2019-06-04 16:11:33 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
/*
|
|
|
|
*
|
|
|
|
* Copyright (C) 2011 Novell Inc.
|
|
|
|
*/
|
|
|
|
|
2015-07-07 22:04:44 +08:00
|
|
|
#include <linux/module.h>
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/file.h>
|
|
|
|
#include <linux/splice.h>
|
|
|
|
#include <linux/xattr.h>
|
|
|
|
#include <linux/security.h>
|
|
|
|
#include <linux/uaccess.h>
|
2017-02-03 02:15:33 +08:00
|
|
|
#include <linux/sched/signal.h>
|
2017-02-03 00:54:15 +08:00
|
|
|
#include <linux/cred.h>
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
#include <linux/namei.h>
|
2015-07-07 22:04:44 +08:00
|
|
|
#include <linux/fdtable.h>
|
|
|
|
#include <linux/ratelimit.h>
|
2017-03-30 20:22:16 +08:00
|
|
|
#include <linux/exportfs.h>
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
#include "overlayfs.h"
|
|
|
|
|
|
|
|
#define OVL_COPY_UP_CHUNK_SIZE (1 << 20)
|
|
|
|
|
2018-07-18 21:44:44 +08:00
|
|
|
static int ovl_ccup_set(const char *buf, const struct kernel_param *param)
|
2015-07-07 22:04:44 +08:00
|
|
|
{
|
2018-07-18 21:44:44 +08:00
|
|
|
pr_warn("overlayfs: \"check_copy_up\" module option is obsolete\n");
|
2015-07-07 22:04:44 +08:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-07-18 21:44:44 +08:00
|
|
|
static int ovl_ccup_get(char *buf, const struct kernel_param *param)
|
2015-07-07 22:04:44 +08:00
|
|
|
{
|
2018-07-18 21:44:44 +08:00
|
|
|
return sprintf(buf, "N\n");
|
2015-07-07 22:04:44 +08:00
|
|
|
}
|
|
|
|
|
2018-07-18 21:44:44 +08:00
|
|
|
module_param_call(check_copy_up, ovl_ccup_set, ovl_ccup_get, NULL, 0644);
|
2019-06-17 15:39:00 +08:00
|
|
|
MODULE_PARM_DESC(check_copy_up, "Obsolete; does nothing");
|
2018-07-18 21:44:44 +08:00
|
|
|
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
int ovl_copy_xattr(struct dentry *old, struct dentry *new)
|
|
|
|
{
|
2015-10-24 20:19:46 +08:00
|
|
|
ssize_t list_size, size, value_size = 0;
|
|
|
|
char *buf, *name, *value = NULL;
|
|
|
|
int uninitialized_var(error);
|
2016-09-16 20:12:11 +08:00
|
|
|
size_t slen;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
2016-09-29 23:48:42 +08:00
|
|
|
if (!(old->d_inode->i_opflags & IOP_XATTR) ||
|
|
|
|
!(new->d_inode->i_opflags & IOP_XATTR))
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
list_size = vfs_listxattr(old, NULL, 0);
|
|
|
|
if (list_size <= 0) {
|
|
|
|
if (list_size == -EOPNOTSUPP)
|
|
|
|
return 0;
|
|
|
|
return list_size;
|
|
|
|
}
|
|
|
|
|
|
|
|
buf = kzalloc(list_size, GFP_KERNEL);
|
|
|
|
if (!buf)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
list_size = vfs_listxattr(old, buf, list_size);
|
|
|
|
if (list_size <= 0) {
|
|
|
|
error = list_size;
|
2015-10-24 20:19:46 +08:00
|
|
|
goto out;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
}
|
|
|
|
|
2016-09-16 20:12:11 +08:00
|
|
|
for (name = buf; list_size; name += slen) {
|
|
|
|
slen = strnlen(name, list_size) + 1;
|
|
|
|
|
|
|
|
/* underlying fs providing us with an broken xattr list? */
|
|
|
|
if (WARN_ON(slen > list_size)) {
|
|
|
|
error = -EIO;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
list_size -= slen;
|
|
|
|
|
2016-08-08 21:08:49 +08:00
|
|
|
if (ovl_is_private_xattr(name))
|
|
|
|
continue;
|
2015-10-24 20:19:46 +08:00
|
|
|
retry:
|
|
|
|
size = vfs_getxattr(old, name, value, value_size);
|
|
|
|
if (size == -ERANGE)
|
|
|
|
size = vfs_getxattr(old, name, NULL, 0);
|
|
|
|
|
2015-11-11 00:08:41 +08:00
|
|
|
if (size < 0) {
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
error = size;
|
2015-10-24 20:19:46 +08:00
|
|
|
break;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
}
|
2015-10-24 20:19:46 +08:00
|
|
|
|
|
|
|
if (size > value_size) {
|
|
|
|
void *new;
|
|
|
|
|
|
|
|
new = krealloc(value, size, GFP_KERNEL);
|
|
|
|
if (!new) {
|
|
|
|
error = -ENOMEM;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
value = new;
|
|
|
|
value_size = size;
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
2016-07-13 22:44:49 +08:00
|
|
|
error = security_inode_copy_up_xattr(name);
|
|
|
|
if (error < 0 && error != -EOPNOTSUPP)
|
|
|
|
break;
|
|
|
|
if (error == 1) {
|
|
|
|
error = 0;
|
|
|
|
continue; /* Discard */
|
|
|
|
}
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
error = vfs_setxattr(new, name, value, size, 0);
|
|
|
|
if (error)
|
2015-10-24 20:19:46 +08:00
|
|
|
break;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
}
|
|
|
|
kfree(value);
|
|
|
|
out:
|
|
|
|
kfree(buf);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ovl_copy_up_data(struct path *old, struct path *new, loff_t len)
|
|
|
|
{
|
|
|
|
struct file *old_file;
|
|
|
|
struct file *new_file;
|
|
|
|
loff_t old_pos = 0;
|
|
|
|
loff_t new_pos = 0;
|
2018-10-30 07:41:49 +08:00
|
|
|
loff_t cloned;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
int error = 0;
|
|
|
|
|
|
|
|
if (len == 0)
|
|
|
|
return 0;
|
|
|
|
|
2015-09-18 18:45:12 +08:00
|
|
|
old_file = ovl_path_open(old, O_LARGEFILE | O_RDONLY);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
