[U-Boot] [PATCH 5/5] lib: Enable private libgcc by default

[Marek Vasut <marex@denx.de>]

On 03/24/2016 07:43 PM, Sergey Kubushyn wrote:
> On Thu, 24 Mar 2016, Sergey Kubushyn wrote:
> 
>> On Thu, 24 Mar 2016, Marek Vasut wrote:
>>
>>>  On 03/24/2016 12:54 AM, Sergey Kubushyn wrote:
>>> >  On Thu, 24 Mar 2016, Marek Vasut wrote:
>>> > > >  On 03/24/2016 12:47 AM, Sergey Kubushyn wrote:
>>> > > >  On Thu, 24 Mar 2016, Marek Vasut wrote:
>>> > > > > > > >  On 03/24/2016 12:08 AM, Tom Rini wrote:
>>> > > > > >  On Wed, Mar 23, 2016 at 04:02:07PM -0700, Sergey Kubushyn
>>> wrote:
>>> > > > > > >  On Wed, 23 Mar 2016, Tom Rini wrote:
>>> > > > > > > > > > > > > >  On Wed, Mar 23, 2016 at 06:08:45PM +0100,
>>> Albert ARIBAUD > > > > > > >  wrote:
>>> > > > > > > > >  Hello Tom,
>>> > > > > > > > > > > > > > > > >  On Wed, 23 Mar 2016 09:22:38 -0400,
>>> Tom Rini > > > > > > > >  <trini@konsulko.com>
>>> > > > > > > > >  wrote:
>>> > > > > > > > > >  On Wed, Mar 23, 2016 at 01:53:35PM +0100, Albert
>>> ARIBAUD > > > > > > > > >  wrote:
>>> > > > > > > > > > >  Hello Marek,
>>> > > > > > > > > > > > > > > > > > > > >  On Sun, 20 Mar 2016 17:15:34
>>> +0100, Marek Vasut > > > > > > > > > >  <marex@denx.de>
>>> > > > > > > > > > >  wrote:
>>> > > > > > > > > > > >  This patch decouples U-Boot binary from the >
>>> > > > > > > > > > >  toolchain on
>>> > > > > > > > > > > >  systems where
>>> > > > > > > > > > > >  private libgcc is available. Instead of
>>> pulling in > > > > > > > > > > >  functions
>>> > > > > > > > > > > >  provided
>>> > > > > > > > > > > >  by the libgcc from the toolchain, U-Boot will
>>> use > > > > > > > > > > >  it's own set
>>> > > > > > > > > > > >  of libgcc
>>> > > > > > > > > > > >  functions. These functions are usually
>>> imported from > > > > > > > > > > >  Linux
>>> > > > > > > > > > > >  kernel, which
>>> > > > > > > > > > > >  also uses it's own libgcc functions instead of
>>> the > > > > > > > > > > >  ones
>>> > > > > > > > > > > >  provided by the
>>> > > > > > > > > > > >  toolchain.
>>> > > > > > > > > > > > > > > > > > > > > > >  This patch solves a
>>> rather common problem. The > > > > > > > > > > >  toolchain can
>>> > > > > > > > > > > >  usually
>>> > > > > > > > > > > >  generate code for many variants of target > >
>>> > > > > > > > > >  architecture and
>>> > > > > > > > > > > >  often even
>>> > > > > > > > > > > >  different endianness. The libgcc on the other
>>> hand > > > > > > > > > > >  is usually
>>> > > > > > > > > > > >  compiled
>>> > > > > > > > > > > >  for one particular configuration and the
>>> functions > > > > > > > > > > >  provided by
>>> > > > > > > > > > > >  it may
>>> > > > > > > > > > > >  or may not be suited for use in U-Boot. This
>>> can > > > > > > > > > > >  manifest in
>>> > > > > > > > > > > >  two ways,
>>> > > > > > > > > > > >  either the U-Boot fails to compile altogether
>>> and > > > > > > > > > > >  linker will
>>> > > > > > > > > > > >  complain
>>> > > > > > > > > > > >  or, in the much worse case, the resulting
>>> U-Boot > > > > > > > > > > >  will build,
>>> > > > > > > > > > > >  but will
>>> > > > > > > > > > > >  misbehave in very subtle and hard to debug ways.
>>> > > > > > > > > > > > > > > > > > > > >  I don't think using private
>>> libgcc by default is a > > > > > > > > > >  good idea.
>>> > > > > > > > > > > > > > > > > > > > >  U-Boot's private libgcc is
>>> not a feature of U-Boot, > > > > > > > > > >  but a fix
>>> > > > > > > > > > >  for some
>>> > > > > > > > > > >  cases where a target cannot properly link with
>>> the > > > > > > > > > >  libgcc
>>> > > > > > > > > > >  provided by
>>> > > > > > > > > > >  the (specific release of the) GCC toolchain in
>>> use. > > > > > > > > > >  Using
>>> > > > > > > > > > >  private libgcc
>>> > > > > > > > > > >  to other cases than these does not fix or
>>> improve > > > > > > > > > >  anything; those
>>> > > > > > > > > > >  other cases were working and did not require any
