Re: [PATCH] powerpc/32: Add missing context synchronisation with CONFIG_VMAP_STACK

[Christophe Leroy]

Michael Ellerman <mpe@xxxxxxxxxxxxxx> a ÃcritÂ:

> Christophe Leroy <christophe.leroy@xxxxxx> writes:
> > After reactivation of data translation by modifying MSR[DR], a isync
> > is required to ensure the translation is effective.
> >
> > Signed-off-by: Christophe Leroy <christophe.leroy@xxxxxx>
> > ---
> > Rebased on powerpc/merge-test
> >
> > @mpe: If not too late:
> > - change to head_32.h should be squashed into "powerpc/32: prepare  
> > for CONFIG_VMAP_STACK"
> > - change to head_32.S should be squashed into "powerpc/32s: Enable  
> > CONFIG_VMAP_STACK"
> >
> > Signed-off-by: Christophe Leroy <christophe.leroy@xxxxxx>
> > ---
> >  arch/powerpc/kernel/head_32.S | 1 +
> >  arch/powerpc/kernel/head_32.h | 2 ++
> >  2 files changed, 3 insertions(+)
> >
> > diff --git a/arch/powerpc/kernel/head_32.h b/arch/powerpc/kernel/head_32.h
> > index 73a035b40dbf..a6a5fbbf8504 100644
> > --- a/arch/powerpc/kernel/head_32.h
> > +++ b/arch/powerpc/kernel/head_32.h
> > @@ -43,6 +43,7 @@
> >  	.ifeq	\for_rtas
> >  	li	r11, MSR_KERNEL & ~(MSR_IR | MSR_RI) /* can take DTLB miss */
> >  	mtmsr	r11
> > +	isync
>
> Actually this one leads to:
>
>   /home/michael/linux/arch/powerpc/kernel/head_8xx.S: Assembler messages:
>   /home/michael/linux/arch/powerpc/kernel/head_8xx.S:151: Error:  
> attempt to move .org backwards
>   make[3]: *** [/home/michael/linux/scripts/Makefile.build:348:  
> arch/powerpc/kernel/head_8xx.o] Error 1
>
> For mpc885_ads_defconfig.
>
> That's the alignment exception overflowing into the program check
> handler:
>
> /* Alignment exception */
> 	. = 0x600
> Alignment:
> 	EXCEPTION_PROLOG handle_dar_dsisr=1
> 	save_dar_dsisr_on_stack r4, r5, r11
> 	li	r6, RPN_PATTERN
> 	mtspr	SPRN_DAR, r6	/* Tag DAR, to be used in DTLB Error */
> 	addi	r3,r1,STACK_FRAME_OVERHEAD
> 	EXC_XFER_STD(0x600, alignment_exception)
>
> /* Program check exception */
> 	EXCEPTION(0x700, ProgramCheck, program_check_exception, EXC_XFER_STD)
>
>
> Can't see an obvious/easy way to fix it.

Argh !

I think the easiest is to move the EXC_XFER_STD(0x600,  
alignment_exception) somewhere else and branch to it. There is some  
space at .0xa00

Christophe

PS: I'm afk until monday