[Qemu-devel] [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Qemu-devel] [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Andrzej Zaborowski]

Revision: 4875
          http://svn.sv.gnu.org/viewvc/?view=rev&root=qemu&revision=4875
Author:   balrog
Date:     2008-07-16 11:31:11 +0000 (Wed, 16 Jul 2008)

Log Message:
-----------
Remove unintended dereference, kills a warning (Jan Kiszka).

Modified Paths:
--------------
    trunk/target-sh4/op.c

Modified: trunk/target-sh4/op.c
===================================================================
--- trunk/target-sh4/op.c	2008-07-16 04:45:12 UTC (rev 4874)
+++ trunk/target-sh4/op.c	2008-07-16 11:31:11 UTC (rev 4875)
@@ -594,8 +594,8 @@
 
 void OPPROTO op_tasb_rN(void)
 {
-    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
-    *(int8_t *) env->gregs[PARAM1] |= 0x80;
+    cond_t((env->gregs[PARAM1] && 0xff) == 0);
+    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
     RETURN();
 }
 

Re: [Qemu-devel] [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Laurent Desnogues]

On Wed, Jul 16, 2008 at 1:31 PM, Andrzej Zaborowski <balrogg@gmail.com> wrote:
>
> Modified: trunk/target-sh4/op.c
> ===================================================================
> --- trunk/target-sh4/op.c       2008-07-16 04:45:12 UTC (rev 4874)
> +++ trunk/target-sh4/op.c       2008-07-16 11:31:11 UTC (rev 4875)
> @@ -594,8 +594,8 @@
>
>  void OPPROTO op_tasb_rN(void)
>  {
> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);

Guess that should be a single '&'.

> +    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
>     RETURN();
>  }

Laurent

Re: [Qemu-devel] [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Andreas Schwab]

Andrzej Zaborowski <balrogg@gmail.com> writes:

>  void OPPROTO op_tasb_rN(void)
>  {
> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);
> +    *(int8_t *) &env->gregs[PARAM1] |= 0x80;

That does not make any sense at all.  The TAS insn operates on memory,
not on a register (atomic operations only make sense on memory anyway).

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

[Qemu-devel] Re: [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 672 bytes --]

Laurent Desnogues wrote:
> On Wed, Jul 16, 2008 at 1:31 PM, Andrzej Zaborowski <balrogg@gmail.com> wrote:
>> Modified: trunk/target-sh4/op.c
>> ===================================================================
>> --- trunk/target-sh4/op.c       2008-07-16 04:45:12 UTC (rev 4874)
>> +++ trunk/target-sh4/op.c       2008-07-16 11:31:11 UTC (rev 4875)
>> @@ -594,8 +594,8 @@
>>
>>  void OPPROTO op_tasb_rN(void)
>>  {
>> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
>> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
>> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);
> 
> Guess that should be a single '&'.

Oh, ooh. Yes, please fix.

Thanks,
Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

Re: [Qemu-devel] [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Laurent Desnogues]

On Wed, Jul 16, 2008 at 2:04 PM, Andreas Schwab <schwab@suse.de> wrote:
> Andrzej Zaborowski <balrogg@gmail.com> writes:
>
>>  void OPPROTO op_tasb_rN(void)
>>  {
>> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
>> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
>> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);
>> +    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
>
> That does not make any sense at all.  The TAS insn operates on memory,
> not on a register (atomic operations only make sense on memory anyway).

SH4 documentation says this:

TAS.B @Rn
          If (Rn) = 0, 1 → T, else 0 → T
          1 → MSB of (Rn)

So indeed it looks like Jan and Andrzej patch is wrong.


Laurent

[Qemu-devel] Re: [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Jan Kiszka]

Laurent Desnogues wrote:
> On Wed, Jul 16, 2008 at 2:04 PM, Andreas Schwab <schwab@suse.de> wrote:
>> Andrzej Zaborowski <balrogg@gmail.com> writes:
>>
>>>  void OPPROTO op_tasb_rN(void)
>>>  {
>>> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
>>> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
>>> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);
>>> +    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
>> That does not make any sense at all.  The TAS insn operates on memory,
>> not on a register (atomic operations only make sense on memory anyway).
> 
> SH4 documentation says this:
> 
> TAS.B @Rn
>           If (Rn) = 0, 1 → T, else 0 → T
>           1 → MSB of (Rn)
> 
> So indeed it looks like Jan and Andrzej patch is wrong.

