linux-kernel - [patch 1/7] tick/broadcast: Serialize access to tick_next

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 17 Nov 2020 14:19:43 +0100
From:   Thomas Gleixner <tglx@...utronix.de>
To:     LKML <linux-kernel@...r.kernel.org>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Frederic Weisbecker <frederic@...nel.org>,
        Yunfeng Ye <yeyunfeng@...wei.com>
Subject: [patch 1/7] tick/broadcast: Serialize access to tick_next_period

tick_broadcast_setup_oneshot() accesses tick_next_period twice without any
serialization. This is wrong in two aspects:

  - Reading it twice might make the broadcast data inconsistent if the
    variable is updated concurrently.

  - On 32bit systems the access might see an partial update

Protect it with jiffies_lock. That's safe as none of the callchains leading
up to this function can create a lock ordering violation:

timer interrupt
  run_local_timers()
    hrtimer_run_queues()
      hrtimer_switch_to_hres()
        tick_init_highres()
	  tick_switch_to_oneshot()
	    tick_broadcast_switch_to_oneshot()
or
     tick_check_oneshot_change()
       tick_nohz_switch_to_nohz()
         tick_switch_to_oneshot()
           tick_broadcast_switch_to_oneshot()

Signed-off-by: Thomas Gleixner <tglx@...utronix.de>
---
 kernel/time/tick-broadcast.c |   23 ++++++++++++++++++++---
 1 file changed, 20 insertions(+), 3 deletions(-)

--- a/kernel/time/tick-broadcast.c
+++ b/kernel/time/tick-broadcast.c
@@ -877,6 +877,22 @@ static void tick_broadcast_init_next_eve
 	}
 }
 
+static inline ktime_t tick_get_next_period(void)
+{
+	ktime_t next;
+
+	/*
+	 * Protect against concurrent updates (store /load tearing on
+	 * 32bit). It does not matter if the time is already in the
+	 * past. The broadcast device which is about to be programmed will
+	 * fire in any case.
+	 */
+	raw_spin_lock(&jiffies_lock);
+	next = tick_next_period;
+	raw_spin_unlock(&jiffies_lock);
+	return next;
+}
+
 /**
  * tick_broadcast_setup_oneshot - setup the broadcast device
  */
@@ -905,10 +921,11 @@ static void tick_broadcast_setup_oneshot
 			   tick_broadcast_oneshot_mask, tmpmask);
 
 		if (was_periodic && !cpumask_empty(tmpmask)) {
+			ktime_t nextevt = tick_get_next_period();
+
 			clockevents_switch_state(bc, CLOCK_EVT_STATE_ONESHOT);
-			tick_broadcast_init_next_event(tmpmask,
-						       tick_next_period);
-			tick_broadcast_set_event(bc, cpu, tick_next_period);
+			tick_broadcast_init_next_event(tmpmask, nextevt);
+			tick_broadcast_set_event(bc, cpu, nextevt);
 		} else
 			bc->next_event = KTIME_MAX;
 	} else {