In RISC-V architecture, when we enable the ftrace_graph tracer on some
functions, the function tracings on other functions will suffer extra
graph tracing work. In essence, graph_ops isn't limited by its func_hash
due to the global ftrace_graph_[regs]_call label. That should be corrected.
What inspires me is the commit 0c0593b45c9b ("x86/ftrace: Make function
graph use ftrace directly") that uses graph_ops::func function to install
return_hooker and makes the function called against its func_hash.
This series of patches makes function graph use ftrace directly for riscv.
If FTRACE_WITH_REGS isn't defined, ftrace_caller keeps ftrace_graph_call
so that it can be replaced with the calling of prepare_ftrace_return by
the enable/disable helper.
As for defining FTRACE_WITH_REGS, ftrace_caller is adjusted to save the
necessary regs against the pt_regs layout, so it can reasonably call the
graph_ops::func function - ftrace_graph_func. And ftrace_graph_[regs]_call
and its enable/disable helper aren't needed.
The tests generated by CONFIG_FTRACE_STARTUP_TEST have passed in the local
qemu-system-riscv64 virt machine. The following is the log during startup.
```
Nov 15 03:07:13 stage4 kernel: Testing tracer function: PASSED
Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace: PASSED
Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace ops #1:
Nov 15 03:07:13 stage4 kernel: (1 0 1 0 0)
Nov 15 03:07:13 stage4 kernel: (1 1 2 0 0)
Nov 15 03:07:13 stage4 kernel: (2 1 3 0 365)
Nov 15 03:07:13 stage4 kernel: (2 2 4 0 399)
Nov 15 03:07:13 stage4 kernel: (3 2 4 0 146071)
Nov 15 03:07:13 stage4 kernel: (3 3 5 0 146105) PASSED
Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace ops #2:
Nov 15 03:07:13 stage4 kernel: (1 0 1 589 0)
Nov 15 03:07:13 stage4 kernel: (1 1 2 635 0)
Nov 15 03:07:13 stage4 kernel: (2 1 3 1 2)
Nov 15 03:07:13 stage4 kernel: (2 2 4 125 126)
Nov 15 03:07:13 stage4 kernel: (3 2 4 146001 146078)
Nov 15 03:07:13 stage4 kernel: (3 3 5 146035 146112) PASSED
Nov 15 03:07:13 stage4 kernel: Testing ftrace recursion: PASSED
Nov 15 03:07:13 stage4 kernel: Testing ftrace recursion safe: PASSED
Nov 15 03:07:13 stage4 kernel: Testing ftrace regs: PASSED
Nov 15 03:07:13 stage4 kernel: Testing tracer nop: PASSED
Nov 15 03:07:13 stage4 kernel: Testing tracer irqsoff: PASSED
Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup:
Nov 15 03:07:13 stage4 kernel: sched: DL replenish lagged too much
Nov 15 03:07:13 stage4 kernel: PASSED
Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup_rt: PASSED
Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup_dl: PASSED
Nov 15 03:07:13 stage4 kernel: Testing tracer function_graph: PASSED
```
Note that the changes of mcount-dyn.S conflicts with this unmerged
commit (riscv: entry: consolidate general regs saving/restoring).
https://lore.kernel.org/linux-riscv/[email protected]
Changes since v1
- fixed the checkpatch warnings in patch 1
Link: https://lore.kernel.org/linux-riscv/[email protected]
Song Shuai (3):
riscv/ftrace: add ftrace_graph_func
riscv/ftrace: SAVE_ALL supports lightweight save
riscv/ftrace: cleanup ftrace_caller and ftrace_regs_caller
arch/riscv/include/asm/ftrace.h | 13 ++-
arch/riscv/kernel/ftrace.c | 35 ++++----
arch/riscv/kernel/mcount-dyn.S | 145 +++++++++++++++++++++++---------
3 files changed, 136 insertions(+), 57 deletions(-)
--
2.20.1
Here implements ftrace_graph_func as the function graph tracing function
with FTRACE_WITH_REGS defined.
function_graph_func gets the point of the parent IP and the frame pointer
from fregs and call prepare_ftrace_return for function graph tracing.
If FTRACE_WITH_REGS isn't defined, the enable/disable helpers of
ftrace_graph_[regs]_call remain for serving the !FTRACE_WITH_REGS
version ftrace_caller.
Signed-off-by: Song Shuai <[email protected]>
---
arch/riscv/include/asm/ftrace.h | 13 +++++++++++--
arch/riscv/kernel/ftrace.c | 16 ++++++++++++++++
2 files changed, 27 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/include/asm/ftrace.h b/arch/riscv/include/asm/ftrace.h
index 04dad3380041..61a8f4be67eb 100644
--- a/arch/riscv/include/asm/ftrace.h
+++ b/arch/riscv/include/asm/ftrace.h
@@ -81,8 +81,17 @@ do { \
struct dyn_ftrace;
int ftrace_init_nop(struct module *mod, struct dyn_ftrace *rec);
#define ftrace_init_nop ftrace_init_nop
-#endif
-#endif
+#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
+struct ftrace_ops;
+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
+#endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
+
+#endif /* __ASSEMBLY__ */
+
+#endif /* CONFIG_DYNAMIC_FTRACE */
#endif /* _ASM_RISCV_FTRACE_H */
diff --git a/arch/riscv/kernel/ftrace.c b/arch/riscv/kernel/ftrace.c
index 2086f6585773..148a5480076b 100644
--- a/arch/riscv/kernel/ftrace.c
+++ b/arch/riscv/kernel/ftrace.c
@@ -196,6 +196,20 @@ void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr,
}
#ifdef CONFIG_DYNAMIC_FTRACE
+
+#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
+
+void ftrace_graph_func(unsigned long ip, unsigned long parent_ip,
+ struct ftrace_ops *op, struct ftrace_regs *fregs)
+{
+ struct pt_regs *regs = arch_ftrace_get_regs(fregs);
+ unsigned long *parent = (unsigned long *)®s->ra;
+
+ prepare_ftrace_return(parent, ip, frame_pointer(regs));
+}
+
+#else /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
+
extern void ftrace_graph_call(void);
extern void ftrace_graph_regs_call(void);
int ftrace_enable_ftrace_graph_caller(void)
@@ -223,5 +237,7 @@ int ftrace_disable_ftrace_graph_caller(void)
return __ftrace_modify_call((unsigned long)&ftrace_graph_regs_call,
(unsigned long)&prepare_ftrace_return, false);
}
+#endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
+
#endif /* CONFIG_DYNAMIC_FTRACE */
#endif /* CONFIG_FUNCTION_GRAPH_TRACER */
--
2.20.1
In order to make the function graph use ftrace directly, ftrace_caller
should be adjusted to save the necessary regs against the pt_regs layout
so it can call ftrace_graph_func reasonably.
