This adds support to boot ARM with the efi loader. The kernel side
is still missing.
Details
Details
Diff Detail
Diff Detail
- Repository
- rS FreeBSD src repository - subversion
- Lint
No Lint Coverage - Unit
No Test Coverage
Event Timeline
Comment Actions
overall looks fine to me
| sys/arm/include/metadata.h | ||
|---|---|---|
| 37–38 | What do you think about making these uint64_ts with the intent that the same metatada could be used everywhere? | |
| sys/boot/efi/boot1/Makefile | ||
| 40 | Out of curiosity why does only arm boot1 need it? | |
| sys/boot/efi/include/arm/efibind.h | ||
| 1–3 | can we share a common efibind.h? | |
| sys/boot/efi/boot1/Makefile | ||
|---|---|---|
| 40 | Also, this should be LDADD+= ${LIBSTAND}, no? | |
| sys/arm/include/metadata.h | ||
|---|---|---|
| 37–38 | They probably should be, we could have > 4GiB of ram with LPAE. | |
| sys/boot/efi/boot1/Makefile | ||
| 40 | It's for the runtime functions the compiler assumes are available. As an alternative may be able to use compiler-rt. I'll add a comment explaining this. I think this is correct, it should have DPADD+= ${LIBSTAND}. LIBSTAND is only correct when DESTDIR is set so may not be in buildenv. | |
| sys/boot/efi/include/arm/efibind.h | ||
| 1–3 | I think we could share large chunks, but would need to investigate further. | |
| sys/boot/efi/loader/arch/arm/reloc.c | ||
|---|---|---|
| 38 | ElfW_Dyn instead? | |
| 71–76 | I think we should use a common reloc.c with a bit of #ifdefery instead, actually - the i386/amd64/arm/arm64 implementations of this file are going to be >80% common, without more divergence over time I think. #if defined(__arm__)
case R_ARM_RELATIVE:
#elif defined(__i386__)
case R_386_RELATIVE:
#endif
/* Address relative to the base address. */
newaddr = (unsigned long *)(ImageBase + rel->r_offset);
*newaddr += ImageBase;
break; | |