linux-kernel - Re: Sv: [PATCH] hrtimer: Interrupt storm on clock

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]

Message-ID: <alpine.DEB.2.21.2102231638540.1861@somnus>
Date:   Tue, 23 Feb 2021 16:53:06 +0100 (CET)
From:   Anna-Maria Behnsen <anna-maria@...utronix.de>
To:     "Beckius, Mikael" <mikael.beckius@...driver.com>
cc:     Thomas Gleixner <tglx@...utronix.de>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: Sv: [PATCH] hrtimer: Interrupt storm on clock_settime

Hi Micke,

On Fri, 12 Feb 2021, Beckius, Mikael wrote:

> Thanks for the update and sorry for the late reply. After long-term
> testing of the patch, storm detection improved, it turns out that a
> similar problem can occur if hrtimer_interrupt runs during
> clock_settime. In this case it seems the offset can get updated and later
> read using hrtimer_update_base in hrtimer_interrupt before
> softirq_expires_next gets updated. As soon as softirq_expires_next gets
> updated by hrtimer_force_reprogram the storm ends. To fix this I made the
> below changes.
> 
> --- a/kernel/time/hrtimer.c
> +++ b/kernel/time/hrtimer.c
> @@ -529,8 +529,10 @@ static ktime_t __hrtimer_next_event_base(struct hrtimer_cpu_base *cpu_base,
>  			if (exclude)
>  				continue;
>  
> -			if (timer->is_soft)
> +			if (timer->is_soft) {
>  				cpu_base->softirq_next_timer = timer;
> +				cpu_base->softirq_expires_next = expires;
> +			}
>  			else
>  				cpu_base->next_timer = timer;
>  		}
> @@ -633,19 +635,6 @@ hrtimer_force_reprogram(struct hrtimer_cpu_base *cpu_base, int skip_equal)
>  	 */
>  	expires_next = __hrtimer_get_next_event(cpu_base, HRTIMER_ACTIVE_ALL);
>  
> -	if (cpu_base->next_timer && cpu_base->next_timer->is_soft) {
> -		/*
> -		 * When the softirq is activated, hrtimer has to be
> -		 * programmed with the first hard hrtimer because soft
> -		 * timer interrupt could occur too late.
> -		 */
> -		if (cpu_base->softirq_activated)
> -			expires_next = __hrtimer_get_next_event(cpu_base,
> -								HRTIMER_ACTIVE_HARD);
> -		else
> -			cpu_base->softirq_expires_next = expires_next;
> -	}
> -
>  	if (skip_equal && expires_next == cpu_base->expires_next)
>  		return;
>  
> --
> 

The proposed change breaks the reprogramming logic. To keep it working
cpu_base::*expires_next is only updated by hrtimer_reprogram() and
hrtimer_interrupt(). I will send you a patch for testing in reply to this
thread. The patch is compile tested only.

Thanks,

	Anna-Maria