if (IS_ERR(old_file))
|
|
|
|
return PTR_ERR(old_file);
|
|
|
|
|
2015-09-18 18:45:12 +08:00
|
|
|
new_file = ovl_path_open(new, O_LARGEFILE | O_WRONLY);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
if (IS_ERR(new_file)) {
|
|
|
|
error = PTR_ERR(new_file);
|
|
|
|
goto out_fput;
|
|
|
|
}
|
|
|
|
|
2016-09-23 16:38:12 +08:00
|
|
|
/* Try to use clone_file_range to clone up within the same fs */
|
2018-10-30 07:41:56 +08:00
|
|
|
cloned = do_clone_file_range(old_file, 0, new_file, 0, len, 0);
|
2018-10-30 07:41:49 +08:00
|
|
|
if (cloned == len)
|
2016-09-23 16:38:12 +08:00
|
|
|
goto out;
|
|
|
|
/* Couldn't clone, so now we try to copy the data */
|
|
|
|
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
/* FIXME: copy up sparse files efficiently */
|
|
|
|
while (len) {
|
|
|
|
size_t this_len = OVL_COPY_UP_CHUNK_SIZE;
|
|
|
|
long bytes;
|
|
|
|
|
|
|
|
if (len < this_len)
|
|
|
|
this_len = len;
|
|
|
|
|
|
|
|
if (signal_pending_state(TASK_KILLABLE, current)) {
|
|
|
|
error = -EINTR;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
bytes = do_splice_direct(old_file, &old_pos,
|
|
|
|
new_file, &new_pos,
|
|
|
|
this_len, SPLICE_F_MOVE);
|
|
|
|
if (bytes <= 0) {
|
|
|
|
error = bytes;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
WARN_ON(old_pos != new_pos);
|
|
|
|
|
|
|
|
len -= bytes;
|
|
|
|
}
|
2016-09-23 16:38:12 +08:00
|
|
|
out:
|
2016-10-31 21:42:14 +08:00
|
|
|
if (!error)
|
|
|
|
error = vfs_fsync(new_file, 0);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
fput(new_file);
|
|
|
|
out_fput:
|
|
|
|
fput(old_file);
|
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
static int ovl_set_size(struct dentry *upperdentry, struct kstat *stat)
|
|
|
|
{
|
|
|
|
struct iattr attr = {
|
|
|
|
.ia_valid = ATTR_SIZE,
|
|
|
|
.ia_size = stat->size,
|
|
|
|
};
|
|
|
|
|
|
|
|
return notify_change(upperdentry, &attr, NULL);
|
|
|
|
}
|
|
|
|
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
static int ovl_set_timestamps(struct dentry *upperdentry, struct kstat *stat)
|
|
|
|
{
|
|
|
|
struct iattr attr = {
|
|
|
|
.ia_valid =
|
|
|
|
ATTR_ATIME | ATTR_MTIME | ATTR_ATIME_SET | ATTR_MTIME_SET,
|
|
|
|
.ia_atime = stat->atime,
|
|
|
|
.ia_mtime = stat->mtime,
|
|
|
|
};
|
|
|
|
|
|
|
|
return notify_change(upperdentry, &attr, NULL);
|
|
|
|
}
|
|
|
|
|
|
|
|
int ovl_set_attr(struct dentry *upperdentry, struct kstat *stat)
|
|
|
|
{
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
if (!S_ISLNK(stat->mode)) {
|
|
|
|
struct iattr attr = {
|
|
|
|
.ia_valid = ATTR_MODE,
|
|
|
|
.ia_mode = stat->mode,
|
|
|
|
};
|
|
|
|
err = notify_change(upperdentry, &attr, NULL);
|
|
|
|
}
|
|
|
|
if (!err) {
|
|
|
|
struct iattr attr = {
|
|
|
|
.ia_valid = ATTR_UID | ATTR_GID,
|
|
|
|
.ia_uid = stat->uid,
|
|
|
|
.ia_gid = stat->gid,
|
|
|
|
};
|
|
|
|
err = notify_change(upperdentry, &attr, NULL);
|
|
|
|
}
|
|
|
|
if (!err)
|
|
|
|
ovl_set_timestamps(upperdentry, stat);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2018-02-02 16:42:03 +08:00
|
|
|
struct ovl_fh *ovl_encode_real_fh(struct dentry *real, bool is_upper)
|
2017-03-30 20:22:16 +08:00
|
|
|
{
|
|
|
|
struct ovl_fh *fh;
|
|
|
|
int fh_type, fh_len, dwords;
|
|
|
|
void *buf;
|
|
|
|
int buflen = MAX_HANDLE_SZ;
|
2018-01-11 14:25:32 +08:00
|
|
|
uuid_t *uuid = &real->d_sb->s_uuid;
|
2017-03-30 20:22:16 +08:00
|
|
|
|
2017-09-14 07:28:29 +08:00
|
|
|
buf = kmalloc(buflen, GFP_KERNEL);
|
2017-03-30 20:22:16 +08:00
|
|
|
if (!buf)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We encode a non-connectable file handle for non-dir, because we
|
|
|
|
* only need to find the lower inode number and we don't want to pay
|
|
|
|
* the price or reconnecting the dentry.
|
|
|
|
*/
|
|
|
|
dwords = buflen >> 2;
|
2018-01-11 14:25:32 +08:00
|
|
|
fh_type = exportfs_encode_fh(real, buf, &dwords, 0);
|
2017-03-30 20:22:16 +08:00
|
|
|
buflen = (dwords << 2);
|
|
|
|
|
|
|
|
fh = ERR_PTR(-EIO);
|
|
|
|
if (WARN_ON(fh_type < 0) ||
|
|
|
|
WARN_ON(buflen > MAX_HANDLE_SZ) ||
|
|
|
|
WARN_ON(fh_type == FILEID_INVALID))
|
|
|
|
goto out;
|
|
|
|
|
2019-11-15 19:33:03 +08:00
|
|
|
/* Make sure the real fid stays 32bit aligned */
|
|
|
|
BUILD_BUG_ON(OVL_FH_FID_OFFSET % 4);
|
|
|
|
BUILD_BUG_ON(MAX_HANDLE_SZ + OVL_FH_FID_OFFSET > 255);
|
|
|
|
fh_len = OVL_FH_FID_OFFSET + buflen;
|
|
|
|
fh = kzalloc(fh_len, GFP_KERNEL);
|
2017-03-30 20:22:16 +08:00
|
|
|
if (!fh) {
|
|
|
|
fh = ERR_PTR(-ENOMEM);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2019-11-15 19:33:03 +08:00
|
|
|
fh->fb.version = OVL_FH_VERSION;
|
|
|
|
fh->fb.magic = OVL_FH_MAGIC;
|
|
|
|
fh->fb.type = fh_type;
|
|
|
|
fh->fb.flags = OVL_FH_FLAG_CPU_ENDIAN;
|
2017-06-21 20:28:38 +08:00
|
|
|
/*
|
|
|
|
* When we will want to decode an overlay dentry from this handle
|
|
|
|
* and all layers are on the same fs, if we get a disconncted real
|
|
|
|
* dentry when we decode fid, the only way to tell if we should assign
|
|
|
|
* it to upperdentry or to lowerstack is by checking this flag.
