On Thu, 21 Oct 2021 19:42:48 +0800
Kefeng Wang <wangkefeng.wang@xxxxxxxxxx> wrote:
> On 2021/10/20 22:07, Jisheng Zhang wrote:
> > From: Jisheng Zhang <jszhang@xxxxxxxxxx>
> >
> > Similar as other architectures such as arm64, x86 and so on, use
> > offsets relative to the exception table entry values rather than
> > absolute addresses for both the exception locationand the fixup.
> >
> > However, RISCV label difference will actually produce two relocations,
> > a pair of R_RISCV_ADD32 and R_RISCV_SUB32. Take below simple code for
> > example:
> >
> > $ cat test.S
> > .section .text
> > 1:
> > nop
> > .section __ex_table,"a"
> > .balign 4
> > .long (1b - .)
> > .previous
> >
> > $ riscv64-linux-gnu-gcc -c test.S
> > $ riscv64-linux-gnu-readelf -r test.o
> > Relocation section '.rela__ex_table' at offset 0x100 contains 2 entries:
> > Offset Info Type Sym. Value Sym. Name + Addend
> > 000000000000 000600000023 R_RISCV_ADD32 0000000000000000 .L1^B1 + 0
> > 000000000000 000500000027 R_RISCV_SUB32 0000000000000000 .L0 + 0
> >
> > The modpost will complain the R_RISCV_SUB32 relocation, so we need to
> > patch modpost.c to skip this relocation for .rela__ex_table section.
> >
> > After this patch, the __ex_table section size of defconfig vmlinux is
> > reduced from 7072 Bytes to 3536 Bytes.
> >
> > Signed-off-by: Jisheng Zhang <jszhang@xxxxxxxxxx>
> > ---
>
>
>
> ...
> > diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c
> > index cb8ab7d91d30..0aa14b5bd124 100644
> > --- a/scripts/mod/modpost.c
> > +++ b/scripts/mod/modpost.c
> > @@ -1830,6 +1830,27 @@ static int addend_mips_rel(struct elf_info *elf, Elf_Shdr *sechdr, Elf_Rela *r)
> > return 0;
> > }
> >
> > +#ifndef EM_RISCV
> > +#define EM_RISCV 243
> > +#endif
> > +
> > +#ifndef R_RISCV_SUB32
> > +#define R_RISCV_SUB32 39
> > +#endif
> > +
> > +static int addend_riscv_rela(struct elf_info *elf, Elf_Shdr *sechdr, Elf_Rela *r)
> > +{
> > + unsigned int r_typ = ELF_R_TYPE(r->r_info);
> > + const char *fromsec;
> > +
> > + fromsec = sech_name(elf, sechdr);
> > + fromsec += strlen(".rela");
> > +
> > + if (!strcmp("__ex_table", fromsec) && r_typ == R_RISCV_SUB32)
> > + return 1; /* skip this */
> > + return 0;
> > +}
> > +
> > static void section_rela(const char *modname, struct elf_info *elf,
> > Elf_Shdr *sechdr)
> > {
> > @@ -1866,6 +1887,12 @@ static void section_rela(const char *modname, struct elf_info *elf,
> > r_sym = ELF_R_SYM(r.r_info);
> > #endif
> > r.r_addend = TO_NATIVE(rela->r_addend);
> > + switch (elf->hdr->e_machine) {
> > + case EM_RISCV:
> > + if (addend_riscv_rela(elf, sechdr, &r))
> directly use
> if (!strcmp("__ex_table", fromsec) && ELF_R_TYPE(r.r_info) == R_RISCV_SUB32)
My local initial patch is as direct as this LoC;) then later I thought we may
need do something more in future. But now I think we can do code
refactoring then. Will send patch v3 soon.
Thanks
>
> > + continue;
> > + break;
> > + }
> > sym = elf->symtab_start + r_sym;
> > /* Skip special sections */
> > if (is_shndx_special(sym->st_shndx))
> > diff --git a/scripts/sorttable.c b/scripts/sorttable.c
> > index 6ee4fa882919..39e86e4acea3 100644
> > --- a/scripts/sorttable.c
> > +++ b/scripts/sorttable.c
> > @@ -346,6 +346,7 @@ static int do_file(char const *const fname, void *addr)
> > case EM_PARISC:
> > case EM_PPC:
> > case EM_PPC64:
> > + case EM_RISCV:
> > custom_sort = sort_relative_table;
> > break;
> > case EM_ARCOMPACT:
> > @@ -353,7 +354,6 @@ static int do_file(char const *const fname, void *addr)
> > case EM_ARM:
> > case EM_MICROBLAZE:
> > case EM_MIPS:
> > - case EM_RISCV:
> > case EM_XTENSA:
> > break;
> > default:
> >
