2019-06-01 02:08:42 -06:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
2013-11-24 07:54:58 -07:00
|
|
|
/*
|
|
|
|
* fs/kernfs/dir.c - kernfs directory implementation
|
|
|
|
*
|
|
|
|
* Copyright (c) 2001-3 Patrick Mochel
|
|
|
|
* Copyright (c) 2007 SUSE Linux Products GmbH
|
|
|
|
* Copyright (c) 2007, 2013 Tejun Heo <tj@kernel.org>
|
|
|
|
*/
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:02:55 -07:00
|
|
|
#include <linux/sched.h>
|
2013-11-28 12:54:33 -07:00
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/namei.h>
|
|
|
|
#include <linux/idr.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/security.h>
|
|
|
|
#include <linux/hash.h>
|
|
|
|
|
|
|
|
#include "kernfs-internal.h"
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
DEFINE_MUTEX(kernfs_mutex);
|
2014-02-07 11:32:07 -07:00
|
|
|
static DEFINE_SPINLOCK(kernfs_rename_lock); /* kn->parent and ->name */
|
|
|
|
static char kernfs_pr_cont_buf[PATH_MAX]; /* protected by rename_lock */
|
2017-07-12 12:49:46 -06:00
|
|
|
static DEFINE_SPINLOCK(kernfs_idr_lock); /* root->ino_idr */
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
#define rb_to_kn(X) rb_entry((X), struct kernfs_node, rb)
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:03:00 -07:00
|
|
|
static bool kernfs_active(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
lockdep_assert_held(&kernfs_mutex);
|
|
|
|
return atomic_read(&kn->active) >= 0;
|
|
|
|
}
|
|
|
|
|
2014-02-03 12:02:59 -07:00
|
|
|
static bool kernfs_lockdep(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_DEBUG_LOCK_ALLOC
|
|
|
|
return kn->flags & KERNFS_LOCKDEP;
|
|
|
|
#else
|
|
|
|
return false;
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2014-02-07 11:32:07 -07:00
|
|
|
static int kernfs_name_locked(struct kernfs_node *kn, char *buf, size_t buflen)
|
|
|
|
{
|
2017-02-08 04:28:55 -07:00
|
|
|
if (!kn)
|
|
|
|
return strlcpy(buf, "(null)", buflen);
|
|
|
|
|
2014-02-07 11:32:07 -07:00
|
|
|
return strlcpy(buf, kn->parent ? kn->name : "/", buflen);
|
|
|
|
}
|
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
/* kernfs_node_depth - compute depth from @from to @to */
|
|
|
|
static size_t kernfs_depth(struct kernfs_node *from, struct kernfs_node *to)
|
2014-02-07 11:32:07 -07:00
|
|
|
{
|
2016-01-29 01:54:04 -07:00
|
|
|
size_t depth = 0;
|
2014-02-07 11:32:07 -07:00
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
while (to->parent && to != from) {
|
|
|
|
depth++;
|
|
|
|
to = to->parent;
|
|
|
|
}
|
|
|
|
return depth;
|
|
|
|
}
|
2014-02-07 11:32:07 -07:00
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
static struct kernfs_node *kernfs_common_ancestor(struct kernfs_node *a,
|
|
|
|
struct kernfs_node *b)
|
|
|
|
{
|
|
|
|
size_t da, db;
|
|
|
|
struct kernfs_root *ra = kernfs_root(a), *rb = kernfs_root(b);
|
|
|
|
|
|
|
|
if (ra != rb)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
da = kernfs_depth(ra->kn, a);
|
|
|
|
db = kernfs_depth(rb->kn, b);
|
|
|
|
|
|
|
|
while (da > db) {
|
|
|
|
a = a->parent;
|
|
|
|
da--;
|
|
|
|
}
|
|
|
|
while (db > da) {
|
|
|
|
b = b->parent;
|
|
|
|
db--;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* worst case b and a will be the same at root */
|
|
|
|
while (b != a) {
|
|
|
|
b = b->parent;
|
|
|
|
a = a->parent;
|
|
|
|
}
|
|
|
|
|
|
|
|
return a;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_path_from_node_locked - find a pseudo-absolute path to @kn_to,
|
|
|
|
* where kn_from is treated as root of the path.
|
|
|
|
* @kn_from: kernfs node which should be treated as root for the path
|
|
|
|
* @kn_to: kernfs node to which path is needed
|
|
|
|
* @buf: buffer to copy the path into
|
|
|
|
* @buflen: size of @buf
|
|
|
|
*
|
|
|
|
* We need to handle couple of scenarios here:
|
|
|
|
* [1] when @kn_from is an ancestor of @kn_to at some level
|
|
|
|
* kn_from: /n1/n2/n3
|
|
|
|
* kn_to: /n1/n2/n3/n4/n5
|
|
|
|
* result: /n4/n5
|
|
|
|
*
|
|
|
|
* [2] when @kn_from is on a different hierarchy and we need to find common
|
|
|
|
* ancestor between @kn_from and @kn_to.
|
|
|
|
* kn_from: /n1/n2/n3/n4
|
|
|
|
* kn_to: /n1/n2/n5
|
|
|
|
* result: /../../n5
|
|
|
|
* OR
|
|
|
|
* kn_from: /n1/n2/n3/n4/n5 [depth=5]
|
|
|
|
* kn_to: /n1/n2/n3 [depth=3]
|
|
|
|
* result: /../..
|
|
|
|
*
|
2017-02-08 04:28:55 -07:00
|
|
|
* [3] when @kn_to is NULL result will be "(null)"
|
|
|
|
*
|
2016-08-10 09:23:44 -06:00
|
|
|
* Returns the length of the full path. If the full length is equal to or
|
|
|
|
* greater than @buflen, @buf contains the truncated path with the trailing
|
|
|
|
* '\0'. On error, -errno is returned.
|
2016-01-29 01:54:04 -07:00
|
|
|
*/
|
|
|
|
static int kernfs_path_from_node_locked(struct kernfs_node *kn_to,
|
|
|
|
struct kernfs_node *kn_from,
|
|
|
|
char *buf, size_t buflen)
|
|
|
|
{
|
|
|
|
struct kernfs_node *kn, *common;
|
|
|
|
const char parent_str[] = "/..";
|
2016-08-10 09:23:44 -06:00
|
|
|
size_t depth_from, depth_to, len = 0;
|
|
|
|
int i, j;
|
2016-01-29 01:54:04 -07:00
|
|
|
|
2017-02-08 04:28:55 -07:00
|
|
|
if (!kn_to)
|
|
|
|
return strlcpy(buf, "(null)", buflen);
|
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
if (!kn_from)
|
|
|
|
kn_from = kernfs_root(kn_to)->kn;
|
|
|
|
|
|
|
|
if (kn_from == kn_to)
|
|
|
|
return strlcpy(buf, "/", buflen);
|
|
|
|
|
|
|
|
common = kernfs_common_ancestor(kn_from, kn_to);
|
|
|
|
if (WARN_ON(!common))
|
2016-08-10 09:23:44 -06:00
|
|
|
return -EINVAL;
|
2016-01-29 01:54:04 -07:00
|
|
|
|
|
|
|
depth_to = kernfs_depth(common, kn_to);
|
|
|
|
depth_from = kernfs_depth(common, kn_from);
|
|
|
|
|
|
|
|
if (buf)
|
|
|
|
buf[0] = '\0';
|
|
|
|
|
|
|
|
for (i = 0; i < depth_from; i++)
|
|
|
|
len += strlcpy(buf + len, parent_str,
|
|
|
|
len < buflen ? buflen - len : 0);
|
|
|
|
|
|
|
|
/* Calculate how many bytes we need for the rest */
|
2016-08-10 09:23:44 -06:00
|
|
|
for (i = depth_to - 1; i >= 0; i--) {
|
|
|
|
for (kn = kn_to, j = 0; j < i; j++)
|
|
|
|
kn = kn->parent;
|
|
|
|
len += strlcpy(buf + len, "/",
|
|
|
|
len < buflen ? buflen - len : 0);
|
|
|
|
len += strlcpy(buf + len, kn->name,
|
|
|
|
len < buflen ? buflen - len : 0);
|
2016-01-29 01:54:04 -07:00
|
|
|
}
|
2014-02-07 11:32:07 -07:00
|
|
|
|
2016-08-10 09:23:44 -06:00
|
|
|
return len;
|
2014-02-07 11:32:07 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_name - obtain the name of a given node
|
|
|
|
* @kn: kernfs_node of interest
|
|
|
|
* @buf: buffer to copy @kn's name into
|
|
|
|
* @buflen: size of @buf
|
|
|
|
*
|
|
|
|
* Copies the name of @kn into @buf of @buflen bytes. The behavior is
|
|
|
|
* similar to strlcpy(). It returns the length of @kn's name and if @buf
|
|
|
|
* isn't long enough, it's filled upto @buflen-1 and nul terminated.
|
|
|
|
*
|
2017-02-08 04:28:55 -07:00
|
|
|
* Fills buffer with "(null)" if @kn is NULL.
|
|
|
|
*
|
2014-02-07 11:32:07 -07:00
|
|
|
* This function can be called from any context.
|
|
|
|
*/
|
|
|
|
int kernfs_name(struct kernfs_node *kn, char *buf, size_t buflen)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&kernfs_rename_lock, flags);
|
|
|
|
ret = kernfs_name_locked(kn, buf, buflen);
|
|
|
|
spin_unlock_irqrestore(&kernfs_rename_lock, flags);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
/**
|
|
|
|
* kernfs_path_from_node - build path of node @to relative to @from.
|
|
|
|
* @from: parent kernfs_node relative to which we need to build the path
|
|
|
|
* @to: kernfs_node of interest
|
|
|
|
* @buf: buffer to copy @to's path into
|
|
|
|
* @buflen: size of @buf
|
|
|
|
*
|
|
|
|
* Builds @to's path relative to @from in @buf. @from and @to must
|
|
|
|
* be on the same kernfs-root. If @from is not parent of @to, then a relative
|
|
|
|
* path (which includes '..'s) as needed to reach from @from to @to is
|
|
|
|
* returned.
|
|
|
|
*
|
2016-08-10 09:23:44 -06:00
|
|
|
* Returns the length of the full path. If the full length is equal to or
|
|
|
|
* greater than @buflen, @buf contains the truncated path with the trailing
|
|
|
|
* '\0'. On error, -errno is returned.
|
2016-01-29 01:54:04 -07:00
|
|
|
*/
|
|
|
|
int kernfs_path_from_node(struct kernfs_node *to, struct kernfs_node *from,
|
|
|
|
char *buf, size_t buflen)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&kernfs_rename_lock, flags);
|
|
|
|
ret = kernfs_path_from_node_locked(to, from, buf, buflen);
|
|
|
|
spin_unlock_irqrestore(&kernfs_rename_lock, flags);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(kernfs_path_from_node);
|
|
|
|
|
2014-02-07 11:32:07 -07:00
|
|
|
/**
|
|
|
|
* pr_cont_kernfs_name - pr_cont name of a kernfs_node
|
|
|
|
* @kn: kernfs_node of interest
|
|
|
|
*
|
|
|
|
* This function can be called from any context.
|
|
|
|
*/
|
|
|
|
void pr_cont_kernfs_name(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&kernfs_rename_lock, flags);
|
|
|
|
|
|
|
|
kernfs_name_locked(kn, kernfs_pr_cont_buf, sizeof(kernfs_pr_cont_buf));
|
|
|
|
pr_cont("%s", kernfs_pr_cont_buf);
|
|
|
|
|
|
|
|
spin_unlock_irqrestore(&kernfs_rename_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* pr_cont_kernfs_path - pr_cont path of a kernfs_node
|
|
|
|
* @kn: kernfs_node of interest
|
|
|
|
*
|
|
|
|
* This function can be called from any context.
