Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1604644 > unrolled thread

Re: [PATCH 06/17] arm64: arch_timer: Get rid of erratum_workaround_set_sne

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH 06/17] arm64: arch_timer: Get rid of  erratum_workaround_set_sne Mark Rutland <mark.rutland@arm.com> - 2017-03-20 15:40 +0100
    Re: [PATCH 06/17] arm64: arch_timer: Get rid of  erratum_workaround_set_sne Marc Zyngier <marc.zyngier@arm.com> - 2017-03-20 18:10 +0100

#1604644 — Re: [PATCH 06/17] arm64: arch_timer: Get rid of erratum_workaround_set_sne

On Mon, Mar 06, 2017 at 11:26:11AM +0000, Marc Zyngier wrote:
> Let's move the handling of workarounds affecting set_next_event
> to the affected function, instead of overwriding the pointers

Nit: overriding

> as an afterthough. Yes, this is an extra indirection on the
> erratum handling path, but the HW is busted anyway.

[...]

> +#define erratum_set_next_event_tval_virt(...)		({BUG_ON(1); 0;})
> +#define erratum_set_next_event_tval_phys(...)		({BUG_ON(1); 0;})

Nit: BUG()

[...]

>  {
> +	if (needs_unstable_timer_counter_workaround())
> +		return erratum_set_next_event_tval_virt(evt, clk);
> +
>  	set_next_event(ARCH_TIMER_VIRT_ACCESS, evt, clk);
>  	return 0;
>  }
> @@ -487,6 +490,9 @@ static int arch_timer_set_next_event_virt(unsigned long evt,
>  static int arch_timer_set_next_event_phys(unsigned long evt,
>  					  struct clock_event_device *clk)
>  {
> +	if (needs_unstable_timer_counter_workaround())
> +		return erratum_set_next_event_tval_phys(evt, clk);
> +
>  	set_next_event(ARCH_TIMER_PHYS_ACCESS, evt, clk);
>  	return 0;
>  }

I take it that the new conditionals are NOPs in the !erratum case, so
this doesn't adversely affect the usual case.

Thanks,
Mark.

[toc] | [next] | [standalone]

#1604807

On 20/03/17 14:06, Mark Rutland wrote:
> On Mon, Mar 06, 2017 at 11:26:11AM +0000, Marc Zyngier wrote:
>> Let's move the handling of workarounds affecting set_next_event
>> to the affected function, instead of overwriding the pointers
> 
> Nit: overriding
> 
>> as an afterthough. Yes, this is an extra indirection on the
>> erratum handling path, but the HW is busted anyway.
> 
> [...]
> 
>> +#define erratum_set_next_event_tval_virt(...)		({BUG_ON(1); 0;})
>> +#define erratum_set_next_event_tval_phys(...)		({BUG_ON(1); 0;})
> 
> Nit: BUG()
> 
> [...]
> 
>>  {
>> +	if (needs_unstable_timer_counter_workaround())
>> +		return erratum_set_next_event_tval_virt(evt, clk);
>> +
>>  	set_next_event(ARCH_TIMER_VIRT_ACCESS, evt, clk);
>>  	return 0;
>>  }
>> @@ -487,6 +490,9 @@ static int arch_timer_set_next_event_virt(unsigned long evt,
>>  static int arch_timer_set_next_event_phys(unsigned long evt,
>>  					  struct clock_event_device *clk)
>>  {
>> +	if (needs_unstable_timer_counter_workaround())
>> +		return erratum_set_next_event_tval_phys(evt, clk);
>> +
>>  	set_next_event(ARCH_TIMER_PHYS_ACCESS, evt, clk);
>>  	return 0;
>>  }
> 
> I take it that the new conditionals are NOPs in the !erratum case, so
> this doesn't adversely affect the usual case.

Indeed, we only get an extra NOP on the normal path. The erratum path
triggers the insertion of a jump targeting the end of the function,
where the workaround gets handled.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web