Cơ chế dọn rác của Unix garbage collector

/*
 * Open file table structure
 */
struct files_struct {
  /*
   * read mostly part
   */
	atomic_t count;
	bool resize_in_progress;
	wait_queue_head_t resize_wait;

	struct fdtable __rcu *fdt;
	struct fdtable fdtab;
  /*
   * written part on a separate cache line in SMP
   */
	spinlock_t file_lock ____cacheline_aligned_in_smp;
	unsigned int next_fd;
	unsigned long close_on_exec_init[1];
	unsigned long open_fds_init[1];
	unsigned long full_fds_bits_init[1];
	struct file __rcu * fd_array[NR_OPEN_DEFAULT];
};

struct file {
	union {
		struct llist_node	fu_llist;
		struct rcu_head 	fu_rcuhead;
	} f_u;
	struct path		f_path;
	struct inode		*f_inode;	/* cached value */
	const struct file_operations	*f_op;

	/*
	 * Protects f_ep, f_flags.
	 * Must not be taken from IRQ context.
	 */
	spinlock_t		f_lock;
	enum rw_hint		f_write_hint;
	atomic_long_t		f_count; // <-- reference counter
	unsigned int 		f_flags;
	fmode_t			f_mode;
	struct mutex		f_pos_lock;
	loff_t			f_pos;
	struct fown_struct	f_owner;
	const struct cred	*f_cred;
	struct file_ra_state	f_ra;

	u64			f_version;
#ifdef CONFIG_SECURITY
	void			*f_security;
#endif
	/* needed for tty driver, and maybe others */
	void			*private_data;

#ifdef CONFIG_EPOLL
	/* Used by fs/eventpoll.c to link all the hooks to this file */
	struct hlist_head	*f_ep;
#endif /* #ifdef CONFIG_EPOLL */
	struct address_space	*f_mapping;
	errseq_t		f_wb_err;
	errseq_t		f_sb_err; /* for syncfs */
} __randomize_layout
  __attribute__((aligned(4)));	/* lest something weird decides that 2 is OK */

int __sock_create(struct net *net, int family, int type, int protocol,
			 struct socket **res, int kern)
{
	int err;
	struct socket *sock;
	const struct net_proto_family *pf;


.......REDACTED


	#ifdef CONFIG_MODULES
	/* Attempt to load a protocol module if the find failed.
	 *
	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
	 * requested real, full-featured networking support upon configuration.
	 * Otherwise module support will break!
	 */
	if (rcu_access_pointer(net_families[family]) == NULL)
		request_module("net-pf-%d", family);
#endif

	rcu_read_lock();
	pf = rcu_dereference(net_families[family]); // [apx1] dereference trả về net_proto_family struct tương ứng với họ socket mà user đã define
	err = -EAFNOSUPPORT;
	if (!pf)
		goto out_release;

	/*
	 * We will call the ->create function, that possibly is in a loadable
	 * module, so we have to bump that loadable module refcnt first.
	 */
	if (!try_module_get(pf->owner))
		goto out_release;

	/* Now protected by module ref count */
	rcu_read_unlock();

	err = pf->create(net, sock, protocol, kern); // [apx2] invoke function "create" của net_proto_family struct đã được xác định
	if (err < 0)
		goto out_module_put;


.......REDACTED


}

static const struct net_proto_family unix_family_ops = {
	.family = PF_UNIX,
	.create = unix_create,
	.owner	= THIS_MODULE,
};

static const struct net_proto_family inet_family_ops = {
	.family = PF_INET,
	.create = inet_create,
	.owner	= THIS_MODULE,
};

static int unix_create(struct net *net, struct socket *sock, int protocol,
		       int kern)
{
	struct sock *sk;

	if (protocol && protocol != PF_UNIX)
		return -EPROTONOSUPPORT;

	sock->state = SS_UNCONNECTED;

	switch (sock->type) {
	case SOCK_STREAM:
		sock->ops = &unix_stream_ops;
		break;
		/*
		 *	Believe it or not BSD has AF_UNIX, SOCK_RAW though
		 *	nothing uses it.
		 */
	case SOCK_RAW:
		sock->type = SOCK_DGRAM;
		fallthrough;
	case SOCK_DGRAM:
		sock->ops = &unix_dgram_ops;
		break;
	case SOCK_SEQPACKET:
		sock->ops = &unix_seqpacket_ops;
		break;
	default:
		return -ESOCKTNOSUPPORT;
	}

	sk = unix_create1(net, sock, kern, sock->type);
	if (IS_ERR(sk))
		return PTR_ERR(sk);

