Release 4.7 net/unix/garbage.c

/*
 * NET3:        Garbage Collector For AF_UNIX sockets
 *
 * Garbage Collector:
 *      Copyright (C) Barak A. Pearlmutter.
 *      Released under the GPL version 2 or later.
 *
 * Chopped about by Alan Cox 22/3/96 to make it fit the AF_UNIX socket problem.
 * If it doesn't work blame me, it worked when Barak sent it.
 *
 * Assumptions:
 *
 *  - object w/ a bit
 *  - free list
 *
 * Current optimizations:
 *
 *  - explicit stack instead of recursion
 *  - tail recurse on first born instead of immediate push/pop
 *  - we gather the stuff that should not be killed into tree
 *    and stack is just a path from root to the current pointer.
 *
 *  Future optimizations:
 *
 *  - don't just push entire root set; process in place
 *
 *      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 of the License, or (at your option) any later version.
 *
 *  Fixes:
 *      Alan Cox        07 Sept 1997    Vmalloc internal stack as needed.
 *                                      Cope with changing max_files.
 *      Al Viro         11 Oct 1998
 *              Graph may have cycles. That is, we can send the descriptor
 *              of foo to bar and vice versa. Current code chokes on that.
 *              Fix: move SCM_RIGHTS ones into the separate list and then
 *              skb_free() them all instead of doing explicit fput's.
 *              Another problem: since fput() may block somebody may
 *              create a new unix_socket when we are in the middle of sweep
 *              phase. Fix: revert the logic wrt MARKED. Mark everything
 *              upon the beginning and unmark non-junk ones.
 *
 *              [12 Oct 1998] AAARGH! New code purges all SCM_RIGHTS
 *              sent to connect()'ed but still not accept()'ed sockets.
 *              Fixed. Old code had slightly different problem here:
 *              extra fput() in situation when we passed the descriptor via
 *              such socket and closed it (descriptor). That would happen on
 *              each unix_gc() until the accept(). Since the struct file in
 *              question would go to the free list and might be reused...
 *              That might be the reason of random oopses on filp_close()
 *              in unrelated processes.
 *
 *      AV              28 Feb 1999
 *              Kill the explicit allocation of stack. Now we keep the tree
 *              with root in dummy + pointer (gc_current) to one of the nodes.
 *              Stack is represented as path from gc_current to dummy. Unmark
 *              now means "add to tree". Push == "make it a son of gc_current".
 *              Pop == "move gc_current to parent". We keep only pointers to
 *              parents (->gc_tree).
 *      AV              1 Mar 1999
 *              Damn. Added missing check for ->dead in listen queues scanning.
 *
 *      Miklos Szeredi 25 Jun 2007
 *              Reimplement with a cycle collecting algorithm. This should
 *              solve several problems with the previous code, like being racy
 *              wrt receive and holding up unrelated socket operations.
 */

#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/socket.h>
#include <linux/un.h>
#include <linux/net.h>
#include <linux/fs.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/file.h>
#include <linux/proc_fs.h>
#include <linux/mutex.h>
#include <linux/wait.h>

#include <net/sock.h>
#include <net/af_unix.h>
#include <net/scm.h>
#include <net/tcp_states.h>

/* Internal data structures and random procedures: */

static LIST_HEAD(gc_inflight_list);
static LIST_HEAD(gc_candidates);
static DEFINE_SPINLOCK(unix_gc_lock);
static DECLARE_WAIT_QUEUE_HEAD(unix_gc_wait);


unsigned int unix_tot_inflight;



struct sock *unix_get_socket(struct file *filp)
{
	struct sock *u_sock = NULL;
	struct inode *inode = file_inode(filp);

	/* Socket ? */
	if (S_ISSOCK(inode->i_mode) && !(filp->f_mode & FMODE_PATH)) {
		struct socket *sock = SOCKET_I(inode);
		struct sock *s = sock->sk;

		/* PF_UNIX ? */
		if (s && sock->ops && sock->ops->family == PF_UNIX)
			u_sock = s;
	}
	return u_sock;
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
pre-git
pre-git
68
73.12%
4
40.00%
al viro
al viro
15
16.13%
3
30.00%
matthew wilcox
matthew wilcox
4
4.30%
1
10.00%
arnaldo carvalho de melo
arnaldo carvalho de melo
4
4.30%
1
10.00%
jason eastman
jason eastman
2
2.15%
1
10.00%
Total
93
100.00%
10
100.00%

