Re: [PATCH v4 03/11] coresight: tmc-etf: Add comment for store ordering

From: Peter Zijlstra <peterz@infradead.org>
To: Leo Yan <leo.yan@linaro.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Ingo Molnar <mingo@redhat.com>,
	Mark Rutland <mark.rutland@arm.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Jiri Olsa <jolsa@redhat.com>, Namhyung Kim <namhyung@kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Borislav Petkov <bp@alien8.de>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
	Mathieu Poirier <mathieu.poirier@linaro.org>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Mike Leach <mike.leach@linaro.org>,
	linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org,
	coresight@lists.linaro.org, linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH v4 03/11] coresight: tmc-etf: Add comment for store ordering
Date: Tue, 13 Jul 2021 14:56:06 +0200	[thread overview]
Message-ID: <20210713125606.GB4170@worktop.programming.kicks-ass.net> (raw)
In-Reply-To: <20210711104105.505728-4-leo.yan@linaro.org>

On Sun, Jul 11, 2021 at 06:40:57PM +0800, Leo Yan wrote:
> AUX ring buffer is required to separate the data store and aux_head
> store, since the function CS_LOCK() has contained memory barrier mb(),
> mb() is a more conservative barrier than smp_wmb() on Arm32/Arm64, thus
> it's needless to add any explicit barrier anymore.
> 
> Add comment to make clear for the barrier usage for ETF.
> 
> Signed-off-by: Leo Yan <leo.yan@linaro.org>
> ---
>  drivers/hwtracing/coresight/coresight-tmc-etf.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etf.c b/drivers/hwtracing/coresight/coresight-tmc-etf.c
> index 45b85edfc690..9a42ee689921 100644
> --- a/drivers/hwtracing/coresight/coresight-tmc-etf.c
> +++ b/drivers/hwtracing/coresight/coresight-tmc-etf.c
> @@ -553,6 +553,12 @@ static unsigned long tmc_update_etf_buffer(struct coresight_device *csdev,
>  	if (buf->snapshot)
>  		handle->head += to_read;
>  
> +	/*
> +	 * AUX ring buffer requires to use memory barrier to separate the trace
> +	 * data store and aux_head store, because CS_LOCK() contains mb() which
> +	 * gives more heavy barrier than smp_wmb(), it's not necessary to
> +	 * explicitly invoke any barrier.
> +	 */
>  	CS_LOCK(drvdata->base);

'more heavy' is not a correctness argument :-)

The argument to make here is that CS_LOCK() ensures completion /
visibility of the hardware buffer.