At least the audience is finally listening. ;)

Is this one better?

Index: qemu/target-sh4/op.c
===================================================================
--- qemu.orig/target-sh4/op.c
+++ qemu/target-sh4/op.c
@@ -594,8 +594,9 @@ void OPPROTO op_shlr16_Rn(void)
 
 void OPPROTO op_tasb_rN(void)
 {
-    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
-    *(int8_t *) env->gregs[PARAM1] |= 0x80;
+    uint8_t val = ldub(env->gregs[PARAM1]);
+    cond_t(val == 0);
+    stb(env->gregs[PARAM1], val | 0x80);
     RETURN();
 }
 

Re: [Qemu-devel] Re: [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[andrzej zaborowski]

2008/7/16 Jan Kiszka <jan.kiszka@web.de>:
> Laurent Desnogues wrote:
>> On Wed, Jul 16, 2008 at 2:04 PM, Andreas Schwab <schwab@suse.de> wrote:
>>> Andrzej Zaborowski <balrogg@gmail.com> writes:
>>>
>>>>  void OPPROTO op_tasb_rN(void)
>>>>  {
>>>> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
>>>> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
>>>> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);
>>>> +    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
>>> That does not make any sense at all.  The TAS insn operates on memory,
>>> not on a register (atomic operations only make sense on memory anyway).
>>
>> SH4 documentation says this:
>>
>> TAS.B @Rn
>>           If (Rn) = 0, 1 → T, else 0 → T
>>           1 → MSB of (Rn)
>>
>> So indeed it looks like Jan and Andrzej patch is wrong.
>
> At least the audience is finally listening. ;)
>
> Is this one better?

I suspect one of these may be more correct, but I haven't seen the
docs.  The below, like the original version, assumes that if the store
generates some kind of trap, the flag is still affected. Otherwise
cond_t needs to be the last.

diff --git a/target-sh4/op.c b/target-sh4/op.c
--- a/target-sh4/op.c
+++ b/target-sh4/op.c
@@ -592,13 +592,6 @@ void OPPROTO op_shlr16_Rn(void)
     RETURN();
 }

-void OPPROTO op_tasb_rN(void)
-{
-    cond_t((env->gregs[PARAM1] & 0xff) == 0);
-    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
-    RETURN();
-}
-
 void OPPROTO op_movl_T0_rN(void)
 {
     env->gregs[PARAM1] = T0;
diff --git a/target-sh4/translate.c b/target-sh4/translate.c
--- a/target-sh4/translate.c
+++ b/target-sh4/translate.c
@@ -1077,7 +1077,12 @@ void _decode_opc(DisasContext * ctx)
        gen_op_shlr16_Rn(REG(B11_8));
        return;
     case 0x401b:               /* tas.b @Rn */
-       gen_op_tasb_rN(REG(B11_8));
+       gen_op_movl_rN_T0(REG(B11_8));
+       gen_op_movl_T0_T1();
+       gen_op_ldub_T0_T0(ctx);
+       gen_op_cmp_eq_imm_T0(0);
+       gen_op_or_imm_T0(0x80);
+       gen_op_stb_T0_T1(ctx);
        return;
     case 0xf00d: /* fsts FPUL,FRn - FPSCR: Nothing */
        gen_op_movl_fpul_FT0();

or

diff --git a/target-sh4/op.c b/target-sh4/op.c
--- a/target-sh4/op.c
+++ b/target-sh4/op.c
@@ -592,13 +592,6 @@ void OPPROTO op_shlr16_Rn(void)
     RETURN();
 }

