GCC 4.6.x miscompiling arm-linux?

[david.jander@protonic.nl (David Jander)]

On Tue, 11 Sep 2012 16:10:14 +0200
Mikael Pettersson <mikpe@it.uu.se> wrote:

> David Jander writes:
>  > On Tue, 11 Sep 2012 14:53:35 +0200
>  > Mikael Pettersson <mikpe@it.uu.se> wrote:
>  > 
>  > > David Jander writes:
>  > >  > On Tue, 11 Sep 2012 13:35:40 +0200
>  > >  > Mikael Pettersson <mikpe@it.uu.se> wrote:
>  > >  > 
>  > >  > > David Jander writes:
>  > >  > >  > > I can make the patches available if there's confirmation that a vanilla
>  > >  > >  > > upstream gcc-4.6.3 doesn't work.
>  > >  > >  > 
>  > >  > >  > I am pretty sure this is the case... do you have a patch series that you can
>  > >  > >  > easily tar and mail to me? I'd like to try those patches with OSELAS, to see
>  > >  > >  > if I can indeed build a gcc-4.6.3 toolchain that generates correct code.... I
>  > >  > >  > already know that I can generate one that doesn't ;-)
>  > >  > >  > 
>  > >  > > ...
>  > >  > >  > > If you're sure no add-on patches were applied, then yes please do.
>  > >  > >  > 
>  > >  > >  > Pretty sure, but not 100%, so I'd like to try your patches first if you don't
>  > >  > >  > mind....
>  > >  > > 
>  > >  > > And I'd like to confirm first. Please tell us the following:
>  > >  > > 
>  > >  > >  > >  > > 2: include the output of gcc -v which tells how that gcc was configured,
>  > >  > 
>  > >  > Using built-in specs.
>  > >  > COLLECT_GCC=/opt/OSELAS.Toolchain-2011.11.1/arm-v5te-linux-gnueabi/gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/bin/arm-v5te-linux-gnueabi-gcc-4.6.3
>  > >  > COLLECT_LTO_WRAPPER=/opt/OSELAS.Toolchain-2011.11.1/arm-v5te-linux-gnueabi/gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/libexec/gcc/arm-v5te-linux-gnueabi/4.6.3/lto-wrapper
>  > >  > Target: arm-v5te-linux-gnueabi
>  > >  > Configured with: /home/djander/ptxdist/OSELAS.Toolchain-2011.11.1/platform-arm-v5te-linux-gnueabi-gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/build-cross/gcc-4.6.3/configure --target=arm-v5te-linux-gnueabi --with-sysroot=/opt/OSELAS.Toolchain-2011.11.1/arm-v5te-linux-gnueabi/gcc-4.6.3
>  > >  > -glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/sysroot-arm-v5te-linux-gnueabi --disable-multilib --with-float=soft --with-fpu=vfp --with-cpu=arm926ej-s --enable-__cxa_atexit --disable-sjlj-exceptions --disable-nls --disable-decimal-float --disable-fixed-point --disable-win32-registry --enable-symvers=gnu --with-pkgversion=OSELAS.Toolchain-2011.11.1 --with-system-zlib --with-gmp=/home/djander/ptxdist/OSELAS.Toolchain-2011.11.1/platform-arm-v5te
>  > >  > -linux-gnueabi-gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/sysroot-host --with-mpfr=/home/djander/ptxdist/OSELAS.Toolchain-2011.11.1/platform-arm-v5te-linux-gnueabi-gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/sysroot-host --prefix=/opt/OSELAS.Toolchain-2011.11.1/arm-v5te-linux-gnueabi/gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized --enable-languages=c,c++ --enable-threads=posix --enable-c99 --enable-long-long --enable-libstdcxx-debug --enable-profile --enable-shared --disable-libssp --enable-checking=release
>  > >  > Thread model: posix
>  > >  > gcc version 4.6.3 (OSELAS.Toolchain-2011.11.1)
>  > >  > 
>  > >  > >  > >  > > 3: give the exact set of gcc options used then compiling the test case.
>  > >  > 