/* Keep the number of times in flight count for the file
 * descriptor if it is for an AF_UNIX socket.
 */



void unix_inflight(struct user_struct *user, struct file *fp)
{
	struct sock *s = unix_get_socket(fp);

	spin_lock(&unix_gc_lock);

	if (s) {
		struct unix_sock *u = unix_sk(s);

		if (atomic_long_inc_return(&u->inflight) == 1) {
			BUG_ON(!list_empty(&u->link));
			list_add_tail(&u->link, &gc_inflight_list);
		} else {
			BUG_ON(list_empty(&u->link));
		}
		unix_tot_inflight++;
	}
	user->unix_inflight++;
	spin_unlock(&unix_gc_lock);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
58
51.79%
1
10.00%
pre-git
pre-git
30
26.79%
3
30.00%
willy tarreau
willy tarreau
12
10.71%
1
10.00%
hannes frederic sowa
hannes frederic sowa
5
4.46%
1
10.00%
david s. miller
david s. miller
3
2.68%
1
10.00%
arnaldo carvalho de melo
arnaldo carvalho de melo
2
1.79%
1
10.00%
pavel emelianov
pavel emelianov
1
0.89%
1
10.00%
al viro
al viro
1
0.89%
1
10.00%
Total
112
100.00%
10
100.00%



void unix_notinflight(struct user_struct *user, struct file *fp)
{
	struct sock *s = unix_get_socket(fp);

	spin_lock(&unix_gc_lock);

	if (s) {
		struct unix_sock *u = unix_sk(s);

		BUG_ON(list_empty(&u->link));

		if (atomic_long_dec_and_test(&u->inflight))
			list_del_init(&u->link);
		unix_tot_inflight--;
	}
	user->unix_inflight--;
	spin_unlock(&unix_gc_lock);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
36
40.00%
1
10.00%
pre-git
pre-git
30
33.33%
3
30.00%
willy tarreau
willy tarreau
12
13.33%
1
10.00%
hannes frederic sowa
hannes frederic sowa
5
5.56%
1
10.00%
david s. miller
david s. miller
3
3.33%
1
10.00%
arnaldo carvalho de melo
arnaldo carvalho de melo
2
2.22%
1
10.00%
al viro
al viro
1
1.11%
1
10.00%
pavel emelianov
pavel emelianov
1
1.11%
1
10.00%
Total
90
100.00%
10
100.00%



static void scan_inflight(struct sock *x, void (*func)(struct unix_sock *),
			  struct sk_buff_head *hitlist)
{
	struct sk_buff *skb;
	struct sk_buff *next;

	spin_lock(&x->sk_receive_queue.lock);
	skb_queue_walk_safe(&x->sk_receive_queue, skb, next) {
		/* Do we have file descriptors ? */
		if (UNIXCB(skb).fp) {
			bool hit = false;
			/* Process the descriptors of this socket */
			int nfd = UNIXCB(skb).fp->count;
			struct file **fp = UNIXCB(skb).fp->fp;

			while (nfd--) {
				/* Get the socket the fd matches if it indeed does so */
				struct sock *sk = unix_get_socket(*fp++);

				if (sk) {
					struct unix_sock *u = unix_sk(sk);

					/* Ignore non-candidates, they could
                                         * have been added to the queues after
                                         * starting the garbage collection
                                         */
					if (test_bit(UNIX_GC_CANDIDATE, &u->gc_flags)) {
						hit = true;

						func(u);
					}
				}
			}
			if (hit && hitlist != NULL) {
				__skb_unlink(skb, &x->sk_receive_queue);
				__skb_queue_tail(hitlist, skb);
			}
		}
	}
	spin_unlock(&x->sk_receive_queue.lock);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
pre-git
pre-git
92
46.23%
5
31.25%
miklos szeredi
miklos szeredi
82
41.21%
2
12.50%
eric dumazet
eric dumazet
7
3.52%
1
6.25%
ilpo jarvinen
ilpo jarvinen
4
2.01%
1
6.25%
jason eastman
jason eastman
4
2.01%
1
6.25%
arnaldo carvalho de melo
arnaldo carvalho de melo
4
2.01%
2
12.50%
linus torvalds
linus torvalds
2
1.01%
1
6.25%
pavel emelianov
pavel emelianov
2
1.01%
1
6.25%
andrew morton
andrew morton
1
0.50%
1
6.25%
david s. miller
david s. miller
1
0.50%
1
6.25%
Total
199
100.00%
16
100.00%



static void scan_children(struct sock *x, void (*func)(struct unix_sock *),
			  struct sk_buff_head *hitlist)
{
	if (x->sk_state != TCP_LISTEN) {
		scan_inflight(x, func, hitlist);
	} else {
		struct sk_buff *skb;
		struct sk_buff *next;
		struct unix_sock *u;
		LIST_HEAD(embryos);