-void OPPROTO op_tasb_rN(void)
-{
-    cond_t((env->gregs[PARAM1] & 0xff) == 0);
-    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
-    RETURN();
-}
-
 void OPPROTO op_movl_T0_rN(void)
 {
     env->gregs[PARAM1] = T0;
diff --git a/target-sh4/op_mem.c b/target-sh4/op_mem.c
--- a/target-sh4/op_mem.c
+++ b/target-sh4/op_mem.c
@@ -76,3 +76,10 @@ void glue(op_stfq_DT0_T1, MEMSUFFIX) (void) {
     glue(stfq, MEMSUFFIX) (T1, DT0);
     RETURN();
 }
+
+void glue(op_tasb_Rn, MEMSUFFIX) (void) {
+    uint8_t val = glue(ldub, MEMSUFFIX) (env->gregs[PARAM1]);
+    cond_t(val == 0);
+    glue(stb, MEMSUFFIX) (env->gregs[PARAM1], val | 0x80);
+    RETURN();
+}
diff --git a/target-sh4/translate.c b/target-sh4/translate.c
--- a/target-sh4/translate.c
+++ b/target-sh4/translate.c
@@ -80,6 +80,10 @@ static void sh4_translate_init()
     gen_op_st##width##_##reg##_T1_raw(); \
   }

+void gen_op_tasb_Rn(DisasContext *ctx, int reg) {
+    gen_op_tasb_Rn_raw(reg);
+}
+
 #else

 #define GEN_OP_LD(width, reg) \
@@ -93,6 +97,13 @@ static void sh4_translate_init()
     else gen_op_st##width##_##reg##_T1_user();\
   }

+void gen_op_tasb_Rn(DisasContext *ctx, int reg) {
+    if (ctx->memidx)
+        gen_op_tasb_Rn_kernel(reg);
+    else
+        gen_op_tasb_Rn_user(reg);
+}
+
 #endif

 GEN_OP_LD(ub, T0)
@@ -1077,7 +1088,7 @@ void _decode_opc(DisasContext * ctx)
        gen_op_shlr16_Rn(REG(B11_8));
        return;
     case 0x401b:               /* tas.b @Rn */
-       gen_op_tasb_rN(REG(B11_8));
+       gen_op_tasb_Rn(ctx, REG(B11_8));
        return;
     case 0xf00d: /* fsts FPUL,FRn - FPSCR: Nothing */
        gen_op_movl_fpul_FT0();

[Qemu-devel] Re: [4875] Remove unintended dereference, kills a warning (Jan Kiszka).

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 4429 bytes --]

