[PATCH v3 1/5] ARM: protect usage of cr_alignment by #ifdef CONFIG_CPU_CP15

[Uwe Kleine-König]

On Wed, Mar 14, 2012 at 11:24:36AM -0400, Nicolas Pitre wrote:
> On Wed, 14 Mar 2012, Uwe Kleine-König wrote:
> 
> > Signed-off-by: Uwe Kleine-König <u.kleine-koenig at pengutronix.de>
> 
> I think you could reduce the amount of #ifdef's significantly here.  
> While the cr_alignment variable and its sister are of no use in a 
> no-CP15 system, you could still leave them there initialized to zero and 
> only conditionally compile out the CP15 modifyers with a dummy version.  
> This is only 2 words wasted in your kernel image to keep the code 
> cleaner.
The main purpose to create this patch is that for v7m I don't compile
entry-armv.S. As cr_alignment is defined there I don't "have" this
variable. So I have to move the definition to a different file to waste
it :-)

> More comments below.
> 
> > index 854bd22..efeb2d0 100644
> > --- a/arch/arm/kernel/head-common.S
> > +++ b/arch/arm/kernel/head-common.S
> > @@ -98,8 +98,10 @@ __mmap_switched:
> >  	str	r9, [r4]			@ Save processor ID
> >  	str	r1, [r5]			@ Save machine type
> >  	str	r2, [r6]			@ Save atags pointer
> > -	bic	r4, r0, #CR_A			@ Clear 'A' bit
> > -	stmia	r7, {r0, r4}			@ Save control register values
> > +	cmp	r7, #0
> > +	itt	ne
> > +	bicne	r4, r0, #CR_A			@ Clear 'A' bit
> > +	stmneia	r7, {r0, r4}			@ Save control register values
> 
> The kernel is compiled with -mimplicit-it so you do not need to specify 
> any it here.
ok
 
> > index a255c39..50d3df8 100644
> > --- a/arch/arm/kernel/setup.c
> > +++ b/arch/arm/kernel/setup.c
> > @@ -472,9 +472,13 @@ static void __init setup_processor(void)
> >  	cpu_cache = *list->cache;
> >  #endif
> >  
> > -	printk("CPU: %s [%08x] revision %d (ARMv%s), cr=%08lx\n",
> > +	printk("CPU: %s [%08x] revision %d (ARMv%s)",
> >  	       cpu_name, read_cpuid_id(), read_cpuid_id() & 15,
> > -	       proc_arch[cpu_architecture()], cr_alignment);
> > +	       proc_arch[cpu_architecture()]);
> > +
> > +#ifdef CONFIG_CPU_CP15
> > +	printk(KERN_CONT ", cr=%08lx\n", cr_alignment);
> > +#endif
> >  
> 
> Again, you could just leave the original display, with cr=00000000 which 
> is a fairly good representation of reality.
Maybe having

	#define cr_alignment 0

for the !CONFIG_CPU_CP15 case would be a nice alternative. This way all
places that want to modify cr_alignment fail to compile, but reading
gives a "fairly good representation of reality".

Having said that I'm not sure about "fairly good". v7m also supports
unaligned accesses. But it's not configured in a cp15 register
(obviously) but in a system register. Quoting ARMARM-v7m:

	Configuration and Control Register, CCR
	[...]
	Bit [3] UNALIGN_TRP	Controls the trapping of unaligned word or
				halfword accesses:
				0 = Trapping disabled.
				1 = Trapping enabled.
				Unaligned load-store multiples and word
				or halfword exclusive accesses always
				fault.

So we need a more general abstraction to have correct and clean code?

But note I didn't check the two different implementations deeply to be
sure not to compare two different things.

> > diff --git a/arch/arm/mm/alignment.c b/arch/arm/mm/alignment.c
> > index caf14dc..119d178 100644
> > --- a/arch/arm/mm/alignment.c
> > +++ b/arch/arm/mm/alignment.c
> > @@ -89,7 +89,11 @@ core_param(alignment, ai_usermode, int, 0600);
> >  /* Return true if and only if the ARMv6 unaligned access model is in use. */
> >  static bool cpu_is_v6_unaligned(void)
> >  {
> > +#ifdef CONFIG_CPU_CP15
> >  	return cpu_architecture() >= CPU_ARCH_ARMv6 && (cr_alignment & CR_U);
> > +#else
> > +	return 0;
> > +#endif
> >  }
> 
> Same here.  With cr_alignment set to zero, you don't need the above 
> #ifdef.
> 
> >  static int safe_usermode(int new_usermode, bool warn)
> > @@ -961,12 +965,14 @@ static int __init alignment_init(void)
> >  		return -ENOMEM;
> >  #endif
> >  
> > +#ifdef CONFIG_CPU_CP15
> >  	if (cpu_is_v6_unaligned()) {
> >  		cr_alignment &= ~CR_A;
> >  		cr_no_alignment &= ~CR_A;
> >  		set_cr(cr_alignment);
> >  		ai_usermode = safe_usermode(ai_usermode, false);
> >  	}
> > +#endif
> 
> Here the #ifdef is probably legitimate.
> 
> >  	hook_fault_code(FAULT_CODE_ALIGNMENT, do_alignment, SIGBUS, BUS_ADRALN,
> >  			"alignment exception");
> > diff --git a/arch/arm/mm/mmu.c b/arch/arm/mm/mmu.c
> > index 94c5a0c..f6dbe1a 100644
> > --- a/arch/arm/mm/mmu.c
> > +++ b/arch/arm/mm/mmu.c
> > @@ -109,8 +109,10 @@ static int __init early_cachepolicy(char *p)
> >  
> >  		if (memcmp(p, cache_policies[i].policy, len) == 0) {
> >  			cachepolicy = i;
> > +#ifdef CONFIG_CPU_CP15
> >  			cr_alignment &= ~cache_policies[i].cr_mask;
> >  			cr_no_alignment &= ~cache_policies[i].cr_mask;
> > +#endif
> 
> Probably here as well.
> 
> >  			break;
> >  		}
> >  	}
> > @@ -128,7 +130,9 @@ static int __init early_cachepolicy(char *p)
> >  		cachepolicy = CPOLICY_WRITEBACK;
> >  	}
> >  	flush_cache_all();
> > +#ifdef CONFIG_CPU_CP15
> >  	set_cr(cr_alignment);
> > +#endif
> 
> However it might be best to provide a dummy set_cr() instead of 
> #ifdef'ing it out everywhere.
What should the dummy set_cr do? Print a runtime warning if called with
!= 0?

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |