using identity_mapping_add & switching MMU state - how ?

[Frank Hofmann]

Hi,


I'm trying to find a way to do an MMU off / on transition.

What I want to do is to call cpu_do_resume() from the hibernation restore 
codepath.

I've succeeded to make this work by adding a test whether the MMU is 
already on when cpu_resume_mmu() is called.

I'm not sure that sort of thing is proper; hence I've been trying to find 
a way to disable the MMU before calling cpu_do_resume().

Can't seem to get this to work though; even though I'm creating a separate 
MMU context that's given 1:1 mappings for all of kernel code/data, 
execution still hangs as soon as I enable the code section below that 
switches the MMU off.


What I have at the moment is code that looks like this:

==============================================================================
[ ... ]
unsigned long notrace __swsusp_arch_restore_image(void)
{
 	extern struct pbe *restore_pblist;
 	struct pbe *pbe;

 	/* __swsusp_pg_dir has been created using pgd_alloc(&init_mm); */

 	cpu_switch_mm(__swsusp_pg_dir, &init_mm);

 	for (pbe = restore_pblist; pbe; pbe = pbe->next)
 		copy_page(pbe->orig_address, pbe->address);

 	flush_tlb_all();
 	flush_cache_all();

 	identity_mapping_add(__swsusp_pg_dir, __pa(_stext), __pa(_etext));
 	identity_mapping_add(__swsusp_pg_dir, __pa(_sdata), __pa(_edata));

 	cpu_switch_mm(__swsusp_pg_dir, &init_mm);

 	flush_tlb_all();
 	flush_cache_all();

 	cpu_proc_fin();		/* turns caches off */

 	/* caller requires v:p offset to calculate physical addresses */
 	return (unsigned long)(PHYS_OFFSET - PAGE_OFFSET);
}

[ ... ]
ENTRY(swsusp_arch_resume)
 	mov	r2, #PSR_I_BIT | PSR_F_BIT | SVC_MODE
 	msr	cpsr_c, r2
 	/*
 	 * Switch stack to a nosavedata region to make sure image restore
 	 * doesn't clobber it underneath itself.
 	 */
 	ldr	sp, =(__swsusp_resume_stk + PAGE_SIZE / 2)
 	bl	__swsusp_arch_restore_image

 	/*
 	 * Restore the CPU registers.
 	 */
 	mov	r1, r0
 	ldr	r0, =(__swsusp_arch_ctx + (NREGS * 4))
/*
  * This is what I'm trying to switch off; yet, doing so makes things hang
  */
#if 0
 	ldr	r2, =cpu_do_resume
 	sub	r2, r1			@ __pa()
 	ldr	r3, =.Lmmu_is_off
 	sub	r3, r1			@ __pa()
 	sub	r0, r1			@ __pa()
 	ldr	lr, =.Lpost_mmu
 	mrc	p15, 0, r1, c1, c0, 0
 	bic	r1, #CR_M
 	mcr	p15, 0, r1, c1, c0, 0	@ MMU OFF

 	mrc	p15, 0, r1, c2, c0, 0	@ queue a dependency on CP15
         sub     pc, r3, r1, lsr #32	@ to local label, phys addr
.ltorg
.align 5
.Lmmu_is_off:
 	mov	pc, r2			@ jump to phys cpu_v6_do_resume
.Lpost_mmu:
#else
 	bl	cpu_do_resume
#endif
 	ldr	r0, =__swsusp_arch_ctx
 	ldmia	r0!, {r1-r11,lr}	@ nonvolatile regs
 	ldr	sp, [r0]		@ stack
 	msr	cpsr, r1
 	msr	spsr, r2

 	mov	r0, #0
 	stmfd	sp!, {r0, lr}
 	bl	cpu_init		@ reinitialize other modes
 	ldmfd	sp!, {r0, pc}
ENDPROC(swsusp_arch_resume)
==============================================================================

I.e. it performs the steps:

 	- flush all caches, tlbs
 	- setup identity mappings for all kernel code & data
 	- switch to a self-contained pagedir
 	- flush again
 	- finish cpu (disable caches)
 	- switch MMU off, re-read config reg to force necessary wait,
 	  and jump to physical address of "self".


As said, things work just fine if simply doing the "bl cpu_do_resume" and 
adding a "MMU already on" check to cpu_resume_mmu().


I must be missing something there; I've been reading the ARM kexec 
postings,

http://lists.infradead.org/pipermail/linux-arm-kernel/2010-July/020183.html

for the basic idea, and used the style from smp.c (alloc a temporary 
pagedir, create identity mappings there). Still, there's something not 
quite right ...


Any ideas what I'm missing ?
Thanks,

FrankH.