2019-06-04 15:11:33 +07:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
2017-07-12 18:17:16 +07:00
|
|
|
/*
|
|
|
|
* Overlayfs NFS export support.
|
|
|
|
*
|
|
|
|
* Amir Goldstein <amir73il@gmail.com>
|
|
|
|
*
|
|
|
|
* Copyright (C) 2017-2018 CTERA Networks. All Rights Reserved.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/cred.h>
|
|
|
|
#include <linux/mount.h>
|
|
|
|
#include <linux/namei.h>
|
|
|
|
#include <linux/xattr.h>
|
|
|
|
#include <linux/exportfs.h>
|
|
|
|
#include <linux/ratelimit.h>
|
|
|
|
#include "overlayfs.h"
|
|
|
|
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
static int ovl_encode_maybe_copy_up(struct dentry *dentry)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (ovl_dentry_upper(dentry))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
err = ovl_want_write(dentry);
|
|
|
|
if (!err) {
|
|
|
|
err = ovl_copy_up(dentry);
|
|
|
|
ovl_drop_write(dentry);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (err) {
|
2019-12-16 18:12:32 +07:00
|
|
|
pr_warn_ratelimited("failed to copy up on encode (%pd2, err=%i)\n",
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
dentry, err);
|
|
|
|
}
|
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Before encoding a non-upper directory file handle from real layer N, we need
|
|
|
|
* to check if it will be possible to reconnect an overlay dentry from the real
|
|
|
|
* lower decoded dentry. This is done by following the overlay ancestry up to a
|
|
|
|
* "layer N connected" ancestor and verifying that all parents along the way are
|
|
|
|
* "layer N connectable". If an ancestor that is NOT "layer N connectable" is
|
|
|
|
* found, we need to copy up an ancestor, which is "layer N connectable", thus
|
|
|
|
* making that ancestor "layer N connected". For example:
|
|
|
|
*
|
|
|
|
* layer 1: /a
|
|
|
|
* layer 2: /a/b/c
|
|
|
|
*
|
|
|
|
* The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
|
|
|
|
* copied up and renamed, upper dir /a will be indexed by lower dir /a from
|
|
|
|
* layer 1. The dir /a from layer 2 will never be indexed, so the algorithm (*)
|
|
|
|
* in ovl_lookup_real_ancestor() will not be able to lookup a connected overlay
|
|
|
|
* dentry from the connected lower dentry /a/b/c.
|
|
|
|
*
|
|
|
|
* To avoid this problem on decode time, we need to copy up an ancestor of
|
|
|
|
* /a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
|
|
|
|
* /a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
|
|
|
|
* and when the time comes to decode the file handle from lower dentry /a/b/c,
|
|
|
|
* ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
|
|
|
|
* a connected overlay dentry will be accomplished.
|
|
|
|
*
|
|
|
|
* (*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup an
|
|
|
|
* entry /a in the lower layers above layer N and find the indexed dir /a from
|
|
|
|
* layer 1. If that improvement is made, then the check for "layer N connected"
|
|
|
|
* will need to verify there are no redirects in lower layers above N. In the
|
|
|
|
* example above, /a will be "layer 2 connectable". However, if layer 2 dir /a
|
|
|
|
* is a target of a layer 1 redirect, then /a will NOT be "layer 2 connectable":
|
|
|
|
*
|
|
|
|
* layer 1: /A (redirect = /a)
|
|
|
|
* layer 2: /a/b/c
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* Return the lowest layer for encoding a connectable file handle */
|
|
|
|
static int ovl_connectable_layer(struct dentry *dentry)
|
|
|
|
{
|
|
|
|
struct ovl_entry *oe = OVL_E(dentry);
|
|
|
|
|
|
|
|
/* We can get overlay root from root of any layer */
|
|
|
|
if (dentry == dentry->d_sb->s_root)
|
|
|
|
return oe->numlower;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If it's an unindexed merge dir, then it's not connectable with any
|
|
|
|
* lower layer
|
|
|
|
*/
|
|
|
|
if (ovl_dentry_upper(dentry) &&
|
|
|
|
!ovl_test_flag(OVL_INDEX, d_inode(dentry)))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
/* We can get upper/overlay path from indexed/lower dentry */
|
|
|
|
return oe->lowerstack[0].layer->idx;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* @dentry is "connected" if all ancestors up to root or a "connected" ancestor
|
|
|
|
* have the same uppermost lower layer as the origin's layer. We may need to
|
|
|
|
* copy up a "connectable" ancestor to make it "connected". A "connected" dentry
|
|
|
|
* cannot become non "connected", so cache positive result in dentry flags.
|
|
|
|
*
|
|
|
|
* Return the connected origin layer or < 0 on error.
|
|
|
|
*/
|
|
|
|
static int ovl_connect_layer(struct dentry *dentry)
|
|
|
|
{
|
|
|
|
struct dentry *next, *parent = NULL;
|
|
|
|
int origin_layer;
|
|
|
|
int err = 0;
|
|
|
|
|
|
|
|
if (WARN_ON(dentry == dentry->d_sb->s_root) ||
|
|
|
|
WARN_ON(!ovl_dentry_lower(dentry)))
|
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
origin_layer = OVL_E(dentry)->lowerstack[0].layer->idx;
|
|
|
|
if (ovl_dentry_test_flag(OVL_E_CONNECTED, dentry))
|
|
|
|
return origin_layer;
|
|
|
|
|
|
|
|
/* Find the topmost origin layer connectable ancestor of @dentry */
|
|
|
|
next = dget(dentry);
|
|
|
|
for (;;) {
|
|
|
|
parent = dget_parent(next);
|
|
|
|
if (WARN_ON(parent == next)) {
|
|
|
|
err = -EIO;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If @parent is not origin layer connectable, then copy up
|
|
|
|
* @next which is origin layer connectable and we are done.
|
|
|
|
*/
|
|
|
|
if (ovl_connectable_layer(parent) < origin_layer) {
|
|
|
|
err = ovl_encode_maybe_copy_up(next);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* If @parent is connected or indexed we are done */
|
|
|
|
if (ovl_dentry_test_flag(OVL_E_CONNECTED, parent) ||
|
|
|
|
ovl_test_flag(OVL_INDEX, d_inode(parent)))
|
|
|
|
break;
|
|
|
|
|
|
|
|
dput(next);
|
|
|
|
next = parent;
|
|
|
|
}
|
|
|
|
|
|
|
|
dput(parent);
|
|
|
|
dput(next);
|
|
|
|
|
|
|
|
if (!err)
|
|
|
|
ovl_dentry_set_flag(OVL_E_CONNECTED, dentry);
|
|
|
|
|
|
|
|
return err ?: origin_layer;
|
|
|
|
}
|
|
|
|
|
2018-01-18 18:14:55 +07:00
|
|
|
/*
|
|
|
|
* We only need to encode origin if there is a chance that the same object was
|
|
|
|
* encoded pre copy up and then we need to stay consistent with the same
|
|
|
|
* encoding also after copy up. If non-pure upper is not indexed, then it was
|
|
|
|
* copied up before NFS export was enabled. In that case we don't need to worry
|
|
|
|
* about staying consistent with pre copy up encoding and we encode an upper
|
|
|
|
* file handle. Overlay root dentry is a private case of non-indexed upper.
