linux-kernel - Re: [PATCH v12 11/11] x86/tsc: use tsc early

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:   Sat, 23 Jun 2018 17:38:45 -0400
From:   Pavel Tatashin <pasha.tatashin@...cle.com>
To:     Thomas Gleixner <tglx@...utronix.de>
Cc:     steven.sistare@...cle.com, daniel.m.jordan@...cle.com,
        linux@...linux.org.uk, schwidefsky@...ibm.com,
        heiko.carstens@...ibm.com, john.stultz@...aro.org,
        sboyd@...eaurora.org, x86@...nel.org, linux-kernel@...r.kernel.org,
        mingo@...hat.com, hpa@...or.com, douly.fnst@...fujitsu.com,
        peterz@...radead.org, prarit@...hat.com, feng.tang@...el.com,
        pmladek@...e.com, gnomes@...rguk.ukuu.org.uk,
        linux-s390@...r.kernel.org
Subject: Re: [PATCH v12 11/11] x86/tsc: use tsc early

On 06/23/2018 12:56 PM, Thomas Gleixner wrote:
> On Thu, 21 Jun 2018, Pavel Tatashin wrote:
>>  /*
>>   * Scheduler clock - returns current time in nanosec units.
>>   */
>> @@ -1354,6 +1364,7 @@ void __init tsc_early_delay_calibrate(void)
>>  	lpj = tsc_khz * 1000;
>>  	do_div(lpj, HZ);
>>  	loops_per_jiffy = lpj;
>> +	sched_clock_early_init(tsc_khz);
>>  }
>>  
>>  void __init tsc_init(void)
>> @@ -1382,6 +1393,7 @@ void __init tsc_init(void)
>>  	if (!tsc_khz) {
>>  		mark_tsc_unstable("could not calculate TSC khz");
>>  		setup_clear_cpu_cap(X86_FEATURE_TSC_DEADLINE_TIMER);
>> +		static_branch_disable(&__use_tsc);
> 
> This wants a proper comment.

Yes, I will add a comment. Basically, this change is for a rare scenario where early in boot we were able to get TSC frequency but failed to do so later in boot when we tried to calculate it more precisely. Though, I am not sure if this is actually possible. However, if this is indeed possible, we will get into the same continuity problem as the one that is solved in "x86/tsc: prepare for early sched_clock". So, I guess we would have to modify either "jiffies_64" or  __sched_clock_offset if we want to take care of this corner case.

Thank you,
Pavel