>>> fix > > > > > > > > > >  in this
>>> > > > > > > > > > >  respect.
>>> > > > > > > > > > > > > > > > > > >  This isn't true, exactly.  If
>>> using clang for example > > > > > > > > >  everyone
>>> > > > > > > > > >  needs to
>>> > > > > > > > > >  enable this code.  We're also using -fno-builtin >
>>> > > > > > > > >  -ffreestanding
>>> > > > > > > > > >  which
>>> > > > > > > > > >  should limit the amount of interference from the >
>>> > > > > > > > >  toolchain.  And
>>> > > > > > > > > >  we get
>>> > > > > > > > > >  that.
>>> > > > > > > > > > > > > > > > >  You mean clang does not produce
>>> self-sustained binaries?
>>> > > > > > > > > > > > > > >  clang does not provide "libgcc", so
>>> there's no -lgcc > > > > > > >  providing
>>> > > > > > > >  all of
>>> > > > > > > >  the functions that are (today) in:
>>> > > > > > > >  _ashldi3.S _ashrdi3.S _divsi3.S  _lshrdi3.S _modsi3.S
>>> > > > > > > >  _udivsi3.S
>>> > > > > > > >  _umodsi3.S div0.S  _uldivmod.S
>>> > > > > > > >  which aside from __modsi3 and __umodsi3 are all
>>> __aeabi_xxx
>>> > > > > > > > > > > > >  There is also _udivmoddi4 pulled from libgcc
>>> for 64-bit > > > > > >  division
>>> > > > > > >  since we
>>> > > > > > >  switched to 64-bit all around ARM. It comes from clock
>>> > > > > > >  calculations for
>>> > > > > > >  video, e.g. from drivers/video/ipu_common.c for i.MX6.
>>> > > > > > > > > > >  Well, this is an example of why we both don't
>>> want libgcc ever > > > > >  nor
>>> > > > > >  do we
>>> > > > > >  want to overly expand what we do offer.  In this case
>>> isn't it > > > > >  an
>>> > > > > >  example of something that should be using lldiv/do_div/etc?
>>> > > > > > > > >  I haven't seen the _udivmoddi4 emitted in my tests.
>>> Linux's libgcc > > > >  copy
>>> > > > >  also doesn't implement the function. Which toolchain do you
>>> use > > > >  and
>>> > > > >  which target did you compile?
>>> > > > > > >  I'm using my own armv7hl-linux-gnueabi toolchain built
>>> for hard > > >  float.
>>> > > >  Linux
>>> > > >  arm libgcc does have arch/arm/lib/div64.S file that provides
>>> > > >  __do_div64()
>>> > > >  function that is used by do_div() from include/asm/div64.h for
>>> > > >  32-bit
>>> > > >  ARM
>>> > > >  platform. Sure, arm64 has neither div64.h nor div64.S. We _DO_
>>> have
>>> > > >  div64.h
>>> > > >  (that is totally different from what Linux provides) but no
>>> div64.S > > >  in
>>> > > >  arch/arm/lib.
>>> > > > >  In that case, we should just import div64.S from Linux on
>>> arm32 and be
>>> > >  done with it ? Since we now have all the necessary macros thanks
>>> to > >  the
>>> > >  first four patches in this series, that should be trivial.
>>> > > > >  What do you think? I can bake a patch real quick, so you can
>>> test it ?
>>> > >  Sure I'll test it, no problems. Just bake the patch :)
>>>
>>>  Done, give it a go please.
>>
>> OK, it didn't work, _udivmoddi4.o is still being pulled from libgcc. I'm
>> analyzing it right now, will come up with more later today.
> 
> OK, it requires a CONFIG_USE_PRIVATE_LIBGCC defined to use private libgcc,
> my bad -- thought it would be automatic. Having that defined makes build
> fail complaining about assembly syntax in div64.S:
> 
> === Cut ===
> arch/arm/lib/div64.S: Assembler messages:
> arch/arm/lib/div64.S:185: Error: bad instruction `arm( orr r2,r2,r1,lsl
> ip)'
> arch/arm/lib/div64.S:186: Error: bad instruction `thumb( lsl r1,r1,ip)'
> arch/arm/lib/div64.S:187: Error: bad instruction `thumb( orr r2,r2,r1)'
> scripts/Makefile.build:316: recipe for target 'arch/arm/lib/div64.o' failed
> make[1]: *** [arch/arm/lib/div64.o] Error 1
> Makefile:1214: recipe for target 'arch/arm/lib' failed
> make: *** [arch/arm/lib] Error 2
> === Cut ===
> 
> Probably something is missing in div64.h? The Linux one is totally
> different. Digging in right now...

Are you building the stuff with all of these 5+1 patches ?

Best regards,
Marek Vasut