"Naveen N. Rao" <naveen.n....@linux.vnet.ibm.com> writes: > entry_*.S now includes a lot more than just kernel entry/exit code. As a > first step at cleaning this up, let's split out the ftrace bits into > separate files. > > No functional changes. > > Suggested-by: Michael Ellerman <m...@ellerman.id.au> > Signed-off-by: Naveen N. Rao <naveen.n....@linux.vnet.ibm.com> > --- > v2: updated commit description. > > arch/powerpc/kernel/Makefile | 2 + > arch/powerpc/kernel/entry_32.S | 107 ----------- > arch/powerpc/kernel/entry_64.S | 380 -------------------------------------- > arch/powerpc/kernel/ftrace_32.S | 118 ++++++++++++ > arch/powerpc/kernel/ftrace_64.S | 391 > ++++++++++++++++++++++++++++++++++++++++
Thanks for having a crack at this. I think I'd actually like to go the whole way, and move all the tracing code into arch/powerpc/trace (or ftrace?). There's the 32 and 64-bit asm, and ftrace.c, and trace_clock.c. And then possibly later some of the ftrace module code could move in there too. And as part of moving the asm, I think we can come up with a better way to organise the code. If you look at the ftrace code in entry_64.S for example the ifdefs look like: #ifdef CONFIG_FUNCTION_TRACER #ifdef CONFIG_DYNAMIC_FTRACE #ifdef CONFIG_FUNCTION_GRAPH_TRACER #endif #else /* CC_USING_MPROFILE_KERNEL */ #ifdef CONFIG_LIVEPATCH #endif #ifdef CONFIG_LIVEPATCH #endif #ifdef CONFIG_LIVEPATCH #endif #ifdef CONFIG_FUNCTION_GRAPH_TRACER #endif #endif /* CC_USING_MPROFILE_KERNEL */ #ifdef CONFIG_LIVEPATCH #endif #else #ifdef CONFIG_FUNCTION_GRAPH_TRACER #endif #endif /* CONFIG_DYNAMIC_FTRACE */ #ifdef CONFIG_FUNCTION_GRAPH_TRACER #else /* CC_USING_MPROFILE_KERNEL */ #endif /* CC_USING_MPROFILE_KERNEL */ #endif /* CONFIG_FUNCTION_GRAPH_TRACER */ #endif /* CONFIG_FUNCTION_TRACER */ Which is not easy to follow. I think the main axis is CC_USING_MPROFILE_KERNEL y/n, so perhaps we split it into two .S files based on that? cheers
