Zong Li <z...@andestech.com> 於 2018年8月15日 週三 下午5:35寫道:
>
> This patch set implements the ftrace function and function graph tracer
> on NDS32 architecture. We try to use C code to do everything, so we don't
> need to separate the implementation to assembly code and C code, we just
> need one ftrace.c file.
>
> In mcount function, we use the prologue/epilogue which generated by compiler
> to help us to save/restore the status. On the other hand, we use naked
> attribute of gcc to avoid to generate the prologue/epilogue when we don't
> need it(such as ftrace_stub and return_to_handler).
>
> We pass the following tests:
>
> 1. FTRACE_STARTUP_TEST: PASS
>
> 2. tools/testing/selftests/ftrace/:
>  # of passed:  28
>  # of failed:  1
>  # of unresolved:  0
>  # of untested:  0
>  # of unsupported:  38
>  # of xfailed:  0
>  # of undefined(test bug):  0
>
> The failed case is func-filter-glob.tc.
> When setting the '*aw*lock' to the set_ftrace_filter in this test case,
> we got the error because we have no these functions such as _raw_spin_lock,
> _raw_spin_trylock and so on in our kernel.
>
> Zong Li (6):
>   nds32/ftrace: Support static function tracer
>   nds32/ftrace: Support static function graph tracer
>   nds32/ftrace: Add RECORD_MCOUNT support
>   nds32/ftrace: Support dynamic function tracer
>   nds32/ftrace: Support dynamic function graph tracer
>   nds32/stack: Get real return address by using ftrace_graph_ret_addr
>
>  arch/nds32/Kconfig              |   4 +
>  arch/nds32/Makefile             |   4 +
>  arch/nds32/include/asm/ftrace.h |  46 +++++++
>  arch/nds32/kernel/Makefile      |   6 +
>  arch/nds32/kernel/ftrace.c      | 297 
> ++++++++++++++++++++++++++++++++++++++++
>  arch/nds32/kernel/stacktrace.c  |   4 +
>  arch/nds32/kernel/traps.c       |  30 +---
>  scripts/recordmcount.pl         |   3 +
>  8 files changed, 370 insertions(+), 24 deletions(-)
>  create mode 100644 arch/nds32/include/asm/ftrace.h
>  create mode 100644 arch/nds32/kernel/ftrace.c

Acked-by: Greentime Hu <greent...@andestech.com>

Reply via email to