Messages in this thread Patch in this message | | | From | Song Shuai <> | Date | Thu, 24 Nov 2022 02:52:06 +0000 | Subject | Re: [PATCH 1/2] riscv/ftrace: add DYNAMIC_FTRACE_WITH_DIRECT_CALLS support |
| |
Guo Ren <guoren@kernel.org> 于2022年11月24日周四 02:08写道: > > On Thu, Nov 24, 2022 at 1:27 AM Song Shuai <suagrfillet@gmail.com> wrote: > > > > Guo Ren <guoren@kernel.org> 于2022年11月23日周三 23:02写道: > > > > > > Cool job, thx. > > > > > > On Wed, Nov 23, 2022 at 10:20 PM Song Shuai <suagrfillet@gmail.com> wrote: > > >> > > >> This patch adds DYNAMIC_FTRACE_WITH_DIRECT_CALLS support for RISC-V. > > >> > > >> select the DYNAMIC_FTRACE_WITH_DIRECT_CALLS to provide the > > >> register_ftrace_direct[_multi] interfaces allowing users to register > > >> the customed trampoline (direct_caller) as the mcount for one or > > >> more target functions. And modify_ftrace_direct[_multi] are also > > >> provided for modifying direct_caller. > > >> > > >> To make the direct_caller and the other ftrace hooks (eg. function/fgraph > > >> tracer, k[ret]probes) co-exist, a temporary register is nominated to > > >> store the address of direct_caller in ftrace_regs_caller. After the > > >> setting of the address direct_caller by direct_ops->func and the > > >> RESTORE_REGS in ftrace_regs_caller, direct_caller will be jumped to > > >> by the `jr` inst. > > >> > > >> Signed-off-by: Song Shuai <suagrfillet@gmail.com> > > >> --- > > >> arch/riscv/Kconfig | 1 + > > >> arch/riscv/include/asm/ftrace.h | 6 ++++++ > > >> arch/riscv/kernel/mcount-dyn.S | 4 ++++ > > >> 3 files changed, 11 insertions(+) > > >> > > >> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig > > >> index 39ec8d628cf6..d083ec08d0b6 100644 > > >> --- a/arch/riscv/Kconfig > > >> +++ b/arch/riscv/Kconfig > > >> @@ -278,6 +278,7 @@ config ARCH_RV64I > > >> select ARCH_SUPPORTS_INT128 if CC_HAS_INT128 > > >> select HAVE_DYNAMIC_FTRACE if !XIP_KERNEL && MMU && $(cc-option,-fpatchable-function-entry=8) > > >> select HAVE_DYNAMIC_FTRACE_WITH_REGS if HAVE_DYNAMIC_FTRACE > > >> + select HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS > > >> select HAVE_FTRACE_MCOUNT_RECORD if !XIP_KERNEL > > >> select HAVE_FUNCTION_GRAPH_TRACER > > >> select HAVE_FUNCTION_TRACER if !XIP_KERNEL && !PREEMPTION > > >> diff --git a/arch/riscv/include/asm/ftrace.h b/arch/riscv/include/asm/ftrace.h > > >> index 01bebb28eabe..be4d57566139 100644 > > >> --- a/arch/riscv/include/asm/ftrace.h > > >> +++ b/arch/riscv/include/asm/ftrace.h > > >> @@ -114,6 +114,12 @@ struct ftrace_regs; > > >> void ftrace_graph_func(unsigned long ip, unsigned long parent_ip, > > >> struct ftrace_ops *op, struct ftrace_regs *fregs); > > >> #define ftrace_graph_func ftrace_graph_func > > >> + > > >> +static inline void arch_ftrace_set_direct_caller(struct pt_regs *regs, unsigned long addr) > > >> +{ > > >> + regs->t1 = addr; > > > > > > How about regs->t0 = addr; ? > > > And delete all mcount-dyn.S modification. > > > > > The direct_caller has the same program layout as the ftrace_caller, which means > > the reg t0 will never be changed when direct_caller returns. > > > > If regs->t0 changes here and ftrace_regs_caller executes `jr t0`, > > direct_caller will enter the dead loop. > ? > > ftrace_regs_caller->call_direct_funcs-> > arch_ftrace_set_direct_caller-> ftrace_regs_caller jr t0. > > Only call_direct_funcs call arch_ftrace_set_direct_caller ! > > static void call_direct_funcs(unsigned long ip, unsigned long pip, > struct ftrace_ops *ops, struct ftrace_regs *fregs) > { > struct pt_regs *regs = ftrace_get_regs(fregs); > unsigned long addr; > > addr = ftrace_find_rec_direct(ip); > if (!addr) > return; > > arch_ftrace_set_direct_caller(regs, addr); > } > When direct_caller and function tracer co-hook a function, the simple diagram is like this:
``` func -> ftrace_regs_caller -> arch_ftrace_ops_list_func -> function_trace_call // write ringbuffer | -> call_direct_funcs // regs->t1 = direct_caller -> t1 !=0 && jr t1 // goto direct_caller ```
And the direct_caller has a similar implement as ftrace_caller.
``` direct_caller: add sp,sp,-? sd t0,?(sp) sd ra,?(sp) call foo ld t0,?(sp) ld ra,?(sp) add sp,sp,? jr t0 // <- back to function entry ```
If we change regs->t0 as direct_caller and execute `jr t0` directly, the direct_caller will always jump to itself due to its last `jr` inst.
``` func -> ftrace_regs_caller -> arch_ftrace_ops_list_func -> function_trace_call // write ringbuffer | -> call_direct_funcs // regs->t0 = direct_caller -> jr t0 // goto direct_caller
direct_caller: ... sd t0,?(sp) ... ld t0,?(s0) ... jr t0 // goto direct_caller always ```
Hope I made it clear. > > > > Actually the reg t0 always saves the address of function entry with 8B > > offset, it should only > > changed by the IPMODIFY ops instead of the direct_ops. > > >> > > >> +} > > >> + > > >> #endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */ > > >> > > >> #endif /* __ASSEMBLY__ */ > > >> diff --git a/arch/riscv/kernel/mcount-dyn.S b/arch/riscv/kernel/mcount-dyn.S > > >> index 466c6ef217b1..b89c85a58569 100644 > > >> --- a/arch/riscv/kernel/mcount-dyn.S > > >> +++ b/arch/riscv/kernel/mcount-dyn.S > > >> @@ -233,6 +233,7 @@ ENDPROC(ftrace_caller) > > >> #else /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */ > > >> ENTRY(ftrace_regs_caller) > > >> SAVE_ABI_REGS 1 > > >> + REG_S x0, PT_T1(sp) > > >> PREPARE_ARGS > > >> > > >> ftrace_regs_call: > > >> @@ -241,7 +242,10 @@ ftrace_regs_call: > > >> > > >> > > >> RESTORE_ABI_REGS 1 > > >> + bnez t1,.Ldirect > > >> jr t0 > > >> +.Ldirect: > > >> + jr t1 > > >> ENDPROC(ftrace_regs_caller) > > >> > > >> ENTRY(ftrace_caller) > > >> -- > > >> 2.20.1 > > >> > > > > > > > > > -- > > > Best Regards > > > Guo Ren > > > > -- > Best Regards > Guo Ren Thanks, Song
| |