SAVE_ALL now saves all the regs according to the pt_regs struct. Here
introduces a lightweight option for SAVE_ALL to save only the necessary
regs for ftrace_caller.
For convenience, the original argument setup for the tracing function in
ftrace_[regs]_caller is killed and appended to the tail of SAVE_ALL.
Signed-off-by: Song Shuai <[email protected]>
---
arch/riscv/kernel/mcount-dyn.S | 110 +++++++++++++++++++++++++++------
1 file changed, 92 insertions(+), 18 deletions(-)
diff --git a/arch/riscv/kernel/mcount-dyn.S b/arch/riscv/kernel/mcount-dyn.S
index d171eca623b6..2f0a280bd7a0 100644
--- a/arch/riscv/kernel/mcount-dyn.S
+++ b/arch/riscv/kernel/mcount-dyn.S
@@ -56,7 +56,51 @@
.endm
#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
- .macro SAVE_ALL
+
+/**
+* SAVE_ALL - save regs against the pt_regs struct
+*
+* @all: tell if saving all the regs
+*
+* If all is set, all the regs will be saved, otherwise only ABI
+* related regs (a0-a7,epc,ra and optional s0) will be saved.
+*
+* For convenience the argument setup for tracing function is appended here.
+* Especially $sp is passed as the 4th argument of the tracing function.
+*
+* After the stack is established,
+*
+* 0(sp) stores the PC of the traced function which can be accessed
+* by &(fregs)->regs->epc in tracing function. Note that the real
+* function entry address should be computed with -FENTRY_RA_OFFSET.
+*
+* 8(sp) stores the function return address (i.e. parent IP) that
+* can be accessed by &(fregs)->regs->ra in tracing function.
+*
+* The other regs are saved at the respective localtion and accessed
+* by the respective pt_regs member.
+*
+* Here is the layout of stack for your reference.
+*
+*
+* =========
+* | pip |
+* PT_SIZE_ON_STACK -> =========
+* + ..... +
+* + t3-t6 +
+* + s2-s11+
+* + a0-a7 + --++++-> ftrace_caller saved
+* + s1 + +
+* + s0 + --+
+* + t0-t2 + +
+* + tp + +
+* + gp + +
+* + sp + +
+* + ra + --+ // parent IP
+* sp -> + epc + --+ // PC of the traced function
+* +++++++++
+**/
+ .macro SAVE_ALL, all=0
addi sp, sp, -SZREG
addi sp, sp, -PT_SIZE_ON_STACK
@@ -67,14 +111,8 @@
REG_S x1, PT_RA(sp)
REG_L x1, PT_EPC(sp)
- REG_S x2, PT_SP(sp)
- REG_S x3, PT_GP(sp)
- REG_S x4, PT_TP(sp)
- REG_S x5, PT_T0(sp)
- REG_S x6, PT_T1(sp)
- REG_S x7, PT_T2(sp)
- REG_S x8, PT_S0(sp)
- REG_S x9, PT_S1(sp)
+ /* always save the ABI regs */
+
REG_S x10, PT_A0(sp)
REG_S x11, PT_A1(sp)
REG_S x12, PT_A2(sp)
@@ -83,6 +121,18 @@
REG_S x15, PT_A5(sp)
REG_S x16, PT_A6(sp)
REG_S x17, PT_A7(sp)
+
+ /* save leftover regs for ftrace_regs_caller*/
+
+ .if \all == 1
+ REG_S x2, PT_SP(sp)
+ REG_S x3, PT_GP(sp)
+ REG_S x4, PT_TP(sp)
+ REG_S x5, PT_T0(sp)
+ REG_S x6, PT_T1(sp)
+ REG_S x7, PT_T2(sp)
+ REG_S x8, PT_S0(sp)
+ REG_S x9, PT_S1(sp)
REG_S x18, PT_S2(sp)
REG_S x19, PT_S3(sp)
REG_S x20, PT_S4(sp)
@@ -97,22 +147,31 @@
REG_S x29, PT_T4(sp)
REG_S x30, PT_T5(sp)
REG_S x31, PT_T6(sp)
+ .else
+
+ /* save s0 for ftrace_caller if FP_TEST defined */
+
+#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
+ REG_S x8, PT_S0(sp)
+#endif
+ .endif
+
+ /* setup 4 args for tracing functions */
+
+ addi a0, ra, -FENTRY_RA_OFFSET // ip
+ la a1, function_trace_op
+ REG_L a2, 0(a1) // op
+ REG_L a1, PT_SIZE_ON_STACK(sp) // parent_ip
+ mv a3, sp // fregs
.endm
- .macro RESTORE_ALL
+ .macro RESTORE_ALL, all=0
REG_L x1, PT_RA(sp)
addi sp, sp, PT_SIZE_ON_STACK
REG_S x1, (sp)
addi sp, sp, -PT_SIZE_ON_STACK
REG_L x1, PT_EPC(sp)
- REG_L x2, PT_SP(sp)
- REG_L x3, PT_GP(sp)
- REG_L x4, PT_TP(sp)
- REG_L x5, PT_T0(sp)
- REG_L x6, PT_T1(sp)
- REG_L x7, PT_T2(sp)
- REG_L x8, PT_S0(sp)
- REG_L x9, PT_S1(sp)
+
REG_L x10, PT_A0(sp)
REG_L x11, PT_A1(sp)
REG_L x12, PT_A2(sp)
@@ -121,6 +180,16 @@
REG_L x15, PT_A5(sp)
REG_L x16, PT_A6(sp)
REG_L x17, PT_A7(sp)
+
+ .if \all == 1
+ REG_L x2, PT_SP(sp)
+ REG_L x3, PT_GP(sp)
+ REG_L x4, PT_TP(sp)
+ REG_L x5, PT_T0(sp)
+ REG_L x6, PT_T1(sp)
+ REG_L x7, PT_T2(sp)
+ REG_L x8, PT_S0(sp)
+ REG_L x9, PT_S1(sp)
REG_L x18, PT_S2(sp)
REG_L x19, PT_S3(sp)
REG_L x20, PT_S4(sp)
@@ -136,6 +205,11 @@
REG_L x30, PT_T5(sp)
REG_L x31, PT_T6(sp)
+ .else
+#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
+ REG_L x8, PT_S0(sp)
+#endif
+ .endif
addi sp, sp, PT_SIZE_ON_STACK
addi sp, sp, SZREG
.endm
--
2.20.1
ftrace_caller and ftrace_regs_caller save their regs with the respective
option of SAVE_ALL, then call the tracing function, especially graph_ops's
ftrace_graph_func. So the ftrace_graph_[regs]_call labels aren't needed
anymore if FTRACE_WITH_REGS is defined.
If FTRACE_WITH_REGS isn't defined, the !FTRACE_WITH_REGS version
ftrace_caller remains with the ftrace_graph_call. So the enable/disable
helpers are revised for serving only this ftrace_graph_call.
Signed-off-by: Song Shuai <[email protected]>
---
arch/riscv/kernel/ftrace.c | 19 ++----------------
arch/riscv/kernel/mcount-dyn.S | 35 +++++++++++++++-------------------
2 files changed, 17 insertions(+), 37 deletions(-)
diff --git a/arch/riscv/kernel/ftrace.c b/arch/riscv/kernel/ftrace.c
index 148a5480076b..2f0bcedc6a2d 100644
--- a/arch/riscv/kernel/ftrace.c
+++ b/arch/riscv/kernel/ftrace.c
@@ -211,30 +211,15 @@ void ftrace_graph_func(unsigned long ip, unsigned long parent_ip,
#else /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
extern void ftrace_graph_call(void);
-extern void ftrace_graph_regs_call(void);
int ftrace_enable_ftrace_graph_caller(void)
{
- int ret;
-
- ret = __ftrace_modify_call((unsigned long)&ftrace_graph_call,
- (unsigned long)&prepare_ftrace_return, true);
- if (ret)
- return ret;
-
- return __ftrace_modify_call((unsigned long)&ftrace_graph_regs_call,
+ return __ftrace_modify_call((unsigned long)&ftrace_graph_call,
(unsigned long)&prepare_ftrace_return, true);
}
int ftrace_disable_ftrace_graph_caller(void)
{
- int ret;
-
- ret = __ftrace_modify_call((unsigned long)&ftrace_graph_call,
- (unsigned long)&prepare_ftrace_return, false);
- if (ret)
- return ret;
-
- return __ftrace_modify_call((unsigned long)&ftrace_graph_regs_call,
+ return __ftrace_modify_call((unsigned long)&ftrace_graph_call,
(unsigned long)&prepare_ftrace_return, false);
}
#endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
diff --git a/arch/riscv/kernel/mcount-dyn.S b/arch/riscv/kernel/mcount-dyn.S
index 2f0a280bd7a0..9e4097c6793d 100644
--- a/arch/riscv/kernel/mcount-dyn.S
+++ b/arch/riscv/kernel/mcount-dyn.S
@@ -215,6 +215,7 @@
.endm
#endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
+#ifndef CONFIG_DYNAMIC_FTRACE_WITH_REGS
ENTRY(ftrace_caller)
SAVE_ABI
@@ -243,33 +244,27 @@ ftrace_graph_call:
ret
ENDPROC(ftrace_caller)
-#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
+#else /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
ENTRY(ftrace_regs_caller)
- SAVE_ALL
-
- addi a0, ra, -FENTRY_RA_OFFSET
- la a1, function_trace_op
- REG_L a2, 0(a1)
- REG_L a1, PT_SIZE_ON_STACK(sp)
- mv a3, sp
+ SAVE_ALL 1
ftrace_regs_call:
.global ftrace_regs_call
call ftrace_stub
-#ifdef CONFIG_FUNCTION_GRAPH_TRACER
- addi a0, sp, PT_RA
- REG_L a1, PT_EPC(sp)
- addi a1, a1, -FENTRY_RA_OFFSET
-#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
- mv a2, s0
-#endif
-ftrace_graph_regs_call:
- .global ftrace_graph_regs_call
- call ftrace_stub
-#endif
- RESTORE_ALL
+ RESTORE_ALL 1
ret
ENDPROC(ftrace_regs_caller)
+
+ENTRY(ftrace_caller)
+ SAVE_ALL 0
+
+ftrace_call:
+ .global ftrace_call
+ call ftrace_stub
+
+ RESTORE_ALL 0
+ ret
+ENDPROC(ftrace_caller)
#endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
--
2.20.1
On Wed, Nov 16, 2022 at 11:13:04AM +0800, Song Shuai wrote:
> In order to make the function graph use ftrace directly, ftrace_caller
> should be adjusted to save the necessary regs against the pt_regs layout
> so it can call ftrace_graph_func reasonably.
