Merge branch 'linus' into perf/core, to pick up fixes

Signed-off-by: Ingo Molnar <mingo@kernel.org>

kernel/audit.c

@@ -54,6 +54,10 @@
 #include <linux/kthread.h>
 #include <linux/kernel.h>
 #include <linux/syscalls.h>
+#include <linux/spinlock.h>
+#include <linux/rcupdate.h>
+#include <linux/mutex.h>
+#include <linux/gfp.h>
 
 #include <linux/audit.h>
 
@@ -90,13 +94,34 @@ static u32	audit_default;
 /* If auditing cannot proceed, audit_failure selects what happens. */
 static u32	audit_failure = AUDIT_FAIL_PRINTK;
 
-/*
- * If audit records are to be written to the netlink socket, audit_pid
- * contains the pid of the auditd process and audit_nlk_portid contains
- * the portid to use to send netlink messages to that process.
+/* private audit network namespace index */
+static unsigned int audit_net_id;
+
+/**
+ * struct audit_net - audit private network namespace data
+ * @sk: communication socket
  */
-int		audit_pid;
-static __u32	audit_nlk_portid;
+struct audit_net {
+	struct sock *sk;
+};
+
+/**
+ * struct auditd_connection - kernel/auditd connection state
+ * @pid: auditd PID
+ * @portid: netlink portid
+ * @net: the associated network namespace
+ * @lock: spinlock to protect write access
+ *
+ * Description:
+ * This struct is RCU protected; you must either hold the RCU lock for reading
+ * or the included spinlock for writing.
+ */
+static struct auditd_connection {
+	int pid;
+	u32 portid;
+	struct net *net;
+	spinlock_t lock;
+} auditd_conn;
 
 /* If audit_rate_limit is non-zero, limit the rate of sending audit records
  * to that number per second.  This prevents DoS attacks, but results in
@@ -123,10 +148,6 @@ u32		audit_sig_sid = 0;
 */
 static atomic_t	audit_lost = ATOMIC_INIT(0);
 
-/* The netlink socket. */
-static struct sock *audit_sock;
-static unsigned int audit_net_id;
-
 /* Hash for inode-based rules */
 struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
 
@@ -139,6 +160,7 @@ static LIST_HEAD(audit_freelist);
 
 /* queue msgs to send via kauditd_task */
 static struct sk_buff_head audit_queue;
+static void kauditd_hold_skb(struct sk_buff *skb);
 /* queue msgs due to temporary unicast send problems */
 static struct sk_buff_head audit_retry_queue;
 /* queue msgs waiting for new auditd connection */
@@ -192,6 +214,43 @@ struct audit_reply {
 	struct sk_buff *skb;
 };
 
+/**
+ * auditd_test_task - Check to see if a given task is an audit daemon
+ * @task: the task to check
+ *
+ * Description:
+ * Return 1 if the task is a registered audit daemon, 0 otherwise.
+ */
+int auditd_test_task(const struct task_struct *task)
+{
+	int rc;
+
+	rcu_read_lock();
+	rc = (auditd_conn.pid && task->tgid == auditd_conn.pid ? 1 : 0);
+	rcu_read_unlock();
+
+	return rc;
+}
+
+/**
+ * audit_get_sk - Return the audit socket for the given network namespace
+ * @net: the destination network namespace
+ *
+ * Description:
+ * Returns the sock pointer if valid, NULL otherwise.  The caller must ensure
+ * that a reference is held for the network namespace while the sock is in use.
+ */
+static struct sock *audit_get_sk(const struct net *net)
+{
+	struct audit_net *aunet;
+
+	if (!net)
+		return NULL;
+
+	aunet = net_generic(net, audit_net_id);
+	return aunet->sk;
+}
+
 static void audit_set_portid(struct audit_buffer *ab, __u32 portid)
 {
 	if (ab) {
@@ -210,9 +269,7 @@ void audit_panic(const char *message)
 			pr_err("%s\n", message);
 		break;
 	case AUDIT_FAIL_PANIC:
-		/* test audit_pid since printk is always losey, why bother? */
-		if (audit_pid)
-			panic("audit: %s\n", message);
+		panic("audit: %s\n", message);
 		break;
 	}
 }
@@ -370,21 +427,87 @@ static int audit_set_failure(u32 state)
 	return audit_do_config_change("audit_failure", &audit_failure, state);
 }
 
-/*
- * For one reason or another this nlh isn't getting delivered to the userspace
- * audit daemon, just send it to printk.
+/**
+ * auditd_set - Set/Reset the auditd connection state
+ * @pid: auditd PID
+ * @portid: auditd netlink portid
+ * @net: auditd network namespace pointer
+ *
+ * Description:
+ * This function will obtain and drop network namespace references as
+ * necessary.
+ */
+static void auditd_set(int pid, u32 portid, struct net *net)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&auditd_conn.lock, flags);
+	auditd_conn.pid = pid;
+	auditd_conn.portid = portid;
+	if (auditd_conn.net)
+		put_net(auditd_conn.net);
+	if (net)
+		auditd_conn.net = get_net(net);
+	else
+		auditd_conn.net = NULL;
+	spin_unlock_irqrestore(&auditd_conn.lock, flags);
+}
+
+/**
+ * auditd_reset - Disconnect the auditd connection
+ *
+ * Description:
+ * Break the auditd/kauditd connection and move all the queued records into the
+ * hold queue in case auditd reconnects.
+ */
+static void auditd_reset(void)
+{
+	struct sk_buff *skb;
+
+	/* if it isn't already broken, break the connection */
+	rcu_read_lock();
+	if (auditd_conn.pid)
+		auditd_set(0, 0, NULL);
+	rcu_read_unlock();
+
+	/* flush all of the main and retry queues to the hold queue */
+	while ((skb = skb_dequeue(&audit_retry_queue)))
+		kauditd_hold_skb(skb);
+	while ((skb = skb_dequeue(&audit_queue)))
+		kauditd_hold_skb(skb);
+}
+
+/**
+ * kauditd_print_skb - Print the audit record to the ring buffer
+ * @skb: audit record
+ *
+ * Whatever the reason, this packet may not make it to the auditd connection
+ * so write it via printk so the information isn't completely lost.
  */
 static void kauditd_printk_skb(struct sk_buff *skb)
 {
 	struct nlmsghdr *nlh = nlmsg_hdr(skb);
 	char *data = nlmsg_data(nlh);
 
-	if (nlh->nlmsg_type != AUDIT_EOE) {
-		if (printk_ratelimit())
-			pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
-		else
-			audit_log_lost("printk limit exceeded");
-	}
+	if (nlh->nlmsg_type != AUDIT_EOE && printk_ratelimit())
+		pr_notice("type=%d %s\n", nlh->nlmsg_type, data);
+}
+
+/**
+ * kauditd_rehold_skb - Handle a audit record send failure in the hold queue
+ * @skb: audit record
+ *
+ * Description:
+ * This should only be used by the kauditd_thread when it fails to flush the
+ * hold queue.
+ */
+static void kauditd_rehold_skb(struct sk_buff *skb)
+{
+	/* put the record back in the queue at the same place */
+	skb_queue_head(&audit_hold_queue, skb);
+
+	/* fail the auditd connection */
+	auditd_reset();
 }
 
 /**
@@ -421,6 +544,9 @@ static void kauditd_hold_skb(struct sk_buff *skb)
 	/* we have no other options - drop the message */
 	audit_log_lost("kauditd hold queue overflow");
 	kfree_skb(skb);
+
+	/* fail the auditd connection */
+	auditd_reset();
 }
 
 /**
@@ -441,51 +567,122 @@ static void kauditd_retry_skb(struct sk_buff *skb)
 }
 
 /**
- * auditd_reset - Disconnect the auditd connection
+ * auditd_send_unicast_skb - Send a record via unicast to auditd
+ * @skb: audit record
  *
  * Description:
- * Break the auditd/kauditd connection and move all the records in the retry
- * queue into the hold queue in case auditd reconnects.  The audit_cmd_mutex
- * must be held when calling this function.
+ * Send a skb to the audit daemon, returns positive/zero values on success and
+ * negative values on failure; in all cases the skb will be consumed by this
+ * function.  If the send results in -ECONNREFUSED the connection with auditd
+ * will be reset.  This function may sleep so callers should not hold any locks
+ * where this would cause a problem.
  */
-static void auditd_reset(void)
+static int auditd_send_unicast_skb(struct sk_buff *skb)
 {
-	struct sk_buff *skb;
+	int rc;
+	u32 portid;
+	struct net *net;
+	struct sock *sk;
 
