Re: [PATCH 04/24] x86/entry: Expose the address of .Lgs_change to traps.c

[Lai Jiangshan]

On 2021/9/2 17:14, Peter Zijlstra wrote:
> On Wed, Sep 01, 2021 at 01:50:05AM +0800, Lai Jiangshan wrote:
>
> > diff --git a/arch/x86/entry/entry_64.S b/arch/x86/entry/entry_64.S
> > index e38a4cf795d9..9164e85b36b8 100644
> > --- a/arch/x86/entry/entry_64.S
> > +++ b/arch/x86/entry/entry_64.S
> > @@ -729,6 +729,7 @@ _ASM_NOKPROBE(common_interrupt_return)
> >   SYM_FUNC_START(asm_load_gs_index)
> >   	FRAME_BEGIN
> >   	swapgs
> > +SYM_INNER_LABEL(asm_load_gs_index_gs_change, SYM_L_GLOBAL)
> >   .Lgs_change:
>
> After all the patches, there's only a single .Lgs_change reference left.
> Can't we convert the _ASM_EXTABLE() entry to use the new label and
> totally remove it?

The label .Lgs_change is still needed in ASM code for extable.

I tried the way as you suggested, and the result is:

warning: objtool: __ex_table+0x0: don't know how to handle non-section reloc symbol asm_load_gs_index_gs_change

But never mind, I have already converted load_gs_index into C code.
I will send it later.

> >   	movl	%edi, %gs
> >   2:	ALTERNATIVE "", "mfence", X86_BUG_SWAPGS_FENCE
> > @@ -1011,7 +1012,7 @@ SYM_CODE_START_LOCAL(error_entry)
> >   	movl	%ecx, %eax			/* zero extend */
> >   	cmpq	%rax, RIP+8(%rsp)
> >   	je	.Lbstep_iret
> > -	cmpq	$.Lgs_change, RIP+8(%rsp)
> > +	cmpq	$asm_load_gs_index_gs_change, RIP+8(%rsp)
> >   	jne	.Lerror_entry_done_lfence