Re: [PATCH v2] module: Ignore RISC-V mapping symbols too

[Jon Hunter]

Hello!

On 07/07/2023 17:00, Palmer Dabbelt wrote:
> RISC-V has an extended form of mapping symbols that we use to encode
> the ISA when it changes in the middle of an ELF.  This trips up modpost
> as a build failure, I haven't yet verified it yet but I believe the
> kallsyms difference should result in stacks looking sane again.
>
> Reported-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> Closes: https://lore.kernel.org/all/9d9e2902-5489-4bf0-d9cb-556c8e5d71c2@xxxxxxxxxxxxx/
> Signed-off-by: Palmer Dabbelt <palmer@xxxxxxxxxxxx>
> ---
> Changes since v1 <20230707054007.32591-1-palmer@xxxxxxxxxxxx/>:
>
> * Drop the unnecessary IS_RISCV define and just inline it.
> ---
>   include/linux/module_symbol.h | 12 +++++++++++-
>   kernel/module/kallsyms.c      |  2 +-
>   scripts/mod/modpost.c         |  2 +-
>   3 files changed, 13 insertions(+), 3 deletions(-)
>
> diff --git a/include/linux/module_symbol.h b/include/linux/module_symbol.h
> index 7ace7ba30203..5b799942b243 100644
> --- a/include/linux/module_symbol.h
> +++ b/include/linux/module_symbol.h
> @@ -3,12 +3,22 @@
>   #define _LINUX_MODULE_SYMBOL_H
>   
>   /* This ignores the intensely annoying "mapping symbols" found in ELF files. */
> -static inline int is_mapping_symbol(const char *str)
> +static inline int is_mapping_symbol(const char *str, int is_riscv)
>   {
>   	if (str[0] == '.' && str[1] == 'L')
>   		return true;
>   	if (str[0] == 'L' && str[1] == '0')
>   		return true;
> +	/*
> +	 * RISC-V defines various special symbols that start with "$".  The
> +	 * mapping symbols, which exist to differentiate between incompatible
> +	 * instruction encodings when disassembling, show up all over the place
> +	 * and are generally not meant to be treated like other symbols.  So
> +	 * just ignore any of the special symbols.
> +	 */
> +	if (is_riscv)
> +		return str[0] == '$';
> +
>   	return str[0] == '$' &&
>   	       (str[1] == 'a' || str[1] == 'd' || str[1] == 't' || str[1] == 'x')
>   	       && (str[2] == '\0' || str[2] == '.');
> diff --git a/kernel/module/kallsyms.c b/kernel/module/kallsyms.c
> index ef73ae7c8909..78a1ffc399d9 100644
> --- a/kernel/module/kallsyms.c
> +++ b/kernel/module/kallsyms.c
> @@ -289,7 +289,7 @@ static const char *find_kallsyms_symbol(struct module *mod,
>   		 * and inserted at a whim.
>   		 */
>   		if (*kallsyms_symbol_name(kallsyms, i) == '\0' ||
> -		    is_mapping_symbol(kallsyms_symbol_name(kallsyms, i)))
> +		    is_mapping_symbol(kallsyms_symbol_name(kallsyms, i), IS_ENABLED(CONFIG_RISCV)))
>   			continue;
>   
>   		if (thisval <= addr && thisval > bestval) {
> diff --git a/scripts/mod/modpost.c b/scripts/mod/modpost.c
> index b29b29707f10..7c71429d6502 100644
> --- a/scripts/mod/modpost.c
> +++ b/scripts/mod/modpost.c
> @@ -1052,7 +1052,7 @@ static inline int is_valid_name(struct elf_info *elf, Elf_Sym *sym)
>   
>   	if (!name || !strlen(name))
>   		return 0;
> -	return !is_mapping_symbol(name);
> +	return !is_mapping_symbol(name, elf->hdr->e_machine == EM_RISCV);
>   }
>   
>   /* Look up the nearest symbol based on the section and the address */


Since this commit was added, our builders are failing with the following 
error ...

kernel/scripts/mod/modpost.c: In function ‘is_valid_name’:
kernel/scripts/mod/modpost.c:1055:57: error: ‘EM_RISCV’ undeclared 
(first use in this function)
  return !is_mapping_symbol(name, elf->hdr->e_machine == EM_RISCV);
                                                         ^

We have had some cases of this before [0] and was fixed by ensuring this 
is defined if it is not.

Thanks
Jon

[0] 
https://patchwork.kernel.org/project/linux-mm/patch/20210924224338.8vuOlQPWr%25akpm@xxxxxxxxxxxxxxxxxxxx/

--
nvpublic