-	/* break the connection */
-	if (audit_sock) {
-		sock_put(audit_sock);
-		audit_sock = NULL;
+	/* NOTE: we can't call netlink_unicast while in the RCU section so
+	 *       take a reference to the network namespace and grab local
+	 *       copies of the namespace, the sock, and the portid; the
+	 *       namespace and sock aren't going to go away while we hold a
+	 *       reference and if the portid does become invalid after the RCU
+	 *       section netlink_unicast() should safely return an error */
+
+	rcu_read_lock();
+	if (!auditd_conn.pid) {
+		rcu_read_unlock();
+		rc = -ECONNREFUSED;
+		goto err;
 	}
-	audit_pid = 0;
-	audit_nlk_portid = 0;
+	net = auditd_conn.net;
+	get_net(net);
+	sk = audit_get_sk(net);
+	portid = auditd_conn.portid;
+	rcu_read_unlock();
 
-	/* flush all of the retry queue to the hold queue */
-	while ((skb = skb_dequeue(&audit_retry_queue)))
-		kauditd_hold_skb(skb);
+	rc = netlink_unicast(sk, skb, portid, 0);
+	put_net(net);
+	if (rc < 0)
+		goto err;
+
+	return rc;
+
+err:
+	if (rc == -ECONNREFUSED)
+		auditd_reset();
+	return rc;
 }
 
 /**
- * kauditd_send_unicast_skb - Send a record via unicast to auditd
- * @skb: audit record
+ * kauditd_send_queue - Helper for kauditd_thread to flush skb queues
+ * @sk: the sending sock
+ * @portid: the netlink destination
+ * @queue: the skb queue to process
+ * @retry_limit: limit on number of netlink unicast failures
+ * @skb_hook: per-skb hook for additional processing
+ * @err_hook: hook called if the skb fails the netlink unicast send
+ *
+ * Description:
+ * Run through the given queue and attempt to send the audit records to auditd,
+ * returns zero on success, negative values on failure.  It is up to the caller
+ * to ensure that the @sk is valid for the duration of this function.
+ *
  */
-static int kauditd_send_unicast_skb(struct sk_buff *skb)
+static int kauditd_send_queue(struct sock *sk, u32 portid,
+			      struct sk_buff_head *queue,
+			      unsigned int retry_limit,
+			      void (*skb_hook)(struct sk_buff *skb),
+			      void (*err_hook)(struct sk_buff *skb))
 {
-	int rc;
+	int rc = 0;
+	struct sk_buff *skb;
+	static unsigned int failed = 0;
 
-	/* if we know nothing is connected, don't even try the netlink call */
-	if (!audit_pid)
-		return -ECONNREFUSED;
+	/* NOTE: kauditd_thread takes care of all our locking, we just use
+	 *       the netlink info passed to us (e.g. sk and portid) */
 
-	/* get an extra skb reference in case we fail to send */
-	skb_get(skb);
-	rc = netlink_unicast(audit_sock, skb, audit_nlk_portid, 0);
-	if (rc >= 0) {
-		consume_skb(skb);
-		rc = 0;
+	while ((skb = skb_dequeue(queue))) {
+		/* call the skb_hook for each skb we touch */
+		if (skb_hook)
+			(*skb_hook)(skb);
+
+		/* can we send to anyone via unicast? */
+		if (!sk) {
+			if (err_hook)
+				(*err_hook)(skb);
+			continue;
+		}
+
+		/* grab an extra skb reference in case of error */
+		skb_get(skb);
+		rc = netlink_unicast(sk, skb, portid, 0);
+		if (rc < 0) {
+			/* fatal failure for our queue flush attempt? */
+			if (++failed >= retry_limit ||
+			    rc == -ECONNREFUSED || rc == -EPERM) {
+				/* yes - error processing for the queue */
+				sk = NULL;
+				if (err_hook)
+					(*err_hook)(skb);
+				if (!skb_hook)
+					goto out;
+				/* keep processing with the skb_hook */
+				continue;
+			} else
+				/* no - requeue to preserve ordering */
+				skb_queue_head(queue, skb);
+		} else {
+			/* it worked - drop the extra reference and continue */
+			consume_skb(skb);
+			failed = 0;
+		}
 	}
 
-	return rc;
+out:
+	return (rc >= 0 ? 0 : rc);
 }
 
 /*
@@ -493,16 +690,19 @@ static int kauditd_send_unicast_skb(struct sk_buff *skb)
  * @skb: audit record
  *
  * Description:
- * This function doesn't consume an skb as might be expected since it has to
- * copy it anyways.
+ * Write a multicast message to anyone listening in the initial network
+ * namespace.  This function doesn't consume an skb as might be expected since
+ * it has to copy it anyways.
  */
 static void kauditd_send_multicast_skb(struct sk_buff *skb)
 {
 	struct sk_buff *copy;
-	struct audit_net *aunet = net_generic(&init_net, audit_net_id);
-	struct sock *sock = aunet->nlsk;
+	struct sock *sock = audit_get_sk(&init_net);
 	struct nlmsghdr *nlh;
 
+	/* NOTE: we are not taking an additional reference for init_net since
+	 *       we don't have to worry about it going away */
+
 	if (!netlink_has_listeners(sock, AUDIT_NLGRP_READLOG))
 		return;
 
@@ -526,149 +726,75 @@ static void kauditd_send_multicast_skb(struct sk_buff *skb)
 }
 
 /**
- * kauditd_wake_condition - Return true when it is time to wake kauditd_thread
- *
- * Description:
- * This function is for use by the wait_event_freezable() call in
- * kauditd_thread().
+ * kauditd_thread - Worker thread to send audit records to userspace
+ * @dummy: unused
  */
-static int kauditd_wake_condition(void)
-{
-	static int pid_last = 0;
-	int rc;
-	int pid = audit_pid;
-
-	/* wake on new messages or a change in the connected auditd */
-	rc = skb_queue_len(&audit_queue) || (pid && pid != pid_last);
-	if (rc)
-		pid_last = pid;
-
-	return rc;
-}
-
 static int kauditd_thread(void *dummy)
 {
 	int rc;
-	int auditd = 0;
-	int reschedule = 0;
-	struct sk_buff *skb;
-	struct nlmsghdr *nlh;
+	u32 portid = 0;
+	struct net *net = NULL;
+	struct sock *sk = NULL;
 
 #define UNICAST_RETRIES 5
-#define AUDITD_BAD(x,y) \
-	((x) == -ECONNREFUSED || (x) == -EPERM || ++(y) >= UNICAST_RETRIES)
-
-	/* NOTE: we do invalidate the auditd connection flag on any sending
-	 * errors, but we only "restore" the connection flag at specific places
-	 * in the loop in order to help ensure proper ordering of audit
-	 * records */
 
 	set_freezable();
 	while (!kthread_should_stop()) {
-		/* NOTE: possible area for future improvement is to look at
-		 *       the hold and retry queues, since only this thread
-		 *       has access to these queues we might be able to do
-		 *       our own queuing and skip some/all of the locking */
-
-		/* NOTE: it might be a fun experiment to split the hold and
-		 *       retry queue handling to another thread, but the
-		 *       synchronization issues and other overhead might kill
-		 *       any performance gains */
+		/* NOTE: see the lock comments in auditd_send_unicast_skb() */
+		rcu_read_lock();
+		if (!auditd_conn.pid) {
+			rcu_read_unlock();
+			goto main_queue;
+		}
+		net = auditd_conn.net;
+		get_net(net);
+		sk = audit_get_sk(net);
+		portid = auditd_conn.portid;
+		rcu_read_unlock();
 