|
|
|
|
*
|
|
|
|
* The following table summarizes the different file handle encodings used for
|
|
|
|
* different overlay object types:
|
|
|
|
*
|
|
|
|
* Object type | Encoding
|
|
|
|
* --------------------------------
|
|
|
|
* Pure upper | U
|
|
|
|
* Non-indexed upper | U
|
2018-01-18 18:15:26 +07:00
|
|
|
* Indexed upper | L (*)
|
|
|
|
* Non-upper | L (*)
|
2018-01-18 18:14:55 +07:00
|
|
|
*
|
|
|
|
* U = upper file handle
|
|
|
|
* L = lower file handle
|
2018-01-18 18:15:26 +07:00
|
|
|
*
|
|
|
|
* (*) Connecting an overlay dir from real lower dentry is not always
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
* possible when there are redirects in lower layers and non-indexed merge dirs.
|
|
|
|
* To mitigate those case, we may copy up the lower dir ancestor before encode
|
|
|
|
* a lower dir file handle.
|
|
|
|
*
|
|
|
|
* Return 0 for upper file handle, > 0 for lower file handle or < 0 on error.
|
2018-01-18 18:14:55 +07:00
|
|
|
*/
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
static int ovl_check_encode_origin(struct dentry *dentry)
|
2018-01-18 18:14:55 +07:00
|
|
|
{
|
2018-01-18 18:15:26 +07:00
|
|
|
struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
|
|
|
|
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
/* Upper file handle for pure upper */
|
2018-01-18 18:14:55 +07:00
|
|
|
if (!ovl_dentry_lower(dentry))
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
return 0;
|
2018-01-18 18:14:55 +07:00
|
|
|
|
2018-01-18 18:15:26 +07:00
|
|
|
/*
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
* Upper file handle for non-indexed upper.
|
2018-01-18 18:15:26 +07:00
|
|
|
*
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
* Root is never indexed, so if there's an upper layer, encode upper for
|
|
|
|
* root.
|
2018-01-18 18:15:26 +07:00
|
|
|
*/
|
2018-01-18 18:14:55 +07:00
|
|
|
if (ovl_dentry_upper(dentry) &&
|
|
|
|
!ovl_test_flag(OVL_INDEX, d_inode(dentry)))
|
2018-01-18 18:15:26 +07:00
|
|
|
return 0;
|
|
|
|
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
/*
|
|
|
|
* Decoding a merge dir, whose origin's ancestor is under a redirected
|
|
|
|
* lower dir or under a non-indexed upper is not always possible.
|
|
|
|
* ovl_connect_layer() will try to make origin's layer "connected" by
|
|
|
|
* copying up a "connectable" ancestor.
|
|
|
|
*/
|
2020-06-04 15:48:19 +07:00
|
|
|
if (d_is_dir(dentry) && ovl_upper_mnt(ofs))
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
return ovl_connect_layer(dentry);
|
2018-01-18 18:15:26 +07:00
|
|
|
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
/* Lower file handle for indexed and non-upper dir/non-dir */
|
|
|
|
return 1;
|
2018-01-18 18:15:26 +07:00
|
|
|
}
|
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
static int ovl_dentry_to_fid(struct dentry *dentry, u32 *fid, int buflen)
|
2017-07-12 18:17:16 +07:00
|
|
|
{
|
|
|
|
struct ovl_fh *fh = NULL;
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
int err, enc_lower;
|
2019-11-15 18:33:03 +07:00
|
|
|
int len;
|
2017-07-12 18:17:16 +07:00
|
|
|
|
2018-01-18 18:15:26 +07:00
|
|
|
/*
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
* Check if we should encode a lower or upper file handle and maybe
|
|
|
|
* copy up an ancestor to make lower file handle connectable.
|
2018-01-18 18:15:26 +07:00
|
|
|
*/
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
err = enc_lower = ovl_check_encode_origin(dentry);
|
|
|
|
if (enc_lower < 0)
|
|
|
|
goto fail;
|
2017-07-12 18:17:16 +07:00
|
|
|
|
ovl: check lower ancestry on encode of lower dir file handle
This change relaxes copy up on encode of merge dir with lower layer > 1
and handles the case of encoding a merge dir with lower layer 1, where an
ancestor is a non-indexed merge dir. In that case, decode of the lower
file handle will not have been possible if the non-indexed ancestor is
redirected before or after encode.
Before encoding a non-upper directory file handle from real layer N, we
need to check if it will be possible to reconnect an overlay dentry from
the real lower decoded dentry. This is done by following the overlay
ancestry up to a "layer N connected" ancestor and verifying that all
parents along the way are "layer N connectable". If an ancestor that is
NOT "layer N connectable" is found, we need to copy up an ancestor, which
is "layer N connectable", thus making that ancestor "layer N connected".
For example:
layer 1: /a
layer 2: /a/b/c
The overlay dentry /a is NOT "layer 2 connectable", because if dir /a is
copied up and renamed, upper dir /a will be indexed by lower dir /a from
layer 1. The dir /a from layer 2 will never be indexed, so the algorithm
in ovl_lookup_real_ancestor() (*) will not be able to lookup a connected
overlay dentry from the connected lower dentry /a/b/c.
To avoid this problem on decode time, we need to copy up an ancestor of
/a/b/c, which is "layer 2 connectable", on encode time. That ancestor is
/a/b. After copy up (and index) of /a/b, it will become "layer 2 connected"
and when the time comes to decode the file handle from lower dentry /a/b/c,
ovl_lookup_real_ancestor() will find the indexed ancestor /a/b and decoding
a connected overlay dentry will be accomplished.
(*) the algorithm in ovl_lookup_real_ancestor() can be improved to lookup
an entry /a in the lower layers above layer N and find the indexed dir /a
from layer 1. If that improvement is made, then the check for "layer N
connected" will need to verify there are no redirects in lower layers above
layer N. In the example above, /a will be "layer 2 connectable". However,
if layer 2 dir /a is a target of a layer 1 redirect, then /a will NOT be
"layer 2 connectable":
layer 1: /A (redirect = /a)
layer 2: /a/b/c
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-01-30 18:31:09 +07:00
|
|
|
/* Encode an upper or lower file handle */
|
2018-02-02 15:42:03 +07:00
|
|
|
fh = ovl_encode_real_fh(enc_lower ? ovl_dentry_lower(dentry) :
|
|
|
|
ovl_dentry_upper(dentry), !enc_lower);
|
2018-01-30 18:54:45 +07:00
|
|
|
if (IS_ERR(fh))
|
2019-09-09 15:29:56 +07:00
|
|
|
return PTR_ERR(fh);
|
2017-07-12 18:17:16 +07:00
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
len = OVL_FH_LEN(fh);
|
2020-05-05 02:35:09 +07:00
|
|
|
if (len <= buflen)
|
|
|
|
memcpy(fid, fh, len);
|
2019-11-15 18:33:03 +07:00
|
|
|
err = len;
|
2017-07-12 18:17:16 +07:00
|
|
|
|
|
|
|
out:
|
|
|
|
kfree(fh);
|
|
|
|
return err;
|
|
|
|
|
|
|
|
fail:
|
2020-05-05 02:35:09 +07:00
|
|
|
pr_warn_ratelimited("failed to encode file handle (%pd2, err=%i)\n",
|
|
|
|
dentry, err);
|
2017-07-12 18:17:16 +07:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2018-02-02 15:42:03 +07:00
|
|
|
static int ovl_encode_fh(struct inode *inode, u32 *fid, int *max_len,
|
|
|
|
struct inode *parent)
|
2017-07-12 18:17:16 +07:00
|
|
|
{
|
|
|
|
struct dentry *dentry;
|
2020-05-05 02:35:09 +07:00
|
|
|
int bytes, buflen = *max_len << 2;
|
2017-07-12 18:17:16 +07:00
|
|
|
|
|
|
|
/* TODO: encode connectable file handles */
|
|
|
|
if (parent)
|
|
|
|
return FILEID_INVALID;
|
|
|
|
|
|
|
|
dentry = d_find_any_alias(inode);
|
|
|
|
if (WARN_ON(!dentry))
|
|
|
|
return FILEID_INVALID;
|
|
|
|
|
2020-05-05 02:35:09 +07:00
|
|
|
bytes = ovl_dentry_to_fid(dentry, fid, buflen);
|
2017-07-12 18:17:16 +07:00
|
|
|
dput(dentry);
|
2019-11-15 18:33:03 +07:00
|
|
|
if (bytes <= 0)
|
|
|
|
return FILEID_INVALID;
|
|
|
|
|
|
|
|
*max_len = bytes >> 2;
|
2020-05-05 02:35:09 +07:00
|
|
|
if (bytes > buflen)
|
|
|
|
return FILEID_INVALID;
|
2019-11-15 18:33:03 +07:00
|
|
|
|
|
|
|
return OVL_FILEID_V1;
|
2017-07-12 18:17:16 +07:00
|
|
|
}
|
|
|
|
|
2018-01-19 06:03:23 +07:00
|
|
|
/*
|
2018-01-20 02:36:20 +07:00
|
|
|
* Find or instantiate an overlay dentry from real dentries and index.