>
> SAVE_ALL now saves all the regs according to the pt_regs struct. Here
> introduces a lightweight option for SAVE_ALL to save only the necessary
> regs for ftrace_caller.
>
> For convenience, the original argument setup for the tracing function in
> ftrace_[regs]_caller is killed and appended to the tail of SAVE_ALL.
>
> Signed-off-by: Song Shuai <[email protected]>
> ---
> arch/riscv/kernel/mcount-dyn.S | 110 +++++++++++++++++++++++++++------
> 1 file changed, 92 insertions(+), 18 deletions(-)
>
> diff --git a/arch/riscv/kernel/mcount-dyn.S b/arch/riscv/kernel/mcount-dyn.S
> index d171eca623b6..2f0a280bd7a0 100644
> --- a/arch/riscv/kernel/mcount-dyn.S
> +++ b/arch/riscv/kernel/mcount-dyn.S
> @@ -56,7 +56,51 @@
> .endm
>
> #ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
> - .macro SAVE_ALL
> +
> +/**
> +* SAVE_ALL - save regs against the pt_regs struct
> +*
> +* @all: tell if saving all the regs
I find it odd to have a macro name that includes 'ALL' in it
to require a parameter 'all' to be set in order to actually save
all. I suggest renaming the macro to something like SAVE_REGS.
> +*
> +* If all is set, all the regs will be saved, otherwise only ABI
> +* related regs (a0-a7,epc,ra and optional s0) will be saved.
> +*
> +* For convenience the argument setup for tracing function is appended here.
> +* Especially $sp is passed as the 4th argument of the tracing function.
> +*
> +* After the stack is established,
> +*
> +* 0(sp) stores the PC of the traced function which can be accessed
> +* by &(fregs)->regs->epc in tracing function. Note that the real
> +* function entry address should be computed with -FENTRY_RA_OFFSET.
> +*
> +* 8(sp) stores the function return address (i.e. parent IP) that
> +* can be accessed by &(fregs)->regs->ra in tracing function.
> +*
> +* The other regs are saved at the respective localtion and accessed
> +* by the respective pt_regs member.
> +*
> +* Here is the layout of stack for your reference.
> +*
> +*
> +* =========
> +* | pip |
> +* PT_SIZE_ON_STACK -> =========
> +* + ..... +
> +* + t3-t6 +
> +* + s2-s11+
> +* + a0-a7 + --++++-> ftrace_caller saved
> +* + s1 + +
> +* + s0 + --+
> +* + t0-t2 + +
> +* + tp + +
> +* + gp + +
> +* + sp + +
> +* + ra + --+ // parent IP
> +* sp -> + epc + --+ // PC of the traced function
> +* +++++++++
> +**/
> + .macro SAVE_ALL, all=0
> addi sp, sp, -SZREG
> addi sp, sp, -PT_SIZE_ON_STACK
>
> @@ -67,14 +111,8 @@
> REG_S x1, PT_RA(sp)
> REG_L x1, PT_EPC(sp)
>
> - REG_S x2, PT_SP(sp)
> - REG_S x3, PT_GP(sp)
> - REG_S x4, PT_TP(sp)
> - REG_S x5, PT_T0(sp)
> - REG_S x6, PT_T1(sp)
> - REG_S x7, PT_T2(sp)
> - REG_S x8, PT_S0(sp)
> - REG_S x9, PT_S1(sp)
> + /* always save the ABI regs */
> +
> REG_S x10, PT_A0(sp)
> REG_S x11, PT_A1(sp)
> REG_S x12, PT_A2(sp)
> @@ -83,6 +121,18 @@
> REG_S x15, PT_A5(sp)
> REG_S x16, PT_A6(sp)
> REG_S x17, PT_A7(sp)
> +
> + /* save leftover regs for ftrace_regs_caller*/
> +
> + .if \all == 1
> + REG_S x2, PT_SP(sp)
> + REG_S x3, PT_GP(sp)
> + REG_S x4, PT_TP(sp)
> + REG_S x5, PT_T0(sp)
> + REG_S x6, PT_T1(sp)
> + REG_S x7, PT_T2(sp)
> + REG_S x8, PT_S0(sp)
> + REG_S x9, PT_S1(sp)
> REG_S x18, PT_S2(sp)
> REG_S x19, PT_S3(sp)
> REG_S x20, PT_S4(sp)
> @@ -97,22 +147,31 @@
> REG_S x29, PT_T4(sp)
> REG_S x30, PT_T5(sp)
> REG_S x31, PT_T6(sp)
> + .else
> +
> + /* save s0 for ftrace_caller if FP_TEST defined */
> +
> +#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
> + REG_S x8, PT_S0(sp)
> +#endif
> + .endif
> +
> + /* setup 4 args for tracing functions */
> +
> + addi a0, ra, -FENTRY_RA_OFFSET // ip
> + la a1, function_trace_op
> + REG_L a2, 0(a1) // op
> + REG_L a1, PT_SIZE_ON_STACK(sp) // parent_ip
> + mv a3, sp // fregs
Please line up the comments.