>  > >  > If I type this in a terminal:
>  > >  > 
>  > >  > 
>  > >  > $ /opt/OSELAS.Toolchain-2011.11.1/arm-v5te-linux-gnueabi/gcc-4.6.3-glibc-2.14.1-binutils-2.21.1a-kernel-2.6.39-sanitized/bin/arm-v5te-linux-gnueabi-gcc-4.6.3 -Os -S -o - -x c -
>  > >  > 
>  > >  > struct flexcan_regs {
>  > >  >         unsigned int mcr;
>  > >  >         unsigned int rxfgmask;
>  > >  > };
>  > >  > 
>  > >  > #define flexcan_read(a) (*(volatile unsigned int *)(a))
>  > >  > #define flexcan_write(v,a)      (*(volatile unsigned int *)(a) = (v))
>  > >  > 
>  > >  > int flexcan_chip_start(int ver, struct flexcan_regs *regs)
>  > >  > {
>  > >  >         flexcan_write(0, &regs->mcr);
>  > >  > 
>  > >  >         if (ver >= 10)
>  > >  >                 flexcan_write(0, &regs->rxfgmask);
>  > >  > 
>  > >  >         return 0;
>  > >  > }
>  > >  > 
>  > >  > 
>  > >  > I get this output after hitting <CTRL-D>:
>  > >  > 
>  > >  >         .cpu arm926ej-s
>  > >  >         .fpu softvfp
>  > >  >         .eabi_attribute 20, 1
>  > >  >         .eabi_attribute 21, 1
>  > >  >         .eabi_attribute 23, 3
>  > >  >         .eabi_attribute 24, 1
>  > >  >         .eabi_attribute 25, 1
>  > >  >         .eabi_attribute 26, 2
>  > >  >         .eabi_attribute 30, 4
>  > >  >         .eabi_attribute 18, 4
>  > >  >         .file   ""
>  > >  >         .text
>  > >  >         .align  2
>  > >  >         .global flexcan_chip_start
>  > >  >         .type   flexcan_chip_start, %function
>  > >  > flexcan_chip_start:
>  > >  >         @ args = 0, pretend = 0, frame = 0
>  > >  >         @ frame_needed = 0, uses_anonymous_args = 0
>  > >  >         @ link register save eliminated.
>  > >  >         mov     r3, #0
>  > >  >         cmp     r0, #9
>  > >  >         str     r3, [r1, #0]
>  > >  >         ldrle   r3, [r1, #4]
>  > >  >         mov     r0, #0
>  > >  >         str     r3, [r1, #4]
>  > >  >         bx      lr
>  > >  >         .size   flexcan_chip_start, .-flexcan_chip_start
>  > >  >         .ident  "GCC: (OSELAS.Toolchain-2011.11.1) 4.6.3"
>  > >  >         .section        .note.GNU-stack,"",%progbits
>  > >  > 
>  > >  > 
>  > >  > Do you need more information?
>  > > 
>  > > No, I can reproduce the bug with vanilla gcc-4.6.3; vanilla 4.7.1 and 4.5.4 are Ok.
>  > > 
>  > > I'll bisect my 4.6.3 patch series to see which patch fixes it.
>  > 
>  > Great. Thanks a lot for your help so far. Looking forward to see what fixes
>  > this issue. Are you implying that you will also file the bug (and possible
>  > patch) with gcc.gnu.org, or do you prefer me to do that?
> 
> This is a known bug: <http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52445>.
> It was reported and fixed in gcc trunk on March 1 this year, but missed
> the gcc-4.6.3 release made the same day (and frozen a week or so before),
> and it hasn't been applied to gcc-4.6.4 branch yet either.
> 
> I've been using and testing the fix in my own gcc-4.6 branch since March 4
> without regressions. I'm attaching my backport of the fix below.
> 
> I'll ping gcc upstream about getting this into gcc-4.6.4.

Thanks a lot for this patch.
I can confirm that the bug is gone, and no other problems have appeared so
far. Successfully recompiled kernel and the entire userspace (ptxdist) with
the patched toolchain, and all seems well.

Best regards,

-- 
David Jander
Protonic Holland.