linux-kernel - Re: v4.10: kernel stack frame pointer .. has bad value (null)

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: <20170306163807.GA20689@amd>
Date:   Mon, 6 Mar 2017 17:38:07 +0100
From:   Pavel Machek <pavel@....cz>
To:     Josh Poimboeuf <jpoimboe@...hat.com>
Cc:     kernel list <linux-kernel@...r.kernel.org>, mingo@...nel.org,
        luto@...nel.org, bp@...en8.de, brgerst@...il.com,
        dvlasenk@...hat.com, hpa@...or.com, torvalds@...ux-foundation.org,
        peterz@...radead.org, tglx@...utronix.de
Subject: Re: v4.10: kernel stack frame pointer .. has bad value (null)

On Thu 2017-03-02 17:45:14, Josh Poimboeuf wrote:
> On Fri, Feb 24, 2017 at 11:04:39PM -0600, Josh Poimboeuf wrote:
> > On Thu, Feb 23, 2017 at 09:10:39PM +0100, Pavel Machek wrote:
> > > Hi!
> > > 
> > > 
> > > > > > > Somehow, startup_32_smp() is on the stack twice.  The stack unwind led
> > > > > > > to the startup_32_smp() frame at 0xf50cdf9c rather than the one at
> > > > > > > 0xf50cdfa8 (which is where it should normally be).  So the question is
> > > > > > > how startup_32_smp() got executed the second time, with the wrong stack
> > > > > > > offset.
> > > > > > 
> > > > > > Not much idea... but this is stack dump, right? Just because some
> > > > > > value is on the stack does not mean it is a return address, no?
> > > > > 
> > > > > Right, but the one at 0xf50cdfa8 is where the startup_32_smp() is
> > > > > *supposed* to be.  If the unwinder had unwinded to that one, it wouldn't
> > > > > have complained.  So it looks to me like the CPU somehow booted twice:
> > > > > the first time at the right stack address, and the second time it
> > > > > somehow ended up with a different stack address.
> > > > > 
> > > > > > And .... startup_32_smp is kind of "interesting" function. Take a
> > > > > > look...
> > > > > 
> > > > > Yes, it's used in bringing up the CPU.
> > > > 
> > > > Can you share your .config?  
> > > 
> > > Here you go...
> > 
> > What version of gcc are you using?
> > 
> > Can you post a disassembly of the first 10 instructions of
> > start_secondary()?
> 
> Pavel, ping?  I'd like to try to get to the bottom of this issue soon.
> 
> I asked for the gcc version and the disassembly of start_secondary()
> because I suspect gcc may have done a funky stack alignment prologue
> which copies the return address on the stack a second time after
> aligning it.

Sorry for the delay. This is on v4.11-rc1, but that should be similar.

pavel@duo:~$ gcc --version
gcc (Debian 4.9.2-10) 4.9.2

And here's the disassemble:

c402d200 <start_secondary>:
c402d200:       57                      push   %edi
c402d201:       8d 7c 24 08             lea    0x8(%esp),%edi
c402d205:       83 e4 f8                and    $0xfffffff8,%esp
c402d208:       ff 77 fc                pushl  -0x4(%edi)
c402d20b:       55                      push   %ebp
c402d20c:       89 e5                   mov    %esp,%ebp
c402d20e:       57                      push   %edi
c402d20f:       56                      push   %esi
c402d210:       83 ec 10                sub    $0x10,%esp
c402d213:       e8 78 78 ff ff          call   c4024a90 <cpu_init>
c402d218:       ff 15 d0 d7 0c c5       call   *0xc50cd7d0
c402d21e:       8b 15 00 53 05 c5       mov    0xc5055300,%edx
c402d224:       8d 75 e8                lea    -0x18(%ebp),%esi
c402d227:       64 a1 f4 c0 1d c5       mov    %fs:0xc51dc0f4,%eax
c402d22d:       89 45 e8                mov    %eax,-0x18(%ebp)
c402d230:       b8 20 00 00 00          mov    $0x20,%eax
c402d235:       ff 52 78                call   *0x78(%edx)
c402d238:       8b 15 00 53 05 c5       mov    0xc5055300,%edx
c402d23e:       ff 52 4c                call   *0x4c(%edx)
c402d241:       e8 ea 2c 00 00          call   c402ff30
<apic_ap_setup>
c402d246:       8b 45 e8                mov    -0x18(%ebp),%eax
c402d249:       e8 42 fb ff ff          call   c402cd90
<smp_store_cpu_info>
c402d24e:       e8 5d 37 fd ff          call   c40009b0
<calibrate_delay>
c402d253:       8b 55 e8                mov    -0x18(%ebp),%edx
c402d256:       b8 00 c0 1d c5          mov    $0xc51dc000,%eax
c402d25b:       8b 0d 88 d6 0b c5       mov    0xc50bd688,%ecx
c402d261:       f6 05 fa fc 13 c5 04    testb  $0x4,0xc513fcfa
c402d268:       8b 14 95 20 52 05 c5    mov
-0x3afaade0(,%edx,4),%edx
c402d26f:       89 8c 10 c4 00 00 00    mov    %ecx,0xc4(%eax,%edx,1)
c402d276:       0f 85 24 01 00 00       jne    c402d3a0
<start_secondary+0x1a0>
c402d27c:       64 a1 f4 c0 1d c5       mov    %fs:0xc51dc0f4,%eax
c402d282:       e8 49 fb ff ff          call   c402cdd0
<set_cpu_sibling_map>

Let me know if I should go back to v4.10 and retry.

Best regards,
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Download attachment "signature.asc" of type "application/pgp-signature" (182 bytes)