Re: [PATCH] arm64: Demote boot and shutdown messages to pr_debug

From: Marc Zyngier <marc.zyngier@arm.com>
To: Mark Rutland <mark.rutland@arm.com>,
	Florian Fainelli <f.fainelli@gmail.com>
Cc: linux-arm-kernel@lists.infradead.org, rmk+kernel@armlinux.org.uk,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will.deacon@arm.com>,
	Ard Biesheuvel <ard.biesheuvel@linaro.org>,
	Julien Thierry <julien.thierry@arm.com>,
	Suzuki K Poulose <Suzuki.Poulose@arm.com>,
	Sudeep Holla <sudeep.holla@arm.com>,
	Rob Herring <robh@kernel.org>,
	Steve Capper <steve.capper@arm.com>,
	Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
	Daniel Thompson <daniel.thompson@linaro.org>,
	open list <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] arm64: Demote boot and shutdown messages to pr_debug
Date: Wed, 1 May 2019 12:01:54 +0100	[thread overview]
Message-ID: <ae251deb-ef34-29b8-edc1-66d987fa6e5c@arm.com> (raw)
In-Reply-To: <20190501104733.GB11740@lakrids.cambridge.arm.com>

On 01/05/2019 11:47, Mark Rutland wrote:
> On Tue, Apr 30, 2019 at 03:38:31PM -0700, Florian Fainelli wrote:
>> Similar to commits c68b0274fb3cf ("ARM: reduce "Booted secondary
>> processor" message to debug level") and 035e787543de7 ("ARM: 8644/1: Reduce "CPU:
>> shutdown" message to debug level"), demote the secondary_start_kernel()
>> and __cpu_die() messages from info, respectively notice to debug. While
>> we are at it, also do this for cpu_psci_cpu_kill() which is redundant
>> with __cpu_die().
>>
>> This helps improve the amount of possible hotplug cycles by around +50%
>> on ARCH_BRCMSTB.
> 
> Could you elaborate on why that matters? 
> 
> e.g. is this just for testing, or does this matter in some shutdown or
> hibernate scenario?
> 
>> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
>> ---
>>  arch/arm64/kernel/psci.c | 2 +-
>>  arch/arm64/kernel/smp.c  | 4 ++--
>>  2 files changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/kernel/psci.c b/arch/arm64/kernel/psci.c
>> index 8cdaf25e99cd..a78581046c80 100644
>> --- a/arch/arm64/kernel/psci.c
>> +++ b/arch/arm64/kernel/psci.c
>> @@ -96,7 +96,7 @@ static int cpu_psci_cpu_kill(unsigned int cpu)
>>  	for (i = 0; i < 10; i++) {
>>  		err = psci_ops.affinity_info(cpu_logical_map(cpu), 0);
>>  		if (err == PSCI_0_2_AFFINITY_LEVEL_OFF) {
>> -			pr_info("CPU%d killed.\n", cpu);
>> +			pr_debug("CPU%d killed.\n", cpu);
>>  			return 0;
>>  		}
>>  
>> diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c
>> index 824de7038967..71fd2b5a3f0e 100644
>> --- a/arch/arm64/kernel/smp.c
>> +++ b/arch/arm64/kernel/smp.c
>> @@ -259,7 +259,7 @@ asmlinkage notrace void secondary_start_kernel(void)
>>  	 * the CPU migration code to notice that the CPU is online
>>  	 * before we continue.
>>  	 */
>> -	pr_info("CPU%u: Booted secondary processor 0x%010lx [0x%08x]\n",
>> +	pr_debug("CPU%u: Booted secondary processor 0x%010lx [0x%08x]\n",
>>  					 cpu, (unsigned long)mpidr,
>>  					 read_cpuid_id());
> 
> I generally agree that we don't need to be verbose, and demoting these
> to debug is fine, but it's a shame that these won't be accessible in
> defconfig.
> 
> I wonder if we should enable DYNAMIC_DEBUG so that we can turn these on
> from the kernel command line, or if we should have something like a
> verbose_hotplug option specifically for these messages.

I really wonder what kind of workload is hotplug-happy enough that these
two messages become a bottleneck. Why isn't turning down the loglevel
enough?

Thanks,

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