2021-02-22 05:27:56

by Zhiyuan Dai

[permalink] [raw]
Subject: [PATCH] mm/kasan: remove volatile keyword

Like volatile, the kernel primitives which make concurrent
access to data safe (spinlocks, mutexes, memory barriers,
etc.) are designed to prevent unwanted optimization.

If they are being used properly, there will be no need to
use volatile as well. If volatile is still necessary,
there is almost certainly a bug in the code somewhere.
In properly-written kernel code, volatile can only serve
to slow things down.

see: Documentation/process/volatile-considered-harmful.rst

Signed-off-by: Zhiyuan Dai <[email protected]>
---
mm/kasan/shadow.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/kasan/shadow.c b/mm/kasan/shadow.c
index 7c2c08c..d5ff9ca 100644
--- a/mm/kasan/shadow.c
+++ b/mm/kasan/shadow.c
@@ -25,13 +25,13 @@

#include "kasan.h"

-bool __kasan_check_read(const volatile void *p, unsigned int size)
+bool __kasan_check_read(const void *p, unsigned int size)
{
return check_memory_region((unsigned long)p, size, false, _RET_IP_);
}
EXPORT_SYMBOL(__kasan_check_read);

-bool __kasan_check_write(const volatile void *p, unsigned int size)
+bool __kasan_check_write(const void *p, unsigned int size)
{
return check_memory_region((unsigned long)p, size, true, _RET_IP_);
}
--
1.8.3.1


2021-02-22 07:42:48

by Dmitry Vyukov

[permalink] [raw]
Subject: Re: [PATCH] mm/kasan: remove volatile keyword

On Mon, Feb 22, 2021 at 6:22 AM Zhiyuan Dai <[email protected]> wrote:
>
> Like volatile, the kernel primitives which make concurrent
> access to data safe (spinlocks, mutexes, memory barriers,
> etc.) are designed to prevent unwanted optimization.
>
> If they are being used properly, there will be no need to
> use volatile as well. If volatile is still necessary,
> there is almost certainly a bug in the code somewhere.
> In properly-written kernel code, volatile can only serve
> to slow things down.
>
> see: Documentation/process/volatile-considered-harmful.rst

Nack.

This function does not require volatile variables. It uses volatile in
the same way as C/C++ atomic functions -- it only supports operating
on volatile variables. The same meaning as for const here. Such
functions need to use all possible type modifiers to support all
possible uses.

Anyway, the function is declared in kasan.h. So you would need to
change the signate there in the first place. But the kernel will
either not compile, or it won't compile in future when somebody adds
__kasan_check_read/write for a volatile variable. Such a change first
requires removing all volatile uses from the entire kernel and banning
volatile.


> Signed-off-by: Zhiyuan Dai <[email protected]>
> ---
> mm/kasan/shadow.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/mm/kasan/shadow.c b/mm/kasan/shadow.c
> index 7c2c08c..d5ff9ca 100644
> --- a/mm/kasan/shadow.c
> +++ b/mm/kasan/shadow.c
> @@ -25,13 +25,13 @@
>
> #include "kasan.h"
>
> -bool __kasan_check_read(const volatile void *p, unsigned int size)
> +bool __kasan_check_read(const void *p, unsigned int size)
> {
> return check_memory_region((unsigned long)p, size, false, _RET_IP_);
> }
> EXPORT_SYMBOL(__kasan_check_read);
>
> -bool __kasan_check_write(const volatile void *p, unsigned int size)
> +bool __kasan_check_write(const void *p, unsigned int size)
> {
> return check_memory_region((unsigned long)p, size, true, _RET_IP_);
> }
> --
> 1.8.3.1
>