|
|
|
|
*/
|
|
|
|
void pr_cont_kernfs_path(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
unsigned long flags;
|
2016-01-29 01:54:04 -07:00
|
|
|
int sz;
|
2014-02-07 11:32:07 -07:00
|
|
|
|
|
|
|
spin_lock_irqsave(&kernfs_rename_lock, flags);
|
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
sz = kernfs_path_from_node_locked(kn, NULL, kernfs_pr_cont_buf,
|
|
|
|
sizeof(kernfs_pr_cont_buf));
|
|
|
|
if (sz < 0) {
|
|
|
|
pr_cont("(error)");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (sz >= sizeof(kernfs_pr_cont_buf)) {
|
|
|
|
pr_cont("(name too long)");
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
pr_cont("%s", kernfs_pr_cont_buf);
|
2014-02-07 11:32:07 -07:00
|
|
|
|
2016-01-29 01:54:04 -07:00
|
|
|
out:
|
2014-02-07 11:32:07 -07:00
|
|
|
spin_unlock_irqrestore(&kernfs_rename_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_get_parent - determine the parent node and pin it
|
|
|
|
* @kn: kernfs_node of interest
|
|
|
|
*
|
|
|
|
* Determines @kn's parent, pins and returns it. This function can be
|
|
|
|
* called from any context.
|
|
|
|
*/
|
|
|
|
struct kernfs_node *kernfs_get_parent(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
struct kernfs_node *parent;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(&kernfs_rename_lock, flags);
|
|
|
|
parent = kn->parent;
|
|
|
|
kernfs_get(parent);
|
|
|
|
spin_unlock_irqrestore(&kernfs_rename_lock, flags);
|
|
|
|
|
|
|
|
return parent;
|
|
|
|
}
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_name_hash
|
2013-11-28 12:54:33 -07:00
|
|
|
* @name: Null terminated string to hash
|
|
|
|
* @ns: Namespace tag to hash
|
|
|
|
*
|
|
|
|
* Returns 31 bit hash of ns + name (so it fits in an off_t )
|
|
|
|
*/
|
2013-12-11 12:11:58 -07:00
|
|
|
static unsigned int kernfs_name_hash(const char *name, const void *ns)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2016-06-10 08:51:30 -06:00
|
|
|
unsigned long hash = init_name_hash(ns);
|
2013-11-28 12:54:33 -07:00
|
|
|
unsigned int len = strlen(name);
|
|
|
|
while (len--)
|
|
|
|
hash = partial_name_hash(*name++, hash);
|
2016-06-10 08:51:30 -06:00
|
|
|
hash = end_name_hash(hash);
|
2013-11-28 12:54:33 -07:00
|
|
|
hash &= 0x7fffffffU;
|
|
|
|
/* Reserve hash numbers 0, 1 and INT_MAX for magic directory entries */
|
2014-03-05 09:10:52 -07:00
|
|
|
if (hash < 2)
|
2013-11-28 12:54:33 -07:00
|
|
|
hash += 2;
|
|
|
|
if (hash >= INT_MAX)
|
|
|
|
hash = INT_MAX - 1;
|
|
|
|
return hash;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static int kernfs_name_compare(unsigned int hash, const char *name,
|
|
|
|
const void *ns, const struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2014-12-05 15:41:33 -07:00
|
|
|
if (hash < kn->hash)
|
|
|
|
return -1;
|
|
|
|
if (hash > kn->hash)
|
|
|
|
return 1;
|
|
|
|
if (ns < kn->ns)
|
|
|
|
return -1;
|
|
|
|
if (ns > kn->ns)
|
|
|
|
return 1;
|
2013-12-11 12:11:54 -07:00
|
|
|
return strcmp(name, kn->name);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static int kernfs_sd_compare(const struct kernfs_node *left,
|
|
|
|
const struct kernfs_node *right)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:58 -07:00
|
|
|
return kernfs_name_compare(left->hash, left->name, left->ns, right);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_link_sibling - link kernfs_node into sibling rbtree
|
2013-12-11 12:11:53 -07:00
|
|
|
* @kn: kernfs_node of interest
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Link @kn into its sibling rbtree which starts from
|
2013-12-11 12:11:54 -07:00
|
|
|
* @kn->parent->dir.children.
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
|
|
|
* Locking:
|
2013-12-11 12:11:57 -07:00
|
|
|
* mutex_lock(kernfs_mutex)
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* 0 on susccess -EEXIST on failure.
|
|
|
|
*/
|
2013-12-11 12:11:58 -07:00
|
|
|
static int kernfs_link_sibling(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:54 -07:00
|
|
|
struct rb_node **node = &kn->parent->dir.children.rb_node;
|
2013-11-28 12:54:33 -07:00
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
|
|
|
|
while (*node) {
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *pos;
|
2013-11-28 12:54:33 -07:00
|
|
|
int result;
|
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
pos = rb_to_kn(*node);
|
2013-11-28 12:54:33 -07:00
|
|
|
parent = *node;
|
2013-12-11 12:11:58 -07:00
|
|
|
result = kernfs_sd_compare(kn, pos);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (result < 0)
|
2013-12-11 12:11:54 -07:00
|
|
|
node = &pos->rb.rb_left;
|
2013-11-28 12:54:33 -07:00
|
|
|
else if (result > 0)
|
2013-12-11 12:11:54 -07:00
|
|
|
node = &pos->rb.rb_right;
|
2013-11-28 12:54:33 -07:00
|
|
|
else
|
|
|
|
return -EEXIST;
|
|
|
|
}
|
2014-04-17 03:52:10 -06:00
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
/* add new node and rebalance the tree */
|
2013-12-11 12:11:54 -07:00
|
|
|
rb_link_node(&kn->rb, parent, node);
|
|
|
|
rb_insert_color(&kn->rb, &kn->parent->dir.children);
|
2014-04-17 03:52:10 -06:00
|
|
|
|
|
|
|
/* successfully added, account subdir number */
|
|
|
|
if (kernfs_type(kn) == KERNFS_DIR)
|
|
|
|
kn->parent->dir.subdirs++;
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_unlink_sibling - unlink kernfs_node from sibling rbtree
|
2013-12-11 12:11:53 -07:00
|
|
|
* @kn: kernfs_node of interest
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
* Try to unlink @kn from its sibling rbtree which starts from
|
|
|
|
* kn->parent->dir.children. Returns %true if @kn was actually
|
|
|
|
* removed, %false if @kn wasn't on the rbtree.
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
|
|
|
* Locking:
|
2013-12-11 12:11:57 -07:00
|
|
|
* mutex_lock(kernfs_mutex)
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
static bool kernfs_unlink_sibling(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
if (RB_EMPTY_NODE(&kn->rb))
|
|
|
|
return false;
|
|
|
|
|
2013-12-11 12:11:56 -07:00
|
|
|
if (kernfs_type(kn) == KERNFS_DIR)
|
2013-12-11 12:11:54 -07:00
|
|
|
kn->parent->dir.subdirs--;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
rb_erase(&kn->rb, &kn->parent->dir.children);
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
RB_CLEAR_NODE(&kn->rb);
|
|
|
|
return true;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_get_active - get an active reference to kernfs_node
|
2013-12-11 12:11:53 -07:00
|
|
|
* @kn: kernfs_node to get an active reference to
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Get an active reference of @kn. This function is noop if @kn
|
2013-11-28 12:54:33 -07:00
|
|
|
* is NULL.
|
|
|
|
*
|
|
|
|
* RETURNS:
|
2013-12-11 12:11:53 -07:00
|
|
|
* Pointer to @kn on success, NULL on failure.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:58 -07:00
|
|
|
struct kernfs_node *kernfs_get_active(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
if (unlikely(!kn))
|
2013-11-28 12:54:33 -07:00
|
|
|
return NULL;
|
|
|
|
|
2014-01-13 15:13:39 -07:00
|
|
|
if (!atomic_inc_unless_negative(&kn->active))
|
|
|
|
return NULL;
|
2014-01-10 06:57:25 -07:00
|
|
|
|
2014-02-03 12:02:59 -07:00
|
|
|
if (kernfs_lockdep(kn))
|
2014-01-13 15:13:39 -07:00
|
|
|
rwsem_acquire_read(&kn->dep_map, 0, 1, _RET_IP_);
|
|
|
|
return kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_put_active - put an active reference to kernfs_node
|
2013-12-11 12:11:53 -07:00
|
|
|
* @kn: kernfs_node to put an active reference to
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Put an active reference to @kn. This function is noop if @kn
|
2013-11-28 12:54:33 -07:00
|
|
|
* is NULL.
|
|
|
|
*/
|
2013-12-11 12:11:58 -07:00
|
|
|
void kernfs_put_active(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2014-02-03 12:02:55 -07:00
|
|
|
struct kernfs_root *root = kernfs_root(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
int v;
|
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
if (unlikely(!kn))
|
2013-11-28 12:54:33 -07:00
|
|
|
return;
|
|
|
|
|
2014-02-03 12:02:59 -07:00
|
|
|
if (kernfs_lockdep(kn))
|
2013-12-11 12:11:53 -07:00
|
|
|
rwsem_release(&kn->dep_map, 1, _RET_IP_);
|
2013-12-11 12:11:54 -07:00
|
|
|
v = atomic_dec_return(&kn->active);
|
2013-12-11 12:11:56 -07:00
|
|
|
if (likely(v != KN_DEACTIVATED_BIAS))
|
2013-11-28 12:54:33 -07:00
|
|
|
return;
|
|
|
|
|
2014-02-03 12:02:55 -07:00
|
|
|
wake_up_all(&root->deactivate_waitq);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2014-02-03 12:03:00 -07:00
|
|
|
* kernfs_drain - drain kernfs_node
|
|
|
|
* @kn: kernfs_node to drain
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2014-02-03 12:03:00 -07:00
|
|
|
* Drain existing usages and nuke all existing mmaps of @kn. Mutiple
|
|
|
|
* removers may invoke this function concurrently on @kn and all will
|
|
|
|
* return after draining is complete.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2014-02-03 12:03:00 -07:00
|
|
|
static void kernfs_drain(struct kernfs_node *kn)
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
__releases(&kernfs_mutex) __acquires(&kernfs_mutex)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2014-02-03 12:02:55 -07:00
|
|
|
struct kernfs_root *root = kernfs_root(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
lockdep_assert_held(&kernfs_mutex);
|
2014-02-03 12:03:00 -07:00
|
|
|
WARN_ON_ONCE(kernfs_active(kn));
|
2014-01-10 06:57:19 -07:00
|
|
|
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2014-02-03 12:02:55 -07:00
|
|
|
|
2014-02-03 12:02:59 -07:00
|
|
|
if (kernfs_lockdep(kn)) {
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
rwsem_acquire(&kn->dep_map, 0, 0, _RET_IP_);
|
|
|
|
if (atomic_read(&kn->active) != KN_DEACTIVATED_BIAS)
|
|
|
|
lock_contended(&kn->dep_map, _RET_IP_);
|
|
|
|
}
|
2014-02-03 12:02:55 -07:00
|
|
|
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
/* but everyone should wait for draining */
|
2014-02-03 12:02:55 -07:00
|
|
|
wait_event(root->deactivate_waitq,
|
|
|
|
atomic_read(&kn->active) == KN_DEACTIVATED_BIAS);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:02:59 -07:00
|
|
|
if (kernfs_lockdep(kn)) {
|
2014-02-03 12:02:54 -07:00
|
|
|
lock_acquired(&kn->dep_map, _RET_IP_);
|
|
|
|
rwsem_release(&kn->dep_map, 1, _RET_IP_);
|
|
|
|
}
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
|
2016-12-27 12:49:03 -07:00
|
|
|
kernfs_drain_open_files(kn);
|
2014-02-03 12:02:57 -07:00
|
|
|
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_get - get a reference count on a kernfs_node
|
|
|
|
* @kn: the target kernfs_node
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
void kernfs_get(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
if (kn) {
|
2013-12-11 12:11:54 -07:00
|
|
|
WARN_ON(!atomic_read(&kn->count));
|
|
|
|
atomic_inc(&kn->count);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(kernfs_get);
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_put - put a reference count on a kernfs_node
|
|
|
|
* @kn: the target kernfs_node
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Put a reference count of @kn and destroy it if it reached zero.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
void kernfs_put(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *parent;
|
2013-11-28 12:54:40 -07:00
|
|
|
struct kernfs_root *root;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2017-07-12 12:49:48 -06:00
|
|
|
/*
|
|
|
|
* kernfs_node is freed with ->count 0, kernfs_find_and_get_node_by_ino
|
|
|
|
* depends on this to filter reused stale node
|
|
|
|
*/
|
2013-12-11 12:11:54 -07:00
|
|
|
if (!kn || !atomic_dec_and_test(&kn->count))
|
2013-11-28 12:54:33 -07:00
|
|
|
return;
|
2013-12-11 12:11:53 -07:00
|
|
|
root = kernfs_root(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
repeat:
|
2014-02-03 12:03:00 -07:00
|
|
|
/*
|
|
|
|
* Moving/renaming is always done while holding reference.