 		/* attempt to flush the hold queue */
-		while (auditd && (skb = skb_dequeue(&audit_hold_queue))) {
-			rc = kauditd_send_unicast_skb(skb);
-			if (rc) {
-				/* requeue to the same spot */
-				skb_queue_head(&audit_hold_queue, skb);
-
-				auditd = 0;
-				if (AUDITD_BAD(rc, reschedule)) {
-					mutex_lock(&audit_cmd_mutex);
-					auditd_reset();
-					mutex_unlock(&audit_cmd_mutex);
-					reschedule = 0;
-				}
-			} else
-				/* we were able to send successfully */
-				reschedule = 0;
+		rc = kauditd_send_queue(sk, portid,
+					&audit_hold_queue, UNICAST_RETRIES,
+					NULL, kauditd_rehold_skb);
+		if (rc < 0) {
+			sk = NULL;
+			goto main_queue;
 		}
 
 		/* attempt to flush the retry queue */
-		while (auditd && (skb = skb_dequeue(&audit_retry_queue))) {
-			rc = kauditd_send_unicast_skb(skb);
-			if (rc) {
-				auditd = 0;
-				if (AUDITD_BAD(rc, reschedule)) {
-					kauditd_hold_skb(skb);
-					mutex_lock(&audit_cmd_mutex);
-					auditd_reset();
-					mutex_unlock(&audit_cmd_mutex);
-					reschedule = 0;
-				} else
-					/* temporary problem (we hope), queue
-					 * to the same spot and retry */
-					skb_queue_head(&audit_retry_queue, skb);
-			} else
-				/* we were able to send successfully */
-				reschedule = 0;
+		rc = kauditd_send_queue(sk, portid,
+					&audit_retry_queue, UNICAST_RETRIES,
+					NULL, kauditd_hold_skb);
+		if (rc < 0) {
+			sk = NULL;
+			goto main_queue;
 		}
 
-		/* standard queue processing, try to be as quick as possible */
-quick_loop:
-		skb = skb_dequeue(&audit_queue);
-		if (skb) {
-			/* setup the netlink header, see the comments in
-			 * kauditd_send_multicast_skb() for length quirks */
-			nlh = nlmsg_hdr(skb);
-			nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
+main_queue:
+		/* process the main queue - do the multicast send and attempt
+		 * unicast, dump failed record sends to the retry queue; if
+		 * sk == NULL due to previous failures we will just do the
+		 * multicast send and move the record to the retry queue */
+		kauditd_send_queue(sk, portid, &audit_queue, 1,
+				   kauditd_send_multicast_skb,
+				   kauditd_retry_skb);
 
-			/* attempt to send to any multicast listeners */
-			kauditd_send_multicast_skb(skb);
-
-			/* attempt to send to auditd, queue on failure */
-			if (auditd) {
-				rc = kauditd_send_unicast_skb(skb);
-				if (rc) {
-					auditd = 0;
-					if (AUDITD_BAD(rc, reschedule)) {
-						mutex_lock(&audit_cmd_mutex);
-						auditd_reset();
-						mutex_unlock(&audit_cmd_mutex);
-						reschedule = 0;
-					}
-
-					/* move to the retry queue */
-					kauditd_retry_skb(skb);
-				} else
-					/* everything is working so go fast! */
-					goto quick_loop;
-			} else if (reschedule)
-				/* we are currently having problems, move to
-				 * the retry queue */
-				kauditd_retry_skb(skb);
-			else
-				/* dump the message via printk and hold it */
-				kauditd_hold_skb(skb);
-		} else {
-			/* we have flushed the backlog so wake everyone */
-			wake_up(&audit_backlog_wait);
-
-			/* if everything is okay with auditd (if present), go
-			 * to sleep until there is something new in the queue
-			 * or we have a change in the connected auditd;
-			 * otherwise simply reschedule to give things a chance
-			 * to recover */
-			if (reschedule) {
-				set_current_state(TASK_INTERRUPTIBLE);
-				schedule();
-			} else
-				wait_event_freezable(kauditd_wait,
-						     kauditd_wake_condition());
-
-			/* update the auditd connection status */
-			auditd = (audit_pid ? 1 : 0);
+		/* drop our netns reference, no auditd sends past this line */
+		if (net) {
+			put_net(net);
+			net = NULL;
 		}
+		sk = NULL;
+
+		/* we have processed all the queues so wake everyone */
+		wake_up(&audit_backlog_wait);
+
+		/* NOTE: we want to wake up if there is anything on the queue,
+		 *       regardless of if an auditd is connected, as we need to
+		 *       do the multicast send and rotate records from the
+		 *       main queue to the retry/hold queues */
+		wait_event_freezable(kauditd_wait,
+				     (skb_queue_len(&audit_queue) ? 1 : 0));
 	}
 
 	return 0;
@@ -678,17 +804,16 @@ int audit_send_list(void *_dest)
 {
 	struct audit_netlink_list *dest = _dest;
 	struct sk_buff *skb;
-	struct net *net = dest->net;
-	struct audit_net *aunet = net_generic(net, audit_net_id);
+	struct sock *sk = audit_get_sk(dest->net);
 
 	/* wait for parent to finish and send an ACK */
 	mutex_lock(&audit_cmd_mutex);
 	mutex_unlock(&audit_cmd_mutex);
 
 	while ((skb = __skb_dequeue(&dest->q)) != NULL)
-		netlink_unicast(aunet->nlsk, skb, dest->portid, 0);
+		netlink_unicast(sk, skb, dest->portid, 0);
 
-	put_net(net);
+	put_net(dest->net);
 	kfree(dest);
 
 	return 0;
@@ -722,16 +847,15 @@ out_kfree_skb:
 static int audit_send_reply_thread(void *arg)
 {
 	struct audit_reply *reply = (struct audit_reply *)arg;
-	struct net *net = reply->net;
-	struct audit_net *aunet = net_generic(net, audit_net_id);
+	struct sock *sk = audit_get_sk(reply->net);
 
 	mutex_lock(&audit_cmd_mutex);
 	mutex_unlock(&audit_cmd_mutex);
 
 	/* Ignore failure. It'll only happen if the sender goes away,
 	   because our timeout is set to infinite. */
-	netlink_unicast(aunet->nlsk , reply->skb, reply->portid, 0);
-	put_net(net);
+	netlink_unicast(sk, reply->skb, reply->portid, 0);
+	put_net(reply->net);
 	kfree(reply);
 	return 0;
 }
@@ -949,12 +1073,12 @@ static int audit_set_feature(struct sk_buff *skb)
 
 static int audit_replace(pid_t pid)
 {
-	struct sk_buff *skb = audit_make_reply(0, 0, AUDIT_REPLACE, 0, 0,
-					       &pid, sizeof(pid));
+	struct sk_buff *skb;
 
+	skb = audit_make_reply(0, 0, AUDIT_REPLACE, 0, 0, &pid, sizeof(pid));
 	if (!skb)
 		return -ENOMEM;
-	return netlink_unicast(audit_sock, skb, audit_nlk_portid, 0);
+	return auditd_send_unicast_skb(skb);
 }
 
 static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
@@ -981,7 +1105,9 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 		memset(&s, 0, sizeof(s));
 		s.enabled		= audit_enabled;
 		s.failure		= audit_failure;
-		s.pid			= audit_pid;
+		rcu_read_lock();
+		s.pid			= auditd_conn.pid;
+		rcu_read_unlock();
 		s.rate_limit		= audit_rate_limit;
 		s.backlog_limit		= audit_backlog_limit;
 		s.lost			= atomic_read(&audit_lost);
@@ -1014,30 +1140,44 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 			 *       from the initial pid namespace, but something
 			 *       to keep in mind if this changes */
 			int new_pid = s.pid;
+			pid_t auditd_pid;
 			pid_t requesting_pid = task_tgid_vnr(current);
 