andrzej zaborowski wrote:
> 2008/7/16 Jan Kiszka <jan.kiszka@web.de>:
>> Laurent Desnogues wrote:
>>> On Wed, Jul 16, 2008 at 2:04 PM, Andreas Schwab <schwab@suse.de> wrote:
>>>> Andrzej Zaborowski <balrogg@gmail.com> writes:
>>>>
>>>>>  void OPPROTO op_tasb_rN(void)
>>>>>  {
>>>>> -    cond_t(*(int8_t *) env->gregs[PARAM1] == 0);
>>>>> -    *(int8_t *) env->gregs[PARAM1] |= 0x80;
>>>>> +    cond_t((env->gregs[PARAM1] && 0xff) == 0);
>>>>> +    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
>>>> That does not make any sense at all.  The TAS insn operates on memory,
>>>> not on a register (atomic operations only make sense on memory anyway).
>>> SH4 documentation says this:
>>>
>>> TAS.B @Rn
>>>           If (Rn) = 0, 1 → T, else 0 → T
>>>           1 → MSB of (Rn)
>>>
>>> So indeed it looks like Jan and Andrzej patch is wrong.
>> At least the audience is finally listening. ;)
>>
>> Is this one better?
> 
> I suspect one of these may be more correct, but I haven't seen the
> docs.  The below, like the original version, assumes that if the store
> generates some kind of trap, the flag is still affected. Otherwise
> cond_t needs to be the last.
> 
> diff --git a/target-sh4/op.c b/target-sh4/op.c
> --- a/target-sh4/op.c
> +++ b/target-sh4/op.c
> @@ -592,13 +592,6 @@ void OPPROTO op_shlr16_Rn(void)
>      RETURN();
>  }
> 
> -void OPPROTO op_tasb_rN(void)
> -{
> -    cond_t((env->gregs[PARAM1] & 0xff) == 0);
> -    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
> -    RETURN();
> -}
> -
>  void OPPROTO op_movl_T0_rN(void)
>  {
>      env->gregs[PARAM1] = T0;
> diff --git a/target-sh4/translate.c b/target-sh4/translate.c
> --- a/target-sh4/translate.c
> +++ b/target-sh4/translate.c
> @@ -1077,7 +1077,12 @@ void _decode_opc(DisasContext * ctx)
>         gen_op_shlr16_Rn(REG(B11_8));
>         return;
>      case 0x401b:               /* tas.b @Rn */
> -       gen_op_tasb_rN(REG(B11_8));
> +       gen_op_movl_rN_T0(REG(B11_8));
> +       gen_op_movl_T0_T1();
> +       gen_op_ldub_T0_T0(ctx);
> +       gen_op_cmp_eq_imm_T0(0);
> +       gen_op_or_imm_T0(0x80);
> +       gen_op_stb_T0_T1(ctx);
>         return;
>      case 0xf00d: /* fsts FPUL,FRn - FPSCR: Nothing */
>         gen_op_movl_fpul_FT0();
> 
> or
> 
> diff --git a/target-sh4/op.c b/target-sh4/op.c
> --- a/target-sh4/op.c
> +++ b/target-sh4/op.c
> @@ -592,13 +592,6 @@ void OPPROTO op_shlr16_Rn(void)
>      RETURN();
>  }
> 
> -void OPPROTO op_tasb_rN(void)
> -{
> -    cond_t((env->gregs[PARAM1] & 0xff) == 0);
> -    *(int8_t *) &env->gregs[PARAM1] |= 0x80;
> -    RETURN();
> -}
> -
>  void OPPROTO op_movl_T0_rN(void)
>  {
>      env->gregs[PARAM1] = T0;
> diff --git a/target-sh4/op_mem.c b/target-sh4/op_mem.c
> --- a/target-sh4/op_mem.c
> +++ b/target-sh4/op_mem.c
> @@ -76,3 +76,10 @@ void glue(op_stfq_DT0_T1, MEMSUFFIX) (void) {
>      glue(stfq, MEMSUFFIX) (T1, DT0);
>      RETURN();
>  }
> +
> +void glue(op_tasb_Rn, MEMSUFFIX) (void) {
> +    uint8_t val = glue(ldub, MEMSUFFIX) (env->gregs[PARAM1]);
> +    cond_t(val == 0);
> +    glue(stb, MEMSUFFIX) (env->gregs[PARAM1], val | 0x80);
> +    RETURN();
> +}
> diff --git a/target-sh4/translate.c b/target-sh4/translate.c
> --- a/target-sh4/translate.c
> +++ b/target-sh4/translate.c
> @@ -80,6 +80,10 @@ static void sh4_translate_init()
>      gen_op_st##width##_##reg##_T1_raw(); \
>    }
> 
> +void gen_op_tasb_Rn(DisasContext *ctx, int reg) {
> +    gen_op_tasb_Rn_raw(reg);
> +}
> +
>  #else
> 
>  #define GEN_OP_LD(width, reg) \
> @@ -93,6 +97,13 @@ static void sh4_translate_init()
>      else gen_op_st##width##_##reg##_T1_user();\
>    }
> 
> +void gen_op_tasb_Rn(DisasContext *ctx, int reg) {
> +    if (ctx->memidx)
> +        gen_op_tasb_Rn_kernel(reg);
> +    else
> +        gen_op_tasb_Rn_user(reg);
> +}
> +
>  #endif
> 
>  GEN_OP_LD(ub, T0)
> @@ -1077,7 +1088,7 @@ void _decode_opc(DisasContext * ctx)
>         gen_op_shlr16_Rn(REG(B11_8));
>         return;
>      case 0x401b:               /* tas.b @Rn */
> -       gen_op_tasb_rN(REG(B11_8));
> +       gen_op_tasb_Rn(ctx, REG(B11_8));
>         return;
>      case 0xf00d: /* fsts FPUL,FRn - FPSCR: Nothing */
>         gen_op_movl_fpul_FT0();

This proposed fix for an open bug is about to be forgotten again. Can
anyone with SH4 experience comment on it?

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]