[musl] clone(),sys_clone() arguments

[musl] clone(),sys_clone() arguments

[erny hombre]

Hello,

I am writing an operating system for an ARM processor (Cortex-A9). For the user programs I want to use musl libc.
I am adding a layer between musl and my os to translate linux system calls into native system calls.
For pthread_create the syscall sys_clone is used. This syscall is not called directly. The library
function clone() is used instead. The arguments of clone() and the raw system call sys_clone differ:
int clone(int (*fn)(void *), void *child_stack, int flags, void *arg, ...
          /* pid_t *ptid, struct user_desc *tls, pid_t *ctid */ );
long sys_clone(unsigned long flags, void *child_stack, void *ptid, void *ctid, struct pt_regs *regs);

Musl uses __clone to reorder clone() arguments to the arguments expected by sys_clone:
/musl-1.2.0/src/thread/arm/clone.s:
__clone:
	stmfd sp!,{r4,r5,r6,r7}
	mov r7,#120
	mov r6,r3
	mov r5,r0
	mov r0,r2
	and r1,r1,#-16
	ldr r2,[sp,#16]
	ldr r3,[sp,#20]
	ldr r4,[sp,#24]
	svc 0				; sys_clone system call
	...

I think that the last two arguments for sys_clone (r3..ctid, r4..regs) are taken in the wrong
order from the clone parameters ([sp,#20]..tls, [sp,#24]..ctid).
Do I miss something or is this a bug ?

Re: [musl] clone(),sys_clone() arguments

[Rich Felker]

On Sat, Jun 27, 2020 at 02:36:20PM +0200, erny hombre wrote:
> Hello,
> 
> I am writing an operating system for an ARM processor (Cortex-A9). For the user programs I want to use musl libc.
> I am adding a layer between musl and my os to translate linux system calls into native system calls.
> For pthread_create the syscall sys_clone is used. This syscall is not called directly. The library
> function clone() is used instead. The arguments of clone() and the raw system call sys_clone differ:
> int clone(int (*fn)(void *), void *child_stack, int flags, void *arg, ...
>           /* pid_t *ptid, struct user_desc *tls, pid_t *ctid */ );
> long sys_clone(unsigned long flags, void *child_stack, void *ptid, void *ctid, struct pt_regs *regs);
> 
> Musl uses __clone to reorder clone() arguments to the arguments expected by sys_clone:
> /musl-1.2.0/src/thread/arm/clone.s:
> __clone:
> 	stmfd sp!,{r4,r5,r6,r7}
> 	mov r7,#120
> 	mov r6,r3
> 	mov r5,r0
> 	mov r0,r2
> 	and r1,r1,#-16
> 	ldr r2,[sp,#16]
> 	ldr r3,[sp,#20]
> 	ldr r4,[sp,#24]
> 	svc 0				; sys_clone system call
> 	...
> 
> I think that the last two arguments for sys_clone (r3..ctid, r4..regs) are taken in the wrong
> order from the clone parameters ([sp,#20]..tls, [sp,#24]..ctid).
> Do I miss something or is this a bug ?

Linux's arch/arm/Kconfig defines CLONE_BACKWARDS, which means the
argument order is (flags, sp, ptid, tp, ctid) -- see kernel/fork.c. I
don't know where you're getting the pt_regs argument from. That's not
part of the syscall interface ever; it is or was a kernel-internal
thing in some kernel versions.

Rich