[PATCH] ARM U-Boot BSS zeroing

[PATCH] ARM U-Boot BSS zeroing

[Leif Lindholm]

[-- Attachment #1: Type: text/plain, Size: 432 bytes --]

The import into the collaborative branch discarded the call to
get_real_bss_start before starting the BSS zeroing operation.

While get_real_bss_start might have been a bit of an ugly hack,
simply discarding it means that we now end up doing unaligned
STR operations, and any platform with MMU disabled fails to boot.

The attached patch prepends a bytewise zeroing loop until
word-aligned, and then continue as before.

/
    Leif

[-- Attachment #2: bss_zero_fix.patch --]
[-- Type: text/x-diff, Size: 592 bytes --]

=== modified file 'grub-core/kern/arm/uboot/startup.S'
--- grub-core/kern/arm/uboot/startup.S	2013-05-03 13:07:39 +0000
+++ grub-core/kern/arm/uboot/startup.S	2013-05-13 14:59:43 +0000
@@ -100,7 +100,13 @@
 	@ Since we _are_ the C run-time, we need to manually zero the BSS
 	@ region before continuing
 	ldr	r0, =EXT_C(__bss_start)	@ zero from here
-	ldr	r1, =EXT_C(_end)		@ to here
+	@ If unaligned, bytewise zero until base address aligned.
+	mov	r1, #0
+1:	tst	r0, #3
+	beq	2f
+	strb	r1, [r0], #1
+	b	1b
+2:	ldr	r1, =EXT_C(_end)		@ to here
 	mov	r2, #0
 1:	str	r2, [r0], #4
 	cmp	r0, r1

Re: [PATCH] ARM U-Boot BSS zeroing

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 792 bytes --]

Could you request savannah membership so you have commit rights to this
branch?
Will you be at LinuxTag?
On 13.05.2013 17:37, Leif Lindholm wrote:

> The import into the collaborative branch discarded the call to
> get_real_bss_start before starting the BSS zeroing operation.
> 
> While get_real_bss_start might have been a bit of an ugly hack,
> simply discarding it means that we now end up doing unaligned
> STR operations, and any platform with MMU disabled fails to boot.
> 
> The attached patch prepends a bytewise zeroing loop until
> word-aligned, and then continue as before.
> 
> /
>     Leif
> 
> 
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: [PATCH] ARM U-Boot BSS zeroing

[Goswin von Brederlow]

On Mon, May 13, 2013 at 03:37:32PM +0000, Leif Lindholm wrote:
> The import into the collaborative branch discarded the call to
> get_real_bss_start before starting the BSS zeroing operation.
> 
> While get_real_bss_start might have been a bit of an ugly hack,
> simply discarding it means that we now end up doing unaligned
> STR operations, and any platform with MMU disabled fails to boot.
> 
> The attached patch prepends a bytewise zeroing loop until
> word-aligned, and then continue as before.
> 
> /
>     Leif

> === modified file 'grub-core/kern/arm/uboot/startup.S'
> --- grub-core/kern/arm/uboot/startup.S	2013-05-03 13:07:39 +0000
> +++ grub-core/kern/arm/uboot/startup.S	2013-05-13 14:59:43 +0000
> @@ -100,7 +100,13 @@
>  	@ Since we _are_ the C run-time, we need to manually zero the BSS
>  	@ region before continuing
>  	ldr	r0, =EXT_C(__bss_start)	@ zero from here
> -	ldr	r1, =EXT_C(_end)		@ to here
> +	@ If unaligned, bytewise zero until base address aligned.
> +	mov	r1, #0
> +1:	tst	r0, #3
> +	beq	2f
> +	strb	r1, [r0], #1
> +	b	1b
> +2:	ldr	r1, =EXT_C(_end)		@ to here
>  	mov	r2, #0
>  1:	str	r2, [r0], #4
>  	cmp	r0, r1

What if I have a 1 byte BSS aligned at 1 byte? You would zero 3 bytes
and overshoot _end and then what happens after the cmp in the last line?

MfG
	Goswin

Re: [PATCH] ARM U-Boot BSS zeroing

[Vladimir 'phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 1795 bytes --]

On May 14, 2013 3:02 PM, "Goswin von Brederlow" <goswin-v-b@web.de> wrote:
>
> On Mon, May 13, 2013 at 03:37:32PM +0000, Leif Lindholm wrote:
> > The import into the collaborative branch discarded the call to
> > get_real_bss_start before starting the BSS zeroing operation.
> >
> > While get_real_bss_start might have been a bit of an ugly hack,
> > simply discarding it means that we now end up doing unaligned
> > STR operations, and any platform with MMU disabled fails to boot.
> >
> > The attached patch prepends a bytewise zeroing loop until
> > word-aligned, and then continue as before.
> >
> > /
> >     Leif
>
> > === modified file 'grub-core/kern/arm/uboot/startup.S'
> > --- grub-core/kern/arm/uboot/startup.S        2013-05-03 13:07:39 +0000
> > +++ grub-core/kern/arm/uboot/startup.S        2013-05-13 14:59:43 +0000
> > @@ -100,7 +100,13 @@
> >       @ Since we _are_ the C run-time, we need to manually zero the BSS
> >       @ region before continuing
> >       ldr     r0, =EXT_C(__bss_start) @ zero from here
> > -     ldr     r1, =EXT_C(_end)                @ to here
> > +     @ If unaligned, bytewise zero until base address aligned.
> > +     mov     r1, #0
> > +1:   tst     r0, #3
> > +     beq     2f
> > +     strb    r1, [r0], #1
> > +     b       1b
> > +2:   ldr     r1, =EXT_C(_end)                @ to here
> >       mov     r2, #0
> >  1:   str     r2, [r0], #4
> >       cmp     r0, r1
>
> What if I have a 1 byte BSS aligned at 1 byte? You would zero 3 bytes
> and overshoot _end and then what happens after the cmp in the last line?
>
Abaolutely irrelevant. our BSS isn't that small

> MfG
>         Goswin
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel

[-- Attachment #2: Type: text/html, Size: 2435 bytes --]