Re: [PATCH 3/5] objtool: Support Clang non-section symbols in ORC generation

[Miroslav Benes]

On Wed, 1 Apr 2020, Josh Poimboeuf wrote:

> When compiling the kernel with AS=clang, objtool produces a lot of
> warnings:
> 
>   warning: objtool: missing symbol for section .text
>   warning: objtool: missing symbol for section .init.text
>   warning: objtool: missing symbol for section .ref.text
> 
> It then fails to generate the ORC table.
> 
> The problem is that objtool assumes text section symbols always exist.
> But the Clang assembler is aggressive about removing them.
> 
> When generating relocations for the ORC table, objtool always tries to
> reference instructions by their section symbol offset.  If the section
> symbol doesn't exist, it bails.
> 
> Do a fallback: when a section symbol isn't available, reference a
> function symbol instead.
> 
> Link: https://github.com/ClangBuiltLinux/linux/issues/669
> Cc: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> Reported-by: Dmitry Golovin <dima@xxxxxxxxxx>
> Tested-by: Nathan Chancellor <natechancellor@xxxxxxxxx>
> Signed-off-by: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
> ---
>  tools/objtool/orc_gen.c | 33 ++++++++++++++++++++++++++-------
>  1 file changed, 26 insertions(+), 7 deletions(-)
> 
> diff --git a/tools/objtool/orc_gen.c b/tools/objtool/orc_gen.c
> index 41e4a2754da4..4c0dabd28000 100644
> --- a/tools/objtool/orc_gen.c
> +++ b/tools/objtool/orc_gen.c
> @@ -88,11 +88,6 @@ static int create_orc_entry(struct elf *elf, struct section *u_sec, struct secti
>  	struct orc_entry *orc;
>  	struct rela *rela;
>  
> -	if (!insn_sec->sym) {
> -		WARN("missing symbol for section %s", insn_sec->name);
> -		return -1;
> -	}
> -
>  	/* populate ORC data */
>  	orc = (struct orc_entry *)u_sec->data->d_buf + idx;
>  	memcpy(orc, o, sizeof(*orc));
> @@ -105,8 +100,32 @@ static int create_orc_entry(struct elf *elf, struct section *u_sec, struct secti
>  	}
>  	memset(rela, 0, sizeof(*rela));
>  
> -	rela->sym = insn_sec->sym;
> -	rela->addend = insn_off;
> +	if (insn_sec->sym) {
> +		rela->sym = insn_sec->sym;
> +		rela->addend = insn_off;
> +	} else {
> +		/*
> +		 * The Clang assembler doesn't produce section symbols, so we
> +		 * have to reference the function symbol instead:
> +		 */
> +		rela->sym = find_symbol_containing(insn_sec, insn_off);
> +		if (!rela->sym) {
> +			/*
> +			 * Hack alert.  This happens when we need to reference
> +			 * the NOP pad insn immediately after the function.
> +			 */
> +			rela->sym = find_symbol_containing(insn_sec,
> +							   insn_off - 1);
> +		}

I suppose there is always just one NOP pad insn, right? Anyway, it would 
be better to get rid of it as you proposed.

> +		if (!rela->sym) {
> +			WARN("missing symbol for insn at offset 0x%lx\n",
> +			     insn_off);
> +			return -1;
> +		}
> +
> +		rela->addend = insn_off - rela->sym->offset;
> +	}
> +
>  	rela->type = R_X86_64_PC32;
>  	rela->offset = idx * sizeof(int);
>  	rela->sec = ip_relasec;

Miroslav