GFP_NOFS);
if (!xent)
return -ENOMEM;
/*
* The extended attribute entries are stored in LNC, so multiple
* look-ups do not involve reading the flash.
*/
xent_key_init(c, &key, host->i_ino, &nm);
err = ubifs_tnc_lookup_nm(c, &key, xent, &nm);
if (err) {
if (err != -ENOENT)
goto out_free;
if (flags & XATTR_REPLACE)
/* We are asked not to create the xattr */
err = -ENODATA;
else
err = create_xattr(c, host, &nm, value, size);
goto out_free;
}
if (flags & XATTR_CREATE) {
/* We are asked not to replace the xattr */
err = -EEXIST;
goto out_free;
}
inode = iget_xattr(c, le64_to_cpu(xent->inum));
if (IS_ERR(inode)) {
err = PTR_ERR(inode);
goto out_free;
}
err = change_xattr(c, host, inode, value, size);
iput(inode);
out_free:
kfree(xent);
return err;
}
ssize_t ubifs_xattr_get(struct inode *host, const char *name, void *buf,
size_t size)
{
struct inode *inode;
struct ubifs_info *c = host->i_sb->s_fs_info;
struct fscrypt_name nm = { .disk_name = FSTR_INIT((char *)name, strlen(name))};
struct ubifs_inode *ui;
struct ubifs_dent_node *xent;
union ubifs_key key;
int err;
if (fname_len(&nm) > UBIFS_MAX_NLEN)
return -ENAMETOOLONG;
xent = kmalloc(UBIFS_MAX_XENT_NODE_SZ, GFP_NOFS);
if (!xent)
return -ENOMEM;
xent_key_init(c, &key, host->i_ino, &nm);
err = ubifs_tnc_lookup_nm(c, &key, xent, &nm);
if (err) {
if (err == -ENOENT)
err = -ENODATA;
goto out_unlock;
}
inode = iget_xattr(c, le64_to_cpu(xent->inum));
if (IS_ERR(inode)) {
err = PTR_ERR(inode);
goto out_unlock;
}
ui = ubifs_inode(inode);
ubifs_assert(inode->i_size == ui->data_len);
ubifs_assert(ubifs_inode(host)->xattr_size > ui->data_len);
mutex_lock(&ui->ui_mutex);
if (buf) {
/* If @buf is %NULL we are supposed to return the length */
if (ui->data_len > size) {
ubifs_err(c, "buffer size %zd, xattr len %d",
size, ui->data_len);
err = -ERANGE;
goto out_iput;
}
memcpy(buf, ui->data, ui->data_len);
}
err = ui->data_len;
out_iput:
mutex_unlock(&ui->ui_mutex);
iput(inode);
out_unlock:
kfree(xent);
return err;
}
static bool xattr_visible(const char *name)
{
/* File encryption related xattrs are for internal use only */
if (strcmp(name, UBIFS_XATTR_NAME_ENCRYPTION_CONTEXT) == 0)
return false;
/* Show trusted namespace only for "power" users */
if (strncmp(name, XATTR_TRUSTED_PREFIX,
XATTR_TRUSTED_PREFIX_LEN) == 0 && !capable(CAP_SYS_ADMIN))
return false;
return true;
}
ssize_t ubifs_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
union ubifs_key key;
struct inode *host = d_inode(dentry);
struct ubifs_info *c = host->i_sb->s_fs_info;
struct ubifs_inode *host_ui = ubifs_inode(host);
struct ubifs_dent_node *xent, *pxent = NULL;
int err, len, written = 0;
struct fscrypt_name nm = {0};
dbg_gen("ino %lu ('%pd'), buffer size %zd", host->i_ino,
dentry, size);
len = host_ui->xattr_names + host_ui->xattr_cnt;
if (!buffer)
/*
* We should return the minimum buffer size which will fit a
* null-terminated list of all the extended attribute names.
*/
return len;
if (len > size)
return -ERANGE;
lowest_xent_key(c, &key, host->i_ino);
while (1) {
xent = ubifs_tnc_next_ent(c, &key, &nm);
if (IS_ERR(xent)) {
err = PTR_ERR(xent);
break;
}
fname_name(&nm) = xent->name;
fname_len(&nm) = le16_to_cpu(xent->nlen);
if (xattr_visible(xent->name)) {
memcpy(buffer + written, fname_name(&nm), fname_len(&nm) + 1);
written += fname_len(&nm) + 1;
}
kfree(pxent);
pxent = xent;
key_read(c, &xent->key, &key);
}
kfree(pxent);
if (err != -ENOENT) {
ubifs_err(c, "cannot find next direntry, error %d", err);
return err;
}
ubifs_assert(written <= size);
return written;
}
static int remove_xattr(struct ubifs_info *c, struct inode *host,
struct inode *inode, const struct fscrypt_name *nm)
{
int err;
struct ubifs_inode *host_ui = ubifs_inode(host);
struct ubifs_inode *ui = ubifs_inode(inode);
struct ubifs_budget_req req = { .dirtied_ino = 2, .mod_dent = 1,
.dirtied_ino_d = ALIGN(host_ui->data_len, 8) };
ubifs_assert(ui->data_len == inode->i_size);
err = ubifs_budget_space(c, &req);
if (err)
return err;
mutex_lock(&host_ui->ui_mutex);
host->i_ctime = current_time(host);
host_ui->xattr_cnt -= 1;
host_ui->xattr_size -= CALC_DENT_SIZE(fname_len(nm));
host_ui->xattr_size -= CALC_XATTR_BYTES(ui->data_len);
host_ui->xattr_names -= fname_len(nm);
err = ubifs_jnl_delete_xattr(c, host, inode, nm);
if (err)
goto out_cancel;
mutex_unlock(&host_ui->ui_mutex);
ubifs_release_budget(c, &req);
return 0;
out_cancel:
host_ui->xattr_cnt += 1;
host_ui->xattr_size += CALC_DENT_SIZE(fname_len(nm));
host_ui->xattr_size += CALC_XATTR_BYTES(ui->data_len);
host_ui->xattr_names += fname_len(nm);
mutex_unlock(&host_ui->ui_mutex);
ubifs_release_budget(c, &req);
make_bad_inode(inode);
return err;
}
/**
* ubifs_evict_xattr_inode - Evict an xattr inode.