|
2018-01-19 06:03:23 +07:00
|
|
|
*/
|
|
|
|
static struct dentry *ovl_obtain_alias(struct super_block *sb,
|
2018-01-20 02:36:20 +07:00
|
|
|
struct dentry *upper_alias,
|
|
|
|
struct ovl_path *lowerpath,
|
|
|
|
struct dentry *index)
|
2018-01-19 06:03:23 +07:00
|
|
|
{
|
2018-01-20 02:33:44 +07:00
|
|
|
struct dentry *lower = lowerpath ? lowerpath->dentry : NULL;
|
2018-01-20 02:36:20 +07:00
|
|
|
struct dentry *upper = upper_alias ?: index;
|
2018-01-19 06:03:23 +07:00
|
|
|
struct dentry *dentry;
|
2018-01-20 02:33:44 +07:00
|
|
|
struct inode *inode;
|
2018-01-19 06:03:23 +07:00
|
|
|
struct ovl_entry *oe;
|
2018-05-08 20:27:21 +07:00
|
|
|
struct ovl_inode_params oip = {
|
|
|
|
.lowerpath = lowerpath,
|
|
|
|
.index = index,
|
|
|
|
.numlower = !!lower
|
|
|
|
};
|
2018-01-19 06:03:23 +07:00
|
|
|
|
2018-01-20 02:36:20 +07:00
|
|
|
/* We get overlay directory dentries with ovl_lookup_real() */
|
|
|
|
if (d_is_dir(upper ?: lower))
|
2018-01-19 06:03:23 +07:00
|
|
|
return ERR_PTR(-EIO);
|
|
|
|
|
2018-05-08 20:27:21 +07:00
|
|
|
oip.upperdentry = dget(upper);
|
|
|
|
inode = ovl_get_inode(sb, &oip);
|
2018-01-19 06:03:23 +07:00
|
|
|
if (IS_ERR(inode)) {
|
|
|
|
dput(upper);
|
|
|
|
return ERR_CAST(inode);
|
|
|
|
}
|
|
|
|
|
2018-05-11 22:49:28 +07:00
|
|
|
if (upper)
|
|
|
|
ovl_set_flag(OVL_UPPERDATA, inode);
|
|
|
|
|
2018-01-19 06:03:23 +07:00
|
|
|
dentry = d_find_any_alias(inode);
|
2019-10-24 08:06:45 +07:00
|
|
|
if (dentry)
|
|
|
|
goto out_iput;
|
|
|
|
|
|
|
|
dentry = d_alloc_anon(inode->i_sb);
|
|
|
|
if (unlikely(!dentry))
|
|
|
|
goto nomem;
|
|
|
|
oe = ovl_alloc_entry(lower ? 1 : 0);
|
|
|
|
if (!oe)
|
|
|
|
goto nomem;
|
|
|
|
|
|
|
|
if (lower) {
|
|
|
|
oe->lowerstack->dentry = dget(lower);
|
|
|
|
oe->lowerstack->layer = lowerpath->layer;
|
2018-01-19 06:03:23 +07:00
|
|
|
}
|
2019-10-24 08:06:45 +07:00
|
|
|
dentry->d_fsdata = oe;
|
|
|
|
if (upper_alias)
|
|
|
|
ovl_dentry_set_upper_alias(dentry);
|
|
|
|
|
2020-03-17 21:04:22 +07:00
|
|
|
ovl_dentry_update_reval(dentry, upper,
|
|
|
|
DCACHE_OP_REVALIDATE | DCACHE_OP_WEAK_REVALIDATE);
|
2018-01-19 06:03:23 +07:00
|
|
|
|
|
|
|
return d_instantiate_anon(dentry, inode);
|
|
|
|
|
|
|
|
nomem:
|
|
|
|
dput(dentry);
|
2019-10-24 08:06:45 +07:00
|
|
|
dentry = ERR_PTR(-ENOMEM);
|
|
|
|
out_iput:
|
|
|
|
iput(inode);
|
|
|
|
return dentry;
|
2018-01-19 06:03:23 +07:00
|
|
|
}
|
|
|
|
|
2018-01-18 03:32:44 +07:00
|
|
|
/* Get the upper or lower dentry in stach whose on layer @idx */
|
|
|
|
static struct dentry *ovl_dentry_real_at(struct dentry *dentry, int idx)
|
|
|
|
{
|
|
|
|
struct ovl_entry *oe = dentry->d_fsdata;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
if (!idx)
|
|
|
|
return ovl_dentry_upper(dentry);
|
|
|
|
|
|
|
|
for (i = 0; i < oe->numlower; i++) {
|
|
|
|
if (oe->lowerstack[i].layer->idx == idx)
|
|
|
|
return oe->lowerstack[i].dentry;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
/*
|
|
|
|
* Lookup a child overlay dentry to get a connected overlay dentry whose real
|
|
|
|
* dentry is @real. If @real is on upper layer, we lookup a child overlay
|
|
|
|
* dentry with the same name as the real dentry. Otherwise, we need to consult
|
|
|
|
* index for lookup.
|
|
|
|
*/
|
|
|
|
static struct dentry *ovl_lookup_real_one(struct dentry *connected,
|
|
|
|
struct dentry *real,
|
2020-01-24 15:46:45 +07:00
|
|
|
const struct ovl_layer *layer)
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
{
|
|
|
|
struct inode *dir = d_inode(connected);
|
|
|
|
struct dentry *this, *parent = NULL;
|
|
|
|
struct name_snapshot name;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup child overlay dentry by real name. The dir mutex protects us
|
|
|
|
* from racing with overlay rename. If the overlay dentry that is above
|
|
|
|
* real has already been moved to a parent that is not under the
|
|
|
|
* connected overlay dir, we return -ECHILD and restart the lookup of
|
|
|
|
* connected real path from the top.
