Re: [PATCH v2] kallsyms: strip ThinLTO postfix ".cfi_jt"

[Sami Tolvanen]

Hi,

On Tue, Jul 27, 2021 at 7:07 AM Padmanabha Srinivasaiah
<treasure4paddy@xxxxxxxxx> wrote:
>
> Clang ThinLTO adds a postfix ".cfi_jt" to a symbols of extern functions.

These symbols are added with CONFIG_CFI_CLANG no matter which LTO mode
is selected, so talking about ThinLTO here isn't quite correct.

> For example this breaks syscall tracer that doesn't expect such postfix,
> so strip out the postfix from the output.
>
> Signed-off-by: Padmanabha Srinivasaiah <treasure4paddy@xxxxxxxxx>
> ---
> Change in v2:
>   - Use existing routine in kallsyms to strip postfix ".cfi_jt" from
>     extern function name.
>   - Modified the commit message accordingly
>
>  kernel/kallsyms.c | 12 ++++++++----
>  1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/kallsyms.c b/kernel/kallsyms.c
> index 0ba87982d017..e9148626ae6c 100644
> --- a/kernel/kallsyms.c
> +++ b/kernel/kallsyms.c
> @@ -166,16 +166,20 @@ static unsigned long kallsyms_sym_address(int idx)
>
>  #if defined(CONFIG_CFI_CLANG) && defined(CONFIG_LTO_CLANG_THIN)
>  /*
> - * LLVM appends a hash to static function names when ThinLTO and CFI are
> - * both enabled, i.e. foo() becomes foo$707af9a22804d33c81801f27dcfe489b.
> - * This causes confusion and potentially breaks user space tools, so we
> - * strip the suffix from expanded symbol names.
> + * LLVM appends a hash to static function names and just ".cfi_jt" postfix
> + * for non-static functions when both ThinLTO and CFI are enabled,

Functions aren't technically speaking renamed to add a .cfi_jt
postfix. Instead, these are separate symbols that point to the CFI
jump table. Perhaps the comment should just say that we want to strip
.cfi_jt from CFI jump table symbols?

> + * i.e. for example foo() becomes foo$707af9a22804d33c81801f27dcfe489b.
> + * This causes confusion and potentially breaks user space tools and
> + * built-in components, so we strip the suffix from expanded symbol names.
>   */
>  static inline bool cleanup_symbol_name(char *s)
>  {
>         char *res;
>
>         res = strrchr(s, '$');
> +       if (!res)
> +               res = strstr(s, ".cfi_jt");
> +
>         if (res)
>                 *res = '\0';

This looks otherwise fine to me, but it's going to conflict with
Nick's earlier patch:

https://lore.kernel.org/lkml/20210707181814.365496-1-ndesaulniers@xxxxxxxxxx/

Could you please rebase this on top of that, and take into account
that we should do this when CONFIG_LTO_CLANG is enabled, not only with
LTO_CLANG_THIN?

Sami