-			if ((!new_pid) && (requesting_pid != audit_pid)) {
-				audit_log_config_change("audit_pid", new_pid, audit_pid, 0);
+			/* test the auditd connection */
+			audit_replace(requesting_pid);
+
+			rcu_read_lock();
+			auditd_pid = auditd_conn.pid;
+			/* only the current auditd can unregister itself */
+			if ((!new_pid) && (requesting_pid != auditd_pid)) {
+				rcu_read_unlock();
+				audit_log_config_change("audit_pid", new_pid,
+							auditd_pid, 0);
 				return -EACCES;
 			}
-			if (audit_pid && new_pid &&
-			    audit_replace(requesting_pid) != -ECONNREFUSED) {
-				audit_log_config_change("audit_pid", new_pid, audit_pid, 0);
+			/* replacing a healthy auditd is not allowed */
+			if (auditd_pid && new_pid) {
+				rcu_read_unlock();
+				audit_log_config_change("audit_pid", new_pid,
+							auditd_pid, 0);
 				return -EEXIST;
 			}
+			rcu_read_unlock();
+
 			if (audit_enabled != AUDIT_OFF)
-				audit_log_config_change("audit_pid", new_pid, audit_pid, 1);
+				audit_log_config_change("audit_pid", new_pid,
+							auditd_pid, 1);
+
 			if (new_pid) {
-				if (audit_sock)
-					sock_put(audit_sock);
-				audit_pid = new_pid;
-				audit_nlk_portid = NETLINK_CB(skb).portid;
-				sock_hold(skb->sk);
-				audit_sock = skb->sk;
-			} else {
+				/* register a new auditd connection */
+				auditd_set(new_pid,
+					   NETLINK_CB(skb).portid,
+					   sock_net(NETLINK_CB(skb).sk));
+				/* try to process any backlog */
+				wake_up_interruptible(&kauditd_wait);
+			} else
+				/* unregister the auditd connection */
 				auditd_reset();
-			}
-			wake_up_interruptible(&kauditd_wait);
 		}
 		if (s.mask & AUDIT_STATUS_RATE_LIMIT) {
 			err = audit_set_rate_limit(s.rate_limit);
@@ -1090,7 +1230,6 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 				if (err)
 					break;
 			}
-			mutex_unlock(&audit_cmd_mutex);
 			audit_log_common_recv_msg(&ab, msg_type);
 			if (msg_type != AUDIT_USER_TTY)
 				audit_log_format(ab, " msg='%.*s'",
@@ -1108,7 +1247,6 @@ static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
 			}
 			audit_set_portid(ab, NETLINK_CB(skb).portid);
 			audit_log_end(ab);
-			mutex_lock(&audit_cmd_mutex);
 		}
 		break;
 	case AUDIT_ADD_RULE:
@@ -1298,26 +1436,26 @@ static int __net_init audit_net_init(struct net *net)
 
 	struct audit_net *aunet = net_generic(net, audit_net_id);
 
-	aunet->nlsk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
-	if (aunet->nlsk == NULL) {
+	aunet->sk = netlink_kernel_create(net, NETLINK_AUDIT, &cfg);
+	if (aunet->sk == NULL) {
 		audit_panic("cannot initialize netlink socket in namespace");
 		return -ENOMEM;
 	}
-	aunet->nlsk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+	aunet->sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
+
 	return 0;
 }
 
 static void __net_exit audit_net_exit(struct net *net)
 {
 	struct audit_net *aunet = net_generic(net, audit_net_id);
-	struct sock *sock = aunet->nlsk;
-	mutex_lock(&audit_cmd_mutex);
-	if (sock == audit_sock)
-		auditd_reset();
-	mutex_unlock(&audit_cmd_mutex);
 
-	netlink_kernel_release(sock);
-	aunet->nlsk = NULL;
+	rcu_read_lock();
+	if (net == auditd_conn.net)
+		auditd_reset();
+	rcu_read_unlock();
+
+	netlink_kernel_release(aunet->sk);
 }
 
 static struct pernet_operations audit_net_ops __net_initdata = {
@@ -1335,20 +1473,24 @@ static int __init audit_init(void)
 	if (audit_initialized == AUDIT_DISABLED)
 		return 0;
 
-	pr_info("initializing netlink subsys (%s)\n",
-		audit_default ? "enabled" : "disabled");
-	register_pernet_subsys(&audit_net_ops);
+	memset(&auditd_conn, 0, sizeof(auditd_conn));
+	spin_lock_init(&auditd_conn.lock);
 
 	skb_queue_head_init(&audit_queue);
 	skb_queue_head_init(&audit_retry_queue);
 	skb_queue_head_init(&audit_hold_queue);
-	audit_initialized = AUDIT_INITIALIZED;
-	audit_enabled = audit_default;
-	audit_ever_enabled |= !!audit_default;
 
 	for (i = 0; i < AUDIT_INODE_BUCKETS; i++)
 		INIT_LIST_HEAD(&audit_inode_hash[i]);
 
+	pr_info("initializing netlink subsys (%s)\n",
+		audit_default ? "enabled" : "disabled");
+	register_pernet_subsys(&audit_net_ops);
+
+	audit_initialized = AUDIT_INITIALIZED;
+	audit_enabled = audit_default;
+	audit_ever_enabled |= !!audit_default;
+
 	kauditd_task = kthread_run(kauditd_thread, NULL, "kauditd");
 	if (IS_ERR(kauditd_task)) {
 		int err = PTR_ERR(kauditd_task);
@@ -1519,20 +1661,16 @@ struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
 	if (unlikely(!audit_filter(type, AUDIT_FILTER_TYPE)))
 		return NULL;
 
-	/* don't ever fail/sleep on these two conditions:
+	/* NOTE: don't ever fail/sleep on these two conditions:
 	 * 1. auditd generated record - since we need auditd to drain the
 	 *    queue; also, when we are checking for auditd, compare PIDs using
 	 *    task_tgid_vnr() since auditd_pid is set in audit_receive_msg()
 	 *    using a PID anchored in the caller's namespace
-	 * 2. audit command message - record types 1000 through 1099 inclusive
-	 *    are command messages/records used to manage the kernel subsystem
-	 *    and the audit userspace, blocking on these messages could cause
-	 *    problems under load so don't do it (note: not all of these
-	 *    command types are valid as record types, but it is quicker to
-	 *    just check two ints than a series of ints in a if/switch stmt) */
-	if (!((audit_pid && audit_pid == task_tgid_vnr(current)) ||
-	      (type >= 1000 && type <= 1099))) {
-		long sleep_time = audit_backlog_wait_time;
+	 * 2. generator holding the audit_cmd_mutex - we don't want to block
+	 *    while holding the mutex */
+	if (!(auditd_test_task(current) ||
+	      (current == __mutex_owner(&audit_cmd_mutex)))) {
+		long stime = audit_backlog_wait_time;
 
 		while (audit_backlog_limit &&
 		       (skb_queue_len(&audit_queue) > audit_backlog_limit)) {
@@ -1541,14 +1679,13 @@ struct audit_buffer *audit_log_start(struct audit_context *ctx, gfp_t gfp_mask,
 
 			/* sleep if we are allowed and we haven't exhausted our
 			 * backlog wait limit */
-			if ((gfp_mask & __GFP_DIRECT_RECLAIM) &&
-			    (sleep_time > 0)) {
+			if (gfpflags_allow_blocking(gfp_mask) && (stime > 0)) {
 				DECLARE_WAITQUEUE(wait, current);
 
 				add_wait_queue_exclusive(&audit_backlog_wait,
 							 &wait);
 				set_current_state(TASK_UNINTERRUPTIBLE);
-				sleep_time = schedule_timeout(sleep_time);
+				stime = schedule_timeout(stime);
 				remove_wait_queue(&audit_backlog_wait, &wait);
 			} else {
 				if (audit_rate_check() && printk_ratelimit())
@@ -2127,15 +2264,27 @@ out:
  */
 void audit_log_end(struct audit_buffer *ab)
 {
+	struct sk_buff *skb;
+	struct nlmsghdr *nlh;
+
 	if (!ab)
 		return;
-	if (!audit_rate_check()) {
-		audit_log_lost("rate limit exceeded");
-	} else {
-		skb_queue_tail(&audit_queue, ab->skb);
-		wake_up_interruptible(&kauditd_wait);
+
+	if (audit_rate_check()) {
+		skb = ab->skb;
 		ab->skb = NULL;
-	}
+
+		/* setup the netlink header, see the comments in
+		 * kauditd_send_multicast_skb() for length quirks */
+		nlh = nlmsg_hdr(skb);
+		nlh->nlmsg_len = skb->len - NLMSG_HDRLEN;
+
+		/* queue the netlink packet and poke the kauditd thread */
+		skb_queue_tail(&audit_queue, skb);
+		wake_up_interruptible(&kauditd_wait);
+	} else
+		audit_log_lost("rate limit exceeded");
+
 	audit_buffer_free(ab);
 }
 

kernel/audit.h

@@ -218,7 +218,7 @@ extern void audit_log_name(struct audit_context *context,
 			   struct audit_names *n, const struct path *path,
 			   int record_num, int *call_panic);
 
-extern int audit_pid;
+extern int auditd_test_task(const struct task_struct *task);
 
 #define AUDIT_INODE_BUCKETS	32
 extern struct list_head audit_inode_hash[AUDIT_INODE_BUCKETS];
@@ -250,10 +250,6 @@ struct audit_netlink_list {
 
 int audit_send_list(void *);
 
-struct audit_net {
-	struct sock *nlsk;
-};
-
 extern int selinux_audit_rule_update(void);
 
 extern struct mutex audit_filter_mutex;
@@ -340,8 +336,7 @@ extern int audit_filter(int msgtype, unsigned int listtype);
 extern int __audit_signal_info(int sig, struct task_struct *t);
 static inline int audit_signal_info(int sig, struct task_struct *t)
 {
-	if (unlikely((audit_pid && t->tgid == audit_pid) ||
-		     (audit_signals && !audit_dummy_context())))
+	if (auditd_test_task(t) || (audit_signals && !audit_dummy_context()))
 		return __audit_signal_info(sig, t);
 	return 0;
 }

kernel/auditsc.c

@@ -762,7 +762,7 @@ static enum audit_state audit_filter_syscall(struct task_struct *tsk,
 	struct audit_entry *e;
 	enum audit_state state;
 
-	if (audit_pid && tsk->tgid == audit_pid)
+	if (auditd_test_task(tsk))
 		return AUDIT_DISABLED;
 
 	rcu_read_lock();
@@ -816,7 +816,7 @@ void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
 {
 	struct audit_names *n;
 
-	if (audit_pid && tsk->tgid == audit_pid)
+	if (auditd_test_task(tsk))
 		return;
 
 	rcu_read_lock();
@@ -2256,7 +2256,7 @@ int __audit_signal_info(int sig, struct task_struct *t)
 	struct audit_context *ctx = tsk->audit_context;
 	kuid_t uid = current_uid(), t_uid = task_uid(t);
 
-	if (audit_pid && t->tgid == audit_pid) {
+	if (auditd_test_task(t)) {
 		if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
 			audit_sig_pid = task_tgid_nr(tsk);
 			if (uid_valid(tsk->loginuid))

kernel/bpf/hashtab.c

@@ -30,18 +30,12 @@ struct bpf_htab {
 		struct pcpu_freelist freelist;
 		struct bpf_lru lru;
 	};
-	void __percpu *extra_elems;
+	struct htab_elem *__percpu *extra_elems;
 	atomic_t count;	/* number of elements in this hashtable */
 	u32 n_buckets;	/* number of hash buckets */
 	u32 elem_size;	/* size of each element in bytes */
 };
 
-enum extra_elem_state {
-	HTAB_NOT_AN_EXTRA_ELEM = 0,
-	HTAB_EXTRA_ELEM_FREE,
-	HTAB_EXTRA_ELEM_USED
-};
-
 /* each htab element is struct htab_elem + key + value */
 struct htab_elem {
 	union {
@@ -56,7 +50,6 @@ struct htab_elem {
 	};
 	union {
 		struct rcu_head rcu;
-		enum extra_elem_state state;
 		struct bpf_lru_node lru_node;
 	};
 	u32 hash;
@@ -77,6 +70,11 @@ static bool htab_is_percpu(const struct bpf_htab *htab)
 		htab->map.map_type == BPF_MAP_TYPE_LRU_PERCPU_HASH;
 }
 
+static bool htab_is_prealloc(const struct bpf_htab *htab)
+{
+	return !(htab->map.map_flags & BPF_F_NO_PREALLOC);
+}
+
 static inline void htab_elem_set_ptr(struct htab_elem *l, u32 key_size,
 				     void __percpu *pptr)
 {
@@ -128,17 +126,20 @@ static struct htab_elem *prealloc_lru_pop(struct bpf_htab *htab, void *key,
 
 static int prealloc_init(struct bpf_htab *htab)
 {
+	u32 num_entries = htab->map.max_entries;
 	int err = -ENOMEM, i;
 
-	htab->elems = bpf_map_area_alloc(htab->elem_size *
-					 htab->map.max_entries);
+	if (!htab_is_percpu(htab) && !htab_is_lru(htab))
+		num_entries += num_possible_cpus();
+
+	htab->elems = bpf_map_area_alloc(htab->elem_size * num_entries);
 	if (!htab->elems)
 		return -ENOMEM;
 
 	if (!htab_is_percpu(htab))
 		goto skip_percpu_elems;
 
-	for (i = 0; i < htab->map.max_entries; i++) {
+	for (i = 0; i < num_entries; i++) {
 		u32 size = round_up(htab->map.value_size, 8);
 		void __percpu *pptr;
 
@@ -166,11 +167,11 @@ skip_percpu_elems:
 	if (htab_is_lru(htab))
 		bpf_lru_populate(&htab->lru, htab->elems,
 				 offsetof(struct htab_elem, lru_node),
-				 htab->elem_size, htab->map.max_entries);
+				 htab->elem_size, num_entries);
 	else
 		pcpu_freelist_populate(&htab->freelist,
 				       htab->elems + offsetof(struct htab_elem, fnode),
-				       htab->elem_size, htab->map.max_entries);
+				       htab->elem_size, num_entries);
 
 	return 0;
 
@@ -191,16 +192,22 @@ static void prealloc_destroy(struct bpf_htab *htab)
 
 static int alloc_extra_elems(struct bpf_htab *htab)
 {
-	void __percpu *pptr;
+	struct htab_elem *__percpu *pptr, *l_new;
+	struct pcpu_freelist_node *l;
 	int cpu;
 
-	pptr = __alloc_percpu_gfp(htab->elem_size, 8, GFP_USER | __GFP_NOWARN);
+	pptr = __alloc_percpu_gfp(sizeof(struct htab_elem *), 8,
+				  GFP_USER | __GFP_NOWARN);
 	if (!pptr)
 		return -ENOMEM;
 
 	for_each_possible_cpu(cpu) {
-		((struct htab_elem *)per_cpu_ptr(pptr, cpu))->state =
-			HTAB_EXTRA_ELEM_FREE;
+		l = pcpu_freelist_pop(&htab->freelist);
+		/* pop will succeed, since prealloc_init()
+		 * preallocated extra num_possible_cpus elements
+		 */
+		l_new = container_of(l, struct htab_elem, fnode);
+		*per_cpu_ptr(pptr, cpu) = l_new;
 	}
 	htab->extra_elems = pptr;
 	return 0;
@@ -342,25 +349,25 @@ static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
 		raw_spin_lock_init(&htab->buckets[i].lock);
 	}
 
-	if (!percpu && !lru) {
-		/* lru itself can remove the least used element, so
-		 * there is no need for an extra elem during map_update.
-		 */
-		err = alloc_extra_elems(htab);
-		if (err)
-			goto free_buckets;
-	}
-
 	if (prealloc) {
 		err = prealloc_init(htab);
 		if (err)
-			goto free_extra_elems;
+			goto free_buckets;
+
+		if (!percpu && !lru) {
+			/* lru itself can remove the least used element, so
+			 * there is no need for an extra elem during map_update.
+			 */
+			err = alloc_extra_elems(htab);
+			if (err)
+				goto free_prealloc;
+		}
 	}
 
 	return &htab->map;
 