|
|
|
|
*/
|
|
|
|
inode_lock_nested(dir, I_MUTEX_PARENT);
|
|
|
|
err = -ECHILD;
|
|
|
|
parent = dget_parent(real);
|
2018-01-18 03:32:44 +07:00
|
|
|
if (ovl_dentry_real_at(connected, layer->idx) != parent)
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
goto fail;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We also need to take a snapshot of real dentry name to protect us
|
|
|
|
* from racing with underlying layer rename. In this case, we don't
|
|
|
|
* care about returning ESTALE, only from dereferencing a free name
|
|
|
|
* pointer because we hold no lock on the real dentry.
|
|
|
|
*/
|
|
|
|
take_dentry_name_snapshot(&name, real);
|
2019-04-27 00:07:27 +07:00
|
|
|
this = lookup_one_len(name.name.name, connected, name.name.len);
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
err = PTR_ERR(this);
|
|
|
|
if (IS_ERR(this)) {
|
|
|
|
goto fail;
|
|
|
|
} else if (!this || !this->d_inode) {
|
|
|
|
dput(this);
|
|
|
|
err = -ENOENT;
|
|
|
|
goto fail;
|
2018-01-18 03:32:44 +07:00
|
|
|
} else if (ovl_dentry_real_at(this, layer->idx) != real) {
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
dput(this);
|
|
|
|
err = -ESTALE;
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
|
|
|
|
out:
|
|
|
|
release_dentry_name_snapshot(&name);
|
|
|
|
dput(parent);
|
|
|
|
inode_unlock(dir);
|
|
|
|
return this;
|
|
|
|
|
|
|
|
fail:
|
2019-12-16 18:12:32 +07:00
|
|
|
pr_warn_ratelimited("failed to lookup one by real (%pd2, layer=%d, connected=%pd2, err=%i)\n",
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
real, layer->idx, connected, err);
|
|
|
|
this = ERR_PTR(err);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2018-01-17 19:40:27 +07:00
|
|
|
static struct dentry *ovl_lookup_real(struct super_block *sb,
|
|
|
|
struct dentry *real,
|
2020-01-24 15:46:45 +07:00
|
|
|
const struct ovl_layer *layer);
|
2018-01-17 19:40:27 +07:00
|
|
|
|
2018-01-18 21:39:13 +07:00
|
|
|
/*
|
|
|
|
* Lookup an indexed or hashed overlay dentry by real inode.
|
|
|
|
*/
|
|
|
|
static struct dentry *ovl_lookup_real_inode(struct super_block *sb,
|
|
|
|
struct dentry *real,
|
2020-01-24 15:46:45 +07:00
|
|
|
const struct ovl_layer *layer)
|
2018-01-18 21:39:13 +07:00
|
|
|
{
|
2018-01-17 19:40:27 +07:00
|
|
|
struct ovl_fs *ofs = sb->s_fs_info;
|
|
|
|
struct dentry *index = NULL;
|
2018-01-18 21:39:13 +07:00
|
|
|
struct dentry *this = NULL;
|
|
|
|
struct inode *inode;
|
|
|
|
|
2018-01-17 19:40:27 +07:00
|
|
|
/*
|
|
|
|
* Decoding upper dir from index is expensive, so first try to lookup
|
|
|
|
* overlay dentry in inode/dcache.
|
|
|
|
*/
|
2018-01-18 21:39:13 +07:00
|
|
|
inode = ovl_lookup_inode(sb, real, !layer->idx);
|
|
|
|
if (IS_ERR(inode))
|
|
|
|
return ERR_CAST(inode);
|
|
|
|
if (inode) {
|
|
|
|
this = d_find_any_alias(inode);
|
|
|
|
iput(inode);
|
|
|
|
}
|
|
|
|
|
2018-01-17 19:40:27 +07:00
|
|
|
/*
|
|
|
|
* For decoded lower dir file handle, lookup index by origin to check
|
|
|
|
* if lower dir was copied up and and/or removed.
|
|
|
|
*/
|
|
|
|
if (!this && layer->idx && ofs->indexdir && !WARN_ON(!d_is_dir(real))) {
|
|
|
|
index = ovl_lookup_index(ofs, NULL, real, false);
|
|
|
|
if (IS_ERR(index))
|
|
|
|
return index;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Get connected upper overlay dir from index */
|
|
|
|
if (index) {
|
|
|
|
struct dentry *upper = ovl_index_upper(ofs, index);
|
|
|
|
|
|
|
|
dput(index);
|
|
|
|
if (IS_ERR_OR_NULL(upper))
|
|
|
|
return upper;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* ovl_lookup_real() in lower layer may call recursively once to
|
|
|
|
* ovl_lookup_real() in upper layer. The first level call walks
|
|
|
|
* back lower parents to the topmost indexed parent. The second
|
|
|
|
* recursive call walks back from indexed upper to the topmost
|
|
|
|
* connected/hashed upper parent (or up to root).
|
|
|
|
*/
|
2019-11-15 19:12:40 +07:00
|
|
|
this = ovl_lookup_real(sb, upper, &ofs->layers[0]);
|
2018-01-17 19:40:27 +07:00
|
|
|
dput(upper);
|
|
|
|
}
|
|
|
|
|
2018-01-30 19:30:50 +07:00
|
|
|
if (IS_ERR_OR_NULL(this))
|
|
|
|
return this;
|
2018-01-18 21:39:13 +07:00
|
|
|
|
|
|
|
if (WARN_ON(ovl_dentry_real_at(this, layer->idx) != real)) {
|
|
|
|
dput(this);
|
|
|
|
this = ERR_PTR(-EIO);
|
|
|
|
}
|
|
|
|
|
|
|
|
return this;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup an indexed or hashed overlay dentry, whose real dentry is an
|
|
|
|
* ancestor of @real.
|
|
|
|
*/
|
|
|
|
static struct dentry *ovl_lookup_real_ancestor(struct super_block *sb,
|
|
|
|
struct dentry *real,
|
2020-01-24 15:46:45 +07:00
|
|
|
const struct ovl_layer *layer)
|
2018-01-18 21:39:13 +07:00
|
|
|
{
|
|
|
|
struct dentry *next, *parent = NULL;
|
|
|
|
struct dentry *ancestor = ERR_PTR(-EIO);
|
|
|
|
|
|
|
|
if (real == layer->mnt->mnt_root)
|
|
|
|
return dget(sb->s_root);
|
|
|
|
|
|
|
|
/* Find the topmost indexed or hashed ancestor */
|
|
|
|
next = dget(real);
|
|
|
|
for (;;) {
|
|
|
|
parent = dget_parent(next);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup a matching overlay dentry in inode/dentry
|
|
|
|
* cache or in index by real inode.
|
|
|
|
*/
|
|
|
|
ancestor = ovl_lookup_real_inode(sb, next, layer);
|
|
|
|
if (ancestor)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (parent == layer->mnt->mnt_root) {
|
|
|
|
ancestor = dget(sb->s_root);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If @real has been moved out of the layer root directory,
|
|
|
|
* we will eventully hit the real fs root. This cannot happen
|
|
|
|
* by legit overlay rename, so we return error in that case.
|
|
|
|
*/
|
|
|
|
if (parent == next) {
|
|
|
|
ancestor = ERR_PTR(-EXDEV);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
dput(next);
|
|
|
|
next = parent;
|
|
|
|
}
|
|
|
|
|
|
|
|
dput(parent);
|
|
|
|
dput(next);
|
|
|
|
|
|
|
|
return ancestor;
|
|
|
|
}
|
|
|
|
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
/*
|
|
|
|
* Lookup a connected overlay dentry whose real dentry is @real.
