11 Jun, 2020

2 commits

  • Currently instrumentation of atomic primitives is done at the architecture
    level, while composites or fallbacks are provided at the generic level.

    The result is that there are no uninstrumented variants of the
    fallbacks. Since there is now need of such variants to isolate text poke
    from any form of instrumentation invert this ordering.

    Doing this means moving the instrumentation into the generic code as
    well as having (for now) two variants of the fallbacks.

    Notes:

    - the various *cond_read* primitives are not proper fallbacks
    and got moved into linux/atomic.c. No arch_ variants are
    generated because the base primitives smp_cond_load*()
    are instrumented.

    - once all architectures are moved over to arch_atomic_ one of the
    fallback variants can be removed and some 2300 lines reclaimed.

    - atomic_{read,set}*() are no longer double-instrumented

    Reported-by: Thomas Gleixner
    Signed-off-by: Peter Zijlstra (Intel)
    Signed-off-by: Thomas Gleixner
    Acked-by: Mark Rutland
    Link: https://lkml.kernel.org/r/20200505134058.769149955@linutronix.de

    Peter Zijlstra
     
  • Use __always_inline for atomic fallback wrappers. When building for size
    (CC_OPTIMIZE_FOR_SIZE), some compilers appear to be less inclined to
    inline even relatively small static inline functions that are assumed to
    be inlinable such as atomic ops. This can cause problems, for example in
    UACCESS regions.

    While the fallback wrappers aren't pure wrappers, they are trivial
    nonetheless, and the function they wrap should determine the final
    inlining policy.

    For x86 tinyconfig we observe:
    - vmlinux baseline: 1315988
    - vmlinux with patch: 1315928 (-60 bytes)

    [ tglx: Cherry-picked from KCSAN ]

    Suggested-by: Mark Rutland
    Signed-off-by: Marco Elver
    Acked-by: Mark Rutland
    Signed-off-by: Paul E. McKenney
    Signed-off-by: Thomas Gleixner

    Marco Elver
     

01 Nov, 2018

2 commits

  • Mark all these scripts executable.

    Cc: Mark Rutland
    Cc: Peter Zijlstra (Intel)
    Cc: Will Deacon
    Cc: linux-arm-kernel@lists.infradead.org
    Cc: Catalin Marinas
    Cc: linuxdrivers@attotech.com
    Cc: dvyukov@google.com
    Cc: boqun.feng@gmail.com
    Cc: arnd@arndb.de
    Cc: aryabinin@virtuozzo.com
    Cc: glider@google.com
    Cc: linux-kernel@vger.kernel.org
    Signed-off-by: Ingo Molnar

    Ingo Molnar
     
  • To minimize repetition, to allow for future rework, and to ensure
    regularity of the various atomic APIs, we'd like to automatically
    generate (the bulk of) a number of headers related to atomics.

    This patch adds the infrastructure to do so, leaving actual conversion
    of headers to subsequent patches. This infrastructure consists of:

    * atomics.tbl - a table describing the functions in the atomics API,
    with names, prototypes, and metadata describing the variants that
    exist (e.g fetch/return, acquire/release/relaxed). Note that the
    return type is dependent on the particular variant.

    * atomic-tbl.sh - a library of routines useful for dealing with
    atomics.tbl (e.g. querying which variants exist, or generating
    argument/parameter lists for a given function variant).

    * gen-atomic-fallback.sh - a script which generates a header of
    fallbacks, covering cases where architecture omit certain functions
    (e.g. omitting relaxed variants).

    * gen-atomic-long.sh - a script which generates wrappers providing the
    atomic_long API atomic of the relevant atomic or atomic64 API,
    ensuring the APIs are consistent.

    * gen-atomic-instrumented.sh - a script which generates atomic* wrappers
    atop of arch_atomic* functions, with automatically generated KASAN
    instrumentation.

    * fallbacks/* - a set of fallback implementations for atomics, which
    should be used when no implementation of a given atomic is provided.
    These are used by gen-atomic-fallback.sh to generate fallbacks, and
    these are also used by other scripts to determine the set of optional
    atomics (as required to generate preprocessor guards correctly).

    Fallbacks may use the following variables:

    ${atomic} atomic prefix: atomic/atomic64/atomic_long, which can be
    used to derive the atomic type, and to prefix functions

    ${int} integer type: int/s64/long

    ${pfx} variant prefix, e.g. fetch_

    ${name} base function name, e.g. add

    ${sfx} variant suffix, e.g. _return

    ${order} order suffix, e.g. _relaxed

    ${atomicname} full name, e.g. atomic64_fetch_add_relaxed

    ${ret} return type of the function, e.g. void

    ${retstmt} a return statement (with a trailing space), unless the
    variant returns void

    ${params} parameter list for the function declaration, e.g.
    "int i, atomic_t *v"

    ${args} argument list for invoking the function, e.g. "i, v"

    ... for clarity, ${ret}, ${retstmt}, ${params}, and ${args} are
    open-coded for fallbacks where these do not vary, or are critical to
    understanding the logic of the fallback.

    The MAINTAINERS entry for the atomic infrastructure is updated to cover
    the new scripts.

    There should be no functional change as a result of this patch.

    Signed-off-by: Mark Rutland
    Signed-off-by: Peter Zijlstra (Intel)
    Cc: linux-arm-kernel@lists.infradead.org
    Cc: catalin.marinas@arm.com
    Cc: Will Deacon
    Cc: linuxdrivers@attotech.com
    Cc: dvyukov@google.com
    Cc: Boqun Feng
    Cc: arnd@arndb.de
    Cc: aryabinin@virtuozzo.com
    Cc: glider@google.com
    Link: http://lkml.kernel.org/r/20180904104830.2975-2-mark.rutland@arm.com
    Signed-off-by: Ingo Molnar

    Mark Rutland