|
|
|
|
*/
|
|
|
|
if (is_upper)
|
2019-11-15 19:33:03 +08:00
|
|
|
fh->fb.flags |= OVL_FH_FLAG_PATH_UPPER;
|
|
|
|
fh->fb.len = fh_len - OVL_FH_WIRE_OFFSET;
|
|
|
|
fh->fb.uuid = *uuid;
|
|
|
|
memcpy(fh->fb.fid, buf, buflen);
|
2017-03-30 20:22:16 +08:00
|
|
|
|
|
|
|
out:
|
|
|
|
kfree(buf);
|
|
|
|
return fh;
|
|
|
|
}
|
|
|
|
|
2018-01-04 01:34:45 +08:00
|
|
|
int ovl_set_origin(struct dentry *dentry, struct dentry *lower,
|
|
|
|
struct dentry *upper)
|
2017-03-30 20:22:16 +08:00
|
|
|
{
|
|
|
|
const struct ovl_fh *fh = NULL;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* When lower layer doesn't support export operations store a 'null' fh,
|
|
|
|
* so we can use the overlay.origin xattr to distignuish between a copy
|
|
|
|
* up and a pure upper inode.
|
|
|
|
*/
|
2017-06-21 20:28:36 +08:00
|
|
|
if (ovl_can_decode_fh(lower->d_sb)) {
|
2018-02-02 16:42:03 +08:00
|
|
|
fh = ovl_encode_real_fh(lower, false);
|
2017-03-30 20:22:16 +08:00
|
|
|
if (IS_ERR(fh))
|
|
|
|
return PTR_ERR(fh);
|
|
|
|
}
|
|
|
|
|
2017-05-18 22:11:24 +08:00
|
|
|
/*
|
|
|
|
* Do not fail when upper doesn't support xattrs.
|
|
|
|
*/
|
2019-11-15 19:33:03 +08:00
|
|
|
err = ovl_check_setxattr(dentry, upper, OVL_XATTR_ORIGIN, fh->buf,
|
|
|
|
fh ? fh->fb.len : 0, 0);
|
2017-03-30 20:22:16 +08:00
|
|
|
kfree(fh);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2018-01-11 20:01:08 +08:00
|
|
|
/* Store file handle of @upper dir in @index dir entry */
|
|
|
|
static int ovl_set_upper_fh(struct dentry *upper, struct dentry *index)
|
|
|
|
{
|
|
|
|
const struct ovl_fh *fh;
|
|
|
|
int err;
|
|
|
|
|
2018-02-02 16:42:03 +08:00
|
|
|
fh = ovl_encode_real_fh(upper, true);
|
2018-01-11 20:01:08 +08:00
|
|
|
if (IS_ERR(fh))
|
|
|
|
return PTR_ERR(fh);
|
|
|
|
|
2019-11-15 19:33:03 +08:00
|
|
|
err = ovl_do_setxattr(index, OVL_XATTR_UPPER, fh->buf, fh->fb.len, 0);
|
2018-01-11 20:01:08 +08:00
|
|
|
|
|
|
|
kfree(fh);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Create and install index entry.
|
|
|
|
*
|
|
|
|
* Caller must hold i_mutex on indexdir.
|
|
|
|
*/
|
|
|
|
static int ovl_create_index(struct dentry *dentry, struct dentry *origin,
|
|
|
|
struct dentry *upper)
|
|
|
|
{
|
|
|
|
struct dentry *indexdir = ovl_indexdir(dentry->d_sb);
|
|
|
|
struct inode *dir = d_inode(indexdir);
|
|
|
|
struct dentry *index = NULL;
|
|
|
|
struct dentry *temp = NULL;
|
|
|
|
struct qstr name = { };
|
|
|
|
int err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For now this is only used for creating index entry for directories,
|
|
|
|
* because non-dir are copied up directly to index and then hardlinked
|
|
|
|
* to upper dir.
|
|
|
|
*
|
|
|
|
* TODO: implement create index for non-dir, so we can call it when
|
|
|
|
* encoding file handle for non-dir in case index does not exist.
|
|
|
|
*/
|
|
|
|
if (WARN_ON(!d_is_dir(dentry)))
|
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
/* Directory not expected to be indexed before copy up */
|
|
|
|
if (WARN_ON(ovl_test_flag(OVL_INDEX, d_inode(dentry))))
|
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
err = ovl_get_index_name(origin, &name);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2018-05-16 22:51:25 +08:00
|
|
|
temp = ovl_create_temp(indexdir, OVL_CATTR(S_IFDIR | 0));
|
2018-05-31 17:06:11 +08:00
|
|
|
err = PTR_ERR(temp);
|
2018-01-11 20:01:08 +08:00
|
|
|
if (IS_ERR(temp))
|
2018-05-31 17:06:11 +08:00
|
|
|
goto free_name;
|
2018-01-11 20:01:08 +08:00
|
|
|
|
|
|
|
err = ovl_set_upper_fh(upper, temp);
|
|
|
|
if (err)
|
2018-05-31 17:06:11 +08:00
|
|
|
goto out;
|
2018-01-11 20:01:08 +08:00
|
|
|
|
|
|
|
index = lookup_one_len(name.name, indexdir, name.len);
|
|
|
|
if (IS_ERR(index)) {
|
|
|
|
err = PTR_ERR(index);
|
|
|
|
} else {
|
|
|
|
err = ovl_do_rename(dir, temp, dir, index, 0);
|
|
|
|
dput(index);
|
|
|
|
}
|
|
|
|
out:
|
2018-05-31 17:06:11 +08:00
|
|
|
if (err)
|
|
|
|
ovl_cleanup(dir, temp);
|
2018-01-11 20:01:08 +08:00
|
|
|
dput(temp);
|
2018-05-31 17:06:11 +08:00
|
|
|
free_name:
|
2018-01-11 20:01:08 +08:00
|
|
|
kfree(name.name);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2017-07-05 03:04:06 +08:00
|
|
|
struct ovl_copy_up_ctx {
|
|
|
|
struct dentry *parent;
|
|
|
|
struct dentry *dentry;
|
|
|
|
struct path lowerpath;
|
|
|
|
struct kstat stat;
|
|
|
|
struct kstat pstat;
|
|
|
|
const char *link;
|
|
|
|
struct dentry *destdir;
|
|
|
|
struct qstr destname;
|
|
|
|
struct dentry *workdir;
|
|
|
|
bool origin;
|
2018-01-11 20:01:08 +08:00
|
|
|
bool indexed;
|
2018-05-11 23:49:27 +08:00
|
|
|
bool metacopy;
|
2017-07-05 03:04:06 +08:00
|
|
|
};
|
|
|
|
|
|
|
|
static int ovl_link_up(struct ovl_copy_up_ctx *c)
|
2017-06-20 20:25:46 +08:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
struct dentry *upper;
|
2017-07-05 03:04:06 +08:00
|
|
|
struct dentry *upperdir = ovl_dentry_upper(c->parent);
|
2017-06-20 20:25:46 +08:00
|
|
|
struct inode *udir = d_inode(upperdir);
|
|
|
|
|
2017-07-05 03:04:06 +08:00
|
|
|
/* Mark parent "impure" because it may now contain non-pure upper */
|
|
|
|
err = ovl_set_impure(c->parent, upperdir);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = ovl_set_nlink_lower(c->dentry);
|
ovl: persistent overlay inode nlink for indexed inodes
With inodes index enabled, an overlay inode nlink counts the union of upper
and non-covered lower hardlinks. During the lifetime of a non-pure upper
inode, the following nlink modifying operations can happen:
1. Lower hardlink copy up
2. Upper hardlink created, unlinked or renamed over
3. Lower hardlink whiteout or renamed over
For the first, copy up case, the union nlink does not change, whether the
operation succeeds or fails, but the upper inode nlink may change.