* @c: UBIFS file-system description object
* @xattr_inum: xattr inode number
*
* When an inode that hosts xattrs is being removed we have to make sure
* that cached inodes of the xattrs also get removed from the inode cache
* otherwise we'd waste memory. This function looks up an inode from the
* inode cache and clears the link counter such that iput() will evict
* the inode.
*/
void ubifs_evict_xattr_inode(struct ubifs_info *c, ino_t xattr_inum)
{
struct inode *inode;
inode = ilookup(c->vfs_sb, xattr_inum);
if (inode) {
clear_nlink(inode);
iput(inode);
}
}
static int ubifs_xattr_remove(struct inode *host, const char *name)
{
struct inode *inode;
struct ubifs_info *c = host->i_sb->s_fs_info;
struct fscrypt_name nm = { .disk_name = FSTR_INIT((char *)name, strlen(name))};
struct ubifs_dent_node *xent;
union ubifs_key key;
int err;
ubifs_assert(inode_is_locked(host));
if (fname_len(&nm) > UBIFS_MAX_NLEN)
return -ENAMETOOLONG;
xent = kmalloc(UBIFS_MAX_XENT_NODE_SZ, GFP_NOFS);
if (!xent)
return -ENOMEM;
xent_key_init(c, &key, host->i_ino, &nm);
err = ubifs_tnc_lookup_nm(c, &key, xent, &nm);
if (err) {
if (err == -ENOENT)
err = -ENODATA;
goto out_free;
}
inode = iget_xattr(c, le64_to_cpu(xent->inum));
if (IS_ERR(inode)) {
err = PTR_ERR(inode);
goto out_free;
}
ubifs_assert(inode->i_nlink == 1);
clear_nlink(inode);
err = remove_xattr(c, host, inode, &nm);
if (err)
set_nlink(inode, 1);
/* If @i_nlink is 0, 'iput()' will delete the inode */
iput(inode);
out_free:
kfree(xent);
return err;
}
#ifdef CONFIG_UBIFS_FS_SECURITY
static int init_xattrs(struct inode *inode, const struct xattr *xattr_array,
void *fs_info)
{
const struct xattr *xattr;
char *name;
int err = 0;
for (xattr = xattr_array; xattr->name != NULL; xattr++) {
name = kmalloc(XATTR_SECURITY_PREFIX_LEN +
strlen(xattr->name) + 1, GFP_NOFS);
if (!name) {
err = -ENOMEM;
break;
}
strcpy(name, XATTR_SECURITY_PREFIX);
strcpy(name + XATTR_SECURITY_PREFIX_LEN, xattr->name);
err = ubifs_xattr_set(inode, name, xattr->value,
xattr->value_len, 0);
kfree(name);
if (err < 0)
break;
}
return err;
}
int ubifs_init_security(struct inode *dentry, struct inode *inode,
const struct qstr *qstr)
{
int err;
err = security_inode_init_security(inode, dentry, qstr,
&init_xattrs, 0);
if (err) {
struct ubifs_info *c = dentry->i_sb->s_fs_info;
ubifs_err(c, "cannot initialize security for inode %lu, error %d",
inode->i_ino, err);
}
return err;
}
#endif
static int xattr_get(const struct xattr_handler *handler,
struct dentry *dentry, struct inode *inode,
const char *name, void *buffer, size_t size)
{
dbg_gen("xattr '%s', ino %lu ('%pd'), buf size %zd", name,
inode->i_ino, dentry, size);
name = xattr_full_name(handler, name);
return ubifs_xattr_get(inode, name, buffer, size);
}
static int xattr_set(const struct xattr_handler *handler,
struct dentry *dentry, struct inode *inode,
const char *name, const void *value,
size_t size, int flags)
{
dbg_gen("xattr '%s', host ino %lu ('%pd'), size %zd",
name, inode->i_ino, dentry, size);
name = xattr_full_name(handler, name);
if (value)
return ubifs_xattr_set(inode, name, value, size, flags);
else
return ubifs_xattr_remove(inode, name);
}
static const struct xattr_handler ubifs_user_xattr_handler = {
.prefix = XATTR_USER_PREFIX,
.get = xattr_get,
.set = xattr_set,
};
static const struct xattr_handler ubifs_trusted_xattr_handler = {
.prefix = XATTR_TRUSTED_PREFIX,
.get = xattr_get,
.set = xattr_set,
};
#ifdef CONFIG_UBIFS_FS_SECURITY
static const struct xattr_handler ubifs_security_xattr_handler = {
.prefix = XATTR_SECURITY_PREFIX,
.get = xattr_get,
.set = xattr_set,
};
#endif
const struct xattr_handler *ubifs_xattr_handlers[] = {
&ubifs_user_xattr_handler,
&ubifs_trusted_xattr_handler,
#ifdef CONFIG_UBIFS_FS_SECURITY
&ubifs_security_xattr_handler,
#endif
NULL
};