[Bug target/104656] New: [12 Regression] trunk 20220222 ftbfs for bpf

[Bug target/104656] New: [12 Regression] trunk 20220222 ftbfs for bpf

[doko at debian dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

            Bug ID: 104656
           Summary: [12 Regression] trunk 20220222 ftbfs for bpf
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: doko at debian dot org
  Target Milestone: ---

trunk 20220222 fails to build for the bpf target:

../../../gcc-source/libgcc/libgcov-driver.c: In function ‘write_topn_counters’:
../../../gcc-source/libgcc/libgcov-driver.c:418:1: error: too many function
arguments for eBPF
  418 | write_topn_counters (const struct gcov_ctr_info *ci_ptr,
      | ^~~~~~~~~~~~~~~~~~~
make[3]: *** [Makefile:928: _gcov_info_to_gcda.o] Error 1

[Bug gcov-profile/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |12.0

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|gcov-profile                |target

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I guess you need to use --disable-gcov when building BPF - maybe that can
somehow be done by default?

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[doko at debian dot org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

--- Comment #2 from Matthias Klose <doko at debian dot org> ---
yes, adding that option lets the build succeed.

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The 5 words of arguments limitation in BPF is very severe, shouldn't the
backend do something smarter in that case and say pass in that case the
arguments but the first 4 words e.g. by invisible reference (push all the
remaining arguments to stack and pass address of the first such argument as a
pointer)?

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The ABI from the Linux kernel side needs to be documented that way.
So someone would need to propose that ABI change.

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[jose.marchesi at oracle dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

Jose E. Marchesi <jose.marchesi at oracle dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jose.marchesi at oracle dot com

--- Comment #5 from Jose E. Marchesi <jose.marchesi at oracle dot com> ---
Yes that would require changing the kernel ABI.  Also, there is the kernel
verifier to deal with: stack accesses are tracked at load-time and they should
be done using certain registers and following certain rules.

For the time being, --disabling-gcov when building for BPF seems like a
reasonable thing to do.  Will write and send a patch in that effect.

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

--- Comment #6 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jose E. Marchesi <jemarch@gcc.gnu.org>:

https://gcc.gnu.org/g:39be73d07bb667fd5d313a3398f5399e091f2c0c

commit r12-7373-g39be73d07bb667fd5d313a3398f5399e091f2c0c
Author: Jose E. Marchesi <jose.marchesi@oracle.com>
Date:   Wed Feb 23 20:48:30 2022 +0100

    bpf: do not --enable-gcov for bpf-*-* targets

    This patch changes the build machinery in order to disable the build
    of GCOV (both compiler and libgcc) in bpf-*-* targets.  The reason for
    this change is that BPF is (currently) too restricted in order to
    support the coverage instrumentalization.

    Tested in bpf-unknown-none and x86_64-linux-gnu targets.

    2022-02-23  Jose E. Marchesi  <jose.marchesi@oracle.com>

    gcc/ChangeLog

            PR target/104656
            * configure.ac: --disable-gcov if targetting bpf-*.
            * configure: Regenerate.

    libgcc/ChangeLog

            PR target/104656
            * configure.ac: --disable-gcov if targetting bpf-*.
            * configure: Regenerate.

[Bug target/104656] [12 Regression] trunk 20220222 ftbfs for bpf

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104656

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Fixed.