Therefore, before copy up, we store the union nlink value relative to the
lower inode nlink in the index inode xattr trusted.overlay.nlink.
For the second, upper hardlink case, the union nlink should be incremented
or decremented IFF the operation succeeds, aligned with nlink change of the
upper inode. Therefore, before link/unlink/rename, we store the union nlink
value relative to the upper inode nlink in the index inode.
For the last, lower cover up case, we simplify things by preceding the
whiteout or cover up with copy up. This makes sure that there is an index
upper inode where the nlink xattr can be stored before the copied up upper
entry is unlink.
Return the overlay inode nlinks for indexed upper inodes on stat(2).
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-06-20 20:35:14 +08:00
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2017-06-20 20:25:46 +08:00
|
|
|
inode_lock_nested(udir, I_MUTEX_PARENT);
|
2017-07-05 03:04:06 +08:00
|
|
|
upper = lookup_one_len(c->dentry->d_name.name, upperdir,
|
|
|
|
c->dentry->d_name.len);
|
2017-06-20 20:25:46 +08:00
|
|
|
err = PTR_ERR(upper);
|
|
|
|
if (!IS_ERR(upper)) {
|
2018-05-16 22:04:00 +08:00
|
|
|
err = ovl_do_link(ovl_dentry_upper(c->dentry), udir, upper);
|
2017-06-20 20:25:46 +08:00
|
|
|
dput(upper);
|
|
|
|
|
2017-07-05 03:04:06 +08:00
|
|
|
if (!err) {
|
|
|
|
/* Restore timestamps on parent (best effort) */
|
|
|
|
ovl_set_timestamps(upperdir, &c->pstat);
|
|
|
|
ovl_dentry_set_upper_alias(c->dentry);
|
|
|
|
}
|
2017-06-20 20:25:46 +08:00
|
|
|
}
|
|
|
|
inode_unlock(udir);
|
2017-10-15 23:00:20 +08:00
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
err = ovl_set_nlink_upper(c->dentry);
|
2017-06-20 20:25:46 +08:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2017-07-05 04:03:18 +08:00
|
|
|
static int ovl_copy_up_inode(struct ovl_copy_up_ctx *c, struct dentry *temp)
|
2017-07-05 04:03:18 +08:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
2019-01-12 02:37:00 +08:00
|
|
|
/*
|
|
|
|
* Copy up data first and then xattrs. Writing data after
|
|
|
|
* xattrs will remove security.capability xattr automatically.
|
|
|
|
*/
|
|
|
|
if (S_ISREG(c->stat.mode) && !c->metacopy) {
|
|
|
|
struct path upperpath, datapath;
|
|
|
|
|
|
|
|
ovl_path_upper(c->dentry, &upperpath);
|
|
|
|
if (WARN_ON(upperpath.dentry != NULL))
|
|
|
|
return -EIO;
|
|
|
|
upperpath.dentry = temp;
|
|
|
|
|
|
|
|
ovl_path_lowerdata(c->dentry, &datapath);
|
|
|
|
err = ovl_copy_up_data(&datapath, &upperpath, c->stat.size);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2017-07-05 04:03:18 +08:00
|
|
|
err = ovl_copy_xattr(c->lowerpath.dentry, temp);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
if (err)
|
2017-05-19 20:16:21 +08:00
|
|
|
return err;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
2017-03-30 20:22:16 +08:00
|
|
|
/*
|
|
|
|
* Store identifier of lower inode in upper inode xattr to
|
|
|
|
* allow lookup of the copy up origin inode.
|
2017-06-28 19:41:22 +08:00
|
|
|
*
|
|
|
|
* Don't set origin when we are breaking the association with a lower
|
|
|
|
* hard link.