		/* For a listening socket collect the queued embryos
                 * and perform a scan on them as well.
                 */
		spin_lock(&x->sk_receive_queue.lock);
		skb_queue_walk_safe(&x->sk_receive_queue, skb, next) {
			u = unix_sk(skb->sk);

			/* An embryo cannot be in-flight, so it's safe
                         * to use the list link.
                         */
			BUG_ON(!list_empty(&u->link));
			list_add_tail(&u->link, &embryos);
		}
		spin_unlock(&x->sk_receive_queue.lock);

		while (!list_empty(&embryos)) {
			u = list_entry(embryos.next, struct unix_sock, link);
			scan_inflight(&u->sk, func, hitlist);
			list_del_init(&u->link);
		}
	}
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
128
71.91%
1
8.33%
pre-git
pre-git
36
20.22%
5
41.67%
ilpo jarvinen
ilpo jarvinen
4
2.25%
1
8.33%
jason eastman
jason eastman
4
2.25%
1
8.33%
david s. miller
david s. miller
4
2.25%
2
16.67%
pavel emelianov
pavel emelianov
1
0.56%
1
8.33%
arnaldo carvalho de melo
arnaldo carvalho de melo
1
0.56%
1
8.33%
Total
178
100.00%
12
100.00%



static void dec_inflight(struct unix_sock *usk)
{
	atomic_long_dec(&usk->inflight);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
15
78.95%
1
33.33%
pavel emelianov
pavel emelianov
3
15.79%
1
33.33%
al viro
al viro
1
5.26%
1
33.33%
Total
19
100.00%
3
100.00%



static void inc_inflight(struct unix_sock *usk)
{
	atomic_long_inc(&usk->inflight);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
15
78.95%
1
33.33%
pavel emelianov
pavel emelianov
3
15.79%
1
33.33%
al viro
al viro
1
5.26%
1
33.33%
Total
19
100.00%
3
100.00%



static void inc_inflight_move_tail(struct unix_sock *u)
{
	atomic_long_inc(&u->inflight);
	/* If this still might be part of a cycle, move it to the end
         * of the list, so that it's checked even if it was already
         * passed over
         */
	if (test_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags))
		list_move_tail(&u->link, &gc_candidates);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
28
65.12%
1
14.29%
eric dumazet
eric dumazet
7
16.28%
1
14.29%
pre-git
pre-git
4
9.30%
2
28.57%
pavel emelianov
pavel emelianov
2
4.65%
1
14.29%
jason eastman
jason eastman
1
2.33%
1
14.29%
al viro
al viro
1
2.33%
1
14.29%
Total
43
100.00%
7
100.00%


static bool gc_in_progress;

#define UNIX_INFLIGHT_TRIGGER_GC 16000



void wait_for_unix_gc(void)
{
	/* If number of inflight sockets is insane,
         * force a garbage collect right now.
         */
	if (unix_tot_inflight > UNIX_INFLIGHT_TRIGGER_GC && !gc_in_progress)
		unix_gc();
	wait_event(unix_gc_wait, gc_in_progress == false);
}

Contributors
Person
Tokens
Prop
Commits
CommitProp
dann frazier
dann frazier
16
55.17%
1
33.33%
eric dumazet
eric dumazet
12
41.38%
1
33.33%
jason eastman
jason eastman
1
3.45%
1
33.33%
Total
29
100.00%
3
100.00%

/* The external entry point: unix_gc() */


void unix_gc(void)
{
	struct unix_sock *u;
	struct unix_sock *next;
	struct sk_buff_head hitlist;
	struct list_head cursor;
	LIST_HEAD(not_cycle_list);

	spin_lock(&unix_gc_lock);

	/* Avoid a recursive GC. */
	if (gc_in_progress)
		goto out;

	gc_in_progress = true;
	/* First, select candidates for garbage collection.  Only
         * in-flight sockets are considered, and from those only ones
         * which don't have any external reference.
         *
         * Holding unix_gc_lock will protect these candidates from
         * being detached, and hence from gaining an external
         * reference.  Since there are no possible receivers, all
         * buffers currently on the candidates' queues stay there
         * during the garbage collection.
         *
         * We also know that no new candidate can be added onto the
         * receive queues.  Other, non candidate sockets _can_ be
         * added to queue, so we must make sure only to touch
         * candidates.
         */
	list_for_each_entry_safe(u, next, &gc_inflight_list, link) {
		long total_refs;
		long inflight_refs;

		total_refs = file_count(u->sk.sk_socket->file);
		inflight_refs = atomic_long_read(&u->inflight);

