linux-brain/fs/btrfs/xattr.c
Filipe Manana e49baccfe2 btrfs: fix warning when creating a directory with smack enabled
commit fd57a98d6f0c98fa295813087f13afb26c224e73 upstream.

When we have smack enabled, during the creation of a directory smack may
attempt to add a "smack transmute" xattr on the inode, which results in
the following warning and trace:

  WARNING: CPU: 3 PID: 2548 at fs/btrfs/transaction.c:537 start_transaction+0x489/0x4f0
  Modules linked in: nft_objref nf_conntrack_netbios_ns (...)
  CPU: 3 PID: 2548 Comm: mkdir Not tainted 5.9.0-rc2smack+ #81
  Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-2.fc32 04/01/2014
  RIP: 0010:start_transaction+0x489/0x4f0
  Code: e9 be fc ff ff (...)
  RSP: 0018:ffffc90001887d10 EFLAGS: 00010202
  RAX: ffff88816f1e0000 RBX: 0000000000000201 RCX: 0000000000000003
  RDX: 0000000000000201 RSI: 0000000000000002 RDI: ffff888177849000
  RBP: ffff888177849000 R08: 0000000000000001 R09: 0000000000000004
  R10: ffffffff825e8f7a R11: 0000000000000003 R12: ffffffffffffffe2
  R13: 0000000000000000 R14: ffff88803d884270 R15: ffff8881680d8000
  FS:  00007f67317b8440(0000) GS:ffff88817bcc0000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007f67247a22a8 CR3: 000000004bfbc002 CR4: 0000000000370ee0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  Call Trace:
   ? slab_free_freelist_hook+0xea/0x1b0
   ? trace_hardirqs_on+0x1c/0xe0
   btrfs_setxattr_trans+0x3c/0xf0
   __vfs_setxattr+0x63/0x80
   smack_d_instantiate+0x2d3/0x360
   security_d_instantiate+0x29/0x40
   d_instantiate_new+0x38/0x90
   btrfs_mkdir+0x1cf/0x1e0
   vfs_mkdir+0x14f/0x200
   do_mkdirat+0x6d/0x110
   do_syscall_64+0x2d/0x40
   entry_SYSCALL_64_after_hwframe+0x44/0xa9
  RIP: 0033:0x7f673196ae6b
  Code: 8b 05 11 (...)
  RSP: 002b:00007ffc3c679b18 EFLAGS: 00000246 ORIG_RAX: 0000000000000053
  RAX: ffffffffffffffda RBX: 00000000000001ff RCX: 00007f673196ae6b
  RDX: 0000000000000000 RSI: 00000000000001ff RDI: 00007ffc3c67a30d
  RBP: 00007ffc3c67a30d R08: 00000000000001ff R09: 0000000000000000
  R10: 000055d3e39fe930 R11: 0000000000000246 R12: 0000000000000000
  R13: 00007ffc3c679cd8 R14: 00007ffc3c67a30d R15: 00007ffc3c679ce0
  irq event stamp: 11029
  hardirqs last  enabled at (11037): [<ffffffff81153fe6>] console_unlock+0x486/0x670
  hardirqs last disabled at (11044): [<ffffffff81153c01>] console_unlock+0xa1/0x670
  softirqs last  enabled at (8864): [<ffffffff81e0102f>] asm_call_on_stack+0xf/0x20
  softirqs last disabled at (8851): [<ffffffff81e0102f>] asm_call_on_stack+0xf/0x20

This happens because at btrfs_mkdir() we call d_instantiate_new() while
holding a transaction handle, which results in the following call chain:

  btrfs_mkdir()
     trans = btrfs_start_transaction(root, 5);

     d_instantiate_new()
        smack_d_instantiate()
            __vfs_setxattr()
                btrfs_setxattr_trans()
                   btrfs_start_transaction()
                      start_transaction()
                         WARN_ON()
                           --> a tansaction start has TRANS_EXTWRITERS
                               set in its type
                         h->orig_rsv = h->block_rsv
                         h->block_rsv = NULL

     btrfs_end_transaction(trans)

Besides the warning triggered at start_transaction, we set the handle's
block_rsv to NULL which may cause some surprises later on.