|
2013-12-11 12:11:54 -07:00
|
|
|
* kn->parent won't change beneath us.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:54 -07:00
|
|
|
parent = kn->parent;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:03:00 -07:00
|
|
|
WARN_ONCE(atomic_read(&kn->active) != KN_DEACTIVATED_BIAS,
|
|
|
|
"kernfs_put: %s/%s: released with incorrect active_ref %d\n",
|
|
|
|
parent ? parent->name : "", kn->name, atomic_read(&kn->active));
|
2013-12-11 12:11:53 -07:00
|
|
|
|
2013-12-11 12:11:56 -07:00
|
|
|
if (kernfs_type(kn) == KERNFS_LINK)
|
2013-12-11 12:11:54 -07:00
|
|
|
kernfs_put(kn->symlink.target_kn);
|
2015-02-13 15:36:31 -07:00
|
|
|
|
|
|
|
kfree_const(kn->name);
|
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
if (kn->iattr) {
|
|
|
|
simple_xattrs_free(&kn->iattr->xattrs);
|
2019-02-05 21:55:42 -07:00
|
|
|
kmem_cache_free(kernfs_iattrs_cache, kn->iattr);
|
2013-11-23 15:40:02 -07:00
|
|
|
}
|
2017-07-12 12:49:46 -06:00
|
|
|
spin_lock(&kernfs_idr_lock);
|
2017-07-12 12:49:50 -06:00
|
|
|
idr_remove(&root->ino_idr, kn->id.ino);
|
2017-07-12 12:49:46 -06:00
|
|
|
spin_unlock(&kernfs_idr_lock);
|
2013-12-11 12:11:57 -07:00
|
|
|
kmem_cache_free(kernfs_node_cache, kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
kn = parent;
|
|
|
|
if (kn) {
|
2013-12-11 12:11:54 -07:00
|
|
|
if (atomic_dec_and_test(&kn->count))
|
2013-11-28 12:54:40 -07:00
|
|
|
goto repeat;
|
|
|
|
} else {
|
2013-12-11 12:11:53 -07:00
|
|
|
/* just released the root kn, free @root too */
|
2017-07-12 12:49:46 -06:00
|
|
|
idr_destroy(&root->ino_idr);
|
2013-11-28 12:54:40 -07:00
|
|
|
kfree(root);
|
|
|
|
}
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(kernfs_put);
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
if (flags & LOOKUP_RCU)
|
|
|
|
return -ECHILD;
|
|
|
|
|
2013-12-11 14:02:59 -07:00
|
|
|
/* Always perform fresh lookup for negatives */
|
2015-03-17 16:25:59 -06:00
|
|
|
if (d_really_is_negative(dentry))
|
2013-12-11 14:02:59 -07:00
|
|
|
goto out_bad_unlocked;
|
|
|
|
|
2017-07-12 12:49:49 -06:00
|
|
|
kn = kernfs_dentry_node(dentry);
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:03:00 -07:00
|
|
|
/* The kernfs node has been deactivated */
|
|
|
|
if (!kernfs_active(kn))
|
2013-11-28 12:54:33 -07:00
|
|
|
goto out_bad;
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
/* The kernfs node has been moved? */
|
2017-07-12 12:49:49 -06:00
|
|
|
if (kernfs_dentry_node(dentry->d_parent) != kn->parent)
|
2013-11-28 12:54:33 -07:00
|
|
|
goto out_bad;
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
/* The kernfs node has been renamed */
|
2013-12-11 12:11:54 -07:00
|
|
|
if (strcmp(dentry->d_name.name, kn->name) != 0)
|
2013-11-28 12:54:33 -07:00
|
|
|
goto out_bad;
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
/* The kernfs node has been moved to a different namespace */
|
2013-12-11 12:11:54 -07:00
|
|
|
if (kn->parent && kernfs_ns_enabled(kn->parent) &&
|
2013-12-11 12:11:55 -07:00
|
|
|
kernfs_info(dentry->d_sb)->ns != kn->ns)
|
2013-11-28 12:54:33 -07:00
|
|
|
goto out_bad;
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
return 1;
|
|
|
|
out_bad:
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-12-11 14:02:59 -07:00
|
|
|
out_bad_unlocked:
|
2013-11-28 12:54:33 -07:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
const struct dentry_operations kernfs_dops = {
|
2013-12-11 12:11:58 -07:00
|
|
|
.d_revalidate = kernfs_dop_revalidate,
|
2013-11-28 12:54:33 -07:00
|
|
|
};
|
|
|
|
|
2014-02-03 12:09:15 -07:00
|
|
|
/**
|
|
|
|
* kernfs_node_from_dentry - determine kernfs_node associated with a dentry
|
|
|
|
* @dentry: the dentry in question
|
|
|
|
*
|
|
|
|
* Return the kernfs_node associated with @dentry. If @dentry is not a
|
|
|
|
* kernfs one, %NULL is returned.
|
|
|
|
*
|
|
|
|
* While the returned kernfs_node will stay accessible as long as @dentry
|
|
|
|
* is accessible, the returned node can be in any state and the caller is
|
|
|
|
* fully responsible for determining what's accessible.
|
|
|
|
*/
|
|
|
|
struct kernfs_node *kernfs_node_from_dentry(struct dentry *dentry)
|
|
|
|
{
|
2017-07-12 12:49:49 -06:00
|
|
|
if (dentry->d_sb->s_op == &kernfs_sops &&
|
|
|
|
!d_really_is_negative(dentry))
|
|
|
|
return kernfs_dentry_node(dentry);
|
2014-02-03 12:09:15 -07:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-01-17 07:58:25 -07:00
|
|
|
static struct kernfs_node *__kernfs_new_node(struct kernfs_root *root,
|
2019-02-22 07:57:18 -07:00
|
|
|
struct kernfs_node *parent,
|
2014-01-17 07:58:25 -07:00
|
|
|
const char *name, umode_t mode,
|
2018-07-20 15:56:47 -06:00
|
|
|
kuid_t uid, kgid_t gid,
|
2014-01-17 07:58:25 -07:00
|
|
|
unsigned flags)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2017-07-12 12:49:47 -06:00
|
|
|
u32 gen;
|
|
|
|
int cursor;
|
2013-11-28 12:54:41 -07:00
|
|
|
int ret;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2015-02-13 15:36:31 -07:00
|
|
|
name = kstrdup_const(name, GFP_KERNEL);
|
|
|
|
if (!name)
|
|
|
|
return NULL;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
kn = kmem_cache_zalloc(kernfs_node_cache, GFP_KERNEL);
|
2013-12-11 12:11:53 -07:00
|
|
|
if (!kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
goto err_out1;
|
|
|
|
|
2017-07-12 12:49:46 -06:00
|
|
|
idr_preload(GFP_KERNEL);
|
|
|
|
spin_lock(&kernfs_idr_lock);
|
2017-07-12 12:49:47 -06:00
|
|
|
cursor = idr_get_cursor(&root->ino_idr);
|
|
|
|
ret = idr_alloc_cyclic(&root->ino_idr, kn, 1, 0, GFP_ATOMIC);
|
|
|
|
if (ret >= 0 && ret < cursor)
|
|
|
|
root->next_generation++;
|
|
|
|
gen = root->next_generation;
|
2017-07-12 12:49:46 -06:00
|
|
|
spin_unlock(&kernfs_idr_lock);
|
|
|
|
idr_preload_end();
|
2013-11-28 12:54:41 -07:00
|
|
|
if (ret < 0)
|
2013-11-28 12:54:33 -07:00
|
|
|
goto err_out2;
|
2017-07-12 12:49:50 -06:00
|
|
|
kn->id.ino = ret;
|
|
|
|
kn->id.generation = gen;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2017-07-12 12:49:48 -06:00
|
|
|
/*
|
2019-04-16 06:17:11 -06:00
|
|
|
* set ino first. This RELEASE is paired with atomic_inc_not_zero in
|
2017-07-12 12:49:48 -06:00
|
|
|
* kernfs_find_and_get_node_by_ino
|
|
|
|
*/
|
2019-04-16 06:17:11 -06:00
|
|
|
atomic_set_release(&kn->count, 1);
|
2014-02-03 12:03:00 -07:00
|
|
|
atomic_set(&kn->active, KN_DEACTIVATED_BIAS);
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
RB_CLEAR_NODE(&kn->rb);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
kn->name = name;
|
|
|
|
kn->mode = mode;
|
2014-02-03 12:03:00 -07:00
|
|
|
kn->flags = flags;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2018-07-20 15:56:47 -06:00
|
|
|
if (!uid_eq(uid, GLOBAL_ROOT_UID) || !gid_eq(gid, GLOBAL_ROOT_GID)) {
|
|
|
|
struct iattr iattr = {
|
|
|
|
.ia_valid = ATTR_UID | ATTR_GID,
|
|
|
|
.ia_uid = uid,
|
|
|
|
.ia_gid = gid,
|
|
|
|
};
|
|
|
|
|
|
|
|
ret = __kernfs_setattr(kn, &iattr);
|
|
|
|
if (ret < 0)
|
|
|
|
goto err_out3;
|
|
|
|
}
|
|
|
|
|
2019-02-22 07:57:18 -07:00
|
|
|
if (parent) {
|
|
|
|
ret = security_kernfs_init_security(parent, kn);
|
|
|
|
if (ret)
|
|
|
|
goto err_out3;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
return kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2018-07-20 15:56:47 -06:00
|
|
|
err_out3:
|
|
|
|
idr_remove(&root->ino_idr, kn->id.ino);
|
2013-11-28 12:54:33 -07:00
|
|
|
err_out2:
|
2013-12-11 12:11:57 -07:00
|
|
|
kmem_cache_free(kernfs_node_cache, kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
err_out1:
|
2015-02-13 15:36:31 -07:00
|
|
|
kfree_const(name);
|
2013-11-28 12:54:33 -07:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2014-01-17 07:58:25 -07:00
|
|
|
struct kernfs_node *kernfs_new_node(struct kernfs_node *parent,
|
|
|
|
const char *name, umode_t mode,
|
2018-07-20 15:56:47 -06:00
|
|
|
kuid_t uid, kgid_t gid,
|
2014-01-17 07:58:25 -07:00
|
|
|
unsigned flags)
|
|
|
|
{
|
|
|
|
struct kernfs_node *kn;
|
|
|
|
|
2019-02-22 07:57:18 -07:00
|
|
|
kn = __kernfs_new_node(kernfs_root(parent), parent,
|
2018-07-20 15:56:47 -06:00
|
|
|
name, mode, uid, gid, flags);
|
2014-01-17 07:58:25 -07:00
|
|
|
if (kn) {
|
|
|
|
kernfs_get(parent);
|
|
|
|
kn->parent = parent;
|
|
|
|
}
|
|
|
|
return kn;
|
|
|
|
}
|
|
|
|
|
2017-07-12 12:49:48 -06:00
|
|
|
/*
|
|
|
|
* kernfs_find_and_get_node_by_ino - get kernfs_node from inode number
|
|
|
|
* @root: the kernfs root
|
|
|
|
* @ino: inode number
|
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* NULL on failure. Return a kernfs node with reference counter incremented
|
|
|
|
*/
|
|
|
|
struct kernfs_node *kernfs_find_and_get_node_by_ino(struct kernfs_root *root,
|
|
|
|
unsigned int ino)
|
|
|
|
{
|
|
|
|
struct kernfs_node *kn;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
kn = idr_find(&root->ino_idr, ino);
|
|
|
|
if (!kn)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Since kernfs_node is freed in RCU, it's possible an old node for ino
|
|
|
|
* is freed, but reused before RCU grace period. But a freed node (see
|
|
|
|
* kernfs_put) or an incompletedly initialized node (see
|
|
|
|
* __kernfs_new_node) should have 'count' 0. We can use this fact to
|
|
|
|
* filter out such node.