-free_extra_elems:
-	free_percpu(htab->extra_elems);
+free_prealloc:
+	prealloc_destroy(htab);
 free_buckets:
 	bpf_map_area_free(htab->buckets);
 free_htab:
@@ -575,12 +582,7 @@ static void htab_elem_free_rcu(struct rcu_head *head)
 
 static void free_htab_elem(struct bpf_htab *htab, struct htab_elem *l)
 {
-	if (l->state == HTAB_EXTRA_ELEM_USED) {
-		l->state = HTAB_EXTRA_ELEM_FREE;
-		return;
-	}
-
-	if (!(htab->map.map_flags & BPF_F_NO_PREALLOC)) {
+	if (htab_is_prealloc(htab)) {
 		pcpu_freelist_push(&htab->freelist, &l->fnode);
 	} else {
 		atomic_dec(&htab->count);
@@ -610,47 +612,43 @@ static void pcpu_copy_value(struct bpf_htab *htab, void __percpu *pptr,
 static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
 					 void *value, u32 key_size, u32 hash,
 					 bool percpu, bool onallcpus,
-					 bool old_elem_exists)
+					 struct htab_elem *old_elem)
 {
 	u32 size = htab->map.value_size;
-	bool prealloc = !(htab->map.map_flags & BPF_F_NO_PREALLOC);
-	struct htab_elem *l_new;
+	bool prealloc = htab_is_prealloc(htab);
+	struct htab_elem *l_new, **pl_new;
 	void __percpu *pptr;
-	int err = 0;
 
 	if (prealloc) {
-		struct pcpu_freelist_node *l;
-
-		l = pcpu_freelist_pop(&htab->freelist);
-		if (!l)
-			err = -E2BIG;
-		else
-			l_new = container_of(l, struct htab_elem, fnode);
-	} else {
-		if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
-			atomic_dec(&htab->count);
-			err = -E2BIG;
+		if (old_elem) {
+			/* if we're updating the existing element,
+			 * use per-cpu extra elems to avoid freelist_pop/push
+			 */
+			pl_new = this_cpu_ptr(htab->extra_elems);
+			l_new = *pl_new;
+			*pl_new = old_elem;
 		} else {
-			l_new = kmalloc(htab->elem_size,
-					GFP_ATOMIC | __GFP_NOWARN);
-			if (!l_new)
-				return ERR_PTR(-ENOMEM);
+			struct pcpu_freelist_node *l;
+
+			l = pcpu_freelist_pop(&htab->freelist);
+			if (!l)
+				return ERR_PTR(-E2BIG);
+			l_new = container_of(l, struct htab_elem, fnode);
 		}
-	}
-
-	if (err) {
-		if (!old_elem_exists)
-			return ERR_PTR(err);
-
-		/* if we're updating the existing element and the hash table
-		 * is full, use per-cpu extra elems
-		 */
-		l_new = this_cpu_ptr(htab->extra_elems);
-		if (l_new->state != HTAB_EXTRA_ELEM_FREE)
-			return ERR_PTR(-E2BIG);
-		l_new->state = HTAB_EXTRA_ELEM_USED;
 	} else {
-		l_new->state = HTAB_NOT_AN_EXTRA_ELEM;
+		if (atomic_inc_return(&htab->count) > htab->map.max_entries)
+			if (!old_elem) {
+				/* when map is full and update() is replacing
+				 * old element, it's ok to allocate, since
+				 * old element will be freed immediately.
+				 * Otherwise return an error
+				 */
+				atomic_dec(&htab->count);
+				return ERR_PTR(-E2BIG);
+			}
+		l_new = kmalloc(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN);
+		if (!l_new)
+			return ERR_PTR(-ENOMEM);
 	}
 
 	memcpy(l_new->key, key, key_size);
@@ -731,7 +729,7 @@ static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
 		goto err;
 
 	l_new = alloc_htab_elem(htab, key, value, key_size, hash, false, false,
-				!!l_old);
+				l_old);
 	if (IS_ERR(l_new)) {
 		/* all pre-allocated elements are in use or memory exhausted */
 		ret = PTR_ERR(l_new);
@@ -744,7 +742,8 @@ static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
 	hlist_nulls_add_head_rcu(&l_new->hash_node, head);
 	if (l_old) {
 		hlist_nulls_del_rcu(&l_old->hash_node);
-		free_htab_elem(htab, l_old);
+		if (!htab_is_prealloc(htab))
+			free_htab_elem(htab, l_old);
 	}
 	ret = 0;
 err:
@@ -856,7 +855,7 @@ static int __htab_percpu_map_update_elem(struct bpf_map *map, void *key,
 				value, onallcpus);
 	} else {
 		l_new = alloc_htab_elem(htab, key, value, key_size,
-					hash, true, onallcpus, false);
+					hash, true, onallcpus, NULL);
 		if (IS_ERR(l_new)) {
 			ret = PTR_ERR(l_new);
 			goto err;
@@ -1024,8 +1023,7 @@ static void delete_all_elements(struct bpf_htab *htab)
 
 		hlist_nulls_for_each_entry_safe(l, n, head, hash_node) {
 			hlist_nulls_del_rcu(&l->hash_node);
-			if (l->state != HTAB_EXTRA_ELEM_USED)
-				htab_elem_free(htab, l);
+			htab_elem_free(htab, l);
 		}
 	}
 }
@@ -1045,7 +1043,7 @@ static void htab_map_free(struct bpf_map *map)
 	 * not have executed. Wait for them.
 	 */
 	rcu_barrier();
-	if (htab->map.map_flags & BPF_F_NO_PREALLOC)
+	if (!htab_is_prealloc(htab))
 		delete_all_elements(htab);
 	else
 		prealloc_destroy(htab);

kernel/cpu.c

@@ -1335,26 +1335,21 @@ static int cpuhp_store_callbacks(enum cpuhp_state state, const char *name,
 	struct cpuhp_step *sp;
 	int ret = 0;
 
-	mutex_lock(&cpuhp_state_mutex);
-
 	if (state == CPUHP_AP_ONLINE_DYN || state == CPUHP_BP_PREPARE_DYN) {
 		ret = cpuhp_reserve_state(state);
 		if (ret < 0)
-			goto out;
+			return ret;
 		state = ret;
 	}
 	sp = cpuhp_get_step(state);
-	if (name && sp->name) {
-		ret = -EBUSY;
-		goto out;
-	}
+	if (name && sp->name)
+		return -EBUSY;
+
 	sp->startup.single = startup;
 	sp->teardown.single = teardown;
 	sp->name = name;
 	sp->multi_instance = multi_instance;
 	INIT_HLIST_HEAD(&sp->list);
-out:
-	mutex_unlock(&cpuhp_state_mutex);
 	return ret;
 }
 
@@ -1428,6 +1423,7 @@ int __cpuhp_state_add_instance(enum cpuhp_state state, struct hlist_node *node,
 		return -EINVAL;
 
 	get_online_cpus();
+	mutex_lock(&cpuhp_state_mutex);
 
 	if (!invoke || !sp->startup.multi)
 		goto add_node;
@@ -1447,16 +1443,14 @@ int __cpuhp_state_add_instance(enum cpuhp_state state, struct hlist_node *node,
 		if (ret) {
 			if (sp->teardown.multi)
 				cpuhp_rollback_install(cpu, state, node);
-			goto err;
+			goto unlock;
 		}
 	}
 add_node:
 	ret = 0;
-	mutex_lock(&cpuhp_state_mutex);
 	hlist_add_head(node, &sp->list);
+unlock:
 	mutex_unlock(&cpuhp_state_mutex);
-
-err:
 	put_online_cpus();
 	return ret;
 }
@@ -1491,6 +1485,7 @@ int __cpuhp_setup_state(enum cpuhp_state state,
 		return -EINVAL;
 
 	get_online_cpus();
+	mutex_lock(&cpuhp_state_mutex);
 