		BUG_ON(inflight_refs < 1);
		BUG_ON(total_refs < inflight_refs);
		if (total_refs == inflight_refs) {
			list_move_tail(&u->link, &gc_candidates);
			__set_bit(UNIX_GC_CANDIDATE, &u->gc_flags);
			__set_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags);
		}
	}

	/* Now remove all internal in-flight reference to children of
         * the candidates.
         */
	list_for_each_entry(u, &gc_candidates, link)
		scan_children(&u->sk, dec_inflight, NULL);

	/* Restore the references for children of all candidates,
         * which have remaining references.  Do this recursively, so
         * only those remain, which form cyclic references.
         *
         * Use a "cursor" link, to make the list traversal safe, even
         * though elements might be moved about.
         */
	list_add(&cursor, &gc_candidates);
	while (cursor.next != &gc_candidates) {
		u = list_entry(cursor.next, struct unix_sock, link);

		/* Move cursor to after the current position. */
		list_move(&cursor, &u->link);

		if (atomic_long_read(&u->inflight) > 0) {
			list_move_tail(&u->link, &not_cycle_list);
			__clear_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags);
			scan_children(&u->sk, inc_inflight_move_tail, NULL);
		}
	}
	list_del(&cursor);

	/* not_cycle_list contains those sockets which do not make up a
         * cycle.  Restore these to the inflight list.
         */
	while (!list_empty(&not_cycle_list)) {
		u = list_entry(not_cycle_list.next, struct unix_sock, link);
		__clear_bit(UNIX_GC_CANDIDATE, &u->gc_flags);
		list_move_tail(&u->link, &gc_inflight_list);
	}

	/* Now gc_candidates contains only garbage.  Restore original
         * inflight counters for these as well, and remove the skbuffs
         * which are creating the cycle(s).
         */
	skb_queue_head_init(&hitlist);
	list_for_each_entry(u, &gc_candidates, link)
	scan_children(&u->sk, inc_inflight, &hitlist);

	spin_unlock(&unix_gc_lock);

	/* Here we are. Hitlist is filled. Die. */
	__skb_queue_purge(&hitlist);

	spin_lock(&unix_gc_lock);

	/* All candidates should have been detached by now. */
	BUG_ON(!list_empty(&gc_candidates));
	gc_in_progress = false;
	wake_up(&unix_gc_wait);

 out:
	spin_unlock(&unix_gc_lock);
}
Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
315
81.40%
2
16.67%
eric dumazet
eric dumazet
28
7.24%
1
8.33%
pre-git
pre-git
26
6.72%
3
25.00%
dann frazier
dann frazier
6
1.55%
1
8.33%
jason eastman
jason eastman
5
1.29%
1
8.33%
al viro
al viro
4
1.03%
1
8.33%
david s. miller
david s. miller
2
0.52%
2
16.67%
linus torvalds
linus torvalds
1
0.26%
1
8.33%
Total
387
100.00%
12
100.00%

Overall Contributors
Person
Tokens
Prop
Commits
CommitProp
miklos szeredi
miklos szeredi
696
55.37%
2
5.71%
pre-git
pre-git
328
26.09%
9
25.71%
eric dumazet
eric dumazet
58
4.61%
2
5.71%
dann frazier
dann frazier
35
2.78%
1
2.86%
al viro
al viro
24
1.91%
4
11.43%
willy tarreau
willy tarreau
24
1.91%
1
2.86%
jason eastman
jason eastman
18
1.43%
1
2.86%
arnaldo carvalho de melo
arnaldo carvalho de melo
17
1.35%
3
8.57%
pavel emelianov
pavel emelianov
15
1.19%
2
5.71%
david s. miller
david s. miller
13
1.03%
3
8.57%
hannes frederic sowa
hannes frederic sowa
10
0.80%
1
2.86%
ilpo jarvinen
ilpo jarvinen
8
0.64%
1
2.86%
matthew wilcox
matthew wilcox
4
0.32%
1
2.86%
arjan van de ven
arjan van de ven
3
0.24%
1
2.86%
linus torvalds
linus torvalds
3
0.24%
2
5.71%
andrew morton
andrew morton
1
0.08%
1
2.86%
Total
1257
100.00%
35
100.00%
  
Directory: net/unix

Information contained on this website is for historical information purposes only and does not indicate or represent copyright ownership.