Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [RFC][PATCH 1/4] sched: Force the address order of each sched
 class descriptor
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     Steven Rostedt <rostedt@goodmis.org>, linux-kernel@vger.kernel.org,
        Kirill Tkhai <tkhai@yandex.ru>,
        Kirill Tkhai <ktkhai@virtuozzo.com>,
        "mingo@redhat.com" <mingo@redhat.com>,
        "juri.lelli@redhat.com" <juri.lelli@redhat.com>,
        "vincent.guittot@linaro.org" <vincent.guittot@linaro.org>,
        "dietmar.eggemann@arm.com" <dietmar.eggemann@arm.com>,
        "bsegall@google.com" <bsegall@google.com>,
        "mgorman@suse.de" <mgorman@suse.de>,
        Ingo Molnar <mingo@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>
References: <20191219214451.340746474@goodmis.org>
 <20191219214558.510271353@goodmis.org>
 <0a957e8d-7af8-613c-11ae-f51b9b241eb7@rasmusvillemoes.dk>
 <20191220100033.GE2844@hirez.programming.kicks-ass.net>
 <1ac359e8-fa7f-7ded-e2f2-9f7d0b23a4e1@rasmusvillemoes.dk>
 <20191220121947.GH2844@hirez.programming.kicks-ass.net>
From:   Rasmus Villemoes <linux@rasmusvillemoes.dk>
Message-ID: <f580513e-b6a1-db1f-4df8-924adb438179@rasmusvillemoes.dk>
Date:   Fri, 20 Dec 2019 15:34:49 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.2.1
MIME-Version: 1.0
In-Reply-To: <20191220121947.GH2844@hirez.programming.kicks-ass.net>
Content-Type: text/plain; charset=windows-1252
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 20/12/2019 13.19, Peter Zijlstra wrote:
> On Fri, Dec 20, 2019 at 11:12:37AM +0100, Rasmus Villemoes wrote:
>> On 20/12/2019 11.00, Peter Zijlstra wrote:
> 
>>>>> +/*
>>>>> + * The order of the sched class addresses are important, as they are
>>>>> + * used to determine the order of the priority of each sched class in
>>>>> + * relation to each other.
>>>>> + */
>>>>> +#define SCHED_DATA				\
>>>>> +	*(__idle_sched_class)			\
>>>>> +	*(__fair_sched_class)			\
>>>>> +	*(__rt_sched_class)			\
>>>>> +	*(__dl_sched_class)			\
>>>>> +	STOP_SCHED_CLASS
>>>
>>> I'm confused, why does that STOP_SCHED_CLASS need magic here at all?
>>> Doesn't the linker deal with empty sections already by making them 0
>>> sized?
>>
>> Yes, but dropping the STOP_SCHED_CLASS define doesn't prevent one from
>> needing some ifdeffery to define highest_sched_class if they are laid
>> out in (higher sched class <-> higher address) order.
> 
> Would not something like:
> 
> 	__begin_sched_classes = .;
> 	*(__idle_sched_class)
> 	*(__fair_sched_class)
> 	*(__rt_sched_class)
> 	*(__dl_sched_class)
> 	*(__stop_sched_class)
> 	__end_sched_classes = .;
> 
> combined with something like:
> 
> extern struct sched_class *__begin_sched_classes;
> extern struct sched_class *__end_sched_classes;

extern const struct sched_class __begin_sched_classes[];

but yes, I get the idea.

> #define sched_class_highest (__end_sched_classes - 1)
> #define sched_class_lowest  (__begin_sched_classes - 1)
> 
> #define for_class_range(class, _from, _to) \
> 	for (class = (_from); class != (_to), class--)
> 
> #define for_each_class(class) \
> 	for_class_range(class, sched_class_highest, sched_class_lowest)
> 
> just work?

Yes, I think so - I was only thinking of the case where all the symbols
would be defined in the linker script, and for this to work you need the
C compiler to subtract the sizeof().

I'd probably not include the -1 in the definition of sched_class_lowest,
but instead put it in the for_each_class definition (i.e. use
sched_class_lowest-1 as _to parameter).

A whole other option is of course to make the whole thing a bona fide C
array defined in sched/core.c, with fair_sched_class being defined as
&sched_classes[1] etc. But that requires giving all the methods extern
linkage. The advantage might be that the compiler can see how much we
iterate over, though I wouldn't expect it to actually unroll the
for_each_class loops five times. So yes, the above is probably the best
way to go.

Rasmus