Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 18 Jul 2018 16:50:57 -0400
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: Re: Segmentation fault in static binaries built with
 recent binutils

On Wed, Jul 18, 2018 at 10:19:28PM +0200, Szabolcs Nagy wrote:
> * Szabolcs Nagy <nsz@...t70.net> [2018-07-18 21:38:34 +0200]:
> > seems like another musl-gcc wrapper issue, if i do the linking
> > manually then i get a working binary, havent yet figured out why
> > 
> > manual linking:
> > 
> > Program Headers:
> >   Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz   Flg Align
> >   LOAD           0x000000 0x0000000000400000 0x0000000000400000 0x0001ec 0x0001ec R   0x1000
> >   LOAD           0x001000 0x0000000000401000 0x0000000000401000 0x0005a4 0x0005a4 R E 0x1000
> >   LOAD           0x002000 0x0000000000402000 0x0000000000402000 0x00004c 0x00004c R   0x1000
> >   LOAD           0x002ff0 0x0000000000403ff0 0x0000000000403ff0 0x000018 0x0002a8 RW  0x1000
> >   NOTE           0x0001c8 0x00000000004001c8 0x00000000004001c8 0x000024 0x000024 R   0x4
> >   GNU_STACK      0x000000 0x0000000000000000 0x0000000000000000 0x000000 0x000000 RW  0x10
> >   GNU_RELRO      0x002ff0 0x0000000000403ff0 0x0000000000403ff0 0x000010 0x000010 R   0x1
> > 
> > musl-gcc linking:
> > 
> > Program Headers:
> >   Type           Offset   VirtAddr           PhysAddr           FileSiz  MemSiz   Flg Align
> >   LOAD           0x001000 0x0000000000401000 0x0000000000401000 0x0005a4 0x0005a4 R E 0x1000
> >   LOAD           0x002000 0x0000000000402000 0x0000000000402000 0x00004c 0x00004c R   0x1000
> >   LOAD           0x002ff0 0x0000000000403ff0 0x0000000000403ff0 0x000018 0x0002a8 RW  0x1000
> >   GNU_STACK      0x000000 0x0000000000000000 0x0000000000000000 0x000000 0x000000 RW  0x10
> >   GNU_RELRO      0x002ff0 0x0000000000403ff0 0x0000000000403ff0 0x000010 0x000010 R   0x1
> 
> the difference between the two cases was --build-id
> 
> --build-id=sha1 works, --build-id=none segfaults
> 
> i assume the note section with the build id happens
> to force ld to keep the initial load segment, but
> that should be there without any note section, so
> it's likely a binutils bug (i see it on 2.30 and
> master branch too)

So am I understanding correctly that binutils ld is considering the
program headers section not to be something that has to live in loaded
memory? And it's putting a whole 4k of padding below the first load
segment? These both seem to be significant bugs.

Rich

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