> .endm
>
> - .macro RESTORE_ALL
> + .macro RESTORE_ALL, all=0
> REG_L x1, PT_RA(sp)
> addi sp, sp, PT_SIZE_ON_STACK
> REG_S x1, (sp)
> addi sp, sp, -PT_SIZE_ON_STACK
> REG_L x1, PT_EPC(sp)
> - REG_L x2, PT_SP(sp)
> - REG_L x3, PT_GP(sp)
> - REG_L x4, PT_TP(sp)
> - REG_L x5, PT_T0(sp)
> - REG_L x6, PT_T1(sp)
> - REG_L x7, PT_T2(sp)
> - REG_L x8, PT_S0(sp)
> - REG_L x9, PT_S1(sp)
> +
> REG_L x10, PT_A0(sp)
> REG_L x11, PT_A1(sp)
> REG_L x12, PT_A2(sp)
> @@ -121,6 +180,16 @@
> REG_L x15, PT_A5(sp)
> REG_L x16, PT_A6(sp)
> REG_L x17, PT_A7(sp)
> +
> + .if \all == 1
> + REG_L x2, PT_SP(sp)
> + REG_L x3, PT_GP(sp)
> + REG_L x4, PT_TP(sp)
> + REG_L x5, PT_T0(sp)
> + REG_L x6, PT_T1(sp)
> + REG_L x7, PT_T2(sp)
> + REG_L x8, PT_S0(sp)
> + REG_L x9, PT_S1(sp)
> REG_L x18, PT_S2(sp)
> REG_L x19, PT_S3(sp)
> REG_L x20, PT_S4(sp)
> @@ -136,6 +205,11 @@
> REG_L x30, PT_T5(sp)
> REG_L x31, PT_T6(sp)
>
> + .else
> +#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
> + REG_L x8, PT_S0(sp)
> +#endif
> + .endif
> addi sp, sp, PT_SIZE_ON_STACK
> addi sp, sp, SZREG
> .endm
> --
> 2.20.1
>
>
Thanks,
drew
On Wed, Nov 16, 2022 at 11:13 AM Song Shuai <[email protected]> wrote:
>
> ftrace_caller and ftrace_regs_caller save their regs with the respective
> option of SAVE_ALL, then call the tracing function, especially graph_ops's
> ftrace_graph_func. So the ftrace_graph_[regs]_call labels aren't needed
> anymore if FTRACE_WITH_REGS is defined.
>
> If FTRACE_WITH_REGS isn't defined, the !FTRACE_WITH_REGS version
> ftrace_caller remains with the ftrace_graph_call. So the enable/disable
> helpers are revised for serving only this ftrace_graph_call.
>
> Signed-off-by: Song Shuai <[email protected]>
> ---
> arch/riscv/kernel/ftrace.c | 19 ++----------------
> arch/riscv/kernel/mcount-dyn.S | 35 +++++++++++++++-------------------
> 2 files changed, 17 insertions(+), 37 deletions(-)
>
> diff --git a/arch/riscv/kernel/ftrace.c b/arch/riscv/kernel/ftrace.c
> index 148a5480076b..2f0bcedc6a2d 100644
> --- a/arch/riscv/kernel/ftrace.c
> +++ b/arch/riscv/kernel/ftrace.c
> @@ -211,30 +211,15 @@ void ftrace_graph_func(unsigned long ip, unsigned long parent_ip,
> #else /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
>
> extern void ftrace_graph_call(void);
> -extern void ftrace_graph_regs_call(void);
> int ftrace_enable_ftrace_graph_caller(void)
> {
> - int ret;
> -
> - ret = __ftrace_modify_call((unsigned long)&ftrace_graph_call,
> - (unsigned long)&prepare_ftrace_return, true);
> - if (ret)
> - return ret;
> -
> - return __ftrace_modify_call((unsigned long)&ftrace_graph_regs_call,
> + return __ftrace_modify_call((unsigned long)&ftrace_graph_call,
> (unsigned long)&prepare_ftrace_return, true);
> }
>
> int ftrace_disable_ftrace_graph_caller(void)
> {
> - int ret;
> -
> - ret = __ftrace_modify_call((unsigned long)&ftrace_graph_call,
> - (unsigned long)&prepare_ftrace_return, false);
> - if (ret)
> - return ret;
> -
> - return __ftrace_modify_call((unsigned long)&ftrace_graph_regs_call,
> + return __ftrace_modify_call((unsigned long)&ftrace_graph_call,
> (unsigned long)&prepare_ftrace_return, false);
> }
> #endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
> diff --git a/arch/riscv/kernel/mcount-dyn.S b/arch/riscv/kernel/mcount-dyn.S
> index 2f0a280bd7a0..9e4097c6793d 100644
> --- a/arch/riscv/kernel/mcount-dyn.S
> +++ b/arch/riscv/kernel/mcount-dyn.S
> @@ -215,6 +215,7 @@
> .endm
> #endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
>
> +#ifndef CONFIG_DYNAMIC_FTRACE_WITH_REGS
> ENTRY(ftrace_caller)
> SAVE_ABI
>
> @@ -243,33 +244,27 @@ ftrace_graph_call:
> ret
> ENDPROC(ftrace_caller)
>
> -#ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
> +#else /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
> ENTRY(ftrace_regs_caller)
> - SAVE_ALL
> -
> - addi a0, ra, -FENTRY_RA_OFFSET
> - la a1, function_trace_op
> - REG_L a2, 0(a1)
> - REG_L a1, PT_SIZE_ON_STACK(sp)
> - mv a3, sp
> + SAVE_ALL 1
I think the patch is broken, we shouldn't split mcount-dyn.S
modification from the patch.
