linux-kernel - Re: [PATCH v8 3/4] intel_idle: Provide the default enter

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

Message-ID: <6bb2ba2e-bb51-4637-b254-2603c38a5537@linux.intel.com>
Date: Tue, 17 Dec 2024 13:42:35 +0100
From: Patryk Wlazlyn <patryk.wlazlyn@...ux.intel.com>
To: "Rafael J. Wysocki" <rafael@...nel.org>
Cc: x86@...nel.org, linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
 rafael.j.wysocki@...el.com, peterz@...radead.org,
 dave.hansen@...ux.intel.com, gautham.shenoy@....com, tglx@...utronix.de,
 len.brown@...el.com, artem.bityutskiy@...ux.intel.com
Subject: Re: [PATCH v8 3/4] intel_idle: Provide the default enter_dead()
 handler

>> Recent Intel platforms require idle driver to provide information about
>> the MWAIT hint used to enter the deepest idle state in the play_dead
>> code.
>>
>> Provide the default enter_dead() handler for all of the platforms and
>> allow overwriting with a custom handler for each platform if needed.
>
> My changelog for this patch:
>
> "A subsequent change is going to make native_play_dead() rely on the
> idle driver to put CPUs going offline into appropriate idle states.
>
> For this reason, provide the default :enter_dead() handler for all of
> the idle states on all platforms supported by intel_idle with an
> option to override it with a custom handler if needed."
>

Ok. I'll apply that in the next version.

>> Signed-off-by: Patryk Wlazlyn <patryk.wlazlyn@...ux.intel.com>
>> ---
>>  drivers/idle/intel_idle.c | 18 ++++++++++++++++--
>>  1 file changed, 16 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c
>> index ac4d8faa3886..c6874a6dbe95 100644
>> --- a/drivers/idle/intel_idle.c
>> +++ b/drivers/idle/intel_idle.c
>> @@ -56,6 +56,7 @@
>>  #include <asm/mwait.h>
>>  #include <asm/spec-ctrl.h>
>>  #include <asm/fpu/api.h>
>> +#include <asm/smp.h>
>>
>>  #define INTEL_IDLE_VERSION "0.5.1"
>>
>> @@ -227,6 +228,16 @@ static __cpuidle int intel_idle_s2idle(struct cpuidle_device *dev,
>>         return 0;
>>  }
>>
>> +static __cpuidle void intel_idle_enter_dead(struct cpuidle_device *dev,
>> +                                           int index)
>> +{
>> +       struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
>> +       struct cpuidle_state *state = &drv->states[index];
>> +       unsigned long eax = flg2MWAIT(state->flags);
>> +
>> +       mwait_play_dead(eax);
>> +}
>> +
>>  /*
>>   * States are indexed by the cstate number,
>>   * which is also the index into the MWAIT hint array.
>> @@ -1798,6 +1809,7 @@ static void __init intel_idle_init_cstates_acpi(struct cpuidle_driver *drv)
>>                         state->flags |= CPUIDLE_FLAG_TIMER_STOP;
>>
>>                 state->enter = intel_idle;
>> +               state->enter_dead = intel_idle_enter_dead;
>>                 state->enter_s2idle = intel_idle_s2idle;
>>         }
>>  }
>> @@ -2143,10 +2155,12 @@ static void __init intel_idle_init_cstates_icpu(struct cpuidle_driver *drv)
>>                 if (intel_idle_max_cstate_reached(cstate))
>>                         break;
>>
>> -               if (!cpuidle_state_table[cstate].enter &&
>> -                   !cpuidle_state_table[cstate].enter_s2idle)
>> +               if (!cpuidle_state_table[cstate].enter)
>
> I don't think that the above change belongs to this patch.  If I'm
> mistaken, it should be mentioned in the changelog and the reason for
> making it should be explained.

Yeah, you are right, removing enter_s2idle check doesn't make much sense.
I think I was changing much more code, but eventually decided not to
change too much in one go and forgot to roll back that one.

With this:
                if (!cpuidle_state_table[cstate].enter &&
                    !cpuidle_state_table[cstate].enter_s2idle)

we would not set enter_dead for states that only provide enter_dead handler, but I doesn't seem to happen in practice, so for the sake of simplicity I am just going to leave the check unchanged.
Unless you think it makes more sense to do:
                if (!cpuidle_state_table[cstate].enter &&
+                   !cpuidle_state_table[cstate].enter_dead &&
                    !cpuidle_state_table[cstate].enter_s2idle)

>
>>                         break;
>>
>> +               if (!cpuidle_state_table[cstate].enter_dead)
>> +                       cpuidle_state_table[cstate].enter_dead = intel_idle_enter_dead;
>> +
>>                 /* If marked as unusable, skip this state. */
>>                 if (cpuidle_state_table[cstate].flags & CPUIDLE_FLAG_UNUSABLE) {
>>                         pr_debug("state %s is disabled\n",
>> --