Question: CLONE_STUB_HEAD stack use

[dismantl]

I am trying to understand how the CLONE_STUB_HEAD shellcode works. I put the bytes through a disassembler and got this:

   0:   eb 44                   jmp    0x46
   2:   b9 00 00 00 00          mov    ecx,0x0      ; stack ptr
   7:   5b                      pop    ebx          ; payload ptr
   8:   83 e1 f0                and    ecx,0xfffffff0   ; align stack address
   b:   83 e9 1c                sub    ecx,0x1c     ; create stack space
   e:   31 c0                   xor    eax,eax
  10:   89 41 0c                mov    DWORD PTR [ecx+0xc],eax    ; 0
  13:   89 59 08                mov    DWORD PTR [ecx+0x8],ebx    ; payload ptr
  16:   c7 41 04 00 00 00 00    mov    DWORD PTR [ecx+0x4],0x0    ; 0
  1d:   be 00 00 00 00          mov    esi,0x0
  22:   ba 00 09 01 00          mov    edx,0x10900  ; clone flags
  27:   bb 00 09 01 00          mov    ebx,0x10900  ; clone flags
  2c:   bf 00 00 00 00          mov    edi,0x0
  31:   89 19                   mov    DWORD PTR [ecx],ebx   ; clone flags
  33:   b8 78 00 00 00          mov    eax,0x78
  38:   cd 80                   int    0x80
  3a:   85 c0                   test   eax,eax
  3c:   74 01                   je     0x3f
  3e:   cc                      int3   
  3f:   bd 00 00 00 00          mov    ebp,0x0
  44:   eb 05                   jmp    0x4b
  46:   e8 b7 ff ff ff          call   0x2

So in addition to putting the clone syscall parameters into the proper registers before the int80, you also create stack space (pointed to by ECX) and add four dwords: zero, a pointer to the caller payload immediately after CLONE_STUB_HEAD, zero, and then the clone flags.

My questions are this: What is the purpose of setting up the stack this way? Would it not suffice just to set the registers and make the syscall? I'm studying thread injection on linux, so thanks in advance for answering my questions.