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  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]
Date:   Mon, 18 Mar 2019 16:44:03 -0700
From:   hpa@...or.com
To:     Matthias Kaehlcke <mka@...omium.org>
CC:     Andy Lutomirski <luto@...nel.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        x86@...nel.org, linux-kernel@...r.kernel.org,
        Nick Desaulniers <ndesaulniers@...gle.com>,
        Manoj Gupta <manojgupta@...omium.org>,
        Tiancong Wang <tcwang@...omium.org>,
        Stephen Hines <srhines@...gle.com>,
        clang-built-linux@...glegroups.com
Subject: Re: [PATCH] lib: Add shared copy of __lshrti3 from libgcc

On March 18, 2019 3:16:39 PM PDT, Matthias Kaehlcke <mka@...omium.org> wrote:
>On Mon, Mar 18, 2019 at 02:50:44PM -0700, hpa@...or.com wrote:
>> On March 18, 2019 2:31:13 PM PDT, Matthias Kaehlcke
><mka@...omium.org> wrote:
>> >On Fri, Mar 15, 2019 at 01:54:50PM -0700, Matthias Kaehlcke wrote:
>> >> The compiler may emit calls to __lshrti3 from the compiler runtime
>> >> library, which results in undefined references:
>> >> 
>> >> arch/x86/kvm/x86.o: In function `mul_u64_u64_shr':
>> >>   include/linux/math64.h:186: undefined reference to `__lshrti3'
>> >> 
>> >> Add a copy of the __lshrti3 libgcc routine (from gcc v4.9.2).
>> >> 
>> >> Include the function for x86 builds with clang, which is the
>> >> environment where the above error was observed.
>> >> 
>> >> Signed-off-by: Matthias Kaehlcke <mka@...omium.org>
>> >
>> >With "Revert "kbuild: use -Oz instead of -Os when using clang"
>> >(https://lore.kernel.org/patchwork/patch/1051932/) the above
>> >error is fixed, a few comments inline for if the patch is
>> >resurrected in the future because __lshrti3 is emitted in a
>> >different context.
>> >
>> >> diff --git a/include/linux/libgcc.h b/include/linux/libgcc.h
>> >> index 32e1e0f4b2d0..a71036471838 100644
>> >> --- a/include/linux/libgcc.h
>> >> +++ b/include/linux/libgcc.h
>> >> @@ -22,15 +22,26 @@
>> >>  #include <asm/byteorder.h>
>> >> 
>> >>  typedef int word_type __attribute__ ((mode (__word__)));
>> >> +typedef int TItype __attribute__ ((mode (TI)));
>> >
>> >Consider using __int128 instead. Definition and use need a
>> >'defined(__SIZEOF_INT128__)' guard  (similar for mode (TI)), since
>> >these 128 bit types aren't supported on all platforms.
>> >
>> >>  #ifdef __BIG_ENDIAN
>> >>  struct DWstruct {
>> >>  	int high, low;
>> >>  };
>> >> +
>> >> +struct DWstruct128 {
>> >> +	long long high, low;
>> >> +};
>> >
>> >This struct isn't needed, struct DWstruct can be used.
>> >
>> >> diff --git a/lib/lshrti3.c b/lib/lshrti3.c
>> >> new file mode 100644
>> >> index 000000000000..2d2123bb3030
>> >> --- /dev/null
>> >> +++ b/lib/lshrti3.c
>> >> @@ -0,0 +1,31 @@
>> >> +// SPDX-License-Identifier: GPL-2.0
>> >> +
>> >> +#include <linux/export.h>
>> >> +#include <linux/libgcc.h>
>> >> +
>> >> +long long __lshrti3(long long u, word_type b)
>> >
>> >use TItype for input/output, which is what gcc does, though the
>above
>> >matches the interface in the documentation.
>> >
>> >> +{
>> >> +	DWunion128 uu, w;
>> >> +	word_type bm;
>> >> +
>> >> +	if (b == 0)
>> >> +		return u;
>> >> +
>> >> +	uu.ll = u;
>> >> +	bm = 64 - b;
>> >> +
>> >> +	if (bm <= 0) {
>> >> +		w.s.high = 0;
>> >> +		w.s.low = (unsigned long long) uu.s.high >> -bm;
>> >
>> >include <linux/types.h> and use u64 instead of unsigned long long.
>> 
>> Ok, now I'm really puzzled.
>> 
>> How could we need a 128-bit shift when the prototype only has 64 bits
>of input?!
>
>Good question, this is the code from libgcc:
>
>TItype
>__lshrti3 (TItype u, shift_count_type b)
>{
>  if (b == 0)
>    return u;
>
>  const DWunion uu = {.ll = u};
>  const shift_count_type bm = (8 * (8)) - b;
>  DWunion w;
>
>  if (bm <= 0)
>    {
>      w.s.high = 0;
>      w.s.low = (UDItype) uu.s.high >> -bm;
>    }
>  else
>    {
>      const UDItype carries = (UDItype) uu.s.high << bm;
>
>      w.s.high = (UDItype) uu.s.high >> b;
>      w.s.low = ((UDItype) uu.s.low >> b) | carries;
>    }
>
>  return w.ll;
>}
>
>
>My compiler knowledge is limited, my guess is that the function is a
>generic implementation, and while a long long is 64-bit wide under
>Linux it could be 128-bit on other platforms.

Yes, long long is just plain wrong.

How could we end up calling this function on 32 bits?!
-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.

Powered by blists - more mailing lists