|
|
|
|
*/
|
|
|
|
if (!atomic_inc_not_zero(&kn->count)) {
|
|
|
|
kn = NULL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The node could be a new node or a reused node. If it's a new node,
|
|
|
|
* we are ok. If it's reused because of RCU (because of
|
|
|
|
* SLAB_TYPESAFE_BY_RCU), the __kernfs_new_node always sets its 'ino'
|
|
|
|
* before 'count'. So if 'count' is uptodate, 'ino' should be uptodate,
|
|
|
|
* hence we can use 'ino' to filter stale node.
|
|
|
|
*/
|
2017-07-12 12:49:50 -06:00
|
|
|
if (kn->id.ino != ino)
|
2017-07-12 12:49:48 -06:00
|
|
|
goto out;
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return kn;
|
|
|
|
out:
|
|
|
|
rcu_read_unlock();
|
|
|
|
kernfs_put(kn);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_add_one - add kernfs_node to parent without warning
|
2013-12-11 12:11:53 -07:00
|
|
|
* @kn: kernfs_node to be added
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2014-01-17 07:58:25 -07:00
|
|
|
* The caller must already have initialized @kn->parent. This
|
|
|
|
* function increments nlink of the parent's inode if @kn is a
|
|
|
|
* directory and link into the children list of the parent.
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* 0 on success, -EEXIST if entry with the given name already
|
|
|
|
* exists.
|
|
|
|
*/
|
2014-02-03 12:02:58 -07:00
|
|
|
int kernfs_add_one(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2014-01-17 07:58:25 -07:00
|
|
|
struct kernfs_node *parent = kn->parent;
|
2013-12-11 12:11:55 -07:00
|
|
|
struct kernfs_iattrs *ps_iattr;
|
2014-02-03 12:02:58 -07:00
|
|
|
bool has_ns;
|
2013-11-28 12:54:33 -07:00
|
|
|
int ret;
|
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
|
|
|
|
ret = -EINVAL;
|
|
|
|
has_ns = kernfs_ns_enabled(parent);
|
|
|
|
if (WARN(has_ns != (bool)kn->ns, KERN_WARNING "kernfs: ns %s in '%s' for '%s'\n",
|
|
|
|
has_ns ? "required" : "invalid", parent->name, kn->name))
|
|
|
|
goto out_unlock;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:56 -07:00
|
|
|
if (kernfs_type(parent) != KERNFS_DIR)
|
2014-02-03 12:02:58 -07:00
|
|
|
goto out_unlock;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
ret = -ENOENT;
|
2015-05-13 15:09:29 -06:00
|
|
|
if (parent->flags & KERNFS_EMPTY_DIR)
|
|
|
|
goto out_unlock;
|
|
|
|
|
2014-02-03 12:09:12 -07:00
|
|
|
if ((parent->flags & KERNFS_ACTIVATED) && !kernfs_active(parent))
|
2014-02-03 12:02:58 -07:00
|
|
|
goto out_unlock;
|
2014-01-13 15:36:03 -07:00
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
kn->hash = kernfs_name_hash(kn->name, kn->ns);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
ret = kernfs_link_sibling(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (ret)
|
2014-02-03 12:02:58 -07:00
|
|
|
goto out_unlock;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
/* Update timestamps on the parent */
|
2013-12-11 12:11:54 -07:00
|
|
|
ps_iattr = parent->iattr;
|
2013-11-28 12:54:33 -07:00
|
|
|
if (ps_iattr) {
|
2019-02-22 07:57:12 -07:00
|
|
|
ktime_get_real_ts64(&ps_iattr->ia_ctime);
|
|
|
|
ps_iattr->ia_mtime = ps_iattr->ia_ctime;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
2014-02-03 12:09:12 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Activate the new node unless CREATE_DEACTIVATED is requested.
|
|
|
|
* If not activated here, the kernfs user is responsible for
|
|
|
|
* activating the node with kernfs_activate(). A node which hasn't
|
|
|
|
* been activated is not visible to userland and its removal won't
|
|
|
|
* trigger deactivation.
|
|
|
|
*/
|
|
|
|
if (!(kernfs_root(kn)->flags & KERNFS_ROOT_CREATE_DEACTIVATED))
|
|
|
|
kernfs_activate(kn);
|
|
|
|
return 0;
|
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
out_unlock:
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2014-02-03 12:02:58 -07:00
|
|
|
return ret;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_find_ns - find kernfs_node with the given name
|
|
|
|
* @parent: kernfs_node to search under
|
2013-11-28 12:54:33 -07:00
|
|
|
* @name: name to look for
|
|
|
|
* @ns: the namespace tag to use
|
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Look for kernfs_node with name @name under @parent. Returns pointer to
|
|
|
|
* the found kernfs_node on success, %NULL on failure.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
static struct kernfs_node *kernfs_find_ns(struct kernfs_node *parent,
|
|
|
|
const unsigned char *name,
|
|
|
|
const void *ns)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:54 -07:00
|
|
|
struct rb_node *node = parent->dir.children.rb_node;
|
2013-11-29 15:19:09 -07:00
|
|
|
bool has_ns = kernfs_ns_enabled(parent);
|
2013-11-28 12:54:33 -07:00
|
|
|
unsigned int hash;
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
lockdep_assert_held(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
if (has_ns != (bool)ns) {
|
2013-12-11 12:11:58 -07:00
|
|
|
WARN(1, KERN_WARNING "kernfs: ns %s in '%s' for '%s'\n",
|
2013-12-11 12:11:54 -07:00
|
|
|
has_ns ? "required" : "invalid", parent->name, name);
|
2013-11-28 12:54:33 -07:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
hash = kernfs_name_hash(name, ns);
|
2013-11-28 12:54:33 -07:00
|
|
|
while (node) {
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
int result;
|
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
kn = rb_to_kn(node);
|
2013-12-11 12:11:58 -07:00
|
|
|
result = kernfs_name_compare(hash, name, ns, kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (result < 0)
|
|
|
|
node = node->rb_left;
|
|
|
|
else if (result > 0)
|
|
|
|
node = node->rb_right;
|
|
|
|
else
|
2013-12-11 12:11:53 -07:00
|
|
|
return kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2015-11-20 13:55:52 -07:00
|
|
|
static struct kernfs_node *kernfs_walk_ns(struct kernfs_node *parent,
|
|
|
|
const unsigned char *path,
|
|
|
|
const void *ns)
|
|
|
|
{
|
2016-01-15 10:30:14 -07:00
|
|
|
size_t len;
|
|
|
|
char *p, *name;
|
2015-11-20 13:55:52 -07:00
|
|
|
|
|
|
|
lockdep_assert_held(&kernfs_mutex);
|
|
|
|
|
2016-01-15 10:30:14 -07:00
|
|
|
/* grab kernfs_rename_lock to piggy back on kernfs_pr_cont_buf */
|
|
|
|
spin_lock_irq(&kernfs_rename_lock);
|
|
|
|
|
|
|
|
len = strlcpy(kernfs_pr_cont_buf, path, sizeof(kernfs_pr_cont_buf));
|
|
|
|
|
|
|
|
if (len >= sizeof(kernfs_pr_cont_buf)) {
|
|
|
|
spin_unlock_irq(&kernfs_rename_lock);
|
2015-11-20 13:55:52 -07:00
|
|
|
return NULL;
|
2016-01-15 10:30:14 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
p = kernfs_pr_cont_buf;
|
2015-11-20 13:55:52 -07:00
|
|
|
|
|
|
|
while ((name = strsep(&p, "/")) && parent) {
|
|
|
|
if (*name == '\0')
|
|
|
|
continue;
|
|
|
|
parent = kernfs_find_ns(parent, name, ns);
|
|
|
|
}
|
|
|
|
|
2016-01-15 10:30:14 -07:00
|
|
|
spin_unlock_irq(&kernfs_rename_lock);
|
|
|
|
|
2015-11-20 13:55:52 -07:00
|
|
|
return parent;
|
|
|
|
}
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
/**
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_find_and_get_ns - find and get kernfs_node with the given name
|
|
|
|
* @parent: kernfs_node to search under
|
2013-11-28 12:54:33 -07:00
|
|
|
* @name: name to look for
|
|
|
|
* @ns: the namespace tag to use
|
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Look for kernfs_node with name @name under @parent and get a reference
|
2013-11-28 12:54:33 -07:00
|
|
|
* if found. This function may sleep and returns pointer to the found
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_node on success, %NULL on failure.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kernfs_find_and_get_ns(struct kernfs_node *parent,
|
|
|
|
const char *name, const void *ns)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-12-11 12:11:53 -07:00
|
|
|
kn = kernfs_find_ns(parent, name, ns);
|
|
|
|
kernfs_get(kn);
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
return kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(kernfs_find_and_get_ns);
|
|
|
|
|
2015-11-20 13:55:52 -07:00
|
|
|
/**
|
|
|
|
* kernfs_walk_and_get_ns - find and get kernfs_node with the given path
|
|
|
|
* @parent: kernfs_node to search under
|
|
|
|
* @path: path to look for
|
|
|
|
* @ns: the namespace tag to use
|
|
|
|
*
|
|
|
|
* Look for kernfs_node with path @path under @parent and get a reference
|
|
|
|
* if found. This function may sleep and returns pointer to the found
|
|
|
|
* kernfs_node on success, %NULL on failure.
|
|
|
|
*/
|
|
|
|
struct kernfs_node *kernfs_walk_and_get_ns(struct kernfs_node *parent,
|
|
|
|
const char *path, const void *ns)
|
|
|
|
{
|
|
|
|
struct kernfs_node *kn;
|
|
|
|
|
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
kn = kernfs_walk_ns(parent, path, ns);
|
|
|
|
kernfs_get(kn);
|
|
|
|
mutex_unlock(&kernfs_mutex);
|
|
|
|
|
|
|
|
return kn;
|
|
|
|
}
|
|
|
|
|
2013-11-28 12:54:40 -07:00
|
|
|
/**
|
|
|
|
* kernfs_create_root - create a new kernfs hierarchy
|
2014-02-03 12:09:09 -07:00
|
|
|
* @scops: optional syscall operations for the hierarchy
|
2014-02-03 12:09:12 -07:00
|
|
|
* @flags: KERNFS_ROOT_* flags
|
2013-11-28 12:54:40 -07:00
|
|
|
* @priv: opaque data associated with the new directory
|
|
|
|
*
|
|
|
|
* Returns the root of the new hierarchy on success, ERR_PTR() value on
|
|
|
|
* failure.
