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: RESEND [PATCH v6 11/12] lib: vdso: Add support for CLOCK_BOOTTIME
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     linux-kernel@vger.kernel.org, James Morse <james.morse@arm.com>,
        Russell King <linux@armlinux.org.uk>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will.deacon@arm.com>,
        Andy Lutomirski <luto@amacapital.net>,
        Dmitry Safonov <dsafonov@virtuozzo.com>,
        John Stultz <john.stultz@linaro.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Laura Abbott <labbott@redhat.com>,
        Kees Cook <keescook@chromium.org>,
        Ard Biesheuvel <ard.biesheuvel@linaro.org>,
        Andy Gross <andy.gross@linaro.org>,
        Kevin Brodsky <kevin.brodsky@arm.com>,
        Andrew Pinski <apinski@cavium.com>,
        linux-arm-kernel@lists.infradead.org,
        Jeremy Linton <Jeremy.Linton@arm.com>,
        android-kernel@android.com, Jinbum Park <jinb.park7@gmail.com>,
        Ingo Molnar <mingo@kernel.org>,
        "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
References: <20181001175845.168430-1-salyzyn@android.com>
 <20181001175845.168430-12-salyzyn@android.com>
 <alpine.DEB.2.21.1810012009010.32062@nanos.tec.linutronix.de>
From:   Mark Salyzyn <salyzyn@android.com>
Message-ID: <a602b661-dde5-399d-6245-cf48a0e43f54@android.com>
Date:   Mon, 1 Oct 2018 13:52:32 -0700
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.9.1
MIME-Version: 1.0
In-Reply-To: <alpine.DEB.2.21.1810012009010.32062@nanos.tec.linutronix.de>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Content-Language: en-GB
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 10/01/2018 11:15 AM, Thomas Gleixner wrote:
> On Mon, 1 Oct 2018, Mark Salyzyn wrote:
>>   
>> +static notrace int do_boottime(const struct vdso_data *vd, struct timespec *ts)
>> +{
>> +	u32 seq, mult, shift;
>> +	u64 nsec, cycle_last;
>> +	vdso_wtm_clock_nsec_t wtm_nsec;
>> +
>> +	/* open coding timespec_add_ns to save a ts->tv_nsec = 0 */
>> +	ts->tv_sec = sec + __iter_div_u64_rem(nsec, NSEC_PER_SEC, &nsec);
>> +	ts->tv_nsec = nsec;
>> +
>> +	return 0;
>> +}
>> +
> Instead of adding yet another copy of the same code you might want to look
> at the rework I did for the x86 vdso in order to support CLOCK_TAI w/o
> running into the issue of the clock switch case being compiled into a jump
> table and then the compile asking for retpoline.
>
>    http://lkml.kernel.org/r/20180917124533.329334911@linutronix.de
>
> Thanks,
>
> 	tglx

Great idea. Thanks!

The point of the first 12 patches is to _align_ the arm code to match 
the assembler for arm64 exactly 1:1. Then switch arm64 assembler to use 
the _same_ code in 'C' as a library. No performance degradation.

Next extend the vdso framework on arm64 to _also_ use that library for 
vDSO32 (arm32 compat on 64-bit). At this point we achieve a 0.4% power 
reduction on Android.

At that point, we would be ready for a rework that fixes all three (ARM 
vDSO, ARM64 vDSO and ARM64 vDSO32) to get the compiler to handle the 
switch statement better. One step at a time.

-- Mark