|
|
|
|
* If @real is on upper layer, we lookup a child overlay dentry with the same
|
|
|
|
* path the real dentry. Otherwise, we need to consult index for lookup.
|
|
|
|
*/
|
|
|
|
static struct dentry *ovl_lookup_real(struct super_block *sb,
|
|
|
|
struct dentry *real,
|
2020-01-24 15:46:45 +07:00
|
|
|
const struct ovl_layer *layer)
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
{
|
|
|
|
struct dentry *connected;
|
|
|
|
int err = 0;
|
|
|
|
|
2018-01-18 21:39:13 +07:00
|
|
|
connected = ovl_lookup_real_ancestor(sb, real, layer);
|
|
|
|
if (IS_ERR(connected))
|
|
|
|
return connected;
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
|
|
|
|
while (!err) {
|
|
|
|
struct dentry *next, *this;
|
|
|
|
struct dentry *parent = NULL;
|
2018-01-18 03:32:44 +07:00
|
|
|
struct dentry *real_connected = ovl_dentry_real_at(connected,
|
|
|
|
layer->idx);
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
|
|
|
|
if (real_connected == real)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/* Find the topmost dentry not yet connected */
|
|
|
|
next = dget(real);
|
|
|
|
for (;;) {
|
|
|
|
parent = dget_parent(next);
|
|
|
|
|
|
|
|
if (parent == real_connected)
|
|
|
|
break;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If real has been moved out of 'real_connected',
|
|
|
|
* we will not find 'real_connected' and hit the layer
|
|
|
|
* root. In that case, we need to restart connecting.
|
|
|
|
* This game can go on forever in the worst case. We
|
|
|
|
* may want to consider taking s_vfs_rename_mutex if
|
|
|
|
* this happens more than once.
|
|
|
|
*/
|
|
|
|
if (parent == layer->mnt->mnt_root) {
|
|
|
|
dput(connected);
|
|
|
|
connected = dget(sb->s_root);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If real file has been moved out of the layer root
|
|
|
|
* directory, we will eventully hit the real fs root.
|
|
|
|
* This cannot happen by legit overlay rename, so we
|
|
|
|
* return error in that case.
|
|
|
|
*/
|
|
|
|
if (parent == next) {
|
|
|
|
err = -EXDEV;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
dput(next);
|
|
|
|
next = parent;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!err) {
|
|
|
|
this = ovl_lookup_real_one(connected, next, layer);
|
|
|
|
if (IS_ERR(this))
|
|
|
|
err = PTR_ERR(this);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup of child in overlay can fail when racing with
|
|
|
|
* overlay rename of child away from 'connected' parent.
|
|
|
|
* In this case, we need to restart the lookup from the
|
|
|
|
* top, because we cannot trust that 'real_connected' is
|
2018-01-18 21:39:13 +07:00
|
|
|
* still an ancestor of 'real'. There is a good chance
|
|
|
|
* that the renamed overlay ancestor is now in cache, so
|
|
|
|
* ovl_lookup_real_ancestor() will find it and we can
|
|
|
|
* continue to connect exactly from where lookup failed.
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
*/
|
|
|
|
if (err == -ECHILD) {
|
2018-01-18 21:39:13 +07:00
|
|
|
this = ovl_lookup_real_ancestor(sb, real,
|
|
|
|
layer);
|
2018-02-05 23:25:16 +07:00
|
|
|
err = PTR_ERR_OR_ZERO(this);
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
}
|
|
|
|
if (!err) {
|
|
|
|
dput(connected);
|
|
|
|
connected = this;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
dput(parent);
|
|
|
|
dput(next);
|
|
|
|
}
|
|
|
|
|
|
|
|
if (err)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
return connected;
|
|
|
|
|
|
|
|
fail:
|
2019-12-16 18:12:32 +07:00
|
|
|
pr_warn_ratelimited("failed to lookup by real (%pd2, layer=%d, connected=%pd2, err=%i)\n",
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
real, layer->idx, connected, err);
|
|
|
|
dput(connected);
|
|
|
|
return ERR_PTR(err);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2018-01-20 02:36:20 +07:00
|
|
|
* Get an overlay dentry from upper/lower real dentries and index.
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
*/
|
|
|
|
static struct dentry *ovl_get_dentry(struct super_block *sb,
|
|
|
|
struct dentry *upper,
|
2018-01-20 02:36:20 +07:00
|
|
|
struct ovl_path *lowerpath,
|
|
|
|
struct dentry *index)
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
{
|
|
|
|
struct ovl_fs *ofs = sb->s_fs_info;
|
2020-01-24 15:46:45 +07:00
|
|
|
const struct ovl_layer *layer = upper ? &ofs->layers[0] : lowerpath->layer;
|
2018-01-20 02:36:20 +07:00
|
|
|
struct dentry *real = upper ?: (index ?: lowerpath->dentry);
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
|
2018-01-20 02:33:44 +07:00
|
|
|
/*
|
2018-01-20 02:36:20 +07:00
|
|
|
* Obtain a disconnected overlay dentry from a non-dir real dentry
|
|
|
|
* and index.
|
2018-01-20 02:33:44 +07:00
|
|
|
*/
|
2018-01-20 02:36:20 +07:00
|
|
|
if (!d_is_dir(real))
|
|
|
|
return ovl_obtain_alias(sb, upper, lowerpath, index);
|
2018-01-20 02:33:44 +07:00
|
|
|
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
/* Removed empty directory? */
|
2018-01-18 03:32:44 +07:00
|
|
|
if ((real->d_flags & DCACHE_DISCONNECTED) || d_unhashed(real))
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
return ERR_PTR(-ENOENT);
|
|
|
|
|
|
|
|
/*
|
2018-01-18 03:32:44 +07:00
|
|
|
* If real dentry is connected and hashed, get a connected overlay
|
|
|
|
* dentry whose real dentry is @real.
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
*/
|
2018-01-18 03:32:44 +07:00
|
|
|
return ovl_lookup_real(sb, real, layer);
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
}
|
|
|
|
|
2018-01-19 06:03:23 +07:00
|
|
|
static struct dentry *ovl_upper_fh_to_d(struct super_block *sb,
|
|
|
|
struct ovl_fh *fh)
|
|
|
|
{
|
|
|
|
struct ovl_fs *ofs = sb->s_fs_info;
|
|
|
|
struct dentry *dentry;
|
|
|
|
struct dentry *upper;
|
|
|
|
|
2020-06-04 15:48:19 +07:00
|
|
|
if (!ovl_upper_mnt(ofs))
|
2018-01-19 06:03:23 +07:00
|
|
|
return ERR_PTR(-EACCES);
|
|
|
|
|
2020-06-04 15:48:19 +07:00
|
|
|
upper = ovl_decode_real_fh(fh, ovl_upper_mnt(ofs), true);
|
2018-01-19 06:03:23 +07:00
|
|
|
if (IS_ERR_OR_NULL(upper))
|
|
|
|
return upper;
|
|
|
|
|
2018-01-20 02:36:20 +07:00
|
|
|
dentry = ovl_get_dentry(sb, upper, NULL, NULL);
|
2018-01-19 06:03:23 +07:00
|
|
|
dput(upper);
|
|
|
|
|
|
|
|
return dentry;
|
|
|
|
}
|
|
|
|
|
2018-01-20 02:33:44 +07:00
|
|
|
static struct dentry *ovl_lower_fh_to_d(struct super_block *sb,
|
|
|
|
struct ovl_fh *fh)
|
|
|
|
{
|
|
|
|
struct ovl_fs *ofs = sb->s_fs_info;
|
|
|
|
struct ovl_path origin = { };
|
|
|
|
struct ovl_path *stack = &origin;
|
|
|
|
struct dentry *dentry = NULL;
|
2018-01-20 02:36:20 +07:00
|
|
|
struct dentry *index = NULL;
|
ovl: lookup in inode cache first when decoding lower file handle
When decoding a lower file handle, we need to check if lower file was
copied up and indexed and if it has a whiteout index, we need to check
if this is an unlinked but open non-dir before returning -ESTALE.
