[PATCH 1/3] arch/x86: Use offsetof

[PATCH 1/3] arch/x86: Use offsetof

[Julia Lawall]

From: Julia Lawall <julia@diku.dk>

In the patch cc154ac64aa8d3396b187f64cef01ce67f433324, Al Viro observed
that the proper way to compute the distance between two structure fields is
to use offsetof() or a cast to a pointer to character.  The same change can
be applied to a few more files.

The change was made using the following semantic patch
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@r@
type T;
T s;
type T1, T2;
identifier fld1, fld2;
typedef uint8_t;
typedef u8;
@@

(
  (char *)&s.fld1 - (char *)&s.fld2
|
  (uint8_t *)&s.fld1 - (uint8_t *)&s.fld2
|
  (u8 *)&s.fld1 - (u8 *)&s.fld2
|
- (T1)&s.fld1 - (T2)&s.fld2
+ offsetof(T,fld1) - offsetof(T,fld2)
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
---

diff -u -p a/arch/x86/kernel/vm86_32.c b/arch/x86/kernel/vm86_32.c
--- a/arch/x86/kernel/vm86_32.c	2007-10-22 11:25:00.000000000 +0200
+++ b/arch/x86/kernel/vm86_32.c	2007-12-26 16:27:15.000000000 +0100
@@ -215,7 +215,9 @@ asmlinkage int sys_vm86old(struct pt_reg
 	ret = -EFAULT;
 	if (tmp)
 		goto out;
-	memset(&info.vm86plus, 0, (int)&info.regs32 - (int)&info.vm86plus);
+	memset(&info.vm86plus, 0,
+	       offsetof(struct kernel_vm86_struct, regs32) -
+	       offsetof(struct kernel_vm86_struct, vm86plus));
 	info.regs32 = &regs;
 	tsk->thread.vm86_info = v86;
 	do_sys_vm86(&info, tsk);

Re: [PATCH 1/3] arch/x86: Use offsetof

[Pavel Machek]

Hi!

> From: Julia Lawall <julia@diku.dk>
> 
> In the patch cc154ac64aa8d3396b187f64cef01ce67f433324, Al Viro observed
> that the proper way to compute the distance between two structure fields is
> to use offsetof() or a cast to a pointer to character.  The same change can
> be applied to a few more files.
> 
> The change was made using the following semantic patch
> (http://www.emn.fr/x-info/coccinelle/)
> 
> // <smpl>
> @r@
> type T;
> T s;
> type T1, T2;
> identifier fld1, fld2;
> typedef uint8_t;
> typedef u8;
> @@
> 
> (
>   (char *)&s.fld1 - (char *)&s.fld2
> |
>   (uint8_t *)&s.fld1 - (uint8_t *)&s.fld2
> |
>   (u8 *)&s.fld1 - (u8 *)&s.fld2
> |
> - (T1)&s.fld1 - (T2)&s.fld2
> + offsetof(T,fld1) - offsetof(T,fld2)
> )
> // </smpl>
> 
> Signed-off-by: Julia Lawall <julia@diku.dk>
> ---
> 
> diff -u -p a/arch/x86/kernel/vm86_32.c b/arch/x86/kernel/vm86_32.c
> --- a/arch/x86/kernel/vm86_32.c	2007-10-22 11:25:00.000000000 +0200
> +++ b/arch/x86/kernel/vm86_32.c	2007-12-26 16:27:15.000000000 +0100
> @@ -215,7 +215,9 @@ asmlinkage int sys_vm86old(struct pt_reg
>  	ret = -EFAULT;
>  	if (tmp)
>  		goto out;
> -	memset(&info.vm86plus, 0, (int)&info.regs32 - (int)&info.vm86plus);
> +	memset(&info.vm86plus, 0,
> +	       offsetof(struct kernel_vm86_struct, regs32) -
> +	       offsetof(struct kernel_vm86_struct, vm86plus));

I do not think this makes it more readable... (int) -> (char *) would
make it portable and readable, AFAICT.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH 1/3] arch/x86: Use offsetof

[H. Peter Anvin]

Pavel Machek wrote:
>>
>> diff -u -p a/arch/x86/kernel/vm86_32.c b/arch/x86/kernel/vm86_32.c
>> --- a/arch/x86/kernel/vm86_32.c	2007-10-22 11:25:00.000000000 +0200
>> +++ b/arch/x86/kernel/vm86_32.c	2007-12-26 16:27:15.000000000 +0100
>> @@ -215,7 +215,9 @@ asmlinkage int sys_vm86old(struct pt_reg
>>  	ret = -EFAULT;
>>  	if (tmp)
>>  		goto out;
>> -	memset(&info.vm86plus, 0, (int)&info.regs32 - (int)&info.vm86plus);
>> +	memset(&info.vm86plus, 0,
>> +	       offsetof(struct kernel_vm86_struct, regs32) -
>> +	       offsetof(struct kernel_vm86_struct, vm86plus));
> 
> I do not think this makes it more readable... (int) -> (char *) would
> make it portable and readable, AFAICT.
> 								Pavel

The right way to do it is:

	memset(&info.vm86plus, 0, sizeof info.vm86plus);

	-hpa

Re: [PATCH 1/3] arch/x86: Use offsetof

[Al Viro]

On Wed, Dec 26, 2007 at 05:01:38PM -0800, H. Peter Anvin wrote:

> The right way to do it is:
> 
> 	memset(&info.vm86plus, 0, sizeof info.vm86plus);

If it's just one field _and_ we don't have padding we want to zero out -
certainly...

