linux-kernel - [PATCH] nohz_full: Fix TICK_DO_TIMER_NONE vs nohz

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>] [day] [month] [year] [list]

Date:   Tue, 27 Sep 2022 13:03:22 +0100
From:   Valentin Schneider <vschneid@...hat.com>
To:     linux-kernel@...r.kernel.org
Cc:     Ingo Molnar <mingo@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
        Daniel Bristot de Oliveira <bristot@...hat.com>,
        Frederic Weisbecker <frederic@...nel.org>,
        "Paul E. McKenney" <paulmck@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Nicholas Piggin <npiggin@...il.com>,
        Marcelo Tosatti <mtosatti@...hat.com>
Subject: [PATCH] nohz_full: Fix TICK_DO_TIMER_NONE vs nohz_full warning

Booting a system with an invalid nohz_full cmdline mask (in my case
nr_cpus=X on the cmdline makes the nohz_full CPUs out of range) triggers:

[    1.209455] WARNING: CPU: 1 PID: 1 at kernel/time/tick-sched.c:191 tick_sched_do_timer+0x90/0xa0
[    1.209455] CPU: 1 PID: 1 Comm: swapper/0 Not tainted 5.19.0-00675-g7e9518baed4c #39
[    1.209455] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
[    1.209455] RIP: 0010:tick_sched_do_timer+0x90/0xa0
[    1.209455] Call Trace:
[    1.209455]  <IRQ>
[    1.209455]  tick_sched_timer+0x2e/0x80
[    1.209455]  __hrtimer_run_queues+0xfc/0x2a0
[    1.209455]  hrtimer_interrupt+0x105/0x240
[    1.209455]  __sysvec_apic_timer_interrupt+0x7a/0x160
[    1.209455]  sysvec_apic_timer_interrupt+0x85/0xb0
[    1.209455]  </IRQ>

This is because nothing checks the actual contents of the mask in
housekeeping_setup(), so in those scenarios we do end up invoking
tick_nohz_full_setup() and thus setting tick_nohz_full_running to true.

However, later on in tick_nohz_init(), this ends up being a no-op:

  for_each_cpu(cpu, tick_nohz_full_mask)
	  ct_cpu_track_user(cpu);

This in turn means we end up with
  tick_nohz_full_running == true
  context_tracking_enabled() == false
IOW
  tick_nohz_full_enabled() == false

Thus, __tick_nohz_idle_stop_tick() can legitimately stop the tick during
idle for the tick_do_timer_cpu, and sets tick_do_timer_cpu to
TICK_DO_TIMER_NONE. This triggers the warning when later on the tick fires
and tick_sched_do_timer() detects the tick_do_timer_cpu was relinquished.

Check the contents of the non_housekeeping_mask after it is
parsed (rcu_init_nohz() does a similar check). For good measure, also
update the check in tick_sched_do_timer() to use tick_nohz_full_enabled().

Fixes: 08ae95f4fd3b ("nohz_full: Allow the boot CPU to be nohz_full")
Signed-off-by: Valentin Schneider <vschneid@...hat.com>
---
 kernel/sched/isolation.c | 3 ++-
 kernel/time/tick-sched.c | 4 +---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/isolation.c b/kernel/sched/isolation.c
index 373d42c707bc..774cd187a1f7 100644
--- a/kernel/sched/isolation.c
+++ b/kernel/sched/isolation.c
@@ -120,7 +120,8 @@ static int __init housekeeping_setup(char *str, unsigned long flags)
 	}
 
 	alloc_bootmem_cpumask_var(&non_housekeeping_mask);
-	if (cpulist_parse(str, non_housekeeping_mask) < 0) {
+	if (cpulist_parse(str, non_housekeeping_mask) < 0 ||
+	    !cpumask_subset(non_housekeeping_mask, cpu_possible_mask)) {
 		pr_warn("Housekeeping: nohz_full= or isolcpus= incorrect CPU range\n");
 		goto free_non_housekeeping_mask;
 	}
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index b0e3c9205946..dae01a6577ab 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -187,9 +187,7 @@ static void tick_sched_do_timer(struct tick_sched *ts, ktime_t now)
 	 * tick_do_timer_cpu never relinquishes.
 	 */
 	if (unlikely(tick_do_timer_cpu == TICK_DO_TIMER_NONE)) {
-#ifdef CONFIG_NO_HZ_FULL
-		WARN_ON_ONCE(tick_nohz_full_running);
-#endif
+		WARN_ON_ONCE(tick_nohz_full_enabled());
 		tick_do_timer_cpu = cpu;
 	}
 #endif
-- 
2.31.1