[PATCH 3/6] i386: Remove syscall assembly codes with 6 arguments

[H.J. Lu]

On Fri, Aug 21, 2015 at 4:57 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Fri, Aug 14, 2015 at 10:34 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>> On Fri, Aug 14, 2015 at 5:17 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
>>> OK for master?
>>>
>>> H.J.
>>> ---
>>> This patch removes i386 assembly implementation for epoll_wait, mmap,
>>> mmap64, semtimeop now that i386 have 6 argument syscall support from
>>> C code and GCC 5 can inline syscalls with 6 argument.  We also compile
>>> epoll_pwait.c, mmap.c, mmap64.c and semtimedop.c with -fomit-frame-pointer
>>> since %ebp may be used to pass the 6th argument to syscall.
>>>
>>> Fo sysdeps/unix/sysv/linux/i386/mmap.c, with -O2 -march=i686
>>> -mtune=generic, GCC 5.2 now generates:
>>>
>>> <__mmap>:
>>>    0:   sub    $0x10,%esp
>>>    3:   mov    0x28(%esp),%eax
>>>    7:   mov    %ebx,(%esp)
>>>    a:   mov    0x18(%esp),%ecx
>>>    e:   mov    %esi,0x4(%esp)
>>>   12:   mov    0x14(%esp),%ebx
>>>   16:   mov    %edi,0x8(%esp)
>>>   1a:   mov    0x1c(%esp),%edx
>>>   1e:   test   $0xfff,%eax
>>>   23:   mov    0x20(%esp),%esi
>>>   27:   mov    %ebp,0xc(%esp)
>>>   2b:   mov    0x24(%esp),%edi
>>>   2f:   jne    60 <__mmap+0x60>
>>>   31:   sar    $0xc,%eax
>>>   34:   mov    %eax,%ebp
>>>   36:   mov    $0xc0,%eax
>>>   3b:   call   *%gs:0x10
>>>   42:   cmp    $0xfffff000,%eax
>>>   47:   ja     65 <__mmap+0x65>
>>>   49:   mov    (%esp),%ebx
>>>   4c:   mov    0x4(%esp),%esi
>>>   50:   mov    0x8(%esp),%edi
>>>   54:   mov    0xc(%esp),%ebp
>>>   58:   add    $0x10,%esp
>>>   5b:   ret
>>>   5c:   lea    0x0(%esi,%eiz,1),%esi
>>>   60:   mov    $0xffffffea,%eax
>>>   65:   mov    (%esp),%ebx
>>>   68:   mov    0x4(%esp),%esi
>>>   6c:   mov    0x8(%esp),%edi
>>>   70:   mov    0xc(%esp),%ebp
>>>   74:   add    $0x10,%esp
>>>   77:   jmp    78 <__mmap+0x78> 78: R_386_PC32  __syscall_error
>>>
>>> vs sysdeps/unix/sysv/linux/i386/mmap.S:
>>>
>>> <__mmap>:
>>>    0:   push   %ebp
>>>    1:   push   %ebx
>>>    2:   push   %esi
>>>    3:   push   %edi
>>>    4:   mov    0x14(%esp),%ebx
>>>    8:   mov    0x18(%esp),%ecx
>>>    c:   mov    0x1c(%esp),%edx
>>>   10:   mov    0x20(%esp),%esi
>>>   14:   mov    0x24(%esp),%edi
>>>   18:   mov    0x28(%esp),%ebp
>>>   1c:   test   $0xfff,%ebp
>>>   22:   mov    $0xffffffea,%eax
>>>   27:   jne    38 <__mmap+0x38>
>>>   29:   shr    $0xc,%ebp
>>>   2c:   mov    $0xc0,%eax
>>>   31:   call   *%gs:0x10
>>>   38:   pop    %edi
>>>   39:   pop    %esi
>>>   3a:   pop    %ebx
>>>   3b:   pop    %ebp
>>>   3c:   cmp    $0xfffff000,%eax
>>>   41:   ja     44 <__mmap+0x44>
>>>   43:   ret
>>>   44:   call   45 <__mmap+0x45> 45: R_386_PC32  __x86.get_pc_thunk.cx
>>>   49:   add    $0x2,%ecx        4b: R_386_GOTPC _GLOBAL_OFFSET_TABLE_
>>>   4f:   mov    0x0(%ecx),%ecx   51: R_386_TLS_GOTIE     __libc_errno
>>>   55:   neg    %eax
>>>   57:   mov    %eax,%gs:(%ecx)
>>>   5a:   or     $0xffffffff,%eax
>>>   5d:   ret
>>>
>>> The C version has:
>>>
>>>    3:   mov    0x28(%esp),%eax
>>> ...
>>>   1e:   test   $0xfff,%eax
>>> ...
>>>   31:   sar    $0xc,%eax
>>>   34:   mov    %eax,%ebp
>>>
>>> is due to missing $ebx register constraint for inline asm.  We have
>>> to use "r" constraint with
>>>
>>> register unsigned int _a6 asm ("ebp") = (unsigned int) (arg6);
>>>
>>> and compiler chose %eax for offset (arg6) in
>>>
>>>   if (offset & ((1 << MMAP_PAGE_SHIFT) - 1))
>>>
>>>         * sysdeps/unix/sysv/linux/i386/Makefile (CFLAGS-epoll_pwait.c):
>>>         Add -fomit-frame-pointer.
>>>         (CFLAGS-mmap.c): Likewise.
>>>         (CFLAGS-mmap64.c): Likewise.
>>>         (CFLAGS-semtimedop.c): Likewise.
>>>         * sysdeps/unix/sysv/linux/i386/mmap.c: New file.
>>>         * sysdeps/unix/sysv/linux/i386/epoll_pwait.S: Remove file.
>>>         * sysdeps/unix/sysv/linux/i386/mmap.S: Likewise.
>>>         * sysdeps/unix/sysv/linux/i386/mmap64.S: Likewise.
>>>         * sysdeps/unix/sysv/linux/i386/semtimedop.S: Likewise.
>>
>> Here is the patch for updated INLINE_SYSCALL_RETURN and
>> INLINE_SYSCALL_ERROR_RETURN.  OK for master?
>
> I will check it in next Monday if there are no objections.
>

This is the patch I will check in next Monday.

-- 
H.J.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-i386-Remove-syscall-assembly-codes-with-6-arguments.patch
Type: text/x-patch
Size: 15914 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20150821/42c80ae6/attachment.bin>