[PATCH V2] gdb/arm-tdep: the "strd rd, [sp, #-imm]!" is the part of function prologue

[Joel Brobecker]

Hello,

On Tue, Jan 14, 2020 at 09:57:35AM +0800, chenzefeng wrote:
> Compiled the follow test case code with "-mtune=cortex-a15 -O2",
> The function test1 dump core:
> (gdb) disassemble test1
> Dump of assembler code for function test1:
>    0x0040062c <+0>:     strd    r4, [sp, #-16]!
> => 0x00400630 <+4>:     add     r4, r0, r1
>    0x00400634 <+8>:     ldr     r0, [pc, #80]   ; 0x40068c <test1+96>
>    0x00400638 <+12>:    mov     r1, r4
>    0x0040063c <+16>:    str     r6, [sp, #8]
>    0x00400640 <+20>:    add     r5, r2, r3
> 
> When make a breakpoint at test1 + <+4>, we can not get a full backtrace:
> (gdb) bt
>  #0  0x00400630 in test1 ()
>  #1  0x004006e8 in show_info ()
>  #2  0x00000000 in ?? ()
> 
> For the arm_analyze_prologue can't deal with the insn
> "strd     r4, [sp, #-16]!",
> which should change the sp register.
> 
> The test case code as follow:
> void test1(int a, int b, int c, int d)
> {
>         int e = a + b;
>         int f = c + d;
>         int g = e / f;
>         printf("a + b = %d\n", e);
>         printf("c + d = %d\n", f);
>         printf("e / f = %d\n", g);
> }
> 
> int show_info(int a, int b, int c)
> {
>         int d;
> 
>         d = a + c;
>         printf("%d, %d %ld %c\n",a , d , c , b);
> 
>         test1(a, b, c, d);
> 
>         return d;
> }
> 
> int main(int argc, char *argv[])
> {
>         return show_info(1, 2, 3);
> }
> ---
>  gdb/ChangeLog  | 4 ++++
>  gdb/arm-tdep.c | 9 +++++++++
>  2 files changed, 13 insertions(+)
> 
> diff --git a/gdb/ChangeLog b/gdb/ChangeLog
> index 731449b6ef..179630b70f 100644
> --- a/gdb/ChangeLog
> +++ b/gdb/ChangeLog
> @@ -1,3 +1,7 @@
> +2020-01-14  Chenzefeng <chenzefeng2@huawei.com>
> +	* arm-tdep.c (arm_analyze_prologue): Add a new insn for function
> +	prologue.

The change looks reasonable, but please resubmit following the same
recommendations as the one I proposed in my previous message to you
(https://www.sourceware.org/ml/gdb-patches/2020-02/msg00018.html)

Let's also take this opportunity for being a little more specific
in the ChangeLog:

        * arm-tdep.c (arm_analyze_prologue): Add handling of
        "strd rd [sp, #imm]!" instructions.

(do you mind doing the same for the other patch as well?)

Thank you!

> +
>  2020-01-13  Chenzefeng <chenzefeng2@huawei.com>
>  
>  	* arm-tdep.c (arm_analyze_prologue): Fix to adapt insn
> diff --git a/gdb/arm-tdep.c b/gdb/arm-tdep.c
> index 8a9f878b69..060acbf285 100644
> --- a/gdb/arm-tdep.c
> +++ b/gdb/arm-tdep.c
> @@ -1550,6 +1550,15 @@ arm_analyze_prologue (struct gdbarch *gdbarch,
>  		       regs[bits (insn, 12, 15)]);
>  	  continue;
>  	}
> +      else if ((insn & 0xffff00f0) == 0xe16d00f0) /* strd rd, [sp, #imm]! */
> +	{
> +          unsigned imm = ((insn & 0xf00) >> 4) + (insn & 0xf);
> +          if (stack.store_would_trash (regs[ARM_SP_REGNUM]))
> +            break;
> +          regs[ARM_SP_REGNUM] = pv_add_constant (regs[ARM_SP_REGNUM], -imm);
> +          stack.store (regs[ARM_SP_REGNUM], 4, regs[bits (insn, 12, 15)]);
> +          continue;
> +	}
>        else if ((insn & 0xffff0000) == 0xe92d0000)
>  	/* stmfd sp!, {..., fp, ip, lr, pc}
>  	   or
> -- 
> 2.12.3

-- 
Joel