So fix this by making btrfs_setxattr_trans() not start a transaction when
we already have a handle on one, stored in current->journal_info, and use
that handle. We are good to use the handle because at btrfs_mkdir() we did
reserve space for the xattr and the inode item.

Reported-by: Casey Schaufler <casey@schaufler-ca.com>
CC: stable@vger.kernel.org # 5.4+
Acked-by: Casey Schaufler <casey@schaufler-ca.com>
Tested-by: Casey Schaufler <casey@schaufler-ca.com>
Link: https://lore.kernel.org/linux-btrfs/434d856f-bd7b-4889-a6ec-e81aaebfa735@schaufler-ca.com/
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2021-03-09 11:09:37 +01:00

501 lines
13 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2007 Red Hat. All rights reserved.
*/
#include <linux/init.h>
#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/rwsem.h>
#include <linux/xattr.h>
#include <linux/security.h>
#include <linux/posix_acl_xattr.h>
#include <linux/iversion.h>
#include <linux/sched/mm.h>
#include "ctree.h"
#include "btrfs_inode.h"
#include "transaction.h"
#include "xattr.h"
#include "disk-io.h"
#include "props.h"
#include "locking.h"
int btrfs_getxattr(struct inode *inode, const char *name,
void *buffer, size_t size)
{
struct btrfs_dir_item *di;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_path *path;
struct extent_buffer *leaf;
int ret = 0;
unsigned long data_ptr;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
/* lookup the xattr by name */
di = btrfs_lookup_xattr(NULL, root, path, btrfs_ino(BTRFS_I(inode)),
name, strlen(name), 0);
if (!di) {
ret = -ENODATA;
goto out;
} else if (IS_ERR(di)) {
ret = PTR_ERR(di);
goto out;
}
leaf = path->nodes[0];
/* if size is 0, that means we want the size of the attr */
if (!size) {
ret = btrfs_dir_data_len(leaf, di);
goto out;
}
/* now get the data out of our dir_item */
if (btrfs_dir_data_len(leaf, di) > size) {
ret = -ERANGE;
goto out;
}
/*
* The way things are packed into the leaf is like this
* |struct btrfs_dir_item|name|data|
* where name is the xattr name, so security.foo, and data is the
* content of the xattr. data_ptr points to the location in memory
* where the data starts in the in memory leaf
*/
data_ptr = (unsigned long)((char *)(di + 1) +
btrfs_dir_name_len(leaf, di));
read_extent_buffer(leaf, buffer, data_ptr,
btrfs_dir_data_len(leaf, di));
ret = btrfs_dir_data_len(leaf, di);
out:
btrfs_free_path(path);
return ret;
}
int btrfs_setxattr(struct btrfs_trans_handle *trans, struct inode *inode,
const char *name, const void *value, size_t size, int flags)
{
struct btrfs_dir_item *di = NULL;
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_fs_info *fs_info = root->fs_info;
struct btrfs_path *path;
size_t name_len = strlen(name);
int ret = 0;
ASSERT(trans);
if (name_len + size > BTRFS_MAX_XATTR_SIZE(root->fs_info))
return -ENOSPC;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->skip_release_on_error = 1;
if (!value) {
di = btrfs_lookup_xattr(trans, root, path,
btrfs_ino(BTRFS_I(inode)), name, name_len, -1);
if (!di && (flags & XATTR_REPLACE))
ret = -ENODATA;
else if (IS_ERR(di))
ret = PTR_ERR(di);
else if (di)
ret = btrfs_delete_one_dir_name(trans, root, path, di);
goto out;
}
/*
* For a replace we can't just do the insert blindly.