To find out if this is an unlinked but open non-dir we need to lookup
an overlay inode in inode cache by lower inode and that requires decoding
the lower file handle before looking in inode cache.
Before this change, if the lower inode turned out to be a directory, we
may have paid an expensive cost to reconnect that lower directory for
nothing.
After this change, we start by decoding a disconnected lower dentry and
using the lower inode for looking up an overlay inode in inode cache.
If we find overlay inode and dentry in cache, we avoid the index lookup
overhead. If we don't find an overlay inode and dentry in cache, then we
only need to decode a connected lower dentry in case the lower dentry is
a non-indexed directory.
The xfstests group overlay/exportfs tests decoding overlayfs file
handles after drop_caches with different states of the file at encode
and decode time. Overall the tests in the group call ovl_lower_fh_to_d()
89 times to decode a lower file handle.
Before this change, the tests called ovl_get_index_fh() 75 times and
reconnect_one() 61 times.
After this change, the tests call ovl_get_index_fh() 70 times and
reconnect_one() 59 times. The 2 cases where reconnect_one() was avoided
are cases where a non-upper directory file handle was encoded, then the
directory removed and then file handle was decoded.
To demonstrate the affect on decoding file handles with hot inode/dentry
cache, the drop_caches call in the tests was disabled. Without
drop_caches, there are no reconnect_one() calls at all before or after
the change. Before the change, there are 75 calls to ovl_get_index_fh(),
exactly as the case with drop_caches. After the change, there are only
10 calls to ovl_get_index_fh().
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-03-09 22:05:55 +07:00
|
|
|
struct inode *inode;
|
2018-01-20 02:33:44 +07:00
|
|
|
int err;
|
|
|
|
|
ovl: lookup in inode cache first when decoding lower file handle
When decoding a lower file handle, we need to check if lower file was
copied up and indexed and if it has a whiteout index, we need to check
if this is an unlinked but open non-dir before returning -ESTALE.
To find out if this is an unlinked but open non-dir we need to lookup
an overlay inode in inode cache by lower inode and that requires decoding
the lower file handle before looking in inode cache.
Before this change, if the lower inode turned out to be a directory, we
may have paid an expensive cost to reconnect that lower directory for
nothing.
After this change, we start by decoding a disconnected lower dentry and
using the lower inode for looking up an overlay inode in inode cache.
If we find overlay inode and dentry in cache, we avoid the index lookup
overhead. If we don't find an overlay inode and dentry in cache, then we
only need to decode a connected lower dentry in case the lower dentry is
a non-indexed directory.
The xfstests group overlay/exportfs tests decoding overlayfs file
handles after drop_caches with different states of the file at encode
and decode time. Overall the tests in the group call ovl_lower_fh_to_d()
89 times to decode a lower file handle.
Before this change, the tests called ovl_get_index_fh() 75 times and
reconnect_one() 61 times.
After this change, the tests call ovl_get_index_fh() 70 times and
reconnect_one() 59 times. The 2 cases where reconnect_one() was avoided
are cases where a non-upper directory file handle was encoded, then the
directory removed and then file handle was decoded.
To demonstrate the affect on decoding file handles with hot inode/dentry
cache, the drop_caches call in the tests was disabled. Without
drop_caches, there are no reconnect_one() calls at all before or after
the change. Before the change, there are 75 calls to ovl_get_index_fh(),
exactly as the case with drop_caches. After the change, there are only
10 calls to ovl_get_index_fh().
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-03-09 22:05:55 +07:00
|
|
|
/* First lookup overlay inode in inode cache by origin fh */
|
|
|
|
err = ovl_check_origin_fh(ofs, fh, false, NULL, &stack);
|
|
|
|
if (err)
|
|
|
|
return ERR_PTR(err);
|
|
|
|
|
|
|
|
if (!d_is_dir(origin.dentry) ||
|
|
|
|
!(origin.dentry->d_flags & DCACHE_DISCONNECTED)) {
|
|
|
|
inode = ovl_lookup_inode(sb, origin.dentry, false);
|
|
|
|
err = PTR_ERR(inode);
|
|
|
|
if (IS_ERR(inode))
|
|
|
|
goto out_err;
|
|
|
|
if (inode) {
|
|
|
|
dentry = d_find_any_alias(inode);
|
|
|
|
iput(inode);
|
|
|
|
if (dentry)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Then lookup indexed upper/whiteout by origin fh */
|
2018-01-20 02:36:20 +07:00
|
|
|
if (ofs->indexdir) {
|
|
|
|
index = ovl_get_index_fh(ofs, fh);
|
|
|
|
err = PTR_ERR(index);
|
2017-12-24 23:28:04 +07:00
|
|
|
if (IS_ERR(index)) {
|
|
|
|
index = NULL;
|
ovl: lookup in inode cache first when decoding lower file handle
When decoding a lower file handle, we need to check if lower file was
copied up and indexed and if it has a whiteout index, we need to check
if this is an unlinked but open non-dir before returning -ESTALE.
To find out if this is an unlinked but open non-dir we need to lookup
an overlay inode in inode cache by lower inode and that requires decoding
the lower file handle before looking in inode cache.
Before this change, if the lower inode turned out to be a directory, we
may have paid an expensive cost to reconnect that lower directory for
nothing.
After this change, we start by decoding a disconnected lower dentry and
using the lower inode for looking up an overlay inode in inode cache.
If we find overlay inode and dentry in cache, we avoid the index lookup
overhead. If we don't find an overlay inode and dentry in cache, then we
only need to decode a connected lower dentry in case the lower dentry is
a non-indexed directory.
The xfstests group overlay/exportfs tests decoding overlayfs file
handles after drop_caches with different states of the file at encode
and decode time. Overall the tests in the group call ovl_lower_fh_to_d()
89 times to decode a lower file handle.
Before this change, the tests called ovl_get_index_fh() 75 times and
reconnect_one() 61 times.
After this change, the tests call ovl_get_index_fh() 70 times and
reconnect_one() 59 times. The 2 cases where reconnect_one() was avoided
are cases where a non-upper directory file handle was encoded, then the
directory removed and then file handle was decoded.
To demonstrate the affect on decoding file handles with hot inode/dentry
cache, the drop_caches call in the tests was disabled. Without
drop_caches, there are no reconnect_one() calls at all before or after
the change. Before the change, there are 75 calls to ovl_get_index_fh(),
exactly as the case with drop_caches. After the change, there are only
10 calls to ovl_get_index_fh().