|
2017-03-30 20:22:16 +08:00
|
|
|
*/
|
2017-06-20 20:25:46 +08:00
|
|
|
if (c->origin) {
|
2017-07-05 04:03:18 +08:00
|
|
|
err = ovl_set_origin(c->dentry, c->lowerpath.dentry, temp);
|
2017-06-28 19:41:22 +08:00
|
|
|
if (err)
|
2017-05-19 20:16:21 +08:00
|
|
|
return err;
|
2017-06-28 19:41:22 +08:00
|
|
|
}
|
2017-03-30 20:22:16 +08:00
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
if (c->metacopy) {
|
|
|
|
err = ovl_check_setxattr(c->dentry, temp, OVL_XATTR_METACOPY,
|
|
|
|
NULL, 0, -EOPNOTSUPP);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:27 +08:00
|
|
|
inode_lock(temp->d_inode);
|
2018-05-11 23:49:28 +08:00
|
|
|
if (c->metacopy)
|
|
|
|
err = ovl_set_size(temp, &c->stat);
|
|
|
|
if (!err)
|
|
|
|
err = ovl_set_attr(temp, &c->stat);
|
2018-05-11 23:49:27 +08:00
|
|
|
inode_unlock(temp->d_inode);
|
|
|
|
|
|
|
|
return err;
|
2017-05-19 20:16:21 +08:00
|
|
|
}
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
struct ovl_cu_creds {
|
|
|
|
const struct cred *old;
|
|
|
|
struct cred *new;
|
|
|
|
};
|
|
|
|
|
|
|
|
static int ovl_prep_cu_creds(struct dentry *dentry, struct ovl_cu_creds *cc)
|
2018-10-08 12:25:23 +08:00
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
cc->old = cc->new = NULL;
|
|
|
|
err = security_inode_copy_up(dentry, &cc->new);
|
2018-10-08 12:25:23 +08:00
|
|
|
if (err < 0)
|
2018-10-27 05:34:39 +08:00
|
|
|
return err;
|
2018-10-08 12:25:23 +08:00
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
if (cc->new)
|
|
|
|
cc->old = override_creds(cc->new);
|
2018-10-08 12:25:23 +08:00
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
return 0;
|
2018-10-08 12:25:23 +08:00
|
|
|
}
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
static void ovl_revert_cu_creds(struct ovl_cu_creds *cc)
|
2018-10-08 12:25:23 +08:00
|
|
|
{
|
2018-10-27 05:34:39 +08:00
|
|
|
if (cc->new) {
|
|
|
|
revert_creds(cc->old);
|
|
|
|
put_cred(cc->new);
|
|
|
|
}
|
2018-10-08 12:25:23 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copyup using workdir to prepare temp file. Used when copying up directories,
|
|
|
|
* special files or when upper fs doesn't support O_TMPFILE.
|
|
|
|
*/
|
|
|
|
static int ovl_copy_up_workdir(struct ovl_copy_up_ctx *c)
|
2017-05-19 20:16:21 +08:00
|
|
|
{
|
2017-06-25 21:37:17 +08:00
|
|
|
struct inode *inode;
|
2018-10-27 05:34:39 +08:00
|
|
|
struct inode *udir = d_inode(c->destdir), *wdir = d_inode(c->workdir);
|
|
|
|
struct dentry *temp, *upper;
|
|
|
|
struct ovl_cu_creds cc;
|
2017-05-19 20:16:21 +08:00
|
|
|
int err;
|
2018-10-27 05:34:39 +08:00
|
|
|
struct ovl_cattr cattr = {
|
|
|
|
/* Can't properly set mode on creation because of the umask */
|
|
|
|
.mode = c->stat.mode & S_IFMT,
|
|
|
|
.rdev = c->stat.rdev,
|
|
|
|
.link = c->link
|
|
|
|
};
|
2017-05-19 20:16:21 +08:00
|
|
|
|
2018-10-08 12:25:23 +08:00
|
|
|
err = ovl_lock_rename_workdir(c->workdir, c->destdir);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
err = ovl_prep_cu_creds(c->dentry, &cc);
|
|
|
|
if (err)
|
|
|
|
goto unlock;
|
|
|
|
|
|
|
|
temp = ovl_create_temp(c->workdir, &cattr);
|
|
|
|
ovl_revert_cu_creds(&cc);
|
|
|
|
|
2018-10-08 12:25:23 +08:00
|
|
|
err = PTR_ERR(temp);
|
2018-05-31 17:06:11 +08:00
|
|
|
if (IS_ERR(temp))
|
2018-10-08 12:25:23 +08:00
|
|
|
goto unlock;
|
2017-05-19 20:16:21 +08:00
|
|
|
|
2017-07-05 04:03:18 +08:00
|
|
|
err = ovl_copy_up_inode(c, temp);
|
2017-05-19 20:16:21 +08:00
|
|
|
if (err)
|
2018-10-08 12:25:23 +08:00
|
|
|
goto cleanup;
|
2017-05-19 20:16:21 +08:00
|
|
|
|
2018-01-11 20:01:08 +08:00
|
|
|
if (S_ISDIR(c->stat.mode) && c->indexed) {
|
|
|
|
err = ovl_create_index(c->dentry, c->lowerpath.dentry, temp);
|
|
|
|
if (err)
|
2018-10-08 12:25:23 +08:00
|
|
|
goto cleanup;
|
2018-01-11 20:01:08 +08:00
|
|
|
}
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
upper = lookup_one_len(c->destname.name, c->destdir, c->destname.len);
|
|
|
|
err = PTR_ERR(upper);
|
|
|
|
if (IS_ERR(upper))
|
|
|
|
goto cleanup;
|
|
|
|
|
|
|
|
err = ovl_do_rename(wdir, temp, udir, upper, 0);
|
|
|
|
dput(upper);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
if (err)
|
2018-10-08 12:25:23 +08:00
|
|
|
goto cleanup;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
if (!c->metacopy)
|
|
|
|
ovl_set_upperdata(d_inode(c->dentry));
|
2017-06-25 21:37:17 +08:00
|
|
|
inode = d_inode(c->dentry);
|
2018-10-27 05:34:39 +08:00
|
|
|
ovl_inode_update(inode, temp);
|
2017-06-25 21:37:17 +08:00
|
|
|
if (S_ISDIR(inode->i_mode))
|
|
|
|
ovl_set_flag(OVL_WHITEOUTS, inode);
|
2018-10-08 12:25:23 +08:00
|
|
|
unlock:
|
|
|
|
unlock_rename(c->workdir, c->destdir);
|
|
|
|
|
|
|
|
return err;
|
|
|
|
|
|
|
|
cleanup:
|
2018-10-27 05:34:39 +08:00
|
|
|
ovl_cleanup(wdir, temp);
|
|
|
|
dput(temp);
|
|
|
|
goto unlock;
|
2018-10-08 12:25:23 +08:00
|
|
|
}
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
/* Copyup using O_TMPFILE which does not require cross dir locking */
|
|
|
|
static int ovl_copy_up_tmpfile(struct ovl_copy_up_ctx *c)
|
2018-10-08 12:25:23 +08:00
|
|
|
{
|
2018-10-27 05:34:39 +08:00
|
|
|
struct inode *udir = d_inode(c->destdir);
|
|
|
|
struct dentry *temp, *upper;
|
|
|
|
struct ovl_cu_creds cc;
|
2018-10-08 12:25:23 +08:00
|
|
|
int err;
|
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
err = ovl_prep_cu_creds(c->dentry, &cc);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2018-10-08 12:25:23 +08:00
|
|
|
|
|
|
|
temp = ovl_do_tmpfile(c->workdir, c->stat.mode);
|
2018-10-27 05:34:39 +08:00
|
|
|
ovl_revert_cu_creds(&cc);
|
2018-10-08 12:25:23 +08:00
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
if (IS_ERR(temp))
|
|
|
|
return PTR_ERR(temp);
|
2018-10-08 12:25:23 +08:00
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
err = ovl_copy_up_inode(c, temp);
|
|
|
|
if (err)
|
|
|
|
goto out_dput;
|
2018-10-08 12:25:23 +08:00
|
|
|
|
|
|
|
inode_lock_nested(udir, I_MUTEX_PARENT);