* Do a lookup first (read-only btrfs_search_slot), and return if xattr
* doesn't exist. If it exists, fall down below to the insert/replace
* path - we can't race with a concurrent xattr delete, because the VFS
* locks the inode's i_mutex before calling setxattr or removexattr.
*/
if (flags & XATTR_REPLACE) {
ASSERT(inode_is_locked(inode));
di = btrfs_lookup_xattr(NULL, root, path,
btrfs_ino(BTRFS_I(inode)), name, name_len, 0);
if (!di)
ret = -ENODATA;
else if (IS_ERR(di))
ret = PTR_ERR(di);
if (ret)
goto out;
btrfs_release_path(path);
di = NULL;
}
ret = btrfs_insert_xattr_item(trans, root, path, btrfs_ino(BTRFS_I(inode)),
name, name_len, value, size);
if (ret == -EOVERFLOW) {
/*
* We have an existing item in a leaf, split_leaf couldn't
* expand it. That item might have or not a dir_item that
* matches our target xattr, so lets check.
*/
ret = 0;
btrfs_assert_tree_locked(path->nodes[0]);
di = btrfs_match_dir_item_name(fs_info, path, name, name_len);
if (!di && !(flags & XATTR_REPLACE)) {
ret = -ENOSPC;
goto out;
}
} else if (ret == -EEXIST) {
ret = 0;
di = btrfs_match_dir_item_name(fs_info, path, name, name_len);
ASSERT(di); /* logic error */
} else if (ret) {
goto out;
}
if (di && (flags & XATTR_CREATE)) {
ret = -EEXIST;
goto out;
}
if (di) {
/*
* We're doing a replace, and it must be atomic, that is, at
* any point in time we have either the old or the new xattr
* value in the tree. We don't want readers (getxattr and
* listxattrs) to miss a value, this is specially important
* for ACLs.
*/
const int slot = path->slots[0];
struct extent_buffer *leaf = path->nodes[0];
const u16 old_data_len = btrfs_dir_data_len(leaf, di);
const u32 item_size = btrfs_item_size_nr(leaf, slot);
const u32 data_size = sizeof(*di) + name_len + size;
struct btrfs_item *item;
unsigned long data_ptr;
char *ptr;
if (size > old_data_len) {
if (btrfs_leaf_free_space(leaf) <
(size - old_data_len)) {
ret = -ENOSPC;
goto out;
}
}
if (old_data_len + name_len + sizeof(*di) == item_size) {
/* No other xattrs packed in the same leaf item. */
if (size > old_data_len)
btrfs_extend_item(path, size - old_data_len);
else if (size < old_data_len)
btrfs_truncate_item(path, data_size, 1);
} else {
/* There are other xattrs packed in the same item. */
ret = btrfs_delete_one_dir_name(trans, root, path, di);
if (ret)
goto out;
btrfs_extend_item(path, data_size);
}
item = btrfs_item_nr(slot);
ptr = btrfs_item_ptr(leaf, slot, char);
ptr += btrfs_item_size(leaf, item) - data_size;
di = (struct btrfs_dir_item *)ptr;
btrfs_set_dir_data_len(leaf, di, size);
data_ptr = ((unsigned long)(di + 1)) + name_len;
write_extent_buffer(leaf, value, data_ptr, size);
btrfs_mark_buffer_dirty(leaf);
} else {
/*
* Insert, and we had space for the xattr, so path->slots[0] is
* where our xattr dir_item is and btrfs_insert_xattr_item()
* filled it.
*/
}
out:
btrfs_free_path(path);
if (!ret)
set_bit(BTRFS_INODE_COPY_EVERYTHING,
&BTRFS_I(inode)->runtime_flags);
return ret;
}
/*
* @value: "" makes the attribute to empty, NULL removes it
*/
int btrfs_setxattr_trans(struct inode *inode, const char *name,
const void *value, size_t size, int flags)
{
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_trans_handle *trans;
const bool start_trans = (current->journal_info == NULL);
int ret;
if (start_trans) {
/*
