kernel_optimize_test/fs/notify/inotify/inotify_fsnotify.c

227 lines
6.5 KiB
C
Raw Normal View History

/*
* fs/inotify_user.c - inotify support for userspace
*
* Authors:
* John McCutchan <ttb@tentacle.dhs.org>
* Robert Love <rml@novell.com>
*
* Copyright (C) 2005 John McCutchan
* Copyright 2006 Hewlett-Packard Development Company, L.P.
*
* Copyright (C) 2009 Eric Paris <Red Hat Inc>
* inotify was largely rewriten to make use of the fsnotify infrastructure
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2, or (at your option) any
* later version.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*/
#include <linux/fs.h> /* struct inode */
#include <linux/fsnotify_backend.h>
#include <linux/inotify.h>
#include <linux/path.h> /* struct path */
#include <linux/slab.h> /* kmem_* */
#include <linux/types.h>
#include <linux/sched.h>
#include "inotify.h"
/*
* Check if 2 events contain the same information. We do not compare private data
* but at this moment that isn't a problem for any know fsnotify listeners.
*/
static bool event_compare(struct fsnotify_event *old, struct fsnotify_event *new)
{
if ((old->mask == new->mask) &&
(old->to_tell == new->to_tell) &&
(old->data_type == new->data_type) &&
(old->name_len == new->name_len)) {
switch (old->data_type) {
case (FSNOTIFY_EVENT_INODE):
/* remember, after old was put on the wait_q we aren't
* allowed to look at the inode any more, only thing
* left to check was if the file_name is the same */
if (!old->name_len ||
!strcmp(old->file_name, new->file_name))
return true;
break;
case (FSNOTIFY_EVENT_PATH):
if ((old->path.mnt == new->path.mnt) &&
(old->path.dentry == new->path.dentry))
return true;
break;
case (FSNOTIFY_EVENT_NONE):
if (old->mask & FS_Q_OVERFLOW)
return true;
else if (old->mask & FS_IN_IGNORED)
return false;
return true;
};
}
return false;
}
static int inotify_merge(struct list_head *list, struct fsnotify_event *event)
{
struct fsnotify_event_holder *last_holder;
struct fsnotify_event *last_event;
int ret = 0;
/* and the list better be locked by something too */
spin_lock(&event->lock);
last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list);
last_event = last_holder->event;
if (event_compare(last_event, event))
ret = -EEXIST;
spin_unlock(&event->lock);
return ret;
}
static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
{
struct fsnotify_mark *entry;
struct inotify_inode_mark_entry *ientry;
struct inode *to_tell;
struct inotify_event_private_data *event_priv;
struct fsnotify_event_private_data *fsn_event_priv;
int wd, ret;
to_tell = event->to_tell;
spin_lock(&to_tell->i_lock);
entry = fsnotify_find_mark_entry(group, to_tell);
spin_unlock(&to_tell->i_lock);
/* race with watch removal? We already passes should_send */
if (unlikely(!entry))
return 0;
ientry = container_of(entry, struct inotify_inode_mark_entry,
fsn_entry);
wd = ientry->wd;
event_priv = kmem_cache_alloc(event_priv_cachep, GFP_KERNEL);
if (unlikely(!event_priv))
return -ENOMEM;
fsn_event_priv = &event_priv->fsnotify_event_priv_data;
fsn_event_priv->group = group;
event_priv->wd = wd;
ret = fsnotify_add_notify_event(group, event, fsn_event_priv, inotify_merge);
if (ret) {
inotify_free_event_priv(fsn_event_priv);
/* EEXIST says we tail matched, EOVERFLOW isn't something
* to report up the stack. */
if ((ret == -EEXIST) ||
(ret == -EOVERFLOW))
ret = 0;
}
/*
* If we hold the entry until after the event is on the queue
* IN_IGNORED won't be able to pass this event in the queue
*/
fsnotify_put_mark(entry);
return ret;
}
static void inotify_freeing_mark(struct fsnotify_mark *entry, struct fsnotify_group *group)
{