|
|
|
|
|
|
|
|
upper = lookup_one_len(c->destname.name, c->destdir, c->destname.len);
|
|
|
|
err = PTR_ERR(upper);
|
2018-10-27 05:34:39 +08:00
|
|
|
if (!IS_ERR(upper)) {
|
|
|
|
err = ovl_do_link(temp, udir, upper);
|
|
|
|
dput(upper);
|
|
|
|
}
|
2018-10-08 12:25:23 +08:00
|
|
|
inode_unlock(udir);
|
|
|
|
|
|
|
|
if (err)
|
2018-10-27 05:34:39 +08:00
|
|
|
goto out_dput;
|
2018-10-08 12:25:23 +08:00
|
|
|
|
|
|
|
if (!c->metacopy)
|
|
|
|
ovl_set_upperdata(d_inode(c->dentry));
|
2018-10-27 05:34:39 +08:00
|
|
|
ovl_inode_update(d_inode(c->dentry), temp);
|
2017-06-25 21:37:17 +08:00
|
|
|
|
2018-10-27 05:34:39 +08:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
out_dput:
|
2017-01-17 12:34:54 +08:00
|
|
|
dput(temp);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Copy up a single dentry
|
|
|
|
*
|
2016-12-16 18:02:56 +08:00
|
|
|
* All renames start with copy up of source if necessary. The actual
|
|
|
|
* rename will only proceed once the copy up was successful. Copy up uses
|
|
|
|
* upper parent i_mutex for exclusion. Since rename can change d_parent it
|
|
|
|
* is possible that the copy up will lock the old parent. At that point
|
|
|
|
* the file will have already been copied up anyway.
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
*/
|
2017-07-05 04:03:18 +08:00
|
|
|
static int ovl_do_copy_up(struct ovl_copy_up_ctx *c)
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
{
|
|
|
|
int err;
|
2017-07-05 04:03:18 +08:00
|
|
|
struct ovl_fs *ofs = c->dentry->d_sb->s_fs_info;
|
2018-01-11 20:01:08 +08:00
|
|
|
bool to_index = false;
|
2017-06-20 20:25:46 +08:00
|
|
|
|
2018-01-11 20:01:08 +08:00
|
|
|
/*
|
|
|
|
* Indexed non-dir is copied up directly to the index entry and then
|
|
|
|
* hardlinked to upper dir. Indexed dir is copied up to indexdir,
|
|
|
|
* then index entry is created and then copied up dir installed.
|
|
|
|
* Copying dir up to indexdir instead of workdir simplifies locking.
|
|
|
|
*/
|
|
|
|
if (ovl_need_index(c->dentry)) {
|
|
|
|
c->indexed = true;
|
|
|
|
if (S_ISDIR(c->stat.mode))
|
|
|
|
c->workdir = ovl_indexdir(c->dentry->d_sb);
|
|
|
|
else
|
|
|
|
to_index = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (S_ISDIR(c->stat.mode) || c->stat.nlink == 1 || to_index)
|
2017-06-20 20:25:46 +08:00
|
|
|
c->origin = true;
|
|
|
|
|
2018-01-11 20:01:08 +08:00
|
|
|
if (to_index) {
|
2017-06-20 20:25:46 +08:00
|
|
|
c->destdir = ovl_indexdir(c->dentry->d_sb);
|
|
|
|
err = ovl_get_index_name(c->lowerpath.dentry, &c->destname);
|
|
|
|
if (err)
|
|
|
|
return err;
|
2017-10-15 23:00:20 +08:00
|
|
|
} else if (WARN_ON(!c->parent)) {
|
|
|
|
/* Disconnected dentry must be copied up to index dir */
|
|
|
|
return -EIO;
|
2017-06-20 20:25:46 +08:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* Mark parent "impure" because it may now contain non-pure
|
|
|
|
* upper
|
|
|
|
*/
|
|
|
|
err = ovl_set_impure(c->parent, c->destdir);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
2017-01-17 12:34:57 +08:00
|
|
|
/* Should we copyup with O_TMPFILE or with workdir? */
|
2018-10-08 12:25:23 +08:00
|
|
|
if (S_ISREG(c->stat.mode) && ofs->tmpfile)
|
|
|
|
err = ovl_copy_up_tmpfile(c);
|
|
|
|
else
|
|
|
|
err = ovl_copy_up_workdir(c);
|
2017-10-15 23:00:20 +08:00
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (c->indexed)
|
2018-01-11 20:01:08 +08:00
|
|
|
ovl_set_flag(OVL_INDEX, d_inode(c->dentry));
|
|
|
|
|
|
|
|
if (to_index) {
|
2017-10-15 23:00:20 +08:00
|
|
|
/* Initialize nlink for copy up of disconnected dentry */
|
|
|
|
err = ovl_set_nlink_upper(c->dentry);
|
|
|
|
} else {
|
2017-06-20 20:25:46 +08:00
|
|
|
struct inode *udir = d_inode(c->destdir);
|
|
|
|
|
|
|
|
/* Restore timestamps on parent (best effort) */
|
|
|
|
inode_lock(udir);
|
|
|
|
ovl_set_timestamps(c->destdir, &c->pstat);
|
|
|
|
inode_unlock(udir);
|
|
|
|
|
|
|
|
ovl_dentry_set_upper_alias(c->dentry);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
}
|
|
|
|
|
2017-10-15 23:00:20 +08:00
|
|
|
out:
|
|
|
|
if (to_index)
|
|
|
|
kfree(c->destname.name);
|
2017-07-05 04:03:18 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:27 +08:00
|
|
|
static bool ovl_need_meta_copy_up(struct dentry *dentry, umode_t mode,
|
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
|
|
|
|
|
|
|
|
if (!ofs->config.metacopy)
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (!S_ISREG(mode))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
if (flags && ((OPEN_FMODE(flags) & FMODE_WRITE) || (flags & O_TRUNC)))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
/* Copy up data of an inode which was copied up metadata only in the past. */
|
|
|
|
static int ovl_copy_up_meta_inode_data(struct ovl_copy_up_ctx *c)
|
|
|
|
{
|
2018-05-11 23:49:30 +08:00
|
|
|
struct path upperpath, datapath;
|
2018-05-11 23:49:28 +08:00
|
|
|
int err;
|
2019-01-31 03:01:57 +08:00
|
|
|
char *capability = NULL;
|
|
|
|
ssize_t uninitialized_var(cap_size);
|
2018-05-11 23:49:28 +08:00
|
|
|
|
|
|
|
ovl_path_upper(c->dentry, &upperpath);
|
|
|
|
if (WARN_ON(upperpath.dentry == NULL))
|
|
|
|
return -EIO;
|
|
|
|
|
2018-05-11 23:49:30 +08:00
|
|
|
ovl_path_lowerdata(c->dentry, &datapath);
|
|
|
|
if (WARN_ON(datapath.dentry == NULL))
|
|
|
|
return -EIO;
|
|
|
|
|
2019-01-31 03:01:57 +08:00
|
|
|
if (c->stat.size) {
|
|
|
|
err = cap_size = ovl_getxattr(upperpath.dentry, XATTR_NAME_CAPS,
|
|
|
|
&capability, 0);
|
|
|
|
if (err < 0 && err != -ENODATA)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:30 +08:00
|
|
|
err = ovl_copy_up_data(&datapath, &upperpath, c->stat.size);
|
2018-05-11 23:49:28 +08:00
|
|
|
if (err)
|
2019-01-31 03:01:57 +08:00
|
|
|
goto out_free;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Writing to upper file will clear security.capability xattr. We
|
|
|
|
* don't want that to happen for normal copy-up operation.