|
|
|
|
*/
|
2014-02-03 12:09:09 -07:00
|
|
|
struct kernfs_root *kernfs_create_root(struct kernfs_syscall_ops *scops,
|
2014-02-03 12:09:12 -07:00
|
|
|
unsigned int flags, void *priv)
|
2013-11-28 12:54:40 -07:00
|
|
|
{
|
|
|
|
struct kernfs_root *root;
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:40 -07:00
|
|
|
|
|
|
|
root = kzalloc(sizeof(*root), GFP_KERNEL);
|
|
|
|
if (!root)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
2017-07-12 12:49:46 -06:00
|
|
|
idr_init(&root->ino_idr);
|
2014-04-09 09:07:30 -06:00
|
|
|
INIT_LIST_HEAD(&root->supers);
|
2017-07-12 12:49:47 -06:00
|
|
|
root->next_generation = 1;
|
2013-11-28 12:54:41 -07:00
|
|
|
|
2019-02-22 07:57:18 -07:00
|
|
|
kn = __kernfs_new_node(root, NULL, "", S_IFDIR | S_IRUGO | S_IXUGO,
|
2018-07-20 15:56:47 -06:00
|
|
|
GLOBAL_ROOT_UID, GLOBAL_ROOT_GID,
|
2014-01-17 07:58:25 -07:00
|
|
|
KERNFS_DIR);
|
2013-12-11 12:11:53 -07:00
|
|
|
if (!kn) {
|
2017-07-12 12:49:46 -06:00
|
|
|
idr_destroy(&root->ino_idr);
|
2013-11-28 12:54:40 -07:00
|
|
|
kfree(root);
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
kn->priv = priv;
|
2013-12-11 12:11:54 -07:00
|
|
|
kn->dir.root = root;
|
2013-11-28 12:54:40 -07:00
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
root->syscall_ops = scops;
|
2014-02-03 12:09:12 -07:00
|
|
|
root->flags = flags;
|
2013-12-11 12:11:53 -07:00
|
|
|
root->kn = kn;
|
2014-02-03 12:02:55 -07:00
|
|
|
init_waitqueue_head(&root->deactivate_waitq);
|
2013-11-28 12:54:40 -07:00
|
|
|
|
2014-02-03 12:09:12 -07:00
|
|
|
if (!(root->flags & KERNFS_ROOT_CREATE_DEACTIVATED))
|
|
|
|
kernfs_activate(kn);
|
|
|
|
|
2013-11-28 12:54:40 -07:00
|
|
|
return root;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_destroy_root - destroy a kernfs hierarchy
|
|
|
|
* @root: root of the hierarchy to destroy
|
|
|
|
*
|
|
|
|
* Destroy the hierarchy anchored at @root by removing all existing
|
|
|
|
* directories and destroying @root.
|
|
|
|
*/
|
|
|
|
void kernfs_destroy_root(struct kernfs_root *root)
|
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
kernfs_remove(root->kn); /* will also free @root */
|
2013-11-28 12:54:40 -07:00
|
|
|
}
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
/**
|
|
|
|
* kernfs_create_dir_ns - create a directory
|
|
|
|
* @parent: parent in which to create a new directory
|
|
|
|
* @name: name of the new directory
|
2013-12-11 14:02:55 -07:00
|
|
|
* @mode: mode of the new directory
|
2018-07-20 15:56:47 -06:00
|
|
|
* @uid: uid of the new directory
|
|
|
|
* @gid: gid of the new directory
|
2013-11-28 12:54:33 -07:00
|
|
|
* @priv: opaque data associated with the new directory
|
|
|
|
* @ns: optional namespace tag of the directory
|
|
|
|
*
|
|
|
|
* Returns the created node on success, ERR_PTR() value on failure.
|
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kernfs_create_dir_ns(struct kernfs_node *parent,
|
2013-12-11 14:02:55 -07:00
|
|
|
const char *name, umode_t mode,
|
2018-07-20 15:56:47 -06:00
|
|
|
kuid_t uid, kgid_t gid,
|
2013-12-11 14:02:55 -07:00
|
|
|
void *priv, const void *ns)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
int rc;
|
|
|
|
|
|
|
|
/* allocate */
|
2018-07-20 15:56:47 -06:00
|
|
|
kn = kernfs_new_node(parent, name, mode | S_IFDIR,
|
|
|
|
uid, gid, KERNFS_DIR);
|
2013-12-11 12:11:53 -07:00
|
|
|
if (!kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
kn->dir.root = parent->dir.root;
|
|
|
|
kn->ns = ns;
|
2013-12-11 12:11:53 -07:00
|
|
|
kn->priv = priv;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
/* link in */
|
2014-02-03 12:02:58 -07:00
|
|
|
rc = kernfs_add_one(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!rc)
|
2013-12-11 12:11:53 -07:00
|
|
|
return kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
kernfs_put(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
return ERR_PTR(rc);
|
|
|
|
}
|
|
|
|
|
2015-05-13 15:09:29 -06:00
|
|
|
/**
|
|
|
|
* kernfs_create_empty_dir - create an always empty directory
|
|
|
|
* @parent: parent in which to create a new directory
|
|
|
|
* @name: name of the new directory
|
|
|
|
*
|
|
|
|
* Returns the created node on success, ERR_PTR() value on failure.
|
|
|
|
*/
|
|
|
|
struct kernfs_node *kernfs_create_empty_dir(struct kernfs_node *parent,
|
|
|
|
const char *name)
|
|
|
|
{
|
|
|
|
struct kernfs_node *kn;
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
/* allocate */
|
2018-07-20 15:56:47 -06:00
|
|
|
kn = kernfs_new_node(parent, name, S_IRUGO|S_IXUGO|S_IFDIR,
|
|
|
|
GLOBAL_ROOT_UID, GLOBAL_ROOT_GID, KERNFS_DIR);
|
2015-05-13 15:09:29 -06:00
|
|
|
if (!kn)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
|
|
|
kn->flags |= KERNFS_EMPTY_DIR;
|
|
|
|
kn->dir.root = parent->dir.root;
|
|
|
|
kn->ns = NULL;
|
|
|
|
kn->priv = NULL;
|
|
|
|
|
|
|
|
/* link in */
|
|
|
|
rc = kernfs_add_one(kn);
|
|
|
|
if (!rc)
|
|
|
|
return kn;
|
|
|
|
|
|
|
|
kernfs_put(kn);
|
|
|
|
return ERR_PTR(rc);
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static struct dentry *kernfs_iop_lookup(struct inode *dir,
|
|
|
|
struct dentry *dentry,
|
|
|
|
unsigned int flags)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 14:02:59 -07:00
|
|
|
struct dentry *ret;
|
2017-07-12 12:49:49 -06:00
|
|
|
struct kernfs_node *parent = dir->i_private;
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
struct inode *inode;
|
|
|
|
const void *ns = NULL;
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
if (kernfs_ns_enabled(parent))
|
2013-12-11 12:11:55 -07:00
|
|
|
ns = kernfs_info(dir->i_sb)->ns;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
kn = kernfs_find_ns(parent, dentry->d_name.name, ns);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
/* no such entry */
|
2014-02-03 12:09:11 -07:00
|
|
|
if (!kn || !kernfs_active(kn)) {
|
2013-12-11 14:02:59 -07:00
|
|
|
ret = NULL;
|
2013-11-28 12:54:33 -07:00
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* attach dentry and inode */
|
2013-12-11 12:11:58 -07:00
|
|
|
inode = kernfs_get_inode(dir->i_sb, kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!inode) {
|
|
|
|
ret = ERR_PTR(-ENOMEM);
|
|
|
|
goto out_unlock;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* instantiate and hash dentry */
|
2014-10-12 20:24:21 -06:00
|
|
|
ret = d_splice_alias(inode, dentry);
|
2013-11-28 12:54:33 -07:00
|
|
|
out_unlock:
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-12-11 14:03:00 -07:00
|
|
|
static int kernfs_iop_mkdir(struct inode *dir, struct dentry *dentry,
|
|
|
|
umode_t mode)
|
|
|
|
{
|
|
|
|
struct kernfs_node *parent = dir->i_private;
|
2014-02-03 12:09:09 -07:00
|
|
|
struct kernfs_syscall_ops *scops = kernfs_root(parent)->syscall_ops;
|
2014-02-03 12:09:08 -07:00
|
|
|
int ret;
|
2013-12-11 14:03:00 -07:00
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
if (!scops || !scops->mkdir)
|
2013-12-11 14:03:00 -07:00
|
|
|
return -EPERM;
|
|
|
|
|
2014-02-03 12:09:08 -07:00
|
|
|
if (!kernfs_get_active(parent))
|
|
|
|
return -ENODEV;
|
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
ret = scops->mkdir(parent, dentry->d_name.name, mode);
|
2014-02-03 12:09:08 -07:00
|
|
|
|
|
|
|
kernfs_put_active(parent);
|
|
|
|
return ret;
|
2013-12-11 14:03:00 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
static int kernfs_iop_rmdir(struct inode *dir, struct dentry *dentry)
|
|
|
|
{
|
2017-07-12 12:49:49 -06:00
|
|
|
struct kernfs_node *kn = kernfs_dentry_node(dentry);
|
2014-02-03 12:09:09 -07:00
|
|
|
struct kernfs_syscall_ops *scops = kernfs_root(kn)->syscall_ops;
|
2014-02-03 12:09:08 -07:00
|
|
|
int ret;
|
2013-12-11 14:03:00 -07:00
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
if (!scops || !scops->rmdir)
|
2013-12-11 14:03:00 -07:00
|
|
|
return -EPERM;
|
|
|
|
|
2014-02-03 12:09:08 -07:00
|
|
|
if (!kernfs_get_active(kn))
|
|
|
|
return -ENODEV;
|
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
ret = scops->rmdir(kn);
|
2014-02-03 12:09:08 -07:00
|
|
|
|
|
|
|
kernfs_put_active(kn);
|
|
|
|
return ret;
|
2013-12-11 14:03:00 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
static int kernfs_iop_rename(struct inode *old_dir, struct dentry *old_dentry,
|
fs: make remaining filesystems use .rename2
This is trivial to do:
- add flags argument to foo_rename()
- check if flags is zero
- assign foo_rename() to .rename2 instead of .rename
This doesn't mean it's impossible to support RENAME_NOREPLACE for these
filesystems, but it is not trivial, like for local filesystems.
RENAME_NOREPLACE must guarantee atomicity (i.e. it shouldn't be possible
for a file to be created on one host while it is overwritten by rename on
another host).
Filesystems converted:
9p, afs, ceph, coda, ecryptfs, kernfs, lustre, ncpfs, nfs, ocfs2, orangefs.
After this, we can get rid of the duplicate interfaces for rename.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: David Howells <dhowells@redhat.com> [AFS]
Acked-by: Mike Marshall <hubcap@omnibond.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: Jan Harkes <jaharkes@cs.cmu.edu>
Cc: Tyler Hicks <tyhicks@canonical.com>
Cc: Oleg Drokin <oleg.drokin@intel.com>
Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: Mark Fasheh <mfasheh@suse.com>
2016-09-27 03:03:58 -06:00
|
|
|
struct inode *new_dir, struct dentry *new_dentry,
|
|
|
|
unsigned int flags)
|
2013-12-11 14:03:00 -07:00
|
|
|
{
|
2017-07-12 12:49:49 -06:00
|
|
|
struct kernfs_node *kn = kernfs_dentry_node(old_dentry);
|
2013-12-11 14:03:00 -07:00
|
|
|
struct kernfs_node *new_parent = new_dir->i_private;
|
2014-02-03 12:09:09 -07:00
|
|
|
struct kernfs_syscall_ops *scops = kernfs_root(kn)->syscall_ops;
|
2014-02-03 12:09:08 -07:00
|
|
|
int ret;
|
2013-12-11 14:03:00 -07:00
|
|
|
|
fs: make remaining filesystems use .rename2
This is trivial to do:
- add flags argument to foo_rename()
- check if flags is zero
- assign foo_rename() to .rename2 instead of .rename
This doesn't mean it's impossible to support RENAME_NOREPLACE for these
filesystems, but it is not trivial, like for local filesystems.
RENAME_NOREPLACE must guarantee atomicity (i.e. it shouldn't be possible
for a file to be created on one host while it is overwritten by rename on
another host).
Filesystems converted:
9p, afs, ceph, coda, ecryptfs, kernfs, lustre, ncpfs, nfs, ocfs2, orangefs.
