qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Qemu-devel] Re: softMMU / MOV translation


From: Clemens Kolbitsch
Subject: [Qemu-devel] Re: softMMU / MOV translation
Date: Wed, 25 Jul 2007 15:22:37 +0200
User-agent: KMail/1.9.6

i think to have found it in translate.c:

/* sign does not matter, except for lidt/lgdt call (TODO: fix it) */
static GenOpFunc *gen_op_ld_T0_A0[3 * 4] = {
    gen_op_ldub_raw_T0_A0,
    gen_op_lduw_raw_T0_A0,
    gen_op_ldl_raw_T0_A0,
    X86_64_ONLY(gen_op_ldq_raw_T0_A0),

#ifndef CONFIG_USER_ONLY
    gen_op_ldub_kernel_T0_A0,
    gen_op_lduw_kernel_T0_A0,
    gen_op_ldl_kernel_T0_A0,
    X86_64_ONLY(gen_op_ldq_kernel_T0_A0),

    gen_op_ldub_user_T0_A0,
    gen_op_lduw_user_T0_A0,
    gen_op_ldl_user_T0_A0,
    X86_64_ONLY(gen_op_ldq_user_T0_A0),
#endif
};


    gen_op_ldl_kernel_T0_A0

looks quite promising. however, i still have not found the code (e.g. in op.c) 
that actually includes the "call" opcode (i find it in the disassembled code 
of the qemu-binary, though).

is it generated somehow by gcc? the only thing that looks similar are the 
ld-functions in softmmu_header:

static inline RES_TYPE glue(glue(ld, USUFFIX), MEMSUFFIX)(target_ulong ptr)

however, if i understand the macro-defs right, these functions are not 
built...

maybe someone could point that out?


On Wednesday 25 July 2007 02:04:47 pm you wrote:
> hi!
> i tried asking this in the irc  but got no answer, hope someone can help me
> here :-)
>
> i'm working on memory-protection for my mather's thesis and have to dig
> into qemu memory management... could someone help me here please? i have
> the following problem:
>
> i'm trying to understand the dynamic translation for the following
> mov-opcode (it's taken from the qemu log with "log asm_in,asm_out"):
>
> IN:
> 0xc011c9f2:  mov    0x60(%esi),%edx
> the hex-code would be "0x8b 0x56 0x60"
> ...
>
> OUT:
> OUT: [size=455]
> 0x08d30fa0:  mov    0x18(%ebp),%edi
> 0x08d30fa3:  add    $0x60,%edi
> 0x08d30fa9:  mov    %edi,%edx
> 0x08d30fab:  mov    %edi,%eax
> 0x08d30fad:  shr    $0x8,%edx
> 0x08d30fb0:  and    $0xfffff003,%eax
> 0x08d30fb5:  and    $0xff0,%edx
> 0x08d30fbb:  lea    0x350(%edx,%ebp,1),%edx
> 0x08d30fc2:  cmp    (%edx),%eax
> 0x08d30fc4:  mov    %edi,%eax
> 0x08d30fc6:  je     0x8d30fd4
> 0x08d30fc8:  push   $0x0
> 0x08d30fca:  call   0x80ee06a         // __ldl_mmu
> 0x08d30fcf:  pop    %edx
> 0x08d30fd0:  mov    %eax,%ebx
> 0x08d30fd2:  jmp    0x8d30fd9
> 0x08d30fd4:  add    0xc(%edx),%eax
> 0x08d30fd7:  mov    (%eax),%ebx               // possibly the output of 
> gen_op_ld_T0_A0
> [ot]+ s->mem_index]();
> 0x08d30fd9:  mov    %ebx,0x8(%ebp)    // this is the output of
> gen_op_mov_reg_T0 [ot][reg](); (translate.c:4005)
> ...
>
>
> therefore, i think the following code in translate.c should be executed:
>
> case 0x8b: /* mov Ev, Gv */
>         if ((b & 1) == 0)
>             ot = OT_BYTE;
>         else
>             ot = OT_WORD + dflag;
>         modrm = ldub_code(s->pc++);
>         reg = ((modrm >> 3) & 7) | rex_r;
>
>         gen_ldst_modrm(s, modrm, ot, OR_TMP0, 0);
>         gen_op_mov_reg_T0[ot][reg]();
>         break;
>
>
> i debugged some time and found out that the last gen_op
>
>         gen_op_mov_reg_T0[ot][reg]();
>
> only produces
>
>         mov    %ebx,0x8(%ebp).
>
> thus, the rest of the OUT-codes is produced by
>
>        gen_ldst_modrm(s, modrm, ot, OR_TMP0, 0);
>
> however, i tried reading through the code but i could not find the function
> where
>
>       0x08d30fca:  call   0x80ee06a           // __ldl_mmu
>
> is inserted into the translation-buffer... i (think to) know why it must be
> inserted, but i just cannot figure out where it is put into the buffer.
>
> if someone could point that out, it would help me a LOT!
> thanks,
>   Clemens






reply via email to

[Prev in Thread] Current Thread [Next in Thread]