* 1 unit for inserting/updating/deleting the xattr
* 1 unit for the inode item update
*/
trans = btrfs_start_transaction(root, 2);
if (IS_ERR(trans))
return PTR_ERR(trans);
} else {
/*
* This can happen when smack is enabled and a directory is being
* created. It happens through d_instantiate_new(), which calls
* smack_d_instantiate(), which in turn calls __vfs_setxattr() to
* set the transmute xattr (XATTR_NAME_SMACKTRANSMUTE) on the
* inode. We have already reserved space for the xattr and inode
* update at btrfs_mkdir(), so just use the transaction handle.
* We don't join or start a transaction, as that will reset the
* block_rsv of the handle and trigger a warning for the start
* case.
*/
ASSERT(strncmp(name, XATTR_SECURITY_PREFIX,
XATTR_SECURITY_PREFIX_LEN) == 0);
trans = current->journal_info;
}
ret = btrfs_setxattr(trans, inode, name, value, size, flags);
if (ret)
goto out;
inode_inc_iversion(inode);
inode->i_ctime = current_time(inode);
ret = btrfs_update_inode(trans, root, inode);
BUG_ON(ret);
out:
if (start_trans)
btrfs_end_transaction(trans);
return ret;
}
ssize_t btrfs_listxattr(struct dentry *dentry, char *buffer, size_t size)
{
struct btrfs_key key;
struct inode *inode = d_inode(dentry);
struct btrfs_root *root = BTRFS_I(inode)->root;
struct btrfs_path *path;
int ret = 0;
size_t total_size = 0, size_left = size;
/*
* ok we want all objects associated with this id.
* NOTE: we set key.offset = 0; because we want to start with the
* first xattr that we find and walk forward
*/
key.objectid = btrfs_ino(BTRFS_I(inode));
key.type = BTRFS_XATTR_ITEM_KEY;
key.offset = 0;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->reada = READA_FORWARD;
/* search for our xattrs */
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto err;
while (1) {
struct extent_buffer *leaf;
int slot;
struct btrfs_dir_item *di;
struct btrfs_key found_key;
u32 item_size;
u32 cur;
leaf = path->nodes[0];
slot = path->slots[0];
/* this is where we start walking through the path */
if (slot >= btrfs_header_nritems(leaf)) {
/*
* if we've reached the last slot in this leaf we need
* to go to the next leaf and reset everything
*/
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto err;
else if (ret > 0)
break;
continue;
}
btrfs_item_key_to_cpu(leaf, &found_key, slot);
/* check to make sure this item is what we want */
if (found_key.objectid != key.objectid)
break;
if (found_key.type > BTRFS_XATTR_ITEM_KEY)
break;
if (found_key.type < BTRFS_XATTR_ITEM_KEY)
goto next_item;
di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
item_size = btrfs_item_size_nr(leaf, slot);
cur = 0;
while (cur < item_size) {
u16 name_len = btrfs_dir_name_len(leaf, di);
u16 data_len = btrfs_dir_data_len(leaf, di);
u32 this_len = sizeof(*di) + name_len + data_len;
unsigned long name_ptr = (unsigned long)(di + 1);
total_size += name_len + 1;
/*
* We are just looking for how big our buffer needs to
* be.
*/
if (!size)
goto next;
if (!buffer || (name_len + 1) > size_left) {
ret = -ERANGE;
goto err;
}
read_extent_buffer(leaf, buffer, name_ptr, name_len);
buffer[name_len] = '\0';
size_left -= name_len + 1;
buffer += name_len + 1;
next:
cur += this_len;
di = (struct btrfs_dir_item *)((char *)di + this_len);
}
next_item:
path->slots[0]++;
}
ret = total_size;