 	ret = cpuhp_store_callbacks(state, name, startup, teardown,
 				    multi_instance);
@@ -1524,6 +1519,7 @@ int __cpuhp_setup_state(enum cpuhp_state state,
 		}
 	}
 out:
+	mutex_unlock(&cpuhp_state_mutex);
 	put_online_cpus();
 	/*
 	 * If the requested state is CPUHP_AP_ONLINE_DYN, return the
@@ -1547,6 +1543,8 @@ int __cpuhp_state_remove_instance(enum cpuhp_state state,
 		return -EINVAL;
 
 	get_online_cpus();
+	mutex_lock(&cpuhp_state_mutex);
+
 	if (!invoke || !cpuhp_get_teardown_cb(state))
 		goto remove;
 	/*
@@ -1563,7 +1561,6 @@ int __cpuhp_state_remove_instance(enum cpuhp_state state,
 	}
 
 remove:
-	mutex_lock(&cpuhp_state_mutex);
 	hlist_del(node);
 	mutex_unlock(&cpuhp_state_mutex);
 	put_online_cpus();
@@ -1571,6 +1568,7 @@ remove:
 	return 0;
 }
 EXPORT_SYMBOL_GPL(__cpuhp_state_remove_instance);
+
 /**
  * __cpuhp_remove_state - Remove the callbacks for an hotplug machine state
  * @state:	The state to remove
@@ -1589,6 +1587,7 @@ void __cpuhp_remove_state(enum cpuhp_state state, bool invoke)
 
 	get_online_cpus();
 
+	mutex_lock(&cpuhp_state_mutex);
 	if (sp->multi_instance) {
 		WARN(!hlist_empty(&sp->list),
 		     "Error: Removing state %d which has instances left.\n",
@@ -1613,6 +1612,7 @@ void __cpuhp_remove_state(enum cpuhp_state state, bool invoke)
 	}
 remove:
 	cpuhp_store_callbacks(state, NULL, NULL, NULL, false);
+	mutex_unlock(&cpuhp_state_mutex);
 	put_online_cpus();
 }
 EXPORT_SYMBOL(__cpuhp_remove_state);

kernel/futex.c

@@ -2815,7 +2815,6 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 {
 	struct hrtimer_sleeper timeout, *to = NULL;
 	struct rt_mutex_waiter rt_waiter;
-	struct rt_mutex *pi_mutex = NULL;
 	struct futex_hash_bucket *hb;
 	union futex_key key2 = FUTEX_KEY_INIT;
 	struct futex_q q = futex_q_init;
@@ -2899,6 +2898,8 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 		if (q.pi_state && (q.pi_state->owner != current)) {
 			spin_lock(q.lock_ptr);
 			ret = fixup_pi_state_owner(uaddr2, &q, current);
+			if (ret && rt_mutex_owner(&q.pi_state->pi_mutex) == current)
+				rt_mutex_unlock(&q.pi_state->pi_mutex);
 			/*
 			 * Drop the reference to the pi state which
 			 * the requeue_pi() code acquired for us.
@@ -2907,6 +2908,8 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 			spin_unlock(q.lock_ptr);
 		}
 	} else {
+		struct rt_mutex *pi_mutex;
+
 		/*
 		 * We have been woken up by futex_unlock_pi(), a timeout, or a
 		 * signal.  futex_unlock_pi() will not destroy the lock_ptr nor
@@ -2930,18 +2933,19 @@ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags,
 		if (res)
 			ret = (res < 0) ? res : 0;
 
+		/*
+		 * If fixup_pi_state_owner() faulted and was unable to handle
+		 * the fault, unlock the rt_mutex and return the fault to
+		 * userspace.
+		 */
+		if (ret && rt_mutex_owner(pi_mutex) == current)
+			rt_mutex_unlock(pi_mutex);
+
 		/* Unqueue and drop the lock. */
 		unqueue_me_pi(&q);
 	}
 
-	/*
-	 * If fixup_pi_state_owner() faulted and was unable to handle the
-	 * fault, unlock the rt_mutex and return the fault to userspace.
-	 */
-	if (ret == -EFAULT) {
-		if (pi_mutex && rt_mutex_owner(pi_mutex) == current)
-			rt_mutex_unlock(pi_mutex);
-	} else if (ret == -EINTR) {
+	if (ret == -EINTR) {
 		/*
 		 * We've already been requeued, but cannot restart by calling
 		 * futex_lock_pi() directly. We could restart this syscall, but

kernel/locking/rwsem-spinlock.c

@@ -213,10 +213,9 @@ int __sched __down_write_common(struct rw_semaphore *sem, int state)
 		 */
 		if (sem->count == 0)
 			break;
-		if (signal_pending_state(state, current)) {
-			ret = -EINTR;
-			goto out;
-		}
+		if (signal_pending_state(state, current))
+			goto out_nolock;
+
 		set_current_state(state);
 		raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
 		schedule();
@@ -224,12 +223,19 @@ int __sched __down_write_common(struct rw_semaphore *sem, int state)
 	}
 	/* got the lock */
 	sem->count = -1;
-out:
 	list_del(&waiter.list);
 
 	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
 
 	return ret;
+
+out_nolock:
+	list_del(&waiter.list);
+	if (!list_empty(&sem->wait_list))
+		__rwsem_do_wake(sem, 1);
+	raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
+
+	return -EINTR;
 }
 
 void __sched __down_write(struct rw_semaphore *sem)

kernel/memremap.c

@@ -247,11 +247,9 @@ static void devm_memremap_pages_release(struct device *dev, void *data)
 	align_start = res->start & ~(SECTION_SIZE - 1);
 	align_size = ALIGN(resource_size(res), SECTION_SIZE);
 
-	lock_device_hotplug();
 	mem_hotplug_begin();
 	arch_remove_memory(align_start, align_size);
 	mem_hotplug_done();
-	unlock_device_hotplug();
 
 	untrack_pfn(NULL, PHYS_PFN(align_start), align_size);
 	pgmap_radix_release(res);
@@ -364,11 +362,9 @@ void *devm_memremap_pages(struct device *dev, struct resource *res,
 	if (error)
 		goto err_pfn_remap;
 
-	lock_device_hotplug();
 	mem_hotplug_begin();
 	error = arch_add_memory(nid, align_start, align_size, true);
 	mem_hotplug_done();
-	unlock_device_hotplug();
 	if (error)
 		goto err_add_memory;
 

kernel/sched/cpufreq_schedutil.c

@@ -584,20 +584,14 @@ static int sugov_start(struct cpufreq_policy *policy)
 	for_each_cpu(cpu, policy->cpus) {
 		struct sugov_cpu *sg_cpu = &per_cpu(sugov_cpu, cpu);
 
+		memset(sg_cpu, 0, sizeof(*sg_cpu));
 		sg_cpu->sg_policy = sg_policy;
-		if (policy_is_shared(policy)) {
-			sg_cpu->util = 0;
-			sg_cpu->max = 0;
-			sg_cpu->flags = SCHED_CPUFREQ_RT;
-			sg_cpu->last_update = 0;
-			sg_cpu->iowait_boost = 0;
-			sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
-			cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
-						     sugov_update_shared);
-		} else {
-			cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
-						     sugov_update_single);
-		}
+		sg_cpu->flags = SCHED_CPUFREQ_RT;
+		sg_cpu->iowait_boost_max = policy->cpuinfo.max_freq;
+		cpufreq_add_update_util_hook(cpu, &sg_cpu->update_util,
+					     policy_is_shared(policy) ?
+							sugov_update_shared :
+							sugov_update_single);
 	}
 	return 0;
 }

kernel/sched/deadline.c

@@ -445,13 +445,13 @@ static void replenish_dl_entity(struct sched_dl_entity *dl_se,
  *
  * This function returns true if:
  *
- *   runtime / (deadline - t) > dl_runtime / dl_period ,
+ *   runtime / (deadline - t) > dl_runtime / dl_deadline ,
  *
  * IOW we can't recycle current parameters.
  *
- * Notice that the bandwidth check is done against the period. For
+ * Notice that the bandwidth check is done against the deadline. For
  * task with deadline equal to period this is the same of using
- * dl_deadline instead of dl_period in the equation above.
+ * dl_period instead of dl_deadline in the equation above.
  */
 static bool dl_entity_overflow(struct sched_dl_entity *dl_se,
 			       struct sched_dl_entity *pi_se, u64 t)
@@ -476,7 +476,7 @@ static bool dl_entity_overflow(struct sched_dl_entity *dl_se,
 	 * of anything below microseconds resolution is actually fiction
 	 * (but still we want to give the user that illusion >;).
 	 */
-	left = (pi_se->dl_period >> DL_SCALE) * (dl_se->runtime >> DL_SCALE);
+	left = (pi_se->dl_deadline >> DL_SCALE) * (dl_se->runtime >> DL_SCALE);
 	right = ((dl_se->deadline - t) >> DL_SCALE) *
 		(pi_se->dl_runtime >> DL_SCALE);
 