>
> ftrace_regs_call:
> .global ftrace_regs_call
> call ftrace_stub
>
> -#ifdef CONFIG_FUNCTION_GRAPH_TRACER
> - addi a0, sp, PT_RA
> - REG_L a1, PT_EPC(sp)
> - addi a1, a1, -FENTRY_RA_OFFSET
> -#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
> - mv a2, s0
> -#endif
> -ftrace_graph_regs_call:
> - .global ftrace_graph_regs_call
> - call ftrace_stub
> -#endif
>
> - RESTORE_ALL
> + RESTORE_ALL 1
> ret
> ENDPROC(ftrace_regs_caller)
> +
> +ENTRY(ftrace_caller)
> + SAVE_ALL 0
> +
> +ftrace_call:
> + .global ftrace_call
> + call ftrace_stub
> +
> + RESTORE_ALL 0
> + ret
> +ENDPROC(ftrace_caller)
> #endif /* CONFIG_DYNAMIC_FTRACE_WITH_REGS */
> --
> 2.20.1
>
--
Best Regards
Guo Ren
I hope this series could base on the following:
https://lore.kernel.org/linux-riscv/[email protected]/
They all modified the mcount-dyn.S.
On Wed, Nov 16, 2022 at 11:13 AM Song Shuai <[email protected]> wrote:
>
> In RISC-V architecture, when we enable the ftrace_graph tracer on some
> functions, the function tracings on other functions will suffer extra
> graph tracing work. In essence, graph_ops isn't limited by its func_hash
> due to the global ftrace_graph_[regs]_call label. That should be corrected.
>
> What inspires me is the commit 0c0593b45c9b ("x86/ftrace: Make function
> graph use ftrace directly") that uses graph_ops::func function to install
> return_hooker and makes the function called against its func_hash.
>
> This series of patches makes function graph use ftrace directly for riscv.
>
> If FTRACE_WITH_REGS isn't defined, ftrace_caller keeps ftrace_graph_call
> so that it can be replaced with the calling of prepare_ftrace_return by
> the enable/disable helper.
>
> As for defining FTRACE_WITH_REGS, ftrace_caller is adjusted to save the
> necessary regs against the pt_regs layout, so it can reasonably call the
> graph_ops::func function - ftrace_graph_func. And ftrace_graph_[regs]_call
> and its enable/disable helper aren't needed.
>
> The tests generated by CONFIG_FTRACE_STARTUP_TEST have passed in the local
> qemu-system-riscv64 virt machine. The following is the log during startup.
>
> ```
> Nov 15 03:07:13 stage4 kernel: Testing tracer function: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace ops #1:
> Nov 15 03:07:13 stage4 kernel: (1 0 1 0 0)
> Nov 15 03:07:13 stage4 kernel: (1 1 2 0 0)
> Nov 15 03:07:13 stage4 kernel: (2 1 3 0 365)
> Nov 15 03:07:13 stage4 kernel: (2 2 4 0 399)
> Nov 15 03:07:13 stage4 kernel: (3 2 4 0 146071)
> Nov 15 03:07:13 stage4 kernel: (3 3 5 0 146105) PASSED
> Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace ops #2:
> Nov 15 03:07:13 stage4 kernel: (1 0 1 589 0)
> Nov 15 03:07:13 stage4 kernel: (1 1 2 635 0)
> Nov 15 03:07:13 stage4 kernel: (2 1 3 1 2)
> Nov 15 03:07:13 stage4 kernel: (2 2 4 125 126)
> Nov 15 03:07:13 stage4 kernel: (3 2 4 146001 146078)
> Nov 15 03:07:13 stage4 kernel: (3 3 5 146035 146112) PASSED
> Nov 15 03:07:13 stage4 kernel: Testing ftrace recursion: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing ftrace recursion safe: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing ftrace regs: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing tracer nop: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing tracer irqsoff: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup:
> Nov 15 03:07:13 stage4 kernel: sched: DL replenish lagged too much
> Nov 15 03:07:13 stage4 kernel: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup_rt: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup_dl: PASSED
> Nov 15 03:07:13 stage4 kernel: Testing tracer function_graph: PASSED
> ```
>
> Note that the changes of mcount-dyn.S conflicts with this unmerged
> commit (riscv: entry: consolidate general regs saving/restoring).
> https://lore.kernel.org/linux-riscv/[email protected]
>
> Changes since v1
> - fixed the checkpatch warnings in patch 1
> Link: https://lore.kernel.org/linux-riscv/[email protected]
>
> Song Shuai (3):
> riscv/ftrace: add ftrace_graph_func
> riscv/ftrace: SAVE_ALL supports lightweight save
> riscv/ftrace: cleanup ftrace_caller and ftrace_regs_caller
>
> arch/riscv/include/asm/ftrace.h | 13 ++-
> arch/riscv/kernel/ftrace.c | 35 ++++----
> arch/riscv/kernel/mcount-dyn.S | 145 +++++++++++++++++++++++---------
> 3 files changed, 136 insertions(+), 57 deletions(-)
>
> --
> 2.20.1
>
--
Best Regards
Guo Ren
On Wed, Nov 16, 2022 at 4:45 PM Andrew Jones <[email protected]> wrote:
>
> On Wed, Nov 16, 2022 at 11:13:04AM +0800, Song Shuai wrote:
> > In order to make the function graph use ftrace directly, ftrace_caller
> > should be adjusted to save the necessary regs against the pt_regs layout
> > so it can call ftrace_graph_func reasonably.
> >
> > SAVE_ALL now saves all the regs according to the pt_regs struct. Here
> > introduces a lightweight option for SAVE_ALL to save only the necessary
> > regs for ftrace_caller.
> >
> > For convenience, the original argument setup for the tracing function in
> > ftrace_[regs]_caller is killed and appended to the tail of SAVE_ALL.
