On Thu, Jan 13, 2022 at 10:54:39AM -0800, Andy Lutomirski wrote:
> On 1/12/22 07:42, Borislav Petkov wrote:
> > On Wed, Jan 12, 2022 at 11:55:41AM +0100, Peter Zijlstra wrote:
> > > Full and proper patch below. Boris, if you could merge in x86/core that
> > > branch should then be ready for a pull req.
> > 
> > I've got this as the final version. Scream if something's wrong.
> 
> AAAAAAAAAAAAAAAAAAAAAHHHHHHHHHHHHHHHHHHHHHH!!!!!!!!!!!

:-)

> > +	/*
> > +	 * There is no _ASM_EXTABLE_TYPE_REG() for ASM, however since this is
> > +	 * ASM the registers are known and we can trivially hard-code them.
> > +	 */
> > +	_ASM_EXTABLE_TYPE(1b, 2b, EX_TYPE_POP_ZERO|EX_REG_DS)
> > +	_ASM_EXTABLE_TYPE(2b, 3b, EX_TYPE_POP_ZERO|EX_REG_ES)
> > +	_ASM_EXTABLE_TYPE(3b, 4b, EX_TYPE_POP_ZERO|EX_REG_FS)
> 
> Aside from POP_ZERO being a bit mystifying to a naive reader...
> 
> >   .endm
> >   .macro RESTORE_ALL_NMI cr3_reg:req pop=0
> > diff --git a/arch/x86/include/asm/extable_fixup_types.h b/arch/x86/include/asm/extable_fixup_types.h
> > index b5ab333e064a..503622627400 100644
> > --- a/arch/x86/include/asm/extable_fixup_types.h
> > +++ b/arch/x86/include/asm/extable_fixup_types.h
> > @@ -16,9 +16,16 @@
> >   #define EX_DATA_FLAG_SHIFT		12
> >   #define EX_DATA_IMM_SHIFT		16
> > +#define EX_DATA_REG(reg)		((reg) << EX_DATA_REG_SHIFT)
> >   #define EX_DATA_FLAG(flag)		((flag) << EX_DATA_FLAG_SHIFT)
> >   #define EX_DATA_IMM(imm)		((imm) << EX_DATA_IMM_SHIFT)
> > +/* segment regs */
> > +#define EX_REG_DS			EX_DATA_REG(8)
> > +#define EX_REG_ES			EX_DATA_REG(9)
> > +#define EX_REG_FS			EX_DATA_REG(10)
> 
> These three seem likely to work

On IRC Andrew also noted that these EX_REG_* things should be __i386__
only. Previosly when they lived as open-coded EX_DATA_REG() usage in
entry_32.S that was implied, but now ...