|
|
|
|
*/
|
|
|
|
if (capability) {
|
|
|
|
err = ovl_do_setxattr(upperpath.dentry, XATTR_NAME_CAPS,
|
|
|
|
capability, cap_size, 0);
|
|
|
|
if (err)
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
|
|
|
|
err = vfs_removexattr(upperpath.dentry, OVL_XATTR_METACOPY);
|
|
|
|
if (err)
|
2019-01-31 03:01:57 +08:00
|
|
|
goto out_free;
|
2018-05-11 23:49:28 +08:00
|
|
|
|
|
|
|
ovl_set_upperdata(d_inode(c->dentry));
|
2019-01-31 03:01:57 +08:00
|
|
|
out_free:
|
|
|
|
kfree(capability);
|
|
|
|
out:
|
2018-05-11 23:49:28 +08:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2017-07-05 04:03:18 +08:00
|
|
|
static int ovl_copy_up_one(struct dentry *parent, struct dentry *dentry,
|
|
|
|
int flags)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
DEFINE_DELAYED_CALL(done);
|
|
|
|
struct path parentpath;
|
|
|
|
struct ovl_copy_up_ctx ctx = {
|
|
|
|
.parent = parent,
|
|
|
|
.dentry = dentry,
|
|
|
|
.workdir = ovl_workdir(dentry),
|
|
|
|
};
|
|
|
|
|
|
|
|
if (WARN_ON(!ctx.workdir))
|
|
|
|
return -EROFS;
|
|
|
|
|
|
|
|
ovl_path_lower(dentry, &ctx.lowerpath);
|
|
|
|
err = vfs_getattr(&ctx.lowerpath, &ctx.stat,
|
|
|
|
STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2018-05-11 23:49:27 +08:00
|
|
|
ctx.metacopy = ovl_need_meta_copy_up(dentry, ctx.stat.mode, flags);
|
|
|
|
|
2017-10-15 23:00:20 +08:00
|
|
|
if (parent) {
|
|
|
|
ovl_path_upper(parent, &parentpath);
|
|
|
|
ctx.destdir = parentpath.dentry;
|
|
|
|
ctx.destname = dentry->d_name;
|
2017-07-05 04:03:18 +08:00
|
|
|
|
2017-10-15 23:00:20 +08:00
|
|
|
err = vfs_getattr(&parentpath, &ctx.pstat,
|
|
|
|
STATX_ATIME | STATX_MTIME,
|
|
|
|
AT_STATX_SYNC_AS_STAT);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
}
|
2017-07-05 04:03:18 +08:00
|
|
|
|
|
|
|
/* maybe truncate regular file. this has no effect on dirs */
|
|
|
|
if (flags & O_TRUNC)
|
|
|
|
ctx.stat.size = 0;
|
|
|
|
|
|
|
|
if (S_ISLNK(ctx.stat.mode)) {
|
|
|
|
ctx.link = vfs_get_link(ctx.lowerpath.dentry, &done);
|
|
|
|
if (IS_ERR(ctx.link))
|
|
|
|
return PTR_ERR(ctx.link);
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
err = ovl_copy_up_start(dentry, flags);
|
2017-07-05 04:03:18 +08:00
|
|
|
/* err < 0: interrupted, err > 0: raced with another copy-up */
|
|
|
|
if (unlikely(err)) {
|
|
|
|
if (err > 0)
|
|
|
|
err = 0;
|
|
|
|
} else {
|
2017-06-20 20:25:46 +08:00
|
|
|
if (!ovl_dentry_upper(dentry))
|
|
|
|
err = ovl_do_copy_up(&ctx);
|
2017-10-15 23:00:20 +08:00
|
|
|
if (!err && parent && !ovl_dentry_has_upper_alias(dentry))
|
2017-07-05 03:04:06 +08:00
|
|
|
err = ovl_link_up(&ctx);
|
2018-05-11 23:49:28 +08:00
|
|
|
if (!err && ovl_dentry_needs_data_copy_up_locked(dentry, flags))
|
|
|
|
err = ovl_copy_up_meta_inode_data(&ctx);
|
2017-07-05 04:03:18 +08:00
|
|
|
ovl_copy_up_end(dentry);
|
|
|
|
}
|
2016-10-04 20:40:45 +08:00
|
|
|
do_delayed_call(&done);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2016-11-13 03:36:03 +08:00
|
|
|
int ovl_copy_up_flags(struct dentry *dentry, int flags)
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
{
|
2016-09-07 01:40:32 +08:00
|
|
|
int err = 0;
|
|
|
|
const struct cred *old_cred = ovl_override_creds(dentry->d_sb);
|
2017-10-15 23:00:20 +08:00
|
|
|
bool disconnected = (dentry->d_flags & DCACHE_DISCONNECTED);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* With NFS export, copy up can get called for a disconnected non-dir.