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-03-09 22:05:55 +07:00
|
|
|
goto out_err;
|
2017-12-24 23:28:04 +07:00
|
|
|
}
|
2018-01-20 02:36:20 +07:00
|
|
|
}
|
|
|
|
|
ovl: lookup in inode cache first when decoding lower file handle
When decoding a lower file handle, we need to check if lower file was
copied up and indexed and if it has a whiteout index, we need to check
if this is an unlinked but open non-dir before returning -ESTALE.
To find out if this is an unlinked but open non-dir we need to lookup
an overlay inode in inode cache by lower inode and that requires decoding
the lower file handle before looking in inode cache.
Before this change, if the lower inode turned out to be a directory, we
may have paid an expensive cost to reconnect that lower directory for
nothing.
After this change, we start by decoding a disconnected lower dentry and
using the lower inode for looking up an overlay inode in inode cache.
If we find overlay inode and dentry in cache, we avoid the index lookup
overhead. If we don't find an overlay inode and dentry in cache, then we
only need to decode a connected lower dentry in case the lower dentry is
a non-indexed directory.
The xfstests group overlay/exportfs tests decoding overlayfs file
handles after drop_caches with different states of the file at encode
and decode time. Overall the tests in the group call ovl_lower_fh_to_d()
89 times to decode a lower file handle.
Before this change, the tests called ovl_get_index_fh() 75 times and
reconnect_one() 61 times.
After this change, the tests call ovl_get_index_fh() 70 times and
reconnect_one() 59 times. The 2 cases where reconnect_one() was avoided
are cases where a non-upper directory file handle was encoded, then the
directory removed and then file handle was decoded.
To demonstrate the affect on decoding file handles with hot inode/dentry
cache, the drop_caches call in the tests was disabled. Without
drop_caches, there are no reconnect_one() calls at all before or after
the change. Before the change, there are 75 calls to ovl_get_index_fh(),
exactly as the case with drop_caches. After the change, there are only
10 calls to ovl_get_index_fh().
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-03-09 22:05:55 +07:00
|
|
|
/* Then try to get a connected upper dir by index */
|
2017-12-24 23:42:16 +07:00
|
|
|
if (index && d_is_dir(index)) {
|
|
|
|
struct dentry *upper = ovl_index_upper(ofs, index);
|
|
|
|
|
|
|
|
err = PTR_ERR(upper);
|
|
|
|
if (IS_ERR_OR_NULL(upper))
|
|
|
|
goto out_err;
|
|
|
|
|
|
|
|
dentry = ovl_get_dentry(sb, upper, NULL, NULL);
|
|
|
|
dput(upper);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2018-11-05 12:50:10 +07:00
|
|
|
/* Find origin.dentry again with ovl_acceptable() layer check */
|
|
|
|
if (d_is_dir(origin.dentry)) {
|
ovl: lookup in inode cache first when decoding lower file handle
When decoding a lower file handle, we need to check if lower file was
copied up and indexed and if it has a whiteout index, we need to check
if this is an unlinked but open non-dir before returning -ESTALE.
To find out if this is an unlinked but open non-dir we need to lookup
an overlay inode in inode cache by lower inode and that requires decoding
the lower file handle before looking in inode cache.
Before this change, if the lower inode turned out to be a directory, we
may have paid an expensive cost to reconnect that lower directory for
nothing.
After this change, we start by decoding a disconnected lower dentry and
using the lower inode for looking up an overlay inode in inode cache.
If we find overlay inode and dentry in cache, we avoid the index lookup
overhead. If we don't find an overlay inode and dentry in cache, then we
only need to decode a connected lower dentry in case the lower dentry is
a non-indexed directory.
The xfstests group overlay/exportfs tests decoding overlayfs file
handles after drop_caches with different states of the file at encode
and decode time. Overall the tests in the group call ovl_lower_fh_to_d()
89 times to decode a lower file handle.
Before this change, the tests called ovl_get_index_fh() 75 times and
reconnect_one() 61 times.
After this change, the tests call ovl_get_index_fh() 70 times and
reconnect_one() 59 times. The 2 cases where reconnect_one() was avoided
are cases where a non-upper directory file handle was encoded, then the
directory removed and then file handle was decoded.
To demonstrate the affect on decoding file handles with hot inode/dentry
cache, the drop_caches call in the tests was disabled. Without
drop_caches, there are no reconnect_one() calls at all before or after
the change. Before the change, there are 75 calls to ovl_get_index_fh(),
exactly as the case with drop_caches. After the change, there are only
10 calls to ovl_get_index_fh().
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-03-09 22:05:55 +07:00
|
|
|
dput(origin.dentry);
|
|
|
|
origin.dentry = NULL;
|
|
|
|
err = ovl_check_origin_fh(ofs, fh, true, NULL, &stack);
|
2018-01-20 02:36:20 +07:00
|
|
|
if (err)
|
|
|
|
goto out_err;
|
ovl: lookup in inode cache first when decoding lower file handle
When decoding a lower file handle, we need to check if lower file was
copied up and indexed and if it has a whiteout index, we need to check
if this is an unlinked but open non-dir before returning -ESTALE.
To find out if this is an unlinked but open non-dir we need to lookup
an overlay inode in inode cache by lower inode and that requires decoding
the lower file handle before looking in inode cache.
Before this change, if the lower inode turned out to be a directory, we
may have paid an expensive cost to reconnect that lower directory for
nothing.
After this change, we start by decoding a disconnected lower dentry and
using the lower inode for looking up an overlay inode in inode cache.
If we find overlay inode and dentry in cache, we avoid the index lookup
overhead. If we don't find an overlay inode and dentry in cache, then we
only need to decode a connected lower dentry in case the lower dentry is
a non-indexed directory.
The xfstests group overlay/exportfs tests decoding overlayfs file
handles after drop_caches with different states of the file at encode
and decode time. Overall the tests in the group call ovl_lower_fh_to_d()
89 times to decode a lower file handle.
Before this change, the tests called ovl_get_index_fh() 75 times and
reconnect_one() 61 times.
After this change, the tests call ovl_get_index_fh() 70 times and
reconnect_one() 59 times. The 2 cases where reconnect_one() was avoided
are cases where a non-upper directory file handle was encoded, then the
directory removed and then file handle was decoded.
To demonstrate the affect on decoding file handles with hot inode/dentry
cache, the drop_caches call in the tests was disabled. Without
drop_caches, there are no reconnect_one() calls at all before or after
the change. Before the change, there are 75 calls to ovl_get_index_fh(),
exactly as the case with drop_caches. After the change, there are only
10 calls to ovl_get_index_fh().