> >
> > Signed-off-by: Song Shuai <[email protected]>
> > ---
> > arch/riscv/kernel/mcount-dyn.S | 110 +++++++++++++++++++++++++++------
> > 1 file changed, 92 insertions(+), 18 deletions(-)
> >
> > diff --git a/arch/riscv/kernel/mcount-dyn.S b/arch/riscv/kernel/mcount-dyn.S
> > index d171eca623b6..2f0a280bd7a0 100644
> > --- a/arch/riscv/kernel/mcount-dyn.S
> > +++ b/arch/riscv/kernel/mcount-dyn.S
> > @@ -56,7 +56,51 @@
> > .endm
> >
> > #ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS
> > - .macro SAVE_ALL
> > +
> > +/**
> > +* SAVE_ALL - save regs against the pt_regs struct
> > +*
> > +* @all: tell if saving all the regs
>
> I find it odd to have a macro name that includes 'ALL' in it
> to require a parameter 'all' to be set in order to actually save
> all. I suggest renaming the macro to something like SAVE_REGS.
SAVE_ABI_REGS <arg>
arg = 0 -> SAVE_ABI in pt_regs
arg = 1 -> SAVE_REGS
SAVE_ABI_REGS keeps the pt_regs layout and needs more stack space than SAVE_ABI.
>
> > +*
> > +* If all is set, all the regs will be saved, otherwise only ABI
> > +* related regs (a0-a7,epc,ra and optional s0) will be saved.
> > +*
> > +* For convenience the argument setup for tracing function is appended here.
> > +* Especially $sp is passed as the 4th argument of the tracing function.
> > +*
> > +* After the stack is established,
> > +*
> > +* 0(sp) stores the PC of the traced function which can be accessed
> > +* by &(fregs)->regs->epc in tracing function. Note that the real
> > +* function entry address should be computed with -FENTRY_RA_OFFSET.
> > +*
> > +* 8(sp) stores the function return address (i.e. parent IP) that
> > +* can be accessed by &(fregs)->regs->ra in tracing function.
> > +*
> > +* The other regs are saved at the respective localtion and accessed
> > +* by the respective pt_regs member.
> > +*
> > +* Here is the layout of stack for your reference.
> > +*
> > +*
> > +* =========
> > +* | pip |
> > +* PT_SIZE_ON_STACK -> =========
> > +* + ..... +
> > +* + t3-t6 +
> > +* + s2-s11+
> > +* + a0-a7 + --++++-> ftrace_caller saved
> > +* + s1 + +
> > +* + s0 + --+
> > +* + t0-t2 + +
> > +* + tp + +
> > +* + gp + +
> > +* + sp + +
> > +* + ra + --+ // parent IP
> > +* sp -> + epc + --+ // PC of the traced function
> > +* +++++++++
> > +**/
> > + .macro SAVE_ALL, all=0
> > addi sp, sp, -SZREG
> > addi sp, sp, -PT_SIZE_ON_STACK
> >
> > @@ -67,14 +111,8 @@
> > REG_S x1, PT_RA(sp)
> > REG_L x1, PT_EPC(sp)
> >
> > - REG_S x2, PT_SP(sp)
> > - REG_S x3, PT_GP(sp)
> > - REG_S x4, PT_TP(sp)
> > - REG_S x5, PT_T0(sp)
> > - REG_S x6, PT_T1(sp)
> > - REG_S x7, PT_T2(sp)
> > - REG_S x8, PT_S0(sp)
> > - REG_S x9, PT_S1(sp)
> > + /* always save the ABI regs */
> > +
> > REG_S x10, PT_A0(sp)
> > REG_S x11, PT_A1(sp)
> > REG_S x12, PT_A2(sp)
> > @@ -83,6 +121,18 @@
> > REG_S x15, PT_A5(sp)
> > REG_S x16, PT_A6(sp)
> > REG_S x17, PT_A7(sp)
> > +
> > + /* save leftover regs for ftrace_regs_caller*/
> > +
> > + .if \all == 1
> > + REG_S x2, PT_SP(sp)
> > + REG_S x3, PT_GP(sp)
> > + REG_S x4, PT_TP(sp)
> > + REG_S x5, PT_T0(sp)
> > + REG_S x6, PT_T1(sp)
> > + REG_S x7, PT_T2(sp)
> > + REG_S x8, PT_S0(sp)
> > + REG_S x9, PT_S1(sp)
> > REG_S x18, PT_S2(sp)
> > REG_S x19, PT_S3(sp)
> > REG_S x20, PT_S4(sp)
> > @@ -97,22 +147,31 @@
> > REG_S x29, PT_T4(sp)
> > REG_S x30, PT_T5(sp)
> > REG_S x31, PT_T6(sp)
> > + .else
> > +
> > + /* save s0 for ftrace_caller if FP_TEST defined */
> > +
> > +#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
> > + REG_S x8, PT_S0(sp)
> > +#endif
> > + .endif
> > +
> > + /* setup 4 args for tracing functions */
> > +
> > + addi a0, ra, -FENTRY_RA_OFFSET // ip
> > + la a1, function_trace_op
> > + REG_L a2, 0(a1) // op
> > + REG_L a1, PT_SIZE_ON_STACK(sp) // parent_ip
> > + mv a3, sp // fregs
>
> Please line up the comments.