|
|
|
|
* In this case, we will copy up lower inode to index dir without
|
|
|
|
* linking it to upper dir.
|
|
|
|
*/
|
|
|
|
if (WARN_ON(disconnected && d_is_dir(dentry)))
|
|
|
|
return -EIO;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
|
|
|
while (!err) {
|
|
|
|
struct dentry *next;
|
2017-10-15 23:00:20 +08:00
|
|
|
struct dentry *parent = NULL;
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
if (ovl_already_copied_up(dentry, flags))
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
break;
|
|
|
|
|
|
|
|
next = dget(dentry);
|
|
|
|
/* find the topmost dentry not yet copied up */
|
2017-10-15 23:00:20 +08:00
|
|
|
for (; !disconnected;) {
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
parent = dget_parent(next);
|
|
|
|
|
2017-06-20 20:25:46 +08:00
|
|
|
if (ovl_dentry_upper(parent))
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
break;
|
|
|
|
|
|
|
|
dput(next);
|
|
|
|
next = parent;
|
|
|
|
}
|
|
|
|
|
2017-07-05 04:03:18 +08:00
|
|
|
err = ovl_copy_up_one(parent, next, flags);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
|
|
|
dput(parent);
|
|
|
|
dput(next);
|
|
|
|
}
|
2016-09-07 01:40:32 +08:00
|
|
|
revert_creds(old_cred);
|
overlay filesystem
Overlayfs allows one, usually read-write, directory tree to be
overlaid onto another, read-only directory tree. All modifications
go to the upper, writable layer.
This type of mechanism is most often used for live CDs but there's a
wide variety of other uses.
The implementation differs from other "union filesystem"
implementations in that after a file is opened all operations go
directly to the underlying, lower or upper, filesystems. This
simplifies the implementation and allows native performance in these
cases.
The dentry tree is duplicated from the underlying filesystems, this
enables fast cached lookups without adding special support into the
VFS. This uses slightly more memory than union mounts, but dentries
are relatively small.
Currently inodes are duplicated as well, but it is a possible
optimization to share inodes for non-directories.
Opening non directories results in the open forwarded to the
underlying filesystem. This makes the behavior very similar to union
mounts (with the same limitations vs. fchmod/fchown on O_RDONLY file
descriptors).
Usage:
mount -t overlayfs overlayfs -olowerdir=/lower,upperdir=/upper/upper,workdir=/upper/work /overlay
The following cotributions have been folded into this patch:
Neil Brown <neilb@suse.de>:
- minimal remount support
- use correct seek function for directories
- initialise is_real before use
- rename ovl_fill_cache to ovl_dir_read
Felix Fietkau <nbd@openwrt.org>:
- fix a deadlock in ovl_dir_read_merged
- fix a deadlock in ovl_remove_whiteouts
Erez Zadok <ezk@fsl.cs.sunysb.edu>
- fix cleanup after WARN_ON
Sedat Dilek <sedat.dilek@googlemail.com>
- fix up permission to confirm to new API
Robin Dong <hao.bigrat@gmail.com>
- fix possible leak in ovl_new_inode
- create new inode in ovl_link
Andy Whitcroft <apw@canonical.com>
- switch to __inode_permission()
- copy up i_uid/i_gid from the underlying inode
AV:
- ovl_copy_up_locked() - dput(ERR_PTR(...)) on two failure exits
- ovl_clear_empty() - one failure exit forgetting to do unlock_rename(),
lack of check for udir being the parent of upper, dropping and regaining
the lock on udir (which would require _another_ check for parent being
right).
- bogus d_drop() in copyup and rename [fix from your mail]
- copyup/remove and copyup/rename races [fix from your mail]
- ovl_dir_fsync() leaving ERR_PTR() in ->realfile
- ovl_entry_free() is pointless - it's just a kfree_rcu()
- fold ovl_do_lookup() into ovl_lookup()
- manually assigning ->d_op is wrong. Just use ->s_d_op.
[patches picked from Miklos]:
* copyup/remove and copyup/rename races
* bogus d_drop() in copyup and rename
Also thanks to the following people for testing and reporting bugs:
Jordi Pujol <jordipujolp@gmail.com>
Andy Whitcroft <apw@canonical.com>
Michal Suchanek <hramrach@centrum.cz>
Felix Fietkau <nbd@openwrt.org>
Erez Zadok <ezk@fsl.cs.sunysb.edu>
Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
2014-10-24 06:14:38 +08:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
2016-11-13 03:36:03 +08:00
|
|
|
|
2018-05-11 23:49:27 +08:00
|
|
|
static bool ovl_open_need_copy_up(struct dentry *dentry, int flags)
|
|
|
|
{
|
|
|
|
/* Copy up of disconnected dentry does not set upper alias */
|
2018-05-11 23:49:28 +08:00
|
|
|
if (ovl_already_copied_up(dentry, flags))
|
2018-05-11 23:49:27 +08:00
|
|
|
return false;
|
|
|
|
|
|
|
|
if (special_file(d_inode(dentry)->i_mode))
|
|
|
|
return false;
|
|
|
|
|
2018-05-11 23:49:28 +08:00
|
|
|
if (!ovl_open_flags_need_copy_up(flags))
|
2018-05-11 23:49:27 +08:00
|
|
|
return false;
|
|
|
|
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2019-01-22 13:01:39 +08:00
|
|
|
int ovl_maybe_copy_up(struct dentry *dentry, int flags)
|
2018-05-11 23:49:27 +08:00
|
|
|
{
|
|
|
|
int err = 0;
|
|
|
|
|
2019-01-22 13:01:39 +08:00
|
|
|
if (ovl_open_need_copy_up(dentry, flags)) {
|
2018-05-11 23:49:27 +08:00
|
|
|
err = ovl_want_write(dentry);
|
|
|
|
if (!err) {
|
2019-01-22 13:01:39 +08:00
|
|
|
err = ovl_copy_up_flags(dentry, flags);
|
2018-05-11 23:49:27 +08:00
|
|
|
ovl_drop_write(dentry);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2018-05-11 23:49:33 +08:00
|
|
|
int ovl_copy_up_with_data(struct dentry *dentry)
|
|
|
|
{
|
|
|
|
return ovl_copy_up_flags(dentry, O_WRONLY);
|
|
|
|
}
|
|
|
|
|
2016-11-13 03:36:03 +08:00
|
|
|
int ovl_copy_up(struct dentry *dentry)
|
|
|
|
{
|
|
|
|
return ovl_copy_up_flags(dentry, 0);
|
|
|
|
}
|