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2018-03-09 22:05:55 +07:00
|
|
|
}
|
|
|
|
if (index) {
|
|
|
|
err = ovl_verify_origin(index, origin.dentry, false);
|
|
|
|
if (err)
|
2017-12-24 23:28:04 +07:00
|
|
|
goto out_err;
|
2018-01-20 02:36:20 +07:00
|
|
|
}
|
2018-01-20 02:33:44 +07:00
|
|
|
|
2018-11-05 12:50:10 +07:00
|
|
|
/* Get a connected non-upper dir or disconnected non-dir */
|
2018-01-20 02:36:20 +07:00
|
|
|
dentry = ovl_get_dentry(sb, NULL, &origin, index);
|
2018-01-20 02:33:44 +07:00
|
|
|
|
2018-01-20 02:36:20 +07:00
|
|
|
out:
|
|
|
|
dput(origin.dentry);
|
|
|
|
dput(index);
|
2018-01-20 02:33:44 +07:00
|
|
|
return dentry;
|
2018-01-20 02:36:20 +07:00
|
|
|
|
|
|
|
out_err:
|
|
|
|
dentry = ERR_PTR(err);
|
|
|
|
goto out;
|
2018-01-20 02:33:44 +07:00
|
|
|
}
|
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
static struct ovl_fh *ovl_fid_to_fh(struct fid *fid, int buflen, int fh_type)
|
|
|
|
{
|
|
|
|
struct ovl_fh *fh;
|
|
|
|
|
|
|
|
/* If on-wire inner fid is aligned - nothing to do */
|
|
|
|
if (fh_type == OVL_FILEID_V1)
|
|
|
|
return (struct ovl_fh *)fid;
|
|
|
|
|
|
|
|
if (fh_type != OVL_FILEID_V0)
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
|
2020-05-06 01:33:31 +07:00
|
|
|
if (buflen <= OVL_FH_WIRE_OFFSET)
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
fh = kzalloc(buflen, GFP_KERNEL);
|
|
|
|
if (!fh)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
|
|
|
/* Copy unaligned inner fh into aligned buffer */
|
|
|
|
memcpy(&fh->fb, fid, buflen - OVL_FH_WIRE_OFFSET);
|
|
|
|
return fh;
|
|
|
|
}
|
|
|
|
|
2018-01-19 06:03:23 +07:00
|
|
|
static struct dentry *ovl_fh_to_dentry(struct super_block *sb, struct fid *fid,
|
|
|
|
int fh_len, int fh_type)
|
|
|
|
{
|
|
|
|
struct dentry *dentry = NULL;
|
2019-11-15 18:33:03 +07:00
|
|
|
struct ovl_fh *fh = NULL;
|
2018-01-19 06:03:23 +07:00
|
|
|
int len = fh_len << 2;
|
|
|
|
unsigned int flags = 0;
|
|
|
|
int err;
|
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
fh = ovl_fid_to_fh(fid, len, fh_type);
|
|
|
|
err = PTR_ERR(fh);
|
|
|
|
if (IS_ERR(fh))
|
2018-01-19 06:03:23 +07:00
|
|
|
goto out_err;
|
|
|
|
|
|
|
|
err = ovl_check_fh_len(fh, len);
|
|
|
|
if (err)
|
|
|
|
goto out_err;
|
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
flags = fh->fb.flags;
|
2018-01-20 02:33:44 +07:00
|
|
|
dentry = (flags & OVL_FH_FLAG_PATH_UPPER) ?
|
|
|
|
ovl_upper_fh_to_d(sb, fh) :
|
|
|
|
ovl_lower_fh_to_d(sb, fh);
|
2018-01-19 06:03:23 +07:00
|
|
|
err = PTR_ERR(dentry);
|
|
|
|
if (IS_ERR(dentry) && err != -ESTALE)
|
|
|
|
goto out_err;
|
|
|
|
|
2019-11-15 18:33:03 +07:00
|
|
|
out:
|
|
|
|
/* We may have needed to re-align OVL_FILEID_V0 */
|
|
|
|
if (!IS_ERR_OR_NULL(fh) && fh != (void *)fid)
|
|
|
|
kfree(fh);
|
|
|
|
|
2018-01-19 06:03:23 +07:00
|
|
|
return dentry;
|
|
|
|
|
|
|
|
out_err:
|
2019-12-16 18:12:32 +07:00
|
|
|
pr_warn_ratelimited("failed to decode file handle (len=%d, type=%d, flags=%x, err=%i)\n",
|
2019-11-15 18:33:03 +07:00
|
|
|
fh_len, fh_type, flags, err);
|
|
|
|
dentry = ERR_PTR(err);
|
|
|
|
goto out;
|
2018-01-19 06:03:23 +07:00
|
|
|
}
|
|
|
|
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
static struct dentry *ovl_fh_to_parent(struct super_block *sb, struct fid *fid,
|
|
|
|
int fh_len, int fh_type)
|
|
|
|
{
|
2019-12-16 18:12:32 +07:00
|
|
|
pr_warn_ratelimited("connectable file handles not supported; use 'no_subtree_check' exportfs option.\n");
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
return ERR_PTR(-EACCES);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int ovl_get_name(struct dentry *parent, char *name,
|
|
|
|
struct dentry *child)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* ovl_fh_to_dentry() returns connected dir overlay dentries and
|
|
|
|
* ovl_fh_to_parent() is not implemented, so we should not get here.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(1);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct dentry *ovl_get_parent(struct dentry *dentry)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* ovl_fh_to_dentry() returns connected dir overlay dentries, so we
|
|
|
|
* should not get here.
|
|
|
|
*/
|
|
|
|
WARN_ON_ONCE(1);
|
|
|
|
return ERR_PTR(-EIO);
|
|
|
|
}
|
|
|
|
|
2017-07-12 18:17:16 +07:00
|
|
|
const struct export_operations ovl_export_operations = {
|
2018-02-02 15:42:03 +07:00
|
|
|
.encode_fh = ovl_encode_fh,
|
2018-01-19 06:03:23 +07:00
|
|
|
.fh_to_dentry = ovl_fh_to_dentry,
|
ovl: decode connected upper dir file handles
Until this change, we decoded upper file handles by instantiating an
overlay dentry from the real upper dentry. This is sufficient to handle
pure upper files, but insufficient to handle merge/impure dirs.
To that end, if decoded real upper dir is connected and hashed, we
lookup an overlay dentry with the same path as the real upper dir.
If decoded real upper is non-dir, we instantiate a disconnected overlay
dentry as before this change.
Because ovl_fh_to_dentry() returns a connected overlay dir dentry,
exportfs never needs to call get_parent() and get_name() to reconnect an
upper overlay dir. Because connectable non-dir file handles are not
supported, exportfs will not be able to use fh_to_parent() and get_name()
methods to reconnect a disconnected non-dir to its parent. Therefore, the
methods get_parent() and get_name() are implemented just to print out a
sanity warning and the method fh_to_parent() is implemented to warn the
user that using the 'subtree_check' exportfs option is not supported.
An alternative approach could have been to implement instantiating of
an overlay directory inode from origin/index and implement get_parent()
and get_name() by calling into underlying fs operations and them
instantiating the overlay parent dir.
The reasons for not choosing the get_parent() approach were:
- Obtaining a disconnected overlay dir dentry would requires a
delicate re-factoring of ovl_lookup() to get a dentry with overlay
parent info. It was preferred to avoid doing that re-factoring unless
it was proven worthy.
- Going down the path of disconnected dir would mean that the (non
trivial) code path of d_splice_alias() could be traveled and that
meant writing more tests and introduces race cases that are very hard
to hit on purpose. Taking the path of connecting overlay dentry by
forward lookup is therefore the safe and boring way to avoid surprises.
The culprits of the chosen "connected overlay dentry" approach:
- We need to take special care to rename of ancestors while connecting
the overlay dentry by real dentry path. These subtleties are usually
handled by generic exportfs and VFS code.
- In a hypothetical workload, we could end up in a loop trying to connect,
interrupted by rename and restarting connect forever.
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
2017-12-28 23:36:16 +07:00
|
|
|
.fh_to_parent = ovl_fh_to_parent,
|
|
|
|
.get_name = ovl_get_name,
|
|
|
|
.get_parent = ovl_get_parent,
|
2017-07-12 18:17:16 +07:00
|
|
|
};
|