@@ -505,10 +505,15 @@ static void update_dl_entity(struct sched_dl_entity *dl_se,
 	}
 }
 
+static inline u64 dl_next_period(struct sched_dl_entity *dl_se)
+{
+	return dl_se->deadline - dl_se->dl_deadline + dl_se->dl_period;
+}
+
 /*
  * If the entity depleted all its runtime, and if we want it to sleep
  * while waiting for some new execution time to become available, we
- * set the bandwidth enforcement timer to the replenishment instant
+ * set the bandwidth replenishment timer to the replenishment instant
  * and try to activate it.
  *
  * Notice that it is important for the caller to know if the timer
@@ -530,7 +535,7 @@ static int start_dl_timer(struct task_struct *p)
 	 * that it is actually coming from rq->clock and not from
 	 * hrtimer's time base reading.
 	 */
-	act = ns_to_ktime(dl_se->deadline);
+	act = ns_to_ktime(dl_next_period(dl_se));
 	now = hrtimer_cb_get_time(timer);
 	delta = ktime_to_ns(now) - rq_clock(rq);
 	act = ktime_add_ns(act, delta);
@@ -638,6 +643,7 @@ static enum hrtimer_restart dl_task_timer(struct hrtimer *timer)
 		lockdep_unpin_lock(&rq->lock, rf.cookie);
 		rq = dl_task_offline_migration(rq, p);
 		rf.cookie = lockdep_pin_lock(&rq->lock);
+		update_rq_clock(rq);
 
 		/*
 		 * Now that the task has been migrated to the new RQ and we
@@ -689,6 +695,37 @@ void init_dl_task_timer(struct sched_dl_entity *dl_se)
 	timer->function = dl_task_timer;
 }
 
+/*
+ * During the activation, CBS checks if it can reuse the current task's
+ * runtime and period. If the deadline of the task is in the past, CBS
+ * cannot use the runtime, and so it replenishes the task. This rule
+ * works fine for implicit deadline tasks (deadline == period), and the
+ * CBS was designed for implicit deadline tasks. However, a task with
+ * constrained deadline (deadine < period) might be awakened after the
+ * deadline, but before the next period. In this case, replenishing the
+ * task would allow it to run for runtime / deadline. As in this case
+ * deadline < period, CBS enables a task to run for more than the
+ * runtime / period. In a very loaded system, this can cause a domino
+ * effect, making other tasks miss their deadlines.
+ *
+ * To avoid this problem, in the activation of a constrained deadline
+ * task after the deadline but before the next period, throttle the
+ * task and set the replenishing timer to the begin of the next period,
+ * unless it is boosted.
+ */
+static inline void dl_check_constrained_dl(struct sched_dl_entity *dl_se)
+{
+	struct task_struct *p = dl_task_of(dl_se);
+	struct rq *rq = rq_of_dl_rq(dl_rq_of_se(dl_se));
+
+	if (dl_time_before(dl_se->deadline, rq_clock(rq)) &&
+	    dl_time_before(rq_clock(rq), dl_next_period(dl_se))) {
+		if (unlikely(dl_se->dl_boosted || !start_dl_timer(p)))
+			return;
+		dl_se->dl_throttled = 1;
+	}
+}
+
 static
 int dl_runtime_exceeded(struct sched_dl_entity *dl_se)
 {
@@ -922,6 +959,11 @@ static void dequeue_dl_entity(struct sched_dl_entity *dl_se)
 	__dequeue_dl_entity(dl_se);
 }
 
+static inline bool dl_is_constrained(struct sched_dl_entity *dl_se)
+{
+	return dl_se->dl_deadline < dl_se->dl_period;
+}
+
 static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 {
 	struct task_struct *pi_task = rt_mutex_get_top_task(p);
@@ -947,6 +989,15 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 		return;
 	}
 
+	/*
+	 * Check if a constrained deadline task was activated
+	 * after the deadline but before the next period.
+	 * If that is the case, the task will be throttled and
+	 * the replenishment timer will be set to the next period.
+	 */
+	if (!p->dl.dl_throttled && dl_is_constrained(&p->dl))
+		dl_check_constrained_dl(&p->dl);
+
 	/*
 	 * If p is throttled, we do nothing. In fact, if it exhausted
 	 * its budget it needs a replenishment and, since it now is on

kernel/sched/loadavg.c

@@ -169,7 +169,7 @@ static inline int calc_load_write_idx(void)
 	 * If the folding window started, make sure we start writing in the
 	 * next idle-delta.
 	 */
-	if (!time_before(jiffies, calc_load_update))
+	if (!time_before(jiffies, READ_ONCE(calc_load_update)))
 		idx++;
 
 	return idx & 1;
@@ -202,8 +202,9 @@ void calc_load_exit_idle(void)
 	struct rq *this_rq = this_rq();
 
 	/*
-	 * If we're still before the sample window, we're done.
+	 * If we're still before the pending sample window, we're done.
 	 */
+	this_rq->calc_load_update = READ_ONCE(calc_load_update);
 	if (time_before(jiffies, this_rq->calc_load_update))
 		return;
 
@@ -212,7 +213,6 @@ void calc_load_exit_idle(void)
 	 * accounted through the nohz accounting, so skip the entire deal and
 	 * sync up for the next window.
 	 */
-	this_rq->calc_load_update = calc_load_update;
 	if (time_before(jiffies, this_rq->calc_load_update + 10))
 		this_rq->calc_load_update += LOAD_FREQ;
 }
@@ -308,13 +308,15 @@ calc_load_n(unsigned long load, unsigned long exp,
  */
 static void calc_global_nohz(void)
 {
+	unsigned long sample_window;
 	long delta, active, n;
 
-	if (!time_before(jiffies, calc_load_update + 10)) {
+	sample_window = READ_ONCE(calc_load_update);
+	if (!time_before(jiffies, sample_window + 10)) {
 		/*
 		 * Catch-up, fold however many we are behind still
 		 */
-		delta = jiffies - calc_load_update - 10;
+		delta = jiffies - sample_window - 10;
 		n = 1 + (delta / LOAD_FREQ);
 
 		active = atomic_long_read(&calc_load_tasks);
@@ -324,7 +326,7 @@ static void calc_global_nohz(void)
 		avenrun[1] = calc_load_n(avenrun[1], EXP_5, active, n);
 		avenrun[2] = calc_load_n(avenrun[2], EXP_15, active, n);
 
-		calc_load_update += n * LOAD_FREQ;
+		WRITE_ONCE(calc_load_update, sample_window + n * LOAD_FREQ);
 	}
 
 	/*
@@ -352,9 +354,11 @@ static inline void calc_global_nohz(void) { }
  */
 void calc_global_load(unsigned long ticks)
 {
+	unsigned long sample_window;
 	long active, delta;
 
-	if (time_before(jiffies, calc_load_update + 10))
+	sample_window = READ_ONCE(calc_load_update);
+	if (time_before(jiffies, sample_window + 10))
 		return;
 
 	/*
@@ -371,7 +375,7 @@ void calc_global_load(unsigned long ticks)
 	avenrun[1] = calc_load(avenrun[1], EXP_5, active);
 	avenrun[2] = calc_load(avenrun[2], EXP_15, active);
 
-	calc_load_update += LOAD_FREQ;
+	WRITE_ONCE(calc_load_update, sample_window + LOAD_FREQ);
 
 	/*
 	 * In case we idled for multiple LOAD_FREQ intervals, catch up in bulk.