Re: [PATCH] RISC-V: Show IPI stats

From: Anup Patel <anup@brainfault.org>
To: Christoph Hellwig <hch@infradead.org>
Cc: Palmer Dabbelt <palmer@sifive.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Atish Patra <atish.patra@wdc.com>,
	linux-riscv@lists.infradead.org,
	"linux-kernel@vger.kernel.org List"
	<linux-kernel@vger.kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH] RISC-V: Show IPI stats
Date: Wed, 12 Sep 2018 10:03:56 +0530	[thread overview]
Message-ID: <CAAhSdy3bGHdCX82r5FDqwRXUZZv-9DZDK9fNGVrxntdOGV36zA@mail.gmail.com> (raw)
In-Reply-To: <20180910134659.GA30774@infradead.org>

On Mon, Sep 10, 2018 at 7:16 PM, Christoph Hellwig <hch@infradead.org> wrote:
> On Fri, Sep 07, 2018 at 06:14:29PM +0530, Anup Patel wrote:
>> This patch provides arch_show_interrupts() implementation to
>> show IPI stats via /proc/interrupts.
>>
>> Now the contents of /proc/interrupts" will look like below:
>>            CPU0       CPU1       CPU2       CPU3
>>   8:         17          7          6         14  SiFive PLIC   8  virtio0
>>  10:         10         10          9         11  SiFive PLIC  10  ttyS0
>> IPI0:       170        673        251         79  Rescheduling interrupts
>> IPI1:         1         12         27          1  Function call interrupts
>> IPI2:         0          0          0          0  CPU wake-up interrupts
>>
>> Signed-off-by: Anup Patel <anup@brainfault.org>
>
> Thanks, this looks pretty sensible to me.  Maybe we want to also show
> timer interrupts if we do this?

Let's not include timer stats here until RISCV INTC driver is
concluded. We can do it as separate patch if required.

>
>> --- a/arch/riscv/kernel/irq.c
>> +++ b/arch/riscv/kernel/irq.c
>> @@ -8,6 +8,7 @@
>>  #include <linux/interrupt.h>
>>  #include <linux/irqchip.h>
>>  #include <linux/irqdomain.h>
>> +#include <linux/seq_file.h>
>>
>>  /*
>>   * Possible interrupt causes:
>> @@ -24,6 +25,14 @@
>>   */
>>  #define INTERRUPT_CAUSE_FLAG (1UL << (__riscv_xlen - 1))
>>
>> +int arch_show_interrupts(struct seq_file *p, int prec)
>> +{
>> +#ifdef CONFIG_SMP
>> +     show_ipi_stats(p, prec);
>> +#endif
>> +     return 0;
>> +}
>
> If we don't also add timer stats I'd just move arch_show_interrupts
> to smp.c and make it conditional.  If we don't this split might make
> more sense.

I understand you want to avoid #ifdef here. We can do same thing
by having empty inline function show_ipi_stats() in asm/smp.h for
!CONFIG_SMP case. This way we can keep arch_show_interrupts()
in kernel/irq.c which is intuitively correct location for
arch_show_interrupts().

>
>> +static const char *ipi_names[IPI_MAX] = {
>> +     [IPI_RESCHEDULE] = "Rescheduling interrupts",
>> +     [IPI_CALL_FUNC] = "Function call interrupts",
>> +     [IPI_CALL_WAKEUP] = "CPU wake-up interrupts",
>> +};
>
> No need for the explicit array size.  Also please use a few tabs to
> align this nicely:
>
> static const char *ipi_names[] = {
>         [IPI_RESCHEDULE]        = "Rescheduling interrupts",
>         [IPI_CALL_FUNC]         = "Function call interrupts",
>         [IPI_CALL_WAKEUP]       = "CPU wake-up interrupts",
> };

Sure, will do.

Regards,
Anup