Will,
On 11/07/2014 11:04 PM, Oleg Nesterov wrote:
On 11/07, AKASHI Takahiro wrote:
--- a/arch/arm/kernel/ptrace.c +++ b/arch/arm/kernel/ptrace.c @@ -853,11 +853,6 @@ long arch_ptrace(struct task_struct *child, long request, datap); break;
case PTRACE_SET_SYSCALL:task_thread_info(child)->syscall = data;ret = 0;break;- #ifdef CONFIG_CRUNCH case PTRACE_GETCRUNCHREGS: ret = ptrace_getcrunchregs(child, datap);
diff --git a/kernel/ptrace.c b/kernel/ptrace.c index 54e7522..d7048fa 100644 --- a/kernel/ptrace.c +++ b/kernel/ptrace.c @@ -1001,6 +1001,12 @@ int ptrace_request(struct task_struct *child, long request, break; } #endif
+#ifdef PTRACE_SET_SYSCALL
- case PTRACE_SET_SYSCALL:
ret = syscall_set_nr(child, task_pt_regs(child), data);break;+#endif
I too do not understand why it makes sense to move PTRACE_SET_SYSCALL into the common kernel/ptrace.c.
I think I explained why we need a new (atomic) interface of changing a system call number while tracing with ptrace. But I don't have a strong preference, either ptrace(SET_SYSCALL) or ptrace(SETREGSET, NT_SYSTEM_CALL).
To me the fact that PTRACE_SET_SYSCALL can be undefined and syscall_set_nr() is very much arch-dependant (but most probably trivial) means that this code should live in arch_ptrace().
Thinking of Oleg's comment above, it doesn't make sense neither to define generic NT_SYSTEM_CALL (user_regset) in uapi/linux/elf.h and implement it in ptrace_regset() in kernel/ptrace.c with arch-defined syscall_(g)set_nr().
Since we should have the same interface on arm and arm64, we'd better implement ptrace(PTRACE_SET_SYSCALL) locally on arm64 for now (as I originally submitted).
-Takahiro AKASHI
In any case, I think it doesn't make sense to pass task_pt_regs(child), this helper can do this itself if it needs struct pt_regs.
Oleg.