Hi Ilias,
On Thu, 9 May 2024 at 14:15, Ilias Apalodimas ilias.apalodimas@linaro.org wrote:
Hi Raymond,
Try not to post the same patches without the fixes that were asked [0], at least not without an explanation.
On Thu, 9 May 2024 at 17:38, Raymond Mao raymond.mao@linaro.org wrote:
Add text section alignment to fix sbsign signing warning 'gaps in the section table may result in different checksums' which causes a failure of efi_image_verify_diges()
Signed-off-by: Raymond Mao raymond.mao@linaro.org
arch/arm/lib/elf_aarch64_efi.lds | 1 + 1 file changed, 1 insertion(+)
diff --git a/arch/arm/lib/elf_aarch64_efi.lds
b/arch/arm/lib/elf_aarch64_efi.lds
index 5dd98091698..bffd9a0447a 100644 --- a/arch/arm/lib/elf_aarch64_efi.lds +++ b/arch/arm/lib/elf_aarch64_efi.lds @@ -28,6 +28,7 @@ SECTIONS *(.dynamic); . = ALIGN(512); }
. = ALIGN(4096);This is not what you want. There's already an ALIGN a few lines below and you must include the symbols for _etext and _text_size. You have to move the rela.* sections after the alignment. IIRC those were inspired by [1], so you can look for guidance there. On top of that riscv will need similar fixes (and any other architecture that uses those)
In practice, I have to add ALIGN(4096) for both `.rela.*` and `.data`,
otherwise it always prompts warning when signing the EFI image as below and finally leads a failure when calling function `efi_image_verify_digest()`.
``` warning: gap in section table: .text : 0x00001000 - 0x00001c00, .data : 0x00002000 - 0x00002200, gaps in the section table may result in different checksums warning: data remaining[7680 vs 12720]: gaps between PE/COFF sections? ```
I am not sure if RISC-V is working, never tried to build with that. But arm32, x86 should be fine - they don't have the `.rela.*` sections and text and data sections are well aligned.
.rela.dyn : { *(.rela.dyn) } .rela.plt : { *(.rela.plt) } .rela.got : { *(.rela.got) }-- 2.25.1
[0] https://lore.kernel.org/u-boot/CAC_iWjK+14L6mEi8GQN-McF-SChvhnjJ79nE1tWLV+kg... [1] https://git.code.sf.net/p/gnu-efi/code
Regards, Raymond