Re: [GIT pull - RFC] locking/kcsan for v5.8

[Will Deacon]

Hi folks,

On Tue, Jun 02, 2020 at 01:05:02PM -0000, Thomas Gleixner wrote:
> Linus,
> 
> please consider to pull the latest locking/kcsan branch from:
> 
>    git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git locking-kcsan-2020-06-02
> 
> up to:  f05e1242fbb2: compiler_types.h: Use unoptimized __unqual_scalar_typeof for sparse

Just in case it's not clear, this pull request also contains the READ_ONCE
rework I've been working on for a while, which bumps the minimum GCC version
and improves code-gen on arm64 when stack protector is enabled:

> Will Deacon (19):
>       sparc32: mm: Fix argument checking in __srmmu_get_nocache()
>       sparc32: mm: Restructure sparc32 MMU page-table layout
>       sparc32: mm: Change pgtable_t type to pte_t * instead of struct page *
>       sparc32: mm: Reduce allocation size for PMD and PTE tables
>       compiler/gcc: Raise minimum GCC version for kernel builds to 4.8
>       linux/compiler.h: Remove redundant '#else'
>       netfilter: Avoid assigning 'const' pointer to non-const pointer
>       net: tls: Avoid assigning 'const' pointer to non-const pointer
>       fault_inject: Don't rely on "return value" from WRITE_ONCE()
>       arm64: csum: Disable KASAN for do_csum()
>       READ_ONCE: Simplify implementations of {READ,WRITE}_ONCE()
>       READ_ONCE: Enforce atomicity for {READ,WRITE}_ONCE() memory accesses
>       READ_ONCE: Drop pointer qualifiers when reading from scalar types
>       locking/barriers: Use '__unqual_scalar_typeof' for load-acquire macros
>       arm64: barrier: Use '__unqual_scalar_typeof' for acquire/release macros
>       gcov: Remove old GCC 3.4 support
>       kcsan: Rework data_race() so that it can be used by READ_ONCE()
>       READ_ONCE: Use data_race() to avoid KCSAN instrumentation
>       READ_ONCE: Fix comment describing 2x32-bit atomicity

The reason it's bundled up with KCSAN is because of conflicts in linux-next
but the series is reasonably mature so it would be a shame if it missed 5.8.

On the off-chance that this lot doesn't make it (it is, after all, an RFC!),
I've stuck the READ_ONCE stuff on its own branch (see below), with some
cherry-picks of fixes that ended up in -tip. The sparc patches are all
sitting in sparc-next, so they'll come in via davem.

Obviously it would better if everything went upstream, but I'm just trying
to avoid throwing the baby out with the bathwater if KCSAN needs a little
longer to bake.

Cheers,

Will

--->8

The following changes since commit 8f3d9f354286745c751374f5f1fcafee6b3f3136:

  Linux 5.7-rc1 (2020-04-12 12:35:55 -0700)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/will/linux.git rwonce/rework

for you to fetch changes up to b398ace5d2ea0b7f00d9f1ce23c647e289c206ca:

  compiler_types.h: Use unoptimized __unqual_scalar_typeof for sparse (2020-06-05 11:40:01 +0100)

----------------------------------------------------------------
Marco Elver (2):
      compiler_types.h: Optimize __unqual_scalar_typeof compilation time
      compiler_types.h: Use unoptimized __unqual_scalar_typeof for sparse

Will Deacon (14):
      compiler/gcc: Raise minimum GCC version for kernel builds to 4.8
      netfilter: Avoid assigning 'const' pointer to non-const pointer
      net: tls: Avoid assigning 'const' pointer to non-const pointer
      fault_inject: Don't rely on "return value" from WRITE_ONCE()
      arm64: csum: Disable KASAN for do_csum()
      READ_ONCE: Simplify implementations of {READ,WRITE}_ONCE()
      READ_ONCE: Enforce atomicity for {READ,WRITE}_ONCE() memory accesses
      READ_ONCE: Drop pointer qualifiers when reading from scalar types
      locking/barriers: Use '__unqual_scalar_typeof' for load-acquire macros
      arm64: barrier: Use '__unqual_scalar_typeof' for acquire/release macros
      gcov: Remove old GCC 3.4 support
      READ_ONCE: Fix comment describing 2x32-bit atomicity
      compiler-types.h: Include naked type in __pick_integer_type() match
      compiler.h: Enforce that READ_ONCE_NOCHECK() access size is sizeof(long)

 Documentation/process/changes.rst |   2 +-
 arch/arm/crypto/Kconfig           |  12 +-
 arch/arm64/include/asm/barrier.h  |  16 +-
 arch/arm64/lib/csum.c             |  20 +-
 crypto/Kconfig                    |   1 -
 drivers/xen/time.c                |   2 +-
 include/asm-generic/barrier.h     |  16 +-
 include/linux/compiler-gcc.h      |   5 +-
 include/linux/compiler.h          | 148 +++++-----
 include/linux/compiler_types.h    |  47 ++++
 init/Kconfig                      |   1 -
 kernel/gcov/Kconfig               |  24 --
 kernel/gcov/Makefile              |   3 +-
 kernel/gcov/gcc_3_4.c             | 573 --------------------------------------
 lib/fault-inject.c                |   4 +-
 net/netfilter/core.c              |   2 +-
 net/tls/tls_main.c                |   2 +-
 scripts/gcc-plugins/Kconfig       |   2 +-
 18 files changed, 161 insertions(+), 719 deletions(-)
 delete mode 100644 kernel/gcov/gcc_3_4.c