err:
btrfs_free_path(path);
return ret;
}
static int btrfs_xattr_handler_get(const struct xattr_handler *handler,
struct dentry *unused, struct inode *inode,
const char *name, void *buffer, size_t size)
{
name = xattr_full_name(handler, name);
return btrfs_getxattr(inode, name, buffer, size);
}
static int btrfs_xattr_handler_set(const struct xattr_handler *handler,
struct dentry *unused, struct inode *inode,
const char *name, const void *buffer,
size_t size, int flags)
{
name = xattr_full_name(handler, name);
return btrfs_setxattr_trans(inode, name, buffer, size, flags);
}
static int btrfs_xattr_handler_set_prop(const struct xattr_handler *handler,
struct dentry *unused, struct inode *inode,
const char *name, const void *value,
size_t size, int flags)
{
int ret;
struct btrfs_trans_handle *trans;
struct btrfs_root *root = BTRFS_I(inode)->root;
name = xattr_full_name(handler, name);
ret = btrfs_validate_prop(name, value, size);
if (ret)
return ret;
trans = btrfs_start_transaction(root, 2);
if (IS_ERR(trans))
return PTR_ERR(trans);
ret = btrfs_set_prop(trans, inode, name, value, size, flags);
if (!ret) {
inode_inc_iversion(inode);
inode->i_ctime = current_time(inode);
ret = btrfs_update_inode(trans, root, inode);
BUG_ON(ret);
}
btrfs_end_transaction(trans);
return ret;
}
static const struct xattr_handler btrfs_security_xattr_handler = {
.prefix = XATTR_SECURITY_PREFIX,
.get = btrfs_xattr_handler_get,
.set = btrfs_xattr_handler_set,
};
static const struct xattr_handler btrfs_trusted_xattr_handler = {
.prefix = XATTR_TRUSTED_PREFIX,
.get = btrfs_xattr_handler_get,
.set = btrfs_xattr_handler_set,
};
static const struct xattr_handler btrfs_user_xattr_handler = {
.prefix = XATTR_USER_PREFIX,
.get = btrfs_xattr_handler_get,
.set = btrfs_xattr_handler_set,
};
static const struct xattr_handler btrfs_btrfs_xattr_handler = {
.prefix = XATTR_BTRFS_PREFIX,
.get = btrfs_xattr_handler_get,
.set = btrfs_xattr_handler_set_prop,
};
const struct xattr_handler *btrfs_xattr_handlers[] = {
&btrfs_security_xattr_handler,
#ifdef CONFIG_BTRFS_FS_POSIX_ACL
&posix_acl_access_xattr_handler,
&posix_acl_default_xattr_handler,
#endif
&btrfs_trusted_xattr_handler,
&btrfs_user_xattr_handler,
&btrfs_btrfs_xattr_handler,
NULL,
};
static int btrfs_initxattrs(struct inode *inode,
const struct xattr *xattr_array, void *fs_private)
{
struct btrfs_trans_handle *trans = fs_private;
const struct xattr *xattr;
unsigned int nofs_flag;
char *name;
int err = 0;
/*
* We're holding a transaction handle, so use a NOFS memory allocation
* context to avoid deadlock if reclaim happens.
*/
nofs_flag = memalloc_nofs_save();
for (xattr = xattr_array; xattr->name != NULL; xattr++) {
name = kmalloc(XATTR_SECURITY_PREFIX_LEN +
strlen(xattr->name) + 1, GFP_KERNEL);
if (!name) {
err = -ENOMEM;
break;
}
strcpy(name, XATTR_SECURITY_PREFIX);
strcpy(name + XATTR_SECURITY_PREFIX_LEN, xattr->name);
err = btrfs_setxattr(trans, inode, name, xattr->value,
xattr->value_len, 0);
kfree(name);
if (err < 0)
break;
}
memalloc_nofs_restore(nofs_flag);
return err;
}
int btrfs_xattr_security_init(struct btrfs_trans_handle *trans,
struct inode *inode, struct inode *dir,
const struct qstr *qstr)
{
return security_inode_init_security(inode, dir, qstr,
&btrfs_initxattrs, trans);
}