Re: [PATCH 1/3] arch/x86: Use offsetof

[H. Peter Anvin]

Al Viro wrote:
> On Wed, Dec 26, 2007 at 05:01:38PM -0800, H. Peter Anvin wrote:
> 
>> The right way to do it is:
>>
>> 	memset(&info.vm86plus, 0, sizeof info.vm86plus);
> 
> If it's just one field _and_ we don't have padding we want to zero out -
> certainly...

It is - [comments removed for clarity]:

struct kernel_vm86_struct {
         struct kernel_vm86_regs regs;
#define VM86_TSS_ESP0 flags
         unsigned long flags;
         unsigned long screen_bitmap;
         unsigned long cpu_type;
         struct revectored_struct int_revectored;
         struct revectored_struct int21_revectored;
         struct vm86plus_info_struct vm86plus;
         struct pt_regs *regs32;
};

Re: [PATCH 1/3] arch/x86: Use offsetof

[Jan Engelhardt]

On Dec 26 2007 17:01, H. Peter Anvin wrote:
>> > @@ -215,7 +215,9 @@ asmlinkage int sys_vm86old(struct pt_reg
>> >   ret = -EFAULT;
>> >   if (tmp)
>> > 		goto out;
>> > -	memset(&info.vm86plus, 0, (int)&info.regs32 - (int)&info.vm86plus);
>> > +	memset(&info.vm86plus, 0,
>> > +	       offsetof(struct kernel_vm86_struct, regs32) -
>> > +	       offsetof(struct kernel_vm86_struct, vm86plus));
>> 
>> I do not think this makes it more readable... (int) -> (char *) would
>> make it portable and readable, AFAICT.
>>         Pavel
>
> The right way to do it is:
>
> 	memset(&info.vm86plus, 0, sizeof info.vm86plus);

Either way, downcasting a pointer to (int) is dangerous,
even if this one occurrence happens to be in 32-bit-only code.

Re: [PATCH 1/3] arch/x86: Use offsetof

[H. Peter Anvin]

Jan Engelhardt wrote:
> On Dec 26 2007 17:01, H. Peter Anvin wrote:
>>>> @@ -215,7 +215,9 @@ asmlinkage int sys_vm86old(struct pt_reg
>>>>   ret = -EFAULT;
>>>>   if (tmp)
>>>> 		goto out;
>>>> -	memset(&info.vm86plus, 0, (int)&info.regs32 - (int)&info.vm86plus);
>>>> +	memset(&info.vm86plus, 0,
>>>> +	       offsetof(struct kernel_vm86_struct, regs32) -
>>>> +	       offsetof(struct kernel_vm86_struct, vm86plus));
>>> I do not think this makes it more readable... (int) -> (char *) would
>>> make it portable and readable, AFAICT.
>>>         Pavel
>> The right way to do it is:
>>
>> 	memset(&info.vm86plus, 0, sizeof info.vm86plus);
> 
> Either way, downcasting a pointer to (int) is dangerous,
> even if this one occurrence happens to be in 32-bit-only code.

Actually, it would be safe (although stupid) in this case since the 
difference would still be 32 bits or less.

Doesn't make it any less wrong.

	-hpa

Re: [PATCH 1/3] arch/x86: Use offsetof

[Ingo Molnar]

* H. Peter Anvin <hpa@zytor.com> wrote:

>>> The right way to do it is:
>>>
>>> 	memset(&info.vm86plus, 0, sizeof info.vm86plus);
>>
>> If it's just one field _and_ we don't have padding we want to zero out -
>> certainly...
>
> It is - [comments removed for clarity]:
>
> struct kernel_vm86_struct {
>         struct kernel_vm86_regs regs;
> #define VM86_TSS_ESP0 flags
>         unsigned long flags;
>         unsigned long screen_bitmap;
>         unsigned long cpu_type;
>         struct revectored_struct int_revectored;
>         struct revectored_struct int21_revectored;
>         struct vm86plus_info_struct vm86plus;
>         struct pt_regs *regs32;
> };

hm, i'm wondering why it was done in such a complex way. Clearing a 
struct field is always done via sizeof. Maybe we lost some alignment 
assumption somewhere along the line?

	Ingo