After this, we can get rid of the duplicate interfaces for rename.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: David Howells <dhowells@redhat.com> [AFS]
Acked-by: Mike Marshall <hubcap@omnibond.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: Jan Harkes <jaharkes@cs.cmu.edu>
Cc: Tyler Hicks <tyhicks@canonical.com>
Cc: Oleg Drokin <oleg.drokin@intel.com>
Cc: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: Mark Fasheh <mfasheh@suse.com>
2016-09-27 03:03:58 -06:00
|
|
|
if (flags)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
if (!scops || !scops->rename)
|
2013-12-11 14:03:00 -07:00
|
|
|
return -EPERM;
|
|
|
|
|
2014-02-03 12:09:08 -07:00
|
|
|
if (!kernfs_get_active(kn))
|
|
|
|
return -ENODEV;
|
|
|
|
|
|
|
|
if (!kernfs_get_active(new_parent)) {
|
|
|
|
kernfs_put_active(kn);
|
|
|
|
return -ENODEV;
|
|
|
|
}
|
|
|
|
|
2014-02-03 12:09:09 -07:00
|
|
|
ret = scops->rename(kn, new_parent, new_dentry->d_name.name);
|
2014-02-03 12:09:08 -07:00
|
|
|
|
|
|
|
kernfs_put_active(new_parent);
|
|
|
|
kernfs_put_active(kn);
|
|
|
|
return ret;
|
2013-12-11 14:03:00 -07:00
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
const struct inode_operations kernfs_dir_iops = {
|
2013-12-11 12:11:58 -07:00
|
|
|
.lookup = kernfs_iop_lookup,
|
|
|
|
.permission = kernfs_iop_permission,
|
|
|
|
.setattr = kernfs_iop_setattr,
|
|
|
|
.getattr = kernfs_iop_getattr,
|
|
|
|
.listxattr = kernfs_iop_listxattr,
|
2013-12-11 14:03:00 -07:00
|
|
|
|
|
|
|
.mkdir = kernfs_iop_mkdir,
|
|
|
|
.rmdir = kernfs_iop_rmdir,
|
|
|
|
.rename = kernfs_iop_rename,
|
2013-11-28 12:54:33 -07:00
|
|
|
};
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static struct kernfs_node *kernfs_leftmost_descendant(struct kernfs_node *pos)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *last;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
while (true) {
|
|
|
|
struct rb_node *rbn;
|
|
|
|
|
|
|
|
last = pos;
|
|
|
|
|
2013-12-11 12:11:56 -07:00
|
|
|
if (kernfs_type(pos) != KERNFS_DIR)
|
2013-11-28 12:54:33 -07:00
|
|
|
break;
|
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
rbn = rb_first(&pos->dir.children);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!rbn)
|
|
|
|
break;
|
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
pos = rb_to_kn(rbn);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
return last;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:58 -07:00
|
|
|
* kernfs_next_descendant_post - find the next descendant for post-order walk
|
2013-11-28 12:54:33 -07:00
|
|
|
* @pos: the current position (%NULL to initiate traversal)
|
2013-12-11 12:11:53 -07:00
|
|
|
* @root: kernfs_node whose descendants to walk
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
|
|
|
* Find the next descendant to visit for post-order traversal of @root's
|
|
|
|
* descendants. @root is included in the iteration and the last node to be
|
|
|
|
* visited.
|
|
|
|
*/
|
2013-12-11 12:11:58 -07:00
|
|
|
static struct kernfs_node *kernfs_next_descendant_post(struct kernfs_node *pos,
|
|
|
|
struct kernfs_node *root)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
|
|
|
struct rb_node *rbn;
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
lockdep_assert_held(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
/* if first iteration, visit leftmost descendant which may be root */
|
|
|
|
if (!pos)
|
2013-12-11 12:11:58 -07:00
|
|
|
return kernfs_leftmost_descendant(root);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
/* if we visited @root, we're done */
|
|
|
|
if (pos == root)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
/* if there's an unvisited sibling, visit its leftmost descendant */
|
2013-12-11 12:11:54 -07:00
|
|
|
rbn = rb_next(&pos->rb);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (rbn)
|
2013-12-11 12:11:58 -07:00
|
|
|
return kernfs_leftmost_descendant(rb_to_kn(rbn));
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
/* no sibling left, visit parent */
|
2013-12-11 12:11:54 -07:00
|
|
|
return pos->parent;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
2014-02-03 12:09:12 -07:00
|
|
|
/**
|
|
|
|
* kernfs_activate - activate a node which started deactivated
|
|
|
|
* @kn: kernfs_node whose subtree is to be activated
|
|
|
|
*
|
|
|
|
* If the root has KERNFS_ROOT_CREATE_DEACTIVATED set, a newly created node
|
|
|
|
* needs to be explicitly activated. A node which hasn't been activated
|
|
|
|
* isn't visible to userland and deactivation is skipped during its
|
|
|
|
* removal. This is useful to construct atomic init sequences where
|
|
|
|
* creation of multiple nodes should either succeed or fail atomically.
|
|
|
|
*
|
|
|
|
* The caller is responsible for ensuring that this function is not called
|
|
|
|
* after kernfs_remove*() is invoked on @kn.
|
|
|
|
*/
|
|
|
|
void kernfs_activate(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
struct kernfs_node *pos;
|
|
|
|
|
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
|
|
|
|
pos = NULL;
|
|
|
|
while ((pos = kernfs_next_descendant_post(pos, kn))) {
|
|
|
|
if (!pos || (pos->flags & KERNFS_ACTIVATED))
|
|
|
|
continue;
|
|
|
|
|
|
|
|
WARN_ON_ONCE(pos->parent && RB_EMPTY_NODE(&pos->rb));
|
|
|
|
WARN_ON_ONCE(atomic_read(&pos->active) != KN_DEACTIVATED_BIAS);
|
|
|
|
|
|
|
|
atomic_sub(KN_DEACTIVATED_BIAS, &pos->active);
|
|
|
|
pos->flags |= KERNFS_ACTIVATED;
|
|
|
|
}
|
|
|
|
|
|
|
|
mutex_unlock(&kernfs_mutex);
|
|
|
|
}
|
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
static void __kernfs_remove(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
struct kernfs_node *pos;
|
|
|
|
|
|
|
|
lockdep_assert_held(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
kernfs, sysfs, driver-core: implement kernfs_remove_self() and its wrappers
Sometimes it's necessary to implement a node which wants to delete
nodes including itself. This isn't straightforward because of kernfs
active reference. While a file operation is in progress, an active
reference is held and kernfs_remove() waits for all such references to
drain before completing. For a self-deleting node, this is a deadlock
as kernfs_remove() ends up waiting for an active reference that itself
is sitting on top of.
This currently is worked around in the sysfs layer using
sysfs_schedule_callback() which makes such removals asynchronous.
While it works, it's rather cumbersome and inherently breaks
synchronicity of the operation - the file operation which triggered
the operation may complete before the removal is finished (or even
started) and the removal may fail asynchronously. If a removal
operation is immmediately followed by another operation which expects
the specific name to be available (e.g. removal followed by rename
onto the same name), there's no way to make the latter operation
reliable.
The thing is there's no inherent reason for this to be asynchrnous.
All that's necessary to do this synchronous is a dedicated operation
which drops its own active ref and deactivates self. This patch
implements kernfs_remove_self() and its wrappers in sysfs and driver
core. kernfs_remove_self() is to be called from one of the file
operations, drops the active ref the task is holding, removes the self
node, and restores active ref to the dead node so that the ref is
balanced afterwards. __kernfs_remove() is updated so that it takes an
early exit if the target node is already fully removed so that the
active ref restored by kernfs_remove_self() after removal doesn't
confuse the deactivation path.
This makes implementing self-deleting nodes very easy. The normal
removal path doesn't even need to be changed to use
kernfs_remove_self() for the self-deleting node. The method can
invoke kernfs_remove_self() on itself before proceeding the normal
removal path. kernfs_remove() invoked on the node by the normal
deletion path will simply be ignored.
This will replace sysfs_schedule_callback(). A subtle feature of
sysfs_schedule_callback() is that it collapses multiple invocations -
even if multiple removals are triggered, the removal callback is run
only once. An equivalent effect can be achieved by testing the return
value of kernfs_remove_self() - only the one which gets %true return
value should proceed with actual deletion. All other instances of
kernfs_remove_self() will wait till the enclosing kernfs operation
which invoked the winning instance of kernfs_remove_self() finishes
and then return %false. This trivially makes all users of
kernfs_remove_self() automatically show correct synchronous behavior
even when there are multiple concurrent operations - all "echo 1 >
delete" instances will finish only after the whole operation is
completed by one of the instances.
Note that manipulation of active ref is implemented in separate public
functions - kernfs_[un]break_active_protection().
kernfs_remove_self() is the only user at the moment but this will be
used to cater to more complex cases.
v2: For !CONFIG_SYSFS, dummy version kernfs_remove_self() was missing
and sysfs_remove_file_self() had incorrect return type. Fix it.
Reported by kbuild test bot.
v3: kernfs_[un]break_active_protection() separated out from
kernfs_remove_self() and exposed as public API.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:03:01 -07:00
|
|
|
/*
|
|
|
|
* Short-circuit if non-root @kn has already finished removal.
|
|
|
|
* This is for kernfs_remove_self() which plays with active ref
|
|
|
|
* after removal.
|
|
|
|
*/
|
|
|
|
if (!kn || (kn->parent && RB_EMPTY_NODE(&kn->rb)))
|
2014-01-13 14:50:31 -07:00
|
|
|
return;
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
pr_debug("kernfs %s: removing\n", kn->name);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:03:00 -07:00
|
|
|
/* prevent any new usage under @kn by deactivating all nodes */
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
pos = NULL;
|
|
|
|
while ((pos = kernfs_next_descendant_post(pos, kn)))
|
2014-02-03 12:03:00 -07:00
|
|
|
if (kernfs_active(pos))
|
|
|
|
atomic_add(KN_DEACTIVATED_BIAS, &pos->active);
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
|
|
|
|
/* deactivate and unlink the subtree node-by-node */
|
2013-11-28 12:54:33 -07:00
|
|
|
do {
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
pos = kernfs_leftmost_descendant(kn);
|
|
|
|
|
|
|
|
/*
|
2014-02-03 12:03:00 -07:00
|
|
|
* kernfs_drain() drops kernfs_mutex temporarily and @pos's
|
|
|
|
* base ref could have been put by someone else by the time
|
|
|
|
* the function returns. Make sure it doesn't go away
|
|
|
|
* underneath us.
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
*/
|
|
|
|
kernfs_get(pos);
|
|
|
|
|
2014-02-03 12:09:12 -07:00
|
|
|
/*
|
|
|
|
* Drain iff @kn was activated. This avoids draining and
|
|
|
|
* its lockdep annotations for nodes which have never been
|
|
|
|
* activated and allows embedding kernfs_remove() in create
|
|
|
|
* error paths without worrying about draining.
|
|
|
|
*/
|
|
|
|
if (kn->flags & KERNFS_ACTIVATED)
|
|
|
|
kernfs_drain(pos);
|
|
|
|
else
|
|
|
|
WARN_ON_ONCE(atomic_read(&kn->active) != KN_DEACTIVATED_BIAS);
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* kernfs_unlink_sibling() succeeds once per node. Use it
|
|
|
|
* to decide who's responsible for cleanups.
|
|
|
|
*/
|
|
|
|
if (!pos->parent || kernfs_unlink_sibling(pos)) {
|
|
|
|
struct kernfs_iattrs *ps_iattr =
|
|
|
|
pos->parent ? pos->parent->iattr : NULL;
|
|
|
|
|
|
|
|
/* update timestamps on the parent */
|
|
|
|
if (ps_iattr) {
|
2019-02-22 07:57:12 -07:00
|
|
|
ktime_get_real_ts64(&ps_iattr->ia_ctime);
|
|
|
|
ps_iattr->ia_mtime = ps_iattr->ia_ctime;
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
}
|
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
kernfs_put(pos);
|
kernfs: restructure removal path to fix possible premature return
The recursive nature of kernfs_remove() means that, even if
kernfs_remove() is not allowed to be called multiple times on the same
node, there may be race conditions between removal of parent and its
descendants. While we can claim that kernfs_remove() shouldn't be
called on one of the descendants while the removal of an ancestor is
in progress, such rule is unnecessarily restrictive and very difficult
to enforce. It's better to simply allow invoking kernfs_remove() as
the caller sees fit as long as the caller ensures that the node is
accessible.
