Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Thu, 14 Jan 2016 17:51:16 -0500
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: string word-at-a-time and atomic.h FAQ on twitter

On Wed, Jan 13, 2016 at 06:30:50PM +0100, Szabolcs Nagy wrote:
> i tried to do it with -fno-lto
> 
> but it seems gcc-6 miscompiles musl with -flto anyway:
> 
> lto incorrectly dead code eliminates _dlstart_c.
> (the libc entry point, _dlstart, is defined in toplevel
> inline asm in ldso/dlstart.c and it jumps to _dlstart_c)

I think the files that receive "crt" treatment (-DCRT?) should be
hard-coded as no-LTO. Even if it weren't for this issue, there are
also potential issues with optimizations that might move code across
relocations and initialization. At the very least there should be a
full asm barrier between such init code and the rest of the program.

> lto breaks symbol binding for environ, _environ, ___environ.
> (they should be weak, without that environ in a main binary
> has different address than in libc.so)
> 
> libc.so built with -flto:
> $ readelf --dyn-syms -W libc.so |grep envi
>     22: 000000000028eb90     8 OBJECT  GLOBAL DEFAULT   15 __environ
>    398: 000000000028eb90     8 OBJECT  GLOBAL PROTECTED   15 ___environ
>   1034: 000000000028eb90     8 OBJECT  GLOBAL PROTECTED   15 _environ
>   1107: 000000000028eb90     8 OBJECT  GLOBAL DEFAULT   15 environ
> 
> libc.so without -flto:
> $ readelf --dyn-syms -W libc.so |grep envi
>     22: 000000000028d2d8     8 OBJECT  GLOBAL DEFAULT   15 __environ
>    398: 000000000028d2d8     8 OBJECT  WEAK   PROTECTED   15 ___environ
>   1034: 000000000028d2d8     8 OBJECT  WEAK   PROTECTED   15 _environ
>   1107: 000000000028d2d8     8 OBJECT  WEAK   DEFAULT   15 environ

This is purely a gcc bug and indicates LTO is not really ready for use
with libc, which means I don't see much sense in trying to support it
yet. Of course that doesn't mean I'm disinterested in fixing actual C
language usage errors (aliasing, etc.) that might break with LTO.
Rather I'm disinterested in LTO-specific hacks for working around bugs
in GCC's LTO. OTOH your patch is nice from the standpoint of being
able to test LTO and find and report the bugs.

Rich

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.