inotify: inotify_destroy_mark_entry could get called twice inotify_destroy_mark_entry could get called twice for the same mark since it is called directly in inotify_rm_watch and when the mark is being destroyed for another reason. As an example assume that the file being watched was just deleted so inotify_destroy_mark_entry would get called from the path fsnotify_inoderemove() -> fsnotify_destroy_marks_by_inode() -> fsnotify_destroy_mark_entry() -> inotify_destroy_mark_entry(). If this happened at the same time as userspace tried to remove a watch via inotify_rm_watch we could attempt to remove the mark from the idr twice and could thus double dec the ref cnt and potentially could be in a use after free/double free situation. The fix is to have inotify_rm_watch use the generic recursive safe fsnotify_destroy_mark_by_entry() so we are sure the inotify_destroy_mark_entry() function can only be called one. This patch also renames the function to inotify_ingored_remove_idr() so it is clear what is actually going on in the function. Hopefully this fixes: [ 20.342058] idr_remove called for id=20 which is not allocated. [ 20.348000] Pid: 1860, comm: udevd Not tainted 2.6.30-tip #1077 [ 20.353933] Call Trace: [ 20.356410] [<ffffffff811a82b7>] idr_remove+0x115/0x18f [ 20.361737] [<ffffffff8134259d>] ? _spin_lock+0x6d/0x75 [ 20.367061] [<ffffffff8111640a>] ? inotify_destroy_mark_entry+0xa3/0xcf [ 20.373771] [<ffffffff8111641e>] inotify_destroy_mark_entry+0xb7/0xcf [ 20.380306] [<ffffffff81115913>] inotify_freeing_mark+0xe/0x10 [ 20.386238] [<ffffffff8111410d>] fsnotify_destroy_mark_by_entry+0x143/0x170 [ 20.393293] [<ffffffff811163a3>] inotify_destroy_mark_entry+0x3c/0xcf [ 20.399829] [<ffffffff811164d1>] sys_inotify_rm_watch+0x9b/0xc6 [ 20.405850] [<ffffffff8100bcdb>] system_call_fastpath+0x16/0x1b Reported-by: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Eric Paris <eparis@redhat.com> Tested-by: Peter Ziljlstra <peterz@infradead.org>
2009-06-13 04:04:26 +08:00
inotify_ignored_and_remove_idr(entry, group);
}
static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode,
struct vfsmount *mnt, __u32 mask, void *data,
int data_type)
{
struct fsnotify_mark *entry;
bool send;
spin_lock(&inode->i_lock);
entry = fsnotify_find_mark_entry(group, inode);
spin_unlock(&inode->i_lock);
if (!entry)
return false;
mask = (mask & ~FS_EVENT_ON_CHILD);
send = (entry->mask & mask);
/* find took a reference */
fsnotify_put_mark(entry);
return send;
}
/*
* This is NEVER supposed to be called. Inotify marks should either have been
* removed from the idr when the watch was removed or in the
* fsnotify_destroy_mark_by_group() call when the inotify instance was being
* torn down. This is only called if the idr is about to be freed but there
* are still marks in it.
*/
static int idr_callback(int id, void *p, void *data)
{
struct fsnotify_mark *entry;
struct inotify_inode_mark_entry *ientry;
static bool warned = false;
if (warned)
return 0;
warned = true;
entry = p;
ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry);
WARN(1, "inotify closing but id=%d for entry=%p in group=%p still in "
"idr. Probably leaking memory\n", id, p, data);
/*
* I'm taking the liberty of assuming that the mark in question is a
* valid address and I'm dereferencing it. This might help to figure
* out why we got here and the panic is no worse than the original
* BUG() that was here.
*/
if (entry)
printk(KERN_WARNING "entry->group=%p inode=%p wd=%d\n",
entry->group, entry->i.inode, ientry->wd);
return 0;
}
static void inotify_free_group_priv(struct fsnotify_group *group)
{
/* ideally the idr is empty and we won't hit the BUG in teh callback */
idr_for_each(&group->inotify_data.idr, idr_callback, group);
idr_remove_all(&group->inotify_data.idr);
idr_destroy(&group->inotify_data.idr);
free_uid(group->inotify_data.user);
}
void inotify_free_event_priv(struct fsnotify_event_private_data *fsn_event_priv)
{
struct inotify_event_private_data *event_priv;
event_priv = container_of(fsn_event_priv, struct inotify_event_private_data,
fsnotify_event_priv_data);
kmem_cache_free(event_priv_cachep, event_priv);
}
const struct fsnotify_ops inotify_fsnotify_ops = {
.handle_event = inotify_handle_event,
.should_send_event = inotify_should_send_event,
.free_group_priv = inotify_free_group_priv,
.free_event_priv = inotify_free_event_priv,
.freeing_mark = inotify_freeing_mark,
};