linux-kernel - Re: [PATCH 05/13] objtool: Increase per-function WARN

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.LSU.2.21.2503171327420.4236@pobox.suse.cz>
Date: Mon, 17 Mar 2025 13:29:10 +0100 (CET)
From: Miroslav Benes <mbenes@...e.cz>
To: Brendan Jackman <jackmanb@...gle.com>
cc: Josh Poimboeuf <jpoimboe@...nel.org>, x86@...nel.org, 
    linux-kernel@...r.kernel.org, Peter Zijlstra <peterz@...radead.org>, 
    Nathan Chancellor <nathan@...nel.org>
Subject: Re: [PATCH 05/13] objtool: Increase per-function WARN_FUNC() rate
 limit

On Mon, 17 Mar 2025, Brendan Jackman wrote:

> On Fri, Mar 14, 2025 at 12:29:03PM -0700, Josh Poimboeuf wrote:
> > Increase the per-function WARN_FUNC() rate limit from 1 to 2.  If the
> > number of warnings for a given function goes beyond 2, print "skipping
> > duplicate warning(s)".  This helps root out additional warnings in a
> > function that might be hiding behind the first one.
> > 
> > Signed-off-by: Josh Poimboeuf <jpoimboe@...nel.org>
> > ---
> >  tools/objtool/check.c                |  2 +-
> >  tools/objtool/include/objtool/elf.h  |  2 +-
> >  tools/objtool/include/objtool/warn.h | 14 +++++++++++---
> >  3 files changed, 13 insertions(+), 5 deletions(-)
> > 
> > diff --git a/tools/objtool/check.c b/tools/objtool/check.c
> > index 6b9ad3afe389..3ddaefe97f55 100644
> > --- a/tools/objtool/check.c
> > +++ b/tools/objtool/check.c
> > @@ -4528,7 +4528,7 @@ static int disas_warned_funcs(struct objtool_file *file)
> >  	char *funcs = NULL, *tmp;
> >  
> >  	for_each_sym(file, sym) {
> > -		if (sym->warned) {
> > +		if (sym->warnings) {
> >  			if (!funcs) {
> >  				funcs = malloc(strlen(sym->name) + 1);
> >  				strcpy(funcs, sym->name);
> > diff --git a/tools/objtool/include/objtool/elf.h b/tools/objtool/include/objtool/elf.h
> > index d7e815c2fd15..223ac1c24b90 100644
> > --- a/tools/objtool/include/objtool/elf.h
> > +++ b/tools/objtool/include/objtool/elf.h
> > @@ -65,10 +65,10 @@ struct symbol {
> >  	u8 return_thunk      : 1;
> >  	u8 fentry            : 1;
> >  	u8 profiling_func    : 1;
> > -	u8 warned	     : 1;
> >  	u8 embedded_insn     : 1;
> >  	u8 local_label       : 1;
> >  	u8 frame_pointer     : 1;
> > +	u8 warnings	     : 2;
> >  	struct list_head pv_target;
> >  	struct reloc *relocs;
> >  };
> > diff --git a/tools/objtool/include/objtool/warn.h b/tools/objtool/include/objtool/warn.h
> > index ac04d3fe4dd9..6180288927fd 100644
> > --- a/tools/objtool/include/objtool/warn.h
> > +++ b/tools/objtool/include/objtool/warn.h
> > @@ -53,14 +53,22 @@ static inline char *offstr(struct section *sec, unsigned long offset)
> >  	free(_str);					\
> >  })
> >  
> > +#define WARN_LIMIT 2
> > +
> >  #define WARN_INSN(insn, format, ...)					\
> >  ({									\
> >  	struct instruction *_insn = (insn);				\
> > -	if (!_insn->sym || !_insn->sym->warned)				\
> > +	BUILD_BUG_ON(WARN_LIMIT > 2);					\
> 
> Shouldn't this be >3? Anyway, I think it would be clearer if the
> coupling was more explicit, e.g:

I think it is correct but I also think that the difference between bits 
and the actual number of "allowed" warnings can be confusing.
 
> diff --git i/tools/objtool/include/objtool/elf.h w/tools/objtool/include/objtool/elf.h
> index 223ac1c24b90..a86e43d2056f 100644
> --- i/tools/objtool/include/objtool/elf.h
> +++ w/tools/objtool/include/objtool/elf.h
> @@ -46,6 +46,8 @@ struct section {
>         struct reloc *relocs;
>  };
> 
> +#define STRUCT_SYMBOL_WARNING_BITS 2
> +
>  struct symbol {
>         struct list_head list;
>         struct rb_node node;
> @@ -68,7 +70,7 @@ struct symbol {
>         u8 embedded_insn     : 1;
>         u8 local_label       : 1;
>         u8 frame_pointer     : 1;
> -       u8 warnings          : 2;
> +       u8 warnings          : STRUCT_SYMBOL_WARNING_BITS;
>         struct list_head pv_target;
>         struct reloc *relocs;
>  };
> diff --git i/tools/objtool/include/objtool/warn.h w/tools/objtool/include/objtool/warn.h
> index e72b9d630551..2fba6866fd2d 100644
> --- i/tools/objtool/include/objtool/warn.h
> +++ w/tools/objtool/include/objtool/warn.h
> @@ -56,6 +56,7 @@ static inline char *offstr(struct section *sec, unsigned long offset)
>  })
> 
>  #define WARN_LIMIT 2
> +static_assert(WARN_LIMIT < (1 << STRUCT_SYMBOL_WARNING_BITS), "symbol.warnings too small");
> 
>  #define WARN_INSN(insn, format, ...)                                   \
>  ({
> 
> 
> Or just drop the bitfield (surely it can't be that important to save a
> byte here?) and use sizeof, e.g:

If I remember correctly, it is important. Josh and Peter spent quite some 
time on optimizing the data structures in the past as every bit counts.
 
> BUILD_BUG_ON(ilog2(WARN_LIMIT) > sizeof(_insn->sym->warnings));

Miroslav