The current behavior in such situations is broken. Whoever enters
removal path first takes the node off the hierarchy and then
deactivates. Following removers either return as soon as it notices
that it's not the first one or can't even find the target node as it
has already been removed from the hierarchy. In both cases, the
following removers may finish prematurely while the nodes which should
be removed and drained are still being processed by the first one.
This patch restructures so that multiple removers, whether through
recursion or direction invocation, always follow the following rules.
* When there are multiple concurrent removers, only one puts the base
ref.
* Regardless of which one puts the base ref, all removers are blocked
until the target node is fully deactivated and removed.
To achieve the above, removal path now first marks all descendants
including self REMOVED and then deactivates and unlinks leftmost
descendant one-by-one. kernfs_deactivate() is called directly from
__kernfs_removal() and drops and regrabs kernfs_mutex for each
descendant to drain active refs. As this means that multiple removers
can enter kernfs_deactivate() for the same node, the function is
updated so that it can handle multiple deactivators of the same node -
only one actually deactivates but all wait till drain completion.
The restructured removal path guarantees that a removed node gets
unlinked only after the node is deactivated and drained. Combined
with proper multiple deactivator handling, this guarantees that any
invocation of kernfs_remove() returns only after the node itself and
all its descendants are deactivated, drained and removed.
v2: Draining separated into a separate loop (used to be in the same
loop as unlink) and done from __kernfs_deactivate(). This is to
allow exposing deactivation as a separate interface later.
Root node removal was broken in v1 patch. Fixed.
v3: Revert most of v2 except for root node removal fix and
simplification of KERNFS_REMOVED setting loop.
v4: Refreshed on top of ("kernfs: make kernfs_deactivate() honor
KERNFS_LOCKDEP flag").
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:02:56 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
kernfs_put(pos);
|
|
|
|
} while (pos != kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_remove - remove a kernfs_node recursively
|
|
|
|
* @kn: the kernfs_node to remove
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Remove @kn along with all its subdirectories and files.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
void kernfs_remove(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2014-02-03 12:02:58 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
__kernfs_remove(kn);
|
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
kernfs, sysfs, driver-core: implement kernfs_remove_self() and its wrappers
Sometimes it's necessary to implement a node which wants to delete
nodes including itself. This isn't straightforward because of kernfs
active reference. While a file operation is in progress, an active
reference is held and kernfs_remove() waits for all such references to
drain before completing. For a self-deleting node, this is a deadlock
as kernfs_remove() ends up waiting for an active reference that itself
is sitting on top of.
This currently is worked around in the sysfs layer using
sysfs_schedule_callback() which makes such removals asynchronous.
While it works, it's rather cumbersome and inherently breaks
synchronicity of the operation - the file operation which triggered
the operation may complete before the removal is finished (or even
started) and the removal may fail asynchronously. If a removal
operation is immmediately followed by another operation which expects
the specific name to be available (e.g. removal followed by rename
onto the same name), there's no way to make the latter operation
reliable.
The thing is there's no inherent reason for this to be asynchrnous.
All that's necessary to do this synchronous is a dedicated operation
which drops its own active ref and deactivates self. This patch
implements kernfs_remove_self() and its wrappers in sysfs and driver
core. kernfs_remove_self() is to be called from one of the file
operations, drops the active ref the task is holding, removes the self
node, and restores active ref to the dead node so that the ref is
balanced afterwards. __kernfs_remove() is updated so that it takes an
early exit if the target node is already fully removed so that the
active ref restored by kernfs_remove_self() after removal doesn't
confuse the deactivation path.
This makes implementing self-deleting nodes very easy. The normal
removal path doesn't even need to be changed to use
kernfs_remove_self() for the self-deleting node. The method can
invoke kernfs_remove_self() on itself before proceeding the normal
removal path. kernfs_remove() invoked on the node by the normal
deletion path will simply be ignored.
This will replace sysfs_schedule_callback(). A subtle feature of
sysfs_schedule_callback() is that it collapses multiple invocations -
even if multiple removals are triggered, the removal callback is run
only once. An equivalent effect can be achieved by testing the return
value of kernfs_remove_self() - only the one which gets %true return
value should proceed with actual deletion. All other instances of
kernfs_remove_self() will wait till the enclosing kernfs operation
which invoked the winning instance of kernfs_remove_self() finishes
and then return %false. This trivially makes all users of
kernfs_remove_self() automatically show correct synchronous behavior
even when there are multiple concurrent operations - all "echo 1 >
delete" instances will finish only after the whole operation is
completed by one of the instances.
Note that manipulation of active ref is implemented in separate public
functions - kernfs_[un]break_active_protection().
kernfs_remove_self() is the only user at the moment but this will be
used to cater to more complex cases.
v2: For !CONFIG_SYSFS, dummy version kernfs_remove_self() was missing
and sysfs_remove_file_self() had incorrect return type. Fix it.
Reported by kbuild test bot.
v3: kernfs_[un]break_active_protection() separated out from
kernfs_remove_self() and exposed as public API.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2014-02-03 12:03:01 -07:00
|
|
|
/**
|
|
|
|
* kernfs_break_active_protection - break out of active protection
|
|
|
|
* @kn: the self kernfs_node
|
|
|
|
*
|
|
|
|
* The caller must be running off of a kernfs operation which is invoked
|
|
|
|
* with an active reference - e.g. one of kernfs_ops. Each invocation of
|
|
|
|
* this function must also be matched with an invocation of
|
|
|
|
* kernfs_unbreak_active_protection().
|
|
|
|
*
|
|
|
|
* This function releases the active reference of @kn the caller is
|
|
|
|
* holding. Once this function is called, @kn may be removed at any point
|
|
|
|
* and the caller is solely responsible for ensuring that the objects it
|
|
|
|
* dereferences are accessible.
|
|
|
|
*/
|
|
|
|
void kernfs_break_active_protection(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Take out ourself out of the active ref dependency chain. If
|
|
|
|
* we're called without an active ref, lockdep will complain.
|
|
|
|
*/
|
|
|
|
kernfs_put_active(kn);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_unbreak_active_protection - undo kernfs_break_active_protection()
|
|
|
|
* @kn: the self kernfs_node
|
|
|
|
*
|
|
|
|
* If kernfs_break_active_protection() was called, this function must be
|
|
|
|
* invoked before finishing the kernfs operation. Note that while this
|
|
|
|
* function restores the active reference, it doesn't and can't actually
|
|
|
|
* restore the active protection - @kn may already or be in the process of
|
|
|
|
* being removed. Once kernfs_break_active_protection() is invoked, that
|
|
|
|
* protection is irreversibly gone for the kernfs operation instance.
|
|
|
|
*
|
|
|
|
* While this function may be called at any point after
|
|
|
|
* kernfs_break_active_protection() is invoked, its most useful location
|
|
|
|
* would be right before the enclosing kernfs operation returns.
|
|
|
|
*/
|
|
|
|
void kernfs_unbreak_active_protection(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* @kn->active could be in any state; however, the increment we do
|
|
|
|
* here will be undone as soon as the enclosing kernfs operation
|
|
|
|
* finishes and this temporary bump can't break anything. If @kn
|
|
|
|
* is alive, nothing changes. If @kn is being deactivated, the
|
|
|
|
* soon-to-follow put will either finish deactivation or restore
|
|
|
|
* deactivated state. If @kn is already removed, the temporary
|
|
|
|
* bump is guaranteed to be gone before @kn is released.
|
|
|
|
*/
|
|
|
|
atomic_inc(&kn->active);
|
|
|
|
if (kernfs_lockdep(kn))
|
|
|
|
rwsem_acquire(&kn->dep_map, 0, 1, _RET_IP_);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_remove_self - remove a kernfs_node from its own method
|
|
|
|
* @kn: the self kernfs_node to remove
|
|
|
|
*
|
|
|
|
* The caller must be running off of a kernfs operation which is invoked
|
|
|
|
* with an active reference - e.g. one of kernfs_ops. This can be used to
|
|
|
|
* implement a file operation which deletes itself.
|
|
|
|
*
|
|
|
|
* For example, the "delete" file for a sysfs device directory can be
|
|
|
|
* implemented by invoking kernfs_remove_self() on the "delete" file
|
|
|
|
* itself. This function breaks the circular dependency of trying to
|
|
|
|
* deactivate self while holding an active ref itself. It isn't necessary
|
|
|
|
* to modify the usual removal path to use kernfs_remove_self(). The
|
|
|
|
* "delete" implementation can simply invoke kernfs_remove_self() on self
|
|
|
|
* before proceeding with the usual removal path. kernfs will ignore later
|
|
|
|
* kernfs_remove() on self.
|
|
|
|
*
|
|
|
|
* kernfs_remove_self() can be called multiple times concurrently on the
|
|
|
|
* same kernfs_node. Only the first one actually performs removal and
|
|
|
|
* returns %true. All others will wait until the kernfs operation which
|
|
|
|
* won self-removal finishes and return %false. Note that the losers wait
|
|
|
|
* for the completion of not only the winning kernfs_remove_self() but also
|
|
|
|
* the whole kernfs_ops which won the arbitration. This can be used to
|
|
|
|
* guarantee, for example, all concurrent writes to a "delete" file to
|
|
|
|
* finish only after the whole operation is complete.
|
|
|
|
*/
|
|
|
|
bool kernfs_remove_self(struct kernfs_node *kn)
|
|
|
|
{
|
|
|
|
bool ret;
|
|
|
|
|
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
kernfs_break_active_protection(kn);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* SUICIDAL is used to arbitrate among competing invocations. Only
|
|
|
|
* the first one will actually perform removal. When the removal
|
|
|
|
* is complete, SUICIDED is set and the active ref is restored
|
|
|
|
* while holding kernfs_mutex. The ones which lost arbitration
|
|
|
|
* waits for SUICDED && drained which can happen only after the
|
|
|
|
* enclosing kernfs operation which executed the winning instance
|
|
|
|
* of kernfs_remove_self() finished.
|
|
|
|
*/
|
|
|
|
if (!(kn->flags & KERNFS_SUICIDAL)) {
|
|
|
|
kn->flags |= KERNFS_SUICIDAL;
|
|
|
|
__kernfs_remove(kn);
|
|
|
|
kn->flags |= KERNFS_SUICIDED;
|
|
|
|
ret = true;
|
|
|
|
} else {
|
|
|
|
wait_queue_head_t *waitq = &kernfs_root(kn)->deactivate_waitq;
|
|
|
|
DEFINE_WAIT(wait);
|
|
|
|
|
|
|
|
while (true) {
|
|
|
|
prepare_to_wait(waitq, &wait, TASK_UNINTERRUPTIBLE);
|
|
|
|
|
|
|
|
if ((kn->flags & KERNFS_SUICIDED) &&
|
|
|
|
atomic_read(&kn->active) == KN_DEACTIVATED_BIAS)
|
|
|
|
break;
|
|
|
|
|
|
|
|
mutex_unlock(&kernfs_mutex);
|
|
|
|
schedule();
|
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
}
|
|
|
|
finish_wait(waitq, &wait);
|
|
|
|
WARN_ON_ONCE(!RB_EMPTY_NODE(&kn->rb));
|
|
|
|
ret = false;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This must be done while holding kernfs_mutex; otherwise, waiting
|
|
|
|
* for SUICIDED && deactivated could finish prematurely.