	return 0;
}

static const struct proto_ops unix_dgram_ops = {
	.family =	PF_UNIX,
	.owner =	THIS_MODULE,
	.release =	unix_release,
	.bind =		unix_bind,
	.connect =	unix_dgram_connect,
	.socketpair =	unix_socketpair,
	.accept =	sock_no_accept,
	.getname =	unix_getname,
	.poll =		unix_dgram_poll,
	.ioctl =	unix_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl =	unix_compat_ioctl,
#endif
	.listen =	sock_no_listen,
	.shutdown =	unix_shutdown,
	.sendmsg =	unix_dgram_sendmsg,
	.read_sock =	unix_read_sock,
	.recvmsg =	unix_dgram_recvmsg,
	.mmap =		sock_no_mmap,
	.sendpage =	sock_no_sendpage,
	.set_peek_off =	unix_set_peek_off,
	.show_fdinfo =	unix_show_fdinfo,
};

static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
			      size_t len)
{
	struct sock *sk = sock->sk;
	struct net *net = sock_net(sk);
	struct unix_sock *u = unix_sk(sk);
	DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name);
	struct sock *other = NULL;
	int namelen = 0; /* fake GCC */
	int err;
	unsigned int hash;
	struct sk_buff *skb;
	long timeo;
	struct scm_cookie scm;
	int data_len = 0;
	int sk_locked;

	wait_for_unix_gc(); // setup signal đợi unix_gc
	err = scm_send(sock, msg, &scm, false); // scm viết tắt của "socket control message"
	if (err < 0)
		return err;

	err = -EOPNOTSUPP;
	if (msg->msg_flags&MSG_OOB)
		goto out;

	if (msg->msg_namelen) {
		err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
		if (err < 0)
			goto out;
		namelen = err;
	} else {
		sunaddr = NULL;
		err = -ENOTCONN;
		other = unix_peer_get(sk);
		if (!other)
			goto out;
	}

	if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr
	    && (err = unix_autobind(sock)) != 0)
		goto out;

	err = -EMSGSIZE;
	if (len > sk->sk_sndbuf - 32)
		goto out;

	if (len > SKB_MAX_ALLOC) {
		data_len = min_t(size_t,
				 len - SKB_MAX_ALLOC,
				 MAX_SKB_FRAGS * PAGE_SIZE);
		data_len = PAGE_ALIGN(data_len);

		BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE);
	}

	skb = sock_alloc_send_pskb(sk, len - data_len, data_len,
				   msg->msg_flags & MSG_DONTWAIT, &err,
				   PAGE_ALLOC_COSTLY_ORDER);
	if (skb == NULL)
		goto out;

	err = unix_scm_to_skb(&scm, skb, true); // parse struct từ scm_cookie sang sk_buff struct
	if (err < 0)
		goto out_free;

	skb_put(skb, len - data_len);
	skb->data_len = data_len;
	skb->len = len;
	err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
	if (err)
		goto out_free;

	timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);

restart:
	if (!other) {
		err = -ECONNRESET;
		if (sunaddr == NULL)
			goto out_free;

		other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
					hash, &err);
		if (other == NULL)
			goto out_free;
	}

	if (sk_filter(other, skb) < 0) { // [pass1] pass sk_buff struct vừa được parse sang cho "other", "other" là đầu nhận thông tin.
		/* Toss the packet but do not return any error to the sender */
		err = len;
		goto out_free;
	}

	sk_locked = 0;
	unix_state_lock(other);
restart_locked:
	err = -EPERM;
	if (!unix_may_send(sk, other))
		goto out_unlock;

	if (unlikely(sock_flag(other, SOCK_DEAD))) {
		/*
		 *	Check with 1003.1g - what should
		 *	datagram error
		 */
		unix_state_unlock(other);
		sock_put(other);

		if (!sk_locked)
			unix_state_lock(sk);

		err = 0;
		if (sk->sk_type == SOCK_SEQPACKET) {
			/* We are here only when racing with unix_release_sock()
			 * is clearing @other. Never change state to TCP_CLOSE
			 * unlike SOCK_DGRAM wants.
			 */
			unix_state_unlock(sk);
			err = -EPIPE;
		} else if (unix_peer(sk) == other) {
			unix_peer(sk) = NULL;
			unix_dgram_peer_wake_disconnect_wakeup(sk, other);

			sk->sk_state = TCP_CLOSE;
			unix_state_unlock(sk);

			unix_dgram_disconnected(sk, other);
			sock_put(other);
			err = -ECONNREFUSED;
		} else {
			unix_state_unlock(sk);
		}

		other = NULL;
		if (err)
			goto out_free;
		goto restart;
	}

	err = -EPIPE;
	if (other->sk_shutdown & RCV_SHUTDOWN)
		goto out_unlock;

	if (sk->sk_type != SOCK_SEQPACKET) {
		err = security_unix_may_send(sk->sk_socket, other->sk_socket);
		if (err)
			goto out_unlock;
	}

