linux-kernel - Re: [patch 18/20] posix-timers: Clarify posix_timer

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:   Thu, 01 Jun 2023 20:43:19 +0200
From:   Thomas Gleixner <tglx@...utronix.de>
To:     Frederic Weisbecker <frederic@...nel.org>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        Anna-Maria Behnsen <anna-maria@...utronix.de>,
        Peter Zijlstra <peterz@...radead.org>,
        Sebastian Siewior <bigeasy@...utronix.de>,
        syzbot+5c54bd3eb218bb595aa9@...kaller.appspotmail.com,
        Dmitry Vyukov <dvyukov@...gle.com>,
        Michael Kerrisk <mtk.manpages@...il.com>
Subject: Re: [patch 18/20] posix-timers: Clarify posix_timer_fn() comments

On Thu, Jun 01 2023 at 15:21, Frederic Weisbecker wrote:
> On Tue, Apr 25, 2023 at 08:49:24PM +0200, Thomas Gleixner wrote:
>> Make the issues vs. SIG_IGN understandable and remove the 15 years old
>> promise that a proper solution is already on the horizon.
>> 
>> Signed-off-by: Thomas Gleixner <tglx@...utronix.de>
>> ---
>>  kernel/time/posix-timers.c |   56 +++++++++++++++++++++------------------------
>>  1 file changed, 27 insertions(+), 29 deletions(-)
>> 
>> --- a/kernel/time/posix-timers.c
>> +++ b/kernel/time/posix-timers.c
>> @@ -325,11 +325,11 @@ int posix_timer_event(struct k_itimer *t
>>  }
>>  
>>  /*
>> - * This function gets called when a POSIX.1b interval timer expires.  It
>> - * is used as a callback from the kernel internal timer.  The
>> - * run_timer_list code ALWAYS calls with interrupts on.
>> -
>> - * This code is for CLOCK_REALTIME* and CLOCK_MONOTONIC* timers.
>> + * This function gets called when a POSIX.1b interval timer expires from
>> + * the HRTIMER soft interrupt with interrupts enabled.
>
> BTW, what arranges for this to be called in softirq with interrupts enabled?
> The modes I see used here are HRTIMER_MODE_ABS or HRTIMER_MODE_REL and not
> their _SOFT counterparts.

Duh. My RT biased brain tricked me.

>> + *
>> + * Handles CLOCK_REALTIME, CLOCK_MONOTONIC, CLOCK_BOOTTIME and CLOCK_TAI
>> + * based timers.
>>   */
>>  static enum hrtimer_restart posix_timer_fn(struct hrtimer *timer)
>>  {
>> @@ -358,34 +359,31 @@ static enum hrtimer_restart posix_timer_
>>  			 * FIXME: What we really want, is to stop this
>>  			 * timer completely and restart it in case the
>>  			 * SIG_IGN is removed. This is a non trivial
>> -			 * change which involves sighand locking
>> -			 * (sigh !), which we don't want to do late in
>> -			 * the release cycle.
>> +			 * change to the signal handling code.
>> +			 *
>> +			 * For now let timers with an interval less than a
>> +			 * jiffie expire every jiffie to avoid softirq
>
> Or rather at least to the next jiffie, right? Because then in the next jiffie
> it gets re-evaluated in case a real signal handler might have been set
> in-between.
>
> Or it could be:
>
>  +                      * For now let timers with an interval less than a
>  +                      * jiffie expire every jiffie (until a real sig handler
>  +			* is found set) to avoid softirq...

Let me rephrase that.

>> +			 * starvation in case of SIG_IGN and a very small
>> +			 * interval, which would put the timer right back
>> +			 * on the softirq pending list. Moving now ahead of
>> +			 * time tricks hrtimer_forward() to expire the
>> +			 * timer later, while it still maintains the
>> +			 * overrun accuracy for the price of a slightly
>> +			 * inconsistency in the timer_gettime() case. This
>> +			 * is at least better than a starved softirq.
> [...]
>>  			 */
>> -#ifdef CONFIG_HIGH_RES_TIMERS
>> -			{
>> +			if (IS_ENABLED(CONFIG_HIGHRES_TIMERS)) {
>>  				ktime_t kj = NSEC_PER_SEC / HZ;
>
> Could be TICK_NSECS?

Yep. Fixed it up.