[PATCH] riscv: entry: Mark ret_from_fork_kernel() and ret_from_fork_user() noinstr

[PATCH] riscv: entry: Mark ret_from_fork_kernel() and ret_from_fork_user() noinstr

[Karl Mehltretter]

ret_from_fork_kernel() and ret_from_fork_user() are traceable and
return after syscall_exit_to_user_mode() enters the RCU extended
quiescent state via user_enter_irqoff() on NO_HZ_FULL CPUs. When
function_graph tracing is enabled with
FTRACE_VALIDATE_RCU_IS_WATCHING, the return hook runs after RCU stops
watching and triggers the ftrace_return_to_handler() warning.

Mark ret_from_fork_kernel() and ret_from_fork_user() noinstr and bracket
the kernel thread callback with instrumentation_begin() and
instrumentation_end().

Fixes: f955aa8723a6 ("riscv: entry: Convert ret_from_fork() to C")
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---
Assisted by an LLM.

Tested on v7.4-rc1 (d48723711bae) under QEMU riscv64 (-smp 4) with
NO_HZ_FULL, FTRACE_VALIDATE_RCU_IS_WATCHING, PROVE_LOCKING,
nohz_full=1-3 rcu_nocbs=1-3 on the kernel command line, and
function_graph tracing during parallel fork bursts across CPUs 0-3.
Without this patch, two consecutive runs trigger:
  RCU not on for: ret_from_fork_user+0x4/0x368
  WARNING: include/linux/trace_recursion.h:132 at ftrace_return_to_handler+0x2e0/0x2f0, CPU#2
followed by an observed kernel paging request Oops in
ftrace_graph_ret_addr() while unwinding the warning backtrace.
With this patch, two consecutive runs complete without the
ftrace_return_to_handler() warning or unwind fault.

 arch/riscv/kernel/process.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
index 34c1d498d1cad..10c1bee77145f 100644
--- a/arch/riscv/kernel/process.c
+++ b/arch/riscv/kernel/process.c
@@ -225,14 +225,17 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 	return 0;
 }
 
-asmlinkage void ret_from_fork_kernel(void *fn_arg, int (*fn)(void *), struct pt_regs *regs)
+asmlinkage void noinstr ret_from_fork_kernel(void *fn_arg, int (*fn)(void *),
+					     struct pt_regs *regs)
 {
+	instrumentation_begin();
 	fn(fn_arg);
+	instrumentation_end();
 
 	syscall_exit_to_user_mode(regs);
 }
 
-asmlinkage void ret_from_fork_user(struct pt_regs *regs)
+asmlinkage void noinstr ret_from_fork_user(struct pt_regs *regs)
 {
 	syscall_exit_to_user_mode(regs);
 }
base-commit: d48723711bae9b8cdc56719c448ba42fef28cbee

-- 
2.39.5 (Apple Git-154)

Re: [PATCH] riscv: entry: Mark ret_from_fork_kernel() and ret_from_fork_user() noinstr

[Jisheng Zhang]

On Sat, Nov 07, 2026 at 08:49:31PM +0100, Karl Mehltretter wrote:
> ret_from_fork_kernel() and ret_from_fork_user() are traceable and
> return after syscall_exit_to_user_mode() enters the RCU extended
> quiescent state via user_enter_irqoff() on NO_HZ_FULL CPUs. When
> function_graph tracing is enabled with
> FTRACE_VALIDATE_RCU_IS_WATCHING, the return hook runs after RCU stops
> watching and triggers the ftrace_return_to_handler() warning.
> 
> Mark ret_from_fork_kernel() and ret_from_fork_user() noinstr and bracket
> the kernel thread callback with instrumentation_begin() and
> instrumentation_end().
> 
> Fixes: f955aa8723a6 ("riscv: entry: Convert ret_from_fork() to C")
> Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
> ---
> Assisted by an LLM.
> 
> Tested on v7.4-rc1 (d48723711bae) under QEMU riscv64 (-smp 4) with
> NO_HZ_FULL, FTRACE_VALIDATE_RCU_IS_WATCHING, PROVE_LOCKING,
> nohz_full=1-3 rcu_nocbs=1-3 on the kernel command line, and
> function_graph tracing during parallel fork bursts across CPUs 0-3.
> Without this patch, two consecutive runs trigger:
>   RCU not on for: ret_from_fork_user+0x4/0x368
>   WARNING: include/linux/trace_recursion.h:132 at ftrace_return_to_handler+0x2e0/0x2f0, CPU#2
> followed by an observed kernel paging request Oops in
> ftrace_graph_ret_addr() while unwinding the warning backtrace.
> With this patch, two consecutive runs complete without the
> ftrace_return_to_handler() warning or unwind fault.
> 
>  arch/riscv/kernel/process.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/riscv/kernel/process.c b/arch/riscv/kernel/process.c
> index 34c1d498d1cad..10c1bee77145f 100644
> --- a/arch/riscv/kernel/process.c
> +++ b/arch/riscv/kernel/process.c
> @@ -225,14 +225,17 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
>  	return 0;
>  }
>  
> -asmlinkage void ret_from_fork_kernel(void *fn_arg, int (*fn)(void *), struct pt_regs *regs)
> +asmlinkage void noinstr ret_from_fork_kernel(void *fn_arg, int (*fn)(void *),
> +					     struct pt_regs *regs)
>  {
> +	instrumentation_begin();

since riscv has no objtool support, this can be removed.

>  	fn(fn_arg);
> +	instrumentation_end();
>  
>  	syscall_exit_to_user_mode(regs);
>  }
>  
> -asmlinkage void ret_from_fork_user(struct pt_regs *regs)
> +asmlinkage void noinstr ret_from_fork_user(struct pt_regs *regs)
>  {
>  	syscall_exit_to_user_mode(regs);
>  }
> base-commit: d48723711bae9b8cdc56719c448ba42fef28cbee
> 
> -- 
> 2.39.5 (Apple Git-154)
> 
> 
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv