Hi Udit
-----Original Message----- From: Udit Kumar udit.kumar@nxp.com Sent: 28 February 2019 09:07 To: Ibai Erkiaga Elorza IBAIE@xilinx.com; u-boot@lists.denx.de Cc: Tom Rini trini@konsulko.com; Albert Aribaud albert.u.boot@aribaud.net; Meenakshi Aggarwal meenakshi.aggarwal@nxp.com; Stefan Roese sr@denx.de; Prabhakar Kushwaha prabhakar.kushwaha@nxp.com Subject: RE: [U-Boot][PATCH] arm: arm64 32bit address relocation
Hi Ibai
-----Original Message----- From: Ibai Erkiaga ibai.erkiaga-elorza@xilinx.com Sent: Wednesday, February 27, 2019 8:18 PM To: u-boot@lists.denx.de Cc: Ibai Erkiaga ibai.erkiaga-elorza@xilinx.com; Udit Kumar udit.kumar@nxp.com; Tom Rini trini@konsulko.com; Albert Aribaud albert.u.boot@aribaud.net; Meenakshi Aggarwal meenakshi.aggarwal@nxp.com; Stefan Roese sr@denx.de; Prabhakar Kushwaha prabhakar.kushwaha@nxp.com Subject: [U-Boot][PATCH] arm: arm64 32bit address relocation
Current relocation code is limited to 21bit PC-relative addressing which might not be enough for bigger code sizes. The following patch increases the addressing to 32bit PC-relative. This feature is specially interesting if U-Boot is build without optimiation (-O0) as the text section is increased significativelly.
Signed-off-by: Ibai Erkiaga ibai.erkiaga-elorza@xilinx.com
arch/arm/lib/relocate_64.S | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-)
diff --git a/arch/arm/lib/relocate_64.S b/arch/arm/lib/relocate_64.S index 7603f52..6e9658b 100644 --- a/arch/arm/lib/relocate_64.S +++ b/arch/arm/lib/relocate_64.S @@ -26,9 +26,10 @@ ENTRY(relocate_code) /* * Copy u-boot from flash to RAM */
- adr x1, __image_copy_start /* x1 <- Run
&__image_copy_start */
- subs x9, x0, x1 /* x8 <- Run to copy offset */
- b.eq relocate_done /* skip relocation */
- adrp x1, __image_copy_start /* x1 <- address bits
[31:12] */
- add x1, x1, :lo12:__image_copy_start/* x1 <- address bits [11:00]
*/
- subs x9, x0, x1 /* x8 <- Run to copy offset */
Should be x9 in comments
The original comment was x8 so did not realize about it. Does take sense to change within my patch? Or should it be done in a different one?
- b.eq relocate_done /* skip relocation */ /*
- Don't ldr x1, __image_copy_start here, since if the code is already
- running at an address other than it was linked to, that
instruction @@ -42,8 +43,10 @@ ENTRY(relocate_code) ldr x1, _TEXT_BASE /* x1 <- Linked &__image_copy_start */ subs x9, x0, x1 /* x9 <- Link to copy offset */
- adr x1, __image_copy_start /* x1 <- Run
&__image_copy_start */
- adr x2, __image_copy_end /* x2 <- Run &__image_copy_end */
- adrp x1, __image_copy_start /* x1 <- address bits
[31:12] */
- add x1, x1, :lo12:__image_copy_start/* x1 <- address bits [11:00]
*/
- adrp x2, __image_copy_end /* x2 <- address bits [31:12]
*/
- add x2, x2, :lo12:__image_copy_end /* x2 <- address bits
[11:00] */ copy_loop: ldp x10, x11, [x1], #16 /* copy from source address [x1] */ stp x10, x11, [x0], #16 /* copy to target address [x0] */ @@ -54,8 +57,10 @@ copy_loop: /* * Fix .rela.dyn relocations */
- adr x2, __rel_dyn_start /* x2 <- Run &__rel_dyn_start */
- adr x3, __rel_dyn_end /* x3 <- Run &__rel_dyn_end */
- adrp x2, __rel_dyn_start /* x2 <- address bits [31:12]
*/
- add x2, x2, :lo12:__rel_dyn_start /* x2 <- address bits [11:00]
*/
- adrp x3, __rel_dyn_end /* x3 <- address bits [31:12]
*/
- add x3, x3, :lo12:__rel_dyn_end /* x3 <- address bits [11:00]
*/ fixloop: ldp x0, x1, [x2], #16 /* (x0,x1) <- (SRC location, fixup) */ ldr x4, [x2], #8 /* x4 <- addend */ -- 1.8.3.1