	/* other == sk && unix_peer(other) != sk if
	 * - unix_peer(sk) == NULL, destination address bound to sk
	 * - unix_peer(sk) == sk by time of get but disconnected before lock
	 */
	if (other != sk &&
	    unlikely(unix_peer(other) != sk &&
	    unix_recvq_full_lockless(other))) {
		if (timeo) {
			timeo = unix_wait_for_peer(other, timeo);

			err = sock_intr_errno(timeo);
			if (signal_pending(current))
				goto out_free;

			goto restart;
		}

		if (!sk_locked) {
			unix_state_unlock(other);
			unix_state_double_lock(sk, other);
		}

		if (unix_peer(sk) != other ||
		    unix_dgram_peer_wake_me(sk, other)) {
			err = -EAGAIN;
			sk_locked = 1;
			goto out_unlock;
		}

		if (!sk_locked) {
			sk_locked = 1;
			goto restart_locked;
		}
	}

	if (unlikely(sk_locked))
		unix_state_unlock(sk);

	if (sock_flag(other, SOCK_RCVTSTAMP))
		__net_timestamp(skb);
	maybe_add_creds(skb, sock, other);
	scm_stat_add(other, skb);
	skb_queue_tail(&other->sk_receive_queue, skb); // [pass2] đưa skb vào queue ở đầu nhận, cho phép socket phía gửi có thể gửi sau đó close luôn, queue ở phía nhận sẽ đợi cho đến khi phía nhận invoke syscall recvmsg()
	unix_state_unlock(other);
	other->sk_data_ready(other);
	sock_put(other);
	scm_destroy(&scm);
	return len;

out_unlock:
	if (sk_locked)
		unix_state_unlock(sk);
	unix_state_unlock(other);
out_free:
	kfree_skb(skb);
out:
	if (other)
		sock_put(other);
	scm_destroy(&scm);
	return err;
}

int __scm_send(struct socket *sock, struct msghdr *msg, struct scm_cookie *p)
{
	struct cmsghdr *cmsg;
	int err;

	for_each_cmsghdr(cmsg, msg) {
		err = -EINVAL;

		/* Verify that cmsg_len is at least sizeof(struct cmsghdr) */
		/* The first check was omitted in <= 2.2.5. The reasoning was
		   that parser checks cmsg_len in any case, so that
		   additional check would be work duplication.
		   But if cmsg_level is not SOL_SOCKET, we do not check
		   for too short ancillary data object at all! Oops.
		   OK, let's add it...
		 */
		if (!CMSG_OK(msg, cmsg))
			goto error;

		if (cmsg->cmsg_level != SOL_SOCKET)
			continue;

		switch (cmsg->cmsg_type)
		{
		case SCM_RIGHTS:
			if (!sock->ops || sock->ops->family != PF_UNIX)
				goto error;
			err=scm_fp_copy(cmsg, &p->fp);
			if (err<0)
				goto error;
			break;


.......REDACTED


}

/* 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));
		}
		/* Paired with READ_ONCE() in wait_for_unix_gc() */
		WRITE_ONCE(unix_tot_inflight, unix_tot_inflight + 1);
	}
	user->unix_inflight++;
	spin_unlock(&unix_gc_lock);
}

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.
	 * Paired with the WRITE_ONCE() in unix_inflight(),
	 * unix_notinflight() and gc_in_progress().
	 */
	if (READ_ONCE(unix_tot_inflight) > UNIX_INFLIGHT_TRIGGER_GC &&
	    !READ_ONCE(gc_in_progress))
		unix_gc();
	wait_event(unix_gc_wait, gc_in_progress == false);
}

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


.......REDACTED


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

	/* Paired with READ_ONCE() in wait_for_unix_gc(). */
	WRITE_ONCE(gc_in_progress, true);


.......REDACTED


}

.......REDACTED


	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); // thêm socket vào gc_candidate queue
			__set_bit(UNIX_GC_CANDIDATE, &u->gc_flags); // đánh dấu bằng bit UNIX_GC_CANDIDATE
			__set_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags); // đánh dấu bằng bit UNIX_GC_MAYBE_CYCLE cho biết có thể socket đang kẹt trong một unbreakable cycle
		}
	}


.......REDACTED

.......REDACTED


	list_for_each_entry(u, &gc_candidates, link)
		scan_children(&u->sk, dec_inflight, NULL);


.......REDACTED

.......REDACTED


	/* 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);


.......REDACTED

.......REDACTED


	/* 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);

	/* 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);
	}


.......REDACTED

.......REDACTED


	/* We need io_uring to clean its registered files, ignore all io_uring
	 * originated skbs. It's fine as io_uring doesn't keep references to
	 * other io_uring instances and so killing all other files in the cycle
	 * will put all io_uring references forcing it to go through normal
	 * release.path eventually putting registered files.
	 */
	skb_queue_walk_safe(&hitlist, skb, next_skb) {
		if (skb->scm_io_uring) {
			__skb_unlink(skb, &hitlist);
			skb_queue_tail(&skb->sk->sk_receive_queue, skb);
		}
	}


.......REDACTED

.......REDACTED


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


.......REDACTED