>
> > .endm
> >
> > - .macro RESTORE_ALL
> > + .macro RESTORE_ALL, all=0
> > REG_L x1, PT_RA(sp)
> > addi sp, sp, PT_SIZE_ON_STACK
> > REG_S x1, (sp)
> > addi sp, sp, -PT_SIZE_ON_STACK
> > REG_L x1, PT_EPC(sp)
> > - REG_L x2, PT_SP(sp)
> > - REG_L x3, PT_GP(sp)
> > - REG_L x4, PT_TP(sp)
> > - REG_L x5, PT_T0(sp)
> > - REG_L x6, PT_T1(sp)
> > - REG_L x7, PT_T2(sp)
> > - REG_L x8, PT_S0(sp)
> > - REG_L x9, PT_S1(sp)
> > +
> > REG_L x10, PT_A0(sp)
> > REG_L x11, PT_A1(sp)
> > REG_L x12, PT_A2(sp)
> > @@ -121,6 +180,16 @@
> > REG_L x15, PT_A5(sp)
> > REG_L x16, PT_A6(sp)
> > REG_L x17, PT_A7(sp)
> > +
> > + .if \all == 1
> > + REG_L x2, PT_SP(sp)
> > + REG_L x3, PT_GP(sp)
> > + REG_L x4, PT_TP(sp)
> > + REG_L x5, PT_T0(sp)
> > + REG_L x6, PT_T1(sp)
> > + REG_L x7, PT_T2(sp)
> > + REG_L x8, PT_S0(sp)
> > + REG_L x9, PT_S1(sp)
> > REG_L x18, PT_S2(sp)
> > REG_L x19, PT_S3(sp)
> > REG_L x20, PT_S4(sp)
> > @@ -136,6 +205,11 @@
> > REG_L x30, PT_T5(sp)
> > REG_L x31, PT_T6(sp)
> >
> > + .else
> > +#ifdef HAVE_FUNCTION_GRAPH_FP_TEST
> > + REG_L x8, PT_S0(sp)
> > +#endif
> > + .endif
> > addi sp, sp, PT_SIZE_ON_STACK
> > addi sp, sp, SZREG
> > .endm
> > --
> > 2.20.1
> >
> >
>
> Thanks,
> drew
--
Best Regards
Guo Ren
Guo Ren <[email protected]> 于2022年11月17日周四 01:46写道：
>
> I hope this series could base on the following:
> https://lore.kernel.org/linux-riscv/[email protected]/
>
> They all modified the mcount-dyn.S.
Ok, I'll adapt these patches based on the series you mentioned.
>
>
> On Wed, Nov 16, 2022 at 11:13 AM Song Shuai <[email protected]> wrote:
> >
> > In RISC-V architecture, when we enable the ftrace_graph tracer on some
> > functions, the function tracings on other functions will suffer extra
> > graph tracing work. In essence, graph_ops isn't limited by its func_hash
> > due to the global ftrace_graph_[regs]_call label. That should be corrected.
> >
> > What inspires me is the commit 0c0593b45c9b ("x86/ftrace: Make function
> > graph use ftrace directly") that uses graph_ops::func function to install
> > return_hooker and makes the function called against its func_hash.
> >
> > This series of patches makes function graph use ftrace directly for riscv.
> >
> > If FTRACE_WITH_REGS isn't defined, ftrace_caller keeps ftrace_graph_call
> > so that it can be replaced with the calling of prepare_ftrace_return by
> > the enable/disable helper.
> >
> > As for defining FTRACE_WITH_REGS, ftrace_caller is adjusted to save the
> > necessary regs against the pt_regs layout, so it can reasonably call the
> > graph_ops::func function - ftrace_graph_func. And ftrace_graph_[regs]_call
> > and its enable/disable helper aren't needed.
> >
> > The tests generated by CONFIG_FTRACE_STARTUP_TEST have passed in the local
> > qemu-system-riscv64 virt machine. The following is the log during startup.
> >
> > ```
> > Nov 15 03:07:13 stage4 kernel: Testing tracer function: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace ops #1:
> > Nov 15 03:07:13 stage4 kernel: (1 0 1 0 0)
> > Nov 15 03:07:13 stage4 kernel: (1 1 2 0 0)
> > Nov 15 03:07:13 stage4 kernel: (2 1 3 0 365)
> > Nov 15 03:07:13 stage4 kernel: (2 2 4 0 399)
> > Nov 15 03:07:13 stage4 kernel: (3 2 4 0 146071)
> > Nov 15 03:07:13 stage4 kernel: (3 3 5 0 146105) PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing dynamic ftrace ops #2:
> > Nov 15 03:07:13 stage4 kernel: (1 0 1 589 0)
> > Nov 15 03:07:13 stage4 kernel: (1 1 2 635 0)
> > Nov 15 03:07:13 stage4 kernel: (2 1 3 1 2)
> > Nov 15 03:07:13 stage4 kernel: (2 2 4 125 126)
> > Nov 15 03:07:13 stage4 kernel: (3 2 4 146001 146078)
> > Nov 15 03:07:13 stage4 kernel: (3 3 5 146035 146112) PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing ftrace recursion: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing ftrace recursion safe: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing ftrace regs: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing tracer nop: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing tracer irqsoff: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup:
> > Nov 15 03:07:13 stage4 kernel: sched: DL replenish lagged too much
> > Nov 15 03:07:13 stage4 kernel: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup_rt: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing tracer wakeup_dl: PASSED
> > Nov 15 03:07:13 stage4 kernel: Testing tracer function_graph: PASSED
> > ```
> >
> > Note that the changes of mcount-dyn.S conflicts with this unmerged
> > commit (riscv: entry: consolidate general regs saving/restoring).
> > https://lore.kernel.org/linux-riscv/[email protected]
> >
> > Changes since v1
> > - fixed the checkpatch warnings in patch 1
> > Link: https://lore.kernel.org/linux-riscv/[email protected]
> >
> > Song Shuai (3):
> > riscv/ftrace: add ftrace_graph_func
> > riscv/ftrace: SAVE_ALL supports lightweight save
> > riscv/ftrace: cleanup ftrace_caller and ftrace_regs_caller
> >
> > arch/riscv/include/asm/ftrace.h | 13 ++-
> > arch/riscv/kernel/ftrace.c | 35 ++++----
> > arch/riscv/kernel/mcount-dyn.S | 145 +++++++++++++++++++++++---------
> > 3 files changed, 136 insertions(+), 57 deletions(-)
> >
> > --
> > 2.20.1
> >
>
>
> --
> Best Regards
> Guo Ren
Thanks,
Song