RE: [PATCH] PER_CPU [3/4] - PER_CPU-init_tss

["Shai Fultheim" <shai@scalex86.org>]

I don't think you have to worry about this.

Ring 1 is never used by Linux so the ss1 and esp1 are NOT used for their real purpose in the task struct.  Instead, data is put
there to be used later for setting up sysenter/sysexit in the 'enable_sep_cpu' (arch/i386/kernel/sysenter.c) function.  So, the
initialization is really unimportant.

I suggest you'll 'grep' the kernel for esp1 (I did that). It has no other usage other than (as said) setting up the
sysenter/sysexit. 

Let me know if that’s helps.

Thanks !
 
-----------------
Shai Fultheim
Scalex86.org


> -----Original Message-----
> From: Andrew Morton [mailto:akpm@osdl.org]
> Sent: Thursday, July 08, 2004 19:08
> To: Shai Fultheim
> Cc: linux-kernel@vger.kernel.org; mort@wildopensource.com; jes@wildopensource.com
> Subject: Re: [PATCH] PER_CPU [3/4] - PER_CPU-init_tss
> 
> "Shai Fultheim" <shai@scalex86.org> wrote:
> >
> >  #define INIT_TSS  {							\
> >   	.esp0		= sizeof(init_stack) + (long)&init_stack,	\
> >   	.ss0		= __KERNEL_DS,					\
> >  -	.esp1		= sizeof(init_tss[0]) + (long)&init_tss[0],	\
> >   	.ss1		= __KERNEL_CS,					\
> >   	.ldt		= GDT_ENTRY_LDT,				\
> >   	.io_bitmap_base	= INVALID_IO_BITMAP_OFFSET,			\
> 
> Why this change?  Is it safe?
>