|
|
|
|
*/
|
|
|
|
kernfs_unbreak_active_protection(kn);
|
|
|
|
|
|
|
|
mutex_unlock(&kernfs_mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
/**
|
2013-12-11 12:11:53 -07:00
|
|
|
* kernfs_remove_by_name_ns - find a kernfs_node by name and remove it
|
|
|
|
* @parent: parent of the target
|
|
|
|
* @name: name of the kernfs_node to remove
|
|
|
|
* @ns: namespace tag of the kernfs_node to remove
|
2013-11-28 12:54:33 -07:00
|
|
|
*
|
2013-12-11 12:11:53 -07:00
|
|
|
* Look for the kernfs_node with @name and @ns under @parent and remove it.
|
|
|
|
* Returns 0 on success, -ENOENT if such entry doesn't exist.
|
2013-11-28 12:54:33 -07:00
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
int kernfs_remove_by_name_ns(struct kernfs_node *parent, const char *name,
|
2013-11-28 12:54:33 -07:00
|
|
|
const void *ns)
|
|
|
|
{
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *kn;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
if (!parent) {
|
2013-12-11 12:11:58 -07:00
|
|
|
WARN(1, KERN_WARNING "kernfs: can not remove '%s', no directory\n",
|
2013-11-28 12:54:33 -07:00
|
|
|
name);
|
|
|
|
return -ENOENT;
|
|
|
|
}
|
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
kn = kernfs_find_ns(parent, name, ns);
|
|
|
|
if (kn)
|
2014-02-03 12:02:58 -07:00
|
|
|
__kernfs_remove(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-03 12:02:58 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
if (kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
return 0;
|
|
|
|
else
|
|
|
|
return -ENOENT;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* kernfs_rename_ns - move and rename a kernfs_node
|
2013-12-11 12:11:53 -07:00
|
|
|
* @kn: target node
|
2013-11-28 12:54:33 -07:00
|
|
|
* @new_parent: new parent to put @sd under
|
|
|
|
* @new_name: new name
|
|
|
|
* @new_ns: new namespace tag
|
|
|
|
*/
|
2013-12-11 12:11:53 -07:00
|
|
|
int kernfs_rename_ns(struct kernfs_node *kn, struct kernfs_node *new_parent,
|
2013-11-28 12:54:33 -07:00
|
|
|
const char *new_name, const void *new_ns)
|
|
|
|
{
|
2014-02-07 11:32:07 -07:00
|
|
|
struct kernfs_node *old_parent;
|
|
|
|
const char *old_name = NULL;
|
2013-11-28 12:54:33 -07:00
|
|
|
int error;
|
|
|
|
|
2014-02-07 11:32:07 -07:00
|
|
|
/* can't move or rename root */
|
|
|
|
if (!kn->parent)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2014-01-13 15:36:03 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
|
|
|
|
2013-12-11 14:02:56 -07:00
|
|
|
error = -ENOENT;
|
2015-05-13 15:09:29 -06:00
|
|
|
if (!kernfs_active(kn) || !kernfs_active(new_parent) ||
|
|
|
|
(new_parent->flags & KERNFS_EMPTY_DIR))
|
2013-12-11 14:02:56 -07:00
|
|
|
goto out;
|
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
error = 0;
|
2013-12-11 12:11:54 -07:00
|
|
|
if ((kn->parent == new_parent) && (kn->ns == new_ns) &&
|
|
|
|
(strcmp(kn->name, new_name) == 0))
|
2014-01-13 15:36:03 -07:00
|
|
|
goto out; /* nothing to rename */
|
2013-11-28 12:54:33 -07:00
|
|
|
|
|
|
|
error = -EEXIST;
|
|
|
|
if (kernfs_find_ns(new_parent, new_name, new_ns))
|
2014-01-13 15:36:03 -07:00
|
|
|
goto out;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
/* rename kernfs_node */
|
2013-12-11 12:11:54 -07:00
|
|
|
if (strcmp(kn->name, new_name) != 0) {
|
2013-11-28 12:54:33 -07:00
|
|
|
error = -ENOMEM;
|
2015-02-13 15:36:27 -07:00
|
|
|
new_name = kstrdup_const(new_name, GFP_KERNEL);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!new_name)
|
2014-01-13 15:36:03 -07:00
|
|
|
goto out;
|
2014-02-07 11:32:07 -07:00
|
|
|
} else {
|
|
|
|
new_name = NULL;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Move to the appropriate place in the appropriate directories rbtree.
|
|
|
|
*/
|
2013-12-11 12:11:58 -07:00
|
|
|
kernfs_unlink_sibling(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
kernfs_get(new_parent);
|
2014-02-07 11:32:07 -07:00
|
|
|
|
|
|
|
/* rename_lock protects ->parent and ->name accessors */
|
|
|
|
spin_lock_irq(&kernfs_rename_lock);
|
|
|
|
|
|
|
|
old_parent = kn->parent;
|
2013-12-11 12:11:54 -07:00
|
|
|
kn->parent = new_parent;
|
2014-02-07 11:32:07 -07:00
|
|
|
|
|
|
|
kn->ns = new_ns;
|
|
|
|
if (new_name) {
|
2015-02-13 15:36:31 -07:00
|
|
|
old_name = kn->name;
|
2014-02-07 11:32:07 -07:00
|
|
|
kn->name = new_name;
|
|
|
|
}
|
|
|
|
|
|
|
|
spin_unlock_irq(&kernfs_rename_lock);
|
|
|
|
|
2014-02-10 15:57:09 -07:00
|
|
|
kn->hash = kernfs_name_hash(kn->name, kn->ns);
|
2013-12-11 12:11:58 -07:00
|
|
|
kernfs_link_sibling(kn);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2014-02-07 11:32:07 -07:00
|
|
|
kernfs_put(old_parent);
|
2015-02-13 15:36:27 -07:00
|
|
|
kfree_const(old_name);
|
2014-02-07 11:32:07 -07:00
|
|
|
|
2013-11-28 12:54:33 -07:00
|
|
|
error = 0;
|
2014-01-13 15:36:03 -07:00
|
|
|
out:
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Relationship between s_mode and the DT_xxx types */
|
2013-12-11 12:11:53 -07:00
|
|
|
static inline unsigned char dt_type(struct kernfs_node *kn)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:54 -07:00
|
|
|
return (kn->mode >> 12) & 15;
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static int kernfs_dir_fop_release(struct inode *inode, struct file *filp)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
|
|
|
kernfs_put(filp->private_data);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static struct kernfs_node *kernfs_dir_pos(const void *ns,
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *parent, loff_t hash, struct kernfs_node *pos)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
|
|
|
if (pos) {
|
2014-02-03 12:03:00 -07:00
|
|
|
int valid = kernfs_active(pos) &&
|
2014-01-13 15:36:03 -07:00
|
|
|
pos->parent == parent && hash == pos->hash;
|
2013-11-28 12:54:33 -07:00
|
|
|
kernfs_put(pos);
|
|
|
|
if (!valid)
|
|
|
|
pos = NULL;
|
|
|
|
}
|
|
|
|
if (!pos && (hash > 1) && (hash < INT_MAX)) {
|
2013-12-11 12:11:54 -07:00
|
|
|
struct rb_node *node = parent->dir.children.rb_node;
|
2013-11-28 12:54:33 -07:00
|
|
|
while (node) {
|
2013-12-11 12:11:53 -07:00
|
|
|
pos = rb_to_kn(node);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
if (hash < pos->hash)
|
2013-11-28 12:54:33 -07:00
|
|
|
node = node->rb_left;
|
2013-12-11 12:11:54 -07:00
|
|
|
else if (hash > pos->hash)
|
2013-11-28 12:54:33 -07:00
|
|
|
node = node->rb_right;
|
|
|
|
else
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2014-02-03 12:09:11 -07:00
|
|
|
/* Skip over entries which are dying/dead or in the wrong namespace */
|
|
|
|
while (pos && (!kernfs_active(pos) || pos->ns != ns)) {
|
2013-12-11 12:11:54 -07:00
|
|
|
struct rb_node *node = rb_next(&pos->rb);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!node)
|
|
|
|
pos = NULL;
|
|
|
|
else
|
2013-12-11 12:11:53 -07:00
|
|
|
pos = rb_to_kn(node);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
|
|
|
return pos;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static struct kernfs_node *kernfs_dir_next_pos(const void *ns,
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *parent, ino_t ino, struct kernfs_node *pos)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
2013-12-11 12:11:58 -07:00
|
|
|
pos = kernfs_dir_pos(ns, parent, ino, pos);
|
2014-02-03 12:09:11 -07:00
|
|
|
if (pos) {
|
2013-11-28 12:54:33 -07:00
|
|
|
do {
|
2013-12-11 12:11:54 -07:00
|
|
|
struct rb_node *node = rb_next(&pos->rb);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!node)
|
|
|
|
pos = NULL;
|
|
|
|
else
|
2013-12-11 12:11:53 -07:00
|
|
|
pos = rb_to_kn(node);
|
2014-02-03 12:09:11 -07:00
|
|
|
} while (pos && (!kernfs_active(pos) || pos->ns != ns));
|
|
|
|
}
|
2013-11-28 12:54:33 -07:00
|
|
|
return pos;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
static int kernfs_fop_readdir(struct file *file, struct dir_context *ctx)
|
2013-11-28 12:54:33 -07:00
|
|
|
{
|
|
|
|
struct dentry *dentry = file->f_path.dentry;
|
2017-07-12 12:49:49 -06:00
|
|
|
struct kernfs_node *parent = kernfs_dentry_node(dentry);
|
2013-12-11 12:11:53 -07:00
|
|
|
struct kernfs_node *pos = file->private_data;
|
2013-11-28 12:54:33 -07:00
|
|
|
const void *ns = NULL;
|
|
|
|
|
|
|
|
if (!dir_emit_dots(file, ctx))
|
|
|
|
return 0;
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:53 -07:00
|
|
|
if (kernfs_ns_enabled(parent))
|
2013-12-11 12:11:55 -07:00
|
|
|
ns = kernfs_info(dentry->d_sb)->ns;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:58 -07:00
|
|
|
for (pos = kernfs_dir_pos(ns, parent, ctx->pos, pos);
|
2013-11-28 12:54:33 -07:00
|
|
|
pos;
|
2013-12-11 12:11:58 -07:00
|
|
|
pos = kernfs_dir_next_pos(ns, parent, ctx->pos, pos)) {
|
2013-12-11 12:11:54 -07:00
|
|
|
const char *name = pos->name;
|
2013-11-28 12:54:33 -07:00
|
|
|
unsigned int type = dt_type(pos);
|
|
|
|
int len = strlen(name);
|
2017-07-12 12:49:50 -06:00
|
|
|
ino_t ino = pos->id.ino;
|
2013-11-28 12:54:33 -07:00
|
|
|
|
2013-12-11 12:11:54 -07:00
|
|
|
ctx->pos = pos->hash;
|
2013-11-28 12:54:33 -07:00
|
|
|
file->private_data = pos;
|
|
|
|
kernfs_get(pos);
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
if (!dir_emit(ctx, name, len, ino, type))
|
|
|
|
return 0;
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_lock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
}
|
2013-12-11 12:11:57 -07:00
|
|
|
mutex_unlock(&kernfs_mutex);
|
2013-11-28 12:54:33 -07:00
|
|
|
file->private_data = NULL;
|
|
|
|
ctx->pos = INT_MAX;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2013-12-11 12:11:57 -07:00
|
|
|
const struct file_operations kernfs_dir_fops = {
|
2013-11-28 12:54:33 -07:00
|
|
|
.read = generic_read_dir,
|
2016-04-20 17:59:01 -06:00
|
|
|
.iterate_shared = kernfs_fop_readdir,
|
2013-12-11 12:11:58 -07:00
|
|
|
.release = kernfs_dir_fop_release,
|
2016-04-20 17:59:01 -06:00
|
|
|
.llseek = generic_file_llseek,
|
2013-11-28 12:54:33 -07:00
|
|
|
};
|