On 08.08.2018 04:08, Stephen Rothwell wrote:
> Hi Kirill,
> 
> On Tue, 07 Aug 2018 18:37:36 +0300 Kirill Tkhai <ktkhai@virtuozzo.com> wrote:
>>
>> This patch kills all CONFIG_SRCU defines and
>> the code under !CONFIG_SRCU.
>>
>> Signed-off-by: Kirill Tkhai <ktkhai@virtuozzo.com>
> 
> So what sort of overheads (in terms of code size and performance) are
> we adding by having SRCU enabled where it used not to be?

SRCU is unconditionally enabled for x86, so I had to use another arch (sparc64)
to check the size difference. The config, I used to compile, is attached, SRCU
was enabled via:

diff --git a/arch/sparc/Kconfig b/arch/sparc/Kconfig
index 2d58c26bff9a..6e9116e356d4 100644
--- a/arch/sparc/Kconfig
+++ b/arch/sparc/Kconfig
@@ -15,6 +15,7 @@ config SPARC
 	select ARCH_MIGHT_HAVE_PC_PARPORT if SPARC64 && PCI
 	select ARCH_MIGHT_HAVE_PC_SERIO
 	select OF
+	select SRCU
 	select OF_PROMTREE
 	select HAVE_IDE
 	select HAVE_OPROFILE

$ size image.srcu.disabled 
   text	   data	    bss	    dec	    hex	filename
5117546	8030506	1968104	15116156	 e6a77c	image.srcu.disabled

$ size image.srcu.enabled
   text	   data	    bss	    dec	    hex	filename
5126175	8064346	1968104	15158625	 e74d61	image.srcu.enabled

The difference is: 15158625-15116156 = 42469 ~41Kb

I have not ideas about performance overhead measurements. If you have ideas,
where they may occur, please say. At the first sight, there should not be
a problem, since SRCU is enabled in x86 by default.